First Commit

deps/jemalloc/include/jemalloc/internal/arena_externs.h

@@ -0,0 +1,104 @@
+#ifndef JEMALLOC_INTERNAL_ARENA_EXTERNS_H
+#define JEMALLOC_INTERNAL_ARENA_EXTERNS_H
+
+#include "jemalloc/internal/bin.h"
+#include "jemalloc/internal/extent_dss.h"
+#include "jemalloc/internal/hook.h"
+#include "jemalloc/internal/pages.h"
+#include "jemalloc/internal/stats.h"
+
+extern ssize_t opt_dirty_decay_ms;
+extern ssize_t opt_muzzy_decay_ms;
+
+extern percpu_arena_mode_t opt_percpu_arena;
+extern const char *percpu_arena_mode_names[];
+
+extern const uint64_t h_steps[SMOOTHSTEP_NSTEPS];
+extern malloc_mutex_t arenas_lock;
+
+extern size_t opt_oversize_threshold;
+extern size_t oversize_threshold;
+
+void arena_basic_stats_merge(tsdn_t *tsdn, arena_t *arena,
+    unsigned *nthreads, const char **dss, ssize_t *dirty_decay_ms,
+    ssize_t *muzzy_decay_ms, size_t *nactive, size_t *ndirty, size_t *nmuzzy);
+void arena_stats_merge(tsdn_t *tsdn, arena_t *arena, unsigned *nthreads,
+    const char **dss, ssize_t *dirty_decay_ms, ssize_t *muzzy_decay_ms,
+    size_t *nactive, size_t *ndirty, size_t *nmuzzy, arena_stats_t *astats,
+    bin_stats_t *bstats, arena_stats_large_t *lstats,
+    arena_stats_extents_t *estats);
+void arena_extents_dirty_dalloc(tsdn_t *tsdn, arena_t *arena,
+    extent_hooks_t **r_extent_hooks, extent_t *extent);
+#ifdef JEMALLOC_JET
+size_t arena_slab_regind(extent_t *slab, szind_t binind, const void *ptr);
+#endif
+extent_t *arena_extent_alloc_large(tsdn_t *tsdn, arena_t *arena,
+    size_t usize, size_t alignment, bool *zero);
+void arena_extent_dalloc_large_prep(tsdn_t *tsdn, arena_t *arena,
+    extent_t *extent);
+void arena_extent_ralloc_large_shrink(tsdn_t *tsdn, arena_t *arena,
+    extent_t *extent, size_t oldsize);
+void arena_extent_ralloc_large_expand(tsdn_t *tsdn, arena_t *arena,
+    extent_t *extent, size_t oldsize);
+ssize_t arena_dirty_decay_ms_get(arena_t *arena);
+bool arena_dirty_decay_ms_set(tsdn_t *tsdn, arena_t *arena, ssize_t decay_ms);
+ssize_t arena_muzzy_decay_ms_get(arena_t *arena);
+bool arena_muzzy_decay_ms_set(tsdn_t *tsdn, arena_t *arena, ssize_t decay_ms);
+void arena_decay(tsdn_t *tsdn, arena_t *arena, bool is_background_thread,
+    bool all);
+void arena_reset(tsd_t *tsd, arena_t *arena);
+void arena_destroy(tsd_t *tsd, arena_t *arena);
+void arena_tcache_fill_small(tsdn_t *tsdn, arena_t *arena, tcache_t *tcache,
+    cache_bin_t *tbin, szind_t binind, uint64_t prof_accumbytes);
+void arena_alloc_junk_small(void *ptr, const bin_info_t *bin_info,
+    bool zero);
+
+typedef void (arena_dalloc_junk_small_t)(void *, const bin_info_t *);
+extern arena_dalloc_junk_small_t *JET_MUTABLE arena_dalloc_junk_small;
+
+void *arena_malloc_hard(tsdn_t *tsdn, arena_t *arena, size_t size,
+    szind_t ind, bool zero);
+void *arena_palloc(tsdn_t *tsdn, arena_t *arena, size_t usize,
+    size_t alignment, bool zero, tcache_t *tcache);
+void arena_prof_promote(tsdn_t *tsdn, void *ptr, size_t usize);
+void arena_dalloc_promoted(tsdn_t *tsdn, void *ptr, tcache_t *tcache,
+    bool slow_path);
+void arena_dalloc_bin_junked_locked(tsdn_t *tsdn, arena_t *arena, bin_t *bin,
+    szind_t binind, extent_t *extent, void *ptr);
+void arena_dalloc_small(tsdn_t *tsdn, void *ptr);
+bool arena_ralloc_no_move(tsdn_t *tsdn, void *ptr, size_t oldsize, size_t size,
+    size_t extra, bool zero, size_t *newsize);
+void *arena_ralloc(tsdn_t *tsdn, arena_t *arena, void *ptr, size_t oldsize,
+    size_t size, size_t alignment, bool zero, tcache_t *tcache,
+    hook_ralloc_args_t *hook_args);
+dss_prec_t arena_dss_prec_get(arena_t *arena);
+bool arena_dss_prec_set(arena_t *arena, dss_prec_t dss_prec);
+ssize_t arena_dirty_decay_ms_default_get(void);
+bool arena_dirty_decay_ms_default_set(ssize_t decay_ms);
+ssize_t arena_muzzy_decay_ms_default_get(void);
+bool arena_muzzy_decay_ms_default_set(ssize_t decay_ms);
+bool arena_retain_grow_limit_get_set(tsd_t *tsd, arena_t *arena,
+    size_t *old_limit, size_t *new_limit);
+unsigned arena_nthreads_get(arena_t *arena, bool internal);
+void arena_nthreads_inc(arena_t *arena, bool internal);
+void arena_nthreads_dec(arena_t *arena, bool internal);
+size_t arena_extent_sn_next(arena_t *arena);
+arena_t *arena_new(tsdn_t *tsdn, unsigned ind, extent_hooks_t *extent_hooks);
+bool arena_init_huge(void);
+bool arena_is_huge(unsigned arena_ind);
+arena_t *arena_choose_huge(tsd_t *tsd);
+bin_t *arena_bin_choose_lock(tsdn_t *tsdn, arena_t *arena, szind_t binind,
+    unsigned *binshard);
+void arena_boot(sc_data_t *sc_data);
+void arena_prefork0(tsdn_t *tsdn, arena_t *arena);
+void arena_prefork1(tsdn_t *tsdn, arena_t *arena);
+void arena_prefork2(tsdn_t *tsdn, arena_t *arena);
+void arena_prefork3(tsdn_t *tsdn, arena_t *arena);
+void arena_prefork4(tsdn_t *tsdn, arena_t *arena);
+void arena_prefork5(tsdn_t *tsdn, arena_t *arena);
+void arena_prefork6(tsdn_t *tsdn, arena_t *arena);
+void arena_prefork7(tsdn_t *tsdn, arena_t *arena);
+void arena_postfork_parent(tsdn_t *tsdn, arena_t *arena);
+void arena_postfork_child(tsdn_t *tsdn, arena_t *arena);
+
+#endif /* JEMALLOC_INTERNAL_ARENA_EXTERNS_H */

deps/jemalloc/include/jemalloc/internal/arena_inlines_a.h

@@ -0,0 +1,57 @@
+#ifndef JEMALLOC_INTERNAL_ARENA_INLINES_A_H
+#define JEMALLOC_INTERNAL_ARENA_INLINES_A_H
+
+static inline unsigned
+arena_ind_get(const arena_t *arena) {
+	return base_ind_get(arena->base);
+}
+
+static inline void
+arena_internal_add(arena_t *arena, size_t size) {
+	atomic_fetch_add_zu(&arena->stats.internal, size, ATOMIC_RELAXED);
+}
+
+static inline void
+arena_internal_sub(arena_t *arena, size_t size) {
+	atomic_fetch_sub_zu(&arena->stats.internal, size, ATOMIC_RELAXED);
+}
+
+static inline size_t
+arena_internal_get(arena_t *arena) {
+	return atomic_load_zu(&arena->stats.internal, ATOMIC_RELAXED);
+}
+
+static inline bool
+arena_prof_accum(tsdn_t *tsdn, arena_t *arena, uint64_t accumbytes) {
+	cassert(config_prof);
+
+	if (likely(prof_interval == 0 || !prof_active_get_unlocked())) {
+		return false;
+	}
+
+	return prof_accum_add(tsdn, &arena->prof_accum, accumbytes);
+}
+
+static inline void
+percpu_arena_update(tsd_t *tsd, unsigned cpu) {
+	assert(have_percpu_arena);
+	arena_t *oldarena = tsd_arena_get(tsd);
+	assert(oldarena != NULL);
+	unsigned oldind = arena_ind_get(oldarena);
+
+	if (oldind != cpu) {
+		unsigned newind = cpu;
+		arena_t *newarena = arena_get(tsd_tsdn(tsd), newind, true);
+		assert(newarena != NULL);
+
+		/* Set new arena/tcache associations. */
+		arena_migrate(tsd, oldind, newind);
+		tcache_t *tcache = tcache_get(tsd);
+		if (tcache != NULL) {
+			tcache_arena_reassociate(tsd_tsdn(tsd), tcache,
+			    newarena);
+		}
+	}
+}
+
+#endif /* JEMALLOC_INTERNAL_ARENA_INLINES_A_H */

deps/jemalloc/include/jemalloc/internal/arena_inlines_b.h

@@ -0,0 +1,427 @@
+#ifndef JEMALLOC_INTERNAL_ARENA_INLINES_B_H
+#define JEMALLOC_INTERNAL_ARENA_INLINES_B_H
+
+#include "jemalloc/internal/jemalloc_internal_types.h"
+#include "jemalloc/internal/mutex.h"
+#include "jemalloc/internal/rtree.h"
+#include "jemalloc/internal/sc.h"
+#include "jemalloc/internal/sz.h"
+#include "jemalloc/internal/ticker.h"
+
+JEMALLOC_ALWAYS_INLINE bool
+arena_has_default_hooks(arena_t *arena) {
+	return (extent_hooks_get(arena) == &extent_hooks_default);
+}
+
+JEMALLOC_ALWAYS_INLINE arena_t *
+arena_choose_maybe_huge(tsd_t *tsd, arena_t *arena, size_t size) {
+	if (arena != NULL) {
+		return arena;
+	}
+
+	/*
+	 * For huge allocations, use the dedicated huge arena if both are true:
+	 * 1) is using auto arena selection (i.e. arena == NULL), and 2) the
+	 * thread is not assigned to a manual arena.
+	 */
+	if (unlikely(size >= oversize_threshold)) {
+		arena_t *tsd_arena = tsd_arena_get(tsd);
+		if (tsd_arena == NULL || arena_is_auto(tsd_arena)) {
+			return arena_choose_huge(tsd);
+		}
+	}
+
+	return arena_choose(tsd, NULL);
+}
+
+JEMALLOC_ALWAYS_INLINE prof_tctx_t *
+arena_prof_tctx_get(tsdn_t *tsdn, const void *ptr, alloc_ctx_t *alloc_ctx) {
+	cassert(config_prof);
+	assert(ptr != NULL);
+
+	/* Static check. */
+	if (alloc_ctx == NULL) {
+		const extent_t *extent = iealloc(tsdn, ptr);
+		if (unlikely(!extent_slab_get(extent))) {
+			return large_prof_tctx_get(tsdn, extent);
+		}
+	} else {
+		if (unlikely(!alloc_ctx->slab)) {
+			return large_prof_tctx_get(tsdn, iealloc(tsdn, ptr));
+		}
+	}
+	return (prof_tctx_t *)(uintptr_t)1U;
+}
+
+JEMALLOC_ALWAYS_INLINE void
+arena_prof_tctx_set(tsdn_t *tsdn, const void *ptr, size_t usize,
+    alloc_ctx_t *alloc_ctx, prof_tctx_t *tctx) {
+	cassert(config_prof);
+	assert(ptr != NULL);
+
+	/* Static check. */
+	if (alloc_ctx == NULL) {
+		extent_t *extent = iealloc(tsdn, ptr);
+		if (unlikely(!extent_slab_get(extent))) {
+			large_prof_tctx_set(tsdn, extent, tctx);
+		}
+	} else {
+		if (unlikely(!alloc_ctx->slab)) {
+			large_prof_tctx_set(tsdn, iealloc(tsdn, ptr), tctx);
+		}
+	}
+}
+
+static inline void
+arena_prof_tctx_reset(tsdn_t *tsdn, const void *ptr, prof_tctx_t *tctx) {
+	cassert(config_prof);
+	assert(ptr != NULL);
+
+	extent_t *extent = iealloc(tsdn, ptr);
+	assert(!extent_slab_get(extent));
+
+	large_prof_tctx_reset(tsdn, extent);
+}
+
+JEMALLOC_ALWAYS_INLINE nstime_t
+arena_prof_alloc_time_get(tsdn_t *tsdn, const void *ptr,
+    alloc_ctx_t *alloc_ctx) {
+	cassert(config_prof);
+	assert(ptr != NULL);
+
+	extent_t *extent = iealloc(tsdn, ptr);
+	/*
+	 * Unlike arena_prof_prof_tctx_{get, set}, we only call this once we're
+	 * sure we have a sampled allocation.
+	 */
+	assert(!extent_slab_get(extent));
+	return large_prof_alloc_time_get(extent);
+}
+
+JEMALLOC_ALWAYS_INLINE void
+arena_prof_alloc_time_set(tsdn_t *tsdn, const void *ptr, alloc_ctx_t *alloc_ctx,
+    nstime_t t) {
+	cassert(config_prof);
+	assert(ptr != NULL);
+
+	extent_t *extent = iealloc(tsdn, ptr);
+	assert(!extent_slab_get(extent));
+	large_prof_alloc_time_set(extent, t);
+}
+
+JEMALLOC_ALWAYS_INLINE void
+arena_decay_ticks(tsdn_t *tsdn, arena_t *arena, unsigned nticks) {
+	tsd_t *tsd;
+	ticker_t *decay_ticker;
+
+	if (unlikely(tsdn_null(tsdn))) {
+		return;
+	}
+	tsd = tsdn_tsd(tsdn);
+	decay_ticker = decay_ticker_get(tsd, arena_ind_get(arena));
+	if (unlikely(decay_ticker == NULL)) {
+		return;
+	}
+	if (unlikely(ticker_ticks(decay_ticker, nticks))) {
+		arena_decay(tsdn, arena, false, false);
+	}
+}
+
+JEMALLOC_ALWAYS_INLINE void
+arena_decay_tick(tsdn_t *tsdn, arena_t *arena) {
+	malloc_mutex_assert_not_owner(tsdn, &arena->decay_dirty.mtx);
+	malloc_mutex_assert_not_owner(tsdn, &arena->decay_muzzy.mtx);
+
+	arena_decay_ticks(tsdn, arena, 1);
+}
+
+/* Purge a single extent to retained / unmapped directly. */
+JEMALLOC_ALWAYS_INLINE void
+arena_decay_extent(tsdn_t *tsdn,arena_t *arena, extent_hooks_t **r_extent_hooks,
+    extent_t *extent) {
+	size_t extent_size = extent_size_get(extent);
+	extent_dalloc_wrapper(tsdn, arena,
+	    r_extent_hooks, extent);
+	if (config_stats) {
+		/* Update stats accordingly. */
+		arena_stats_lock(tsdn, &arena->stats);
+		arena_stats_add_u64(tsdn, &arena->stats,
+		    &arena->decay_dirty.stats->nmadvise, 1);
+		arena_stats_add_u64(tsdn, &arena->stats,
+		    &arena->decay_dirty.stats->purged, extent_size >> LG_PAGE);
+		arena_stats_sub_zu(tsdn, &arena->stats, &arena->stats.mapped,
+		    extent_size);
+		arena_stats_unlock(tsdn, &arena->stats);
+	}
+}
+
+JEMALLOC_ALWAYS_INLINE void *
+arena_malloc(tsdn_t *tsdn, arena_t *arena, size_t size, szind_t ind, bool zero,
+    tcache_t *tcache, bool slow_path) {
+	assert(!tsdn_null(tsdn) || tcache == NULL);
+
+	if (likely(tcache != NULL)) {
+		if (likely(size <= SC_SMALL_MAXCLASS)) {
+			return tcache_alloc_small(tsdn_tsd(tsdn), arena,
+			    tcache, size, ind, zero, slow_path);
+		}
+		if (likely(size <= tcache_maxclass)) {
+			return tcache_alloc_large(tsdn_tsd(tsdn), arena,
+			    tcache, size, ind, zero, slow_path);
+		}
+		/* (size > tcache_maxclass) case falls through. */
+		assert(size > tcache_maxclass);
+	}
+
+	return arena_malloc_hard(tsdn, arena, size, ind, zero);
+}
+
+JEMALLOC_ALWAYS_INLINE arena_t *
+arena_aalloc(tsdn_t *tsdn, const void *ptr) {
+	return extent_arena_get(iealloc(tsdn, ptr));
+}
+
+JEMALLOC_ALWAYS_INLINE size_t
+arena_salloc(tsdn_t *tsdn, const void *ptr) {
+	assert(ptr != NULL);
+
+	rtree_ctx_t rtree_ctx_fallback;
+	rtree_ctx_t *rtree_ctx = tsdn_rtree_ctx(tsdn, &rtree_ctx_fallback);
+
+	szind_t szind = rtree_szind_read(tsdn, &extents_rtree, rtree_ctx,
+	    (uintptr_t)ptr, true);
+	assert(szind != SC_NSIZES);
+
+	return sz_index2size(szind);
+}
+
+JEMALLOC_ALWAYS_INLINE size_t
+arena_vsalloc(tsdn_t *tsdn, const void *ptr) {
+	/*
+	 * Return 0 if ptr is not within an extent managed by jemalloc.  This
+	 * function has two extra costs relative to isalloc():
+	 * - The rtree calls cannot claim to be dependent lookups, which induces
+	 *   rtree lookup load dependencies.
+	 * - The lookup may fail, so there is an extra branch to check for
+	 *   failure.
+	 */
+
+	rtree_ctx_t rtree_ctx_fallback;
+	rtree_ctx_t *rtree_ctx = tsdn_rtree_ctx(tsdn, &rtree_ctx_fallback);
+
+	extent_t *extent;
+	szind_t szind;
+	if (rtree_extent_szind_read(tsdn, &extents_rtree, rtree_ctx,
+	    (uintptr_t)ptr, false, &extent, &szind)) {
+		return 0;
+	}
+
+	if (extent == NULL) {
+		return 0;
+	}
+	assert(extent_state_get(extent) == extent_state_active);
+	/* Only slab members should be looked up via interior pointers. */
+	assert(extent_addr_get(extent) == ptr || extent_slab_get(extent));
+
+	assert(szind != SC_NSIZES);
+
+	return sz_index2size(szind);
+}
+
+static inline void
+arena_dalloc_large_no_tcache(tsdn_t *tsdn, void *ptr, szind_t szind) {
+	if (config_prof && unlikely(szind < SC_NBINS)) {
+		arena_dalloc_promoted(tsdn, ptr, NULL, true);
+	} else {
+		extent_t *extent = iealloc(tsdn, ptr);
+		large_dalloc(tsdn, extent);
+	}
+}
+
+static inline void
+arena_dalloc_no_tcache(tsdn_t *tsdn, void *ptr) {
+	assert(ptr != NULL);
+
+	rtree_ctx_t rtree_ctx_fallback;
+	rtree_ctx_t *rtree_ctx = tsdn_rtree_ctx(tsdn, &rtree_ctx_fallback);
+
+	szind_t szind;
+	bool slab;
+	rtree_szind_slab_read(tsdn, &extents_rtree, rtree_ctx, (uintptr_t)ptr,
+	    true, &szind, &slab);
+
+	if (config_debug) {
+		extent_t *extent = rtree_extent_read(tsdn, &extents_rtree,
+		    rtree_ctx, (uintptr_t)ptr, true);
+		assert(szind == extent_szind_get(extent));
+		assert(szind < SC_NSIZES);
+		assert(slab == extent_slab_get(extent));
+	}
+
+	if (likely(slab)) {
+		/* Small allocation. */
+		arena_dalloc_small(tsdn, ptr);
+	} else {
+		arena_dalloc_large_no_tcache(tsdn, ptr, szind);
+	}
+}
+
+JEMALLOC_ALWAYS_INLINE void
+arena_dalloc_large(tsdn_t *tsdn, void *ptr, tcache_t *tcache, szind_t szind,
+    bool slow_path) {
+	if (szind < nhbins) {
+		if (config_prof && unlikely(szind < SC_NBINS)) {
+			arena_dalloc_promoted(tsdn, ptr, tcache, slow_path);
+		} else {
+			tcache_dalloc_large(tsdn_tsd(tsdn), tcache, ptr, szind,
+			    slow_path);
+		}
+	} else {
+		extent_t *extent = iealloc(tsdn, ptr);
+		large_dalloc(tsdn, extent);
+	}
+}
+
+JEMALLOC_ALWAYS_INLINE void
+arena_dalloc(tsdn_t *tsdn, void *ptr, tcache_t *tcache,
+    alloc_ctx_t *alloc_ctx, bool slow_path) {
+	assert(!tsdn_null(tsdn) || tcache == NULL);
+	assert(ptr != NULL);
+
+	if (unlikely(tcache == NULL)) {
+		arena_dalloc_no_tcache(tsdn, ptr);
+		return;
+	}
+
+	szind_t szind;
+	bool slab;
+	rtree_ctx_t *rtree_ctx;
+	if (alloc_ctx != NULL) {
+		szind = alloc_ctx->szind;
+		slab = alloc_ctx->slab;
+		assert(szind != SC_NSIZES);
+	} else {
+		rtree_ctx = tsd_rtree_ctx(tsdn_tsd(tsdn));
+		rtree_szind_slab_read(tsdn, &extents_rtree, rtree_ctx,
+		    (uintptr_t)ptr, true, &szind, &slab);
+	}
+
+	if (config_debug) {
+		rtree_ctx = tsd_rtree_ctx(tsdn_tsd(tsdn));
+		extent_t *extent = rtree_extent_read(tsdn, &extents_rtree,
+		    rtree_ctx, (uintptr_t)ptr, true);
+		assert(szind == extent_szind_get(extent));
+		assert(szind < SC_NSIZES);
+		assert(slab == extent_slab_get(extent));
+	}
+
+	if (likely(slab)) {
+		/* Small allocation. */
+		tcache_dalloc_small(tsdn_tsd(tsdn), tcache, ptr, szind,
+		    slow_path);
+	} else {
+		arena_dalloc_large(tsdn, ptr, tcache, szind, slow_path);
+	}
+}
+
+static inline void
+arena_sdalloc_no_tcache(tsdn_t *tsdn, void *ptr, size_t size) {
+	assert(ptr != NULL);
+	assert(size <= SC_LARGE_MAXCLASS);
+
+	szind_t szind;
+	bool slab;
+	if (!config_prof || !opt_prof) {
+		/*
+		 * There is no risk of being confused by a promoted sampled
+		 * object, so base szind and slab on the given size.
+		 */
+		szind = sz_size2index(size);
+		slab = (szind < SC_NBINS);
+	}
+
+	if ((config_prof && opt_prof) || config_debug) {
+		rtree_ctx_t rtree_ctx_fallback;
+		rtree_ctx_t *rtree_ctx = tsdn_rtree_ctx(tsdn,
+		    &rtree_ctx_fallback);
+
+		rtree_szind_slab_read(tsdn, &extents_rtree, rtree_ctx,
+		    (uintptr_t)ptr, true, &szind, &slab);
+
+		assert(szind == sz_size2index(size));
+		assert((config_prof && opt_prof) || slab == (szind < SC_NBINS));
+
+		if (config_debug) {
+			extent_t *extent = rtree_extent_read(tsdn,
+			    &extents_rtree, rtree_ctx, (uintptr_t)ptr, true);
+			assert(szind == extent_szind_get(extent));
+			assert(slab == extent_slab_get(extent));
+		}
+	}
+
+	if (likely(slab)) {
+		/* Small allocation. */
+		arena_dalloc_small(tsdn, ptr);
+	} else {
+		arena_dalloc_large_no_tcache(tsdn, ptr, szind);
+	}
+}
+
+JEMALLOC_ALWAYS_INLINE void
+arena_sdalloc(tsdn_t *tsdn, void *ptr, size_t size, tcache_t *tcache,
+    alloc_ctx_t *alloc_ctx, bool slow_path) {
+	assert(!tsdn_null(tsdn) || tcache == NULL);
+	assert(ptr != NULL);
+	assert(size <= SC_LARGE_MAXCLASS);
+
+	if (unlikely(tcache == NULL)) {
+		arena_sdalloc_no_tcache(tsdn, ptr, size);
+		return;
+	}
+
+	szind_t szind;
+	bool slab;
+	alloc_ctx_t local_ctx;
+	if (config_prof && opt_prof) {
+		if (alloc_ctx == NULL) {
+			/* Uncommon case and should be a static check. */
+			rtree_ctx_t rtree_ctx_fallback;
+			rtree_ctx_t *rtree_ctx = tsdn_rtree_ctx(tsdn,
+			    &rtree_ctx_fallback);
+			rtree_szind_slab_read(tsdn, &extents_rtree, rtree_ctx,
+			    (uintptr_t)ptr, true, &local_ctx.szind,
+			    &local_ctx.slab);
+			assert(local_ctx.szind == sz_size2index(size));
+			alloc_ctx = &local_ctx;
+		}
+		slab = alloc_ctx->slab;
+		szind = alloc_ctx->szind;
+	} else {
+		/*
+		 * There is no risk of being confused by a promoted sampled
+		 * object, so base szind and slab on the given size.
+		 */
+		szind = sz_size2index(size);
+		slab = (szind < SC_NBINS);
+	}
+
+	if (config_debug) {
+		rtree_ctx_t *rtree_ctx = tsd_rtree_ctx(tsdn_tsd(tsdn));
+		rtree_szind_slab_read(tsdn, &extents_rtree, rtree_ctx,
+		    (uintptr_t)ptr, true, &szind, &slab);
+		extent_t *extent = rtree_extent_read(tsdn,
+		    &extents_rtree, rtree_ctx, (uintptr_t)ptr, true);
+		assert(szind == extent_szind_get(extent));
+		assert(slab == extent_slab_get(extent));
+	}
+
+	if (likely(slab)) {
+		/* Small allocation. */
+		tcache_dalloc_small(tsdn_tsd(tsdn), tcache, ptr, szind,
+		    slow_path);
+	} else {
+		arena_dalloc_large(tsdn, ptr, tcache, szind, slow_path);
+	}
+}
+
+#endif /* JEMALLOC_INTERNAL_ARENA_INLINES_B_H */

deps/jemalloc/include/jemalloc/internal/arena_stats.h

@@ -0,0 +1,271 @@
+#ifndef JEMALLOC_INTERNAL_ARENA_STATS_H
+#define JEMALLOC_INTERNAL_ARENA_STATS_H
+
+#include "jemalloc/internal/atomic.h"
+#include "jemalloc/internal/mutex.h"
+#include "jemalloc/internal/mutex_prof.h"
+#include "jemalloc/internal/sc.h"
+
+JEMALLOC_DIAGNOSTIC_DISABLE_SPURIOUS
+
+/*
+ * In those architectures that support 64-bit atomics, we use atomic updates for
+ * our 64-bit values.  Otherwise, we use a plain uint64_t and synchronize
+ * externally.
+ */
+#ifdef JEMALLOC_ATOMIC_U64
+typedef atomic_u64_t arena_stats_u64_t;
+#else
+/* Must hold the arena stats mutex while reading atomically. */
+typedef uint64_t arena_stats_u64_t;
+#endif
+
+typedef struct arena_stats_large_s arena_stats_large_t;
+struct arena_stats_large_s {
+	/*
+	 * Total number of allocation/deallocation requests served directly by
+	 * the arena.
+	 */
+	arena_stats_u64_t	nmalloc;
+	arena_stats_u64_t	ndalloc;
+
+	/*
+	 * Number of allocation requests that correspond to this size class.
+	 * This includes requests served by tcache, though tcache only
+	 * periodically merges into this counter.
+	 */
+	arena_stats_u64_t	nrequests; /* Partially derived. */
+	/*
+	 * Number of tcache fills / flushes for large (similarly, periodically
+	 * merged).  Note that there is no large tcache batch-fill currently
+	 * (i.e. only fill 1 at a time); however flush may be batched.
+	 */
+	arena_stats_u64_t	nfills; /* Partially derived. */
+	arena_stats_u64_t	nflushes; /* Partially derived. */
+
+	/* Current number of allocations of this size class. */
+	size_t		curlextents; /* Derived. */
+};
+
+typedef struct arena_stats_decay_s arena_stats_decay_t;
+struct arena_stats_decay_s {
+	/* Total number of purge sweeps. */
+	arena_stats_u64_t	npurge;
+	/* Total number of madvise calls made. */
+	arena_stats_u64_t	nmadvise;
+	/* Total number of pages purged. */
+	arena_stats_u64_t	purged;
+};
+
+typedef struct arena_stats_extents_s arena_stats_extents_t;
+struct arena_stats_extents_s {
+	/*
+	 * Stats for a given index in the range [0, SC_NPSIZES] in an extents_t.
+	 * We track both bytes and # of extents: two extents in the same bucket
+	 * may have different sizes if adjacent size classes differ by more than
+	 * a page, so bytes cannot always be derived from # of extents.
+	 */
+	atomic_zu_t ndirty;
+	atomic_zu_t dirty_bytes;
+	atomic_zu_t nmuzzy;
+	atomic_zu_t muzzy_bytes;
+	atomic_zu_t nretained;
+	atomic_zu_t retained_bytes;
+};
+
+/*
+ * Arena stats.  Note that fields marked "derived" are not directly maintained
+ * within the arena code; rather their values are derived during stats merge
+ * requests.
+ */
+typedef struct arena_stats_s arena_stats_t;
+struct arena_stats_s {
+#ifndef JEMALLOC_ATOMIC_U64
+	malloc_mutex_t		mtx;
+#endif
+
+	/* Number of bytes currently mapped, excluding retained memory. */
+	atomic_zu_t		mapped; /* Partially derived. */
+
+	/*
+	 * Number of unused virtual memory bytes currently retained.  Retained
+	 * bytes are technically mapped (though always decommitted or purged),
+	 * but they are excluded from the mapped statistic (above).
+	 */
+	atomic_zu_t		retained; /* Derived. */
+
+	/* Number of extent_t structs allocated by base, but not being used. */
+	atomic_zu_t		extent_avail;
+
+	arena_stats_decay_t	decay_dirty;
+	arena_stats_decay_t	decay_muzzy;
+
+	atomic_zu_t		base; /* Derived. */
+	atomic_zu_t		internal;
+	atomic_zu_t		resident; /* Derived. */
+	atomic_zu_t		metadata_thp;
+
+	atomic_zu_t		allocated_large; /* Derived. */
+	arena_stats_u64_t	nmalloc_large; /* Derived. */
+	arena_stats_u64_t	ndalloc_large; /* Derived. */
+	arena_stats_u64_t	nfills_large; /* Derived. */
+	arena_stats_u64_t	nflushes_large; /* Derived. */
+	arena_stats_u64_t	nrequests_large; /* Derived. */
+
+	/* VM space had to be leaked (undocumented).  Normally 0. */
+	atomic_zu_t		abandoned_vm;
+
+	/* Number of bytes cached in tcache associated with this arena. */
+	atomic_zu_t		tcache_bytes; /* Derived. */
+
+	mutex_prof_data_t mutex_prof_data[mutex_prof_num_arena_mutexes];
+
+	/* One element for each large size class. */
+	arena_stats_large_t	lstats[SC_NSIZES - SC_NBINS];
+
+	/* Arena uptime. */
+	nstime_t		uptime;
+};
+
+static inline bool
+arena_stats_init(tsdn_t *tsdn, arena_stats_t *arena_stats) {
+	if (config_debug) {
+		for (size_t i = 0; i < sizeof(arena_stats_t); i++) {
+			assert(((char *)arena_stats)[i] == 0);
+		}
+	}
+#ifndef JEMALLOC_ATOMIC_U64
+	if (malloc_mutex_init(&arena_stats->mtx, "arena_stats",
+	    WITNESS_RANK_ARENA_STATS, malloc_mutex_rank_exclusive)) {
+		return true;
+	}
+#endif
+	/* Memory is zeroed, so there is no need to clear stats. */
+	return false;
+}
+
+static inline void
+arena_stats_lock(tsdn_t *tsdn, arena_stats_t *arena_stats) {
+#ifndef JEMALLOC_ATOMIC_U64
+	malloc_mutex_lock(tsdn, &arena_stats->mtx);
+#endif
+}
+
+static inline void
+arena_stats_unlock(tsdn_t *tsdn, arena_stats_t *arena_stats) {
+#ifndef JEMALLOC_ATOMIC_U64
+	malloc_mutex_unlock(tsdn, &arena_stats->mtx);
+#endif
+}
+
+static inline uint64_t
+arena_stats_read_u64(tsdn_t *tsdn, arena_stats_t *arena_stats,
+    arena_stats_u64_t *p) {
+#ifdef JEMALLOC_ATOMIC_U64
+	return atomic_load_u64(p, ATOMIC_RELAXED);
+#else
+	malloc_mutex_assert_owner(tsdn, &arena_stats->mtx);
+	return *p;
+#endif
+}
+
+static inline void
+arena_stats_add_u64(tsdn_t *tsdn, arena_stats_t *arena_stats,
+    arena_stats_u64_t *p, uint64_t x) {
+#ifdef JEMALLOC_ATOMIC_U64
+	atomic_fetch_add_u64(p, x, ATOMIC_RELAXED);
+#else
+	malloc_mutex_assert_owner(tsdn, &arena_stats->mtx);
+	*p += x;
+#endif
+}
+
+static inline void
+arena_stats_sub_u64(tsdn_t *tsdn, arena_stats_t *arena_stats,
+    arena_stats_u64_t *p, uint64_t x) {
+#ifdef JEMALLOC_ATOMIC_U64
+	uint64_t r = atomic_fetch_sub_u64(p, x, ATOMIC_RELAXED);
+	assert(r - x <= r);
+#else
+	malloc_mutex_assert_owner(tsdn, &arena_stats->mtx);
+	*p -= x;
+	assert(*p + x >= *p);
+#endif
+}
+
+/*
+ * Non-atomically sets *dst += src.  *dst needs external synchronization.
+ * This lets us avoid the cost of a fetch_add when its unnecessary (note that
+ * the types here are atomic).
+ */
+static inline void
+arena_stats_accum_u64(arena_stats_u64_t *dst, uint64_t src) {
+#ifdef JEMALLOC_ATOMIC_U64
+	uint64_t cur_dst = atomic_load_u64(dst, ATOMIC_RELAXED);
+	atomic_store_u64(dst, src + cur_dst, ATOMIC_RELAXED);
+#else
+	*dst += src;
+#endif
+}
+
+static inline size_t
+arena_stats_read_zu(tsdn_t *tsdn, arena_stats_t *arena_stats,
+    atomic_zu_t *p) {
+#ifdef JEMALLOC_ATOMIC_U64
+	return atomic_load_zu(p, ATOMIC_RELAXED);
+#else
+	malloc_mutex_assert_owner(tsdn, &arena_stats->mtx);
+	return atomic_load_zu(p, ATOMIC_RELAXED);
+#endif
+}
+
+static inline void
+arena_stats_add_zu(tsdn_t *tsdn, arena_stats_t *arena_stats,
+    atomic_zu_t *p, size_t x) {
+#ifdef JEMALLOC_ATOMIC_U64
+	atomic_fetch_add_zu(p, x, ATOMIC_RELAXED);
+#else
+	malloc_mutex_assert_owner(tsdn, &arena_stats->mtx);
+	size_t cur = atomic_load_zu(p, ATOMIC_RELAXED);
+	atomic_store_zu(p, cur + x, ATOMIC_RELAXED);
+#endif
+}
+
+static inline void
+arena_stats_sub_zu(tsdn_t *tsdn, arena_stats_t *arena_stats,
+    atomic_zu_t *p, size_t x) {
+#ifdef JEMALLOC_ATOMIC_U64
+	size_t r = atomic_fetch_sub_zu(p, x, ATOMIC_RELAXED);
+	assert(r - x <= r);
+#else
+	malloc_mutex_assert_owner(tsdn, &arena_stats->mtx);
+	size_t cur = atomic_load_zu(p, ATOMIC_RELAXED);
+	atomic_store_zu(p, cur - x, ATOMIC_RELAXED);
+#endif
+}
+
+/* Like the _u64 variant, needs an externally synchronized *dst. */
+static inline void
+arena_stats_accum_zu(atomic_zu_t *dst, size_t src) {
+	size_t cur_dst = atomic_load_zu(dst, ATOMIC_RELAXED);
+	atomic_store_zu(dst, src + cur_dst, ATOMIC_RELAXED);
+}
+
+static inline void
+arena_stats_large_flush_nrequests_add(tsdn_t *tsdn, arena_stats_t *arena_stats,
+    szind_t szind, uint64_t nrequests) {
+	arena_stats_lock(tsdn, arena_stats);
+	arena_stats_large_t *lstats = &arena_stats->lstats[szind - SC_NBINS];
+	arena_stats_add_u64(tsdn, arena_stats, &lstats->nrequests, nrequests);
+	arena_stats_add_u64(tsdn, arena_stats, &lstats->nflushes, 1);
+	arena_stats_unlock(tsdn, arena_stats);
+}
+
+static inline void
+arena_stats_mapped_add(tsdn_t *tsdn, arena_stats_t *arena_stats, size_t size) {
+	arena_stats_lock(tsdn, arena_stats);
+	arena_stats_add_zu(tsdn, arena_stats, &arena_stats->mapped, size);
+	arena_stats_unlock(tsdn, arena_stats);
+}
+
+#endif /* JEMALLOC_INTERNAL_ARENA_STATS_H */

deps/jemalloc/include/jemalloc/internal/arena_structs_a.h

@@ -0,0 +1,11 @@
+#ifndef JEMALLOC_INTERNAL_ARENA_STRUCTS_A_H
+#define JEMALLOC_INTERNAL_ARENA_STRUCTS_A_H
+
+#include "jemalloc/internal/bitmap.h"
+
+struct arena_slab_data_s {
+	/* Per region allocated/deallocated bitmap. */
+	bitmap_t	bitmap[BITMAP_GROUPS_MAX];
+};
+
+#endif /* JEMALLOC_INTERNAL_ARENA_STRUCTS_A_H */

deps/jemalloc/include/jemalloc/internal/arena_structs_b.h

@@ -0,0 +1,232 @@
+#ifndef JEMALLOC_INTERNAL_ARENA_STRUCTS_B_H
+#define JEMALLOC_INTERNAL_ARENA_STRUCTS_B_H
+
+#include "jemalloc/internal/arena_stats.h"
+#include "jemalloc/internal/atomic.h"
+#include "jemalloc/internal/bin.h"
+#include "jemalloc/internal/bitmap.h"
+#include "jemalloc/internal/extent_dss.h"
+#include "jemalloc/internal/jemalloc_internal_types.h"
+#include "jemalloc/internal/mutex.h"
+#include "jemalloc/internal/nstime.h"
+#include "jemalloc/internal/ql.h"
+#include "jemalloc/internal/sc.h"
+#include "jemalloc/internal/smoothstep.h"
+#include "jemalloc/internal/ticker.h"
+
+struct arena_decay_s {
+	/* Synchronizes all non-atomic fields. */
+	malloc_mutex_t		mtx;
+	/*
+	 * True if a thread is currently purging the extents associated with
+	 * this decay structure.
+	 */
+	bool			purging;
+	/*
+	 * Approximate time in milliseconds from the creation of a set of unused
+	 * dirty pages until an equivalent set of unused dirty pages is purged
+	 * and/or reused.
+	 */
+	atomic_zd_t		time_ms;
+	/* time / SMOOTHSTEP_NSTEPS. */
+	nstime_t		interval;
+	/*
+	 * Time at which the current decay interval logically started.  We do
+	 * not actually advance to a new epoch until sometime after it starts
+	 * because of scheduling and computation delays, and it is even possible
+	 * to completely skip epochs.  In all cases, during epoch advancement we
+	 * merge all relevant activity into the most recently recorded epoch.
+	 */
+	nstime_t		epoch;
+	/* Deadline randomness generator. */
+	uint64_t		jitter_state;
+	/*
+	 * Deadline for current epoch.  This is the sum of interval and per
+	 * epoch jitter which is a uniform random variable in [0..interval).
+	 * Epochs always advance by precise multiples of interval, but we
+	 * randomize the deadline to reduce the likelihood of arenas purging in
+	 * lockstep.
+	 */
+	nstime_t		deadline;
+	/*
+	 * Number of unpurged pages at beginning of current epoch.  During epoch
+	 * advancement we use the delta between arena->decay_*.nunpurged and
+	 * extents_npages_get(&arena->extents_*) to determine how many dirty
+	 * pages, if any, were generated.
+	 */
+	size_t			nunpurged;
+	/*
+	 * Trailing log of how many unused dirty pages were generated during
+	 * each of the past SMOOTHSTEP_NSTEPS decay epochs, where the last
+	 * element is the most recent epoch.  Corresponding epoch times are
+	 * relative to epoch.
+	 */
+	size_t			backlog[SMOOTHSTEP_NSTEPS];
+
+	/*
+	 * Pointer to associated stats.  These stats are embedded directly in
+	 * the arena's stats due to how stats structures are shared between the
+	 * arena and ctl code.
+	 *
+	 * Synchronization: Same as associated arena's stats field. */
+	arena_stats_decay_t	*stats;
+	/* Peak number of pages in associated extents.  Used for debug only. */
+	uint64_t		ceil_npages;
+};
+
+struct arena_s {
+	/*
+	 * Number of threads currently assigned to this arena.  Each thread has
+	 * two distinct assignments, one for application-serving allocation, and
+	 * the other for internal metadata allocation.  Internal metadata must
+	 * not be allocated from arenas explicitly created via the arenas.create
+	 * mallctl, because the arena.<i>.reset mallctl indiscriminately
+	 * discards all allocations for the affected arena.
+	 *
+	 *   0: Application allocation.
+	 *   1: Internal metadata allocation.
+	 *
+	 * Synchronization: atomic.
+	 */
+	atomic_u_t		nthreads[2];
+
+	/* Next bin shard for binding new threads. Synchronization: atomic. */
+	atomic_u_t		binshard_next;
+
+	/*
+	 * When percpu_arena is enabled, to amortize the cost of reading /
+	 * updating the current CPU id, track the most recent thread accessing
+	 * this arena, and only read CPU if there is a mismatch.
+	 */
+	tsdn_t		*last_thd;
+
+	/* Synchronization: internal. */
+	arena_stats_t		stats;
+
+	/*
+	 * Lists of tcaches and cache_bin_array_descriptors for extant threads
+	 * associated with this arena.  Stats from these are merged
+	 * incrementally, and at exit if opt_stats_print is enabled.
+	 *
+	 * Synchronization: tcache_ql_mtx.
+	 */
+	ql_head(tcache_t)			tcache_ql;
+	ql_head(cache_bin_array_descriptor_t)	cache_bin_array_descriptor_ql;
+	malloc_mutex_t				tcache_ql_mtx;
+
+	/* Synchronization: internal. */
+	prof_accum_t		prof_accum;
+
+	/*
+	 * PRNG state for cache index randomization of large allocation base
+	 * pointers.
+	 *
+	 * Synchronization: atomic.
+	 */
+	atomic_zu_t		offset_state;
+
+	/*
+	 * Extent serial number generator state.
+	 *
+	 * Synchronization: atomic.
+	 */
+	atomic_zu_t		extent_sn_next;
+
+	/*
+	 * Represents a dss_prec_t, but atomically.
+	 *
+	 * Synchronization: atomic.
+	 */
+	atomic_u_t		dss_prec;
+
+	/*
+	 * Number of pages in active extents.
+	 *
+	 * Synchronization: atomic.
+	 */
+	atomic_zu_t		nactive;
+
+	/*
+	 * Extant large allocations.
+	 *
+	 * Synchronization: large_mtx.
+	 */
+	extent_list_t		large;
+	/* Synchronizes all large allocation/update/deallocation. */
+	malloc_mutex_t		large_mtx;
+
+	/*
+	 * Collections of extents that were previously allocated.  These are
+	 * used when allocating extents, in an attempt to re-use address space.
+	 *
+	 * Synchronization: internal.
+	 */
+	extents_t		extents_dirty;
+	extents_t		extents_muzzy;
+	extents_t		extents_retained;
+
+	/*
+	 * Decay-based purging state, responsible for scheduling extent state
+	 * transitions.
+	 *
+	 * Synchronization: internal.
+	 */
+	arena_decay_t		decay_dirty; /* dirty --> muzzy */
+	arena_decay_t		decay_muzzy; /* muzzy --> retained */
+
+	/*
+	 * Next extent size class in a growing series to use when satisfying a
+	 * request via the extent hooks (only if opt_retain).  This limits the
+	 * number of disjoint virtual memory ranges so that extent merging can
+	 * be effective even if multiple arenas' extent allocation requests are
+	 * highly interleaved.
+	 *
+	 * retain_grow_limit is the max allowed size ind to expand (unless the
+	 * required size is greater).  Default is no limit, and controlled
+	 * through mallctl only.
+	 *
+	 * Synchronization: extent_grow_mtx
+	 */
+	pszind_t		extent_grow_next;
+	pszind_t		retain_grow_limit;
+	malloc_mutex_t		extent_grow_mtx;
+
+	/*
+	 * Available extent structures that were allocated via
+	 * base_alloc_extent().
+	 *
+	 * Synchronization: extent_avail_mtx.
+	 */
+	extent_tree_t		extent_avail;
+	atomic_zu_t		extent_avail_cnt;
+	malloc_mutex_t		extent_avail_mtx;
+
+	/*
+	 * bins is used to store heaps of free regions.
+	 *
+	 * Synchronization: internal.
+	 */
+	bins_t			bins[SC_NBINS];
+
+	/*
+	 * Base allocator, from which arena metadata are allocated.
+	 *
+	 * Synchronization: internal.
+	 */
+	base_t			*base;
+	/* Used to determine uptime.  Read-only after initialization. */
+	nstime_t		create_time;
+};
+
+/* Used in conjunction with tsd for fast arena-related context lookup. */
+struct arena_tdata_s {
+	ticker_t		decay_ticker;
+};
+
+/* Used to pass rtree lookup context down the path. */
+struct alloc_ctx_s {
+	szind_t szind;
+	bool slab;
+};
+
+#endif /* JEMALLOC_INTERNAL_ARENA_STRUCTS_B_H */

deps/jemalloc/include/jemalloc/internal/arena_types.h

@@ -0,0 +1,51 @@
+#ifndef JEMALLOC_INTERNAL_ARENA_TYPES_H
+#define JEMALLOC_INTERNAL_ARENA_TYPES_H
+
+#include "jemalloc/internal/sc.h"
+
+/* Maximum number of regions in one slab. */
+#define LG_SLAB_MAXREGS		(LG_PAGE - SC_LG_TINY_MIN)
+#define SLAB_MAXREGS		(1U << LG_SLAB_MAXREGS)
+
+/* Default decay times in milliseconds. */
+#define DIRTY_DECAY_MS_DEFAULT	ZD(10 * 1000)
+#define MUZZY_DECAY_MS_DEFAULT	(0)
+/* Number of event ticks between time checks. */
+#define DECAY_NTICKS_PER_UPDATE	1000
+
+typedef struct arena_slab_data_s arena_slab_data_t;
+typedef struct arena_decay_s arena_decay_t;
+typedef struct arena_s arena_t;
+typedef struct arena_tdata_s arena_tdata_t;
+typedef struct alloc_ctx_s alloc_ctx_t;
+
+typedef enum {
+	percpu_arena_mode_names_base   = 0, /* Used for options processing. */
+
+	/*
+	 * *_uninit are used only during bootstrapping, and must correspond
+	 * to initialized variant plus percpu_arena_mode_enabled_base.
+	 */
+	percpu_arena_uninit            = 0,
+	per_phycpu_arena_uninit        = 1,
+
+	/* All non-disabled modes must come after percpu_arena_disabled. */
+	percpu_arena_disabled          = 2,
+
+	percpu_arena_mode_names_limit  = 3, /* Used for options processing. */
+	percpu_arena_mode_enabled_base = 3,
+
+	percpu_arena                   = 3,
+	per_phycpu_arena               = 4  /* Hyper threads share arena. */
+} percpu_arena_mode_t;
+
+#define PERCPU_ARENA_ENABLED(m)	((m) >= percpu_arena_mode_enabled_base)
+#define PERCPU_ARENA_DEFAULT	percpu_arena_disabled
+
+/*
+ * When allocation_size >= oversize_threshold, use the dedicated huge arena
+ * (unless have explicitly spicified arena index).  0 disables the feature.
+ */
+#define OVERSIZE_THRESHOLD_DEFAULT (8 << 20)
+
+#endif /* JEMALLOC_INTERNAL_ARENA_TYPES_H */

deps/jemalloc/include/jemalloc/internal/assert.h

@@ -0,0 +1,56 @@
+#include "jemalloc/internal/malloc_io.h"
+#include "jemalloc/internal/util.h"
+
+/*
+ * Define a custom assert() in order to reduce the chances of deadlock during
+ * assertion failure.
+ */
+#ifndef assert
+#define assert(e) do {							\
+	if (unlikely(config_debug && !(e))) {				\
+		malloc_printf(						\
+		    "<jemalloc>: %s:%d: Failed assertion: \"%s\"\n",	\
+		    __FILE__, __LINE__, #e);				\
+		abort();						\
+	}								\
+} while (0)
+#endif
+
+#ifndef not_reached
+#define not_reached() do {						\
+	if (config_debug) {						\
+		malloc_printf(						\
+		    "<jemalloc>: %s:%d: Unreachable code reached\n",	\
+		    __FILE__, __LINE__);				\
+		abort();						\
+	}								\
+	unreachable();							\
+} while (0)
+#endif
+
+#ifndef not_implemented
+#define not_implemented() do {						\
+	if (config_debug) {						\
+		malloc_printf("<jemalloc>: %s:%d: Not implemented\n",	\
+		    __FILE__, __LINE__);				\
+		abort();						\
+	}								\
+} while (0)
+#endif
+
+#ifndef assert_not_implemented
+#define assert_not_implemented(e) do {					\
+	if (unlikely(config_debug && !(e))) {				\
+		not_implemented();					\
+	}								\
+} while (0)
+#endif
+
+/* Use to assert a particular configuration, e.g., cassert(config_debug). */
+#ifndef cassert
+#define cassert(c) do {							\
+	if (unlikely(!(c))) {						\
+		not_reached();						\
+	}								\
+} while (0)
+#endif

deps/jemalloc/include/jemalloc/internal/atomic.h

@@ -0,0 +1,86 @@
+#ifndef JEMALLOC_INTERNAL_ATOMIC_H
+#define JEMALLOC_INTERNAL_ATOMIC_H
+
+#define ATOMIC_INLINE JEMALLOC_ALWAYS_INLINE
+
+#define JEMALLOC_U8_ATOMICS
+#if defined(JEMALLOC_GCC_ATOMIC_ATOMICS)
+#  include "jemalloc/internal/atomic_gcc_atomic.h"
+#  if !defined(JEMALLOC_GCC_U8_ATOMIC_ATOMICS)
+#    undef JEMALLOC_U8_ATOMICS
+#  endif
+#elif defined(JEMALLOC_GCC_SYNC_ATOMICS)
+#  include "jemalloc/internal/atomic_gcc_sync.h"
+#  if !defined(JEMALLOC_GCC_U8_SYNC_ATOMICS)
+#    undef JEMALLOC_U8_ATOMICS
+#  endif
+#elif defined(_MSC_VER)
+#  include "jemalloc/internal/atomic_msvc.h"
+#elif defined(JEMALLOC_C11_ATOMICS)
+#  include "jemalloc/internal/atomic_c11.h"
+#else
+#  error "Don't have atomics implemented on this platform."
+#endif
+
+/*
+ * This header gives more or less a backport of C11 atomics. The user can write
+ * JEMALLOC_GENERATE_ATOMICS(type, short_type, lg_sizeof_type); to generate
+ * counterparts of the C11 atomic functions for type, as so:
+ *   JEMALLOC_GENERATE_ATOMICS(int *, pi, 3);
+ * and then write things like:
+ *   int *some_ptr;
+ *   atomic_pi_t atomic_ptr_to_int;
+ *   atomic_store_pi(&atomic_ptr_to_int, some_ptr, ATOMIC_RELAXED);
+ *   int *prev_value = atomic_exchange_pi(&ptr_to_int, NULL, ATOMIC_ACQ_REL);
+ *   assert(some_ptr == prev_value);
+ * and expect things to work in the obvious way.
+ *
+ * Also included (with naming differences to avoid conflicts with the standard
+ * library):
+ *   atomic_fence(atomic_memory_order_t) (mimics C11's atomic_thread_fence).
+ *   ATOMIC_INIT (mimics C11's ATOMIC_VAR_INIT).
+ */
+
+/*
+ * Pure convenience, so that we don't have to type "atomic_memory_order_"
+ * quite so often.
+ */
+#define ATOMIC_RELAXED atomic_memory_order_relaxed
+#define ATOMIC_ACQUIRE atomic_memory_order_acquire
+#define ATOMIC_RELEASE atomic_memory_order_release
+#define ATOMIC_ACQ_REL atomic_memory_order_acq_rel
+#define ATOMIC_SEQ_CST atomic_memory_order_seq_cst
+
+/*
+ * Not all platforms have 64-bit atomics.  If we do, this #define exposes that
+ * fact.
+ */
+#if (LG_SIZEOF_PTR == 3 || LG_SIZEOF_INT == 3)
+#  define JEMALLOC_ATOMIC_U64
+#endif
+
+JEMALLOC_GENERATE_ATOMICS(void *, p, LG_SIZEOF_PTR)
+
+/*
+ * There's no actual guarantee that sizeof(bool) == 1, but it's true on the only
+ * platform that actually needs to know the size, MSVC.
+ */
+JEMALLOC_GENERATE_ATOMICS(bool, b, 0)
+
+JEMALLOC_GENERATE_INT_ATOMICS(unsigned, u, LG_SIZEOF_INT)
+
+JEMALLOC_GENERATE_INT_ATOMICS(size_t, zu, LG_SIZEOF_PTR)
+
+JEMALLOC_GENERATE_INT_ATOMICS(ssize_t, zd, LG_SIZEOF_PTR)
+
+JEMALLOC_GENERATE_INT_ATOMICS(uint8_t, u8, 0)
+
+JEMALLOC_GENERATE_INT_ATOMICS(uint32_t, u32, 2)
+
+#ifdef JEMALLOC_ATOMIC_U64
+JEMALLOC_GENERATE_INT_ATOMICS(uint64_t, u64, 3)
+#endif
+
+#undef ATOMIC_INLINE
+
+#endif /* JEMALLOC_INTERNAL_ATOMIC_H */

deps/jemalloc/include/jemalloc/internal/atomic_c11.h

@@ -0,0 +1,97 @@
+#ifndef JEMALLOC_INTERNAL_ATOMIC_C11_H
+#define JEMALLOC_INTERNAL_ATOMIC_C11_H
+
+#include <stdatomic.h>
+
+#define ATOMIC_INIT(...) ATOMIC_VAR_INIT(__VA_ARGS__)
+
+#define atomic_memory_order_t memory_order
+#define atomic_memory_order_relaxed memory_order_relaxed
+#define atomic_memory_order_acquire memory_order_acquire
+#define atomic_memory_order_release memory_order_release
+#define atomic_memory_order_acq_rel memory_order_acq_rel
+#define atomic_memory_order_seq_cst memory_order_seq_cst
+
+#define atomic_fence atomic_thread_fence
+
+#define JEMALLOC_GENERATE_ATOMICS(type, short_type,			\
+    /* unused */ lg_size)						\
+typedef _Atomic(type) atomic_##short_type##_t;				\
+									\
+ATOMIC_INLINE type							\
+atomic_load_##short_type(const atomic_##short_type##_t *a,		\
+    atomic_memory_order_t mo) {						\
+	/*								\
+	 * A strict interpretation of the C standard prevents		\
+	 * atomic_load from taking a const argument, but it's		\
+	 * convenient for our purposes. This cast is a workaround.	\
+	 */								\
+	atomic_##short_type##_t* a_nonconst =				\
+	    (atomic_##short_type##_t*)a;				\
+	return atomic_load_explicit(a_nonconst, mo);			\
+}									\
+									\
+ATOMIC_INLINE void							\
+atomic_store_##short_type(atomic_##short_type##_t *a,			\
+    type val, atomic_memory_order_t mo) {				\
+	atomic_store_explicit(a, val, mo);				\
+}									\
+									\
+ATOMIC_INLINE type							\
+atomic_exchange_##short_type(atomic_##short_type##_t *a, type val,	\
+    atomic_memory_order_t mo) {						\
+	return atomic_exchange_explicit(a, val, mo);			\
+}									\
+									\
+ATOMIC_INLINE bool							\
+atomic_compare_exchange_weak_##short_type(atomic_##short_type##_t *a,	\
+    type *expected, type desired, atomic_memory_order_t success_mo,	\
+    atomic_memory_order_t failure_mo) {					\
+	return atomic_compare_exchange_weak_explicit(a, expected,	\
+	    desired, success_mo, failure_mo);				\
+}									\
+									\
+ATOMIC_INLINE bool							\
+atomic_compare_exchange_strong_##short_type(atomic_##short_type##_t *a,	\
+    type *expected, type desired, atomic_memory_order_t success_mo,	\
+    atomic_memory_order_t failure_mo) {					\
+	return atomic_compare_exchange_strong_explicit(a, expected,	\
+	    desired, success_mo, failure_mo);				\
+}
+
+/*
+ * Integral types have some special operations available that non-integral ones
+ * lack.
+ */
+#define JEMALLOC_GENERATE_INT_ATOMICS(type, short_type, 		\
+    /* unused */ lg_size)						\
+JEMALLOC_GENERATE_ATOMICS(type, short_type, /* unused */ lg_size)	\
+									\
+ATOMIC_INLINE type							\
+atomic_fetch_add_##short_type(atomic_##short_type##_t *a,		\
+    type val, atomic_memory_order_t mo) {				\
+	return atomic_fetch_add_explicit(a, val, mo);			\
+}									\
+									\
+ATOMIC_INLINE type							\
+atomic_fetch_sub_##short_type(atomic_##short_type##_t *a,		\
+    type val, atomic_memory_order_t mo) {				\
+	return atomic_fetch_sub_explicit(a, val, mo);			\
+}									\
+ATOMIC_INLINE type							\
+atomic_fetch_and_##short_type(atomic_##short_type##_t *a,		\
+    type val, atomic_memory_order_t mo) {				\
+	return atomic_fetch_and_explicit(a, val, mo);			\
+}									\
+ATOMIC_INLINE type							\
+atomic_fetch_or_##short_type(atomic_##short_type##_t *a,		\
+    type val, atomic_memory_order_t mo) {				\
+	return atomic_fetch_or_explicit(a, val, mo);			\
+}									\
+ATOMIC_INLINE type							\
+atomic_fetch_xor_##short_type(atomic_##short_type##_t *a,		\
+    type val, atomic_memory_order_t mo) {				\
+	return atomic_fetch_xor_explicit(a, val, mo);			\
+}
+
+#endif /* JEMALLOC_INTERNAL_ATOMIC_C11_H */

deps/jemalloc/include/jemalloc/internal/atomic_gcc_atomic.h

@@ -0,0 +1,129 @@
+#ifndef JEMALLOC_INTERNAL_ATOMIC_GCC_ATOMIC_H
+#define JEMALLOC_INTERNAL_ATOMIC_GCC_ATOMIC_H
+
+#include "jemalloc/internal/assert.h"
+
+#define ATOMIC_INIT(...) {__VA_ARGS__}
+
+typedef enum {
+	atomic_memory_order_relaxed,
+	atomic_memory_order_acquire,
+	atomic_memory_order_release,
+	atomic_memory_order_acq_rel,
+	atomic_memory_order_seq_cst
+} atomic_memory_order_t;
+
+ATOMIC_INLINE int
+atomic_enum_to_builtin(atomic_memory_order_t mo) {
+	switch (mo) {
+	case atomic_memory_order_relaxed:
+		return __ATOMIC_RELAXED;
+	case atomic_memory_order_acquire:
+		return __ATOMIC_ACQUIRE;
+	case atomic_memory_order_release:
+		return __ATOMIC_RELEASE;
+	case atomic_memory_order_acq_rel:
+		return __ATOMIC_ACQ_REL;
+	case atomic_memory_order_seq_cst:
+		return __ATOMIC_SEQ_CST;
+	}
+	/* Can't happen; the switch is exhaustive. */
+	not_reached();
+}
+
+ATOMIC_INLINE void
+atomic_fence(atomic_memory_order_t mo) {
+	__atomic_thread_fence(atomic_enum_to_builtin(mo));
+}
+
+#define JEMALLOC_GENERATE_ATOMICS(type, short_type,			\
+    /* unused */ lg_size)						\
+typedef struct {							\
+	type repr;							\
+} atomic_##short_type##_t;						\
+									\
+ATOMIC_INLINE type							\
+atomic_load_##short_type(const atomic_##short_type##_t *a,		\
+    atomic_memory_order_t mo) {						\
+	type result;							\
+	__atomic_load(&a->repr, &result, atomic_enum_to_builtin(mo));	\
+	return result;							\
+}									\
+									\
+ATOMIC_INLINE void							\
+atomic_store_##short_type(atomic_##short_type##_t *a, type val,		\
+    atomic_memory_order_t mo) {						\
+	__atomic_store(&a->repr, &val, atomic_enum_to_builtin(mo));	\
+}									\
+									\
+ATOMIC_INLINE type							\
+atomic_exchange_##short_type(atomic_##short_type##_t *a, type val,	\
+    atomic_memory_order_t mo) {						\
+	type result;							\
+	__atomic_exchange(&a->repr, &val, &result,			\
+	    atomic_enum_to_builtin(mo));				\
+	return result;							\
+}									\
+									\
+ATOMIC_INLINE bool							\
+atomic_compare_exchange_weak_##short_type(atomic_##short_type##_t *a,	\
+    UNUSED type *expected, type desired,				\
+    atomic_memory_order_t success_mo,					\
+    atomic_memory_order_t failure_mo) {					\
+	return __atomic_compare_exchange(&a->repr, expected, &desired,	\
+	    true, atomic_enum_to_builtin(success_mo),			\
+	    atomic_enum_to_builtin(failure_mo));			\
+}									\
+									\
+ATOMIC_INLINE bool							\
+atomic_compare_exchange_strong_##short_type(atomic_##short_type##_t *a,	\
+    UNUSED type *expected, type desired,				\
+    atomic_memory_order_t success_mo,					\
+    atomic_memory_order_t failure_mo) {					\
+	return __atomic_compare_exchange(&a->repr, expected, &desired,	\
+	    false,							\
+	    atomic_enum_to_builtin(success_mo),				\
+	    atomic_enum_to_builtin(failure_mo));			\
+}
+
+
+#define JEMALLOC_GENERATE_INT_ATOMICS(type, short_type,			\
+    /* unused */ lg_size)						\
+JEMALLOC_GENERATE_ATOMICS(type, short_type, /* unused */ lg_size)	\
+									\
+ATOMIC_INLINE type							\
+atomic_fetch_add_##short_type(atomic_##short_type##_t *a, type val,	\
+    atomic_memory_order_t mo) {						\
+	return __atomic_fetch_add(&a->repr, val,			\
+	    atomic_enum_to_builtin(mo));				\
+}									\
+									\
+ATOMIC_INLINE type							\
+atomic_fetch_sub_##short_type(atomic_##short_type##_t *a, type val,	\
+    atomic_memory_order_t mo) {						\
+	return __atomic_fetch_sub(&a->repr, val,			\
+	    atomic_enum_to_builtin(mo));				\
+}									\
+									\
+ATOMIC_INLINE type							\
+atomic_fetch_and_##short_type(atomic_##short_type##_t *a, type val,	\
+    atomic_memory_order_t mo) {						\
+	return __atomic_fetch_and(&a->repr, val,			\
+	    atomic_enum_to_builtin(mo));				\
+}									\
+									\
+ATOMIC_INLINE type							\
+atomic_fetch_or_##short_type(atomic_##short_type##_t *a, type val,	\
+    atomic_memory_order_t mo) {						\
+	return __atomic_fetch_or(&a->repr, val,				\
+	    atomic_enum_to_builtin(mo));				\
+}									\
+									\
+ATOMIC_INLINE type							\
+atomic_fetch_xor_##short_type(atomic_##short_type##_t *a, type val,	\
+    atomic_memory_order_t mo) {						\
+	return __atomic_fetch_xor(&a->repr, val,			\
+	    atomic_enum_to_builtin(mo));				\
+}
+
+#endif /* JEMALLOC_INTERNAL_ATOMIC_GCC_ATOMIC_H */

deps/jemalloc/include/jemalloc/internal/atomic_gcc_sync.h

@@ -0,0 +1,195 @@
+#ifndef JEMALLOC_INTERNAL_ATOMIC_GCC_SYNC_H
+#define JEMALLOC_INTERNAL_ATOMIC_GCC_SYNC_H
+
+#define ATOMIC_INIT(...) {__VA_ARGS__}
+
+typedef enum {
+	atomic_memory_order_relaxed,
+	atomic_memory_order_acquire,
+	atomic_memory_order_release,
+	atomic_memory_order_acq_rel,
+	atomic_memory_order_seq_cst
+} atomic_memory_order_t;
+
+ATOMIC_INLINE void
+atomic_fence(atomic_memory_order_t mo) {
+	/* Easy cases first: no barrier, and full barrier. */
+	if (mo == atomic_memory_order_relaxed) {
+		asm volatile("" ::: "memory");
+		return;
+	}
+	if (mo == atomic_memory_order_seq_cst) {
+		asm volatile("" ::: "memory");
+		__sync_synchronize();
+		asm volatile("" ::: "memory");
+		return;
+	}
+	asm volatile("" ::: "memory");
+#  if defined(__i386__) || defined(__x86_64__)
+	/* This is implicit on x86. */
+#  elif defined(__ppc64__)
+	asm volatile("lwsync");
+#  elif defined(__ppc__)
+	asm volatile("sync");
+#  elif defined(__sparc__) && defined(__arch64__)
+	if (mo == atomic_memory_order_acquire) {
+		asm volatile("membar #LoadLoad | #LoadStore");
+	} else if (mo == atomic_memory_order_release) {
+		asm volatile("membar #LoadStore | #StoreStore");
+	} else {
+		asm volatile("membar #LoadLoad | #LoadStore | #StoreStore");
+	}
+#  else
+	__sync_synchronize();
+#  endif
+	asm volatile("" ::: "memory");
+}
+
+/*
+ * A correct implementation of seq_cst loads and stores on weakly ordered
+ * architectures could do either of the following:
+ *   1. store() is weak-fence -> store -> strong fence, load() is load ->
+ *      strong-fence.
+ *   2. store() is strong-fence -> store, load() is strong-fence -> load ->
+ *      weak-fence.
+ * The tricky thing is, load() and store() above can be the load or store
+ * portions of a gcc __sync builtin, so we have to follow GCC's lead, which
+ * means going with strategy 2.
+ * On strongly ordered architectures, the natural strategy is to stick a strong
+ * fence after seq_cst stores, and have naked loads.  So we want the strong
+ * fences in different places on different architectures.
+ * atomic_pre_sc_load_fence and atomic_post_sc_store_fence allow us to
+ * accomplish this.
+ */
+
+ATOMIC_INLINE void
+atomic_pre_sc_load_fence() {
+#  if defined(__i386__) || defined(__x86_64__) ||			\
+    (defined(__sparc__) && defined(__arch64__))
+	atomic_fence(atomic_memory_order_relaxed);
+#  else
+	atomic_fence(atomic_memory_order_seq_cst);
+#  endif
+}
+
+ATOMIC_INLINE void
+atomic_post_sc_store_fence() {
+#  if defined(__i386__) || defined(__x86_64__) ||			\
+    (defined(__sparc__) && defined(__arch64__))
+	atomic_fence(atomic_memory_order_seq_cst);
+#  else
+	atomic_fence(atomic_memory_order_relaxed);
+#  endif
+
+}
+
+#define JEMALLOC_GENERATE_ATOMICS(type, short_type,			\
+    /* unused */ lg_size)						\
+typedef struct {							\
+	type volatile repr;						\
+} atomic_##short_type##_t;						\
+									\
+ATOMIC_INLINE type							\
+atomic_load_##short_type(const atomic_##short_type##_t *a,		\
+    atomic_memory_order_t mo) {						\
+	if (mo == atomic_memory_order_seq_cst) {			\
+		atomic_pre_sc_load_fence();				\
+	}								\
+	type result = a->repr;						\
+	if (mo != atomic_memory_order_relaxed) {			\
+		atomic_fence(atomic_memory_order_acquire);		\
+	}								\
+	return result;							\
+}									\
+									\
+ATOMIC_INLINE void							\
+atomic_store_##short_type(atomic_##short_type##_t *a,			\
+    type val, atomic_memory_order_t mo) {				\
+	if (mo != atomic_memory_order_relaxed) {			\
+		atomic_fence(atomic_memory_order_release);		\
+	}								\
+	a->repr = val;							\
+	if (mo == atomic_memory_order_seq_cst) {			\
+		atomic_post_sc_store_fence();				\
+	}								\
+}									\
+									\
+ATOMIC_INLINE type							\
+atomic_exchange_##short_type(atomic_##short_type##_t *a, type val, \
+    atomic_memory_order_t mo) {                  					 \
+	/*								\
+	 * Because of FreeBSD, we care about gcc 4.2, which doesn't have\
+	 * an atomic exchange builtin.  We fake it with a CAS loop.	\
+	 */								\
+	while (true) {							\
+		type old = a->repr;					\
+		if (__sync_bool_compare_and_swap(&a->repr, old, val)) {	\
+			return old;					\
+		}							\
+	}								\
+}									\
+									\
+ATOMIC_INLINE bool							\
+atomic_compare_exchange_weak_##short_type(atomic_##short_type##_t *a,	\
+    type *expected, type desired,                                     \
+    atomic_memory_order_t success_mo,                          \
+    atomic_memory_order_t failure_mo) {				                \
+	type prev = __sync_val_compare_and_swap(&a->repr, *expected,	\
+	    desired);							\
+	if (prev == *expected) {					\
+		return true;						\
+	} else {							\
+		*expected = prev;					\
+		return false;						\
+	}								\
+}									\
+ATOMIC_INLINE bool							\
+atomic_compare_exchange_strong_##short_type(atomic_##short_type##_t *a,	\
+    type *expected, type desired,                                       \
+    atomic_memory_order_t success_mo,                            \
+    atomic_memory_order_t failure_mo) {                          \
+	type prev = __sync_val_compare_and_swap(&a->repr, *expected,	\
+	    desired);							\
+	if (prev == *expected) {					\
+		return true;						\
+	} else {							\
+		*expected = prev;					\
+		return false;						\
+	}								\
+}
+
+#define JEMALLOC_GENERATE_INT_ATOMICS(type, short_type,			\
+    /* unused */ lg_size)						\
+JEMALLOC_GENERATE_ATOMICS(type, short_type, /* unused */ lg_size)	\
+									\
+ATOMIC_INLINE type							\
+atomic_fetch_add_##short_type(atomic_##short_type##_t *a, type val,	\
+    atomic_memory_order_t mo) {						\
+	return __sync_fetch_and_add(&a->repr, val);			\
+}									\
+									\
+ATOMIC_INLINE type							\
+atomic_fetch_sub_##short_type(atomic_##short_type##_t *a, type val,	\
+    atomic_memory_order_t mo) {						\
+	return __sync_fetch_and_sub(&a->repr, val);			\
+}									\
+									\
+ATOMIC_INLINE type							\
+atomic_fetch_and_##short_type(atomic_##short_type##_t *a, type val,	\
+    atomic_memory_order_t mo) {						\
+	return __sync_fetch_and_and(&a->repr, val);			\
+}									\
+									\
+ATOMIC_INLINE type							\
+atomic_fetch_or_##short_type(atomic_##short_type##_t *a, type val,	\
+    atomic_memory_order_t mo) {						\
+	return __sync_fetch_and_or(&a->repr, val);			\
+}									\
+									\
+ATOMIC_INLINE type							\
+atomic_fetch_xor_##short_type(atomic_##short_type##_t *a, type val,	\
+    atomic_memory_order_t mo) {						\
+	return __sync_fetch_and_xor(&a->repr, val);			\
+}
+
+#endif /* JEMALLOC_INTERNAL_ATOMIC_GCC_SYNC_H */

deps/jemalloc/include/jemalloc/internal/atomic_msvc.h

@@ -0,0 +1,158 @@
+#ifndef JEMALLOC_INTERNAL_ATOMIC_MSVC_H
+#define JEMALLOC_INTERNAL_ATOMIC_MSVC_H
+
+#define ATOMIC_INIT(...) {__VA_ARGS__}
+
+typedef enum {
+	atomic_memory_order_relaxed,
+	atomic_memory_order_acquire,
+	atomic_memory_order_release,
+	atomic_memory_order_acq_rel,
+	atomic_memory_order_seq_cst
+} atomic_memory_order_t;
+
+typedef char atomic_repr_0_t;
+typedef short atomic_repr_1_t;
+typedef long atomic_repr_2_t;
+typedef __int64 atomic_repr_3_t;
+
+ATOMIC_INLINE void
+atomic_fence(atomic_memory_order_t mo) {
+	_ReadWriteBarrier();
+#  if defined(_M_ARM) || defined(_M_ARM64)
+	/* ARM needs a barrier for everything but relaxed. */
+	if (mo != atomic_memory_order_relaxed) {
+		MemoryBarrier();
+	}
+#  elif defined(_M_IX86) || defined (_M_X64)
+	/* x86 needs a barrier only for seq_cst. */
+	if (mo == atomic_memory_order_seq_cst) {
+		MemoryBarrier();
+	}
+#  else
+#  error "Don't know how to create atomics for this platform for MSVC."
+#  endif
+	_ReadWriteBarrier();
+}
+
+#define ATOMIC_INTERLOCKED_REPR(lg_size) atomic_repr_ ## lg_size ## _t
+
+#define ATOMIC_CONCAT(a, b) ATOMIC_RAW_CONCAT(a, b)
+#define ATOMIC_RAW_CONCAT(a, b) a ## b
+
+#define ATOMIC_INTERLOCKED_NAME(base_name, lg_size) ATOMIC_CONCAT(	\
+    base_name, ATOMIC_INTERLOCKED_SUFFIX(lg_size))
+
+#define ATOMIC_INTERLOCKED_SUFFIX(lg_size)				\
+    ATOMIC_CONCAT(ATOMIC_INTERLOCKED_SUFFIX_, lg_size)
+
+#define ATOMIC_INTERLOCKED_SUFFIX_0 8
+#define ATOMIC_INTERLOCKED_SUFFIX_1 16
+#define ATOMIC_INTERLOCKED_SUFFIX_2
+#define ATOMIC_INTERLOCKED_SUFFIX_3 64
+
+#define JEMALLOC_GENERATE_ATOMICS(type, short_type, lg_size)		\
+typedef struct {							\
+	ATOMIC_INTERLOCKED_REPR(lg_size) repr;				\
+} atomic_##short_type##_t;						\
+									\
+ATOMIC_INLINE type							\
+atomic_load_##short_type(const atomic_##short_type##_t *a,		\
+    atomic_memory_order_t mo) {						\
+	ATOMIC_INTERLOCKED_REPR(lg_size) ret = a->repr;			\
+	if (mo != atomic_memory_order_relaxed) {			\
+		atomic_fence(atomic_memory_order_acquire);		\
+	}								\
+	return (type) ret;						\
+}									\
+									\
+ATOMIC_INLINE void							\
+atomic_store_##short_type(atomic_##short_type##_t *a,			\
+    type val, atomic_memory_order_t mo) {				\
+	if (mo != atomic_memory_order_relaxed) {			\
+		atomic_fence(atomic_memory_order_release);		\
+	}								\
+	a->repr = (ATOMIC_INTERLOCKED_REPR(lg_size)) val;		\
+	if (mo == atomic_memory_order_seq_cst) {			\
+		atomic_fence(atomic_memory_order_seq_cst);		\
+	}								\
+}									\
+									\
+ATOMIC_INLINE type							\
+atomic_exchange_##short_type(atomic_##short_type##_t *a, type val,	\
+    atomic_memory_order_t mo) {						\
+	return (type)ATOMIC_INTERLOCKED_NAME(_InterlockedExchange,	\
+	    lg_size)(&a->repr, (ATOMIC_INTERLOCKED_REPR(lg_size))val);	\
+}									\
+									\
+ATOMIC_INLINE bool							\
+atomic_compare_exchange_weak_##short_type(atomic_##short_type##_t *a,	\
+    type *expected, type desired, atomic_memory_order_t success_mo,	\
+    atomic_memory_order_t failure_mo) {					\
+	ATOMIC_INTERLOCKED_REPR(lg_size) e =				\
+	    (ATOMIC_INTERLOCKED_REPR(lg_size))*expected;		\
+	ATOMIC_INTERLOCKED_REPR(lg_size) d =				\
+	    (ATOMIC_INTERLOCKED_REPR(lg_size))desired;			\
+	ATOMIC_INTERLOCKED_REPR(lg_size) old =				\
+	    ATOMIC_INTERLOCKED_NAME(_InterlockedCompareExchange, 	\
+		lg_size)(&a->repr, d, e);				\
+	if (old == e) {							\
+		return true;						\
+	} else {							\
+		*expected = (type)old;					\
+		return false;						\
+	}								\
+}									\
+									\
+ATOMIC_INLINE bool							\
+atomic_compare_exchange_strong_##short_type(atomic_##short_type##_t *a,	\
+    type *expected, type desired, atomic_memory_order_t success_mo,	\
+    atomic_memory_order_t failure_mo) {					\
+	/* We implement the weak version with strong semantics. */	\
+	return atomic_compare_exchange_weak_##short_type(a, expected,	\
+	    desired, success_mo, failure_mo);				\
+}
+
+
+#define JEMALLOC_GENERATE_INT_ATOMICS(type, short_type, lg_size)	\
+JEMALLOC_GENERATE_ATOMICS(type, short_type, lg_size)			\
+									\
+ATOMIC_INLINE type							\
+atomic_fetch_add_##short_type(atomic_##short_type##_t *a,		\
+    type val, atomic_memory_order_t mo) {				\
+	return (type)ATOMIC_INTERLOCKED_NAME(_InterlockedExchangeAdd,	\
+	    lg_size)(&a->repr, (ATOMIC_INTERLOCKED_REPR(lg_size))val);	\
+}									\
+									\
+ATOMIC_INLINE type							\
+atomic_fetch_sub_##short_type(atomic_##short_type##_t *a,		\
+    type val, atomic_memory_order_t mo) {				\
+	/*								\
+	 * MSVC warns on negation of unsigned operands, but for us it	\
+	 * gives exactly the right semantics (MAX_TYPE + 1 - operand).	\
+	 */								\
+	__pragma(warning(push))						\
+	__pragma(warning(disable: 4146))				\
+	return atomic_fetch_add_##short_type(a, -val, mo);		\
+	__pragma(warning(pop))						\
+}									\
+ATOMIC_INLINE type							\
+atomic_fetch_and_##short_type(atomic_##short_type##_t *a,		\
+    type val, atomic_memory_order_t mo) {				\
+	return (type)ATOMIC_INTERLOCKED_NAME(_InterlockedAnd, lg_size)(	\
+	    &a->repr, (ATOMIC_INTERLOCKED_REPR(lg_size))val);		\
+}									\
+ATOMIC_INLINE type							\
+atomic_fetch_or_##short_type(atomic_##short_type##_t *a,		\
+    type val, atomic_memory_order_t mo) {				\
+	return (type)ATOMIC_INTERLOCKED_NAME(_InterlockedOr, lg_size)(	\
+	    &a->repr, (ATOMIC_INTERLOCKED_REPR(lg_size))val);		\
+}									\
+ATOMIC_INLINE type							\
+atomic_fetch_xor_##short_type(atomic_##short_type##_t *a,		\
+    type val, atomic_memory_order_t mo) {				\
+	return (type)ATOMIC_INTERLOCKED_NAME(_InterlockedXor, lg_size)(	\
+	    &a->repr, (ATOMIC_INTERLOCKED_REPR(lg_size))val);		\
+}
+
+#endif /* JEMALLOC_INTERNAL_ATOMIC_MSVC_H */

deps/jemalloc/include/jemalloc/internal/background_thread_externs.h

@@ -0,0 +1,32 @@
+#ifndef JEMALLOC_INTERNAL_BACKGROUND_THREAD_EXTERNS_H
+#define JEMALLOC_INTERNAL_BACKGROUND_THREAD_EXTERNS_H
+
+extern bool opt_background_thread;
+extern size_t opt_max_background_threads;
+extern malloc_mutex_t background_thread_lock;
+extern atomic_b_t background_thread_enabled_state;
+extern size_t n_background_threads;
+extern size_t max_background_threads;
+extern background_thread_info_t *background_thread_info;
+
+bool background_thread_create(tsd_t *tsd, unsigned arena_ind);
+bool background_threads_enable(tsd_t *tsd);
+bool background_threads_disable(tsd_t *tsd);
+void background_thread_interval_check(tsdn_t *tsdn, arena_t *arena,
+    arena_decay_t *decay, size_t npages_new);
+void background_thread_prefork0(tsdn_t *tsdn);
+void background_thread_prefork1(tsdn_t *tsdn);
+void background_thread_postfork_parent(tsdn_t *tsdn);
+void background_thread_postfork_child(tsdn_t *tsdn);
+bool background_thread_stats_read(tsdn_t *tsdn,
+    background_thread_stats_t *stats);
+void background_thread_ctl_init(tsdn_t *tsdn);
+
+#ifdef JEMALLOC_PTHREAD_CREATE_WRAPPER
+extern int pthread_create_wrapper(pthread_t *__restrict, const pthread_attr_t *,
+    void *(*)(void *), void *__restrict);
+#endif
+bool background_thread_boot0(void);
+bool background_thread_boot1(tsdn_t *tsdn);
+
+#endif /* JEMALLOC_INTERNAL_BACKGROUND_THREAD_EXTERNS_H */

deps/jemalloc/include/jemalloc/internal/background_thread_inlines.h

@@ -0,0 +1,62 @@
+#ifndef JEMALLOC_INTERNAL_BACKGROUND_THREAD_INLINES_H
+#define JEMALLOC_INTERNAL_BACKGROUND_THREAD_INLINES_H
+
+JEMALLOC_ALWAYS_INLINE bool
+background_thread_enabled(void) {
+	return atomic_load_b(&background_thread_enabled_state, ATOMIC_RELAXED);
+}
+
+JEMALLOC_ALWAYS_INLINE void
+background_thread_enabled_set(tsdn_t *tsdn, bool state) {
+	malloc_mutex_assert_owner(tsdn, &background_thread_lock);
+	atomic_store_b(&background_thread_enabled_state, state, ATOMIC_RELAXED);
+}
+
+JEMALLOC_ALWAYS_INLINE background_thread_info_t *
+arena_background_thread_info_get(arena_t *arena) {
+	unsigned arena_ind = arena_ind_get(arena);
+	return &background_thread_info[arena_ind % max_background_threads];
+}
+
+JEMALLOC_ALWAYS_INLINE background_thread_info_t *
+background_thread_info_get(size_t ind) {
+	return &background_thread_info[ind % max_background_threads];
+}
+
+JEMALLOC_ALWAYS_INLINE uint64_t
+background_thread_wakeup_time_get(background_thread_info_t *info) {
+	uint64_t next_wakeup = nstime_ns(&info->next_wakeup);
+	assert(atomic_load_b(&info->indefinite_sleep, ATOMIC_ACQUIRE) ==
+	    (next_wakeup == BACKGROUND_THREAD_INDEFINITE_SLEEP));
+	return next_wakeup;
+}
+
+JEMALLOC_ALWAYS_INLINE void
+background_thread_wakeup_time_set(tsdn_t *tsdn, background_thread_info_t *info,
+    uint64_t wakeup_time) {
+	malloc_mutex_assert_owner(tsdn, &info->mtx);
+	atomic_store_b(&info->indefinite_sleep,
+	    wakeup_time == BACKGROUND_THREAD_INDEFINITE_SLEEP, ATOMIC_RELEASE);
+	nstime_init(&info->next_wakeup, wakeup_time);
+}
+
+JEMALLOC_ALWAYS_INLINE bool
+background_thread_indefinite_sleep(background_thread_info_t *info) {
+	return atomic_load_b(&info->indefinite_sleep, ATOMIC_ACQUIRE);
+}
+
+JEMALLOC_ALWAYS_INLINE void
+arena_background_thread_inactivity_check(tsdn_t *tsdn, arena_t *arena,
+    bool is_background_thread) {
+	if (!background_thread_enabled() || is_background_thread) {
+		return;
+	}
+	background_thread_info_t *info =
+	    arena_background_thread_info_get(arena);
+	if (background_thread_indefinite_sleep(info)) {
+		background_thread_interval_check(tsdn, arena,
+		    &arena->decay_dirty, 0);
+	}
+}
+
+#endif /* JEMALLOC_INTERNAL_BACKGROUND_THREAD_INLINES_H */

deps/jemalloc/include/jemalloc/internal/background_thread_structs.h

@@ -0,0 +1,54 @@
+#ifndef JEMALLOC_INTERNAL_BACKGROUND_THREAD_STRUCTS_H
+#define JEMALLOC_INTERNAL_BACKGROUND_THREAD_STRUCTS_H
+
+/* This file really combines "structs" and "types", but only transitionally. */
+
+#if defined(JEMALLOC_BACKGROUND_THREAD) || defined(JEMALLOC_LAZY_LOCK)
+#  define JEMALLOC_PTHREAD_CREATE_WRAPPER
+#endif
+
+#define BACKGROUND_THREAD_INDEFINITE_SLEEP UINT64_MAX
+#define MAX_BACKGROUND_THREAD_LIMIT MALLOCX_ARENA_LIMIT
+#define DEFAULT_NUM_BACKGROUND_THREAD 4
+
+typedef enum {
+	background_thread_stopped,
+	background_thread_started,
+	/* Thread waits on the global lock when paused (for arena_reset). */
+	background_thread_paused,
+} background_thread_state_t;
+
+struct background_thread_info_s {
+#ifdef JEMALLOC_BACKGROUND_THREAD
+	/* Background thread is pthread specific. */
+	pthread_t		thread;
+	pthread_cond_t		cond;
+#endif
+	malloc_mutex_t		mtx;
+	background_thread_state_t	state;
+	/* When true, it means no wakeup scheduled. */
+	atomic_b_t		indefinite_sleep;
+	/* Next scheduled wakeup time (absolute time in ns). */
+	nstime_t		next_wakeup;
+	/*
+	 *  Since the last background thread run, newly added number of pages
+	 *  that need to be purged by the next wakeup.  This is adjusted on
+	 *  epoch advance, and is used to determine whether we should signal the
+	 *  background thread to wake up earlier.
+	 */
+	size_t			npages_to_purge_new;
+	/* Stats: total number of runs since started. */
+	uint64_t		tot_n_runs;
+	/* Stats: total sleep time since started. */
+	nstime_t		tot_sleep_time;
+};
+typedef struct background_thread_info_s background_thread_info_t;
+
+struct background_thread_stats_s {
+	size_t num_threads;
+	uint64_t num_runs;
+	nstime_t run_interval;
+};
+typedef struct background_thread_stats_s background_thread_stats_t;
+
+#endif /* JEMALLOC_INTERNAL_BACKGROUND_THREAD_STRUCTS_H */

deps/jemalloc/include/jemalloc/internal/base.h

@@ -0,0 +1,26 @@
+/******************************************************************************/
+#ifdef JEMALLOC_H_TYPES
+
+#endif /* JEMALLOC_H_TYPES */
+/******************************************************************************/
+#ifdef JEMALLOC_H_STRUCTS
+
+#endif /* JEMALLOC_H_STRUCTS */
+/******************************************************************************/
+#ifdef JEMALLOC_H_EXTERNS
+
+void	*base_alloc(size_t size);
+void	*base_calloc(size_t number, size_t size);
+extent_node_t *base_node_alloc(void);
+void	base_node_dealloc(extent_node_t *node);
+bool	base_boot(void);
+void	base_prefork(void);
+void	base_postfork_parent(void);
+void	base_postfork_child(void);
+
+#endif /* JEMALLOC_H_EXTERNS */
+/******************************************************************************/
+#ifdef JEMALLOC_H_INLINES
+
+#endif /* JEMALLOC_H_INLINES */
+/******************************************************************************/

deps/jemalloc/include/jemalloc/internal/base_externs.h

@@ -0,0 +1,22 @@
+#ifndef JEMALLOC_INTERNAL_BASE_EXTERNS_H
+#define JEMALLOC_INTERNAL_BASE_EXTERNS_H
+
+extern metadata_thp_mode_t opt_metadata_thp;
+extern const char *metadata_thp_mode_names[];
+
+base_t *b0get(void);
+base_t *base_new(tsdn_t *tsdn, unsigned ind, extent_hooks_t *extent_hooks);
+void base_delete(tsdn_t *tsdn, base_t *base);
+extent_hooks_t *base_extent_hooks_get(base_t *base);
+extent_hooks_t *base_extent_hooks_set(base_t *base,
+    extent_hooks_t *extent_hooks);
+void *base_alloc(tsdn_t *tsdn, base_t *base, size_t size, size_t alignment);
+extent_t *base_alloc_extent(tsdn_t *tsdn, base_t *base);
+void base_stats_get(tsdn_t *tsdn, base_t *base, size_t *allocated,
+    size_t *resident, size_t *mapped, size_t *n_thp);
+void base_prefork(tsdn_t *tsdn, base_t *base);
+void base_postfork_parent(tsdn_t *tsdn, base_t *base);
+void base_postfork_child(tsdn_t *tsdn, base_t *base);
+bool base_boot(tsdn_t *tsdn);
+
+#endif /* JEMALLOC_INTERNAL_BASE_EXTERNS_H */

deps/jemalloc/include/jemalloc/internal/base_inlines.h

@@ -0,0 +1,13 @@
+#ifndef JEMALLOC_INTERNAL_BASE_INLINES_H
+#define JEMALLOC_INTERNAL_BASE_INLINES_H
+
+static inline unsigned
+base_ind_get(const base_t *base) {
+	return base->ind;
+}
+
+static inline bool
+metadata_thp_enabled(void) {
+	return (opt_metadata_thp != metadata_thp_disabled);
+}
+#endif /* JEMALLOC_INTERNAL_BASE_INLINES_H */

deps/jemalloc/include/jemalloc/internal/base_structs.h

@@ -0,0 +1,59 @@
+#ifndef JEMALLOC_INTERNAL_BASE_STRUCTS_H
+#define JEMALLOC_INTERNAL_BASE_STRUCTS_H
+
+#include "jemalloc/internal/jemalloc_internal_types.h"
+#include "jemalloc/internal/mutex.h"
+#include "jemalloc/internal/sc.h"
+
+/* Embedded at the beginning of every block of base-managed virtual memory. */
+struct base_block_s {
+	/* Total size of block's virtual memory mapping. */
+	size_t		size;
+
+	/* Next block in list of base's blocks. */
+	base_block_t	*next;
+
+	/* Tracks unused trailing space. */
+	extent_t	extent;
+};
+
+struct base_s {
+	/* Associated arena's index within the arenas array. */
+	unsigned	ind;
+
+	/*
+	 * User-configurable extent hook functions.  Points to an
+	 * extent_hooks_t.
+	 */
+	atomic_p_t	extent_hooks;
+
+	/* Protects base_alloc() and base_stats_get() operations. */
+	malloc_mutex_t	mtx;
+
+	/* Using THP when true (metadata_thp auto mode). */
+	bool		auto_thp_switched;
+	/*
+	 * Most recent size class in the series of increasingly large base
+	 * extents.  Logarithmic spacing between subsequent allocations ensures
+	 * that the total number of distinct mappings remains small.
+	 */
+	pszind_t	pind_last;
+
+	/* Serial number generation state. */
+	size_t		extent_sn_next;
+
+	/* Chain of all blocks associated with base. */
+	base_block_t	*blocks;
+
+	/* Heap of extents that track unused trailing space within blocks. */
+	extent_heap_t	avail[SC_NSIZES];
+
+	/* Stats, only maintained if config_stats. */
+	size_t		allocated;
+	size_t		resident;
+	size_t		mapped;
+	/* Number of THP regions touched. */
+	size_t		n_thp;
+};
+
+#endif /* JEMALLOC_INTERNAL_BASE_STRUCTS_H */

deps/jemalloc/include/jemalloc/internal/base_types.h

@@ -0,0 +1,33 @@
+#ifndef JEMALLOC_INTERNAL_BASE_TYPES_H
+#define JEMALLOC_INTERNAL_BASE_TYPES_H
+
+typedef struct base_block_s base_block_t;
+typedef struct base_s base_t;
+
+#define METADATA_THP_DEFAULT metadata_thp_disabled
+
+/*
+ * In auto mode, arenas switch to huge pages for the base allocator on the
+ * second base block.  a0 switches to thp on the 5th block (after 20 megabytes
+ * of metadata), since more metadata (e.g. rtree nodes) come from a0's base.
+ */
+
+#define BASE_AUTO_THP_THRESHOLD    2
+#define BASE_AUTO_THP_THRESHOLD_A0 5
+
+typedef enum {
+	metadata_thp_disabled   = 0,
+	/*
+	 * Lazily enable hugepage for metadata. To avoid high RSS caused by THP
+	 * + low usage arena (i.e. THP becomes a significant percentage), the
+	 * "auto" option only starts using THP after a base allocator used up
+	 * the first THP region.  Starting from the second hugepage (in a single
+	 * arena), "auto" behaves the same as "always", i.e. madvise hugepage
+	 * right away.
+	 */
+	metadata_thp_auto       = 1,
+	metadata_thp_always     = 2,
+	metadata_thp_mode_limit = 3
+} metadata_thp_mode_t;
+
+#endif /* JEMALLOC_INTERNAL_BASE_TYPES_H */

deps/jemalloc/include/jemalloc/internal/bin.h

@@ -0,0 +1,123 @@
+#ifndef JEMALLOC_INTERNAL_BIN_H
+#define JEMALLOC_INTERNAL_BIN_H
+
+#include "jemalloc/internal/bin_stats.h"
+#include "jemalloc/internal/bin_types.h"
+#include "jemalloc/internal/extent_types.h"
+#include "jemalloc/internal/extent_structs.h"
+#include "jemalloc/internal/mutex.h"
+#include "jemalloc/internal/sc.h"
+
+/*
+ * A bin contains a set of extents that are currently being used for slab
+ * allocations.
+ */
+
+/*
+ * Read-only information associated with each element of arena_t's bins array
+ * is stored separately, partly to reduce memory usage (only one copy, rather
+ * than one per arena), but mainly to avoid false cacheline sharing.
+ *
+ * Each slab has the following layout:
+ *
+ *   /--------------------\
+ *   | region 0           |
+ *   |--------------------|
+ *   | region 1           |
+ *   |--------------------|
+ *   | ...                |
+ *   | ...                |
+ *   | ...                |
+ *   |--------------------|
+ *   | region nregs-1     |
+ *   \--------------------/
+ */
+typedef struct bin_info_s bin_info_t;
+struct bin_info_s {
+	/* Size of regions in a slab for this bin's size class. */
+	size_t			reg_size;
+
+	/* Total size of a slab for this bin's size class. */
+	size_t			slab_size;
+
+	/* Total number of regions in a slab for this bin's size class. */
+	uint32_t		nregs;
+
+	/* Number of sharded bins in each arena for this size class. */
+	uint32_t		n_shards;
+
+	/*
+	 * Metadata used to manipulate bitmaps for slabs associated with this
+	 * bin.
+	 */
+	bitmap_info_t		bitmap_info;
+};
+
+extern bin_info_t bin_infos[SC_NBINS];
+
+typedef struct bin_s bin_t;
+struct bin_s {
+	/* All operations on bin_t fields require lock ownership. */
+	malloc_mutex_t		lock;
+
+	/*
+	 * Current slab being used to service allocations of this bin's size
+	 * class.  slabcur is independent of slabs_{nonfull,full}; whenever
+	 * slabcur is reassigned, the previous slab must be deallocated or
+	 * inserted into slabs_{nonfull,full}.
+	 */
+	extent_t		*slabcur;
+
+	/*
+	 * Heap of non-full slabs.  This heap is used to assure that new
+	 * allocations come from the non-full slab that is oldest/lowest in
+	 * memory.
+	 */
+	extent_heap_t		slabs_nonfull;
+
+	/* List used to track full slabs. */
+	extent_list_t		slabs_full;
+
+	/* Bin statistics. */
+	bin_stats_t	stats;
+};
+
+/* A set of sharded bins of the same size class. */
+typedef struct bins_s bins_t;
+struct bins_s {
+	/* Sharded bins.  Dynamically sized. */
+	bin_t *bin_shards;
+};
+
+void bin_shard_sizes_boot(unsigned bin_shards[SC_NBINS]);
+bool bin_update_shard_size(unsigned bin_shards[SC_NBINS], size_t start_size,
+    size_t end_size, size_t nshards);
+void bin_boot(sc_data_t *sc_data, unsigned bin_shard_sizes[SC_NBINS]);
+
+/* Initializes a bin to empty.  Returns true on error. */
+bool bin_init(bin_t *bin);
+
+/* Forking. */
+void bin_prefork(tsdn_t *tsdn, bin_t *bin);
+void bin_postfork_parent(tsdn_t *tsdn, bin_t *bin);
+void bin_postfork_child(tsdn_t *tsdn, bin_t *bin);
+
+/* Stats. */
+static inline void
+bin_stats_merge(tsdn_t *tsdn, bin_stats_t *dst_bin_stats, bin_t *bin) {
+	malloc_mutex_lock(tsdn, &bin->lock);
+	malloc_mutex_prof_accum(tsdn, &dst_bin_stats->mutex_data, &bin->lock);
+	dst_bin_stats->nmalloc += bin->stats.nmalloc;
+	dst_bin_stats->ndalloc += bin->stats.ndalloc;
+	dst_bin_stats->nrequests += bin->stats.nrequests;
+	dst_bin_stats->curregs += bin->stats.curregs;
+	dst_bin_stats->nfills += bin->stats.nfills;
+	dst_bin_stats->nflushes += bin->stats.nflushes;
+	dst_bin_stats->nslabs += bin->stats.nslabs;
+	dst_bin_stats->reslabs += bin->stats.reslabs;
+	dst_bin_stats->curslabs += bin->stats.curslabs;
+	dst_bin_stats->nonfull_slabs += bin->stats.nonfull_slabs;
+	malloc_mutex_unlock(tsdn, &bin->lock);
+}
+
+#endif /* JEMALLOC_INTERNAL_BIN_H */

deps/jemalloc/include/jemalloc/internal/bin_stats.h

@@ -0,0 +1,54 @@
+#ifndef JEMALLOC_INTERNAL_BIN_STATS_H
+#define JEMALLOC_INTERNAL_BIN_STATS_H
+
+#include "jemalloc/internal/mutex_prof.h"
+
+typedef struct bin_stats_s bin_stats_t;
+struct bin_stats_s {
+	/*
+	 * Total number of allocation/deallocation requests served directly by
+	 * the bin.  Note that tcache may allocate an object, then recycle it
+	 * many times, resulting many increments to nrequests, but only one
+	 * each to nmalloc and ndalloc.
+	 */
+	uint64_t	nmalloc;
+	uint64_t	ndalloc;
+
+	/*
+	 * Number of allocation requests that correspond to the size of this
+	 * bin.  This includes requests served by tcache, though tcache only
+	 * periodically merges into this counter.
+	 */
+	uint64_t	nrequests;
+
+	/*
+	 * Current number of regions of this size class, including regions
+	 * currently cached by tcache.
+	 */
+	size_t		curregs;
+
+	/* Number of tcache fills from this bin. */
+	uint64_t	nfills;
+
+	/* Number of tcache flushes to this bin. */
+	uint64_t	nflushes;
+
+	/* Total number of slabs created for this bin's size class. */
+	uint64_t	nslabs;
+
+	/*
+	 * Total number of slabs reused by extracting them from the slabs heap
+	 * for this bin's size class.
+	 */
+	uint64_t	reslabs;
+
+	/* Current number of slabs in this bin. */
+	size_t		curslabs;
+
+	/* Current size of nonfull slabs heap in this bin. */
+	size_t		nonfull_slabs;
+
+	mutex_prof_data_t mutex_data;
+};
+
+#endif /* JEMALLOC_INTERNAL_BIN_STATS_H */

deps/jemalloc/include/jemalloc/internal/bin_types.h

@@ -0,0 +1,17 @@
+#ifndef JEMALLOC_INTERNAL_BIN_TYPES_H
+#define JEMALLOC_INTERNAL_BIN_TYPES_H
+
+#include "jemalloc/internal/sc.h"
+
+#define BIN_SHARDS_MAX (1 << EXTENT_BITS_BINSHARD_WIDTH)
+#define N_BIN_SHARDS_DEFAULT 1
+
+/* Used in TSD static initializer only. Real init in arena_bind(). */
+#define TSD_BINSHARDS_ZERO_INITIALIZER {{UINT8_MAX}}
+
+typedef struct tsd_binshards_s tsd_binshards_t;
+struct tsd_binshards_s {
+	uint8_t binshard[SC_NBINS];
+};
+
+#endif /* JEMALLOC_INTERNAL_BIN_TYPES_H */

deps/jemalloc/include/jemalloc/internal/bit_util.h

@@ -0,0 +1,239 @@
+#ifndef JEMALLOC_INTERNAL_BIT_UTIL_H
+#define JEMALLOC_INTERNAL_BIT_UTIL_H
+
+#include "jemalloc/internal/assert.h"
+
+#define BIT_UTIL_INLINE static inline
+
+/* Sanity check. */
+#if !defined(JEMALLOC_INTERNAL_FFSLL) || !defined(JEMALLOC_INTERNAL_FFSL) \
+    || !defined(JEMALLOC_INTERNAL_FFS)
+#  error JEMALLOC_INTERNAL_FFS{,L,LL} should have been defined by configure
+#endif
+
+
+BIT_UTIL_INLINE unsigned
+ffs_llu(unsigned long long bitmap) {
+	return JEMALLOC_INTERNAL_FFSLL(bitmap);
+}
+
+BIT_UTIL_INLINE unsigned
+ffs_lu(unsigned long bitmap) {
+	return JEMALLOC_INTERNAL_FFSL(bitmap);
+}
+
+BIT_UTIL_INLINE unsigned
+ffs_u(unsigned bitmap) {
+	return JEMALLOC_INTERNAL_FFS(bitmap);
+}
+
+#ifdef JEMALLOC_INTERNAL_POPCOUNTL
+BIT_UTIL_INLINE unsigned
+popcount_lu(unsigned long bitmap) {
+  return JEMALLOC_INTERNAL_POPCOUNTL(bitmap);
+}
+#endif
+
+/*
+ * Clears first unset bit in bitmap, and returns
+ * place of bit.  bitmap *must not* be 0.
+ */
+
+BIT_UTIL_INLINE size_t
+cfs_lu(unsigned long* bitmap) {
+	size_t bit = ffs_lu(*bitmap) - 1;
+	*bitmap ^= ZU(1) << bit;
+	return bit;
+}
+
+BIT_UTIL_INLINE unsigned
+ffs_zu(size_t bitmap) {
+#if LG_SIZEOF_PTR == LG_SIZEOF_INT
+	return ffs_u(bitmap);
+#elif LG_SIZEOF_PTR == LG_SIZEOF_LONG
+	return ffs_lu(bitmap);
+#elif LG_SIZEOF_PTR == LG_SIZEOF_LONG_LONG
+	return ffs_llu(bitmap);
+#else
+#error No implementation for size_t ffs()
+#endif
+}
+
+BIT_UTIL_INLINE unsigned
+ffs_u64(uint64_t bitmap) {
+#if LG_SIZEOF_LONG == 3
+	return ffs_lu(bitmap);
+#elif LG_SIZEOF_LONG_LONG == 3
+	return ffs_llu(bitmap);
+#else
+#error No implementation for 64-bit ffs()
+#endif
+}
+
+BIT_UTIL_INLINE unsigned
+ffs_u32(uint32_t bitmap) {
+#if LG_SIZEOF_INT == 2
+	return ffs_u(bitmap);
+#else
+#error No implementation for 32-bit ffs()
+#endif
+	return ffs_u(bitmap);
+}
+
+BIT_UTIL_INLINE uint64_t
+pow2_ceil_u64(uint64_t x) {
+#if (defined(__amd64__) || defined(__x86_64__) || defined(JEMALLOC_HAVE_BUILTIN_CLZ))
+	if(unlikely(x <= 1)) {
+		return x;
+	}
+	size_t msb_on_index;
+#if (defined(__amd64__) || defined(__x86_64__))
+	asm ("bsrq %1, %0"
+			: "=r"(msb_on_index) // Outputs.
+			: "r"(x-1)           // Inputs.
+		);
+#elif (defined(JEMALLOC_HAVE_BUILTIN_CLZ))
+	msb_on_index = (63 ^ __builtin_clzll(x - 1));
+#endif
+	assert(msb_on_index < 63);
+	return 1ULL << (msb_on_index + 1);
+#else
+	x--;
+	x |= x >> 1;
+	x |= x >> 2;
+	x |= x >> 4;
+	x |= x >> 8;
+	x |= x >> 16;
+	x |= x >> 32;
+	x++;
+	return x;
+#endif
+}
+
+BIT_UTIL_INLINE uint32_t
+pow2_ceil_u32(uint32_t x) {
+#if ((defined(__i386__) || defined(JEMALLOC_HAVE_BUILTIN_CLZ)) && (!defined(__s390__)))
+	if(unlikely(x <= 1)) {
+		return x;
+	}
+	size_t msb_on_index;
+#if (defined(__i386__))
+	asm ("bsr %1, %0"
+			: "=r"(msb_on_index) // Outputs.
+			: "r"(x-1)           // Inputs.
+		);
+#elif (defined(JEMALLOC_HAVE_BUILTIN_CLZ))
+	msb_on_index = (31 ^ __builtin_clz(x - 1));
+#endif
+	assert(msb_on_index < 31);
+	return 1U << (msb_on_index + 1);
+#else
+	x--;
+	x |= x >> 1;
+	x |= x >> 2;
+	x |= x >> 4;
+	x |= x >> 8;
+	x |= x >> 16;
+	x++;
+	return x;
+#endif
+}
+
+/* Compute the smallest power of 2 that is >= x. */
+BIT_UTIL_INLINE size_t
+pow2_ceil_zu(size_t x) {
+#if (LG_SIZEOF_PTR == 3)
+	return pow2_ceil_u64(x);
+#else
+	return pow2_ceil_u32(x);
+#endif
+}
+
+#if (defined(__i386__) || defined(__amd64__) || defined(__x86_64__))
+BIT_UTIL_INLINE unsigned
+lg_floor(size_t x) {
+	size_t ret;
+	assert(x != 0);
+
+	asm ("bsr %1, %0"
+	    : "=r"(ret) // Outputs.
+	    : "r"(x)    // Inputs.
+	    );
+	assert(ret < UINT_MAX);
+	return (unsigned)ret;
+}
+#elif (defined(_MSC_VER))
+BIT_UTIL_INLINE unsigned
+lg_floor(size_t x) {
+	unsigned long ret;
+
+	assert(x != 0);
+
+#if (LG_SIZEOF_PTR == 3)
+	_BitScanReverse64(&ret, x);
+#elif (LG_SIZEOF_PTR == 2)
+	_BitScanReverse(&ret, x);
+#else
+#  error "Unsupported type size for lg_floor()"
+#endif
+	assert(ret < UINT_MAX);
+	return (unsigned)ret;
+}
+#elif (defined(JEMALLOC_HAVE_BUILTIN_CLZ))
+BIT_UTIL_INLINE unsigned
+lg_floor(size_t x) {
+	assert(x != 0);
+
+#if (LG_SIZEOF_PTR == LG_SIZEOF_INT)
+	return ((8 << LG_SIZEOF_PTR) - 1) - __builtin_clz(x);
+#elif (LG_SIZEOF_PTR == LG_SIZEOF_LONG)
+	return ((8 << LG_SIZEOF_PTR) - 1) - __builtin_clzl(x);
+#else
+#  error "Unsupported type size for lg_floor()"
+#endif
+}
+#else
+BIT_UTIL_INLINE unsigned
+lg_floor(size_t x) {
+	assert(x != 0);
+
+	x |= (x >> 1);
+	x |= (x >> 2);
+	x |= (x >> 4);
+	x |= (x >> 8);
+	x |= (x >> 16);
+#if (LG_SIZEOF_PTR == 3)
+	x |= (x >> 32);
+#endif
+	if (x == SIZE_T_MAX) {
+		return (8 << LG_SIZEOF_PTR) - 1;
+	}
+	x++;
+	return ffs_zu(x) - 2;
+}
+#endif
+
+BIT_UTIL_INLINE unsigned
+lg_ceil(size_t x) {
+	return lg_floor(x) + ((x & (x - 1)) == 0 ? 0 : 1);
+}
+
+#undef BIT_UTIL_INLINE
+
+/* A compile-time version of lg_floor and lg_ceil. */
+#define LG_FLOOR_1(x) 0
+#define LG_FLOOR_2(x) (x < (1ULL << 1) ? LG_FLOOR_1(x) : 1 + LG_FLOOR_1(x >> 1))
+#define LG_FLOOR_4(x) (x < (1ULL << 2) ? LG_FLOOR_2(x) : 2 + LG_FLOOR_2(x >> 2))
+#define LG_FLOOR_8(x) (x < (1ULL << 4) ? LG_FLOOR_4(x) : 4 + LG_FLOOR_4(x >> 4))
+#define LG_FLOOR_16(x) (x < (1ULL << 8) ? LG_FLOOR_8(x) : 8 + LG_FLOOR_8(x >> 8))
+#define LG_FLOOR_32(x) (x < (1ULL << 16) ? LG_FLOOR_16(x) : 16 + LG_FLOOR_16(x >> 16))
+#define LG_FLOOR_64(x) (x < (1ULL << 32) ? LG_FLOOR_32(x) : 32 + LG_FLOOR_32(x >> 32))
+#if LG_SIZEOF_PTR == 2
+#  define LG_FLOOR(x) LG_FLOOR_32((x))
+#else
+#  define LG_FLOOR(x) LG_FLOOR_64((x))
+#endif
+
+#define LG_CEIL(x) (LG_FLOOR(x) + (((x) & ((x) - 1)) == 0 ? 0 : 1))
+
+#endif /* JEMALLOC_INTERNAL_BIT_UTIL_H */

deps/jemalloc/include/jemalloc/internal/bitmap.h

@@ -0,0 +1,369 @@
+#ifndef JEMALLOC_INTERNAL_BITMAP_H
+#define JEMALLOC_INTERNAL_BITMAP_H
+
+#include "jemalloc/internal/arena_types.h"
+#include "jemalloc/internal/bit_util.h"
+#include "jemalloc/internal/sc.h"
+
+typedef unsigned long bitmap_t;
+#define LG_SIZEOF_BITMAP	LG_SIZEOF_LONG
+
+/* Maximum bitmap bit count is 2^LG_BITMAP_MAXBITS. */
+#if LG_SLAB_MAXREGS > LG_CEIL(SC_NSIZES)
+/* Maximum bitmap bit count is determined by maximum regions per slab. */
+#  define LG_BITMAP_MAXBITS	LG_SLAB_MAXREGS
+#else
+/* Maximum bitmap bit count is determined by number of extent size classes. */
+#  define LG_BITMAP_MAXBITS	LG_CEIL(SC_NSIZES)
+#endif
+#define BITMAP_MAXBITS		(ZU(1) << LG_BITMAP_MAXBITS)
+
+/* Number of bits per group. */
+#define LG_BITMAP_GROUP_NBITS		(LG_SIZEOF_BITMAP + 3)
+#define BITMAP_GROUP_NBITS		(1U << LG_BITMAP_GROUP_NBITS)
+#define BITMAP_GROUP_NBITS_MASK		(BITMAP_GROUP_NBITS-1)
+
+/*
+ * Do some analysis on how big the bitmap is before we use a tree.  For a brute
+ * force linear search, if we would have to call ffs_lu() more than 2^3 times,
+ * use a tree instead.
+ */
+#if LG_BITMAP_MAXBITS - LG_BITMAP_GROUP_NBITS > 3
+#  define BITMAP_USE_TREE
+#endif
+
+/* Number of groups required to store a given number of bits. */
+#define BITMAP_BITS2GROUPS(nbits)					\
+    (((nbits) + BITMAP_GROUP_NBITS_MASK) >> LG_BITMAP_GROUP_NBITS)
+
+/*
+ * Number of groups required at a particular level for a given number of bits.
+ */
+#define BITMAP_GROUPS_L0(nbits)						\
+    BITMAP_BITS2GROUPS(nbits)
+#define BITMAP_GROUPS_L1(nbits)						\
+    BITMAP_BITS2GROUPS(BITMAP_BITS2GROUPS(nbits))
+#define BITMAP_GROUPS_L2(nbits)						\
+    BITMAP_BITS2GROUPS(BITMAP_BITS2GROUPS(BITMAP_BITS2GROUPS((nbits))))
+#define BITMAP_GROUPS_L3(nbits)						\
+    BITMAP_BITS2GROUPS(BITMAP_BITS2GROUPS(BITMAP_BITS2GROUPS(		\
+	BITMAP_BITS2GROUPS((nbits)))))
+#define BITMAP_GROUPS_L4(nbits)						\
+    BITMAP_BITS2GROUPS(BITMAP_BITS2GROUPS(BITMAP_BITS2GROUPS(		\
+	BITMAP_BITS2GROUPS(BITMAP_BITS2GROUPS((nbits))))))
+
+/*
+ * Assuming the number of levels, number of groups required for a given number
+ * of bits.
+ */
+#define BITMAP_GROUPS_1_LEVEL(nbits)					\
+    BITMAP_GROUPS_L0(nbits)
+#define BITMAP_GROUPS_2_LEVEL(nbits)					\
+    (BITMAP_GROUPS_1_LEVEL(nbits) + BITMAP_GROUPS_L1(nbits))
+#define BITMAP_GROUPS_3_LEVEL(nbits)					\
+    (BITMAP_GROUPS_2_LEVEL(nbits) + BITMAP_GROUPS_L2(nbits))
+#define BITMAP_GROUPS_4_LEVEL(nbits)					\
+    (BITMAP_GROUPS_3_LEVEL(nbits) + BITMAP_GROUPS_L3(nbits))
+#define BITMAP_GROUPS_5_LEVEL(nbits)					\
+    (BITMAP_GROUPS_4_LEVEL(nbits) + BITMAP_GROUPS_L4(nbits))
+
+/*
+ * Maximum number of groups required to support LG_BITMAP_MAXBITS.
+ */
+#ifdef BITMAP_USE_TREE
+
+#if LG_BITMAP_MAXBITS <= LG_BITMAP_GROUP_NBITS
+#  define BITMAP_GROUPS(nbits)	BITMAP_GROUPS_1_LEVEL(nbits)
+#  define BITMAP_GROUPS_MAX	BITMAP_GROUPS_1_LEVEL(BITMAP_MAXBITS)
+#elif LG_BITMAP_MAXBITS <= LG_BITMAP_GROUP_NBITS * 2
+#  define BITMAP_GROUPS(nbits)	BITMAP_GROUPS_2_LEVEL(nbits)
+#  define BITMAP_GROUPS_MAX	BITMAP_GROUPS_2_LEVEL(BITMAP_MAXBITS)
+#elif LG_BITMAP_MAXBITS <= LG_BITMAP_GROUP_NBITS * 3
+#  define BITMAP_GROUPS(nbits)	BITMAP_GROUPS_3_LEVEL(nbits)
+#  define BITMAP_GROUPS_MAX	BITMAP_GROUPS_3_LEVEL(BITMAP_MAXBITS)
+#elif LG_BITMAP_MAXBITS <= LG_BITMAP_GROUP_NBITS * 4
+#  define BITMAP_GROUPS(nbits)	BITMAP_GROUPS_4_LEVEL(nbits)
+#  define BITMAP_GROUPS_MAX	BITMAP_GROUPS_4_LEVEL(BITMAP_MAXBITS)
+#elif LG_BITMAP_MAXBITS <= LG_BITMAP_GROUP_NBITS * 5
+#  define BITMAP_GROUPS(nbits)	BITMAP_GROUPS_5_LEVEL(nbits)
+#  define BITMAP_GROUPS_MAX	BITMAP_GROUPS_5_LEVEL(BITMAP_MAXBITS)
+#else
+#  error "Unsupported bitmap size"
+#endif
+
+/*
+ * Maximum number of levels possible.  This could be statically computed based
+ * on LG_BITMAP_MAXBITS:
+ *
+ * #define BITMAP_MAX_LEVELS \
+ *     (LG_BITMAP_MAXBITS / LG_SIZEOF_BITMAP) \
+ *     + !!(LG_BITMAP_MAXBITS % LG_SIZEOF_BITMAP)
+ *
+ * However, that would not allow the generic BITMAP_INFO_INITIALIZER() macro, so
+ * instead hardcode BITMAP_MAX_LEVELS to the largest number supported by the
+ * various cascading macros.  The only additional cost this incurs is some
+ * unused trailing entries in bitmap_info_t structures; the bitmaps themselves
+ * are not impacted.
+ */
+#define BITMAP_MAX_LEVELS	5
+
+#define BITMAP_INFO_INITIALIZER(nbits) {				\
+	/* nbits. */							\
+	nbits,								\
+	/* nlevels. */							\
+	(BITMAP_GROUPS_L0(nbits) > BITMAP_GROUPS_L1(nbits)) +		\
+	    (BITMAP_GROUPS_L1(nbits) > BITMAP_GROUPS_L2(nbits)) +	\
+	    (BITMAP_GROUPS_L2(nbits) > BITMAP_GROUPS_L3(nbits)) +	\
+	    (BITMAP_GROUPS_L3(nbits) > BITMAP_GROUPS_L4(nbits)) + 1,	\
+	/* levels. */							\
+	{								\
+		{0},							\
+		{BITMAP_GROUPS_L0(nbits)},				\
+		{BITMAP_GROUPS_L1(nbits) + BITMAP_GROUPS_L0(nbits)},	\
+		{BITMAP_GROUPS_L2(nbits) + BITMAP_GROUPS_L1(nbits) +	\
+		    BITMAP_GROUPS_L0(nbits)},				\
+		{BITMAP_GROUPS_L3(nbits) + BITMAP_GROUPS_L2(nbits) +	\
+		    BITMAP_GROUPS_L1(nbits) + BITMAP_GROUPS_L0(nbits)},	\
+		{BITMAP_GROUPS_L4(nbits) + BITMAP_GROUPS_L3(nbits) +	\
+		     BITMAP_GROUPS_L2(nbits) + BITMAP_GROUPS_L1(nbits)	\
+		     + BITMAP_GROUPS_L0(nbits)}				\
+	}								\
+}
+
+#else /* BITMAP_USE_TREE */
+
+#define BITMAP_GROUPS(nbits)	BITMAP_BITS2GROUPS(nbits)
+#define BITMAP_GROUPS_MAX	BITMAP_BITS2GROUPS(BITMAP_MAXBITS)
+
+#define BITMAP_INFO_INITIALIZER(nbits) {				\
+	/* nbits. */							\
+	nbits,								\
+	/* ngroups. */							\
+	BITMAP_BITS2GROUPS(nbits)					\
+}
+
+#endif /* BITMAP_USE_TREE */
+
+typedef struct bitmap_level_s {
+	/* Offset of this level's groups within the array of groups. */
+	size_t group_offset;
+} bitmap_level_t;
+
+typedef struct bitmap_info_s {
+	/* Logical number of bits in bitmap (stored at bottom level). */
+	size_t nbits;
+
+#ifdef BITMAP_USE_TREE
+	/* Number of levels necessary for nbits. */
+	unsigned nlevels;
+
+	/*
+	 * Only the first (nlevels+1) elements are used, and levels are ordered
+	 * bottom to top (e.g. the bottom level is stored in levels[0]).
+	 */
+	bitmap_level_t levels[BITMAP_MAX_LEVELS+1];
+#else /* BITMAP_USE_TREE */
+	/* Number of groups necessary for nbits. */
+	size_t ngroups;
+#endif /* BITMAP_USE_TREE */
+} bitmap_info_t;
+
+void bitmap_info_init(bitmap_info_t *binfo, size_t nbits);
+void bitmap_init(bitmap_t *bitmap, const bitmap_info_t *binfo, bool fill);
+size_t bitmap_size(const bitmap_info_t *binfo);
+
+static inline bool
+bitmap_full(bitmap_t *bitmap, const bitmap_info_t *binfo) {
+#ifdef BITMAP_USE_TREE
+	size_t rgoff = binfo->levels[binfo->nlevels].group_offset - 1;
+	bitmap_t rg = bitmap[rgoff];
+	/* The bitmap is full iff the root group is 0. */
+	return (rg == 0);
+#else
+	size_t i;
+
+	for (i = 0; i < binfo->ngroups; i++) {
+		if (bitmap[i] != 0) {
+			return false;
+		}
+	}
+	return true;
+#endif
+}
+
+static inline bool
+bitmap_get(bitmap_t *bitmap, const bitmap_info_t *binfo, size_t bit) {
+	size_t goff;
+	bitmap_t g;
+
+	assert(bit < binfo->nbits);
+	goff = bit >> LG_BITMAP_GROUP_NBITS;
+	g = bitmap[goff];
+	return !(g & (ZU(1) << (bit & BITMAP_GROUP_NBITS_MASK)));
+}
+
+static inline void
+bitmap_set(bitmap_t *bitmap, const bitmap_info_t *binfo, size_t bit) {
+	size_t goff;
+	bitmap_t *gp;
+	bitmap_t g;
+
+	assert(bit < binfo->nbits);
+	assert(!bitmap_get(bitmap, binfo, bit));
+	goff = bit >> LG_BITMAP_GROUP_NBITS;
+	gp = &bitmap[goff];
+	g = *gp;
+	assert(g & (ZU(1) << (bit & BITMAP_GROUP_NBITS_MASK)));
+	g ^= ZU(1) << (bit & BITMAP_GROUP_NBITS_MASK);
+	*gp = g;
+	assert(bitmap_get(bitmap, binfo, bit));
+#ifdef BITMAP_USE_TREE
+	/* Propagate group state transitions up the tree. */
+	if (g == 0) {
+		unsigned i;
+		for (i = 1; i < binfo->nlevels; i++) {
+			bit = goff;
+			goff = bit >> LG_BITMAP_GROUP_NBITS;
+			gp = &bitmap[binfo->levels[i].group_offset + goff];
+			g = *gp;
+			assert(g & (ZU(1) << (bit & BITMAP_GROUP_NBITS_MASK)));
+			g ^= ZU(1) << (bit & BITMAP_GROUP_NBITS_MASK);
+			*gp = g;
+			if (g != 0) {
+				break;
+			}
+		}
+	}
+#endif
+}
+
+/* ffu: find first unset >= bit. */
+static inline size_t
+bitmap_ffu(const bitmap_t *bitmap, const bitmap_info_t *binfo, size_t min_bit) {
+	assert(min_bit < binfo->nbits);
+
+#ifdef BITMAP_USE_TREE
+	size_t bit = 0;
+	for (unsigned level = binfo->nlevels; level--;) {
+		size_t lg_bits_per_group = (LG_BITMAP_GROUP_NBITS * (level +
+		    1));
+		bitmap_t group = bitmap[binfo->levels[level].group_offset + (bit
+		    >> lg_bits_per_group)];
+		unsigned group_nmask = (unsigned)(((min_bit > bit) ? (min_bit -
+		    bit) : 0) >> (lg_bits_per_group - LG_BITMAP_GROUP_NBITS));
+		assert(group_nmask <= BITMAP_GROUP_NBITS);
+		bitmap_t group_mask = ~((1LU << group_nmask) - 1);
+		bitmap_t group_masked = group & group_mask;
+		if (group_masked == 0LU) {
+			if (group == 0LU) {
+				return binfo->nbits;
+			}
+			/*
+			 * min_bit was preceded by one or more unset bits in
+			 * this group, but there are no other unset bits in this
+			 * group.  Try again starting at the first bit of the
+			 * next sibling.  This will recurse at most once per
+			 * non-root level.
+			 */
+			size_t sib_base = bit + (ZU(1) << lg_bits_per_group);
+			assert(sib_base > min_bit);
+			assert(sib_base > bit);
+			if (sib_base >= binfo->nbits) {
+				return binfo->nbits;
+			}
+			return bitmap_ffu(bitmap, binfo, sib_base);
+		}
+		bit += ((size_t)(ffs_lu(group_masked) - 1)) <<
+		    (lg_bits_per_group - LG_BITMAP_GROUP_NBITS);
+	}
+	assert(bit >= min_bit);
+	assert(bit < binfo->nbits);
+	return bit;
+#else
+	size_t i = min_bit >> LG_BITMAP_GROUP_NBITS;
+	bitmap_t g = bitmap[i] & ~((1LU << (min_bit & BITMAP_GROUP_NBITS_MASK))
+	    - 1);
+	size_t bit;
+	do {
+		bit = ffs_lu(g);
+		if (bit != 0) {
+			return (i << LG_BITMAP_GROUP_NBITS) + (bit - 1);
+		}
+		i++;
+		g = bitmap[i];
+	} while (i < binfo->ngroups);
+	return binfo->nbits;
+#endif
+}
+
+/* sfu: set first unset. */
+static inline size_t
+bitmap_sfu(bitmap_t *bitmap, const bitmap_info_t *binfo) {
+	size_t bit;
+	bitmap_t g;
+	unsigned i;
+
+	assert(!bitmap_full(bitmap, binfo));
+
+#ifdef BITMAP_USE_TREE
+	i = binfo->nlevels - 1;
+	g = bitmap[binfo->levels[i].group_offset];
+	bit = ffs_lu(g) - 1;
+	while (i > 0) {
+		i--;
+		g = bitmap[binfo->levels[i].group_offset + bit];
+		bit = (bit << LG_BITMAP_GROUP_NBITS) + (ffs_lu(g) - 1);
+	}
+#else
+	i = 0;
+	g = bitmap[0];
+	while ((bit = ffs_lu(g)) == 0) {
+		i++;
+		g = bitmap[i];
+	}
+	bit = (i << LG_BITMAP_GROUP_NBITS) + (bit - 1);
+#endif
+	bitmap_set(bitmap, binfo, bit);
+	return bit;
+}
+
+static inline void
+bitmap_unset(bitmap_t *bitmap, const bitmap_info_t *binfo, size_t bit) {
+	size_t goff;
+	bitmap_t *gp;
+	bitmap_t g;
+	UNUSED bool propagate;
+
+	assert(bit < binfo->nbits);
+	assert(bitmap_get(bitmap, binfo, bit));
+	goff = bit >> LG_BITMAP_GROUP_NBITS;
+	gp = &bitmap[goff];
+	g = *gp;
+	propagate = (g == 0);
+	assert((g & (ZU(1) << (bit & BITMAP_GROUP_NBITS_MASK))) == 0);
+	g ^= ZU(1) << (bit & BITMAP_GROUP_NBITS_MASK);
+	*gp = g;
+	assert(!bitmap_get(bitmap, binfo, bit));
+#ifdef BITMAP_USE_TREE
+	/* Propagate group state transitions up the tree. */
+	if (propagate) {
+		unsigned i;
+		for (i = 1; i < binfo->nlevels; i++) {
+			bit = goff;
+			goff = bit >> LG_BITMAP_GROUP_NBITS;
+			gp = &bitmap[binfo->levels[i].group_offset + goff];
+			g = *gp;
+			propagate = (g == 0);
+			assert((g & (ZU(1) << (bit & BITMAP_GROUP_NBITS_MASK)))
+			    == 0);
+			g ^= ZU(1) << (bit & BITMAP_GROUP_NBITS_MASK);
+			*gp = g;
+			if (!propagate) {
+				break;
+			}
+		}
+	}
+#endif /* BITMAP_USE_TREE */
+}
+
+#endif /* JEMALLOC_INTERNAL_BITMAP_H */

deps/jemalloc/include/jemalloc/internal/cache_bin.h

@@ -0,0 +1,131 @@
+#ifndef JEMALLOC_INTERNAL_CACHE_BIN_H
+#define JEMALLOC_INTERNAL_CACHE_BIN_H
+
+#include "jemalloc/internal/ql.h"
+
+/*
+ * The cache_bins are the mechanism that the tcache and the arena use to
+ * communicate.  The tcache fills from and flushes to the arena by passing a
+ * cache_bin_t to fill/flush.  When the arena needs to pull stats from the
+ * tcaches associated with it, it does so by iterating over its
+ * cache_bin_array_descriptor_t objects and reading out per-bin stats it
+ * contains.  This makes it so that the arena need not know about the existence
+ * of the tcache at all.
+ */
+
+
+/*
+ * The count of the number of cached allocations in a bin.  We make this signed
+ * so that negative numbers can encode "invalid" states (e.g. a low water mark
+ * of -1 for a cache that has been depleted).
+ */
+typedef int32_t cache_bin_sz_t;
+
+typedef struct cache_bin_stats_s cache_bin_stats_t;
+struct cache_bin_stats_s {
+	/*
+	 * Number of allocation requests that corresponded to the size of this
+	 * bin.
+	 */
+	uint64_t nrequests;
+};
+
+/*
+ * Read-only information associated with each element of tcache_t's tbins array
+ * is stored separately, mainly to reduce memory usage.
+ */
+typedef struct cache_bin_info_s cache_bin_info_t;
+struct cache_bin_info_s {
+	/* Upper limit on ncached. */
+	cache_bin_sz_t ncached_max;
+};
+
+typedef struct cache_bin_s cache_bin_t;
+struct cache_bin_s {
+	/* Min # cached since last GC. */
+	cache_bin_sz_t low_water;
+	/* # of cached objects. */
+	cache_bin_sz_t ncached;
+	/*
+	 * ncached and stats are both modified frequently.  Let's keep them
+	 * close so that they have a higher chance of being on the same
+	 * cacheline, thus less write-backs.
+	 */
+	cache_bin_stats_t tstats;
+	/*
+	 * Stack of available objects.
+	 *
+	 * To make use of adjacent cacheline prefetch, the items in the avail
+	 * stack goes to higher address for newer allocations.  avail points
+	 * just above the available space, which means that
+	 * avail[-ncached, ... -1] are available items and the lowest item will
+	 * be allocated first.
+	 */
+	void **avail;
+};
+
+typedef struct cache_bin_array_descriptor_s cache_bin_array_descriptor_t;
+struct cache_bin_array_descriptor_s {
+	/*
+	 * The arena keeps a list of the cache bins associated with it, for
+	 * stats collection.
+	 */
+	ql_elm(cache_bin_array_descriptor_t) link;
+	/* Pointers to the tcache bins. */
+	cache_bin_t *bins_small;
+	cache_bin_t *bins_large;
+};
+
+static inline void
+cache_bin_array_descriptor_init(cache_bin_array_descriptor_t *descriptor,
+    cache_bin_t *bins_small, cache_bin_t *bins_large) {
+	ql_elm_new(descriptor, link);
+	descriptor->bins_small = bins_small;
+	descriptor->bins_large = bins_large;
+}
+
+JEMALLOC_ALWAYS_INLINE void *
+cache_bin_alloc_easy(cache_bin_t *bin, bool *success) {
+	void *ret;
+
+	bin->ncached--;
+
+	/*
+	 * Check for both bin->ncached == 0 and ncached < low_water
+	 * in a single branch.
+	 */
+	if (unlikely(bin->ncached <= bin->low_water)) {
+		bin->low_water = bin->ncached;
+		if (bin->ncached == -1) {
+			bin->ncached = 0;
+			*success = false;
+			return NULL;
+		}
+	}
+
+	/*
+	 * success (instead of ret) should be checked upon the return of this
+	 * function.  We avoid checking (ret == NULL) because there is never a
+	 * null stored on the avail stack (which is unknown to the compiler),
+	 * and eagerly checking ret would cause pipeline stall (waiting for the
+	 * cacheline).
+	 */
+	*success = true;
+	ret = *(bin->avail - (bin->ncached + 1));
+
+	return ret;
+}
+
+JEMALLOC_ALWAYS_INLINE bool
+cache_bin_dalloc_easy(cache_bin_t *bin, cache_bin_info_t *bin_info, void *ptr) {
+	if (unlikely(bin->ncached == bin_info->ncached_max)) {
+		return false;
+	}
+	assert(bin->ncached < bin_info->ncached_max);
+	bin->ncached++;
+	*(bin->avail - bin->ncached) = ptr;
+
+	return true;
+}
+
+#endif /* JEMALLOC_INTERNAL_CACHE_BIN_H */

deps/jemalloc/include/jemalloc/internal/chunk.h

@@ -0,0 +1,63 @@
+/******************************************************************************/
+#ifdef JEMALLOC_H_TYPES
+
+/*
+ * Size and alignment of memory chunks that are allocated by the OS's virtual
+ * memory system.
+ */
+#define	LG_CHUNK_DEFAULT	22
+
+/* Return the chunk address for allocation address a. */
+#define	CHUNK_ADDR2BASE(a)						\
+	((void *)((uintptr_t)(a) & ~chunksize_mask))
+
+/* Return the chunk offset of address a. */
+#define	CHUNK_ADDR2OFFSET(a)						\
+	((size_t)((uintptr_t)(a) & chunksize_mask))
+
+/* Return the smallest chunk multiple that is >= s. */
+#define	CHUNK_CEILING(s)						\
+	(((s) + chunksize_mask) & ~chunksize_mask)
+
+#endif /* JEMALLOC_H_TYPES */
+/******************************************************************************/
+#ifdef JEMALLOC_H_STRUCTS
+
+#endif /* JEMALLOC_H_STRUCTS */
+/******************************************************************************/
+#ifdef JEMALLOC_H_EXTERNS
+
+extern size_t		opt_lg_chunk;
+extern const char	*opt_dss;
+
+/* Protects stats_chunks; currently not used for any other purpose. */
+extern malloc_mutex_t	chunks_mtx;
+/* Chunk statistics. */
+extern chunk_stats_t	stats_chunks;
+
+extern rtree_t		*chunks_rtree;
+
+extern size_t		chunksize;
+extern size_t		chunksize_mask; /* (chunksize - 1). */
+extern size_t		chunk_npages;
+extern size_t		map_bias; /* Number of arena chunk header pages. */
+extern size_t		arena_maxclass; /* Max size class for arenas. */
+
+void	*chunk_alloc(size_t size, size_t alignment, bool base, bool *zero,
+    dss_prec_t dss_prec);
+void	chunk_unmap(void *chunk, size_t size);
+void	chunk_dealloc(void *chunk, size_t size, bool unmap);
+bool	chunk_boot(void);
+void	chunk_prefork(void);
+void	chunk_postfork_parent(void);
+void	chunk_postfork_child(void);
+
+#endif /* JEMALLOC_H_EXTERNS */
+/******************************************************************************/
+#ifdef JEMALLOC_H_INLINES
+
+#endif /* JEMALLOC_H_INLINES */
+/******************************************************************************/
+
+#include "jemalloc/internal/chunk_dss.h"
+#include "jemalloc/internal/chunk_mmap.h"

deps/jemalloc/include/jemalloc/internal/chunk_dss.h

@@ -0,0 +1,38 @@
+/******************************************************************************/
+#ifdef JEMALLOC_H_TYPES
+
+typedef enum {
+	dss_prec_disabled  = 0,
+	dss_prec_primary   = 1,
+	dss_prec_secondary = 2,
+
+	dss_prec_limit     = 3
+} dss_prec_t;
+#define	DSS_PREC_DEFAULT	dss_prec_secondary
+#define	DSS_DEFAULT		"secondary"
+
+#endif /* JEMALLOC_H_TYPES */
+/******************************************************************************/
+#ifdef JEMALLOC_H_STRUCTS
+
+extern const char *dss_prec_names[];
+
+#endif /* JEMALLOC_H_STRUCTS */
+/******************************************************************************/
+#ifdef JEMALLOC_H_EXTERNS
+
+dss_prec_t	chunk_dss_prec_get(void);
+bool	chunk_dss_prec_set(dss_prec_t dss_prec);
+void	*chunk_alloc_dss(size_t size, size_t alignment, bool *zero);
+bool	chunk_in_dss(void *chunk);
+bool	chunk_dss_boot(void);
+void	chunk_dss_prefork(void);
+void	chunk_dss_postfork_parent(void);
+void	chunk_dss_postfork_child(void);
+
+#endif /* JEMALLOC_H_EXTERNS */
+/******************************************************************************/
+#ifdef JEMALLOC_H_INLINES
+
+#endif /* JEMALLOC_H_INLINES */
+/******************************************************************************/

deps/jemalloc/include/jemalloc/internal/chunk_mmap.h

@@ -0,0 +1,22 @@
+/******************************************************************************/
+#ifdef JEMALLOC_H_TYPES
+
+#endif /* JEMALLOC_H_TYPES */
+/******************************************************************************/
+#ifdef JEMALLOC_H_STRUCTS
+
+#endif /* JEMALLOC_H_STRUCTS */
+/******************************************************************************/
+#ifdef JEMALLOC_H_EXTERNS
+
+bool	pages_purge(void *addr, size_t length);
+
+void	*chunk_alloc_mmap(size_t size, size_t alignment, bool *zero);
+bool	chunk_dealloc_mmap(void *chunk, size_t size);
+
+#endif /* JEMALLOC_H_EXTERNS */
+/******************************************************************************/
+#ifdef JEMALLOC_H_INLINES
+
+#endif /* JEMALLOC_H_INLINES */
+/******************************************************************************/

deps/jemalloc/include/jemalloc/internal/ckh.h

@@ -0,0 +1,101 @@
+#ifndef JEMALLOC_INTERNAL_CKH_H
+#define JEMALLOC_INTERNAL_CKH_H
+
+#include "jemalloc/internal/tsd.h"
+
+/* Cuckoo hashing implementation.  Skip to the end for the interface. */
+
+/******************************************************************************/
+/* INTERNAL DEFINITIONS -- IGNORE */
+/******************************************************************************/
+
+/* Maintain counters used to get an idea of performance. */
+/* #define CKH_COUNT */
+/* Print counter values in ckh_delete() (requires CKH_COUNT). */
+/* #define CKH_VERBOSE */
+
+/*
+ * There are 2^LG_CKH_BUCKET_CELLS cells in each hash table bucket.  Try to fit
+ * one bucket per L1 cache line.
+ */
+#define LG_CKH_BUCKET_CELLS (LG_CACHELINE - LG_SIZEOF_PTR - 1)
+
+/* Typedefs to allow easy function pointer passing. */
+typedef void ckh_hash_t (const void *, size_t[2]);
+typedef bool ckh_keycomp_t (const void *, const void *);
+
+/* Hash table cell. */
+typedef struct {
+	const void *key;
+	const void *data;
+} ckhc_t;
+
+/* The hash table itself. */
+typedef struct {
+#ifdef CKH_COUNT
+	/* Counters used to get an idea of performance. */
+	uint64_t ngrows;
+	uint64_t nshrinks;
+	uint64_t nshrinkfails;
+	uint64_t ninserts;
+	uint64_t nrelocs;
+#endif
+
+	/* Used for pseudo-random number generation. */
+	uint64_t prng_state;
+
+	/* Total number of items. */
+	size_t count;
+
+	/*
+	 * Minimum and current number of hash table buckets.  There are
+	 * 2^LG_CKH_BUCKET_CELLS cells per bucket.
+	 */
+	unsigned lg_minbuckets;
+	unsigned lg_curbuckets;
+
+	/* Hash and comparison functions. */
+	ckh_hash_t *hash;
+	ckh_keycomp_t *keycomp;
+
+	/* Hash table with 2^lg_curbuckets buckets. */
+	ckhc_t *tab;
+} ckh_t;
+
+/******************************************************************************/
+/* BEGIN PUBLIC API */
+/******************************************************************************/
+
+/* Lifetime management.  Minitems is the initial capacity. */
+bool ckh_new(tsd_t *tsd, ckh_t *ckh, size_t minitems, ckh_hash_t *hash,
+    ckh_keycomp_t *keycomp);
+void ckh_delete(tsd_t *tsd, ckh_t *ckh);
+
+/* Get the number of elements in the set. */
+size_t ckh_count(ckh_t *ckh);
+
+/*
+ * To iterate over the elements in the table, initialize *tabind to 0 and call
+ * this function until it returns true.  Each call that returns false will
+ * update *key and *data to the next element in the table, assuming the pointers
+ * are non-NULL.
+ */
+bool ckh_iter(ckh_t *ckh, size_t *tabind, void **key, void **data);
+
+/*
+ * Basic hash table operations -- insert, removal, lookup.  For ckh_remove and
+ * ckh_search, key or data can be NULL.  The hash-table only stores pointers to
+ * the key and value, and doesn't do any lifetime management.
+ */
+bool ckh_insert(tsd_t *tsd, ckh_t *ckh, const void *key, const void *data);
+bool ckh_remove(tsd_t *tsd, ckh_t *ckh, const void *searchkey, void **key,
+    void **data);
+bool ckh_search(ckh_t *ckh, const void *searchkey, void **key, void **data);
+
+/* Some useful hash and comparison functions for strings and pointers. */
+void ckh_string_hash(const void *key, size_t r_hash[2]);
+bool ckh_string_keycomp(const void *k1, const void *k2);
+void ckh_pointer_hash(const void *key, size_t r_hash[2]);
+bool ckh_pointer_keycomp(const void *k1, const void *k2);
+
+#endif /* JEMALLOC_INTERNAL_CKH_H */

deps/jemalloc/include/jemalloc/internal/ctl.h

@@ -0,0 +1,134 @@
+#ifndef JEMALLOC_INTERNAL_CTL_H
+#define JEMALLOC_INTERNAL_CTL_H
+
+#include "jemalloc/internal/jemalloc_internal_types.h"
+#include "jemalloc/internal/malloc_io.h"
+#include "jemalloc/internal/mutex_prof.h"
+#include "jemalloc/internal/ql.h"
+#include "jemalloc/internal/sc.h"
+#include "jemalloc/internal/stats.h"
+
+/* Maximum ctl tree depth. */
+#define CTL_MAX_DEPTH	7
+
+typedef struct ctl_node_s {
+	bool named;
+} ctl_node_t;
+
+typedef struct ctl_named_node_s {
+	ctl_node_t node;
+	const char *name;
+	/* If (nchildren == 0), this is a terminal node. */
+	size_t nchildren;
+	const ctl_node_t *children;
+	int (*ctl)(tsd_t *, const size_t *, size_t, void *, size_t *, void *,
+	    size_t);
+} ctl_named_node_t;
+
+typedef struct ctl_indexed_node_s {
+	struct ctl_node_s node;
+	const ctl_named_node_t *(*index)(tsdn_t *, const size_t *, size_t,
+	    size_t);
+} ctl_indexed_node_t;
+
+typedef struct ctl_arena_stats_s {
+	arena_stats_t astats;
+
+	/* Aggregate stats for small size classes, based on bin stats. */
+	size_t allocated_small;
+	uint64_t nmalloc_small;
+	uint64_t ndalloc_small;
+	uint64_t nrequests_small;
+	uint64_t nfills_small;
+	uint64_t nflushes_small;
+
+	bin_stats_t bstats[SC_NBINS];
+	arena_stats_large_t lstats[SC_NSIZES - SC_NBINS];
+	arena_stats_extents_t estats[SC_NPSIZES];
+} ctl_arena_stats_t;
+
+typedef struct ctl_stats_s {
+	size_t allocated;
+	size_t active;
+	size_t metadata;
+	size_t metadata_thp;
+	size_t resident;
+	size_t mapped;
+	size_t retained;
+
+	background_thread_stats_t background_thread;
+	mutex_prof_data_t mutex_prof_data[mutex_prof_num_global_mutexes];
+} ctl_stats_t;
+
+typedef struct ctl_arena_s ctl_arena_t;
+struct ctl_arena_s {
+	unsigned arena_ind;
+	bool initialized;
+	ql_elm(ctl_arena_t) destroyed_link;
+
+	/* Basic stats, supported even if !config_stats. */
+	unsigned nthreads;
+	const char *dss;
+	ssize_t dirty_decay_ms;
+	ssize_t muzzy_decay_ms;
+	size_t pactive;
+	size_t pdirty;
+	size_t pmuzzy;
+
+	/* NULL if !config_stats. */
+	ctl_arena_stats_t *astats;
+};
+
+typedef struct ctl_arenas_s {
+	uint64_t epoch;
+	unsigned narenas;
+	ql_head(ctl_arena_t) destroyed;
+
+	/*
+	 * Element 0 corresponds to merged stats for extant arenas (accessed via
+	 * MALLCTL_ARENAS_ALL), element 1 corresponds to merged stats for
+	 * destroyed arenas (accessed via MALLCTL_ARENAS_DESTROYED), and the
+	 * remaining MALLOCX_ARENA_LIMIT elements correspond to arenas.
+	 */
+	ctl_arena_t *arenas[2 + MALLOCX_ARENA_LIMIT];
+} ctl_arenas_t;
+
+int ctl_byname(tsd_t *tsd, const char *name, void *oldp, size_t *oldlenp,
+    void *newp, size_t newlen);
+int ctl_nametomib(tsd_t *tsd, const char *name, size_t *mibp, size_t *miblenp);
+
+int ctl_bymib(tsd_t *tsd, const size_t *mib, size_t miblen, void *oldp,
+    size_t *oldlenp, void *newp, size_t newlen);
+bool ctl_boot(void);
+void ctl_prefork(tsdn_t *tsdn);
+void ctl_postfork_parent(tsdn_t *tsdn);
+void ctl_postfork_child(tsdn_t *tsdn);
+
+#define xmallctl(name, oldp, oldlenp, newp, newlen) do {		\
+	if (je_mallctl(name, oldp, oldlenp, newp, newlen)		\
+	    != 0) {							\
+		malloc_printf(						\
+		    "<jemalloc>: Failure in xmallctl(\"%s\", ...)\n",	\
+		    name);						\
+		abort();						\
+	}								\
+} while (0)
+
+#define xmallctlnametomib(name, mibp, miblenp) do {			\
+	if (je_mallctlnametomib(name, mibp, miblenp) != 0) {		\
+		malloc_printf("<jemalloc>: Failure in "			\
+		    "xmallctlnametomib(\"%s\", ...)\n", name);		\
+		abort();						\
+	}								\
+} while (0)
+
+#define xmallctlbymib(mib, miblen, oldp, oldlenp, newp, newlen) do {	\
+	if (je_mallctlbymib(mib, miblen, oldp, oldlenp, newp,		\
+	    newlen) != 0) {						\
+		malloc_write(						\
+		    "<jemalloc>: Failure in xmallctlbymib()\n");	\
+		abort();						\
+	}								\
+} while (0)
+
+#endif /* JEMALLOC_INTERNAL_CTL_H */

deps/jemalloc/include/jemalloc/internal/div.h

@@ -0,0 +1,41 @@
+#ifndef JEMALLOC_INTERNAL_DIV_H
+#define JEMALLOC_INTERNAL_DIV_H
+
+#include "jemalloc/internal/assert.h"
+
+/*
+ * This module does the division that computes the index of a region in a slab,
+ * given its offset relative to the base.
+ * That is, given a divisor d, an n = i * d (all integers), we'll return i.
+ * We do some pre-computation to do this more quickly than a CPU division
+ * instruction.
+ * We bound n < 2^32, and don't support dividing by one.
+ */
+
+typedef struct div_info_s div_info_t;
+struct div_info_s {
+	uint32_t magic;
+#ifdef JEMALLOC_DEBUG
+	size_t d;
+#endif
+};
+
+void div_init(div_info_t *div_info, size_t divisor);
+
+static inline size_t
+div_compute(div_info_t *div_info, size_t n) {
+	assert(n <= (uint32_t)-1);
+	/*
+	 * This generates, e.g. mov; imul; shr on x86-64. On a 32-bit machine,
+	 * the compilers I tried were all smart enough to turn this into the
+	 * appropriate "get the high 32 bits of the result of a multiply" (e.g.
+	 * mul; mov edx eax; on x86, umull on arm, etc.).
+	 */
+	size_t i = ((uint64_t)n * (uint64_t)div_info->magic) >> 32;
+#ifdef JEMALLOC_DEBUG
+	assert(i * div_info->d == n);
+#endif
+	return i;
+}
+
+#endif /* JEMALLOC_INTERNAL_DIV_H */

deps/jemalloc/include/jemalloc/internal/emitter.h

@@ -0,0 +1,486 @@
+#ifndef JEMALLOC_INTERNAL_EMITTER_H
+#define JEMALLOC_INTERNAL_EMITTER_H
+
+#include "jemalloc/internal/ql.h"
+
+typedef enum emitter_output_e emitter_output_t;
+enum emitter_output_e {
+	emitter_output_json,
+	emitter_output_table
+};
+
+typedef enum emitter_justify_e emitter_justify_t;
+enum emitter_justify_e {
+	emitter_justify_left,
+	emitter_justify_right,
+	/* Not for users; just to pass to internal functions. */
+	emitter_justify_none
+};
+
+typedef enum emitter_type_e emitter_type_t;
+enum emitter_type_e {
+	emitter_type_bool,
+	emitter_type_int,
+	emitter_type_unsigned,
+	emitter_type_uint32,
+	emitter_type_uint64,
+	emitter_type_size,
+	emitter_type_ssize,
+	emitter_type_string,
+	/*
+	 * A title is a column title in a table; it's just a string, but it's
+	 * not quoted.
+	 */
+	emitter_type_title,
+};
+
+typedef struct emitter_col_s emitter_col_t;
+struct emitter_col_s {
+	/* Filled in by the user. */
+	emitter_justify_t justify;
+	int width;
+	emitter_type_t type;
+	union {
+		bool bool_val;
+		int int_val;
+		unsigned unsigned_val;
+		uint32_t uint32_val;
+		uint32_t uint32_t_val;
+		uint64_t uint64_val;
+		uint64_t uint64_t_val;
+		size_t size_val;
+		ssize_t ssize_val;
+		const char *str_val;
+	};
+
+	/* Filled in by initialization. */
+	ql_elm(emitter_col_t) link;
+};
+
+typedef struct emitter_row_s emitter_row_t;
+struct emitter_row_s {
+	ql_head(emitter_col_t) cols;
+};
+
+typedef struct emitter_s emitter_t;
+struct emitter_s {
+	emitter_output_t output;
+	/* The output information. */
+	void (*write_cb)(void *, const char *);
+	void *cbopaque;
+	int nesting_depth;
+	/* True if we've already emitted a value at the given depth. */
+	bool item_at_depth;
+	/* True if we emitted a key and will emit corresponding value next. */
+	bool emitted_key;
+};
+
+/* Internal convenience function.  Write to the emitter the given string. */
+JEMALLOC_FORMAT_PRINTF(2, 3)
+static inline void
+emitter_printf(emitter_t *emitter, const char *format, ...) {
+	va_list ap;
+
+	va_start(ap, format);
+	malloc_vcprintf(emitter->write_cb, emitter->cbopaque, format, ap);
+	va_end(ap);
+}
+
+static inline const char * JEMALLOC_FORMAT_ARG(3)
+emitter_gen_fmt(char *out_fmt, size_t out_size, const char *fmt_specifier,
+    emitter_justify_t justify, int width) {
+	size_t written;
+	fmt_specifier++;
+	if (justify == emitter_justify_none) {
+		written = malloc_snprintf(out_fmt, out_size,
+		    "%%%s", fmt_specifier);
+	} else if (justify == emitter_justify_left) {
+		written = malloc_snprintf(out_fmt, out_size,
+		    "%%-%d%s", width, fmt_specifier);
+	} else {
+		written = malloc_snprintf(out_fmt, out_size,
+		    "%%%d%s", width, fmt_specifier);
+	}
+	/* Only happens in case of bad format string, which *we* choose. */
+	assert(written <  out_size);
+	return out_fmt;
+}
+
+/*
+ * Internal.  Emit the given value type in the relevant encoding (so that the
+ * bool true gets mapped to json "true", but the string "true" gets mapped to
+ * json "\"true\"", for instance.
+ *
+ * Width is ignored if justify is emitter_justify_none.
+ */
+static inline void
+emitter_print_value(emitter_t *emitter, emitter_justify_t justify, int width,
+    emitter_type_t value_type, const void *value) {
+	size_t str_written;
+#define BUF_SIZE 256
+#define FMT_SIZE 10
+	/*
+	 * We dynamically generate a format string to emit, to let us use the
+	 * snprintf machinery.  This is kinda hacky, but gets the job done
+	 * quickly without having to think about the various snprintf edge
+	 * cases.
+	 */
+	char fmt[FMT_SIZE];
+	char buf[BUF_SIZE];
+
+#define EMIT_SIMPLE(type, format)					\
+	emitter_printf(emitter,						\
+	    emitter_gen_fmt(fmt, FMT_SIZE, format, justify, width),	\
+	    *(const type *)value);
+
+	switch (value_type) {
+	case emitter_type_bool:
+		emitter_printf(emitter, 
+		    emitter_gen_fmt(fmt, FMT_SIZE, "%s", justify, width),
+		    *(const bool *)value ?  "true" : "false");
+		break;
+	case emitter_type_int:
+		EMIT_SIMPLE(int, "%d")
+		break;
+	case emitter_type_unsigned:
+		EMIT_SIMPLE(unsigned, "%u")
+		break;
+	case emitter_type_ssize:
+		EMIT_SIMPLE(ssize_t, "%zd")
+		break;
+	case emitter_type_size:
+		EMIT_SIMPLE(size_t, "%zu")
+		break;
+	case emitter_type_string:
+		str_written = malloc_snprintf(buf, BUF_SIZE, "\"%s\"",
+		    *(const char *const *)value);
+		/*
+		 * We control the strings we output; we shouldn't get anything
+		 * anywhere near the fmt size.
+		 */
+		assert(str_written < BUF_SIZE);
+		emitter_printf(emitter, 
+		    emitter_gen_fmt(fmt, FMT_SIZE, "%s", justify, width), buf);
+		break;
+	case emitter_type_uint32:
+		EMIT_SIMPLE(uint32_t, "%" FMTu32)
+		break;
+	case emitter_type_uint64:
+		EMIT_SIMPLE(uint64_t, "%" FMTu64)
+		break;
+	case emitter_type_title:
+		EMIT_SIMPLE(char *const, "%s");
+		break;
+	default:
+		unreachable();
+	}
+#undef BUF_SIZE
+#undef FMT_SIZE
+}
+
+
+/* Internal functions.  In json mode, tracks nesting state. */
+static inline void
+emitter_nest_inc(emitter_t *emitter) {
+	emitter->nesting_depth++;
+	emitter->item_at_depth = false;
+}
+
+static inline void
+emitter_nest_dec(emitter_t *emitter) {
+	emitter->nesting_depth--;
+	emitter->item_at_depth = true;
+}
+
+static inline void
+emitter_indent(emitter_t *emitter) {
+	int amount = emitter->nesting_depth;
+	const char *indent_str;
+	if (emitter->output == emitter_output_json) {
+		indent_str = "\t";
+	} else {
+		amount *= 2;
+		indent_str = " ";
+	}
+	for (int i = 0; i < amount; i++) {
+		emitter_printf(emitter, "%s", indent_str);
+	}
+}
+
+static inline void
+emitter_json_key_prefix(emitter_t *emitter) {
+	if (emitter->emitted_key) {
+		emitter->emitted_key = false;
+		return;
+	}
+	emitter_printf(emitter, "%s\n", emitter->item_at_depth ? "," : "");
+	emitter_indent(emitter);
+}
+
+/******************************************************************************/
+/* Public functions for emitter_t. */
+
+static inline void
+emitter_init(emitter_t *emitter, emitter_output_t emitter_output,
+    void (*write_cb)(void *, const char *), void *cbopaque) {
+	emitter->output = emitter_output;
+	emitter->write_cb = write_cb;
+	emitter->cbopaque = cbopaque;
+	emitter->item_at_depth = false;
+	emitter->emitted_key = false; 
+	emitter->nesting_depth = 0;
+}
+
+/******************************************************************************/
+/* JSON public API. */
+
+/* 
+ * Emits a key (e.g. as appears in an object). The next json entity emitted will
+ * be the corresponding value.
+ */
+static inline void
+emitter_json_key(emitter_t *emitter, const char *json_key) {
+	if (emitter->output == emitter_output_json) {
+		emitter_json_key_prefix(emitter);
+		emitter_printf(emitter, "\"%s\": ", json_key);
+		emitter->emitted_key = true;
+	}
+}
+
+static inline void
+emitter_json_value(emitter_t *emitter, emitter_type_t value_type,
+    const void *value) {
+	if (emitter->output == emitter_output_json) {
+		emitter_json_key_prefix(emitter);
+		emitter_print_value(emitter, emitter_justify_none, -1,
+		    value_type, value);
+		emitter->item_at_depth = true;
+	}
+}
+
+/* Shorthand for calling emitter_json_key and then emitter_json_value. */
+static inline void
+emitter_json_kv(emitter_t *emitter, const char *json_key,
+    emitter_type_t value_type, const void *value) {
+	emitter_json_key(emitter, json_key);
+	emitter_json_value(emitter, value_type, value);
+}
+
+static inline void
+emitter_json_array_begin(emitter_t *emitter) {
+	if (emitter->output == emitter_output_json) {
+		emitter_json_key_prefix(emitter);
+		emitter_printf(emitter, "[");
+		emitter_nest_inc(emitter);
+	}
+}
+
+/* Shorthand for calling emitter_json_key and then emitter_json_array_begin. */
+static inline void
+emitter_json_array_kv_begin(emitter_t *emitter, const char *json_key) {
+	emitter_json_key(emitter, json_key);
+	emitter_json_array_begin(emitter);
+}
+
+static inline void
+emitter_json_array_end(emitter_t *emitter) {
+	if (emitter->output == emitter_output_json) {
+		assert(emitter->nesting_depth > 0);
+		emitter_nest_dec(emitter);
+		emitter_printf(emitter, "\n");
+		emitter_indent(emitter);
+		emitter_printf(emitter, "]");
+	}
+}
+
+static inline void
+emitter_json_object_begin(emitter_t *emitter) {
+	if (emitter->output == emitter_output_json) {
+		emitter_json_key_prefix(emitter);
+		emitter_printf(emitter, "{");
+		emitter_nest_inc(emitter);
+	}
+}
+
+/* Shorthand for calling emitter_json_key and then emitter_json_object_begin. */
+static inline void
+emitter_json_object_kv_begin(emitter_t *emitter, const char *json_key) {
+	emitter_json_key(emitter, json_key);
+	emitter_json_object_begin(emitter);
+}
+
+static inline void
+emitter_json_object_end(emitter_t *emitter) {
+	if (emitter->output == emitter_output_json) {
+		assert(emitter->nesting_depth > 0);
+		emitter_nest_dec(emitter);
+		emitter_printf(emitter, "\n");
+		emitter_indent(emitter);
+		emitter_printf(emitter, "}");
+	}
+}
+
+
+/******************************************************************************/
+/* Table public API. */
+
+static inline void
+emitter_table_dict_begin(emitter_t *emitter, const char *table_key) {
+	if (emitter->output == emitter_output_table) {
+		emitter_indent(emitter);
+		emitter_printf(emitter, "%s\n", table_key);
+		emitter_nest_inc(emitter);
+	}
+}
+
+static inline void
+emitter_table_dict_end(emitter_t *emitter) {
+	if (emitter->output == emitter_output_table) {
+		emitter_nest_dec(emitter);
+	}
+}
+
+static inline void
+emitter_table_kv_note(emitter_t *emitter, const char *table_key,
+    emitter_type_t value_type, const void *value,
+    const char *table_note_key, emitter_type_t table_note_value_type,
+    const void *table_note_value) {
+	if (emitter->output == emitter_output_table) {
+		emitter_indent(emitter);
+		emitter_printf(emitter, "%s: ", table_key);
+		emitter_print_value(emitter, emitter_justify_none, -1,
+		    value_type, value);
+		if (table_note_key != NULL) {
+			emitter_printf(emitter, " (%s: ", table_note_key);
+			emitter_print_value(emitter, emitter_justify_none, -1,
+			    table_note_value_type, table_note_value);
+			emitter_printf(emitter, ")");
+		}
+		emitter_printf(emitter, "\n");
+	}
+	emitter->item_at_depth = true;
+}
+
+static inline void
+emitter_table_kv(emitter_t *emitter, const char *table_key,
+    emitter_type_t value_type, const void *value) {
+	emitter_table_kv_note(emitter, table_key, value_type, value, NULL,
+	    emitter_type_bool, NULL);
+}
+
+
+/* Write to the emitter the given string, but only in table mode. */
+JEMALLOC_FORMAT_PRINTF(2, 3)
+static inline void
+emitter_table_printf(emitter_t *emitter, const char *format, ...) {
+	if (emitter->output == emitter_output_table) {
+		va_list ap;
+		va_start(ap, format);
+		malloc_vcprintf(emitter->write_cb, emitter->cbopaque, format, ap);
+		va_end(ap);
+	}
+}
+
+static inline void
+emitter_table_row(emitter_t *emitter, emitter_row_t *row) {
+	if (emitter->output != emitter_output_table) {
+		return;
+	}
+	emitter_col_t *col;
+	ql_foreach(col, &row->cols, link) {
+		emitter_print_value(emitter, col->justify, col->width,
+		    col->type, (const void *)&col->bool_val);
+	}
+	emitter_table_printf(emitter, "\n");
+}
+
+static inline void
+emitter_row_init(emitter_row_t *row) {
+	ql_new(&row->cols);
+}
+
+static inline void
+emitter_col_init(emitter_col_t *col, emitter_row_t *row) {
+	ql_elm_new(col, link);
+	ql_tail_insert(&row->cols, col, link);
+}
+
+
+/******************************************************************************/
+/*
+ * Generalized public API. Emits using either JSON or table, according to
+ * settings in the emitter_t. */
+
+/*
+ * Note emits a different kv pair as well, but only in table mode.  Omits the
+ * note if table_note_key is NULL.
+ */
+static inline void
+emitter_kv_note(emitter_t *emitter, const char *json_key, const char *table_key,
+    emitter_type_t value_type, const void *value,
+    const char *table_note_key, emitter_type_t table_note_value_type,
+    const void *table_note_value) {
+	if (emitter->output == emitter_output_json) {
+		emitter_json_key(emitter, json_key);
+		emitter_json_value(emitter, value_type, value);
+	} else {
+		emitter_table_kv_note(emitter, table_key, value_type, value,
+		    table_note_key, table_note_value_type, table_note_value);
+	}
+	emitter->item_at_depth = true;
+}
+
+static inline void
+emitter_kv(emitter_t *emitter, const char *json_key, const char *table_key,
+    emitter_type_t value_type, const void *value) {
+	emitter_kv_note(emitter, json_key, table_key, value_type, value, NULL,
+	    emitter_type_bool, NULL);
+}
+
+static inline void
+emitter_dict_begin(emitter_t *emitter, const char *json_key,
+    const char *table_header) {
+	if (emitter->output == emitter_output_json) {
+		emitter_json_key(emitter, json_key);
+		emitter_json_object_begin(emitter);
+	} else {
+		emitter_table_dict_begin(emitter, table_header);
+	}
+}
+
+static inline void
+emitter_dict_end(emitter_t *emitter) {
+	if (emitter->output == emitter_output_json) {
+		emitter_json_object_end(emitter);
+	} else {
+		emitter_table_dict_end(emitter);
+	}
+}
+
+static inline void
+emitter_begin(emitter_t *emitter) {
+	if (emitter->output == emitter_output_json) {
+		assert(emitter->nesting_depth == 0);
+		emitter_printf(emitter, "{");
+		emitter_nest_inc(emitter);
+	} else {
+		/*
+		 * This guarantees that we always call write_cb at least once.
+		 * This is useful if some invariant is established by each call
+		 * to write_cb, but doesn't hold initially: e.g., some buffer
+		 * holds a null-terminated string.
+		 */
+		emitter_printf(emitter, "%s", "");
+	}
+}
+
+static inline void
+emitter_end(emitter_t *emitter) {
+	if (emitter->output == emitter_output_json) {
+		assert(emitter->nesting_depth == 1);
+		emitter_nest_dec(emitter);
+		emitter_printf(emitter, "\n}\n");
+	}
+}
+
+#endif /* JEMALLOC_INTERNAL_EMITTER_H */

deps/jemalloc/include/jemalloc/internal/extent.h

@@ -0,0 +1,46 @@
+/******************************************************************************/
+#ifdef JEMALLOC_H_TYPES
+
+typedef struct extent_node_s extent_node_t;
+
+#endif /* JEMALLOC_H_TYPES */
+/******************************************************************************/
+#ifdef JEMALLOC_H_STRUCTS
+
+/* Tree of extents. */
+struct extent_node_s {
+	/* Linkage for the size/address-ordered tree. */
+	rb_node(extent_node_t)	link_szad;
+
+	/* Linkage for the address-ordered tree. */
+	rb_node(extent_node_t)	link_ad;
+
+	/* Profile counters, used for huge objects. */
+	prof_ctx_t		*prof_ctx;
+
+	/* Pointer to the extent that this tree node is responsible for. */
+	void			*addr;
+
+	/* Total region size. */
+	size_t			size;
+
+	/* True if zero-filled; used by chunk recycling code. */
+	bool			zeroed;
+};
+typedef rb_tree(extent_node_t) extent_tree_t;
+
+#endif /* JEMALLOC_H_STRUCTS */
+/******************************************************************************/
+#ifdef JEMALLOC_H_EXTERNS
+
+rb_proto(, extent_tree_szad_, extent_tree_t, extent_node_t)
+
+rb_proto(, extent_tree_ad_, extent_tree_t, extent_node_t)
+
+#endif /* JEMALLOC_H_EXTERNS */
+/******************************************************************************/
+#ifdef JEMALLOC_H_INLINES
+
+#endif /* JEMALLOC_H_INLINES */
+/******************************************************************************/
+

deps/jemalloc/include/jemalloc/internal/extent_dss.h

@@ -0,0 +1,26 @@
+#ifndef JEMALLOC_INTERNAL_EXTENT_DSS_H
+#define JEMALLOC_INTERNAL_EXTENT_DSS_H
+
+typedef enum {
+	dss_prec_disabled  = 0,
+	dss_prec_primary   = 1,
+	dss_prec_secondary = 2,
+
+	dss_prec_limit     = 3
+} dss_prec_t;
+#define DSS_PREC_DEFAULT dss_prec_secondary
+#define DSS_DEFAULT "secondary"
+
+extern const char *dss_prec_names[];
+
+extern const char *opt_dss;
+
+dss_prec_t extent_dss_prec_get(void);
+bool extent_dss_prec_set(dss_prec_t dss_prec);
+void *extent_alloc_dss(tsdn_t *tsdn, arena_t *arena, void *new_addr,
+    size_t size, size_t alignment, bool *zero, bool *commit);
+bool extent_in_dss(void *addr);
+bool extent_dss_mergeable(void *addr_a, void *addr_b);
+void extent_dss_boot(void);
+
+#endif /* JEMALLOC_INTERNAL_EXTENT_DSS_H */

deps/jemalloc/include/jemalloc/internal/extent_externs.h

@@ -0,0 +1,83 @@
+#ifndef JEMALLOC_INTERNAL_EXTENT_EXTERNS_H
+#define JEMALLOC_INTERNAL_EXTENT_EXTERNS_H
+
+#include "jemalloc/internal/mutex.h"
+#include "jemalloc/internal/mutex_pool.h"
+#include "jemalloc/internal/ph.h"
+#include "jemalloc/internal/rtree.h"
+
+extern size_t opt_lg_extent_max_active_fit;
+
+extern rtree_t extents_rtree;
+extern const extent_hooks_t extent_hooks_default;
+extern mutex_pool_t extent_mutex_pool;
+
+extent_t *extent_alloc(tsdn_t *tsdn, arena_t *arena);
+void extent_dalloc(tsdn_t *tsdn, arena_t *arena, extent_t *extent);
+
+extent_hooks_t *extent_hooks_get(arena_t *arena);
+extent_hooks_t *extent_hooks_set(tsd_t *tsd, arena_t *arena,
+    extent_hooks_t *extent_hooks);
+
+#ifdef JEMALLOC_JET
+size_t extent_size_quantize_floor(size_t size);
+size_t extent_size_quantize_ceil(size_t size);
+#endif
+
+ph_proto(, extent_avail_, extent_tree_t, extent_t)
+ph_proto(, extent_heap_, extent_heap_t, extent_t)
+
+bool extents_init(tsdn_t *tsdn, extents_t *extents, extent_state_t state,
+    bool delay_coalesce);
+extent_state_t extents_state_get(const extents_t *extents);
+size_t extents_npages_get(extents_t *extents);
+/* Get the number of extents in the given page size index. */
+size_t extents_nextents_get(extents_t *extents, pszind_t ind);
+/* Get the sum total bytes of the extents in the given page size index. */
+size_t extents_nbytes_get(extents_t *extents, pszind_t ind);
+extent_t *extents_alloc(tsdn_t *tsdn, arena_t *arena,
+    extent_hooks_t **r_extent_hooks, extents_t *extents, void *new_addr,
+    size_t size, size_t pad, size_t alignment, bool slab, szind_t szind,
+    bool *zero, bool *commit);
+void extents_dalloc(tsdn_t *tsdn, arena_t *arena,
+    extent_hooks_t **r_extent_hooks, extents_t *extents, extent_t *extent);
+extent_t *extents_evict(tsdn_t *tsdn, arena_t *arena,
+    extent_hooks_t **r_extent_hooks, extents_t *extents, size_t npages_min);
+void extents_prefork(tsdn_t *tsdn, extents_t *extents);
+void extents_postfork_parent(tsdn_t *tsdn, extents_t *extents);
+void extents_postfork_child(tsdn_t *tsdn, extents_t *extents);
+extent_t *extent_alloc_wrapper(tsdn_t *tsdn, arena_t *arena,
+    extent_hooks_t **r_extent_hooks, void *new_addr, size_t size, size_t pad,
+    size_t alignment, bool slab, szind_t szind, bool *zero, bool *commit);
+void extent_dalloc_gap(tsdn_t *tsdn, arena_t *arena, extent_t *extent);
+void extent_dalloc_wrapper(tsdn_t *tsdn, arena_t *arena,
+    extent_hooks_t **r_extent_hooks, extent_t *extent);
+void extent_destroy_wrapper(tsdn_t *tsdn, arena_t *arena,
+    extent_hooks_t **r_extent_hooks, extent_t *extent);
+bool extent_commit_wrapper(tsdn_t *tsdn, arena_t *arena,
+    extent_hooks_t **r_extent_hooks, extent_t *extent, size_t offset,
+    size_t length);
+bool extent_decommit_wrapper(tsdn_t *tsdn, arena_t *arena,
+    extent_hooks_t **r_extent_hooks, extent_t *extent, size_t offset,
+    size_t length);
+bool extent_purge_lazy_wrapper(tsdn_t *tsdn, arena_t *arena,
+    extent_hooks_t **r_extent_hooks, extent_t *extent, size_t offset,
+    size_t length);
+bool extent_purge_forced_wrapper(tsdn_t *tsdn, arena_t *arena,
+    extent_hooks_t **r_extent_hooks, extent_t *extent, size_t offset,
+    size_t length);
+extent_t *extent_split_wrapper(tsdn_t *tsdn, arena_t *arena,
+    extent_hooks_t **r_extent_hooks, extent_t *extent, size_t size_a,
+    szind_t szind_a, bool slab_a, size_t size_b, szind_t szind_b, bool slab_b);
+bool extent_merge_wrapper(tsdn_t *tsdn, arena_t *arena,
+    extent_hooks_t **r_extent_hooks, extent_t *a, extent_t *b);
+
+bool extent_boot(void);
+
+void extent_util_stats_get(tsdn_t *tsdn, const void *ptr,
+    size_t *nfree, size_t *nregs, size_t *size);
+void extent_util_stats_verbose_get(tsdn_t *tsdn, const void *ptr,
+    size_t *nfree, size_t *nregs, size_t *size,
+    size_t *bin_nfree, size_t *bin_nregs, void **slabcur_addr);
+
+#endif /* JEMALLOC_INTERNAL_EXTENT_EXTERNS_H */

deps/jemalloc/include/jemalloc/internal/extent_inlines.h

@@ -0,0 +1,501 @@
+#ifndef JEMALLOC_INTERNAL_EXTENT_INLINES_H
+#define JEMALLOC_INTERNAL_EXTENT_INLINES_H
+
+#include "jemalloc/internal/mutex.h"
+#include "jemalloc/internal/mutex_pool.h"
+#include "jemalloc/internal/pages.h"
+#include "jemalloc/internal/prng.h"
+#include "jemalloc/internal/ql.h"
+#include "jemalloc/internal/sc.h"
+#include "jemalloc/internal/sz.h"
+
+static inline void
+extent_lock(tsdn_t *tsdn, extent_t *extent) {
+	assert(extent != NULL);
+	mutex_pool_lock(tsdn, &extent_mutex_pool, (uintptr_t)extent);
+}
+
+static inline void
+extent_unlock(tsdn_t *tsdn, extent_t *extent) {
+	assert(extent != NULL);
+	mutex_pool_unlock(tsdn, &extent_mutex_pool, (uintptr_t)extent);
+}
+
+static inline void
+extent_lock2(tsdn_t *tsdn, extent_t *extent1, extent_t *extent2) {
+	assert(extent1 != NULL && extent2 != NULL);
+	mutex_pool_lock2(tsdn, &extent_mutex_pool, (uintptr_t)extent1,
+	    (uintptr_t)extent2);
+}
+
+static inline void
+extent_unlock2(tsdn_t *tsdn, extent_t *extent1, extent_t *extent2) {
+	assert(extent1 != NULL && extent2 != NULL);
+	mutex_pool_unlock2(tsdn, &extent_mutex_pool, (uintptr_t)extent1,
+	    (uintptr_t)extent2);
+}
+
+static inline unsigned
+extent_arena_ind_get(const extent_t *extent) {
+	unsigned arena_ind = (unsigned)((extent->e_bits &
+	    EXTENT_BITS_ARENA_MASK) >> EXTENT_BITS_ARENA_SHIFT);
+	assert(arena_ind < MALLOCX_ARENA_LIMIT);
+
+	return arena_ind;
+}
+
+static inline arena_t *
+extent_arena_get(const extent_t *extent) {
+	unsigned arena_ind = extent_arena_ind_get(extent);
+
+	return (arena_t *)atomic_load_p(&arenas[arena_ind], ATOMIC_ACQUIRE);
+}
+
+static inline szind_t
+extent_szind_get_maybe_invalid(const extent_t *extent) {
+	szind_t szind = (szind_t)((extent->e_bits & EXTENT_BITS_SZIND_MASK) >>
+	    EXTENT_BITS_SZIND_SHIFT);
+	assert(szind <= SC_NSIZES);
+	return szind;
+}
+
+static inline szind_t
+extent_szind_get(const extent_t *extent) {
+	szind_t szind = extent_szind_get_maybe_invalid(extent);
+	assert(szind < SC_NSIZES); /* Never call when "invalid". */
+	return szind;
+}
+
+static inline size_t
+extent_usize_get(const extent_t *extent) {
+	return sz_index2size(extent_szind_get(extent));
+}
+
+static inline unsigned
+extent_binshard_get(const extent_t *extent) {
+	unsigned binshard = (unsigned)((extent->e_bits &
+	    EXTENT_BITS_BINSHARD_MASK) >> EXTENT_BITS_BINSHARD_SHIFT);
+	assert(binshard < bin_infos[extent_szind_get(extent)].n_shards);
+	return binshard;
+}
+
+static inline size_t
+extent_sn_get(const extent_t *extent) {
+	return (size_t)((extent->e_bits & EXTENT_BITS_SN_MASK) >>
+	    EXTENT_BITS_SN_SHIFT);
+}
+
+static inline extent_state_t
+extent_state_get(const extent_t *extent) {
+	return (extent_state_t)((extent->e_bits & EXTENT_BITS_STATE_MASK) >>
+	    EXTENT_BITS_STATE_SHIFT);
+}
+
+static inline bool
+extent_zeroed_get(const extent_t *extent) {
+	return (bool)((extent->e_bits & EXTENT_BITS_ZEROED_MASK) >>
+	    EXTENT_BITS_ZEROED_SHIFT);
+}
+
+static inline bool
+extent_committed_get(const extent_t *extent) {
+	return (bool)((extent->e_bits & EXTENT_BITS_COMMITTED_MASK) >>
+	    EXTENT_BITS_COMMITTED_SHIFT);
+}
+
+static inline bool
+extent_dumpable_get(const extent_t *extent) {
+	return (bool)((extent->e_bits & EXTENT_BITS_DUMPABLE_MASK) >>
+	    EXTENT_BITS_DUMPABLE_SHIFT);
+}
+
+static inline bool
+extent_slab_get(const extent_t *extent) {
+	return (bool)((extent->e_bits & EXTENT_BITS_SLAB_MASK) >>
+	    EXTENT_BITS_SLAB_SHIFT);
+}
+
+static inline unsigned
+extent_nfree_get(const extent_t *extent) {
+	assert(extent_slab_get(extent));
+	return (unsigned)((extent->e_bits & EXTENT_BITS_NFREE_MASK) >>
+	    EXTENT_BITS_NFREE_SHIFT);
+}
+
+static inline void *
+extent_base_get(const extent_t *extent) {
+	assert(extent->e_addr == PAGE_ADDR2BASE(extent->e_addr) ||
+	    !extent_slab_get(extent));
+	return PAGE_ADDR2BASE(extent->e_addr);
+}
+
+static inline void *
+extent_addr_get(const extent_t *extent) {
+	assert(extent->e_addr == PAGE_ADDR2BASE(extent->e_addr) ||
+	    !extent_slab_get(extent));
+	return extent->e_addr;
+}
+
+static inline size_t
+extent_size_get(const extent_t *extent) {
+	return (extent->e_size_esn & EXTENT_SIZE_MASK);
+}
+
+static inline size_t
+extent_esn_get(const extent_t *extent) {
+	return (extent->e_size_esn & EXTENT_ESN_MASK);
+}
+
+static inline size_t
+extent_bsize_get(const extent_t *extent) {
+	return extent->e_bsize;
+}
+
+static inline void *
+extent_before_get(const extent_t *extent) {
+	return (void *)((uintptr_t)extent_base_get(extent) - PAGE);
+}
+
+static inline void *
+extent_last_get(const extent_t *extent) {
+	return (void *)((uintptr_t)extent_base_get(extent) +
+	    extent_size_get(extent) - PAGE);
+}
+
+static inline void *
+extent_past_get(const extent_t *extent) {
+	return (void *)((uintptr_t)extent_base_get(extent) +
+	    extent_size_get(extent));
+}
+
+static inline arena_slab_data_t *
+extent_slab_data_get(extent_t *extent) {
+	assert(extent_slab_get(extent));
+	return &extent->e_slab_data;
+}
+
+static inline const arena_slab_data_t *
+extent_slab_data_get_const(const extent_t *extent) {
+	assert(extent_slab_get(extent));
+	return &extent->e_slab_data;
+}
+
+static inline prof_tctx_t *
+extent_prof_tctx_get(const extent_t *extent) {
+	return (prof_tctx_t *)atomic_load_p(&extent->e_prof_tctx,
+	    ATOMIC_ACQUIRE);
+}
+
+static inline nstime_t
+extent_prof_alloc_time_get(const extent_t *extent) {
+	return extent->e_alloc_time;
+}
+
+static inline void
+extent_arena_set(extent_t *extent, arena_t *arena) {
+	unsigned arena_ind = (arena != NULL) ? arena_ind_get(arena) : ((1U <<
+	    MALLOCX_ARENA_BITS) - 1);
+	extent->e_bits = (extent->e_bits & ~EXTENT_BITS_ARENA_MASK) |
+	    ((uint64_t)arena_ind << EXTENT_BITS_ARENA_SHIFT);
+}
+
+static inline void
+extent_binshard_set(extent_t *extent, unsigned binshard) {
+	/* The assertion assumes szind is set already. */
+	assert(binshard < bin_infos[extent_szind_get(extent)].n_shards);
+	extent->e_bits = (extent->e_bits & ~EXTENT_BITS_BINSHARD_MASK) |
+	    ((uint64_t)binshard << EXTENT_BITS_BINSHARD_SHIFT);
+}
+
+static inline void
+extent_addr_set(extent_t *extent, void *addr) {
+	extent->e_addr = addr;
+}
+
+static inline void
+extent_addr_randomize(tsdn_t *tsdn, extent_t *extent, size_t alignment) {
+	assert(extent_base_get(extent) == extent_addr_get(extent));
+
+	if (alignment < PAGE) {
+		unsigned lg_range = LG_PAGE -
+		    lg_floor(CACHELINE_CEILING(alignment));
+		size_t r;
+		if (!tsdn_null(tsdn)) {
+			tsd_t *tsd = tsdn_tsd(tsdn);
+			r = (size_t)prng_lg_range_u64(
+			    tsd_offset_statep_get(tsd), lg_range);
+		} else {
+			r = prng_lg_range_zu(
+			    &extent_arena_get(extent)->offset_state,
+			    lg_range, true);
+		}
+		uintptr_t random_offset = ((uintptr_t)r) << (LG_PAGE -
+		    lg_range);
+		extent->e_addr = (void *)((uintptr_t)extent->e_addr +
+		    random_offset);
+		assert(ALIGNMENT_ADDR2BASE(extent->e_addr, alignment) ==
+		    extent->e_addr);
+	}
+}
+
+static inline void
+extent_size_set(extent_t *extent, size_t size) {
+	assert((size & ~EXTENT_SIZE_MASK) == 0);
+	extent->e_size_esn = size | (extent->e_size_esn & ~EXTENT_SIZE_MASK);
+}
+
+static inline void
+extent_esn_set(extent_t *extent, size_t esn) {
+	extent->e_size_esn = (extent->e_size_esn & ~EXTENT_ESN_MASK) | (esn &
+	    EXTENT_ESN_MASK);
+}
+
+static inline void
+extent_bsize_set(extent_t *extent, size_t bsize) {
+	extent->e_bsize = bsize;
+}
+
+static inline void
+extent_szind_set(extent_t *extent, szind_t szind) {
+	assert(szind <= SC_NSIZES); /* SC_NSIZES means "invalid". */
+	extent->e_bits = (extent->e_bits & ~EXTENT_BITS_SZIND_MASK) |
+	    ((uint64_t)szind << EXTENT_BITS_SZIND_SHIFT);
+}
+
+static inline void
+extent_nfree_set(extent_t *extent, unsigned nfree) {
+	assert(extent_slab_get(extent));
+	extent->e_bits = (extent->e_bits & ~EXTENT_BITS_NFREE_MASK) |
+	    ((uint64_t)nfree << EXTENT_BITS_NFREE_SHIFT);
+}
+
+static inline void
+extent_nfree_binshard_set(extent_t *extent, unsigned nfree, unsigned binshard) {
+	/* The assertion assumes szind is set already. */
+	assert(binshard < bin_infos[extent_szind_get(extent)].n_shards);
+	extent->e_bits = (extent->e_bits &
+	    (~EXTENT_BITS_NFREE_MASK & ~EXTENT_BITS_BINSHARD_MASK)) |
+	    ((uint64_t)binshard << EXTENT_BITS_BINSHARD_SHIFT) |
+	    ((uint64_t)nfree << EXTENT_BITS_NFREE_SHIFT);
+}
+
+static inline void
+extent_nfree_inc(extent_t *extent) {
+	assert(extent_slab_get(extent));
+	extent->e_bits += ((uint64_t)1U << EXTENT_BITS_NFREE_SHIFT);
+}
+
+static inline void
+extent_nfree_dec(extent_t *extent) {
+	assert(extent_slab_get(extent));
+	extent->e_bits -= ((uint64_t)1U << EXTENT_BITS_NFREE_SHIFT);
+}
+
+static inline void
+extent_nfree_sub(extent_t *extent, uint64_t n) {
+	assert(extent_slab_get(extent));
+	extent->e_bits -= (n << EXTENT_BITS_NFREE_SHIFT);
+}
+
+static inline void
+extent_sn_set(extent_t *extent, size_t sn) {
+	extent->e_bits = (extent->e_bits & ~EXTENT_BITS_SN_MASK) |
+	    ((uint64_t)sn << EXTENT_BITS_SN_SHIFT);
+}
+
+static inline void
+extent_state_set(extent_t *extent, extent_state_t state) {
+	extent->e_bits = (extent->e_bits & ~EXTENT_BITS_STATE_MASK) |
+	    ((uint64_t)state << EXTENT_BITS_STATE_SHIFT);
+}
+
+static inline void
+extent_zeroed_set(extent_t *extent, bool zeroed) {
+	extent->e_bits = (extent->e_bits & ~EXTENT_BITS_ZEROED_MASK) |
+	    ((uint64_t)zeroed << EXTENT_BITS_ZEROED_SHIFT);
+}
+
+static inline void
+extent_committed_set(extent_t *extent, bool committed) {
+	extent->e_bits = (extent->e_bits & ~EXTENT_BITS_COMMITTED_MASK) |
+	    ((uint64_t)committed << EXTENT_BITS_COMMITTED_SHIFT);
+}
+
+static inline void
+extent_dumpable_set(extent_t *extent, bool dumpable) {
+	extent->e_bits = (extent->e_bits & ~EXTENT_BITS_DUMPABLE_MASK) |
+	    ((uint64_t)dumpable << EXTENT_BITS_DUMPABLE_SHIFT);
+}
+
+static inline void
+extent_slab_set(extent_t *extent, bool slab) {
+	extent->e_bits = (extent->e_bits & ~EXTENT_BITS_SLAB_MASK) |
+	    ((uint64_t)slab << EXTENT_BITS_SLAB_SHIFT);
+}
+
+static inline void
+extent_prof_tctx_set(extent_t *extent, prof_tctx_t *tctx) {
+	atomic_store_p(&extent->e_prof_tctx, tctx, ATOMIC_RELEASE);
+}
+
+static inline void
+extent_prof_alloc_time_set(extent_t *extent, nstime_t t) {
+	nstime_copy(&extent->e_alloc_time, &t);
+}
+
+static inline bool
+extent_is_head_get(extent_t *extent) {
+	if (maps_coalesce) {
+		not_reached();
+	}
+
+	return (bool)((extent->e_bits & EXTENT_BITS_IS_HEAD_MASK) >>
+	    EXTENT_BITS_IS_HEAD_SHIFT);
+}
+
+static inline void
+extent_is_head_set(extent_t *extent, bool is_head) {
+	if (maps_coalesce) {
+		not_reached();
+	}
+
+	extent->e_bits = (extent->e_bits & ~EXTENT_BITS_IS_HEAD_MASK) |
+	    ((uint64_t)is_head << EXTENT_BITS_IS_HEAD_SHIFT);
+}
+
+static inline void
+extent_init(extent_t *extent, arena_t *arena, void *addr, size_t size,
+    bool slab, szind_t szind, size_t sn, extent_state_t state, bool zeroed,
+    bool committed, bool dumpable, extent_head_state_t is_head) {
+	assert(addr == PAGE_ADDR2BASE(addr) || !slab);
+
+	extent_arena_set(extent, arena);
+	extent_addr_set(extent, addr);
+	extent_size_set(extent, size);
+	extent_slab_set(extent, slab);
+	extent_szind_set(extent, szind);
+	extent_sn_set(extent, sn);
+	extent_state_set(extent, state);
+	extent_zeroed_set(extent, zeroed);
+	extent_committed_set(extent, committed);
+	extent_dumpable_set(extent, dumpable);
+	ql_elm_new(extent, ql_link);
+	if (!maps_coalesce) {
+		extent_is_head_set(extent, (is_head == EXTENT_IS_HEAD) ? true :
+		    false);
+	}
+	if (config_prof) {
+		extent_prof_tctx_set(extent, NULL);
+	}
+}
+
+static inline void
+extent_binit(extent_t *extent, void *addr, size_t bsize, size_t sn) {
+	extent_arena_set(extent, NULL);
+	extent_addr_set(extent, addr);
+	extent_bsize_set(extent, bsize);
+	extent_slab_set(extent, false);
+	extent_szind_set(extent, SC_NSIZES);
+	extent_sn_set(extent, sn);
+	extent_state_set(extent, extent_state_active);
+	extent_zeroed_set(extent, true);
+	extent_committed_set(extent, true);
+	extent_dumpable_set(extent, true);
+}
+
+static inline void
+extent_list_init(extent_list_t *list) {
+	ql_new(list);
+}
+
+static inline extent_t *
+extent_list_first(const extent_list_t *list) {
+	return ql_first(list);
+}
+
+static inline extent_t *
+extent_list_last(const extent_list_t *list) {
+	return ql_last(list, ql_link);
+}
+
+static inline void
+extent_list_append(extent_list_t *list, extent_t *extent) {
+	ql_tail_insert(list, extent, ql_link);
+}
+
+static inline void
+extent_list_prepend(extent_list_t *list, extent_t *extent) {
+	ql_head_insert(list, extent, ql_link);
+}
+
+static inline void
+extent_list_replace(extent_list_t *list, extent_t *to_remove,
+    extent_t *to_insert) {
+	ql_after_insert(to_remove, to_insert, ql_link);
+	ql_remove(list, to_remove, ql_link);
+}
+
+static inline void
+extent_list_remove(extent_list_t *list, extent_t *extent) {
+	ql_remove(list, extent, ql_link);
+}
+
+static inline int
+extent_sn_comp(const extent_t *a, const extent_t *b) {
+	size_t a_sn = extent_sn_get(a);
+	size_t b_sn = extent_sn_get(b);
+
+	return (a_sn > b_sn) - (a_sn < b_sn);
+}
+
+static inline int
+extent_esn_comp(const extent_t *a, const extent_t *b) {
+	size_t a_esn = extent_esn_get(a);
+	size_t b_esn = extent_esn_get(b);
+
+	return (a_esn > b_esn) - (a_esn < b_esn);
+}
+
+static inline int
+extent_ad_comp(const extent_t *a, const extent_t *b) {
+	uintptr_t a_addr = (uintptr_t)extent_addr_get(a);
+	uintptr_t b_addr = (uintptr_t)extent_addr_get(b);
+
+	return (a_addr > b_addr) - (a_addr < b_addr);
+}
+
+static inline int
+extent_ead_comp(const extent_t *a, const extent_t *b) {
+	uintptr_t a_eaddr = (uintptr_t)a;
+	uintptr_t b_eaddr = (uintptr_t)b;
+
+	return (a_eaddr > b_eaddr) - (a_eaddr < b_eaddr);
+}
+
+static inline int
+extent_snad_comp(const extent_t *a, const extent_t *b) {
+	int ret;
+
+	ret = extent_sn_comp(a, b);
+	if (ret != 0) {
+		return ret;
+	}
+
+	ret = extent_ad_comp(a, b);
+	return ret;
+}
+
+static inline int
+extent_esnead_comp(const extent_t *a, const extent_t *b) {
+	int ret;
+
+	ret = extent_esn_comp(a, b);
+	if (ret != 0) {
+		return ret;
+	}
+
+	ret = extent_ead_comp(a, b);
+	return ret;
+}
+
+#endif /* JEMALLOC_INTERNAL_EXTENT_INLINES_H */

deps/jemalloc/include/jemalloc/internal/extent_mmap.h

@@ -0,0 +1,10 @@
+#ifndef JEMALLOC_INTERNAL_EXTENT_MMAP_EXTERNS_H
+#define JEMALLOC_INTERNAL_EXTENT_MMAP_EXTERNS_H
+
+extern bool opt_retain;
+
+void *extent_alloc_mmap(void *new_addr, size_t size, size_t alignment,
+    bool *zero, bool *commit);
+bool extent_dalloc_mmap(void *addr, size_t size);
+
+#endif /* JEMALLOC_INTERNAL_EXTENT_MMAP_EXTERNS_H */

deps/jemalloc/include/jemalloc/internal/extent_structs.h

@@ -0,0 +1,256 @@
+#ifndef JEMALLOC_INTERNAL_EXTENT_STRUCTS_H
+#define JEMALLOC_INTERNAL_EXTENT_STRUCTS_H
+
+#include "jemalloc/internal/atomic.h"
+#include "jemalloc/internal/bit_util.h"
+#include "jemalloc/internal/bitmap.h"
+#include "jemalloc/internal/mutex.h"
+#include "jemalloc/internal/ql.h"
+#include "jemalloc/internal/ph.h"
+#include "jemalloc/internal/sc.h"
+
+typedef enum {
+	extent_state_active   = 0,
+	extent_state_dirty    = 1,
+	extent_state_muzzy    = 2,
+	extent_state_retained = 3
+} extent_state_t;
+
+/* Extent (span of pages).  Use accessor functions for e_* fields. */
+struct extent_s {
+	/*
+	 * Bitfield containing several fields:
+	 *
+	 * a: arena_ind
+	 * b: slab
+	 * c: committed
+	 * d: dumpable
+	 * z: zeroed
+	 * t: state
+	 * i: szind
+	 * f: nfree
+	 * s: bin_shard
+	 * n: sn
+	 *
+	 * nnnnnnnn ... nnnnnnss ssssffff ffffffii iiiiiitt zdcbaaaa aaaaaaaa
+	 *
+	 * arena_ind: Arena from which this extent came, or all 1 bits if
+	 *            unassociated.
+	 *
+	 * slab: The slab flag indicates whether the extent is used for a slab
+	 *       of small regions.  This helps differentiate small size classes,
+	 *       and it indicates whether interior pointers can be looked up via
+	 *       iealloc().
+	 *
+	 * committed: The committed flag indicates whether physical memory is
+	 *            committed to the extent, whether explicitly or implicitly
+	 *            as on a system that overcommits and satisfies physical
+	 *            memory needs on demand via soft page faults.
+	 *
+	 * dumpable: The dumpable flag indicates whether or not we've set the
+	 *           memory in question to be dumpable.  Note that this
+	 *           interacts somewhat subtly with user-specified extent hooks,
+	 *           since we don't know if *they* are fiddling with
+	 *           dumpability (in which case, we don't want to undo whatever
+	 *           they're doing).  To deal with this scenario, we:
+	 *             - Make dumpable false only for memory allocated with the
+	 *               default hooks.
+	 *             - Only allow memory to go from non-dumpable to dumpable,
+	 *               and only once.
+	 *             - Never make the OS call to allow dumping when the
+	 *               dumpable bit is already set.
+	 *           These three constraints mean that we will never
+	 *           accidentally dump user memory that the user meant to set
+	 *           nondumpable with their extent hooks.
+	 *
+	 *
+	 * zeroed: The zeroed flag is used by extent recycling code to track
+	 *         whether memory is zero-filled.
+	 *
+	 * state: The state flag is an extent_state_t.
+	 *
+	 * szind: The szind flag indicates usable size class index for
+	 *        allocations residing in this extent, regardless of whether the
+	 *        extent is a slab.  Extent size and usable size often differ
+	 *        even for non-slabs, either due to sz_large_pad or promotion of
+	 *        sampled small regions.
+	 *
+	 * nfree: Number of free regions in slab.
+	 *
+	 * bin_shard: the shard of the bin from which this extent came.
+	 *
+	 * sn: Serial number (potentially non-unique).
+	 *
+	 *     Serial numbers may wrap around if !opt_retain, but as long as
+	 *     comparison functions fall back on address comparison for equal
+	 *     serial numbers, stable (if imperfect) ordering is maintained.
+	 *
+	 *     Serial numbers may not be unique even in the absence of
+	 *     wrap-around, e.g. when splitting an extent and assigning the same
+	 *     serial number to both resulting adjacent extents.
+	 */
+	uint64_t		e_bits;
+#define MASK(CURRENT_FIELD_WIDTH, CURRENT_FIELD_SHIFT) ((((((uint64_t)0x1U) << (CURRENT_FIELD_WIDTH)) - 1)) << (CURRENT_FIELD_SHIFT))
+
+#define EXTENT_BITS_ARENA_WIDTH  MALLOCX_ARENA_BITS
+#define EXTENT_BITS_ARENA_SHIFT  0
+#define EXTENT_BITS_ARENA_MASK  MASK(EXTENT_BITS_ARENA_WIDTH, EXTENT_BITS_ARENA_SHIFT)
+
+#define EXTENT_BITS_SLAB_WIDTH  1
+#define EXTENT_BITS_SLAB_SHIFT  (EXTENT_BITS_ARENA_WIDTH + EXTENT_BITS_ARENA_SHIFT)
+#define EXTENT_BITS_SLAB_MASK  MASK(EXTENT_BITS_SLAB_WIDTH, EXTENT_BITS_SLAB_SHIFT)
+
+#define EXTENT_BITS_COMMITTED_WIDTH  1
+#define EXTENT_BITS_COMMITTED_SHIFT  (EXTENT_BITS_SLAB_WIDTH + EXTENT_BITS_SLAB_SHIFT)
+#define EXTENT_BITS_COMMITTED_MASK  MASK(EXTENT_BITS_COMMITTED_WIDTH, EXTENT_BITS_COMMITTED_SHIFT)
+
+#define EXTENT_BITS_DUMPABLE_WIDTH  1
+#define EXTENT_BITS_DUMPABLE_SHIFT  (EXTENT_BITS_COMMITTED_WIDTH + EXTENT_BITS_COMMITTED_SHIFT)
+#define EXTENT_BITS_DUMPABLE_MASK  MASK(EXTENT_BITS_DUMPABLE_WIDTH, EXTENT_BITS_DUMPABLE_SHIFT)
+
+#define EXTENT_BITS_ZEROED_WIDTH  1
+#define EXTENT_BITS_ZEROED_SHIFT  (EXTENT_BITS_DUMPABLE_WIDTH + EXTENT_BITS_DUMPABLE_SHIFT)
+#define EXTENT_BITS_ZEROED_MASK  MASK(EXTENT_BITS_ZEROED_WIDTH, EXTENT_BITS_ZEROED_SHIFT)
+
+#define EXTENT_BITS_STATE_WIDTH  2
+#define EXTENT_BITS_STATE_SHIFT  (EXTENT_BITS_ZEROED_WIDTH + EXTENT_BITS_ZEROED_SHIFT)
+#define EXTENT_BITS_STATE_MASK  MASK(EXTENT_BITS_STATE_WIDTH, EXTENT_BITS_STATE_SHIFT)
+
+#define EXTENT_BITS_SZIND_WIDTH  LG_CEIL(SC_NSIZES)
+#define EXTENT_BITS_SZIND_SHIFT  (EXTENT_BITS_STATE_WIDTH + EXTENT_BITS_STATE_SHIFT)
+#define EXTENT_BITS_SZIND_MASK  MASK(EXTENT_BITS_SZIND_WIDTH, EXTENT_BITS_SZIND_SHIFT)
+
+#define EXTENT_BITS_NFREE_WIDTH  (LG_SLAB_MAXREGS + 1)
+#define EXTENT_BITS_NFREE_SHIFT  (EXTENT_BITS_SZIND_WIDTH + EXTENT_BITS_SZIND_SHIFT)
+#define EXTENT_BITS_NFREE_MASK  MASK(EXTENT_BITS_NFREE_WIDTH, EXTENT_BITS_NFREE_SHIFT)
+
+#define EXTENT_BITS_BINSHARD_WIDTH  6
+#define EXTENT_BITS_BINSHARD_SHIFT  (EXTENT_BITS_NFREE_WIDTH + EXTENT_BITS_NFREE_SHIFT)
+#define EXTENT_BITS_BINSHARD_MASK  MASK(EXTENT_BITS_BINSHARD_WIDTH, EXTENT_BITS_BINSHARD_SHIFT)
+
+#define EXTENT_BITS_IS_HEAD_WIDTH 1
+#define EXTENT_BITS_IS_HEAD_SHIFT  (EXTENT_BITS_BINSHARD_WIDTH + EXTENT_BITS_BINSHARD_SHIFT)
+#define EXTENT_BITS_IS_HEAD_MASK  MASK(EXTENT_BITS_IS_HEAD_WIDTH, EXTENT_BITS_IS_HEAD_SHIFT)
+
+#define EXTENT_BITS_SN_SHIFT   (EXTENT_BITS_IS_HEAD_WIDTH + EXTENT_BITS_IS_HEAD_SHIFT)
+#define EXTENT_BITS_SN_MASK  (UINT64_MAX << EXTENT_BITS_SN_SHIFT)
+
+	/* Pointer to the extent that this structure is responsible for. */
+	void			*e_addr;
+
+	union {
+		/*
+		 * Extent size and serial number associated with the extent
+		 * structure (different than the serial number for the extent at
+		 * e_addr).
+		 *
+		 * ssssssss [...] ssssssss ssssnnnn nnnnnnnn
+		 */
+		size_t			e_size_esn;
+	#define EXTENT_SIZE_MASK	((size_t)~(PAGE-1))
+	#define EXTENT_ESN_MASK		((size_t)PAGE-1)
+		/* Base extent size, which may not be a multiple of PAGE. */
+		size_t			e_bsize;
+	};
+
+	/*
+	 * List linkage, used by a variety of lists:
+	 * - bin_t's slabs_full
+	 * - extents_t's LRU
+	 * - stashed dirty extents
+	 * - arena's large allocations
+	 */
+	ql_elm(extent_t)	ql_link;
+
+	/*
+	 * Linkage for per size class sn/address-ordered heaps, and
+	 * for extent_avail
+	 */
+	phn(extent_t)		ph_link;
+
+	union {
+		/* Small region slab metadata. */
+		arena_slab_data_t	e_slab_data;
+
+		/* Profiling data, used for large objects. */
+		struct {
+			/* Time when this was allocated. */
+			nstime_t		e_alloc_time;
+			/* Points to a prof_tctx_t. */
+			atomic_p_t		e_prof_tctx;
+		};
+	};
+};
+typedef ql_head(extent_t) extent_list_t;
+typedef ph(extent_t) extent_tree_t;
+typedef ph(extent_t) extent_heap_t;
+
+/* Quantized collection of extents, with built-in LRU queue. */
+struct extents_s {
+	malloc_mutex_t		mtx;
+
+	/*
+	 * Quantized per size class heaps of extents.
+	 *
+	 * Synchronization: mtx.
+	 */
+	extent_heap_t		heaps[SC_NPSIZES + 1];
+	atomic_zu_t		nextents[SC_NPSIZES + 1];
+	atomic_zu_t		nbytes[SC_NPSIZES + 1];
+
+	/*
+	 * Bitmap for which set bits correspond to non-empty heaps.
+	 *
+	 * Synchronization: mtx.
+	 */
+	bitmap_t		bitmap[BITMAP_GROUPS(SC_NPSIZES + 1)];
+
+	/*
+	 * LRU of all extents in heaps.
+	 *
+	 * Synchronization: mtx.
+	 */
+	extent_list_t		lru;
+
+	/*
+	 * Page sum for all extents in heaps.
+	 *
+	 * The synchronization here is a little tricky.  Modifications to npages
+	 * must hold mtx, but reads need not (though, a reader who sees npages
+	 * without holding the mutex can't assume anything about the rest of the
+	 * state of the extents_t).
+	 */
+	atomic_zu_t		npages;
+
+	/* All stored extents must be in the same state. */
+	extent_state_t		state;
+
+	/*
+	 * If true, delay coalescing until eviction; otherwise coalesce during
+	 * deallocation.
+	 */
+	bool			delay_coalesce;
+};
+
+/*
+ * The following two structs are for experimental purposes. See
+ * experimental_utilization_query_ctl and
+ * experimental_utilization_batch_query_ctl in src/ctl.c.
+ */
+
+struct extent_util_stats_s {
+	size_t nfree;
+	size_t nregs;
+	size_t size;
+};
+
+struct extent_util_stats_verbose_s {
+	void *slabcur_addr;
+	size_t nfree;
+	size_t nregs;
+	size_t size;
+	size_t bin_nfree;
+	size_t bin_nregs;
+};
+
+#endif /* JEMALLOC_INTERNAL_EXTENT_STRUCTS_H */

deps/jemalloc/include/jemalloc/internal/extent_types.h

@@ -0,0 +1,23 @@
+#ifndef JEMALLOC_INTERNAL_EXTENT_TYPES_H
+#define JEMALLOC_INTERNAL_EXTENT_TYPES_H
+
+typedef struct extent_s extent_t;
+typedef struct extents_s extents_t;
+
+typedef struct extent_util_stats_s extent_util_stats_t;
+typedef struct extent_util_stats_verbose_s extent_util_stats_verbose_t;
+
+#define EXTENT_HOOKS_INITIALIZER	NULL
+
+/*
+ * When reuse (and split) an active extent, (1U << opt_lg_extent_max_active_fit)
+ * is the max ratio between the size of the active extent and the new extent.
+ */
+#define LG_EXTENT_MAX_ACTIVE_FIT_DEFAULT 6
+
+typedef enum {
+	EXTENT_NOT_HEAD,
+	EXTENT_IS_HEAD   /* Only relevant for Windows && opt.retain. */
+} extent_head_state_t;
+
+#endif /* JEMALLOC_INTERNAL_EXTENT_TYPES_H */

deps/jemalloc/include/jemalloc/internal/hash.h

@@ -0,0 +1,319 @@
+#ifndef JEMALLOC_INTERNAL_HASH_H
+#define JEMALLOC_INTERNAL_HASH_H
+
+#include "jemalloc/internal/assert.h"
+
+/*
+ * The following hash function is based on MurmurHash3, placed into the public
+ * domain by Austin Appleby.  See https://github.com/aappleby/smhasher for
+ * details.
+ */
+
+/******************************************************************************/
+/* Internal implementation. */
+static inline uint32_t
+hash_rotl_32(uint32_t x, int8_t r) {
+	return ((x << r) | (x >> (32 - r)));
+}
+
+static inline uint64_t
+hash_rotl_64(uint64_t x, int8_t r) {
+	return ((x << r) | (x >> (64 - r)));
+}
+
+static inline uint32_t
+hash_get_block_32(const uint32_t *p, int i) {
+	/* Handle unaligned read. */
+	if (unlikely((uintptr_t)p & (sizeof(uint32_t)-1)) != 0) {
+		uint32_t ret;
+
+		memcpy(&ret, (uint8_t *)(p + i), sizeof(uint32_t));
+		return ret;
+	}
+
+	return p[i];
+}
+
+static inline uint64_t
+hash_get_block_64(const uint64_t *p, int i) {
+	/* Handle unaligned read. */
+	if (unlikely((uintptr_t)p & (sizeof(uint64_t)-1)) != 0) {
+		uint64_t ret;
+
+		memcpy(&ret, (uint8_t *)(p + i), sizeof(uint64_t));
+		return ret;
+	}
+
+	return p[i];
+}
+
+static inline uint32_t
+hash_fmix_32(uint32_t h) {
+	h ^= h >> 16;
+	h *= 0x85ebca6b;
+	h ^= h >> 13;
+	h *= 0xc2b2ae35;
+	h ^= h >> 16;
+
+	return h;
+}
+
+static inline uint64_t
+hash_fmix_64(uint64_t k) {
+	k ^= k >> 33;
+	k *= KQU(0xff51afd7ed558ccd);
+	k ^= k >> 33;
+	k *= KQU(0xc4ceb9fe1a85ec53);
+	k ^= k >> 33;
+
+	return k;
+}
+
+static inline uint32_t
+hash_x86_32(const void *key, int len, uint32_t seed) {
+	const uint8_t *data = (const uint8_t *) key;
+	const int nblocks = len / 4;
+
+	uint32_t h1 = seed;
+
+	const uint32_t c1 = 0xcc9e2d51;
+	const uint32_t c2 = 0x1b873593;
+
+	/* body */
+	{
+		const uint32_t *blocks = (const uint32_t *) (data + nblocks*4);
+		int i;
+
+		for (i = -nblocks; i; i++) {
+			uint32_t k1 = hash_get_block_32(blocks, i);
+
+			k1 *= c1;
+			k1 = hash_rotl_32(k1, 15);
+			k1 *= c2;
+
+			h1 ^= k1;
+			h1 = hash_rotl_32(h1, 13);
+			h1 = h1*5 + 0xe6546b64;
+		}
+	}
+
+	/* tail */
+	{
+		const uint8_t *tail = (const uint8_t *) (data + nblocks*4);
+
+		uint32_t k1 = 0;
+
+		switch (len & 3) {
+		case 3: k1 ^= tail[2] << 16; JEMALLOC_FALLTHROUGH
+		case 2: k1 ^= tail[1] << 8; JEMALLOC_FALLTHROUGH
+		case 1: k1 ^= tail[0]; k1 *= c1; k1 = hash_rotl_32(k1, 15);
+			k1 *= c2; h1 ^= k1;
+		}
+	}
+
+	/* finalization */
+	h1 ^= len;
+
+	h1 = hash_fmix_32(h1);
+
+	return h1;
+}
+
+static inline void
+hash_x86_128(const void *key, const int len, uint32_t seed,
+    uint64_t r_out[2]) {
+	const uint8_t * data = (const uint8_t *) key;
+	const int nblocks = len / 16;
+
+	uint32_t h1 = seed;
+	uint32_t h2 = seed;
+	uint32_t h3 = seed;
+	uint32_t h4 = seed;
+
+	const uint32_t c1 = 0x239b961b;
+	const uint32_t c2 = 0xab0e9789;
+	const uint32_t c3 = 0x38b34ae5;
+	const uint32_t c4 = 0xa1e38b93;
+
+	/* body */
+	{
+		const uint32_t *blocks = (const uint32_t *) (data + nblocks*16);
+		int i;
+
+		for (i = -nblocks; i; i++) {
+			uint32_t k1 = hash_get_block_32(blocks, i*4 + 0);
+			uint32_t k2 = hash_get_block_32(blocks, i*4 + 1);
+			uint32_t k3 = hash_get_block_32(blocks, i*4 + 2);
+			uint32_t k4 = hash_get_block_32(blocks, i*4 + 3);
+
+			k1 *= c1; k1 = hash_rotl_32(k1, 15); k1 *= c2; h1 ^= k1;
+
+			h1 = hash_rotl_32(h1, 19); h1 += h2;
+			h1 = h1*5 + 0x561ccd1b;
+
+			k2 *= c2; k2 = hash_rotl_32(k2, 16); k2 *= c3; h2 ^= k2;
+
+			h2 = hash_rotl_32(h2, 17); h2 += h3;
+			h2 = h2*5 + 0x0bcaa747;
+
+			k3 *= c3; k3 = hash_rotl_32(k3, 17); k3 *= c4; h3 ^= k3;
+
+			h3 = hash_rotl_32(h3, 15); h3 += h4;
+			h3 = h3*5 + 0x96cd1c35;
+
+			k4 *= c4; k4 = hash_rotl_32(k4, 18); k4 *= c1; h4 ^= k4;
+
+			h4 = hash_rotl_32(h4, 13); h4 += h1;
+			h4 = h4*5 + 0x32ac3b17;
+		}
+	}
+
+	/* tail */
+	{
+		const uint8_t *tail = (const uint8_t *) (data + nblocks*16);
+		uint32_t k1 = 0;
+		uint32_t k2 = 0;
+		uint32_t k3 = 0;
+		uint32_t k4 = 0;
+
+		switch (len & 15) {
+		case 15: k4 ^= tail[14] << 16; JEMALLOC_FALLTHROUGH
+		case 14: k4 ^= tail[13] << 8; JEMALLOC_FALLTHROUGH
+		case 13: k4 ^= tail[12] << 0;
+			k4 *= c4; k4 = hash_rotl_32(k4, 18); k4 *= c1; h4 ^= k4;
+      JEMALLOC_FALLTHROUGH
+		case 12: k3 ^= tail[11] << 24; JEMALLOC_FALLTHROUGH
+		case 11: k3 ^= tail[10] << 16; JEMALLOC_FALLTHROUGH
+		case 10: k3 ^= tail[ 9] << 8; JEMALLOC_FALLTHROUGH
+		case  9: k3 ^= tail[ 8] << 0;
+		     k3 *= c3; k3 = hash_rotl_32(k3, 17); k3 *= c4; h3 ^= k3;
+         JEMALLOC_FALLTHROUGH
+		case  8: k2 ^= tail[ 7] << 24; JEMALLOC_FALLTHROUGH
+		case  7: k2 ^= tail[ 6] << 16; JEMALLOC_FALLTHROUGH
+		case  6: k2 ^= tail[ 5] << 8; JEMALLOC_FALLTHROUGH
+		case  5: k2 ^= tail[ 4] << 0;
+			k2 *= c2; k2 = hash_rotl_32(k2, 16); k2 *= c3; h2 ^= k2;
+      JEMALLOC_FALLTHROUGH
+		case  4: k1 ^= tail[ 3] << 24; JEMALLOC_FALLTHROUGH
+		case  3: k1 ^= tail[ 2] << 16; JEMALLOC_FALLTHROUGH
+		case  2: k1 ^= tail[ 1] << 8; JEMALLOC_FALLTHROUGH
+		case  1: k1 ^= tail[ 0] << 0;
+			k1 *= c1; k1 = hash_rotl_32(k1, 15); k1 *= c2; h1 ^= k1;
+      JEMALLOC_FALLTHROUGH
+		}
+	}
+
+	/* finalization */
+	h1 ^= len; h2 ^= len; h3 ^= len; h4 ^= len;
+
+	h1 += h2; h1 += h3; h1 += h4;
+	h2 += h1; h3 += h1; h4 += h1;
+
+	h1 = hash_fmix_32(h1);
+	h2 = hash_fmix_32(h2);
+	h3 = hash_fmix_32(h3);
+	h4 = hash_fmix_32(h4);
+
+	h1 += h2; h1 += h3; h1 += h4;
+	h2 += h1; h3 += h1; h4 += h1;
+
+	r_out[0] = (((uint64_t) h2) << 32) | h1;
+	r_out[1] = (((uint64_t) h4) << 32) | h3;
+}
+
+static inline void
+hash_x64_128(const void *key, const int len, const uint32_t seed,
+    uint64_t r_out[2]) {
+	const uint8_t *data = (const uint8_t *) key;
+	const int nblocks = len / 16;
+
+	uint64_t h1 = seed;
+	uint64_t h2 = seed;
+
+	const uint64_t c1 = KQU(0x87c37b91114253d5);
+	const uint64_t c2 = KQU(0x4cf5ad432745937f);
+
+	/* body */
+	{
+		const uint64_t *blocks = (const uint64_t *) (data);
+		int i;
+
+		for (i = 0; i < nblocks; i++) {
+			uint64_t k1 = hash_get_block_64(blocks, i*2 + 0);
+			uint64_t k2 = hash_get_block_64(blocks, i*2 + 1);
+
+			k1 *= c1; k1 = hash_rotl_64(k1, 31); k1 *= c2; h1 ^= k1;
+
+			h1 = hash_rotl_64(h1, 27); h1 += h2;
+			h1 = h1*5 + 0x52dce729;
+
+			k2 *= c2; k2 = hash_rotl_64(k2, 33); k2 *= c1; h2 ^= k2;
+
+			h2 = hash_rotl_64(h2, 31); h2 += h1;
+			h2 = h2*5 + 0x38495ab5;
+		}
+	}
+
+	/* tail */
+	{
+		const uint8_t *tail = (const uint8_t*)(data + nblocks*16);
+		uint64_t k1 = 0;
+		uint64_t k2 = 0;
+
+		switch (len & 15) {
+		case 15: k2 ^= ((uint64_t)(tail[14])) << 48; JEMALLOC_FALLTHROUGH
+		case 14: k2 ^= ((uint64_t)(tail[13])) << 40; JEMALLOC_FALLTHROUGH
+		case 13: k2 ^= ((uint64_t)(tail[12])) << 32; JEMALLOC_FALLTHROUGH
+		case 12: k2 ^= ((uint64_t)(tail[11])) << 24; JEMALLOC_FALLTHROUGH
+		case 11: k2 ^= ((uint64_t)(tail[10])) << 16; JEMALLOC_FALLTHROUGH
+		case 10: k2 ^= ((uint64_t)(tail[ 9])) << 8;  JEMALLOC_FALLTHROUGH
+		case  9: k2 ^= ((uint64_t)(tail[ 8])) << 0;
+			k2 *= c2; k2 = hash_rotl_64(k2, 33); k2 *= c1; h2 ^= k2;
+			JEMALLOC_FALLTHROUGH
+		case  8: k1 ^= ((uint64_t)(tail[ 7])) << 56; JEMALLOC_FALLTHROUGH
+		case  7: k1 ^= ((uint64_t)(tail[ 6])) << 48; JEMALLOC_FALLTHROUGH
+		case  6: k1 ^= ((uint64_t)(tail[ 5])) << 40; JEMALLOC_FALLTHROUGH
+		case  5: k1 ^= ((uint64_t)(tail[ 4])) << 32; JEMALLOC_FALLTHROUGH
+		case  4: k1 ^= ((uint64_t)(tail[ 3])) << 24; JEMALLOC_FALLTHROUGH
+		case  3: k1 ^= ((uint64_t)(tail[ 2])) << 16; JEMALLOC_FALLTHROUGH
+		case  2: k1 ^= ((uint64_t)(tail[ 1])) << 8;  JEMALLOC_FALLTHROUGH
+		case  1: k1 ^= ((uint64_t)(tail[ 0])) << 0;
+			k1 *= c1; k1 = hash_rotl_64(k1, 31); k1 *= c2; h1 ^= k1;
+		}
+	}
+
+	/* finalization */
+	h1 ^= len; h2 ^= len;
+
+	h1 += h2;
+	h2 += h1;
+
+	h1 = hash_fmix_64(h1);
+	h2 = hash_fmix_64(h2);
+
+	h1 += h2;
+	h2 += h1;
+
+	r_out[0] = h1;
+	r_out[1] = h2;
+}
+
+/******************************************************************************/
+/* API. */
+static inline void
+hash(const void *key, size_t len, const uint32_t seed, size_t r_hash[2]) {
+	assert(len <= INT_MAX); /* Unfortunate implementation limitation. */
+
+#if (LG_SIZEOF_PTR == 3 && !defined(JEMALLOC_BIG_ENDIAN))
+	hash_x64_128(key, (int)len, seed, (uint64_t *)r_hash);
+#else
+	{
+		uint64_t hashes[2];
+		hash_x86_128(key, (int)len, seed, hashes);
+		r_hash[0] = (size_t)hashes[0];
+		r_hash[1] = (size_t)hashes[1];
+	}
+#endif
+}
+
+#endif /* JEMALLOC_INTERNAL_HASH_H */

deps/jemalloc/include/jemalloc/internal/hook.h

@@ -0,0 +1,163 @@
+#ifndef JEMALLOC_INTERNAL_HOOK_H
+#define JEMALLOC_INTERNAL_HOOK_H
+
+#include "jemalloc/internal/tsd.h"
+
+/*
+ * This API is *extremely* experimental, and may get ripped out, changed in API-
+ * and ABI-incompatible ways, be insufficiently or incorrectly documented, etc.
+ *
+ * It allows hooking the stateful parts of the API to see changes as they
+ * happen.
+ *
+ * Allocation hooks are called after the allocation is done, free hooks are
+ * called before the free is done, and expand hooks are called after the
+ * allocation is expanded.
+ *
+ * For realloc and rallocx, if the expansion happens in place, the expansion
+ * hook is called.  If it is moved, then the alloc hook is called on the new
+ * location, and then the free hook is called on the old location (i.e. both
+ * hooks are invoked in between the alloc and the dalloc).
+ *
+ * If we return NULL from OOM, then usize might not be trustworthy.  Calling
+ * realloc(NULL, size) only calls the alloc hook, and calling realloc(ptr, 0)
+ * only calls the free hook.  (Calling realloc(NULL, 0) is treated as malloc(0),
+ * and only calls the alloc hook).
+ *
+ * Reentrancy:
+ *   Reentrancy is guarded against from within the hook implementation.  If you
+ *   call allocator functions from within a hook, the hooks will not be invoked
+ *   again.
+ * Threading:
+ *   The installation of a hook synchronizes with all its uses.  If you can
+ *   prove the installation of a hook happens-before a jemalloc entry point,
+ *   then the hook will get invoked (unless there's a racing removal).
+ *
+ *   Hook insertion appears to be atomic at a per-thread level (i.e. if a thread
+ *   allocates and has the alloc hook invoked, then a subsequent free on the
+ *   same thread will also have the free hook invoked).
+ *
+ *   The *removal* of a hook does *not* block until all threads are done with
+ *   the hook.  Hook authors have to be resilient to this, and need some
+ *   out-of-band mechanism for cleaning up any dynamically allocated memory
+ *   associated with their hook.
+ * Ordering:
+ *   Order of hook execution is unspecified, and may be different than insertion
+ *   order.
+ */
+
+#define HOOK_MAX 4
+
+enum hook_alloc_e {
+	hook_alloc_malloc,
+	hook_alloc_posix_memalign,
+	hook_alloc_aligned_alloc,
+	hook_alloc_calloc,
+	hook_alloc_memalign,
+	hook_alloc_valloc,
+	hook_alloc_mallocx,
+
+	/* The reallocating functions have both alloc and dalloc variants */
+	hook_alloc_realloc,
+	hook_alloc_rallocx,
+};
+/*
+ * We put the enum typedef after the enum, since this file may get included by
+ * jemalloc_cpp.cpp, and C++ disallows enum forward declarations.
+ */
+typedef enum hook_alloc_e hook_alloc_t;
+
+enum hook_dalloc_e {
+	hook_dalloc_free,
+	hook_dalloc_dallocx,
+	hook_dalloc_sdallocx,
+
+	/*
+	 * The dalloc halves of reallocation (not called if in-place expansion
+	 * happens).
+	 */
+	hook_dalloc_realloc,
+	hook_dalloc_rallocx,
+};
+typedef enum hook_dalloc_e hook_dalloc_t;
+
+
+enum hook_expand_e {
+	hook_expand_realloc,
+	hook_expand_rallocx,
+	hook_expand_xallocx,
+};
+typedef enum hook_expand_e hook_expand_t;
+
+typedef void (*hook_alloc)(
+    void *extra, hook_alloc_t type, void *result, uintptr_t result_raw,
+    uintptr_t args_raw[3]);
+
+typedef void (*hook_dalloc)(
+    void *extra, hook_dalloc_t type, void *address, uintptr_t args_raw[3]);
+
+typedef void (*hook_expand)(
+    void *extra, hook_expand_t type, void *address, size_t old_usize,
+    size_t new_usize, uintptr_t result_raw, uintptr_t args_raw[4]);
+
+typedef struct hooks_s hooks_t;
+struct hooks_s {
+	hook_alloc alloc_hook;
+	hook_dalloc dalloc_hook;
+	hook_expand expand_hook;
+	void *extra;
+};
+
+/*
+ * Begin implementation details; everything above this point might one day live
+ * in a public API.  Everything below this point never will.
+ */
+
+/*
+ * The realloc pathways haven't gotten any refactoring love in a while, and it's
+ * fairly difficult to pass information from the entry point to the hooks.  We
+ * put the informaiton the hooks will need into a struct to encapsulate
+ * everything.
+ *
+ * Much of these pathways are force-inlined, so that the compiler can avoid
+ * materializing this struct until we hit an extern arena function.  For fairly
+ * goofy reasons, *many* of the realloc paths hit an extern arena function.
+ * These paths are cold enough that it doesn't matter; eventually, we should
+ * rewrite the realloc code to make the expand-in-place and the
+ * free-then-realloc paths more orthogonal, at which point we don't need to
+ * spread the hook logic all over the place.
+ */
+typedef struct hook_ralloc_args_s hook_ralloc_args_t;
+struct hook_ralloc_args_s {
+	/* I.e. as opposed to rallocx. */
+	bool is_realloc;
+	/*
+	 * The expand hook takes 4 arguments, even if only 3 are actually used;
+	 * we add an extra one in case the user decides to memcpy without
+	 * looking too closely at the hooked function.
+	 */
+	uintptr_t args[4];
+};
+
+/*
+ * Returns an opaque handle to be used when removing the hook.  NULL means that
+ * we couldn't install the hook.
+ */
+bool hook_boot();
+
+void *hook_install(tsdn_t *tsdn, hooks_t *hooks);
+/* Uninstalls the hook with the handle previously returned from hook_install. */
+void hook_remove(tsdn_t *tsdn, void *opaque);
+
+/* Hooks */
+
+void hook_invoke_alloc(hook_alloc_t type, void *result, uintptr_t result_raw,
+    uintptr_t args_raw[3]);
+
+void hook_invoke_dalloc(hook_dalloc_t type, void *address,
+    uintptr_t args_raw[3]);
+
+void hook_invoke_expand(hook_expand_t type, void *address, size_t old_usize,
+    size_t new_usize, uintptr_t result_raw, uintptr_t args_raw[4]);
+
+#endif /* JEMALLOC_INTERNAL_HOOK_H */

deps/jemalloc/include/jemalloc/internal/hooks.h

@@ -0,0 +1,19 @@
+#ifndef JEMALLOC_INTERNAL_HOOKS_H
+#define JEMALLOC_INTERNAL_HOOKS_H
+
+extern JEMALLOC_EXPORT void (*hooks_arena_new_hook)();
+extern JEMALLOC_EXPORT void (*hooks_libc_hook)();
+
+#define JEMALLOC_HOOK(fn, hook) ((void)(hook != NULL && (hook(), 0)), fn)
+
+#define open JEMALLOC_HOOK(open, hooks_libc_hook)
+#define read JEMALLOC_HOOK(read, hooks_libc_hook)
+#define write JEMALLOC_HOOK(write, hooks_libc_hook)
+#define readlink JEMALLOC_HOOK(readlink, hooks_libc_hook)
+#define close JEMALLOC_HOOK(close, hooks_libc_hook)
+#define creat JEMALLOC_HOOK(creat, hooks_libc_hook)
+#define secure_getenv JEMALLOC_HOOK(secure_getenv, hooks_libc_hook)
+/* Note that this is undef'd and re-define'd in src/prof.c. */
+#define _Unwind_Backtrace JEMALLOC_HOOK(_Unwind_Backtrace, hooks_libc_hook)
+
+#endif /* JEMALLOC_INTERNAL_HOOKS_H */

deps/jemalloc/include/jemalloc/internal/huge.h

@@ -0,0 +1,46 @@
+/******************************************************************************/
+#ifdef JEMALLOC_H_TYPES
+
+#endif /* JEMALLOC_H_TYPES */
+/******************************************************************************/
+#ifdef JEMALLOC_H_STRUCTS
+
+#endif /* JEMALLOC_H_STRUCTS */
+/******************************************************************************/
+#ifdef JEMALLOC_H_EXTERNS
+
+/* Huge allocation statistics. */
+extern uint64_t		huge_nmalloc;
+extern uint64_t		huge_ndalloc;
+extern size_t		huge_allocated;
+
+/* Protects chunk-related data structures. */
+extern malloc_mutex_t	huge_mtx;
+
+void	*huge_malloc(size_t size, bool zero, dss_prec_t dss_prec);
+void	*huge_palloc(size_t size, size_t alignment, bool zero,
+    dss_prec_t dss_prec);
+bool	huge_ralloc_no_move(void *ptr, size_t oldsize, size_t size,
+    size_t extra);
+void	*huge_ralloc(void *ptr, size_t oldsize, size_t size, size_t extra,
+    size_t alignment, bool zero, bool try_tcache_dalloc, dss_prec_t dss_prec);
+#ifdef JEMALLOC_JET
+typedef void (huge_dalloc_junk_t)(void *, size_t);
+extern huge_dalloc_junk_t *huge_dalloc_junk;
+#endif
+void	huge_dalloc(void *ptr, bool unmap);
+size_t	huge_salloc(const void *ptr);
+dss_prec_t	huge_dss_prec_get(arena_t *arena);
+prof_ctx_t	*huge_prof_ctx_get(const void *ptr);
+void	huge_prof_ctx_set(const void *ptr, prof_ctx_t *ctx);
+bool	huge_boot(void);
+void	huge_prefork(void);
+void	huge_postfork_parent(void);
+void	huge_postfork_child(void);
+
+#endif /* JEMALLOC_H_EXTERNS */
+/******************************************************************************/
+#ifdef JEMALLOC_H_INLINES
+
+#endif /* JEMALLOC_H_INLINES */
+/******************************************************************************/

deps/jemalloc/include/jemalloc/internal/jemalloc_internal_decls.h

@@ -0,0 +1,94 @@
+#ifndef JEMALLOC_INTERNAL_DECLS_H
+#define JEMALLOC_INTERNAL_DECLS_H
+
+#include <math.h>
+#ifdef _WIN32
+#  include <windows.h>
+#  include "msvc_compat/windows_extra.h"
+#  ifdef _WIN64
+#    if LG_VADDR <= 32
+#      error Generate the headers using x64 vcargs
+#    endif
+#  else
+#    if LG_VADDR > 32
+#      undef LG_VADDR
+#      define LG_VADDR 32
+#    endif
+#  endif
+#else
+#  include <sys/param.h>
+#  include <sys/mman.h>
+#  if !defined(__pnacl__) && !defined(__native_client__)
+#    include <sys/syscall.h>
+#    if !defined(SYS_write) && defined(__NR_write)
+#      define SYS_write __NR_write
+#    endif
+#    if defined(SYS_open) && defined(__aarch64__)
+       /* Android headers may define SYS_open to __NR_open even though
+        * __NR_open may not exist on AArch64 (superseded by __NR_openat). */
+#      undef SYS_open
+#    endif
+#    include <sys/uio.h>
+#  endif
+#  include <pthread.h>
+#  ifdef __FreeBSD__
+#  include <pthread_np.h>
+#  endif
+#  include <signal.h>
+#  ifdef JEMALLOC_OS_UNFAIR_LOCK
+#    include <os/lock.h>
+#  endif
+#  ifdef JEMALLOC_GLIBC_MALLOC_HOOK
+#    include <sched.h>
+#  endif
+#  include <errno.h>
+#  include <sys/time.h>
+#  include <time.h>
+#  ifdef JEMALLOC_HAVE_MACH_ABSOLUTE_TIME
+#    include <mach/mach_time.h>
+#  endif
+#endif
+#include <sys/types.h>
+
+#include <limits.h>
+#ifndef SIZE_T_MAX
+#  define SIZE_T_MAX	SIZE_MAX
+#endif
+#ifndef SSIZE_MAX
+#  define SSIZE_MAX	((ssize_t)(SIZE_T_MAX >> 1))
+#endif
+#include <stdarg.h>
+#include <stdbool.h>
+#include <stdio.h>
+#include <stdlib.h>
+#include <stdint.h>
+#include <stddef.h>
+#ifndef offsetof
+#  define offsetof(type, member)	((size_t)&(((type *)NULL)->member))
+#endif
+#include <string.h>
+#include <strings.h>
+#include <ctype.h>
+#ifdef _MSC_VER
+#  include <io.h>
+typedef intptr_t ssize_t;
+#  define PATH_MAX 1024
+#  define STDERR_FILENO 2
+#  define __func__ __FUNCTION__
+#  ifdef JEMALLOC_HAS_RESTRICT
+#    define restrict __restrict
+#  endif
+/* Disable warnings about deprecated system functions. */
+#  pragma warning(disable: 4996)
+#if _MSC_VER < 1800
+static int
+isblank(int c) {
+	return (c == '\t' || c == ' ');
+}
+#endif
+#else
+#  include <unistd.h>
+#endif
+#include <fcntl.h>
+
+#endif /* JEMALLOC_INTERNAL_H */

deps/jemalloc/include/jemalloc/internal/jemalloc_internal_externs.h

@@ -0,0 +1,57 @@
+#ifndef JEMALLOC_INTERNAL_EXTERNS_H
+#define JEMALLOC_INTERNAL_EXTERNS_H
+
+#include "jemalloc/internal/atomic.h"
+#include "jemalloc/internal/tsd_types.h"
+
+/* TSD checks this to set thread local slow state accordingly. */
+extern bool malloc_slow;
+
+/* Run-time options. */
+extern bool opt_abort;
+extern bool opt_abort_conf;
+extern bool opt_confirm_conf;
+extern const char *opt_junk;
+extern bool opt_junk_alloc;
+extern bool opt_junk_free;
+extern bool opt_utrace;
+extern bool opt_xmalloc;
+extern bool opt_zero;
+extern unsigned opt_narenas;
+
+/* Number of CPUs. */
+extern unsigned ncpus;
+
+/* Number of arenas used for automatic multiplexing of threads and arenas. */
+extern unsigned narenas_auto;
+
+/* Base index for manual arenas. */
+extern unsigned manual_arena_base;
+
+/*
+ * Arenas that are used to service external requests.  Not all elements of the
+ * arenas array are necessarily used; arenas are created lazily as needed.
+ */
+extern atomic_p_t arenas[];
+
+void *a0malloc(size_t size);
+void a0dalloc(void *ptr);
+void *bootstrap_malloc(size_t size);
+void *bootstrap_calloc(size_t num, size_t size);
+void bootstrap_free(void *ptr);
+void arena_set(unsigned ind, arena_t *arena);
+unsigned narenas_total_get(void);
+arena_t *arena_init(tsdn_t *tsdn, unsigned ind, extent_hooks_t *extent_hooks);
+arena_tdata_t *arena_tdata_get_hard(tsd_t *tsd, unsigned ind);
+arena_t *arena_choose_hard(tsd_t *tsd, bool internal);
+void arena_migrate(tsd_t *tsd, unsigned oldind, unsigned newind);
+void iarena_cleanup(tsd_t *tsd);
+void arena_cleanup(tsd_t *tsd);
+void arenas_tdata_cleanup(tsd_t *tsd);
+void jemalloc_prefork(void);
+void jemalloc_postfork_parent(void);
+void jemalloc_postfork_child(void);
+bool malloc_initialized(void);
+void je_sdallocx_noflags(void *ptr, size_t size);
+
+#endif /* JEMALLOC_INTERNAL_EXTERNS_H */

deps/jemalloc/include/jemalloc/internal/jemalloc_internal_includes.h

@@ -0,0 +1,94 @@
+#ifndef JEMALLOC_INTERNAL_INCLUDES_H
+#define JEMALLOC_INTERNAL_INCLUDES_H
+
+/*
+ * jemalloc can conceptually be broken into components (arena, tcache, etc.),
+ * but there are circular dependencies that cannot be broken without
+ * substantial performance degradation.
+ *
+ * Historically, we dealt with this by each header into four sections (types,
+ * structs, externs, and inlines), and included each header file multiple times
+ * in this file, picking out the portion we want on each pass using the
+ * following #defines:
+ *   JEMALLOC_H_TYPES   : Preprocessor-defined constants and psuedo-opaque data
+ *                        types.
+ *   JEMALLOC_H_STRUCTS : Data structures.
+ *   JEMALLOC_H_EXTERNS : Extern data declarations and function prototypes.
+ *   JEMALLOC_H_INLINES : Inline functions.
+ *
+ * We're moving toward a world in which the dependencies are explicit; each file
+ * will #include the headers it depends on (rather than relying on them being
+ * implicitly available via this file including every header file in the
+ * project).
+ *
+ * We're now in an intermediate state: we've broken up the header files to avoid
+ * having to include each one multiple times, but have not yet moved the
+ * dependency information into the header files (i.e. we still rely on the
+ * ordering in this file to ensure all a header's dependencies are available in
+ * its translation unit).  Each component is now broken up into multiple header
+ * files, corresponding to the sections above (e.g. instead of "foo.h", we now
+ * have "foo_types.h", "foo_structs.h", "foo_externs.h", "foo_inlines.h").
+ *
+ * Those files which have been converted to explicitly include their
+ * inter-component dependencies are now in the initial HERMETIC HEADERS
+ * section.  All headers may still rely on jemalloc_preamble.h (which, by fiat,
+ * must be included first in every translation unit) for system headers and
+ * global jemalloc definitions, however.
+ */
+
+/******************************************************************************/
+/* TYPES */
+/******************************************************************************/
+
+#include "jemalloc/internal/extent_types.h"
+#include "jemalloc/internal/base_types.h"
+#include "jemalloc/internal/arena_types.h"
+#include "jemalloc/internal/tcache_types.h"
+#include "jemalloc/internal/prof_types.h"
+
+/******************************************************************************/
+/* STRUCTS */
+/******************************************************************************/
+
+#include "jemalloc/internal/arena_structs_a.h"
+#include "jemalloc/internal/extent_structs.h"
+#include "jemalloc/internal/base_structs.h"
+#include "jemalloc/internal/prof_structs.h"
+#include "jemalloc/internal/arena_structs_b.h"
+#include "jemalloc/internal/tcache_structs.h"
+#include "jemalloc/internal/background_thread_structs.h"
+
+/******************************************************************************/
+/* EXTERNS */
+/******************************************************************************/
+
+#include "jemalloc/internal/jemalloc_internal_externs.h"
+#include "jemalloc/internal/extent_externs.h"
+#include "jemalloc/internal/base_externs.h"
+#include "jemalloc/internal/arena_externs.h"
+#include "jemalloc/internal/large_externs.h"
+#include "jemalloc/internal/tcache_externs.h"
+#include "jemalloc/internal/prof_externs.h"
+#include "jemalloc/internal/background_thread_externs.h"
+
+/******************************************************************************/
+/* INLINES */
+/******************************************************************************/
+
+#include "jemalloc/internal/jemalloc_internal_inlines_a.h"
+#include "jemalloc/internal/base_inlines.h"
+/*
+ * Include portions of arena code interleaved with tcache code in order to
+ * resolve circular dependencies.
+ */
+#include "jemalloc/internal/prof_inlines_a.h"
+#include "jemalloc/internal/arena_inlines_a.h"
+#include "jemalloc/internal/extent_inlines.h"
+#include "jemalloc/internal/jemalloc_internal_inlines_b.h"
+#include "jemalloc/internal/tcache_inlines.h"
+#include "jemalloc/internal/arena_inlines_b.h"
+#include "jemalloc/internal/jemalloc_internal_inlines_c.h"
+#include "jemalloc/internal/prof_inlines_b.h"
+#include "jemalloc/internal/background_thread_inlines.h"
+
+#endif /* JEMALLOC_INTERNAL_INCLUDES_H */

deps/jemalloc/include/jemalloc/internal/jemalloc_internal_inlines_a.h

@@ -0,0 +1,174 @@
+#ifndef JEMALLOC_INTERNAL_INLINES_A_H
+#define JEMALLOC_INTERNAL_INLINES_A_H
+
+#include "jemalloc/internal/atomic.h"
+#include "jemalloc/internal/bit_util.h"
+#include "jemalloc/internal/jemalloc_internal_types.h"
+#include "jemalloc/internal/sc.h"
+#include "jemalloc/internal/ticker.h"
+
+JEMALLOC_ALWAYS_INLINE malloc_cpuid_t
+malloc_getcpu(void) {
+	assert(have_percpu_arena);
+#if defined(_WIN32)
+	return GetCurrentProcessorNumber();
+#elif defined(JEMALLOC_HAVE_SCHED_GETCPU)
+	return (malloc_cpuid_t)sched_getcpu();
+#else
+	not_reached();
+	return -1;
+#endif
+}
+
+/* Return the chosen arena index based on current cpu. */
+JEMALLOC_ALWAYS_INLINE unsigned
+percpu_arena_choose(void) {
+	assert(have_percpu_arena && PERCPU_ARENA_ENABLED(opt_percpu_arena));
+
+	malloc_cpuid_t cpuid = malloc_getcpu();
+	assert(cpuid >= 0);
+
+	unsigned arena_ind;
+	if ((opt_percpu_arena == percpu_arena) || ((unsigned)cpuid < ncpus /
+	    2)) {
+		arena_ind = cpuid;
+	} else {
+		assert(opt_percpu_arena == per_phycpu_arena);
+		/* Hyper threads on the same physical CPU share arena. */
+		arena_ind = cpuid - ncpus / 2;
+	}
+
+	return arena_ind;
+}
+
+/* Return the limit of percpu auto arena range, i.e. arenas[0...ind_limit). */
+JEMALLOC_ALWAYS_INLINE unsigned
+percpu_arena_ind_limit(percpu_arena_mode_t mode) {
+	assert(have_percpu_arena && PERCPU_ARENA_ENABLED(mode));
+	if (mode == per_phycpu_arena && ncpus > 1) {
+		if (ncpus % 2) {
+			/* This likely means a misconfig. */
+			return ncpus / 2 + 1;
+		}
+		return ncpus / 2;
+	} else {
+		return ncpus;
+	}
+}
+
+static inline arena_tdata_t *
+arena_tdata_get(tsd_t *tsd, unsigned ind, bool refresh_if_missing) {
+	arena_tdata_t *tdata;
+	arena_tdata_t *arenas_tdata = tsd_arenas_tdata_get(tsd);
+
+	if (unlikely(arenas_tdata == NULL)) {
+		/* arenas_tdata hasn't been initialized yet. */
+		return arena_tdata_get_hard(tsd, ind);
+	}
+	if (unlikely(ind >= tsd_narenas_tdata_get(tsd))) {
+		/*
+		 * ind is invalid, cache is old (too small), or tdata to be
+		 * initialized.
+		 */
+		return (refresh_if_missing ? arena_tdata_get_hard(tsd, ind) :
+		    NULL);
+	}
+
+	tdata = &arenas_tdata[ind];
+	if (likely(tdata != NULL) || !refresh_if_missing) {
+		return tdata;
+	}
+	return arena_tdata_get_hard(tsd, ind);
+}
+
+static inline arena_t *
+arena_get(tsdn_t *tsdn, unsigned ind, bool init_if_missing) {
+	arena_t *ret;
+
+	assert(ind < MALLOCX_ARENA_LIMIT);
+
+	ret = (arena_t *)atomic_load_p(&arenas[ind], ATOMIC_ACQUIRE);
+	if (unlikely(ret == NULL)) {
+		if (init_if_missing) {
+			ret = arena_init(tsdn, ind,
+			    (extent_hooks_t *)&extent_hooks_default);
+		}
+	}
+	return ret;
+}
+
+static inline ticker_t *
+decay_ticker_get(tsd_t *tsd, unsigned ind) {
+	arena_tdata_t *tdata;
+
+	tdata = arena_tdata_get(tsd, ind, true);
+	if (unlikely(tdata == NULL)) {
+		return NULL;
+	}
+	return &tdata->decay_ticker;
+}
+
+JEMALLOC_ALWAYS_INLINE cache_bin_t *
+tcache_small_bin_get(tcache_t *tcache, szind_t binind) {
+	assert(binind < SC_NBINS);
+	return &tcache->bins_small[binind];
+}
+
+JEMALLOC_ALWAYS_INLINE cache_bin_t *
+tcache_large_bin_get(tcache_t *tcache, szind_t binind) {
+	assert(binind >= SC_NBINS &&binind < nhbins);
+	return &tcache->bins_large[binind - SC_NBINS];
+}
+
+JEMALLOC_ALWAYS_INLINE bool
+tcache_available(tsd_t *tsd) {
+	/*
+	 * Thread specific auto tcache might be unavailable if: 1) during tcache
+	 * initialization, or 2) disabled through thread.tcache.enabled mallctl
+	 * or config options.  This check covers all cases.
+	 */
+	if (likely(tsd_tcache_enabled_get(tsd))) {
+		/* Associated arena == NULL implies tcache init in progress. */
+		assert(tsd_tcachep_get(tsd)->arena == NULL ||
+		    tcache_small_bin_get(tsd_tcachep_get(tsd), 0)->avail !=
+		    NULL);
+		return true;
+	}
+
+	return false;
+}
+
+JEMALLOC_ALWAYS_INLINE tcache_t *
+tcache_get(tsd_t *tsd) {
+	if (!tcache_available(tsd)) {
+		return NULL;
+	}
+
+	return tsd_tcachep_get(tsd);
+}
+
+static inline void
+pre_reentrancy(tsd_t *tsd, arena_t *arena) {
+	/* arena is the current context.  Reentry from a0 is not allowed. */
+	assert(arena != arena_get(tsd_tsdn(tsd), 0, false));
+
+	bool fast = tsd_fast(tsd);
+	assert(tsd_reentrancy_level_get(tsd) < INT8_MAX);
+	++*tsd_reentrancy_levelp_get(tsd);
+	if (fast) {
+		/* Prepare slow path for reentrancy. */
+		tsd_slow_update(tsd);
+		assert(tsd_state_get(tsd) == tsd_state_nominal_slow);
+	}
+}
+
+static inline void
+post_reentrancy(tsd_t *tsd) {
+	int8_t *reentrancy_level = tsd_reentrancy_levelp_get(tsd);
+	assert(*reentrancy_level > 0);
+	if (--*reentrancy_level == 0) {
+		tsd_slow_update(tsd);
+	}
+}
+
+#endif /* JEMALLOC_INTERNAL_INLINES_A_H */

deps/jemalloc/include/jemalloc/internal/jemalloc_internal_inlines_b.h

@@ -0,0 +1,87 @@
+#ifndef JEMALLOC_INTERNAL_INLINES_B_H
+#define JEMALLOC_INTERNAL_INLINES_B_H
+
+#include "jemalloc/internal/rtree.h"
+
+/* Choose an arena based on a per-thread value. */
+static inline arena_t *
+arena_choose_impl(tsd_t *tsd, arena_t *arena, bool internal) {
+	arena_t *ret;
+
+	if (arena != NULL) {
+		return arena;
+	}
+
+	/* During reentrancy, arena 0 is the safest bet. */
+	if (unlikely(tsd_reentrancy_level_get(tsd) > 0)) {
+		return arena_get(tsd_tsdn(tsd), 0, true);
+	}
+
+	ret = internal ? tsd_iarena_get(tsd) : tsd_arena_get(tsd);
+	if (unlikely(ret == NULL)) {
+		ret = arena_choose_hard(tsd, internal);
+		assert(ret);
+		if (tcache_available(tsd)) {
+			tcache_t *tcache = tcache_get(tsd);
+			if (tcache->arena != NULL) {
+				/* See comments in tcache_data_init().*/
+				assert(tcache->arena ==
+				    arena_get(tsd_tsdn(tsd), 0, false));
+				if (tcache->arena != ret) {
+					tcache_arena_reassociate(tsd_tsdn(tsd),
+					    tcache, ret);
+				}
+			} else {
+				tcache_arena_associate(tsd_tsdn(tsd), tcache,
+				    ret);
+			}
+		}
+	}
+
+	/*
+	 * Note that for percpu arena, if the current arena is outside of the
+	 * auto percpu arena range, (i.e. thread is assigned to a manually
+	 * managed arena), then percpu arena is skipped.
+	 */
+	if (have_percpu_arena && PERCPU_ARENA_ENABLED(opt_percpu_arena) &&
+	    !internal && (arena_ind_get(ret) <
+	    percpu_arena_ind_limit(opt_percpu_arena)) && (ret->last_thd !=
+	    tsd_tsdn(tsd))) {
+		unsigned ind = percpu_arena_choose();
+		if (arena_ind_get(ret) != ind) {
+			percpu_arena_update(tsd, ind);
+			ret = tsd_arena_get(tsd);
+		}
+		ret->last_thd = tsd_tsdn(tsd);
+	}
+
+	return ret;
+}
+
+static inline arena_t *
+arena_choose(tsd_t *tsd, arena_t *arena) {
+	return arena_choose_impl(tsd, arena, false);
+}
+
+static inline arena_t *
+arena_ichoose(tsd_t *tsd, arena_t *arena) {
+	return arena_choose_impl(tsd, arena, true);
+}
+
+static inline bool
+arena_is_auto(arena_t *arena) {
+	assert(narenas_auto > 0);
+
+	return (arena_ind_get(arena) < manual_arena_base);
+}
+
+JEMALLOC_ALWAYS_INLINE extent_t *
+iealloc(tsdn_t *tsdn, const void *ptr) {
+	rtree_ctx_t rtree_ctx_fallback;
+	rtree_ctx_t *rtree_ctx = tsdn_rtree_ctx(tsdn, &rtree_ctx_fallback);
+
+	return rtree_extent_read(tsdn, &extents_rtree, rtree_ctx,
+	    (uintptr_t)ptr, true);
+}
+
+#endif /* JEMALLOC_INTERNAL_INLINES_B_H */

deps/jemalloc/include/jemalloc/internal/jemalloc_internal_inlines_c.h

@@ -0,0 +1,222 @@
+#ifndef JEMALLOC_INTERNAL_INLINES_C_H
+#define JEMALLOC_INTERNAL_INLINES_C_H
+
+#include "jemalloc/internal/hook.h"
+#include "jemalloc/internal/jemalloc_internal_types.h"
+#include "jemalloc/internal/sz.h"
+#include "jemalloc/internal/witness.h"
+
+/*
+ * Translating the names of the 'i' functions:
+ *   Abbreviations used in the first part of the function name (before
+ *   alloc/dalloc) describe what that function accomplishes:
+ *     a: arena (query)
+ *     s: size (query, or sized deallocation)
+ *     e: extent (query)
+ *     p: aligned (allocates)
+ *     vs: size (query, without knowing that the pointer is into the heap)
+ *     r: rallocx implementation
+ *     x: xallocx implementation
+ *   Abbreviations used in the second part of the function name (after
+ *   alloc/dalloc) describe the arguments it takes
+ *     z: whether to return zeroed memory
+ *     t: accepts a tcache_t * parameter
+ *     m: accepts an arena_t * parameter
+ */
+
+JEMALLOC_ALWAYS_INLINE arena_t *
+iaalloc(tsdn_t *tsdn, const void *ptr) {
+	assert(ptr != NULL);
+
+	return arena_aalloc(tsdn, ptr);
+}
+
+JEMALLOC_ALWAYS_INLINE size_t
+isalloc(tsdn_t *tsdn, const void *ptr) {
+	assert(ptr != NULL);
+
+	return arena_salloc(tsdn, ptr);
+}
+
+JEMALLOC_ALWAYS_INLINE void *
+iallocztm(tsdn_t *tsdn, size_t size, szind_t ind, bool zero, tcache_t *tcache,
+    bool is_internal, arena_t *arena, bool slow_path) {
+	void *ret;
+
+	assert(!is_internal || tcache == NULL);
+	assert(!is_internal || arena == NULL || arena_is_auto(arena));
+	if (!tsdn_null(tsdn) && tsd_reentrancy_level_get(tsdn_tsd(tsdn)) == 0) {
+		witness_assert_depth_to_rank(tsdn_witness_tsdp_get(tsdn),
+		    WITNESS_RANK_CORE, 0);
+	}
+
+	ret = arena_malloc(tsdn, arena, size, ind, zero, tcache, slow_path);
+	if (config_stats && is_internal && likely(ret != NULL)) {
+		arena_internal_add(iaalloc(tsdn, ret), isalloc(tsdn, ret));
+	}
+	return ret;
+}
+
+JEMALLOC_ALWAYS_INLINE void *
+ialloc(tsd_t *tsd, size_t size, szind_t ind, bool zero, bool slow_path) {
+	return iallocztm(tsd_tsdn(tsd), size, ind, zero, tcache_get(tsd), false,
+	    NULL, slow_path);
+}
+
+JEMALLOC_ALWAYS_INLINE void *
+ipallocztm(tsdn_t *tsdn, size_t usize, size_t alignment, bool zero,
+    tcache_t *tcache, bool is_internal, arena_t *arena) {
+	void *ret;
+
+	assert(usize != 0);
+	assert(usize == sz_sa2u(usize, alignment));
+	assert(!is_internal || tcache == NULL);
+	assert(!is_internal || arena == NULL || arena_is_auto(arena));
+	witness_assert_depth_to_rank(tsdn_witness_tsdp_get(tsdn),
+	    WITNESS_RANK_CORE, 0);
+
+	ret = arena_palloc(tsdn, arena, usize, alignment, zero, tcache);
+	assert(ALIGNMENT_ADDR2BASE(ret, alignment) == ret);
+	if (config_stats && is_internal && likely(ret != NULL)) {
+		arena_internal_add(iaalloc(tsdn, ret), isalloc(tsdn, ret));
+	}
+	return ret;
+}
+
+JEMALLOC_ALWAYS_INLINE void *
+ipalloct(tsdn_t *tsdn, size_t usize, size_t alignment, bool zero,
+    tcache_t *tcache, arena_t *arena) {
+	return ipallocztm(tsdn, usize, alignment, zero, tcache, false, arena);
+}
+
+JEMALLOC_ALWAYS_INLINE void *
+ipalloc(tsd_t *tsd, size_t usize, size_t alignment, bool zero) {
+	return ipallocztm(tsd_tsdn(tsd), usize, alignment, zero,
+	    tcache_get(tsd), false, NULL);
+}
+
+JEMALLOC_ALWAYS_INLINE size_t
+ivsalloc(tsdn_t *tsdn, const void *ptr) {
+	return arena_vsalloc(tsdn, ptr);
+}
+
+JEMALLOC_ALWAYS_INLINE void
+idalloctm(tsdn_t *tsdn, void *ptr, tcache_t *tcache, alloc_ctx_t *alloc_ctx,
+    bool is_internal, bool slow_path) {
+	assert(ptr != NULL);
+	assert(!is_internal || tcache == NULL);
+	assert(!is_internal || arena_is_auto(iaalloc(tsdn, ptr)));
+	witness_assert_depth_to_rank(tsdn_witness_tsdp_get(tsdn),
+	    WITNESS_RANK_CORE, 0);
+	if (config_stats && is_internal) {
+		arena_internal_sub(iaalloc(tsdn, ptr), isalloc(tsdn, ptr));
+	}
+	if (!is_internal && !tsdn_null(tsdn) &&
+	    tsd_reentrancy_level_get(tsdn_tsd(tsdn)) != 0) {
+		assert(tcache == NULL);
+	}
+	arena_dalloc(tsdn, ptr, tcache, alloc_ctx, slow_path);
+}
+
+JEMALLOC_ALWAYS_INLINE void
+idalloc(tsd_t *tsd, void *ptr) {
+	idalloctm(tsd_tsdn(tsd), ptr, tcache_get(tsd), NULL, false, true);
+}
+
+JEMALLOC_ALWAYS_INLINE void
+isdalloct(tsdn_t *tsdn, void *ptr, size_t size, tcache_t *tcache,
+    alloc_ctx_t *alloc_ctx, bool slow_path) {
+	witness_assert_depth_to_rank(tsdn_witness_tsdp_get(tsdn),
+	    WITNESS_RANK_CORE, 0);
+	arena_sdalloc(tsdn, ptr, size, tcache, alloc_ctx, slow_path);
+}
+
+JEMALLOC_ALWAYS_INLINE void *
+iralloct_realign(tsdn_t *tsdn, void *ptr, size_t oldsize, size_t size,
+    size_t alignment, bool zero, tcache_t *tcache, arena_t *arena,
+    hook_ralloc_args_t *hook_args) {
+	witness_assert_depth_to_rank(tsdn_witness_tsdp_get(tsdn),
+	    WITNESS_RANK_CORE, 0);
+	void *p;
+	size_t usize, copysize;
+
+	usize = sz_sa2u(size, alignment);
+	if (unlikely(usize == 0 || usize > SC_LARGE_MAXCLASS)) {
+		return NULL;
+	}
+	p = ipalloct(tsdn, usize, alignment, zero, tcache, arena);
+	if (p == NULL) {
+		return NULL;
+	}
+	/*
+	 * Copy at most size bytes (not size+extra), since the caller has no
+	 * expectation that the extra bytes will be reliably preserved.
+	 */
+	copysize = (size < oldsize) ? size : oldsize;
+	memcpy(p, ptr, copysize);
+	hook_invoke_alloc(hook_args->is_realloc
+	    ? hook_alloc_realloc : hook_alloc_rallocx, p, (uintptr_t)p,
+	    hook_args->args);
+	hook_invoke_dalloc(hook_args->is_realloc
+	    ? hook_dalloc_realloc : hook_dalloc_rallocx, ptr, hook_args->args);
+	isdalloct(tsdn, ptr, oldsize, tcache, NULL, true);
+	return p;
+}
+
+/*
+ * is_realloc threads through the knowledge of whether or not this call comes
+ * from je_realloc (as opposed to je_rallocx); this ensures that we pass the
+ * correct entry point into any hooks.
+ * Note that these functions are all force-inlined, so no actual bool gets
+ * passed-around anywhere.
+ */
+JEMALLOC_ALWAYS_INLINE void *
+iralloct(tsdn_t *tsdn, void *ptr, size_t oldsize, size_t size, size_t alignment,
+    bool zero, tcache_t *tcache, arena_t *arena, hook_ralloc_args_t *hook_args)
+{
+	assert(ptr != NULL);
+	assert(size != 0);
+	witness_assert_depth_to_rank(tsdn_witness_tsdp_get(tsdn),
+	    WITNESS_RANK_CORE, 0);
+
+	if (alignment != 0 && ((uintptr_t)ptr & ((uintptr_t)alignment-1))
+	    != 0) {
+		/*
+		 * Existing object alignment is inadequate; allocate new space
+		 * and copy.
+		 */
+		return iralloct_realign(tsdn, ptr, oldsize, size, alignment,
+		    zero, tcache, arena, hook_args);
+	}
+
+	return arena_ralloc(tsdn, arena, ptr, oldsize, size, alignment, zero,
+	    tcache, hook_args);
+}
+
+JEMALLOC_ALWAYS_INLINE void *
+iralloc(tsd_t *tsd, void *ptr, size_t oldsize, size_t size, size_t alignment,
+    bool zero, hook_ralloc_args_t *hook_args) {
+	return iralloct(tsd_tsdn(tsd), ptr, oldsize, size, alignment, zero,
+	    tcache_get(tsd), NULL, hook_args);
+}
+
+JEMALLOC_ALWAYS_INLINE bool
+ixalloc(tsdn_t *tsdn, void *ptr, size_t oldsize, size_t size, size_t extra,
+    size_t alignment, bool zero, size_t *newsize) {
+	assert(ptr != NULL);
+	assert(size != 0);
+	witness_assert_depth_to_rank(tsdn_witness_tsdp_get(tsdn),
+	    WITNESS_RANK_CORE, 0);
+
+	if (alignment != 0 && ((uintptr_t)ptr & ((uintptr_t)alignment-1))
+	    != 0) {
+		/* Existing object alignment is inadequate. */
+		*newsize = oldsize;
+		return true;
+	}
+
+	return arena_ralloc_no_move(tsdn, ptr, oldsize, size, extra, zero,
+	    newsize);
+}
+
+#endif /* JEMALLOC_INTERNAL_INLINES_C_H */

deps/jemalloc/include/jemalloc/internal/jemalloc_internal_macros.h

@@ -0,0 +1,114 @@
+#ifndef JEMALLOC_INTERNAL_MACROS_H
+#define JEMALLOC_INTERNAL_MACROS_H
+
+#ifdef JEMALLOC_DEBUG
+#  define JEMALLOC_ALWAYS_INLINE static inline
+#else
+#  define JEMALLOC_ALWAYS_INLINE JEMALLOC_ATTR(always_inline) static inline
+#endif
+#ifdef _MSC_VER
+#  define inline _inline
+#endif
+
+#define UNUSED JEMALLOC_ATTR(unused)
+
+#define ZU(z)	((size_t)z)
+#define ZD(z)	((ssize_t)z)
+#define QU(q)	((uint64_t)q)
+#define QD(q)	((int64_t)q)
+
+#define KZU(z)	ZU(z##ULL)
+#define KZD(z)	ZD(z##LL)
+#define KQU(q)	QU(q##ULL)
+#define KQD(q)	QI(q##LL)
+
+#ifndef __DECONST
+#  define	__DECONST(type, var)	((type)(uintptr_t)(const void *)(var))
+#endif
+
+#if !defined(JEMALLOC_HAS_RESTRICT) || defined(__cplusplus)
+#  define restrict
+#endif
+
+/* Various function pointers are static and immutable except during testing. */
+#ifdef JEMALLOC_JET
+#  define JET_MUTABLE
+#else
+#  define JET_MUTABLE const
+#endif
+
+#define JEMALLOC_VA_ARGS_HEAD(head, ...) head
+#define JEMALLOC_VA_ARGS_TAIL(head, ...) __VA_ARGS__
+
+#if (defined(__GNUC__) || defined(__GNUG__)) && !defined(__clang__) \
+  && defined(JEMALLOC_HAVE_ATTR) && (__GNUC__ >= 7)
+#define JEMALLOC_FALLTHROUGH JEMALLOC_ATTR(fallthrough);
+#else
+#define JEMALLOC_FALLTHROUGH /* falls through */
+#endif
+
+/* Diagnostic suppression macros */
+#if defined(_MSC_VER) && !defined(__clang__)
+#  define JEMALLOC_DIAGNOSTIC_PUSH __pragma(warning(push))
+#  define JEMALLOC_DIAGNOSTIC_POP __pragma(warning(pop))
+#  define JEMALLOC_DIAGNOSTIC_IGNORE(W) __pragma(warning(disable:W))
+#  define JEMALLOC_DIAGNOSTIC_IGNORE_MISSING_STRUCT_FIELD_INITIALIZERS
+#  define JEMALLOC_DIAGNOSTIC_IGNORE_TYPE_LIMITS
+#  define JEMALLOC_DIAGNOSTIC_IGNORE_ALLOC_SIZE_LARGER_THAN
+#  define JEMALLOC_DIAGNOSTIC_DISABLE_SPURIOUS
+/* #pragma GCC diagnostic first appeared in gcc 4.6. */
+#elif (defined(__GNUC__) && ((__GNUC__ > 4) || ((__GNUC__ == 4) && \
+  (__GNUC_MINOR__ > 5)))) || defined(__clang__)
+/*
+ * The JEMALLOC_PRAGMA__ macro is an implementation detail of the GCC and Clang
+ * diagnostic suppression macros and should not be used anywhere else.
+ */
+#  define JEMALLOC_PRAGMA__(X) _Pragma(#X)
+#  define JEMALLOC_DIAGNOSTIC_PUSH JEMALLOC_PRAGMA__(GCC diagnostic push)
+#  define JEMALLOC_DIAGNOSTIC_POP JEMALLOC_PRAGMA__(GCC diagnostic pop)
+#  define JEMALLOC_DIAGNOSTIC_IGNORE(W) \
+     JEMALLOC_PRAGMA__(GCC diagnostic ignored W)
+
+/*
+ * The -Wmissing-field-initializers warning is buggy in GCC versions < 5.1 and
+ * all clang versions up to version 7 (currently trunk, unreleased).  This macro
+ * suppresses the warning for the affected compiler versions only.
+ */
+#  if ((defined(__GNUC__) && !defined(__clang__)) && (__GNUC__ < 5)) || \
+     defined(__clang__)
+#    define JEMALLOC_DIAGNOSTIC_IGNORE_MISSING_STRUCT_FIELD_INITIALIZERS  \
+          JEMALLOC_DIAGNOSTIC_IGNORE("-Wmissing-field-initializers")
+#  else
+#    define JEMALLOC_DIAGNOSTIC_IGNORE_MISSING_STRUCT_FIELD_INITIALIZERS
+#  endif
+
+#  define JEMALLOC_DIAGNOSTIC_IGNORE_TYPE_LIMITS  \
+     JEMALLOC_DIAGNOSTIC_IGNORE("-Wtype-limits")
+#  define JEMALLOC_DIAGNOSTIC_IGNORE_UNUSED_PARAMETER \
+     JEMALLOC_DIAGNOSTIC_IGNORE("-Wunused-parameter")
+#  if defined(__GNUC__) && !defined(__clang__) && (__GNUC__ >= 7)
+#    define JEMALLOC_DIAGNOSTIC_IGNORE_ALLOC_SIZE_LARGER_THAN \
+       JEMALLOC_DIAGNOSTIC_IGNORE("-Walloc-size-larger-than=")
+#  else
+#    define JEMALLOC_DIAGNOSTIC_IGNORE_ALLOC_SIZE_LARGER_THAN
+#  endif
+#  define JEMALLOC_DIAGNOSTIC_DISABLE_SPURIOUS \
+  JEMALLOC_DIAGNOSTIC_PUSH \
+  JEMALLOC_DIAGNOSTIC_IGNORE_UNUSED_PARAMETER
+#else
+#  define JEMALLOC_DIAGNOSTIC_PUSH
+#  define JEMALLOC_DIAGNOSTIC_POP
+#  define JEMALLOC_DIAGNOSTIC_IGNORE(W)
+#  define JEMALLOC_DIAGNOSTIC_IGNORE_MISSING_STRUCT_FIELD_INITIALIZERS
+#  define JEMALLOC_DIAGNOSTIC_IGNORE_TYPE_LIMITS
+#  define JEMALLOC_DIAGNOSTIC_IGNORE_ALLOC_SIZE_LARGER_THAN
+#  define JEMALLOC_DIAGNOSTIC_DISABLE_SPURIOUS
+#endif
+
+/*
+ * Disables spurious diagnostics for all headers.  Since these headers are not
+ * included by users directly, it does not affect their diagnostic settings.
+ */
+JEMALLOC_DIAGNOSTIC_DISABLE_SPURIOUS
+
+#endif /* JEMALLOC_INTERNAL_MACROS_H */

deps/jemalloc/include/jemalloc/internal/jemalloc_internal_types.h

@@ -0,0 +1,114 @@
+#ifndef JEMALLOC_INTERNAL_TYPES_H
+#define JEMALLOC_INTERNAL_TYPES_H
+
+#include "jemalloc/internal/quantum.h"
+
+/* Page size index type. */
+typedef unsigned pszind_t;
+
+/* Size class index type. */
+typedef unsigned szind_t;
+
+/* Processor / core id type. */
+typedef int malloc_cpuid_t;
+
+/*
+ * Flags bits:
+ *
+ * a: arena
+ * t: tcache
+ * 0: unused
+ * z: zero
+ * n: alignment
+ *
+ * aaaaaaaa aaaatttt tttttttt 0znnnnnn
+ */
+#define MALLOCX_ARENA_BITS	12
+#define MALLOCX_TCACHE_BITS	12
+#define MALLOCX_LG_ALIGN_BITS	6
+#define MALLOCX_ARENA_SHIFT	20
+#define MALLOCX_TCACHE_SHIFT	8
+#define MALLOCX_ARENA_MASK \
+    (((1 << MALLOCX_ARENA_BITS) - 1) << MALLOCX_ARENA_SHIFT)
+/* NB: Arena index bias decreases the maximum number of arenas by 1. */
+#define MALLOCX_ARENA_LIMIT	((1 << MALLOCX_ARENA_BITS) - 1)
+#define MALLOCX_TCACHE_MASK \
+    (((1 << MALLOCX_TCACHE_BITS) - 1) << MALLOCX_TCACHE_SHIFT)
+#define MALLOCX_TCACHE_MAX	((1 << MALLOCX_TCACHE_BITS) - 3)
+#define MALLOCX_LG_ALIGN_MASK	((1 << MALLOCX_LG_ALIGN_BITS) - 1)
+/* Use MALLOCX_ALIGN_GET() if alignment may not be specified in flags. */
+#define MALLOCX_ALIGN_GET_SPECIFIED(flags)				\
+    (ZU(1) << (flags & MALLOCX_LG_ALIGN_MASK))
+#define MALLOCX_ALIGN_GET(flags)					\
+    (MALLOCX_ALIGN_GET_SPECIFIED(flags) & (SIZE_T_MAX-1))
+#define MALLOCX_ZERO_GET(flags)						\
+    ((bool)(flags & MALLOCX_ZERO))
+
+#define MALLOCX_TCACHE_GET(flags)					\
+    (((unsigned)((flags & MALLOCX_TCACHE_MASK) >> MALLOCX_TCACHE_SHIFT)) - 2)
+#define MALLOCX_ARENA_GET(flags)					\
+    (((unsigned)(((unsigned)flags) >> MALLOCX_ARENA_SHIFT)) - 1)
+
+/* Smallest size class to support. */
+#define TINY_MIN		(1U << LG_TINY_MIN)
+
+#define LONG			((size_t)(1U << LG_SIZEOF_LONG))
+#define LONG_MASK		(LONG - 1)
+
+/* Return the smallest long multiple that is >= a. */
+#define LONG_CEILING(a)							\
+	(((a) + LONG_MASK) & ~LONG_MASK)
+
+#define SIZEOF_PTR		(1U << LG_SIZEOF_PTR)
+#define PTR_MASK		(SIZEOF_PTR - 1)
+
+/* Return the smallest (void *) multiple that is >= a. */
+#define PTR_CEILING(a)							\
+	(((a) + PTR_MASK) & ~PTR_MASK)
+
+/*
+ * Maximum size of L1 cache line.  This is used to avoid cache line aliasing.
+ * In addition, this controls the spacing of cacheline-spaced size classes.
+ *
+ * CACHELINE cannot be based on LG_CACHELINE because __declspec(align()) can
+ * only handle raw constants.
+ */
+#define LG_CACHELINE		6
+#define CACHELINE		64
+#define CACHELINE_MASK		(CACHELINE - 1)
+
+/* Return the smallest cacheline multiple that is >= s. */
+#define CACHELINE_CEILING(s)						\
+	(((s) + CACHELINE_MASK) & ~CACHELINE_MASK)
+
+/* Return the nearest aligned address at or below a. */
+#define ALIGNMENT_ADDR2BASE(a, alignment)				\
+	((void *)((uintptr_t)(a) & ((~(alignment)) + 1)))
+
+/* Return the offset between a and the nearest aligned address at or below a. */
+#define ALIGNMENT_ADDR2OFFSET(a, alignment)				\
+	((size_t)((uintptr_t)(a) & (alignment - 1)))
+
+/* Return the smallest alignment multiple that is >= s. */
+#define ALIGNMENT_CEILING(s, alignment)					\
+	(((s) + (alignment - 1)) & ((~(alignment)) + 1))
+
+/* Declare a variable-length array. */
+#if __STDC_VERSION__ < 199901L
+#  ifdef _MSC_VER
+#    include <malloc.h>
+#    define alloca _alloca
+#  else
+#    ifdef JEMALLOC_HAS_ALLOCA_H
+#      include <alloca.h>
+#    else
+#      include <stdlib.h>
+#    endif
+#  endif
+#  define VARIABLE_ARRAY(type, name, count) \
+	type *name = alloca(sizeof(type) * (count))
+#else
+#  define VARIABLE_ARRAY(type, name, count) type name[(count)]
+#endif
+
+#endif /* JEMALLOC_INTERNAL_TYPES_H */

deps/jemalloc/include/jemalloc/internal/jemalloc_preamble.h

@@ -0,0 +1,213 @@
+#ifndef JEMALLOC_PREAMBLE_H
+#define JEMALLOC_PREAMBLE_H
+
+#include "jemalloc_internal_defs.h"
+#include "jemalloc/internal/jemalloc_internal_decls.h"
+
+#ifdef JEMALLOC_UTRACE
+#include <sys/ktrace.h>
+#endif
+
+#define JEMALLOC_NO_DEMANGLE
+#ifdef JEMALLOC_JET
+#  undef JEMALLOC_IS_MALLOC
+#  define JEMALLOC_N(n) jet_##n
+#  include "jemalloc/internal/public_namespace.h"
+#  define JEMALLOC_NO_RENAME
+#  include "../jemalloc.h"
+#  undef JEMALLOC_NO_RENAME
+#else
+#  define JEMALLOC_N(n) je_##n
+#  include "../jemalloc.h"
+#endif
+
+#if defined(JEMALLOC_OSATOMIC)
+#include <libkern/OSAtomic.h>
+#endif
+
+#ifdef JEMALLOC_ZONE
+#include <mach/mach_error.h>
+#include <mach/mach_init.h>
+#include <mach/vm_map.h>
+#endif
+
+#include "jemalloc/internal/jemalloc_internal_macros.h"
+
+/*
+ * Note that the ordering matters here; the hook itself is name-mangled.  We
+ * want the inclusion of hooks to happen early, so that we hook as much as
+ * possible.
+ */
+#ifndef JEMALLOC_NO_PRIVATE_NAMESPACE
+#  ifndef JEMALLOC_JET
+#    include "jemalloc/internal/private_namespace.h"
+#  else
+#    include "jemalloc/internal/private_namespace_jet.h"
+#  endif
+#endif
+#include "jemalloc/internal/test_hooks.h"
+
+#ifdef JEMALLOC_DEFINE_MADVISE_FREE
+#  define JEMALLOC_MADV_FREE 8
+#endif
+
+static const bool config_debug =
+#ifdef JEMALLOC_DEBUG
+    true
+#else
+    false
+#endif
+    ;
+static const bool have_dss =
+#ifdef JEMALLOC_DSS
+    true
+#else
+    false
+#endif
+    ;
+static const bool have_madvise_huge =
+#ifdef JEMALLOC_HAVE_MADVISE_HUGE
+    true
+#else
+    false
+#endif
+    ;
+static const bool config_fill =
+#ifdef JEMALLOC_FILL
+    true
+#else
+    false
+#endif
+    ;
+static const bool config_lazy_lock =
+#ifdef JEMALLOC_LAZY_LOCK
+    true
+#else
+    false
+#endif
+    ;
+static const char * const config_malloc_conf = JEMALLOC_CONFIG_MALLOC_CONF;
+static const bool config_prof =
+#ifdef JEMALLOC_PROF
+    true
+#else
+    false
+#endif
+    ;
+static const bool config_prof_libgcc =
+#ifdef JEMALLOC_PROF_LIBGCC
+    true
+#else
+    false
+#endif
+    ;
+static const bool config_prof_libunwind =
+#ifdef JEMALLOC_PROF_LIBUNWIND
+    true
+#else
+    false
+#endif
+    ;
+static const bool maps_coalesce =
+#ifdef JEMALLOC_MAPS_COALESCE
+    true
+#else
+    false
+#endif
+    ;
+static const bool config_stats =
+#ifdef JEMALLOC_STATS
+    true
+#else
+    false
+#endif
+    ;
+static const bool config_tls =
+#ifdef JEMALLOC_TLS
+    true
+#else
+    false
+#endif
+    ;
+static const bool config_utrace =
+#ifdef JEMALLOC_UTRACE
+    true
+#else
+    false
+#endif
+    ;
+static const bool config_xmalloc =
+#ifdef JEMALLOC_XMALLOC
+    true
+#else
+    false
+#endif
+    ;
+static const bool config_cache_oblivious =
+#ifdef JEMALLOC_CACHE_OBLIVIOUS
+    true
+#else
+    false
+#endif
+    ;
+/*
+ * Undocumented, for jemalloc development use only at the moment.  See the note
+ * in jemalloc/internal/log.h.
+ */
+static const bool config_log =
+#ifdef JEMALLOC_LOG
+    true
+#else
+    false
+#endif
+    ;
+/*
+ * Are extra safety checks enabled; things like checking the size of sized
+ * deallocations, double-frees, etc.
+ */
+static const bool config_opt_safety_checks =
+#ifdef JEMALLOC_OPT_SAFETY_CHECKS
+    true
+#elif defined(JEMALLOC_DEBUG)
+    /*
+     * This lets us only guard safety checks by one flag instead of two; fast
+     * checks can guard solely by config_opt_safety_checks and run in debug mode
+     * too.
+     */
+    true
+#else
+    false
+#endif
+    ;
+
+#if defined(_WIN32) || defined(JEMALLOC_HAVE_SCHED_GETCPU)
+/* Currently percpu_arena depends on sched_getcpu. */
+#define JEMALLOC_PERCPU_ARENA
+#endif
+static const bool have_percpu_arena =
+#ifdef JEMALLOC_PERCPU_ARENA
+    true
+#else
+    false
+#endif
+    ;
+/*
+ * Undocumented, and not recommended; the application should take full
+ * responsibility for tracking provenance.
+ */
+static const bool force_ivsalloc =
+#ifdef JEMALLOC_FORCE_IVSALLOC
+    true
+#else
+    false
+#endif
+    ;
+static const bool have_background_thread =
+#ifdef JEMALLOC_BACKGROUND_THREAD
+    true
+#else
+    false
+#endif
+    ;
+
+#endif /* JEMALLOC_PREAMBLE_H */

deps/jemalloc/include/jemalloc/internal/jemalloc_preamble.h.in

@@ -0,0 +1,179 @@
+#ifndef JEMALLOC_PREAMBLE_H
+#define JEMALLOC_PREAMBLE_H
+
+#include "jemalloc_internal_defs.h"
+#include "jemalloc/internal/jemalloc_internal_decls.h"
+
+#ifdef JEMALLOC_UTRACE
+#include <sys/ktrace.h>
+#endif
+
+#define JEMALLOC_NO_DEMANGLE
+#ifdef JEMALLOC_JET
+#  undef JEMALLOC_IS_MALLOC
+#  define JEMALLOC_N(n) jet_##n
+#  include "jemalloc/internal/public_namespace.h"
+#  define JEMALLOC_NO_RENAME
+#  include "../jemalloc@install_suffix@.h"
+#  undef JEMALLOC_NO_RENAME
+#else
+#  define JEMALLOC_N(n) @private_namespace@##n
+#  include "../jemalloc@install_suffix@.h"
+#endif
+
+#if (defined(JEMALLOC_OSATOMIC) || defined(JEMALLOC_OSSPIN))
+#include <libkern/OSAtomic.h>
+#endif
+
+#ifdef JEMALLOC_ZONE
+#include <mach/mach_error.h>
+#include <mach/mach_init.h>
+#include <mach/vm_map.h>
+#endif
+
+#include "jemalloc/internal/jemalloc_internal_macros.h"
+
+/*
+ * Note that the ordering matters here; the hook itself is name-mangled.  We
+ * want the inclusion of hooks to happen early, so that we hook as much as
+ * possible.
+ */
+#ifndef JEMALLOC_NO_PRIVATE_NAMESPACE
+#  ifndef JEMALLOC_JET
+#    include "jemalloc/internal/private_namespace.h"
+#  else
+#    include "jemalloc/internal/private_namespace_jet.h"
+#  endif
+#endif
+#include "jemalloc/internal/hooks.h"
+
+static const bool config_debug =
+#ifdef JEMALLOC_DEBUG
+    true
+#else
+    false
+#endif
+    ;
+static const bool have_dss =
+#ifdef JEMALLOC_DSS
+    true
+#else
+    false
+#endif
+    ;
+static const bool config_fill =
+#ifdef JEMALLOC_FILL
+    true
+#else
+    false
+#endif
+    ;
+static const bool config_lazy_lock =
+#ifdef JEMALLOC_LAZY_LOCK
+    true
+#else
+    false
+#endif
+    ;
+static const char * const config_malloc_conf = JEMALLOC_CONFIG_MALLOC_CONF;
+static const bool config_prof =
+#ifdef JEMALLOC_PROF
+    true
+#else
+    false
+#endif
+    ;
+static const bool config_prof_libgcc =
+#ifdef JEMALLOC_PROF_LIBGCC
+    true
+#else
+    false
+#endif
+    ;
+static const bool config_prof_libunwind =
+#ifdef JEMALLOC_PROF_LIBUNWIND
+    true
+#else
+    false
+#endif
+    ;
+static const bool maps_coalesce =
+#ifdef JEMALLOC_MAPS_COALESCE
+    true
+#else
+    false
+#endif
+    ;
+static const bool config_stats =
+#ifdef JEMALLOC_STATS
+    true
+#else
+    false
+#endif
+    ;
+static const bool config_thp =
+#ifdef JEMALLOC_THP
+    true
+#else
+    false
+#endif
+    ;
+static const bool config_tls =
+#ifdef JEMALLOC_TLS
+    true
+#else
+    false
+#endif
+    ;
+static const bool config_utrace =
+#ifdef JEMALLOC_UTRACE
+    true
+#else
+    false
+#endif
+    ;
+static const bool config_xmalloc =
+#ifdef JEMALLOC_XMALLOC
+    true
+#else
+    false
+#endif
+    ;
+static const bool config_cache_oblivious =
+#ifdef JEMALLOC_CACHE_OBLIVIOUS
+    true
+#else
+    false
+#endif
+    ;
+#ifdef JEMALLOC_HAVE_SCHED_GETCPU
+/* Currently percpu_arena depends on sched_getcpu. */
+#define JEMALLOC_PERCPU_ARENA
+#endif
+static const bool have_percpu_arena =
+#ifdef JEMALLOC_PERCPU_ARENA
+    true
+#else
+    false
+#endif
+    ;
+/*
+ * Undocumented, and not recommended; the application should take full
+ * responsibility for tracking provenance.
+ */
+static const bool force_ivsalloc =
+#ifdef JEMALLOC_FORCE_IVSALLOC
+    true
+#else
+    false
+#endif
+    ;
+static const bool have_background_thread =
+#ifdef JEMALLOC_BACKGROUND_THREAD
+    true
+#else
+    false
+#endif
+    ;
+
+#endif /* JEMALLOC_PREAMBLE_H */

deps/jemalloc/include/jemalloc/internal/large_externs.h

@@ -0,0 +1,32 @@
+#ifndef JEMALLOC_INTERNAL_LARGE_EXTERNS_H
+#define JEMALLOC_INTERNAL_LARGE_EXTERNS_H
+
+#include "jemalloc/internal/hook.h"
+
+void *large_malloc(tsdn_t *tsdn, arena_t *arena, size_t usize, bool zero);
+void *large_palloc(tsdn_t *tsdn, arena_t *arena, size_t usize, size_t alignment,
+    bool zero);
+bool large_ralloc_no_move(tsdn_t *tsdn, extent_t *extent, size_t usize_min,
+    size_t usize_max, bool zero);
+void *large_ralloc(tsdn_t *tsdn, arena_t *arena, void *ptr, size_t usize,
+    size_t alignment, bool zero, tcache_t *tcache,
+    hook_ralloc_args_t *hook_args);
+
+typedef void (large_dalloc_junk_t)(void *, size_t);
+extern large_dalloc_junk_t *JET_MUTABLE large_dalloc_junk;
+
+typedef void (large_dalloc_maybe_junk_t)(void *, size_t);
+extern large_dalloc_maybe_junk_t *JET_MUTABLE large_dalloc_maybe_junk;
+
+void large_dalloc_prep_junked_locked(tsdn_t *tsdn, extent_t *extent);
+void large_dalloc_finish(tsdn_t *tsdn, extent_t *extent);
+void large_dalloc(tsdn_t *tsdn, extent_t *extent);
+size_t large_salloc(tsdn_t *tsdn, const extent_t *extent);
+prof_tctx_t *large_prof_tctx_get(tsdn_t *tsdn, const extent_t *extent);
+void large_prof_tctx_set(tsdn_t *tsdn, extent_t *extent, prof_tctx_t *tctx);
+void large_prof_tctx_reset(tsdn_t *tsdn, extent_t *extent);
+
+nstime_t large_prof_alloc_time_get(const extent_t *extent);
+void large_prof_alloc_time_set(extent_t *extent, nstime_t time);
+
+#endif /* JEMALLOC_INTERNAL_LARGE_EXTERNS_H */

deps/jemalloc/include/jemalloc/internal/log.h

@@ -0,0 +1,115 @@
+#ifndef JEMALLOC_INTERNAL_LOG_H
+#define JEMALLOC_INTERNAL_LOG_H
+
+#include "jemalloc/internal/atomic.h"
+#include "jemalloc/internal/malloc_io.h"
+#include "jemalloc/internal/mutex.h"
+
+#ifdef JEMALLOC_LOG
+#  define JEMALLOC_LOG_VAR_BUFSIZE 1000
+#else
+#  define JEMALLOC_LOG_VAR_BUFSIZE 1
+#endif
+
+#define JEMALLOC_LOG_BUFSIZE 4096
+
+/*
+ * The log malloc_conf option is a '|'-delimited list of log_var name segments
+ * which should be logged.  The names are themselves hierarchical, with '.' as
+ * the delimiter (a "segment" is just a prefix in the log namespace).  So, if
+ * you have:
+ *
+ * log("arena", "log msg for arena"); // 1
+ * log("arena.a", "log msg for arena.a"); // 2
+ * log("arena.b", "log msg for arena.b"); // 3
+ * log("arena.a.a", "log msg for arena.a.a"); // 4
+ * log("extent.a", "log msg for extent.a"); // 5
+ * log("extent.b", "log msg for extent.b"); // 6
+ *
+ * And your malloc_conf option is "log=arena.a|extent", then lines 2, 4, 5, and
+ * 6 will print at runtime.  You can enable logging from all log vars by
+ * writing "log=.".
+ *
+ * None of this should be regarded as a stable API for right now.  It's intended
+ * as a debugging interface, to let us keep around some of our printf-debugging
+ * statements.
+ */
+
+extern char log_var_names[JEMALLOC_LOG_VAR_BUFSIZE];
+extern atomic_b_t log_init_done;
+
+typedef struct log_var_s log_var_t;
+struct log_var_s {
+	/*
+	 * Lowest bit is "inited", second lowest is "enabled".  Putting them in
+	 * a single word lets us avoid any fences on weak architectures.
+	 */
+	atomic_u_t state;
+	const char *name;
+};
+
+#define LOG_NOT_INITIALIZED 0U
+#define LOG_INITIALIZED_NOT_ENABLED 1U
+#define LOG_ENABLED 2U
+
+#define LOG_VAR_INIT(name_str) {ATOMIC_INIT(LOG_NOT_INITIALIZED), name_str}
+
+/*
+ * Returns the value we should assume for state (which is not necessarily
+ * accurate; if logging is done before logging has finished initializing, then
+ * we default to doing the safe thing by logging everything).
+ */
+unsigned log_var_update_state(log_var_t *log_var);
+
+/* We factor out the metadata management to allow us to test more easily. */
+#define log_do_begin(log_var)						\
+if (config_log) {							\
+	unsigned log_state = atomic_load_u(&(log_var).state,		\
+	    ATOMIC_RELAXED);						\
+	if (unlikely(log_state == LOG_NOT_INITIALIZED)) {		\
+		log_state = log_var_update_state(&(log_var));		\
+		assert(log_state != LOG_NOT_INITIALIZED);		\
+	}								\
+	if (log_state == LOG_ENABLED) {					\
+		{
+			/* User code executes here. */
+#define log_do_end(log_var)						\
+		}							\
+	}								\
+}
+
+/*
+ * MSVC has some preprocessor bugs in its expansion of __VA_ARGS__ during
+ * preprocessing.  To work around this, we take all potential extra arguments in
+ * a var-args functions.  Since a varargs macro needs at least one argument in
+ * the "...", we accept the format string there, and require that the first
+ * argument in this "..." is a const char *.
+ */
+static inline void
+log_impl_varargs(const char *name, ...) {
+	char buf[JEMALLOC_LOG_BUFSIZE];
+	va_list ap;
+
+	va_start(ap, name);
+	const char *format = va_arg(ap, const char *);
+	size_t dst_offset = 0;
+	dst_offset += malloc_snprintf(buf, JEMALLOC_LOG_BUFSIZE, "%s: ", name);
+	dst_offset += malloc_vsnprintf(buf + dst_offset,
+	    JEMALLOC_LOG_BUFSIZE - dst_offset, format, ap);
+	dst_offset += malloc_snprintf(buf + dst_offset,
+	    JEMALLOC_LOG_BUFSIZE - dst_offset, "\n");
+	va_end(ap);
+
+	malloc_write(buf);
+}
+
+/* Call as log("log.var.str", "format_string %d", arg_for_format_string); */
+#define LOG(log_var_str, ...)						\
+do {									\
+	static log_var_t log_var = LOG_VAR_INIT(log_var_str);		\
+	log_do_begin(log_var)						\
+		log_impl_varargs((log_var).name, __VA_ARGS__);		\
+	log_do_end(log_var)						\
+} while (0)
+
+#endif /* JEMALLOC_INTERNAL_LOG_H */

deps/jemalloc/include/jemalloc/internal/malloc_io.h

@@ -0,0 +1,102 @@
+#ifndef JEMALLOC_INTERNAL_MALLOC_IO_H
+#define JEMALLOC_INTERNAL_MALLOC_IO_H
+
+#ifdef _WIN32
+#  ifdef _WIN64
+#    define FMT64_PREFIX "ll"
+#    define FMTPTR_PREFIX "ll"
+#  else
+#    define FMT64_PREFIX "ll"
+#    define FMTPTR_PREFIX ""
+#  endif
+#  define FMTd32 "d"
+#  define FMTu32 "u"
+#  define FMTx32 "x"
+#  define FMTd64 FMT64_PREFIX "d"
+#  define FMTu64 FMT64_PREFIX "u"
+#  define FMTx64 FMT64_PREFIX "x"
+#  define FMTdPTR FMTPTR_PREFIX "d"
+#  define FMTuPTR FMTPTR_PREFIX "u"
+#  define FMTxPTR FMTPTR_PREFIX "x"
+#else
+#  include <inttypes.h>
+#  define FMTd32 PRId32
+#  define FMTu32 PRIu32
+#  define FMTx32 PRIx32
+#  define FMTd64 PRId64
+#  define FMTu64 PRIu64
+#  define FMTx64 PRIx64
+#  define FMTdPTR PRIdPTR
+#  define FMTuPTR PRIuPTR
+#  define FMTxPTR PRIxPTR
+#endif
+
+/* Size of stack-allocated buffer passed to buferror(). */
+#define BUFERROR_BUF		64
+
+/*
+ * Size of stack-allocated buffer used by malloc_{,v,vc}printf().  This must be
+ * large enough for all possible uses within jemalloc.
+ */
+#define MALLOC_PRINTF_BUFSIZE	4096
+
+int buferror(int err, char *buf, size_t buflen);
+uintmax_t malloc_strtoumax(const char *restrict nptr, char **restrict endptr,
+    int base);
+void malloc_write(const char *s);
+
+/*
+ * malloc_vsnprintf() supports a subset of snprintf(3) that avoids floating
+ * point math.
+ */
+size_t malloc_vsnprintf(char *str, size_t size, const char *format,
+    va_list ap);
+size_t malloc_snprintf(char *str, size_t size, const char *format, ...)
+    JEMALLOC_FORMAT_PRINTF(3, 4);
+/*
+ * The caller can set write_cb to null to choose to print with the
+ * je_malloc_message hook.
+ */
+void malloc_vcprintf(void (*write_cb)(void *, const char *), void *cbopaque,
+    const char *format, va_list ap);
+void malloc_cprintf(void (*write_cb)(void *, const char *), void *cbopaque,
+    const char *format, ...) JEMALLOC_FORMAT_PRINTF(3, 4);
+void malloc_printf(const char *format, ...) JEMALLOC_FORMAT_PRINTF(1, 2);
+
+static inline ssize_t
+malloc_write_fd(int fd, const void *buf, size_t count) {
+#if defined(JEMALLOC_USE_SYSCALL) && defined(SYS_write)
+	/*
+	 * Use syscall(2) rather than write(2) when possible in order to avoid
+	 * the possibility of memory allocation within libc.  This is necessary
+	 * on FreeBSD; most operating systems do not have this problem though.
+	 *
+	 * syscall() returns long or int, depending on platform, so capture the
+	 * result in the widest plausible type to avoid compiler warnings.
+	 */
+	long result = syscall(SYS_write, fd, buf, count);
+#else
+	ssize_t result = (ssize_t)write(fd, buf,
+#ifdef _WIN32
+	    (unsigned int)
+#endif
+	    count);
+#endif
+	return (ssize_t)result;
+}
+
+static inline ssize_t
+malloc_read_fd(int fd, void *buf, size_t count) {
+#if defined(JEMALLOC_USE_SYSCALL) && defined(SYS_read)
+	long result = syscall(SYS_read, fd, buf, count);
+#else
+	ssize_t result = read(fd, buf,
+#ifdef _WIN32
+	    (unsigned int)
+#endif
+	    count);
+#endif
+	return (ssize_t)result;
+}
+
+#endif /* JEMALLOC_INTERNAL_MALLOC_IO_H */

deps/jemalloc/include/jemalloc/internal/mb.h

@@ -0,0 +1,115 @@
+/******************************************************************************/
+#ifdef JEMALLOC_H_TYPES
+
+#endif /* JEMALLOC_H_TYPES */
+/******************************************************************************/
+#ifdef JEMALLOC_H_STRUCTS
+
+#endif /* JEMALLOC_H_STRUCTS */
+/******************************************************************************/
+#ifdef JEMALLOC_H_EXTERNS
+
+#endif /* JEMALLOC_H_EXTERNS */
+/******************************************************************************/
+#ifdef JEMALLOC_H_INLINES
+
+#ifndef JEMALLOC_ENABLE_INLINE
+void	mb_write(void);
+#endif
+
+#if (defined(JEMALLOC_ENABLE_INLINE) || defined(JEMALLOC_MB_C_))
+#ifdef __i386__
+/*
+ * According to the Intel Architecture Software Developer's Manual, current
+ * processors execute instructions in order from the perspective of other
+ * processors in a multiprocessor system, but 1) Intel reserves the right to
+ * change that, and 2) the compiler's optimizer could re-order instructions if
+ * there weren't some form of barrier.  Therefore, even if running on an
+ * architecture that does not need memory barriers (everything through at least
+ * i686), an "optimizer barrier" is necessary.
+ */
+JEMALLOC_INLINE void
+mb_write(void)
+{
+
+#  if 0
+	/* This is a true memory barrier. */
+	asm volatile ("pusha;"
+	    "xor  %%eax,%%eax;"
+	    "cpuid;"
+	    "popa;"
+	    : /* Outputs. */
+	    : /* Inputs. */
+	    : "memory" /* Clobbers. */
+	    );
+#else
+	/*
+	 * This is hopefully enough to keep the compiler from reordering
+	 * instructions around this one.
+	 */
+	asm volatile ("nop;"
+	    : /* Outputs. */
+	    : /* Inputs. */
+	    : "memory" /* Clobbers. */
+	    );
+#endif
+}
+#elif (defined(__amd64__) || defined(__x86_64__))
+JEMALLOC_INLINE void
+mb_write(void)
+{
+
+	asm volatile ("sfence"
+	    : /* Outputs. */
+	    : /* Inputs. */
+	    : "memory" /* Clobbers. */
+	    );
+}
+#elif defined(__powerpc__)
+JEMALLOC_INLINE void
+mb_write(void)
+{
+
+	asm volatile ("eieio"
+	    : /* Outputs. */
+	    : /* Inputs. */
+	    : "memory" /* Clobbers. */
+	    );
+}
+#elif defined(__sparc64__)
+JEMALLOC_INLINE void
+mb_write(void)
+{
+
+	asm volatile ("membar #StoreStore"
+	    : /* Outputs. */
+	    : /* Inputs. */
+	    : "memory" /* Clobbers. */
+	    );
+}
+#elif defined(__tile__)
+JEMALLOC_INLINE void
+mb_write(void)
+{
+
+	__sync_synchronize();
+}
+#else
+/*
+ * This is much slower than a simple memory barrier, but the semantics of mutex
+ * unlock make this work.
+ */
+JEMALLOC_INLINE void
+mb_write(void)
+{
+	malloc_mutex_t mtx;
+
+	malloc_mutex_init(&mtx);
+	malloc_mutex_lock(&mtx);
+	malloc_mutex_unlock(&mtx);
+}
+#endif
+#endif
+
+#endif /* JEMALLOC_H_INLINES */
+/******************************************************************************/

deps/jemalloc/include/jemalloc/internal/mutex.h

@@ -0,0 +1,288 @@
+#ifndef JEMALLOC_INTERNAL_MUTEX_H
+#define JEMALLOC_INTERNAL_MUTEX_H
+
+#include "jemalloc/internal/atomic.h"
+#include "jemalloc/internal/mutex_prof.h"
+#include "jemalloc/internal/tsd.h"
+#include "jemalloc/internal/witness.h"
+
+typedef enum {
+	/* Can only acquire one mutex of a given witness rank at a time. */
+	malloc_mutex_rank_exclusive,
+	/*
+	 * Can acquire multiple mutexes of the same witness rank, but in
+	 * address-ascending order only.
+	 */
+	malloc_mutex_address_ordered
+} malloc_mutex_lock_order_t;
+
+typedef struct malloc_mutex_s malloc_mutex_t;
+struct malloc_mutex_s {
+	union {
+		struct {
+			/*
+			 * prof_data is defined first to reduce cacheline
+			 * bouncing: the data is not touched by the mutex holder
+			 * during unlocking, while might be modified by
+			 * contenders.  Having it before the mutex itself could
+			 * avoid prefetching a modified cacheline (for the
+			 * unlocking thread).
+			 */
+			mutex_prof_data_t	prof_data;
+#ifdef _WIN32
+#  if _WIN32_WINNT >= 0x0600
+			SRWLOCK         	lock;
+#  else
+			CRITICAL_SECTION	lock;
+#  endif
+#elif (defined(JEMALLOC_OS_UNFAIR_LOCK))
+			os_unfair_lock		lock;
+#elif (defined(JEMALLOC_MUTEX_INIT_CB))
+			pthread_mutex_t		lock;
+			malloc_mutex_t		*postponed_next;
+#else
+			pthread_mutex_t		lock;
+#endif
+			/* 
+			 * Hint flag to avoid exclusive cache line contention
+			 * during spin waiting
+			 */
+			atomic_b_t		locked;
+		};
+		/*
+		 * We only touch witness when configured w/ debug.  However we
+		 * keep the field in a union when !debug so that we don't have
+		 * to pollute the code base with #ifdefs, while avoid paying the
+		 * memory cost.
+		 */
+#if !defined(JEMALLOC_DEBUG)
+		witness_t			witness;
+		malloc_mutex_lock_order_t	lock_order;
+#endif
+	};
+
+#if defined(JEMALLOC_DEBUG)
+	witness_t			witness;
+	malloc_mutex_lock_order_t	lock_order;
+#endif
+};
+
+/*
+ * Based on benchmark results, a fixed spin with this amount of retries works
+ * well for our critical sections.
+ */
+#define MALLOC_MUTEX_MAX_SPIN 250
+
+#ifdef _WIN32
+#  if _WIN32_WINNT >= 0x0600
+#    define MALLOC_MUTEX_LOCK(m)    AcquireSRWLockExclusive(&(m)->lock)
+#    define MALLOC_MUTEX_UNLOCK(m)  ReleaseSRWLockExclusive(&(m)->lock)
+#    define MALLOC_MUTEX_TRYLOCK(m) (!TryAcquireSRWLockExclusive(&(m)->lock))
+#  else
+#    define MALLOC_MUTEX_LOCK(m)    EnterCriticalSection(&(m)->lock)
+#    define MALLOC_MUTEX_UNLOCK(m)  LeaveCriticalSection(&(m)->lock)
+#    define MALLOC_MUTEX_TRYLOCK(m) (!TryEnterCriticalSection(&(m)->lock))
+#  endif
+#elif (defined(JEMALLOC_OS_UNFAIR_LOCK))
+#    define MALLOC_MUTEX_LOCK(m)    os_unfair_lock_lock(&(m)->lock)
+#    define MALLOC_MUTEX_UNLOCK(m)  os_unfair_lock_unlock(&(m)->lock)
+#    define MALLOC_MUTEX_TRYLOCK(m) (!os_unfair_lock_trylock(&(m)->lock))
+#else
+#    define MALLOC_MUTEX_LOCK(m)    pthread_mutex_lock(&(m)->lock)
+#    define MALLOC_MUTEX_UNLOCK(m)  pthread_mutex_unlock(&(m)->lock)
+#    define MALLOC_MUTEX_TRYLOCK(m) (pthread_mutex_trylock(&(m)->lock) != 0)
+#endif
+
+#define LOCK_PROF_DATA_INITIALIZER					\
+    {NSTIME_ZERO_INITIALIZER, NSTIME_ZERO_INITIALIZER, 0, 0, 0,		\
+	    ATOMIC_INIT(0), 0, NULL, 0}
+
+#ifdef _WIN32
+#  define MALLOC_MUTEX_INITIALIZER
+#elif (defined(JEMALLOC_OS_UNFAIR_LOCK))
+#  if defined(JEMALLOC_DEBUG)
+#    define MALLOC_MUTEX_INITIALIZER					\
+  {{{LOCK_PROF_DATA_INITIALIZER, OS_UNFAIR_LOCK_INIT, ATOMIC_INIT(false)}}, \
+         WITNESS_INITIALIZER("mutex", WITNESS_RANK_OMIT), 0}
+#  else
+#    define MALLOC_MUTEX_INITIALIZER                      \
+  {{{LOCK_PROF_DATA_INITIALIZER, OS_UNFAIR_LOCK_INIT, ATOMIC_INIT(false)}},  \
+      WITNESS_INITIALIZER("mutex", WITNESS_RANK_OMIT)}
+#  endif
+#elif (defined(JEMALLOC_MUTEX_INIT_CB))
+#  if (defined(JEMALLOC_DEBUG))
+#     define MALLOC_MUTEX_INITIALIZER					\
+      {{{LOCK_PROF_DATA_INITIALIZER, PTHREAD_MUTEX_INITIALIZER, NULL, ATOMIC_INIT(false)}},	\
+           WITNESS_INITIALIZER("mutex", WITNESS_RANK_OMIT), 0}
+#  else
+#     define MALLOC_MUTEX_INITIALIZER					\
+      {{{LOCK_PROF_DATA_INITIALIZER, PTHREAD_MUTEX_INITIALIZER, NULL, ATOMIC_INIT(false)}},	\
+           WITNESS_INITIALIZER("mutex", WITNESS_RANK_OMIT)}
+#  endif
+
+#else
+#    define MALLOC_MUTEX_TYPE PTHREAD_MUTEX_DEFAULT
+#  if defined(JEMALLOC_DEBUG)
+#    define MALLOC_MUTEX_INITIALIZER					\
+     {{{LOCK_PROF_DATA_INITIALIZER, PTHREAD_MUTEX_INITIALIZER, ATOMIC_INIT(false)}}, \
+           WITNESS_INITIALIZER("mutex", WITNESS_RANK_OMIT), 0}
+#  else
+#    define MALLOC_MUTEX_INITIALIZER                          \
+     {{{LOCK_PROF_DATA_INITIALIZER, PTHREAD_MUTEX_INITIALIZER, ATOMIC_INIT(false)}},	\
+      WITNESS_INITIALIZER("mutex", WITNESS_RANK_OMIT)}
+#  endif
+#endif
+
+#ifdef JEMALLOC_LAZY_LOCK
+extern bool isthreaded;
+#else
+#  undef isthreaded /* Undo private_namespace.h definition. */
+#  define isthreaded true
+#endif
+
+bool malloc_mutex_init(malloc_mutex_t *mutex, const char *name,
+    witness_rank_t rank, malloc_mutex_lock_order_t lock_order);
+void malloc_mutex_prefork(tsdn_t *tsdn, malloc_mutex_t *mutex);
+void malloc_mutex_postfork_parent(tsdn_t *tsdn, malloc_mutex_t *mutex);
+void malloc_mutex_postfork_child(tsdn_t *tsdn, malloc_mutex_t *mutex);
+bool malloc_mutex_boot(void);
+void malloc_mutex_prof_data_reset(tsdn_t *tsdn, malloc_mutex_t *mutex);
+
+void malloc_mutex_lock_slow(malloc_mutex_t *mutex);
+
+static inline void
+malloc_mutex_lock_final(malloc_mutex_t *mutex) {
+	MALLOC_MUTEX_LOCK(mutex);
+	atomic_store_b(&mutex->locked, true, ATOMIC_RELAXED);
+}
+
+static inline bool
+malloc_mutex_trylock_final(malloc_mutex_t *mutex) {
+	return MALLOC_MUTEX_TRYLOCK(mutex);
+}
+
+static inline void
+mutex_owner_stats_update(tsdn_t *tsdn, malloc_mutex_t *mutex) {
+	if (config_stats) {
+		mutex_prof_data_t *data = &mutex->prof_data;
+		data->n_lock_ops++;
+		if (data->prev_owner != tsdn) {
+			data->prev_owner = tsdn;
+			data->n_owner_switches++;
+		}
+	}
+}
+
+/* Trylock: return false if the lock is successfully acquired. */
+static inline bool
+malloc_mutex_trylock(tsdn_t *tsdn, malloc_mutex_t *mutex) {
+	witness_assert_not_owner(tsdn_witness_tsdp_get(tsdn), &mutex->witness);
+	if (isthreaded) {
+		if (malloc_mutex_trylock_final(mutex)) {
+			atomic_store_b(&mutex->locked, true, ATOMIC_RELAXED);
+			return true;
+		}
+		mutex_owner_stats_update(tsdn, mutex);
+	}
+	witness_lock(tsdn_witness_tsdp_get(tsdn), &mutex->witness);
+
+	return false;
+}
+
+/* Aggregate lock prof data. */
+static inline void
+malloc_mutex_prof_merge(mutex_prof_data_t *sum, mutex_prof_data_t *data) {
+	nstime_add(&sum->tot_wait_time, &data->tot_wait_time);
+	if (nstime_compare(&sum->max_wait_time, &data->max_wait_time) < 0) {
+		nstime_copy(&sum->max_wait_time, &data->max_wait_time);
+	}
+
+	sum->n_wait_times += data->n_wait_times;
+	sum->n_spin_acquired += data->n_spin_acquired;
+
+	if (sum->max_n_thds < data->max_n_thds) {
+		sum->max_n_thds = data->max_n_thds;
+	}
+	uint32_t cur_n_waiting_thds = atomic_load_u32(&sum->n_waiting_thds,
+	    ATOMIC_RELAXED);
+	uint32_t new_n_waiting_thds = cur_n_waiting_thds + atomic_load_u32(
+	    &data->n_waiting_thds, ATOMIC_RELAXED);
+	atomic_store_u32(&sum->n_waiting_thds, new_n_waiting_thds,
+	    ATOMIC_RELAXED);
+	sum->n_owner_switches += data->n_owner_switches;
+	sum->n_lock_ops += data->n_lock_ops;
+}
+
+static inline void
+malloc_mutex_lock(tsdn_t *tsdn, malloc_mutex_t *mutex) {
+	witness_assert_not_owner(tsdn_witness_tsdp_get(tsdn), &mutex->witness);
+	if (isthreaded) {
+		if (malloc_mutex_trylock_final(mutex)) {
+			malloc_mutex_lock_slow(mutex);
+			atomic_store_b(&mutex->locked, true, ATOMIC_RELAXED);
+		}
+		mutex_owner_stats_update(tsdn, mutex);
+	}
+	witness_lock(tsdn_witness_tsdp_get(tsdn), &mutex->witness);
+}
+
+static inline void
+malloc_mutex_unlock(tsdn_t *tsdn, malloc_mutex_t *mutex) {
+	atomic_store_b(&mutex->locked, false, ATOMIC_RELAXED);
+	witness_unlock(tsdn_witness_tsdp_get(tsdn), &mutex->witness);
+	if (isthreaded) {
+		MALLOC_MUTEX_UNLOCK(mutex);
+	}
+}
+
+static inline void
+malloc_mutex_assert_owner(tsdn_t *tsdn, malloc_mutex_t *mutex) {
+	witness_assert_owner(tsdn_witness_tsdp_get(tsdn), &mutex->witness);
+}
+
+static inline void
+malloc_mutex_assert_not_owner(tsdn_t *tsdn, malloc_mutex_t *mutex) {
+	witness_assert_not_owner(tsdn_witness_tsdp_get(tsdn), &mutex->witness);
+}
+
+/* Copy the prof data from mutex for processing. */
+static inline void
+malloc_mutex_prof_read(tsdn_t *tsdn, mutex_prof_data_t *data,
+    malloc_mutex_t *mutex) {
+	mutex_prof_data_t *source = &mutex->prof_data;
+	/* Can only read holding the mutex. */
+	malloc_mutex_assert_owner(tsdn, mutex);
+
+	/*
+	 * Not *really* allowed (we shouldn't be doing non-atomic loads of
+	 * atomic data), but the mutex protection makes this safe, and writing
+	 * a member-for-member copy is tedious for this situation.
+	 */
+	*data = *source;
+	/* n_wait_thds is not reported (modified w/o locking). */
+	atomic_store_u32(&data->n_waiting_thds, 0, ATOMIC_RELAXED);
+}
+
+static inline void
+malloc_mutex_prof_accum(tsdn_t *tsdn, mutex_prof_data_t *data,
+    malloc_mutex_t *mutex) {
+	mutex_prof_data_t *source = &mutex->prof_data;
+	/* Can only read holding the mutex. */
+	malloc_mutex_assert_owner(tsdn, mutex);
+
+	nstime_add(&data->tot_wait_time, &source->tot_wait_time);
+	if (nstime_compare(&source->max_wait_time, &data->max_wait_time) > 0) {
+		nstime_copy(&data->max_wait_time, &source->max_wait_time);
+	}
+	data->n_wait_times += source->n_wait_times;
+	data->n_spin_acquired += source->n_spin_acquired;
+	if (data->max_n_thds < source->max_n_thds) {
+		data->max_n_thds = source->max_n_thds;
+	}
+	/* n_wait_thds is not reported. */
+	atomic_store_u32(&data->n_waiting_thds, 0, ATOMIC_RELAXED);
+	data->n_owner_switches += source->n_owner_switches;
+	data->n_lock_ops += source->n_lock_ops;
+}
+
+#endif /* JEMALLOC_INTERNAL_MUTEX_H */

deps/jemalloc/include/jemalloc/internal/mutex_pool.h

@@ -0,0 +1,94 @@
+#ifndef JEMALLOC_INTERNAL_MUTEX_POOL_H
+#define JEMALLOC_INTERNAL_MUTEX_POOL_H
+
+#include "jemalloc/internal/hash.h"
+#include "jemalloc/internal/mutex.h"
+#include "jemalloc/internal/witness.h"
+
+/* We do mod reductions by this value, so it should be kept a power of 2. */
+#define MUTEX_POOL_SIZE 256
+
+typedef struct mutex_pool_s mutex_pool_t;
+struct mutex_pool_s {
+	malloc_mutex_t mutexes[MUTEX_POOL_SIZE];
+};
+
+bool mutex_pool_init(mutex_pool_t *pool, const char *name, witness_rank_t rank);
+
+/* Internal helper - not meant to be called outside this module. */
+static inline malloc_mutex_t *
+mutex_pool_mutex(mutex_pool_t *pool, uintptr_t key) {
+	size_t hash_result[2];
+	hash(&key, sizeof(key), 0xd50dcc1b, hash_result);
+	return &pool->mutexes[hash_result[0] % MUTEX_POOL_SIZE];
+}
+
+static inline void
+mutex_pool_assert_not_held(tsdn_t *tsdn, mutex_pool_t *pool) {
+	for (int i = 0; i < MUTEX_POOL_SIZE; i++) {
+		malloc_mutex_assert_not_owner(tsdn, &pool->mutexes[i]);
+	}
+}
+
+/*
+ * Note that a mutex pool doesn't work exactly the way an embdedded mutex would.
+ * You're not allowed to acquire mutexes in the pool one at a time.  You have to
+ * acquire all the mutexes you'll need in a single function call, and then
+ * release them all in a single function call.
+ */
+
+static inline void
+mutex_pool_lock(tsdn_t *tsdn, mutex_pool_t *pool, uintptr_t key) {
+	mutex_pool_assert_not_held(tsdn, pool);
+
+	malloc_mutex_t *mutex = mutex_pool_mutex(pool, key);
+	malloc_mutex_lock(tsdn, mutex);
+}
+
+static inline void
+mutex_pool_unlock(tsdn_t *tsdn, mutex_pool_t *pool, uintptr_t key) {
+	malloc_mutex_t *mutex = mutex_pool_mutex(pool, key);
+	malloc_mutex_unlock(tsdn, mutex);
+
+	mutex_pool_assert_not_held(tsdn, pool);
+}
+
+static inline void
+mutex_pool_lock2(tsdn_t *tsdn, mutex_pool_t *pool, uintptr_t key1,
+    uintptr_t key2) {
+	mutex_pool_assert_not_held(tsdn, pool);
+
+	malloc_mutex_t *mutex1 = mutex_pool_mutex(pool, key1);
+	malloc_mutex_t *mutex2 = mutex_pool_mutex(pool, key2);
+	if ((uintptr_t)mutex1 < (uintptr_t)mutex2) {
+		malloc_mutex_lock(tsdn, mutex1);
+		malloc_mutex_lock(tsdn, mutex2);
+	} else if ((uintptr_t)mutex1 == (uintptr_t)mutex2) {
+		malloc_mutex_lock(tsdn, mutex1);
+	} else {
+		malloc_mutex_lock(tsdn, mutex2);
+		malloc_mutex_lock(tsdn, mutex1);
+	}
+}
+
+static inline void
+mutex_pool_unlock2(tsdn_t *tsdn, mutex_pool_t *pool, uintptr_t key1,
+    uintptr_t key2) {
+	malloc_mutex_t *mutex1 = mutex_pool_mutex(pool, key1);
+	malloc_mutex_t *mutex2 = mutex_pool_mutex(pool, key2);
+	if (mutex1 == mutex2) {
+		malloc_mutex_unlock(tsdn, mutex1);
+	} else {
+		malloc_mutex_unlock(tsdn, mutex1);
+		malloc_mutex_unlock(tsdn, mutex2);
+	}
+
+	mutex_pool_assert_not_held(tsdn, pool);
+}
+
+static inline void
+mutex_pool_assert_owner(tsdn_t *tsdn, mutex_pool_t *pool, uintptr_t key) {
+	malloc_mutex_assert_owner(tsdn, mutex_pool_mutex(pool, key));
+}
+
+#endif /* JEMALLOC_INTERNAL_MUTEX_POOL_H */

deps/jemalloc/include/jemalloc/internal/mutex_prof.h

@@ -0,0 +1,108 @@
+#ifndef JEMALLOC_INTERNAL_MUTEX_PROF_H
+#define JEMALLOC_INTERNAL_MUTEX_PROF_H
+
+#include "jemalloc/internal/atomic.h"
+#include "jemalloc/internal/nstime.h"
+#include "jemalloc/internal/tsd_types.h"
+
+#define MUTEX_PROF_GLOBAL_MUTEXES					\
+    OP(background_thread)						\
+    OP(ctl)								\
+    OP(prof)
+
+typedef enum {
+#define OP(mtx) global_prof_mutex_##mtx,
+	MUTEX_PROF_GLOBAL_MUTEXES
+#undef OP
+	mutex_prof_num_global_mutexes
+} mutex_prof_global_ind_t;
+
+#define MUTEX_PROF_ARENA_MUTEXES					\
+    OP(large)								\
+    OP(extent_avail)							\
+    OP(extents_dirty)							\
+    OP(extents_muzzy)							\
+    OP(extents_retained)						\
+    OP(decay_dirty)							\
+    OP(decay_muzzy)							\
+    OP(base)								\
+    OP(tcache_list)
+
+typedef enum {
+#define OP(mtx) arena_prof_mutex_##mtx,
+	MUTEX_PROF_ARENA_MUTEXES
+#undef OP
+	mutex_prof_num_arena_mutexes
+} mutex_prof_arena_ind_t;
+
+/*
+ * The forth parameter is a boolean value that is true for derived rate counters
+ * and false for real ones.
+ */
+#define MUTEX_PROF_UINT64_COUNTERS					\
+    OP(num_ops, uint64_t, "n_lock_ops", false, num_ops)					\
+    OP(num_ops_ps, uint64_t, "(#/sec)", true, num_ops)				\
+    OP(num_wait, uint64_t, "n_waiting", false, num_wait)				\
+    OP(num_wait_ps, uint64_t, "(#/sec)", true, num_wait)				\
+    OP(num_spin_acq, uint64_t, "n_spin_acq", false, num_spin_acq)			\
+    OP(num_spin_acq_ps, uint64_t, "(#/sec)", true, num_spin_acq)			\
+    OP(num_owner_switch, uint64_t, "n_owner_switch", false, num_owner_switch)		\
+    OP(num_owner_switch_ps, uint64_t, "(#/sec)", true, num_owner_switch)	\
+    OP(total_wait_time, uint64_t, "total_wait_ns", false, total_wait_time)		\
+    OP(total_wait_time_ps, uint64_t, "(#/sec)", true, total_wait_time)		\
+    OP(max_wait_time, uint64_t, "max_wait_ns", false, max_wait_time)
+
+#define MUTEX_PROF_UINT32_COUNTERS					\
+    OP(max_num_thds, uint32_t, "max_n_thds", false, max_num_thds)
+
+#define MUTEX_PROF_COUNTERS						\
+		MUTEX_PROF_UINT64_COUNTERS				\
+		MUTEX_PROF_UINT32_COUNTERS
+
+#define OP(counter, type, human, derived, base_counter) mutex_counter_##counter,
+
+#define COUNTER_ENUM(counter_list, t)					\
+		typedef enum {						\
+			counter_list					\
+			mutex_prof_num_##t##_counters			\
+		} mutex_prof_##t##_counter_ind_t;
+
+COUNTER_ENUM(MUTEX_PROF_UINT64_COUNTERS, uint64_t)
+COUNTER_ENUM(MUTEX_PROF_UINT32_COUNTERS, uint32_t)
+
+#undef COUNTER_ENUM
+#undef OP
+
+typedef struct {
+	/*
+	 * Counters touched on the slow path, i.e. when there is lock
+	 * contention.  We update them once we have the lock.
+	 */
+	/* Total time (in nano seconds) spent waiting on this mutex. */
+	nstime_t		tot_wait_time;
+	/* Max time (in nano seconds) spent on a single lock operation. */
+	nstime_t		max_wait_time;
+	/* # of times have to wait for this mutex (after spinning). */
+	uint64_t		n_wait_times;
+	/* # of times acquired the mutex through local spinning. */
+	uint64_t		n_spin_acquired;
+	/* Max # of threads waiting for the mutex at the same time. */
+	uint32_t		max_n_thds;
+	/* Current # of threads waiting on the lock.  Atomic synced. */
+	atomic_u32_t		n_waiting_thds;
+
+	/*
+	 * Data touched on the fast path.  These are modified right after we
+	 * grab the lock, so it's placed closest to the end (i.e. right before
+	 * the lock) so that we have a higher chance of them being on the same
+	 * cacheline.
+	 */
+	/* # of times the mutex holder is different than the previous one. */
+	uint64_t		n_owner_switches;
+	/* Previous mutex holder, to facilitate n_owner_switches. */
+	tsdn_t			*prev_owner;
+	/* # of lock() operations in total. */
+	uint64_t		n_lock_ops;
+} mutex_prof_data_t;
+
+#endif /* JEMALLOC_INTERNAL_MUTEX_PROF_H */

deps/jemalloc/include/jemalloc/internal/nstime.h

@@ -0,0 +1,34 @@
+#ifndef JEMALLOC_INTERNAL_NSTIME_H
+#define JEMALLOC_INTERNAL_NSTIME_H
+
+/* Maximum supported number of seconds (~584 years). */
+#define NSTIME_SEC_MAX KQU(18446744072)
+#define NSTIME_ZERO_INITIALIZER {0}
+
+typedef struct {
+	uint64_t ns;
+} nstime_t;
+
+void nstime_init(nstime_t *time, uint64_t ns);
+void nstime_init2(nstime_t *time, uint64_t sec, uint64_t nsec);
+uint64_t nstime_ns(const nstime_t *time);
+uint64_t nstime_sec(const nstime_t *time);
+uint64_t nstime_msec(const nstime_t *time);
+uint64_t nstime_nsec(const nstime_t *time);
+void nstime_copy(nstime_t *time, const nstime_t *source);
+int nstime_compare(const nstime_t *a, const nstime_t *b);
+void nstime_add(nstime_t *time, const nstime_t *addend);
+void nstime_iadd(nstime_t *time, uint64_t addend);
+void nstime_subtract(nstime_t *time, const nstime_t *subtrahend);
+void nstime_isubtract(nstime_t *time, uint64_t subtrahend);
+void nstime_imultiply(nstime_t *time, uint64_t multiplier);
+void nstime_idivide(nstime_t *time, uint64_t divisor);
+uint64_t nstime_divide(const nstime_t *time, const nstime_t *divisor);
+
+typedef bool (nstime_monotonic_t)(void);
+extern nstime_monotonic_t *JET_MUTABLE nstime_monotonic;
+
+typedef bool (nstime_update_t)(nstime_t *);
+extern nstime_update_t *JET_MUTABLE nstime_update;
+
+#endif /* JEMALLOC_INTERNAL_NSTIME_H */

deps/jemalloc/include/jemalloc/internal/pages.h

@@ -0,0 +1,88 @@
+#ifndef JEMALLOC_INTERNAL_PAGES_EXTERNS_H
+#define JEMALLOC_INTERNAL_PAGES_EXTERNS_H
+
+/* Page size.  LG_PAGE is determined by the configure script. */
+#ifdef PAGE_MASK
+#  undef PAGE_MASK
+#endif
+#define PAGE		((size_t)(1U << LG_PAGE))
+#define PAGE_MASK	((size_t)(PAGE - 1))
+/* Return the page base address for the page containing address a. */
+#define PAGE_ADDR2BASE(a)						\
+	((void *)((uintptr_t)(a) & ~PAGE_MASK))
+/* Return the smallest pagesize multiple that is >= s. */
+#define PAGE_CEILING(s)							\
+	(((s) + PAGE_MASK) & ~PAGE_MASK)
+
+/* Huge page size.  LG_HUGEPAGE is determined by the configure script. */
+#define HUGEPAGE	((size_t)(1U << LG_HUGEPAGE))
+#define HUGEPAGE_MASK	((size_t)(HUGEPAGE - 1))
+/* Return the huge page base address for the huge page containing address a. */
+#define HUGEPAGE_ADDR2BASE(a)						\
+	((void *)((uintptr_t)(a) & ~HUGEPAGE_MASK))
+/* Return the smallest pagesize multiple that is >= s. */
+#define HUGEPAGE_CEILING(s)						\
+	(((s) + HUGEPAGE_MASK) & ~HUGEPAGE_MASK)
+
+/* PAGES_CAN_PURGE_LAZY is defined if lazy purging is supported. */
+#if defined(_WIN32) || defined(JEMALLOC_PURGE_MADVISE_FREE)
+#  define PAGES_CAN_PURGE_LAZY
+#endif
+/*
+ * PAGES_CAN_PURGE_FORCED is defined if forced purging is supported.
+ *
+ * The only supported way to hard-purge on Windows is to decommit and then
+ * re-commit, but doing so is racy, and if re-commit fails it's a pain to
+ * propagate the "poisoned" memory state.  Since we typically decommit as the
+ * next step after purging on Windows anyway, there's no point in adding such
+ * complexity.
+ */
+#if !defined(_WIN32) && ((defined(JEMALLOC_PURGE_MADVISE_DONTNEED) && \
+    defined(JEMALLOC_PURGE_MADVISE_DONTNEED_ZEROS)) || \
+    defined(JEMALLOC_MAPS_COALESCE))
+#  define PAGES_CAN_PURGE_FORCED
+#endif
+
+static const bool pages_can_purge_lazy =
+#ifdef PAGES_CAN_PURGE_LAZY
+    true
+#else
+    false
+#endif
+    ;
+static const bool pages_can_purge_forced =
+#ifdef PAGES_CAN_PURGE_FORCED
+    true
+#else
+    false
+#endif
+    ;
+
+typedef enum {
+	thp_mode_default       = 0, /* Do not change hugepage settings. */
+	thp_mode_always        = 1, /* Always set MADV_HUGEPAGE. */
+	thp_mode_never         = 2, /* Always set MADV_NOHUGEPAGE. */
+
+	thp_mode_names_limit   = 3, /* Used for option processing. */
+	thp_mode_not_supported = 3  /* No THP support detected. */
+} thp_mode_t;
+
+#define THP_MODE_DEFAULT thp_mode_default
+extern thp_mode_t opt_thp;
+extern thp_mode_t init_system_thp_mode; /* Initial system wide state. */
+extern const char *thp_mode_names[];
+
+void *pages_map(void *addr, size_t size, size_t alignment, bool *commit);
+void pages_unmap(void *addr, size_t size);
+bool pages_commit(void *addr, size_t size);
+bool pages_decommit(void *addr, size_t size);
+bool pages_purge_lazy(void *addr, size_t size);
+bool pages_purge_forced(void *addr, size_t size);
+bool pages_huge(void *addr, size_t size);
+bool pages_nohuge(void *addr, size_t size);
+bool pages_dontdump(void *addr, size_t size);
+bool pages_dodump(void *addr, size_t size);
+bool pages_boot(void);
+void pages_set_thp_state (void *ptr, size_t size);
+
+#endif /* JEMALLOC_INTERNAL_PAGES_EXTERNS_H */

deps/jemalloc/include/jemalloc/internal/ph.h

@@ -0,0 +1,391 @@
+/*
+ * A Pairing Heap implementation.
+ *
+ * "The Pairing Heap: A New Form of Self-Adjusting Heap"
+ * https://www.cs.cmu.edu/~sleator/papers/pairing-heaps.pdf
+ *
+ * With auxiliary twopass list, described in a follow on paper.
+ *
+ * "Pairing Heaps: Experiments and Analysis"
+ * http://citeseerx.ist.psu.edu/viewdoc/download?doi=10.1.1.106.2988&rep=rep1&type=pdf
+ *
+ *******************************************************************************
+ */
+
+#ifndef PH_H_
+#define PH_H_
+
+/* Node structure. */
+#define phn(a_type)							\
+struct {								\
+	a_type	*phn_prev;						\
+	a_type	*phn_next;						\
+	a_type	*phn_lchild;						\
+}
+
+/* Root structure. */
+#define ph(a_type)							\
+struct {								\
+	a_type	*ph_root;						\
+}
+
+/* Internal utility macros. */
+#define phn_lchild_get(a_type, a_field, a_phn)				\
+	(a_phn->a_field.phn_lchild)
+#define phn_lchild_set(a_type, a_field, a_phn, a_lchild) do {		\
+	a_phn->a_field.phn_lchild = a_lchild;				\
+} while (0)
+
+#define phn_next_get(a_type, a_field, a_phn)				\
+	(a_phn->a_field.phn_next)
+#define phn_prev_set(a_type, a_field, a_phn, a_prev) do {		\
+	a_phn->a_field.phn_prev = a_prev;				\
+} while (0)
+
+#define phn_prev_get(a_type, a_field, a_phn)				\
+	(a_phn->a_field.phn_prev)
+#define phn_next_set(a_type, a_field, a_phn, a_next) do {		\
+	a_phn->a_field.phn_next = a_next;				\
+} while (0)
+
+#define phn_merge_ordered(a_type, a_field, a_phn0, a_phn1, a_cmp) do {	\
+	a_type *phn0child;						\
+									\
+	assert(a_phn0 != NULL);						\
+	assert(a_phn1 != NULL);						\
+	assert(a_cmp(a_phn0, a_phn1) <= 0);				\
+									\
+	phn_prev_set(a_type, a_field, a_phn1, a_phn0);			\
+	phn0child = phn_lchild_get(a_type, a_field, a_phn0);		\
+	phn_next_set(a_type, a_field, a_phn1, phn0child);		\
+	if (phn0child != NULL) {					\
+		phn_prev_set(a_type, a_field, phn0child, a_phn1);	\
+	}								\
+	phn_lchild_set(a_type, a_field, a_phn0, a_phn1);		\
+} while (0)
+
+#define phn_merge(a_type, a_field, a_phn0, a_phn1, a_cmp, r_phn) do {	\
+	if (a_phn0 == NULL) {						\
+		r_phn = a_phn1;						\
+	} else if (a_phn1 == NULL) {					\
+		r_phn = a_phn0;						\
+	} else if (a_cmp(a_phn0, a_phn1) < 0) {				\
+		phn_merge_ordered(a_type, a_field, a_phn0, a_phn1,	\
+		    a_cmp);						\
+		r_phn = a_phn0;						\
+	} else {							\
+		phn_merge_ordered(a_type, a_field, a_phn1, a_phn0,	\
+		    a_cmp);						\
+		r_phn = a_phn1;						\
+	}								\
+} while (0)
+
+#define ph_merge_siblings(a_type, a_field, a_phn, a_cmp, r_phn) do {	\
+	a_type *head = NULL;						\
+	a_type *tail = NULL;						\
+	a_type *phn0 = a_phn;						\
+	a_type *phn1 = phn_next_get(a_type, a_field, phn0);		\
+									\
+	/*								\
+	 * Multipass merge, wherein the first two elements of a FIFO	\
+	 * are repeatedly merged, and each result is appended to the	\
+	 * singly linked FIFO, until the FIFO contains only a single	\
+	 * element.  We start with a sibling list but no reference to	\
+	 * its tail, so we do a single pass over the sibling list to	\
+	 * populate the FIFO.						\
+	 */								\
+	if (phn1 != NULL) {						\
+		a_type *phnrest = phn_next_get(a_type, a_field, phn1);	\
+		if (phnrest != NULL) {					\
+			phn_prev_set(a_type, a_field, phnrest, NULL);	\
+		}							\
+		phn_prev_set(a_type, a_field, phn0, NULL);		\
+		phn_next_set(a_type, a_field, phn0, NULL);		\
+		phn_prev_set(a_type, a_field, phn1, NULL);		\
+		phn_next_set(a_type, a_field, phn1, NULL);		\
+		phn_merge(a_type, a_field, phn0, phn1, a_cmp, phn0);	\
+		head = tail = phn0;					\
+		phn0 = phnrest;						\
+		while (phn0 != NULL) {					\
+			phn1 = phn_next_get(a_type, a_field, phn0);	\
+			if (phn1 != NULL) {				\
+				phnrest = phn_next_get(a_type, a_field,	\
+				    phn1);				\
+				if (phnrest != NULL) {			\
+					phn_prev_set(a_type, a_field,	\
+					    phnrest, NULL);		\
+				}					\
+				phn_prev_set(a_type, a_field, phn0,	\
+				    NULL);				\
+				phn_next_set(a_type, a_field, phn0,	\
+				    NULL);				\
+				phn_prev_set(a_type, a_field, phn1,	\
+				    NULL);				\
+				phn_next_set(a_type, a_field, phn1,	\
+				    NULL);				\
+				phn_merge(a_type, a_field, phn0, phn1,	\
+				    a_cmp, phn0);			\
+				phn_next_set(a_type, a_field, tail,	\
+				    phn0);				\
+				tail = phn0;				\
+				phn0 = phnrest;				\
+			} else {					\
+				phn_next_set(a_type, a_field, tail,	\
+				    phn0);				\
+				tail = phn0;				\
+				phn0 = NULL;				\
+			}						\
+		}							\
+		phn0 = head;						\
+		phn1 = phn_next_get(a_type, a_field, phn0);		\
+		if (phn1 != NULL) {					\
+			while (true) {					\
+				head = phn_next_get(a_type, a_field,	\
+				    phn1);				\
+				assert(phn_prev_get(a_type, a_field,	\
+				    phn0) == NULL);			\
+				phn_next_set(a_type, a_field, phn0,	\
+				    NULL);				\
+				assert(phn_prev_get(a_type, a_field,	\
+				    phn1) == NULL);			\
+				phn_next_set(a_type, a_field, phn1,	\
+				    NULL);				\
+				phn_merge(a_type, a_field, phn0, phn1,	\
+				    a_cmp, phn0);			\
+				if (head == NULL) {			\
+					break;				\
+				}					\
+				phn_next_set(a_type, a_field, tail,	\
+				    phn0);				\
+				tail = phn0;				\
+				phn0 = head;				\
+				phn1 = phn_next_get(a_type, a_field,	\
+				    phn0);				\
+			}						\
+		}							\
+	}								\
+	r_phn = phn0;							\
+} while (0)
+
+#define ph_merge_aux(a_type, a_field, a_ph, a_cmp) do {			\
+	a_type *phn = phn_next_get(a_type, a_field, a_ph->ph_root);	\
+	if (phn != NULL) {						\
+		phn_prev_set(a_type, a_field, a_ph->ph_root, NULL);	\
+		phn_next_set(a_type, a_field, a_ph->ph_root, NULL);	\
+		phn_prev_set(a_type, a_field, phn, NULL);		\
+		ph_merge_siblings(a_type, a_field, phn, a_cmp, phn);	\
+		assert(phn_next_get(a_type, a_field, phn) == NULL);	\
+		phn_merge(a_type, a_field, a_ph->ph_root, phn, a_cmp,	\
+		    a_ph->ph_root);					\
+	}								\
+} while (0)
+
+#define ph_merge_children(a_type, a_field, a_phn, a_cmp, r_phn) do {	\
+	a_type *lchild = phn_lchild_get(a_type, a_field, a_phn);	\
+	if (lchild == NULL) {						\
+		r_phn = NULL;						\
+	} else {							\
+		ph_merge_siblings(a_type, a_field, lchild, a_cmp,	\
+		    r_phn);						\
+	}								\
+} while (0)
+
+/*
+ * The ph_proto() macro generates function prototypes that correspond to the
+ * functions generated by an equivalently parameterized call to ph_gen().
+ */
+#define ph_proto(a_attr, a_prefix, a_ph_type, a_type)			\
+a_attr void	a_prefix##new(a_ph_type *ph);				\
+a_attr bool	a_prefix##empty(a_ph_type *ph);				\
+a_attr a_type	*a_prefix##first(a_ph_type *ph);			\
+a_attr a_type	*a_prefix##any(a_ph_type *ph);				\
+a_attr void	a_prefix##insert(a_ph_type *ph, a_type *phn);		\
+a_attr a_type	*a_prefix##remove_first(a_ph_type *ph);			\
+a_attr a_type	*a_prefix##remove_any(a_ph_type *ph);			\
+a_attr void	a_prefix##remove(a_ph_type *ph, a_type *phn);
+
+/*
+ * The ph_gen() macro generates a type-specific pairing heap implementation,
+ * based on the above cpp macros.
+ */
+#define ph_gen(a_attr, a_prefix, a_ph_type, a_type, a_field, a_cmp)	\
+a_attr void								\
+a_prefix##new(a_ph_type *ph) {						\
+	memset(ph, 0, sizeof(ph(a_type)));				\
+}									\
+a_attr bool								\
+a_prefix##empty(a_ph_type *ph) {					\
+	return (ph->ph_root == NULL);					\
+}									\
+a_attr a_type *								\
+a_prefix##first(a_ph_type *ph) {					\
+	if (ph->ph_root == NULL) {					\
+		return NULL;						\
+	}								\
+	ph_merge_aux(a_type, a_field, ph, a_cmp);			\
+	return ph->ph_root;						\
+}									\
+a_attr a_type *								\
+a_prefix##any(a_ph_type *ph) {						\
+	if (ph->ph_root == NULL) {					\
+		return NULL;						\
+	}								\
+	a_type *aux = phn_next_get(a_type, a_field, ph->ph_root);	\
+	if (aux != NULL) {						\
+		return aux;						\
+	}								\
+	return ph->ph_root;						\
+}									\
+a_attr void								\
+a_prefix##insert(a_ph_type *ph, a_type *phn) {				\
+	memset(&phn->a_field, 0, sizeof(phn(a_type)));			\
+									\
+	/*								\
+	 * Treat the root as an aux list during insertion, and lazily	\
+	 * merge during a_prefix##remove_first().  For elements that	\
+	 * are inserted, then removed via a_prefix##remove() before the	\
+	 * aux list is ever processed, this makes insert/remove		\
+	 * constant-time, whereas eager merging would make insert	\
+	 * O(log n).							\
+	 */								\
+	if (ph->ph_root == NULL) {					\
+		ph->ph_root = phn;					\
+	} else {							\
+		phn_next_set(a_type, a_field, phn, phn_next_get(a_type,	\
+		    a_field, ph->ph_root));				\
+		if (phn_next_get(a_type, a_field, ph->ph_root) !=	\
+		    NULL) {						\
+			phn_prev_set(a_type, a_field,			\
+			    phn_next_get(a_type, a_field, ph->ph_root),	\
+			    phn);					\
+		}							\
+		phn_prev_set(a_type, a_field, phn, ph->ph_root);	\
+		phn_next_set(a_type, a_field, ph->ph_root, phn);	\
+	}								\
+}									\
+a_attr a_type *								\
+a_prefix##remove_first(a_ph_type *ph) {					\
+	a_type *ret;							\
+									\
+	if (ph->ph_root == NULL) {					\
+		return NULL;						\
+	}								\
+	ph_merge_aux(a_type, a_field, ph, a_cmp);			\
+									\
+	ret = ph->ph_root;						\
+									\
+	ph_merge_children(a_type, a_field, ph->ph_root, a_cmp,		\
+	    ph->ph_root);						\
+									\
+	return ret;							\
+}									\
+a_attr a_type *								\
+a_prefix##remove_any(a_ph_type *ph) {					\
+	/*								\
+	 * Remove the most recently inserted aux list element, or the	\
+	 * root if the aux list is empty.  This has the effect of	\
+	 * behaving as a LIFO (and insertion/removal is therefore	\
+	 * constant-time) if a_prefix##[remove_]first() are never	\
+	 * called.							\
+	 */								\
+	if (ph->ph_root == NULL) {					\
+		return NULL;						\
+	}								\
+	a_type *ret = phn_next_get(a_type, a_field, ph->ph_root);	\
+	if (ret != NULL) {						\
+		a_type *aux = phn_next_get(a_type, a_field, ret);	\
+		phn_next_set(a_type, a_field, ph->ph_root, aux);	\
+		if (aux != NULL) {					\
+			phn_prev_set(a_type, a_field, aux,		\
+			    ph->ph_root);				\
+		}							\
+		return ret;						\
+	}								\
+	ret = ph->ph_root;						\
+	ph_merge_children(a_type, a_field, ph->ph_root, a_cmp,		\
+	    ph->ph_root);						\
+	return ret;							\
+}									\
+a_attr void								\
+a_prefix##remove(a_ph_type *ph, a_type *phn) {				\
+	a_type *replace, *parent;					\
+									\
+	if (ph->ph_root == phn) {					\
+		/*							\
+		 * We can delete from aux list without merging it, but	\
+		 * we need to merge if we are dealing with the root	\
+		 * node and it has children.				\
+		 */							\
+		if (phn_lchild_get(a_type, a_field, phn) == NULL) {	\
+			ph->ph_root = phn_next_get(a_type, a_field,	\
+			    phn);					\
+			if (ph->ph_root != NULL) {			\
+				phn_prev_set(a_type, a_field,		\
+				    ph->ph_root, NULL);			\
+			}						\
+			return;						\
+		}							\
+		ph_merge_aux(a_type, a_field, ph, a_cmp);		\
+		if (ph->ph_root == phn) {				\
+			ph_merge_children(a_type, a_field, ph->ph_root,	\
+			    a_cmp, ph->ph_root);			\
+			return;						\
+		}							\
+	}								\
+									\
+	/* Get parent (if phn is leftmost child) before mutating. */	\
+	if ((parent = phn_prev_get(a_type, a_field, phn)) != NULL) {	\
+		if (phn_lchild_get(a_type, a_field, parent) != phn) {	\
+			parent = NULL;					\
+		}							\
+	}								\
+	/* Find a possible replacement node, and link to parent. */	\
+	ph_merge_children(a_type, a_field, phn, a_cmp, replace);	\
+	/* Set next/prev for sibling linked list. */			\
+	if (replace != NULL) {						\
+		if (parent != NULL) {					\
+			phn_prev_set(a_type, a_field, replace, parent);	\
+			phn_lchild_set(a_type, a_field, parent,		\
+			    replace);					\
+		} else {						\
+			phn_prev_set(a_type, a_field, replace,		\
+			    phn_prev_get(a_type, a_field, phn));	\
+			if (phn_prev_get(a_type, a_field, phn) !=	\
+			    NULL) {					\
+				phn_next_set(a_type, a_field,		\
+				    phn_prev_get(a_type, a_field, phn),	\
+				    replace);				\
+			}						\
+		}							\
+		phn_next_set(a_type, a_field, replace,			\
+		    phn_next_get(a_type, a_field, phn));		\
+		if (phn_next_get(a_type, a_field, phn) != NULL) {	\
+			phn_prev_set(a_type, a_field,			\
+			    phn_next_get(a_type, a_field, phn),		\
+			    replace);					\
+		}							\
+	} else {							\
+		if (parent != NULL) {					\
+			a_type *next = phn_next_get(a_type, a_field,	\
+			    phn);					\
+			phn_lchild_set(a_type, a_field, parent, next);	\
+			if (next != NULL) {				\
+				phn_prev_set(a_type, a_field, next,	\
+				    parent);				\
+			}						\
+		} else {						\
+			assert(phn_prev_get(a_type, a_field, phn) !=	\
+			    NULL);					\
+			phn_next_set(a_type, a_field,			\
+			    phn_prev_get(a_type, a_field, phn),		\
+			    phn_next_get(a_type, a_field, phn));	\
+		}							\
+		if (phn_next_get(a_type, a_field, phn) != NULL) {	\
+			phn_prev_set(a_type, a_field,			\
+			    phn_next_get(a_type, a_field, phn),		\
+			    phn_prev_get(a_type, a_field, phn));	\
+		}							\
+	}								\
+}
+
+#endif /* PH_H_ */

deps/jemalloc/include/jemalloc/internal/private_namespace.h

@@ -0,0 +1,422 @@
+#define a0dalloc JEMALLOC_N(a0dalloc)
+#define a0malloc JEMALLOC_N(a0malloc)
+#define arena_choose_hard JEMALLOC_N(arena_choose_hard)
+#define arena_cleanup JEMALLOC_N(arena_cleanup)
+#define arena_init JEMALLOC_N(arena_init)
+#define arena_migrate JEMALLOC_N(arena_migrate)
+#define arenas JEMALLOC_N(arenas)
+#define arena_set JEMALLOC_N(arena_set)
+#define arenas_lock JEMALLOC_N(arenas_lock)
+#define arenas_tdata_cleanup JEMALLOC_N(arenas_tdata_cleanup)
+#define arena_tdata_get_hard JEMALLOC_N(arena_tdata_get_hard)
+#define bootstrap_calloc JEMALLOC_N(bootstrap_calloc)
+#define bootstrap_free JEMALLOC_N(bootstrap_free)
+#define bootstrap_malloc JEMALLOC_N(bootstrap_malloc)
+#define free_default JEMALLOC_N(free_default)
+#define iarena_cleanup JEMALLOC_N(iarena_cleanup)
+#define jemalloc_postfork_child JEMALLOC_N(jemalloc_postfork_child)
+#define jemalloc_postfork_parent JEMALLOC_N(jemalloc_postfork_parent)
+#define jemalloc_prefork JEMALLOC_N(jemalloc_prefork)
+#define je_sdallocx_noflags JEMALLOC_N(je_sdallocx_noflags)
+#define malloc_default JEMALLOC_N(malloc_default)
+#define malloc_initialized JEMALLOC_N(malloc_initialized)
+#define malloc_slow JEMALLOC_N(malloc_slow)
+#define manual_arena_base JEMALLOC_N(manual_arena_base)
+#define narenas_auto JEMALLOC_N(narenas_auto)
+#define narenas_total_get JEMALLOC_N(narenas_total_get)
+#define ncpus JEMALLOC_N(ncpus)
+#define opt_abort JEMALLOC_N(opt_abort)
+#define opt_abort_conf JEMALLOC_N(opt_abort_conf)
+#define opt_confirm_conf JEMALLOC_N(opt_confirm_conf)
+#define opt_junk JEMALLOC_N(opt_junk)
+#define opt_junk_alloc JEMALLOC_N(opt_junk_alloc)
+#define opt_junk_free JEMALLOC_N(opt_junk_free)
+#define opt_narenas JEMALLOC_N(opt_narenas)
+#define opt_utrace JEMALLOC_N(opt_utrace)
+#define opt_xmalloc JEMALLOC_N(opt_xmalloc)
+#define opt_zero JEMALLOC_N(opt_zero)
+#define sdallocx_default JEMALLOC_N(sdallocx_default)
+#define arena_alloc_junk_small JEMALLOC_N(arena_alloc_junk_small)
+#define arena_basic_stats_merge JEMALLOC_N(arena_basic_stats_merge)
+#define arena_bin_choose_lock JEMALLOC_N(arena_bin_choose_lock)
+#define arena_boot JEMALLOC_N(arena_boot)
+#define arena_choose_huge JEMALLOC_N(arena_choose_huge)
+#define arena_dalloc_bin_junked_locked JEMALLOC_N(arena_dalloc_bin_junked_locked)
+#define arena_dalloc_junk_small JEMALLOC_N(arena_dalloc_junk_small)
+#define arena_dalloc_promoted JEMALLOC_N(arena_dalloc_promoted)
+#define arena_dalloc_small JEMALLOC_N(arena_dalloc_small)
+#define arena_decay JEMALLOC_N(arena_decay)
+#define arena_destroy JEMALLOC_N(arena_destroy)
+#define arena_dirty_decay_ms_default_get JEMALLOC_N(arena_dirty_decay_ms_default_get)
+#define arena_dirty_decay_ms_default_set JEMALLOC_N(arena_dirty_decay_ms_default_set)
+#define arena_dirty_decay_ms_get JEMALLOC_N(arena_dirty_decay_ms_get)
+#define arena_dirty_decay_ms_set JEMALLOC_N(arena_dirty_decay_ms_set)
+#define arena_dss_prec_get JEMALLOC_N(arena_dss_prec_get)
+#define arena_dss_prec_set JEMALLOC_N(arena_dss_prec_set)
+#define arena_extent_alloc_large JEMALLOC_N(arena_extent_alloc_large)
+#define arena_extent_dalloc_large_prep JEMALLOC_N(arena_extent_dalloc_large_prep)
+#define arena_extent_ralloc_large_expand JEMALLOC_N(arena_extent_ralloc_large_expand)
+#define arena_extent_ralloc_large_shrink JEMALLOC_N(arena_extent_ralloc_large_shrink)
+#define arena_extents_dirty_dalloc JEMALLOC_N(arena_extents_dirty_dalloc)
+#define arena_extent_sn_next JEMALLOC_N(arena_extent_sn_next)
+#define arena_init_huge JEMALLOC_N(arena_init_huge)
+#define arena_is_huge JEMALLOC_N(arena_is_huge)
+#define arena_malloc_hard JEMALLOC_N(arena_malloc_hard)
+#define arena_muzzy_decay_ms_default_get JEMALLOC_N(arena_muzzy_decay_ms_default_get)
+#define arena_muzzy_decay_ms_default_set JEMALLOC_N(arena_muzzy_decay_ms_default_set)
+#define arena_muzzy_decay_ms_get JEMALLOC_N(arena_muzzy_decay_ms_get)
+#define arena_muzzy_decay_ms_set JEMALLOC_N(arena_muzzy_decay_ms_set)
+#define arena_new JEMALLOC_N(arena_new)
+#define arena_nthreads_dec JEMALLOC_N(arena_nthreads_dec)
+#define arena_nthreads_get JEMALLOC_N(arena_nthreads_get)
+#define arena_nthreads_inc JEMALLOC_N(arena_nthreads_inc)
+#define arena_palloc JEMALLOC_N(arena_palloc)
+#define arena_postfork_child JEMALLOC_N(arena_postfork_child)
+#define arena_postfork_parent JEMALLOC_N(arena_postfork_parent)
+#define arena_prefork0 JEMALLOC_N(arena_prefork0)
+#define arena_prefork1 JEMALLOC_N(arena_prefork1)
+#define arena_prefork2 JEMALLOC_N(arena_prefork2)
+#define arena_prefork3 JEMALLOC_N(arena_prefork3)
+#define arena_prefork4 JEMALLOC_N(arena_prefork4)
+#define arena_prefork5 JEMALLOC_N(arena_prefork5)
+#define arena_prefork6 JEMALLOC_N(arena_prefork6)
+#define arena_prefork7 JEMALLOC_N(arena_prefork7)
+#define arena_prof_promote JEMALLOC_N(arena_prof_promote)
+#define arena_ralloc JEMALLOC_N(arena_ralloc)
+#define arena_ralloc_no_move JEMALLOC_N(arena_ralloc_no_move)
+#define arena_reset JEMALLOC_N(arena_reset)
+#define arena_retain_grow_limit_get_set JEMALLOC_N(arena_retain_grow_limit_get_set)
+#define arena_stats_merge JEMALLOC_N(arena_stats_merge)
+#define arena_tcache_fill_small JEMALLOC_N(arena_tcache_fill_small)
+#define h_steps JEMALLOC_N(h_steps)
+#define opt_dirty_decay_ms JEMALLOC_N(opt_dirty_decay_ms)
+#define opt_muzzy_decay_ms JEMALLOC_N(opt_muzzy_decay_ms)
+#define opt_oversize_threshold JEMALLOC_N(opt_oversize_threshold)
+#define opt_percpu_arena JEMALLOC_N(opt_percpu_arena)
+#define oversize_threshold JEMALLOC_N(oversize_threshold)
+#define percpu_arena_mode_names JEMALLOC_N(percpu_arena_mode_names)
+#define background_thread_boot0 JEMALLOC_N(background_thread_boot0)
+#define background_thread_boot1 JEMALLOC_N(background_thread_boot1)
+#define background_thread_create JEMALLOC_N(background_thread_create)
+#define background_thread_ctl_init JEMALLOC_N(background_thread_ctl_init)
+#define background_thread_enabled_state JEMALLOC_N(background_thread_enabled_state)
+#define background_thread_info JEMALLOC_N(background_thread_info)
+#define background_thread_interval_check JEMALLOC_N(background_thread_interval_check)
+#define background_thread_lock JEMALLOC_N(background_thread_lock)
+#define background_thread_postfork_child JEMALLOC_N(background_thread_postfork_child)
+#define background_thread_postfork_parent JEMALLOC_N(background_thread_postfork_parent)
+#define background_thread_prefork0 JEMALLOC_N(background_thread_prefork0)
+#define background_thread_prefork1 JEMALLOC_N(background_thread_prefork1)
+#define background_threads_disable JEMALLOC_N(background_threads_disable)
+#define background_threads_enable JEMALLOC_N(background_threads_enable)
+#define background_thread_stats_read JEMALLOC_N(background_thread_stats_read)
+#define max_background_threads JEMALLOC_N(max_background_threads)
+#define n_background_threads JEMALLOC_N(n_background_threads)
+#define opt_background_thread JEMALLOC_N(opt_background_thread)
+#define opt_max_background_threads JEMALLOC_N(opt_max_background_threads)
+#define pthread_create_wrapper JEMALLOC_N(pthread_create_wrapper)
+#define b0get JEMALLOC_N(b0get)
+#define base_alloc JEMALLOC_N(base_alloc)
+#define base_alloc_extent JEMALLOC_N(base_alloc_extent)
+#define base_boot JEMALLOC_N(base_boot)
+#define base_delete JEMALLOC_N(base_delete)
+#define base_extent_hooks_get JEMALLOC_N(base_extent_hooks_get)
+#define base_extent_hooks_set JEMALLOC_N(base_extent_hooks_set)
+#define base_new JEMALLOC_N(base_new)
+#define base_postfork_child JEMALLOC_N(base_postfork_child)
+#define base_postfork_parent JEMALLOC_N(base_postfork_parent)
+#define base_prefork JEMALLOC_N(base_prefork)
+#define base_stats_get JEMALLOC_N(base_stats_get)
+#define metadata_thp_mode_names JEMALLOC_N(metadata_thp_mode_names)
+#define opt_metadata_thp JEMALLOC_N(opt_metadata_thp)
+#define bin_boot JEMALLOC_N(bin_boot)
+#define bin_infos JEMALLOC_N(bin_infos)
+#define bin_init JEMALLOC_N(bin_init)
+#define bin_postfork_child JEMALLOC_N(bin_postfork_child)
+#define bin_postfork_parent JEMALLOC_N(bin_postfork_parent)
+#define bin_prefork JEMALLOC_N(bin_prefork)
+#define bin_shard_sizes_boot JEMALLOC_N(bin_shard_sizes_boot)
+#define bin_update_shard_size JEMALLOC_N(bin_update_shard_size)
+#define bitmap_info_init JEMALLOC_N(bitmap_info_init)
+#define bitmap_init JEMALLOC_N(bitmap_init)
+#define bitmap_size JEMALLOC_N(bitmap_size)
+#define ckh_count JEMALLOC_N(ckh_count)
+#define ckh_delete JEMALLOC_N(ckh_delete)
+#define ckh_insert JEMALLOC_N(ckh_insert)
+#define ckh_iter JEMALLOC_N(ckh_iter)
+#define ckh_new JEMALLOC_N(ckh_new)
+#define ckh_pointer_hash JEMALLOC_N(ckh_pointer_hash)
+#define ckh_pointer_keycomp JEMALLOC_N(ckh_pointer_keycomp)
+#define ckh_remove JEMALLOC_N(ckh_remove)
+#define ckh_search JEMALLOC_N(ckh_search)
+#define ckh_string_hash JEMALLOC_N(ckh_string_hash)
+#define ckh_string_keycomp JEMALLOC_N(ckh_string_keycomp)
+#define ctl_boot JEMALLOC_N(ctl_boot)
+#define ctl_bymib JEMALLOC_N(ctl_bymib)
+#define ctl_byname JEMALLOC_N(ctl_byname)
+#define ctl_nametomib JEMALLOC_N(ctl_nametomib)
+#define ctl_postfork_child JEMALLOC_N(ctl_postfork_child)
+#define ctl_postfork_parent JEMALLOC_N(ctl_postfork_parent)
+#define ctl_prefork JEMALLOC_N(ctl_prefork)
+#define div_init JEMALLOC_N(div_init)
+#define extent_alloc JEMALLOC_N(extent_alloc)
+#define extent_alloc_wrapper JEMALLOC_N(extent_alloc_wrapper)
+#define extent_avail_any JEMALLOC_N(extent_avail_any)
+#define extent_avail_empty JEMALLOC_N(extent_avail_empty)
+#define extent_avail_first JEMALLOC_N(extent_avail_first)
+#define extent_avail_insert JEMALLOC_N(extent_avail_insert)
+#define extent_avail_new JEMALLOC_N(extent_avail_new)
+#define extent_avail_remove JEMALLOC_N(extent_avail_remove)
+#define extent_avail_remove_any JEMALLOC_N(extent_avail_remove_any)
+#define extent_avail_remove_first JEMALLOC_N(extent_avail_remove_first)
+#define extent_boot JEMALLOC_N(extent_boot)
+#define extent_commit_wrapper JEMALLOC_N(extent_commit_wrapper)
+#define extent_dalloc JEMALLOC_N(extent_dalloc)
+#define extent_dalloc_gap JEMALLOC_N(extent_dalloc_gap)
+#define extent_dalloc_wrapper JEMALLOC_N(extent_dalloc_wrapper)
+#define extent_decommit_wrapper JEMALLOC_N(extent_decommit_wrapper)
+#define extent_destroy_wrapper JEMALLOC_N(extent_destroy_wrapper)
+#define extent_heap_any JEMALLOC_N(extent_heap_any)
+#define extent_heap_empty JEMALLOC_N(extent_heap_empty)
+#define extent_heap_first JEMALLOC_N(extent_heap_first)
+#define extent_heap_insert JEMALLOC_N(extent_heap_insert)
+#define extent_heap_new JEMALLOC_N(extent_heap_new)
+#define extent_heap_remove JEMALLOC_N(extent_heap_remove)
+#define extent_heap_remove_any JEMALLOC_N(extent_heap_remove_any)
+#define extent_heap_remove_first JEMALLOC_N(extent_heap_remove_first)
+#define extent_hooks_default JEMALLOC_N(extent_hooks_default)
+#define extent_hooks_get JEMALLOC_N(extent_hooks_get)
+#define extent_hooks_set JEMALLOC_N(extent_hooks_set)
+#define extent_merge_wrapper JEMALLOC_N(extent_merge_wrapper)
+#define extent_mutex_pool JEMALLOC_N(extent_mutex_pool)
+#define extent_purge_forced_wrapper JEMALLOC_N(extent_purge_forced_wrapper)
+#define extent_purge_lazy_wrapper JEMALLOC_N(extent_purge_lazy_wrapper)
+#define extents_alloc JEMALLOC_N(extents_alloc)
+#define extents_dalloc JEMALLOC_N(extents_dalloc)
+#define extents_evict JEMALLOC_N(extents_evict)
+#define extents_init JEMALLOC_N(extents_init)
+#define extents_nbytes_get JEMALLOC_N(extents_nbytes_get)
+#define extents_nextents_get JEMALLOC_N(extents_nextents_get)
+#define extents_npages_get JEMALLOC_N(extents_npages_get)
+#define extent_split_wrapper JEMALLOC_N(extent_split_wrapper)
+#define extents_postfork_child JEMALLOC_N(extents_postfork_child)
+#define extents_postfork_parent JEMALLOC_N(extents_postfork_parent)
+#define extents_prefork JEMALLOC_N(extents_prefork)
+#define extents_rtree JEMALLOC_N(extents_rtree)
+#define extents_state_get JEMALLOC_N(extents_state_get)
+#define extent_util_stats_get JEMALLOC_N(extent_util_stats_get)
+#define extent_util_stats_verbose_get JEMALLOC_N(extent_util_stats_verbose_get)
+#define opt_lg_extent_max_active_fit JEMALLOC_N(opt_lg_extent_max_active_fit)
+#define dss_prec_names JEMALLOC_N(dss_prec_names)
+#define extent_alloc_dss JEMALLOC_N(extent_alloc_dss)
+#define extent_dss_boot JEMALLOC_N(extent_dss_boot)
+#define extent_dss_mergeable JEMALLOC_N(extent_dss_mergeable)
+#define extent_dss_prec_get JEMALLOC_N(extent_dss_prec_get)
+#define extent_dss_prec_set JEMALLOC_N(extent_dss_prec_set)
+#define extent_in_dss JEMALLOC_N(extent_in_dss)
+#define opt_dss JEMALLOC_N(opt_dss)
+#define extent_alloc_mmap JEMALLOC_N(extent_alloc_mmap)
+#define extent_dalloc_mmap JEMALLOC_N(extent_dalloc_mmap)
+#define opt_retain JEMALLOC_N(opt_retain)
+#define hook_boot JEMALLOC_N(hook_boot)
+#define hook_install JEMALLOC_N(hook_install)
+#define hook_invoke_alloc JEMALLOC_N(hook_invoke_alloc)
+#define hook_invoke_dalloc JEMALLOC_N(hook_invoke_dalloc)
+#define hook_invoke_expand JEMALLOC_N(hook_invoke_expand)
+#define hook_remove JEMALLOC_N(hook_remove)
+#define large_dalloc JEMALLOC_N(large_dalloc)
+#define large_dalloc_finish JEMALLOC_N(large_dalloc_finish)
+#define large_dalloc_junk JEMALLOC_N(large_dalloc_junk)
+#define large_dalloc_maybe_junk JEMALLOC_N(large_dalloc_maybe_junk)
+#define large_dalloc_prep_junked_locked JEMALLOC_N(large_dalloc_prep_junked_locked)
+#define large_malloc JEMALLOC_N(large_malloc)
+#define large_palloc JEMALLOC_N(large_palloc)
+#define large_prof_alloc_time_get JEMALLOC_N(large_prof_alloc_time_get)
+#define large_prof_alloc_time_set JEMALLOC_N(large_prof_alloc_time_set)
+#define large_prof_tctx_get JEMALLOC_N(large_prof_tctx_get)
+#define large_prof_tctx_reset JEMALLOC_N(large_prof_tctx_reset)
+#define large_prof_tctx_set JEMALLOC_N(large_prof_tctx_set)
+#define large_ralloc JEMALLOC_N(large_ralloc)
+#define large_ralloc_no_move JEMALLOC_N(large_ralloc_no_move)
+#define large_salloc JEMALLOC_N(large_salloc)
+#define log_init_done JEMALLOC_N(log_init_done)
+#define log_var_names JEMALLOC_N(log_var_names)
+#define log_var_update_state JEMALLOC_N(log_var_update_state)
+#define buferror JEMALLOC_N(buferror)
+#define malloc_cprintf JEMALLOC_N(malloc_cprintf)
+#define malloc_printf JEMALLOC_N(malloc_printf)
+#define malloc_snprintf JEMALLOC_N(malloc_snprintf)
+#define malloc_strtoumax JEMALLOC_N(malloc_strtoumax)
+#define malloc_vcprintf JEMALLOC_N(malloc_vcprintf)
+#define malloc_vsnprintf JEMALLOC_N(malloc_vsnprintf)
+#define malloc_write JEMALLOC_N(malloc_write)
+#define malloc_mutex_boot JEMALLOC_N(malloc_mutex_boot)
+#define malloc_mutex_init JEMALLOC_N(malloc_mutex_init)
+#define malloc_mutex_lock_slow JEMALLOC_N(malloc_mutex_lock_slow)
+#define malloc_mutex_postfork_child JEMALLOC_N(malloc_mutex_postfork_child)
+#define malloc_mutex_postfork_parent JEMALLOC_N(malloc_mutex_postfork_parent)
+#define malloc_mutex_prefork JEMALLOC_N(malloc_mutex_prefork)
+#define malloc_mutex_prof_data_reset JEMALLOC_N(malloc_mutex_prof_data_reset)
+#define mutex_pool_init JEMALLOC_N(mutex_pool_init)
+#define nstime_add JEMALLOC_N(nstime_add)
+#define nstime_compare JEMALLOC_N(nstime_compare)
+#define nstime_copy JEMALLOC_N(nstime_copy)
+#define nstime_divide JEMALLOC_N(nstime_divide)
+#define nstime_iadd JEMALLOC_N(nstime_iadd)
+#define nstime_idivide JEMALLOC_N(nstime_idivide)
+#define nstime_imultiply JEMALLOC_N(nstime_imultiply)
+#define nstime_init JEMALLOC_N(nstime_init)
+#define nstime_init2 JEMALLOC_N(nstime_init2)
+#define nstime_isubtract JEMALLOC_N(nstime_isubtract)
+#define nstime_monotonic JEMALLOC_N(nstime_monotonic)
+#define nstime_msec JEMALLOC_N(nstime_msec)
+#define nstime_ns JEMALLOC_N(nstime_ns)
+#define nstime_nsec JEMALLOC_N(nstime_nsec)
+#define nstime_sec JEMALLOC_N(nstime_sec)
+#define nstime_subtract JEMALLOC_N(nstime_subtract)
+#define nstime_update JEMALLOC_N(nstime_update)
+#define init_system_thp_mode JEMALLOC_N(init_system_thp_mode)
+#define opt_thp JEMALLOC_N(opt_thp)
+#define pages_boot JEMALLOC_N(pages_boot)
+#define pages_commit JEMALLOC_N(pages_commit)
+#define pages_decommit JEMALLOC_N(pages_decommit)
+#define pages_dodump JEMALLOC_N(pages_dodump)
+#define pages_dontdump JEMALLOC_N(pages_dontdump)
+#define pages_huge JEMALLOC_N(pages_huge)
+#define pages_map JEMALLOC_N(pages_map)
+#define pages_nohuge JEMALLOC_N(pages_nohuge)
+#define pages_purge_forced JEMALLOC_N(pages_purge_forced)
+#define pages_purge_lazy JEMALLOC_N(pages_purge_lazy)
+#define pages_set_thp_state JEMALLOC_N(pages_set_thp_state)
+#define pages_unmap JEMALLOC_N(pages_unmap)
+#define thp_mode_names JEMALLOC_N(thp_mode_names)
+#define bt2gctx_mtx JEMALLOC_N(bt2gctx_mtx)
+#define bt_init JEMALLOC_N(bt_init)
+#define lg_prof_sample JEMALLOC_N(lg_prof_sample)
+#define opt_lg_prof_interval JEMALLOC_N(opt_lg_prof_interval)
+#define opt_lg_prof_sample JEMALLOC_N(opt_lg_prof_sample)
+#define opt_prof JEMALLOC_N(opt_prof)
+#define opt_prof_accum JEMALLOC_N(opt_prof_accum)
+#define opt_prof_active JEMALLOC_N(opt_prof_active)
+#define opt_prof_final JEMALLOC_N(opt_prof_final)
+#define opt_prof_gdump JEMALLOC_N(opt_prof_gdump)
+#define opt_prof_leak JEMALLOC_N(opt_prof_leak)
+#define opt_prof_log JEMALLOC_N(opt_prof_log)
+#define opt_prof_prefix JEMALLOC_N(opt_prof_prefix)
+#define opt_prof_thread_active_init JEMALLOC_N(opt_prof_thread_active_init)
+#define prof_accum_init JEMALLOC_N(prof_accum_init)
+#define prof_active JEMALLOC_N(prof_active)
+#define prof_active_get JEMALLOC_N(prof_active_get)
+#define prof_active_set JEMALLOC_N(prof_active_set)
+#define prof_alloc_rollback JEMALLOC_N(prof_alloc_rollback)
+#define prof_backtrace JEMALLOC_N(prof_backtrace)
+#define prof_boot0 JEMALLOC_N(prof_boot0)
+#define prof_boot1 JEMALLOC_N(prof_boot1)
+#define prof_boot2 JEMALLOC_N(prof_boot2)
+#define prof_dump_header JEMALLOC_N(prof_dump_header)
+#define prof_dump_open JEMALLOC_N(prof_dump_open)
+#define prof_free_sampled_object JEMALLOC_N(prof_free_sampled_object)
+#define prof_gdump JEMALLOC_N(prof_gdump)
+#define prof_gdump_get JEMALLOC_N(prof_gdump_get)
+#define prof_gdump_set JEMALLOC_N(prof_gdump_set)
+#define prof_gdump_val JEMALLOC_N(prof_gdump_val)
+#define prof_idump JEMALLOC_N(prof_idump)
+#define prof_interval JEMALLOC_N(prof_interval)
+#define prof_logging_state JEMALLOC_N(prof_logging_state)
+#define prof_log_start JEMALLOC_N(prof_log_start)
+#define prof_log_stop JEMALLOC_N(prof_log_stop)
+#define prof_lookup JEMALLOC_N(prof_lookup)
+#define prof_malloc_sample_object JEMALLOC_N(prof_malloc_sample_object)
+#define prof_mdump JEMALLOC_N(prof_mdump)
+#define prof_postfork_child JEMALLOC_N(prof_postfork_child)
+#define prof_postfork_parent JEMALLOC_N(prof_postfork_parent)
+#define prof_prefork0 JEMALLOC_N(prof_prefork0)
+#define prof_prefork1 JEMALLOC_N(prof_prefork1)
+#define prof_reset JEMALLOC_N(prof_reset)
+#define prof_sample_threshold_update JEMALLOC_N(prof_sample_threshold_update)
+#define prof_tdata_cleanup JEMALLOC_N(prof_tdata_cleanup)
+#define prof_tdata_init JEMALLOC_N(prof_tdata_init)
+#define prof_tdata_reinit JEMALLOC_N(prof_tdata_reinit)
+#define prof_thread_active_get JEMALLOC_N(prof_thread_active_get)
+#define prof_thread_active_init_get JEMALLOC_N(prof_thread_active_init_get)
+#define prof_thread_active_init_set JEMALLOC_N(prof_thread_active_init_set)
+#define prof_thread_active_set JEMALLOC_N(prof_thread_active_set)
+#define prof_thread_name_get JEMALLOC_N(prof_thread_name_get)
+#define prof_thread_name_set JEMALLOC_N(prof_thread_name_set)
+#define rtree_ctx_data_init JEMALLOC_N(rtree_ctx_data_init)
+#define rtree_leaf_alloc JEMALLOC_N(rtree_leaf_alloc)
+#define rtree_leaf_dalloc JEMALLOC_N(rtree_leaf_dalloc)
+#define rtree_leaf_elm_lookup_hard JEMALLOC_N(rtree_leaf_elm_lookup_hard)
+#define rtree_new JEMALLOC_N(rtree_new)
+#define rtree_node_alloc JEMALLOC_N(rtree_node_alloc)
+#define rtree_node_dalloc JEMALLOC_N(rtree_node_dalloc)
+#define safety_check_fail JEMALLOC_N(safety_check_fail)
+#define safety_check_set_abort JEMALLOC_N(safety_check_set_abort)
+#define arena_mutex_names JEMALLOC_N(arena_mutex_names)
+#define global_mutex_names JEMALLOC_N(global_mutex_names)
+#define opt_stats_print JEMALLOC_N(opt_stats_print)
+#define opt_stats_print_opts JEMALLOC_N(opt_stats_print_opts)
+#define stats_print JEMALLOC_N(stats_print)
+#define sc_boot JEMALLOC_N(sc_boot)
+#define sc_data_global JEMALLOC_N(sc_data_global)
+#define sc_data_init JEMALLOC_N(sc_data_init)
+#define sc_data_update_slab_size JEMALLOC_N(sc_data_update_slab_size)
+#define sz_boot JEMALLOC_N(sz_boot)
+#define sz_index2size_tab JEMALLOC_N(sz_index2size_tab)
+#define sz_pind2sz_tab JEMALLOC_N(sz_pind2sz_tab)
+#define sz_size2index_tab JEMALLOC_N(sz_size2index_tab)
+#define nhbins JEMALLOC_N(nhbins)
+#define opt_lg_tcache_max JEMALLOC_N(opt_lg_tcache_max)
+#define opt_tcache JEMALLOC_N(opt_tcache)
+#define tcache_alloc_small_hard JEMALLOC_N(tcache_alloc_small_hard)
+#define tcache_arena_associate JEMALLOC_N(tcache_arena_associate)
+#define tcache_arena_reassociate JEMALLOC_N(tcache_arena_reassociate)
+#define tcache_bin_flush_large JEMALLOC_N(tcache_bin_flush_large)
+#define tcache_bin_flush_small JEMALLOC_N(tcache_bin_flush_small)
+#define tcache_bin_info JEMALLOC_N(tcache_bin_info)
+#define tcache_boot JEMALLOC_N(tcache_boot)
+#define tcache_cleanup JEMALLOC_N(tcache_cleanup)
+#define tcache_create_explicit JEMALLOC_N(tcache_create_explicit)
+#define tcache_event_hard JEMALLOC_N(tcache_event_hard)
+#define tcache_flush JEMALLOC_N(tcache_flush)
+#define tcache_maxclass JEMALLOC_N(tcache_maxclass)
+#define tcache_postfork_child JEMALLOC_N(tcache_postfork_child)
+#define tcache_postfork_parent JEMALLOC_N(tcache_postfork_parent)
+#define tcache_prefork JEMALLOC_N(tcache_prefork)
+#define tcaches JEMALLOC_N(tcaches)
+#define tcache_salloc JEMALLOC_N(tcache_salloc)
+#define tcaches_create JEMALLOC_N(tcaches_create)
+#define tcaches_destroy JEMALLOC_N(tcaches_destroy)
+#define tcaches_flush JEMALLOC_N(tcaches_flush)
+#define tcache_stats_merge JEMALLOC_N(tcache_stats_merge)
+#define tsd_tcache_data_init JEMALLOC_N(tsd_tcache_data_init)
+#define tsd_tcache_enabled_data_init JEMALLOC_N(tsd_tcache_enabled_data_init)
+#define test_hooks_arena_new_hook JEMALLOC_N(test_hooks_arena_new_hook)
+#define test_hooks_libc_hook JEMALLOC_N(test_hooks_libc_hook)
+#define malloc_tsd_boot0 JEMALLOC_N(malloc_tsd_boot0)
+#define malloc_tsd_boot1 JEMALLOC_N(malloc_tsd_boot1)
+#define malloc_tsd_cleanup_register JEMALLOC_N(malloc_tsd_cleanup_register)
+#define malloc_tsd_dalloc JEMALLOC_N(malloc_tsd_dalloc)
+#define malloc_tsd_malloc JEMALLOC_N(malloc_tsd_malloc)
+#define tsd_booted JEMALLOC_N(tsd_booted)
+#define tsd_cleanup JEMALLOC_N(tsd_cleanup)
+#define tsd_fetch_slow JEMALLOC_N(tsd_fetch_slow)
+#define tsd_global_slow JEMALLOC_N(tsd_global_slow)
+#define tsd_global_slow_dec JEMALLOC_N(tsd_global_slow_dec)
+#define tsd_global_slow_inc JEMALLOC_N(tsd_global_slow_inc)
+#define tsd_postfork_child JEMALLOC_N(tsd_postfork_child)
+#define tsd_postfork_parent JEMALLOC_N(tsd_postfork_parent)
+#define tsd_prefork JEMALLOC_N(tsd_prefork)
+#define tsd_slow_update JEMALLOC_N(tsd_slow_update)
+#define tsd_state_set JEMALLOC_N(tsd_state_set)
+#define tsd_tls JEMALLOC_N(tsd_tls)
+#define tsd_tsd JEMALLOC_N(tsd_tsd)
+#define witness_depth_error JEMALLOC_N(witness_depth_error)
+#define witnesses_cleanup JEMALLOC_N(witnesses_cleanup)
+#define witness_init JEMALLOC_N(witness_init)
+#define witness_lock_error JEMALLOC_N(witness_lock_error)
+#define witness_not_owner_error JEMALLOC_N(witness_not_owner_error)
+#define witness_owner_error JEMALLOC_N(witness_owner_error)
+#define witness_postfork_child JEMALLOC_N(witness_postfork_child)
+#define witness_postfork_parent JEMALLOC_N(witness_postfork_parent)
+#define witness_prefork JEMALLOC_N(witness_prefork)

deps/jemalloc/include/jemalloc/internal/prng.h

@@ -0,0 +1,185 @@
+#ifndef JEMALLOC_INTERNAL_PRNG_H
+#define JEMALLOC_INTERNAL_PRNG_H
+
+#include "jemalloc/internal/atomic.h"
+#include "jemalloc/internal/bit_util.h"
+
+/*
+ * Simple linear congruential pseudo-random number generator:
+ *
+ *   prng(y) = (a*x + c) % m
+ *
+ * where the following constants ensure maximal period:
+ *
+ *   a == Odd number (relatively prime to 2^n), and (a-1) is a multiple of 4.
+ *   c == Odd number (relatively prime to 2^n).
+ *   m == 2^32
+ *
+ * See Knuth's TAOCP 3rd Ed., Vol. 2, pg. 17 for details on these constraints.
+ *
+ * This choice of m has the disadvantage that the quality of the bits is
+ * proportional to bit position.  For example, the lowest bit has a cycle of 2,
+ * the next has a cycle of 4, etc.  For this reason, we prefer to use the upper
+ * bits.
+ */
+
+/******************************************************************************/
+/* INTERNAL DEFINITIONS -- IGNORE */
+/******************************************************************************/
+#define PRNG_A_32	UINT32_C(1103515241)
+#define PRNG_C_32	UINT32_C(12347)
+
+#define PRNG_A_64	UINT64_C(6364136223846793005)
+#define PRNG_C_64	UINT64_C(1442695040888963407)
+
+JEMALLOC_ALWAYS_INLINE uint32_t
+prng_state_next_u32(uint32_t state) {
+	return (state * PRNG_A_32) + PRNG_C_32;
+}
+
+JEMALLOC_ALWAYS_INLINE uint64_t
+prng_state_next_u64(uint64_t state) {
+	return (state * PRNG_A_64) + PRNG_C_64;
+}
+
+JEMALLOC_ALWAYS_INLINE size_t
+prng_state_next_zu(size_t state) {
+#if LG_SIZEOF_PTR == 2
+	return (state * PRNG_A_32) + PRNG_C_32;
+#elif LG_SIZEOF_PTR == 3
+	return (state * PRNG_A_64) + PRNG_C_64;
+#else
+#error Unsupported pointer size
+#endif
+}
+
+/******************************************************************************/
+/* BEGIN PUBLIC API */
+/******************************************************************************/
+
+/*
+ * The prng_lg_range functions give a uniform int in the half-open range [0,
+ * 2**lg_range).  If atomic is true, they do so safely from multiple threads.
+ * Multithreaded 64-bit prngs aren't supported.
+ */
+
+JEMALLOC_ALWAYS_INLINE uint32_t
+prng_lg_range_u32(atomic_u32_t *state, unsigned lg_range, bool atomic) {
+	uint32_t ret, state0, state1;
+
+	assert(lg_range > 0);
+	assert(lg_range <= 32);
+
+	state0 = atomic_load_u32(state, ATOMIC_RELAXED);
+
+	if (atomic) {
+		do {
+			state1 = prng_state_next_u32(state0);
+		} while (!atomic_compare_exchange_weak_u32(state, &state0,
+		    state1, ATOMIC_RELAXED, ATOMIC_RELAXED));
+	} else {
+		state1 = prng_state_next_u32(state0);
+		atomic_store_u32(state, state1, ATOMIC_RELAXED);
+	}
+	ret = state1 >> (32 - lg_range);
+
+	return ret;
+}
+
+JEMALLOC_ALWAYS_INLINE uint64_t
+prng_lg_range_u64(uint64_t *state, unsigned lg_range) {
+	uint64_t ret, state1;
+
+	assert(lg_range > 0);
+	assert(lg_range <= 64);
+
+	state1 = prng_state_next_u64(*state);
+	*state = state1;
+	ret = state1 >> (64 - lg_range);
+
+	return ret;
+}
+
+JEMALLOC_ALWAYS_INLINE size_t
+prng_lg_range_zu(atomic_zu_t *state, unsigned lg_range, bool atomic) {
+	size_t ret, state0, state1;
+
+	assert(lg_range > 0);
+	assert(lg_range <= ZU(1) << (3 + LG_SIZEOF_PTR));
+
+	state0 = atomic_load_zu(state, ATOMIC_RELAXED);
+
+	if (atomic) {
+		do {
+			state1 = prng_state_next_zu(state0);
+		} while (atomic_compare_exchange_weak_zu(state, &state0,
+		    state1, ATOMIC_RELAXED, ATOMIC_RELAXED));
+	} else {
+		state1 = prng_state_next_zu(state0);
+		atomic_store_zu(state, state1, ATOMIC_RELAXED);
+	}
+	ret = state1 >> ((ZU(1) << (3 + LG_SIZEOF_PTR)) - lg_range);
+
+	return ret;
+}
+
+/*
+ * The prng_range functions behave like the prng_lg_range, but return a result
+ * in [0, range) instead of [0, 2**lg_range).
+ */
+
+JEMALLOC_ALWAYS_INLINE uint32_t
+prng_range_u32(atomic_u32_t *state, uint32_t range, bool atomic) {
+	uint32_t ret;
+	unsigned lg_range;
+
+	assert(range > 1);
+
+	/* Compute the ceiling of lg(range). */
+	lg_range = ffs_u32(pow2_ceil_u32(range)) - 1;
+
+	/* Generate a result in [0..range) via repeated trial. */
+	do {
+		ret = prng_lg_range_u32(state, lg_range, atomic);
+	} while (ret >= range);
+
+	return ret;
+}
+
+JEMALLOC_ALWAYS_INLINE uint64_t
+prng_range_u64(uint64_t *state, uint64_t range) {
+	uint64_t ret;
+	unsigned lg_range;
+
+	assert(range > 1);
+
+	/* Compute the ceiling of lg(range). */
+	lg_range = ffs_u64(pow2_ceil_u64(range)) - 1;
+
+	/* Generate a result in [0..range) via repeated trial. */
+	do {
+		ret = prng_lg_range_u64(state, lg_range);
+	} while (ret >= range);
+
+	return ret;
+}
+
+JEMALLOC_ALWAYS_INLINE size_t
+prng_range_zu(atomic_zu_t *state, size_t range, bool atomic) {
+	size_t ret;
+	unsigned lg_range;
+
+	assert(range > 1);
+
+	/* Compute the ceiling of lg(range). */
+	lg_range = ffs_u64(pow2_ceil_u64(range)) - 1;
+
+	/* Generate a result in [0..range) via repeated trial. */
+	do {
+		ret = prng_lg_range_zu(state, lg_range, atomic);
+	} while (ret >= range);
+
+	return ret;
+}
+
+#endif /* JEMALLOC_INTERNAL_PRNG_H */

deps/jemalloc/include/jemalloc/internal/prof.h

@@ -0,0 +1,613 @@
+/******************************************************************************/
+#ifdef JEMALLOC_H_TYPES
+
+typedef struct prof_bt_s prof_bt_t;
+typedef struct prof_cnt_s prof_cnt_t;
+typedef struct prof_thr_cnt_s prof_thr_cnt_t;
+typedef struct prof_ctx_s prof_ctx_t;
+typedef struct prof_tdata_s prof_tdata_t;
+
+/* Option defaults. */
+#ifdef JEMALLOC_PROF
+#  define PROF_PREFIX_DEFAULT		"jeprof"
+#else
+#  define PROF_PREFIX_DEFAULT		""
+#endif
+#define	LG_PROF_SAMPLE_DEFAULT		19
+#define	LG_PROF_INTERVAL_DEFAULT	-1
+
+/*
+ * Hard limit on stack backtrace depth.  The version of prof_backtrace() that
+ * is based on __builtin_return_address() necessarily has a hard-coded number
+ * of backtrace frame handlers, and should be kept in sync with this setting.
+ */
+#define	PROF_BT_MAX			128
+
+/* Maximum number of backtraces to store in each per thread LRU cache. */
+#define	PROF_TCMAX			1024
+
+/* Initial hash table size. */
+#define	PROF_CKH_MINITEMS		64
+
+/* Size of memory buffer to use when writing dump files. */
+#define	PROF_DUMP_BUFSIZE		65536
+
+/* Size of stack-allocated buffer used by prof_printf(). */
+#define	PROF_PRINTF_BUFSIZE		128
+
+/*
+ * Number of mutexes shared among all ctx's.  No space is allocated for these
+ * unless profiling is enabled, so it's okay to over-provision.
+ */
+#define	PROF_NCTX_LOCKS			1024
+
+/*
+ * prof_tdata pointers close to NULL are used to encode state information that
+ * is used for cleaning up during thread shutdown.
+ */
+#define	PROF_TDATA_STATE_REINCARNATED	((prof_tdata_t *)(uintptr_t)1)
+#define	PROF_TDATA_STATE_PURGATORY	((prof_tdata_t *)(uintptr_t)2)
+#define	PROF_TDATA_STATE_MAX		PROF_TDATA_STATE_PURGATORY
+
+#endif /* JEMALLOC_H_TYPES */
+/******************************************************************************/
+#ifdef JEMALLOC_H_STRUCTS
+
+struct prof_bt_s {
+	/* Backtrace, stored as len program counters. */
+	void		**vec;
+	unsigned	len;
+};
+
+#ifdef JEMALLOC_PROF_LIBGCC
+/* Data structure passed to libgcc _Unwind_Backtrace() callback functions. */
+typedef struct {
+	prof_bt_t	*bt;
+	unsigned	nignore;
+	unsigned	max;
+} prof_unwind_data_t;
+#endif
+
+struct prof_cnt_s {
+	/*
+	 * Profiling counters.  An allocation/deallocation pair can operate on
+	 * different prof_thr_cnt_t objects that are linked into the same
+	 * prof_ctx_t cnts_ql, so it is possible for the cur* counters to go
+	 * negative.  In principle it is possible for the *bytes counters to
+	 * overflow/underflow, but a general solution would require something
+	 * like 128-bit counters; this implementation doesn't bother to solve
+	 * that problem.
+	 */
+	int64_t		curobjs;
+	int64_t		curbytes;
+	uint64_t	accumobjs;
+	uint64_t	accumbytes;
+};
+
+struct prof_thr_cnt_s {
+	/* Linkage into prof_ctx_t's cnts_ql. */
+	ql_elm(prof_thr_cnt_t)	cnts_link;
+
+	/* Linkage into thread's LRU. */
+	ql_elm(prof_thr_cnt_t)	lru_link;
+
+	/*
+	 * Associated context.  If a thread frees an object that it did not
+	 * allocate, it is possible that the context is not cached in the
+	 * thread's hash table, in which case it must be able to look up the
+	 * context, insert a new prof_thr_cnt_t into the thread's hash table,
+	 * and link it into the prof_ctx_t's cnts_ql.
+	 */
+	prof_ctx_t		*ctx;
+
+	/*
+	 * Threads use memory barriers to update the counters.  Since there is
+	 * only ever one writer, the only challenge is for the reader to get a
+	 * consistent read of the counters.
+	 *
+	 * The writer uses this series of operations:
+	 *
+	 * 1) Increment epoch to an odd number.
+	 * 2) Update counters.
+	 * 3) Increment epoch to an even number.
+	 *
+	 * The reader must assure 1) that the epoch is even while it reads the
+	 * counters, and 2) that the epoch doesn't change between the time it
+	 * starts and finishes reading the counters.
+	 */
+	unsigned		epoch;
+
+	/* Profiling counters. */
+	prof_cnt_t		cnts;
+};
+
+struct prof_ctx_s {
+	/* Associated backtrace. */
+	prof_bt_t		*bt;
+
+	/* Protects nlimbo, cnt_merged, and cnts_ql. */
+	malloc_mutex_t		*lock;
+
+	/*
+	 * Number of threads that currently cause this ctx to be in a state of
+	 * limbo due to one of:
+	 *   - Initializing per thread counters associated with this ctx.
+	 *   - Preparing to destroy this ctx.
+	 *   - Dumping a heap profile that includes this ctx.
+	 * nlimbo must be 1 (single destroyer) in order to safely destroy the
+	 * ctx.
+	 */
+	unsigned		nlimbo;
+
+	/* Temporary storage for summation during dump. */
+	prof_cnt_t		cnt_summed;
+
+	/* When threads exit, they merge their stats into cnt_merged. */
+	prof_cnt_t		cnt_merged;
+
+	/*
+	 * List of profile counters, one for each thread that has allocated in
+	 * this context.
+	 */
+	ql_head(prof_thr_cnt_t)	cnts_ql;
+
+	/* Linkage for list of contexts to be dumped. */
+	ql_elm(prof_ctx_t)	dump_link;
+};
+typedef ql_head(prof_ctx_t) prof_ctx_list_t;
+
+struct prof_tdata_s {
+	/*
+	 * Hash of (prof_bt_t *)-->(prof_thr_cnt_t *).  Each thread keeps a
+	 * cache of backtraces, with associated thread-specific prof_thr_cnt_t
+	 * objects.  Other threads may read the prof_thr_cnt_t contents, but no
+	 * others will ever write them.
+	 *
+	 * Upon thread exit, the thread must merge all the prof_thr_cnt_t
+	 * counter data into the associated prof_ctx_t objects, and unlink/free
+	 * the prof_thr_cnt_t objects.
+	 */
+	ckh_t			bt2cnt;
+
+	/* LRU for contents of bt2cnt. */
+	ql_head(prof_thr_cnt_t)	lru_ql;
+
+	/* Backtrace vector, used for calls to prof_backtrace(). */
+	void			**vec;
+
+	/* Sampling state. */
+	uint64_t		prng_state;
+	uint64_t		threshold;
+	uint64_t		accum;
+
+	/* State used to avoid dumping while operating on prof internals. */
+	bool			enq;
+	bool			enq_idump;
+	bool			enq_gdump;
+};
+
+#endif /* JEMALLOC_H_STRUCTS */
+/******************************************************************************/
+#ifdef JEMALLOC_H_EXTERNS
+
+extern bool	opt_prof;
+/*
+ * Even if opt_prof is true, sampling can be temporarily disabled by setting
+ * opt_prof_active to false.  No locking is used when updating opt_prof_active,
+ * so there are no guarantees regarding how long it will take for all threads
+ * to notice state changes.
+ */
+extern bool	opt_prof_active;
+extern size_t	opt_lg_prof_sample;   /* Mean bytes between samples. */
+extern ssize_t	opt_lg_prof_interval; /* lg(prof_interval). */
+extern bool	opt_prof_gdump;       /* High-water memory dumping. */
+extern bool	opt_prof_final;       /* Final profile dumping. */
+extern bool	opt_prof_leak;        /* Dump leak summary at exit. */
+extern bool	opt_prof_accum;       /* Report cumulative bytes. */
+extern char	opt_prof_prefix[
+    /* Minimize memory bloat for non-prof builds. */
+#ifdef JEMALLOC_PROF
+    PATH_MAX +
+#endif
+    1];
+
+/*
+ * Profile dump interval, measured in bytes allocated.  Each arena triggers a
+ * profile dump when it reaches this threshold.  The effect is that the
+ * interval between profile dumps averages prof_interval, though the actual
+ * interval between dumps will tend to be sporadic, and the interval will be a
+ * maximum of approximately (prof_interval * narenas).
+ */
+extern uint64_t	prof_interval;
+
+/*
+ * If true, promote small sampled objects to large objects, since small run
+ * headers do not have embedded profile context pointers.
+ */
+extern bool	prof_promote;
+
+void	bt_init(prof_bt_t *bt, void **vec);
+void	prof_backtrace(prof_bt_t *bt, unsigned nignore);
+prof_thr_cnt_t	*prof_lookup(prof_bt_t *bt);
+#ifdef JEMALLOC_JET
+size_t	prof_bt_count(void);
+typedef int (prof_dump_open_t)(bool, const char *);
+extern prof_dump_open_t *prof_dump_open;
+#endif
+void	prof_idump(void);
+bool	prof_mdump(const char *filename);
+void	prof_gdump(void);
+prof_tdata_t	*prof_tdata_init(void);
+void	prof_tdata_cleanup(void *arg);
+void	prof_boot0(void);
+void	prof_boot1(void);
+bool	prof_boot2(void);
+void	prof_prefork(void);
+void	prof_postfork_parent(void);
+void	prof_postfork_child(void);
+
+#endif /* JEMALLOC_H_EXTERNS */
+/******************************************************************************/
+#ifdef JEMALLOC_H_INLINES
+
+#define	PROF_ALLOC_PREP(nignore, size, ret) do {			\
+	prof_tdata_t *prof_tdata;					\
+	prof_bt_t bt;							\
+									\
+	assert(size == s2u(size));					\
+									\
+	prof_tdata = prof_tdata_get(true);				\
+	if ((uintptr_t)prof_tdata <= (uintptr_t)PROF_TDATA_STATE_MAX) {	\
+		if (prof_tdata != NULL)					\
+			ret = (prof_thr_cnt_t *)(uintptr_t)1U;		\
+		else							\
+			ret = NULL;					\
+		break;							\
+	}								\
+									\
+	if (opt_prof_active == false) {					\
+		/* Sampling is currently inactive, so avoid sampling. */\
+		ret = (prof_thr_cnt_t *)(uintptr_t)1U;			\
+	} else if (opt_lg_prof_sample == 0) {				\
+		/* Don't bother with sampling logic, since sampling   */\
+		/* interval is 1.                                     */\
+		bt_init(&bt, prof_tdata->vec);				\
+		prof_backtrace(&bt, nignore);				\
+		ret = prof_lookup(&bt);					\
+	} else {							\
+		if (prof_tdata->threshold == 0) {			\
+			/* Initialize.  Seed the prng differently for */\
+			/* each thread.                               */\
+			prof_tdata->prng_state =			\
+			    (uint64_t)(uintptr_t)&size;			\
+			prof_sample_threshold_update(prof_tdata);	\
+		}							\
+									\
+		/* Determine whether to capture a backtrace based on  */\
+		/* whether size is enough for prof_accum to reach     */\
+		/* prof_tdata->threshold.  However, delay updating    */\
+		/* these variables until prof_{m,re}alloc(), because  */\
+		/* we don't know for sure that the allocation will    */\
+		/* succeed.                                           */\
+		/*                                                    */\
+		/* Use subtraction rather than addition to avoid      */\
+		/* potential integer overflow.                        */\
+		if (size >= prof_tdata->threshold -			\
+		    prof_tdata->accum) {				\
+			bt_init(&bt, prof_tdata->vec);			\
+			prof_backtrace(&bt, nignore);			\
+			ret = prof_lookup(&bt);				\
+		} else							\
+			ret = (prof_thr_cnt_t *)(uintptr_t)1U;		\
+	}								\
+} while (0)
+
+#ifndef JEMALLOC_ENABLE_INLINE
+malloc_tsd_protos(JEMALLOC_ATTR(unused), prof_tdata, prof_tdata_t *)
+
+prof_tdata_t	*prof_tdata_get(bool create);
+void	prof_sample_threshold_update(prof_tdata_t *prof_tdata);
+prof_ctx_t	*prof_ctx_get(const void *ptr);
+void	prof_ctx_set(const void *ptr, size_t usize, prof_ctx_t *ctx);
+bool	prof_sample_accum_update(size_t size);
+void	prof_malloc(const void *ptr, size_t usize, prof_thr_cnt_t *cnt);
+void	prof_realloc(const void *ptr, size_t usize, prof_thr_cnt_t *cnt,
+    size_t old_usize, prof_ctx_t *old_ctx);
+void	prof_free(const void *ptr, size_t size);
+#endif
+
+#if (defined(JEMALLOC_ENABLE_INLINE) || defined(JEMALLOC_PROF_C_))
+/* Thread-specific backtrace cache, used to reduce bt2ctx contention. */
+malloc_tsd_externs(prof_tdata, prof_tdata_t *)
+malloc_tsd_funcs(JEMALLOC_INLINE, prof_tdata, prof_tdata_t *, NULL,
+    prof_tdata_cleanup)
+
+JEMALLOC_INLINE prof_tdata_t *
+prof_tdata_get(bool create)
+{
+	prof_tdata_t *prof_tdata;
+
+	cassert(config_prof);
+
+	prof_tdata = *prof_tdata_tsd_get();
+	if (create && prof_tdata == NULL)
+		prof_tdata = prof_tdata_init();
+
+	return (prof_tdata);
+}
+
+JEMALLOC_INLINE void
+prof_sample_threshold_update(prof_tdata_t *prof_tdata)
+{
+	/*
+	 * The body of this function is compiled out unless heap profiling is
+	 * enabled, so that it is possible to compile jemalloc with floating
+	 * point support completely disabled.  Avoiding floating point code is
+	 * important on memory-constrained systems, but it also enables a
+	 * workaround for versions of glibc that don't properly save/restore
+	 * floating point registers during dynamic lazy symbol loading (which
+	 * internally calls into whatever malloc implementation happens to be
+	 * integrated into the application).  Note that some compilers (e.g.
+	 * gcc 4.8) may use floating point registers for fast memory moves, so
+	 * jemalloc must be compiled with such optimizations disabled (e.g.
+	 * -mno-sse) in order for the workaround to be complete.
+	 */
+#ifdef JEMALLOC_PROF
+	uint64_t r;
+	double u;
+
+	cassert(config_prof);
+
+	/*
+	 * Compute sample threshold as a geometrically distributed random
+	 * variable with mean (2^opt_lg_prof_sample).
+	 *
+	 *                         __        __
+	 *                         |  log(u)  |                     1
+	 * prof_tdata->threshold = | -------- |, where p = -------------------
+	 *                         | log(1-p) |             opt_lg_prof_sample
+	 *                                                 2
+	 *
+	 * For more information on the math, see:
+	 *
+	 *   Non-Uniform Random Variate Generation
+	 *   Luc Devroye
+	 *   Springer-Verlag, New York, 1986
+	 *   pp 500
+	 *   (http://luc.devroye.org/rnbookindex.html)
+	 */
+	prng64(r, 53, prof_tdata->prng_state,
+	    UINT64_C(6364136223846793005), UINT64_C(1442695040888963407));
+	u = (double)r * (1.0/9007199254740992.0L);
+	prof_tdata->threshold = (uint64_t)(log(u) /
+	    log(1.0 - (1.0 / (double)((uint64_t)1U << opt_lg_prof_sample))))
+	    + (uint64_t)1U;
+#endif
+}
+
+JEMALLOC_INLINE prof_ctx_t *
+prof_ctx_get(const void *ptr)
+{
+	prof_ctx_t *ret;
+	arena_chunk_t *chunk;
+
+	cassert(config_prof);
+	assert(ptr != NULL);
+
+	chunk = (arena_chunk_t *)CHUNK_ADDR2BASE(ptr);
+	if (chunk != ptr) {
+		/* Region. */
+		ret = arena_prof_ctx_get(ptr);
+	} else
+		ret = huge_prof_ctx_get(ptr);
+
+	return (ret);
+}
+
+JEMALLOC_INLINE void
+prof_ctx_set(const void *ptr, size_t usize, prof_ctx_t *ctx)
+{
+	arena_chunk_t *chunk;
+
+	cassert(config_prof);
+	assert(ptr != NULL);
+
+	chunk = (arena_chunk_t *)CHUNK_ADDR2BASE(ptr);
+	if (chunk != ptr) {
+		/* Region. */
+		arena_prof_ctx_set(ptr, usize, ctx);
+	} else
+		huge_prof_ctx_set(ptr, ctx);
+}
+
+JEMALLOC_INLINE bool
+prof_sample_accum_update(size_t size)
+{
+	prof_tdata_t *prof_tdata;
+
+	cassert(config_prof);
+	/* Sampling logic is unnecessary if the interval is 1. */
+	assert(opt_lg_prof_sample != 0);
+
+	prof_tdata = prof_tdata_get(false);
+	if ((uintptr_t)prof_tdata <= (uintptr_t)PROF_TDATA_STATE_MAX)
+		return (true);
+
+	/* Take care to avoid integer overflow. */
+	if (size >= prof_tdata->threshold - prof_tdata->accum) {
+		prof_tdata->accum -= (prof_tdata->threshold - size);
+		/* Compute new sample threshold. */
+		prof_sample_threshold_update(prof_tdata);
+		while (prof_tdata->accum >= prof_tdata->threshold) {
+			prof_tdata->accum -= prof_tdata->threshold;
+			prof_sample_threshold_update(prof_tdata);
+		}
+		return (false);
+	} else {
+		prof_tdata->accum += size;
+		return (true);
+	}
+}
+
+JEMALLOC_INLINE void
+prof_malloc(const void *ptr, size_t usize, prof_thr_cnt_t *cnt)
+{
+
+	cassert(config_prof);
+	assert(ptr != NULL);
+	assert(usize == isalloc(ptr, true));
+
+	if (opt_lg_prof_sample != 0) {
+		if (prof_sample_accum_update(usize)) {
+			/*
+			 * Don't sample.  For malloc()-like allocation, it is
+			 * always possible to tell in advance how large an
+			 * object's usable size will be, so there should never
+			 * be a difference between the usize passed to
+			 * PROF_ALLOC_PREP() and prof_malloc().
+			 */
+			assert((uintptr_t)cnt == (uintptr_t)1U);
+		}
+	}
+
+	if ((uintptr_t)cnt > (uintptr_t)1U) {
+		prof_ctx_set(ptr, usize, cnt->ctx);
+
+		cnt->epoch++;
+		/*********/
+		mb_write();
+		/*********/
+		cnt->cnts.curobjs++;
+		cnt->cnts.curbytes += usize;
+		if (opt_prof_accum) {
+			cnt->cnts.accumobjs++;
+			cnt->cnts.accumbytes += usize;
+		}
+		/*********/
+		mb_write();
+		/*********/
+		cnt->epoch++;
+		/*********/
+		mb_write();
+		/*********/
+	} else
+		prof_ctx_set(ptr, usize, (prof_ctx_t *)(uintptr_t)1U);
+}
+
+JEMALLOC_INLINE void
+prof_realloc(const void *ptr, size_t usize, prof_thr_cnt_t *cnt,
+    size_t old_usize, prof_ctx_t *old_ctx)
+{
+	prof_thr_cnt_t *told_cnt;
+
+	cassert(config_prof);
+	assert(ptr != NULL || (uintptr_t)cnt <= (uintptr_t)1U);
+
+	if (ptr != NULL) {
+		assert(usize == isalloc(ptr, true));
+		if (opt_lg_prof_sample != 0) {
+			if (prof_sample_accum_update(usize)) {
+				/*
+				 * Don't sample.  The usize passed to
+				 * PROF_ALLOC_PREP() was larger than what
+				 * actually got allocated, so a backtrace was
+				 * captured for this allocation, even though
+				 * its actual usize was insufficient to cross
+				 * the sample threshold.
+				 */
+				cnt = (prof_thr_cnt_t *)(uintptr_t)1U;
+			}
+		}
+	}
+
+	if ((uintptr_t)old_ctx > (uintptr_t)1U) {
+		told_cnt = prof_lookup(old_ctx->bt);
+		if (told_cnt == NULL) {
+			/*
+			 * It's too late to propagate OOM for this realloc(),
+			 * so operate directly on old_cnt->ctx->cnt_merged.
+			 */
+			malloc_mutex_lock(old_ctx->lock);
+			old_ctx->cnt_merged.curobjs--;
+			old_ctx->cnt_merged.curbytes -= old_usize;
+			malloc_mutex_unlock(old_ctx->lock);
+			told_cnt = (prof_thr_cnt_t *)(uintptr_t)1U;
+		}
+	} else
+		told_cnt = (prof_thr_cnt_t *)(uintptr_t)1U;
+
+	if ((uintptr_t)told_cnt > (uintptr_t)1U)
+		told_cnt->epoch++;
+	if ((uintptr_t)cnt > (uintptr_t)1U) {
+		prof_ctx_set(ptr, usize, cnt->ctx);
+		cnt->epoch++;
+	} else if (ptr != NULL)
+		prof_ctx_set(ptr, usize, (prof_ctx_t *)(uintptr_t)1U);
+	/*********/
+	mb_write();
+	/*********/
+	if ((uintptr_t)told_cnt > (uintptr_t)1U) {
+		told_cnt->cnts.curobjs--;
+		told_cnt->cnts.curbytes -= old_usize;
+	}
+	if ((uintptr_t)cnt > (uintptr_t)1U) {
+		cnt->cnts.curobjs++;
+		cnt->cnts.curbytes += usize;
+		if (opt_prof_accum) {
+			cnt->cnts.accumobjs++;
+			cnt->cnts.accumbytes += usize;
+		}
+	}
+	/*********/
+	mb_write();
+	/*********/
+	if ((uintptr_t)told_cnt > (uintptr_t)1U)
+		told_cnt->epoch++;
+	if ((uintptr_t)cnt > (uintptr_t)1U)
+		cnt->epoch++;
+	/*********/
+	mb_write(); /* Not strictly necessary. */
+}
+
+JEMALLOC_INLINE void
+prof_free(const void *ptr, size_t size)
+{
+	prof_ctx_t *ctx = prof_ctx_get(ptr);
+
+	cassert(config_prof);
+
+	if ((uintptr_t)ctx > (uintptr_t)1) {
+		prof_thr_cnt_t *tcnt;
+		assert(size == isalloc(ptr, true));
+		tcnt = prof_lookup(ctx->bt);
+
+		if (tcnt != NULL) {
+			tcnt->epoch++;
+			/*********/
+			mb_write();
+			/*********/
+			tcnt->cnts.curobjs--;
+			tcnt->cnts.curbytes -= size;
+			/*********/
+			mb_write();
+			/*********/
+			tcnt->epoch++;
+			/*********/
+			mb_write();
+			/*********/
+		} else {
+			/*
+			 * OOM during free() cannot be propagated, so operate
+			 * directly on cnt->ctx->cnt_merged.
+			 */
+			malloc_mutex_lock(ctx->lock);
+			ctx->cnt_merged.curobjs--;
+			ctx->cnt_merged.curbytes -= size;
+			malloc_mutex_unlock(ctx->lock);
+		}
+	}
+}
+#endif
+
+#endif /* JEMALLOC_H_INLINES */
+/******************************************************************************/

deps/jemalloc/include/jemalloc/internal/prof_externs.h

@@ -0,0 +1,105 @@
+#ifndef JEMALLOC_INTERNAL_PROF_EXTERNS_H
+#define JEMALLOC_INTERNAL_PROF_EXTERNS_H
+
+#include "jemalloc/internal/mutex.h"
+
+extern malloc_mutex_t	bt2gctx_mtx;
+
+extern bool	opt_prof;
+extern bool	opt_prof_active;
+extern bool	opt_prof_thread_active_init;
+extern size_t	opt_lg_prof_sample;   /* Mean bytes between samples. */
+extern ssize_t	opt_lg_prof_interval; /* lg(prof_interval). */
+extern bool	opt_prof_gdump;       /* High-water memory dumping. */
+extern bool	opt_prof_final;       /* Final profile dumping. */
+extern bool	opt_prof_leak;        /* Dump leak summary at exit. */
+extern bool	opt_prof_accum;       /* Report cumulative bytes. */
+extern bool	opt_prof_log;	      /* Turn logging on at boot. */
+extern char	opt_prof_prefix[
+    /* Minimize memory bloat for non-prof builds. */
+#ifdef JEMALLOC_PROF
+    PATH_MAX +
+#endif
+    1];
+
+/* Accessed via prof_active_[gs]et{_unlocked,}(). */
+extern bool	prof_active;
+
+/* Accessed via prof_gdump_[gs]et{_unlocked,}(). */
+extern bool	prof_gdump_val;
+
+/*
+ * Profile dump interval, measured in bytes allocated.  Each arena triggers a
+ * profile dump when it reaches this threshold.  The effect is that the
+ * interval between profile dumps averages prof_interval, though the actual
+ * interval between dumps will tend to be sporadic, and the interval will be a
+ * maximum of approximately (prof_interval * narenas).
+ */
+extern uint64_t	prof_interval;
+
+/*
+ * Initialized as opt_lg_prof_sample, and potentially modified during profiling
+ * resets.
+ */
+extern size_t	lg_prof_sample;
+
+void prof_alloc_rollback(tsd_t *tsd, prof_tctx_t *tctx, bool updated);
+void prof_malloc_sample_object(tsdn_t *tsdn, const void *ptr, size_t usize,
+    prof_tctx_t *tctx);
+void prof_free_sampled_object(tsd_t *tsd, const void *ptr, size_t usize,
+    prof_tctx_t *tctx);
+void bt_init(prof_bt_t *bt, void **vec);
+void prof_backtrace(prof_bt_t *bt);
+prof_tctx_t *prof_lookup(tsd_t *tsd, prof_bt_t *bt);
+#ifdef JEMALLOC_JET
+size_t prof_tdata_count(void);
+size_t prof_bt_count(void);
+#endif
+typedef int (prof_dump_open_t)(bool, const char *);
+extern prof_dump_open_t *JET_MUTABLE prof_dump_open;
+
+typedef bool (prof_dump_header_t)(tsdn_t *, bool, const prof_cnt_t *);
+extern prof_dump_header_t *JET_MUTABLE prof_dump_header;
+#ifdef JEMALLOC_JET
+void prof_cnt_all(uint64_t *curobjs, uint64_t *curbytes, uint64_t *accumobjs,
+    uint64_t *accumbytes);
+#endif
+bool prof_accum_init(tsdn_t *tsdn, prof_accum_t *prof_accum);
+void prof_idump(tsdn_t *tsdn);
+bool prof_mdump(tsd_t *tsd, const char *filename);
+void prof_gdump(tsdn_t *tsdn);
+prof_tdata_t *prof_tdata_init(tsd_t *tsd);
+prof_tdata_t *prof_tdata_reinit(tsd_t *tsd, prof_tdata_t *tdata);
+void prof_reset(tsd_t *tsd, size_t lg_sample);
+void prof_tdata_cleanup(tsd_t *tsd);
+bool prof_active_get(tsdn_t *tsdn);
+bool prof_active_set(tsdn_t *tsdn, bool active);
+const char *prof_thread_name_get(tsd_t *tsd);
+int prof_thread_name_set(tsd_t *tsd, const char *thread_name);
+bool prof_thread_active_get(tsd_t *tsd);
+bool prof_thread_active_set(tsd_t *tsd, bool active);
+bool prof_thread_active_init_get(tsdn_t *tsdn);
+bool prof_thread_active_init_set(tsdn_t *tsdn, bool active_init);
+bool prof_gdump_get(tsdn_t *tsdn);
+bool prof_gdump_set(tsdn_t *tsdn, bool active);
+void prof_boot0(void);
+void prof_boot1(void);
+bool prof_boot2(tsd_t *tsd);
+void prof_prefork0(tsdn_t *tsdn);
+void prof_prefork1(tsdn_t *tsdn);
+void prof_postfork_parent(tsdn_t *tsdn);
+void prof_postfork_child(tsdn_t *tsdn);
+void prof_sample_threshold_update(prof_tdata_t *tdata);
+
+bool prof_log_start(tsdn_t *tsdn, const char *filename);
+bool prof_log_stop(tsdn_t *tsdn);
+#ifdef JEMALLOC_JET
+size_t prof_log_bt_count(void);
+size_t prof_log_alloc_count(void);
+size_t prof_log_thr_count(void);
+bool prof_log_is_logging(void);
+bool prof_log_rep_check(void);
+void prof_log_dummy_set(bool new_value);
+#endif
+
+#endif /* JEMALLOC_INTERNAL_PROF_EXTERNS_H */

deps/jemalloc/include/jemalloc/internal/prof_inlines_a.h

@@ -0,0 +1,85 @@
+#ifndef JEMALLOC_INTERNAL_PROF_INLINES_A_H
+#define JEMALLOC_INTERNAL_PROF_INLINES_A_H
+
+#include "jemalloc/internal/mutex.h"
+
+static inline bool
+prof_accum_add(tsdn_t *tsdn, prof_accum_t *prof_accum,
+    uint64_t accumbytes) {
+	cassert(config_prof);
+
+	bool overflow;
+	uint64_t a0, a1;
+
+	/*
+	 * If the application allocates fast enough (and/or if idump is slow
+	 * enough), extreme overflow here (a1 >= prof_interval * 2) can cause
+	 * idump trigger coalescing.  This is an intentional mechanism that
+	 * avoids rate-limiting allocation.
+	 */
+#ifdef JEMALLOC_ATOMIC_U64
+	a0 = atomic_load_u64(&prof_accum->accumbytes, ATOMIC_RELAXED);
+	do {
+		a1 = a0 + accumbytes;
+		assert(a1 >= a0);
+		overflow = (a1 >= prof_interval);
+		if (overflow) {
+			a1 %= prof_interval;
+		}
+	} while (!atomic_compare_exchange_weak_u64(&prof_accum->accumbytes, &a0,
+	    a1, ATOMIC_RELAXED, ATOMIC_RELAXED));
+#else
+	malloc_mutex_lock(tsdn, &prof_accum->mtx);
+	a0 = prof_accum->accumbytes;
+	a1 = a0 + accumbytes;
+	overflow = (a1 >= prof_interval);
+	if (overflow) {
+		a1 %= prof_interval;
+	}
+	prof_accum->accumbytes = a1;
+	malloc_mutex_unlock(tsdn, &prof_accum->mtx);
+#endif
+	return overflow;
+}
+
+static inline void
+prof_accum_cancel(tsdn_t *tsdn, prof_accum_t *prof_accum,
+    size_t usize) {
+	cassert(config_prof);
+
+	/*
+	 * Cancel out as much of the excessive prof_accumbytes increase as
+	 * possible without underflowing.  Interval-triggered dumps occur
+	 * slightly more often than intended as a result of incomplete
+	 * canceling.
+	 */
+	uint64_t a0, a1;
+#ifdef JEMALLOC_ATOMIC_U64
+	a0 = atomic_load_u64(&prof_accum->accumbytes, ATOMIC_RELAXED);
+	do {
+		a1 = (a0 >= SC_LARGE_MINCLASS - usize)
+		    ? a0 - (SC_LARGE_MINCLASS - usize) : 0;
+	} while (!atomic_compare_exchange_weak_u64(&prof_accum->accumbytes, &a0,
+	    a1, ATOMIC_RELAXED, ATOMIC_RELAXED));
+#else
+	malloc_mutex_lock(tsdn, &prof_accum->mtx);
+	a0 = prof_accum->accumbytes;
+	a1 = (a0 >= SC_LARGE_MINCLASS - usize)
+	    ?  a0 - (SC_LARGE_MINCLASS - usize) : 0;
+	prof_accum->accumbytes = a1;
+	malloc_mutex_unlock(tsdn, &prof_accum->mtx);
+#endif
+}
+
+JEMALLOC_ALWAYS_INLINE bool
+prof_active_get_unlocked(void) {
+	/*
+	 * Even if opt_prof is true, sampling can be temporarily disabled by
+	 * setting prof_active to false.  No locking is used when reading
+	 * prof_active in the fast path, so there are no guarantees regarding
+	 * how long it will take for all threads to notice state changes.
+	 */
+	return prof_active;
+}
+
+#endif /* JEMALLOC_INTERNAL_PROF_INLINES_A_H */

deps/jemalloc/include/jemalloc/internal/prof_inlines_b.h

@@ -0,0 +1,250 @@
+#ifndef JEMALLOC_INTERNAL_PROF_INLINES_B_H
+#define JEMALLOC_INTERNAL_PROF_INLINES_B_H
+
+#include "jemalloc/internal/safety_check.h"
+#include "jemalloc/internal/sz.h"
+
+JEMALLOC_ALWAYS_INLINE bool
+prof_gdump_get_unlocked(void) {
+	/*
+	 * No locking is used when reading prof_gdump_val in the fast path, so
+	 * there are no guarantees regarding how long it will take for all
+	 * threads to notice state changes.
+	 */
+	return prof_gdump_val;
+}
+
+JEMALLOC_ALWAYS_INLINE prof_tdata_t *
+prof_tdata_get(tsd_t *tsd, bool create) {
+	prof_tdata_t *tdata;
+
+	cassert(config_prof);
+
+	tdata = tsd_prof_tdata_get(tsd);
+	if (create) {
+		if (unlikely(tdata == NULL)) {
+			if (tsd_nominal(tsd)) {
+				tdata = prof_tdata_init(tsd);
+				tsd_prof_tdata_set(tsd, tdata);
+			}
+		} else if (unlikely(tdata->expired)) {
+			tdata = prof_tdata_reinit(tsd, tdata);
+			tsd_prof_tdata_set(tsd, tdata);
+		}
+		assert(tdata == NULL || tdata->attached);
+	}
+
+	return tdata;
+}
+
+JEMALLOC_ALWAYS_INLINE prof_tctx_t *
+prof_tctx_get(tsdn_t *tsdn, const void *ptr, alloc_ctx_t *alloc_ctx) {
+	cassert(config_prof);
+	assert(ptr != NULL);
+
+	return arena_prof_tctx_get(tsdn, ptr, alloc_ctx);
+}
+
+JEMALLOC_ALWAYS_INLINE void
+prof_tctx_set(tsdn_t *tsdn, const void *ptr, size_t usize,
+    alloc_ctx_t *alloc_ctx, prof_tctx_t *tctx) {
+	cassert(config_prof);
+	assert(ptr != NULL);
+
+	arena_prof_tctx_set(tsdn, ptr, usize, alloc_ctx, tctx);
+}
+
+JEMALLOC_ALWAYS_INLINE void
+prof_tctx_reset(tsdn_t *tsdn, const void *ptr, prof_tctx_t *tctx) {
+	cassert(config_prof);
+	assert(ptr != NULL);
+
+	arena_prof_tctx_reset(tsdn, ptr, tctx);
+}
+
+JEMALLOC_ALWAYS_INLINE nstime_t
+prof_alloc_time_get(tsdn_t *tsdn, const void *ptr, alloc_ctx_t *alloc_ctx) {
+	cassert(config_prof);
+	assert(ptr != NULL);
+
+	return arena_prof_alloc_time_get(tsdn, ptr, alloc_ctx);
+}
+
+JEMALLOC_ALWAYS_INLINE void
+prof_alloc_time_set(tsdn_t *tsdn, const void *ptr, alloc_ctx_t *alloc_ctx,
+    nstime_t t) {
+	cassert(config_prof);
+	assert(ptr != NULL);
+
+	arena_prof_alloc_time_set(tsdn, ptr, alloc_ctx, t);
+}
+
+JEMALLOC_ALWAYS_INLINE bool
+prof_sample_check(tsd_t *tsd, size_t usize, bool update) {
+	ssize_t check = update ? 0 : usize;
+
+	int64_t bytes_until_sample = tsd_bytes_until_sample_get(tsd);
+	if (update) {
+		bytes_until_sample -= usize;
+		if (tsd_nominal(tsd)) {
+			tsd_bytes_until_sample_set(tsd, bytes_until_sample);
+		}
+	}
+	if (likely(bytes_until_sample >= check)) {
+		return true;
+	}
+
+	return false;
+}
+
+JEMALLOC_ALWAYS_INLINE bool
+prof_sample_accum_update(tsd_t *tsd, size_t usize, bool update,
+			 prof_tdata_t **tdata_out) {
+	prof_tdata_t *tdata;
+
+	cassert(config_prof);
+
+	/* Fastpath: no need to load tdata */
+	if (likely(prof_sample_check(tsd, usize, update))) {
+		return true;
+	}
+
+	bool booted = tsd_prof_tdata_get(tsd);
+	tdata = prof_tdata_get(tsd, true);
+	if (unlikely((uintptr_t)tdata <= (uintptr_t)PROF_TDATA_STATE_MAX)) {
+		tdata = NULL;
+	}
+
+	if (tdata_out != NULL) {
+		*tdata_out = tdata;
+	}
+
+	if (unlikely(tdata == NULL)) {
+		return true;
+	}
+
+	/*
+	 * If this was the first creation of tdata, then
+	 * prof_tdata_get() reset bytes_until_sample, so decrement and
+	 * check it again
+	 */
+	if (!booted && prof_sample_check(tsd, usize, update)) {
+		return true;
+	}
+
+	if (tsd_reentrancy_level_get(tsd) > 0) {
+		return true;
+	}
+	/* Compute new sample threshold. */
+	if (update) {
+		prof_sample_threshold_update(tdata);
+	}
+	return !tdata->active;
+}
+
+JEMALLOC_ALWAYS_INLINE prof_tctx_t *
+prof_alloc_prep(tsd_t *tsd, size_t usize, bool prof_active, bool update) {
+	prof_tctx_t *ret;
+	prof_tdata_t *tdata;
+	prof_bt_t bt;
+
+	assert(usize == sz_s2u(usize));
+
+	if (!prof_active || likely(prof_sample_accum_update(tsd, usize, update,
+	    &tdata))) {
+		ret = (prof_tctx_t *)(uintptr_t)1U;
+	} else {
+		bt_init(&bt, tdata->vec);
+		prof_backtrace(&bt);
+		ret = prof_lookup(tsd, &bt);
+	}
+
+	return ret;
+}
+
+JEMALLOC_ALWAYS_INLINE void
+prof_malloc(tsdn_t *tsdn, const void *ptr, size_t usize, alloc_ctx_t *alloc_ctx,
+    prof_tctx_t *tctx) {
+	cassert(config_prof);
+	assert(ptr != NULL);
+	assert(usize == isalloc(tsdn, ptr));
+
+	if (unlikely((uintptr_t)tctx > (uintptr_t)1U)) {
+		prof_malloc_sample_object(tsdn, ptr, usize, tctx);
+	} else {
+		prof_tctx_set(tsdn, ptr, usize, alloc_ctx,
+		    (prof_tctx_t *)(uintptr_t)1U);
+	}
+}
+
+JEMALLOC_ALWAYS_INLINE void
+prof_realloc(tsd_t *tsd, const void *ptr, size_t usize, prof_tctx_t *tctx,
+    bool prof_active, bool updated, const void *old_ptr, size_t old_usize,
+    prof_tctx_t *old_tctx) {
+	bool sampled, old_sampled, moved;
+
+	cassert(config_prof);
+	assert(ptr != NULL || (uintptr_t)tctx <= (uintptr_t)1U);
+
+	if (prof_active && !updated && ptr != NULL) {
+		assert(usize == isalloc(tsd_tsdn(tsd), ptr));
+		if (prof_sample_accum_update(tsd, usize, true, NULL)) {
+			/*
+			 * Don't sample.  The usize passed to prof_alloc_prep()
+			 * was larger than what actually got allocated, so a
+			 * backtrace was captured for this allocation, even
+			 * though its actual usize was insufficient to cross the
+			 * sample threshold.
+			 */
+			prof_alloc_rollback(tsd, tctx, true);
+			tctx = (prof_tctx_t *)(uintptr_t)1U;
+		}
+	}
+
+	sampled = ((uintptr_t)tctx > (uintptr_t)1U);
+	old_sampled = ((uintptr_t)old_tctx > (uintptr_t)1U);
+	moved = (ptr != old_ptr);
+
+	if (unlikely(sampled)) {
+		prof_malloc_sample_object(tsd_tsdn(tsd), ptr, usize, tctx);
+	} else if (moved) {
+		prof_tctx_set(tsd_tsdn(tsd), ptr, usize, NULL,
+		    (prof_tctx_t *)(uintptr_t)1U);
+	} else if (unlikely(old_sampled)) {
+		/*
+		 * prof_tctx_set() would work for the !moved case as well, but
+		 * prof_tctx_reset() is slightly cheaper, and the proper thing
+		 * to do here in the presence of explicit knowledge re: moved
+		 * state.
+		 */
+		prof_tctx_reset(tsd_tsdn(tsd), ptr, tctx);
+	} else {
+		assert((uintptr_t)prof_tctx_get(tsd_tsdn(tsd), ptr, NULL) ==
+		    (uintptr_t)1U);
+	}
+
+	/*
+	 * The prof_free_sampled_object() call must come after the
+	 * prof_malloc_sample_object() call, because tctx and old_tctx may be
+	 * the same, in which case reversing the call order could cause the tctx
+	 * to be prematurely destroyed as a side effect of momentarily zeroed
+	 * counters.
+	 */
+	if (unlikely(old_sampled)) {
+		prof_free_sampled_object(tsd, ptr, old_usize, old_tctx);
+	}
+}
+
+JEMALLOC_ALWAYS_INLINE void
+prof_free(tsd_t *tsd, const void *ptr, size_t usize, alloc_ctx_t *alloc_ctx) {
+	prof_tctx_t *tctx = prof_tctx_get(tsd_tsdn(tsd), ptr, alloc_ctx);
+
+	cassert(config_prof);
+	assert(usize == isalloc(tsd_tsdn(tsd), ptr));
+
+	if (unlikely((uintptr_t)tctx > (uintptr_t)1U)) {
+		prof_free_sampled_object(tsd, ptr, usize, tctx);
+	}
+}
+
+#endif /* JEMALLOC_INTERNAL_PROF_INLINES_B_H */

deps/jemalloc/include/jemalloc/internal/prof_structs.h

@@ -0,0 +1,200 @@
+#ifndef JEMALLOC_INTERNAL_PROF_STRUCTS_H
+#define JEMALLOC_INTERNAL_PROF_STRUCTS_H
+
+#include "jemalloc/internal/ckh.h"
+#include "jemalloc/internal/mutex.h"
+#include "jemalloc/internal/prng.h"
+#include "jemalloc/internal/rb.h"
+
+struct prof_bt_s {
+	/* Backtrace, stored as len program counters. */
+	void		**vec;
+	unsigned	len;
+};
+
+#ifdef JEMALLOC_PROF_LIBGCC
+/* Data structure passed to libgcc _Unwind_Backtrace() callback functions. */
+typedef struct {
+	prof_bt_t	*bt;
+	unsigned	max;
+} prof_unwind_data_t;
+#endif
+
+struct prof_accum_s {
+#ifndef JEMALLOC_ATOMIC_U64
+	malloc_mutex_t	mtx;
+	uint64_t	accumbytes;
+#else
+	atomic_u64_t	accumbytes;
+#endif
+};
+
+struct prof_cnt_s {
+	/* Profiling counters. */
+	uint64_t	curobjs;
+	uint64_t	curbytes;
+	uint64_t	accumobjs;
+	uint64_t	accumbytes;
+};
+
+typedef enum {
+	prof_tctx_state_initializing,
+	prof_tctx_state_nominal,
+	prof_tctx_state_dumping,
+	prof_tctx_state_purgatory /* Dumper must finish destroying. */
+} prof_tctx_state_t;
+
+struct prof_tctx_s {
+	/* Thread data for thread that performed the allocation. */
+	prof_tdata_t		*tdata;
+
+	/*
+	 * Copy of tdata->thr_{uid,discrim}, necessary because tdata may be
+	 * defunct during teardown.
+	 */
+	uint64_t		thr_uid;
+	uint64_t		thr_discrim;
+
+	/* Profiling counters, protected by tdata->lock. */
+	prof_cnt_t		cnts;
+
+	/* Associated global context. */
+	prof_gctx_t		*gctx;
+
+	/*
+	 * UID that distinguishes multiple tctx's created by the same thread,
+	 * but coexisting in gctx->tctxs.  There are two ways that such
+	 * coexistence can occur:
+	 * - A dumper thread can cause a tctx to be retained in the purgatory
+	 *   state.
+	 * - Although a single "producer" thread must create all tctx's which
+	 *   share the same thr_uid, multiple "consumers" can each concurrently
+	 *   execute portions of prof_tctx_destroy().  prof_tctx_destroy() only
+	 *   gets called once each time cnts.cur{objs,bytes} drop to 0, but this
+	 *   threshold can be hit again before the first consumer finishes
+	 *   executing prof_tctx_destroy().
+	 */
+	uint64_t		tctx_uid;
+
+	/* Linkage into gctx's tctxs. */
+	rb_node(prof_tctx_t)	tctx_link;
+
+	/*
+	 * True during prof_alloc_prep()..prof_malloc_sample_object(), prevents
+	 * sample vs destroy race.
+	 */
+	bool			prepared;
+
+	/* Current dump-related state, protected by gctx->lock. */
+	prof_tctx_state_t	state;
+
+	/*
+	 * Copy of cnts snapshotted during early dump phase, protected by
+	 * dump_mtx.
+	 */
+	prof_cnt_t		dump_cnts;
+};
+typedef rb_tree(prof_tctx_t) prof_tctx_tree_t;
+
+struct prof_gctx_s {
+	/* Protects nlimbo, cnt_summed, and tctxs. */
+	malloc_mutex_t		*lock;
+
+	/*
+	 * Number of threads that currently cause this gctx to be in a state of
+	 * limbo due to one of:
+	 *   - Initializing this gctx.
+	 *   - Initializing per thread counters associated with this gctx.
+	 *   - Preparing to destroy this gctx.
+	 *   - Dumping a heap profile that includes this gctx.
+	 * nlimbo must be 1 (single destroyer) in order to safely destroy the
+	 * gctx.
+	 */
+	unsigned		nlimbo;
+
+	/*
+	 * Tree of profile counters, one for each thread that has allocated in
+	 * this context.
+	 */
+	prof_tctx_tree_t	tctxs;
+
+	/* Linkage for tree of contexts to be dumped. */
+	rb_node(prof_gctx_t)	dump_link;
+
+	/* Temporary storage for summation during dump. */
+	prof_cnt_t		cnt_summed;
+
+	/* Associated backtrace. */
+	prof_bt_t		bt;
+
+	/* Backtrace vector, variable size, referred to by bt. */
+	void			*vec[1];
+};
+typedef rb_tree(prof_gctx_t) prof_gctx_tree_t;
+
+struct prof_tdata_s {
+	malloc_mutex_t		*lock;
+
+	/* Monotonically increasing unique thread identifier. */
+	uint64_t		thr_uid;
+
+	/*
+	 * Monotonically increasing discriminator among tdata structures
+	 * associated with the same thr_uid.
+	 */
+	uint64_t		thr_discrim;
+
+	/* Included in heap profile dumps if non-NULL. */
+	char			*thread_name;
+
+	bool			attached;
+	bool			expired;
+
+	rb_node(prof_tdata_t)	tdata_link;
+
+	/*
+	 * Counter used to initialize prof_tctx_t's tctx_uid.  No locking is
+	 * necessary when incrementing this field, because only one thread ever
+	 * does so.
+	 */
+	uint64_t		tctx_uid_next;
+
+	/*
+	 * Hash of (prof_bt_t *)-->(prof_tctx_t *).  Each thread tracks
+	 * backtraces for which it has non-zero allocation/deallocation counters
+	 * associated with thread-specific prof_tctx_t objects.  Other threads
+	 * may write to prof_tctx_t contents when freeing associated objects.
+	 */
+	ckh_t			bt2tctx;
+
+	/* Sampling state. */
+	uint64_t		prng_state;
+
+	/* State used to avoid dumping while operating on prof internals. */
+	bool			enq;
+	bool			enq_idump;
+	bool			enq_gdump;
+
+	/*
+	 * Set to true during an early dump phase for tdata's which are
+	 * currently being dumped.  New threads' tdata's have this initialized
+	 * to false so that they aren't accidentally included in later dump
+	 * phases.
+	 */
+	bool			dumping;
+
+	/*
+	 * True if profiling is active for this tdata's thread
+	 * (thread.prof.active mallctl).
+	 */
+	bool			active;
+
+	/* Temporary storage for summation during dump. */
+	prof_cnt_t		cnt_summed;
+
+	/* Backtrace vector, used for calls to prof_backtrace(). */
+	void			*vec[PROF_BT_MAX];
+};
+typedef rb_tree(prof_tdata_t) prof_tdata_tree_t;
+
+#endif /* JEMALLOC_INTERNAL_PROF_STRUCTS_H */

deps/jemalloc/include/jemalloc/internal/prof_types.h

@@ -0,0 +1,56 @@
+#ifndef JEMALLOC_INTERNAL_PROF_TYPES_H
+#define JEMALLOC_INTERNAL_PROF_TYPES_H
+
+typedef struct prof_bt_s prof_bt_t;
+typedef struct prof_accum_s prof_accum_t;
+typedef struct prof_cnt_s prof_cnt_t;
+typedef struct prof_tctx_s prof_tctx_t;
+typedef struct prof_gctx_s prof_gctx_t;
+typedef struct prof_tdata_s prof_tdata_t;
+
+/* Option defaults. */
+#ifdef JEMALLOC_PROF
+#  define PROF_PREFIX_DEFAULT		"jeprof"
+#else
+#  define PROF_PREFIX_DEFAULT		""
+#endif
+#define LG_PROF_SAMPLE_DEFAULT		19
+#define LG_PROF_INTERVAL_DEFAULT	-1
+
+/*
+ * Hard limit on stack backtrace depth.  The version of prof_backtrace() that
+ * is based on __builtin_return_address() necessarily has a hard-coded number
+ * of backtrace frame handlers, and should be kept in sync with this setting.
+ */
+#define PROF_BT_MAX			128
+
+/* Initial hash table size. */
+#define PROF_CKH_MINITEMS		64
+
+/* Size of memory buffer to use when writing dump files. */
+#define PROF_DUMP_BUFSIZE		65536
+
+/* Size of stack-allocated buffer used by prof_printf(). */
+#define PROF_PRINTF_BUFSIZE		128
+
+/*
+ * Number of mutexes shared among all gctx's.  No space is allocated for these
+ * unless profiling is enabled, so it's okay to over-provision.
+ */
+#define PROF_NCTX_LOCKS			1024
+
+/*
+ * Number of mutexes shared among all tdata's.  No space is allocated for these
+ * unless profiling is enabled, so it's okay to over-provision.
+ */
+#define PROF_NTDATA_LOCKS		256
+
+/*
+ * prof_tdata pointers close to NULL are used to encode state information that
+ * is used for cleaning up during thread shutdown.
+ */
+#define PROF_TDATA_STATE_REINCARNATED	((prof_tdata_t *)(uintptr_t)1)
+#define PROF_TDATA_STATE_PURGATORY	((prof_tdata_t *)(uintptr_t)2)
+#define PROF_TDATA_STATE_MAX		PROF_TDATA_STATE_PURGATORY
+
+#endif /* JEMALLOC_INTERNAL_PROF_TYPES_H */

deps/jemalloc/include/jemalloc/internal/ql.h

@@ -0,0 +1,88 @@
+#ifndef JEMALLOC_INTERNAL_QL_H
+#define JEMALLOC_INTERNAL_QL_H
+
+#include "jemalloc/internal/qr.h"
+
+/* List definitions. */
+#define ql_head(a_type)							\
+struct {								\
+	a_type *qlh_first;						\
+}
+
+#define ql_head_initializer(a_head) {NULL}
+
+#define ql_elm(a_type)	qr(a_type)
+
+/* List functions. */
+#define ql_new(a_head) do {						\
+	(a_head)->qlh_first = NULL;					\
+} while (0)
+
+#define ql_elm_new(a_elm, a_field) qr_new((a_elm), a_field)
+
+#define ql_first(a_head) ((a_head)->qlh_first)
+
+#define ql_last(a_head, a_field)					\
+	((ql_first(a_head) != NULL)					\
+	    ? qr_prev(ql_first(a_head), a_field) : NULL)
+
+#define ql_next(a_head, a_elm, a_field)					\
+	((ql_last(a_head, a_field) != (a_elm))				\
+	    ? qr_next((a_elm), a_field)	: NULL)
+
+#define ql_prev(a_head, a_elm, a_field)					\
+	((ql_first(a_head) != (a_elm)) ? qr_prev((a_elm), a_field)	\
+				       : NULL)
+
+#define ql_before_insert(a_head, a_qlelm, a_elm, a_field) do {		\
+	qr_before_insert((a_qlelm), (a_elm), a_field);			\
+	if (ql_first(a_head) == (a_qlelm)) {				\
+		ql_first(a_head) = (a_elm);				\
+	}								\
+} while (0)
+
+#define ql_after_insert(a_qlelm, a_elm, a_field)			\
+	qr_after_insert((a_qlelm), (a_elm), a_field)
+
+#define ql_head_insert(a_head, a_elm, a_field) do {			\
+	if (ql_first(a_head) != NULL) {					\
+		qr_before_insert(ql_first(a_head), (a_elm), a_field);	\
+	}								\
+	ql_first(a_head) = (a_elm);					\
+} while (0)
+
+#define ql_tail_insert(a_head, a_elm, a_field) do {			\
+	if (ql_first(a_head) != NULL) {					\
+		qr_before_insert(ql_first(a_head), (a_elm), a_field);	\
+	}								\
+	ql_first(a_head) = qr_next((a_elm), a_field);			\
+} while (0)
+
+#define ql_remove(a_head, a_elm, a_field) do {				\
+	if (ql_first(a_head) == (a_elm)) {				\
+		ql_first(a_head) = qr_next(ql_first(a_head), a_field);	\
+	}								\
+	if (ql_first(a_head) != (a_elm)) {				\
+		qr_remove((a_elm), a_field);				\
+	} else {							\
+		ql_first(a_head) = NULL;				\
+	}								\
+} while (0)
+
+#define ql_head_remove(a_head, a_type, a_field) do {			\
+	a_type *t = ql_first(a_head);					\
+	ql_remove((a_head), t, a_field);				\
+} while (0)
+
+#define ql_tail_remove(a_head, a_type, a_field) do {			\
+	a_type *t = ql_last(a_head, a_field);				\
+	ql_remove((a_head), t, a_field);				\
+} while (0)
+
+#define ql_foreach(a_var, a_head, a_field)				\
+	qr_foreach((a_var), ql_first(a_head), a_field)
+
+#define ql_reverse_foreach(a_var, a_head, a_field)			\
+	qr_reverse_foreach((a_var), ql_first(a_head), a_field)
+
+#endif /* JEMALLOC_INTERNAL_QL_H */

deps/jemalloc/include/jemalloc/internal/qr.h

@@ -0,0 +1,72 @@
+#ifndef JEMALLOC_INTERNAL_QR_H
+#define JEMALLOC_INTERNAL_QR_H
+
+/* Ring definitions. */
+#define qr(a_type)							\
+struct {								\
+	a_type	*qre_next;						\
+	a_type	*qre_prev;						\
+}
+
+/* Ring functions. */
+#define qr_new(a_qr, a_field) do {					\
+	(a_qr)->a_field.qre_next = (a_qr);				\
+	(a_qr)->a_field.qre_prev = (a_qr);				\
+} while (0)
+
+#define qr_next(a_qr, a_field) ((a_qr)->a_field.qre_next)
+
+#define qr_prev(a_qr, a_field) ((a_qr)->a_field.qre_prev)
+
+#define qr_before_insert(a_qrelm, a_qr, a_field) do {			\
+	(a_qr)->a_field.qre_prev = (a_qrelm)->a_field.qre_prev;		\
+	(a_qr)->a_field.qre_next = (a_qrelm);				\
+	(a_qr)->a_field.qre_prev->a_field.qre_next = (a_qr);		\
+	(a_qrelm)->a_field.qre_prev = (a_qr);				\
+} while (0)
+
+#define qr_after_insert(a_qrelm, a_qr, a_field) do {			\
+	(a_qr)->a_field.qre_next = (a_qrelm)->a_field.qre_next;		\
+	(a_qr)->a_field.qre_prev = (a_qrelm);				\
+	(a_qr)->a_field.qre_next->a_field.qre_prev = (a_qr);		\
+	(a_qrelm)->a_field.qre_next = (a_qr);				\
+} while (0)
+
+#define qr_meld(a_qr_a, a_qr_b, a_type, a_field) do {			\
+	a_type *t;							\
+	(a_qr_a)->a_field.qre_prev->a_field.qre_next = (a_qr_b);	\
+	(a_qr_b)->a_field.qre_prev->a_field.qre_next = (a_qr_a);	\
+	t = (a_qr_a)->a_field.qre_prev;					\
+	(a_qr_a)->a_field.qre_prev = (a_qr_b)->a_field.qre_prev;	\
+	(a_qr_b)->a_field.qre_prev = t;					\
+} while (0)
+
+/*
+ * qr_meld() and qr_split() are functionally equivalent, so there's no need to
+ * have two copies of the code.
+ */
+#define qr_split(a_qr_a, a_qr_b, a_type, a_field)			\
+	qr_meld((a_qr_a), (a_qr_b), a_type, a_field)
+
+#define qr_remove(a_qr, a_field) do {					\
+	(a_qr)->a_field.qre_prev->a_field.qre_next			\
+	    = (a_qr)->a_field.qre_next;					\
+	(a_qr)->a_field.qre_next->a_field.qre_prev			\
+	    = (a_qr)->a_field.qre_prev;					\
+	(a_qr)->a_field.qre_next = (a_qr);				\
+	(a_qr)->a_field.qre_prev = (a_qr);				\
+} while (0)
+
+#define qr_foreach(var, a_qr, a_field)					\
+	for ((var) = (a_qr);						\
+	    (var) != NULL;						\
+	    (var) = (((var)->a_field.qre_next != (a_qr))		\
+	    ? (var)->a_field.qre_next : NULL))
+
+#define qr_reverse_foreach(var, a_qr, a_field)				\
+	for ((var) = ((a_qr) != NULL) ? qr_prev(a_qr, a_field) : NULL;	\
+	    (var) != NULL;						\
+	    (var) = (((var) != (a_qr))					\
+	    ? (var)->a_field.qre_prev : NULL))
+
+#endif /* JEMALLOC_INTERNAL_QR_H */

deps/jemalloc/include/jemalloc/internal/quantum.h

@@ -0,0 +1,77 @@
+#ifndef JEMALLOC_INTERNAL_QUANTUM_H
+#define JEMALLOC_INTERNAL_QUANTUM_H
+
+/*
+ * Minimum allocation alignment is 2^LG_QUANTUM bytes (ignoring tiny size
+ * classes).
+ */
+#ifndef LG_QUANTUM
+#  if (defined(__i386__) || defined(_M_IX86))
+#    define LG_QUANTUM		4
+#  endif
+#  ifdef __ia64__
+#    define LG_QUANTUM		4
+#  endif
+#  ifdef __alpha__
+#    define LG_QUANTUM		4
+#  endif
+#  if (defined(__sparc64__) || defined(__sparcv9) || defined(__sparc_v9__))
+#    define LG_QUANTUM		4
+#  endif
+#  if (defined(__amd64__) || defined(__x86_64__) || defined(_M_X64))
+#    define LG_QUANTUM		4
+#  endif
+#  ifdef __arm__
+#    define LG_QUANTUM		3
+#  endif
+#  ifdef __aarch64__
+#    define LG_QUANTUM		4
+#  endif
+#  ifdef __hppa__
+#    define LG_QUANTUM		4
+#  endif
+#  ifdef __m68k__
+#    define LG_QUANTUM		3
+#  endif
+#  ifdef __mips__
+#    define LG_QUANTUM		3
+#  endif
+#  ifdef __nios2__
+#    define LG_QUANTUM		3
+#  endif
+#  ifdef __or1k__
+#    define LG_QUANTUM		3
+#  endif
+#  ifdef __powerpc__
+#    define LG_QUANTUM		4
+#  endif
+#  if defined(__riscv) || defined(__riscv__)
+#    define LG_QUANTUM		4
+#  endif
+#  ifdef __s390__
+#    define LG_QUANTUM		4
+#  endif
+#  if (defined (__SH3E__) || defined(__SH4_SINGLE__) || defined(__SH4__) || \
+	defined(__SH4_SINGLE_ONLY__))
+#    define LG_QUANTUM		4
+#  endif
+#  ifdef __tile__
+#    define LG_QUANTUM		4
+#  endif
+#  ifdef __le32__
+#    define LG_QUANTUM		4
+#  endif
+#  ifndef LG_QUANTUM
+#    error "Unknown minimum alignment for architecture; specify via "
+	 "--with-lg-quantum"
+#  endif
+#endif
+
+#define QUANTUM			((size_t)(1U << LG_QUANTUM))
+#define QUANTUM_MASK		(QUANTUM - 1)
+
+/* Return the smallest quantum multiple that is >= a. */
+#define QUANTUM_CEILING(a)						\
+	(((a) + QUANTUM_MASK) & ~QUANTUM_MASK)
+
+#endif /* JEMALLOC_INTERNAL_QUANTUM_H */

deps/jemalloc/include/jemalloc/internal/quarantine.h

@@ -0,0 +1,67 @@
+/******************************************************************************/
+#ifdef JEMALLOC_H_TYPES
+
+typedef struct quarantine_obj_s quarantine_obj_t;
+typedef struct quarantine_s quarantine_t;
+
+/* Default per thread quarantine size if valgrind is enabled. */
+#define	JEMALLOC_VALGRIND_QUARANTINE_DEFAULT	(ZU(1) << 24)
+
+#endif /* JEMALLOC_H_TYPES */
+/******************************************************************************/
+#ifdef JEMALLOC_H_STRUCTS
+
+struct quarantine_obj_s {
+	void	*ptr;
+	size_t	usize;
+};
+
+struct quarantine_s {
+	size_t			curbytes;
+	size_t			curobjs;
+	size_t			first;
+#define	LG_MAXOBJS_INIT 10
+	size_t			lg_maxobjs;
+	quarantine_obj_t	objs[1]; /* Dynamically sized ring buffer. */
+};
+
+#endif /* JEMALLOC_H_STRUCTS */
+/******************************************************************************/
+#ifdef JEMALLOC_H_EXTERNS
+
+quarantine_t	*quarantine_init(size_t lg_maxobjs);
+void	quarantine(void *ptr);
+void	quarantine_cleanup(void *arg);
+bool	quarantine_boot(void);
+
+#endif /* JEMALLOC_H_EXTERNS */
+/******************************************************************************/
+#ifdef JEMALLOC_H_INLINES
+
+#ifndef JEMALLOC_ENABLE_INLINE
+malloc_tsd_protos(JEMALLOC_ATTR(unused), quarantine, quarantine_t *)
+
+void	quarantine_alloc_hook(void);
+#endif
+
+#if (defined(JEMALLOC_ENABLE_INLINE) || defined(JEMALLOC_QUARANTINE_C_))
+malloc_tsd_externs(quarantine, quarantine_t *)
+malloc_tsd_funcs(JEMALLOC_ALWAYS_INLINE, quarantine, quarantine_t *, NULL,
+    quarantine_cleanup)
+
+JEMALLOC_ALWAYS_INLINE void
+quarantine_alloc_hook(void)
+{
+	quarantine_t *quarantine;
+
+	assert(config_fill && opt_quarantine);
+
+	quarantine = *quarantine_tsd_get();
+	if (quarantine == NULL)
+		quarantine_init(LG_MAXOBJS_INIT);
+}
+#endif
+
+#endif /* JEMALLOC_H_INLINES */
+/******************************************************************************/
+

deps/jemalloc/include/jemalloc/internal/rtree.h

@@ -0,0 +1,528 @@
+#ifndef JEMALLOC_INTERNAL_RTREE_H
+#define JEMALLOC_INTERNAL_RTREE_H
+
+#include "jemalloc/internal/atomic.h"
+#include "jemalloc/internal/mutex.h"
+#include "jemalloc/internal/rtree_tsd.h"
+#include "jemalloc/internal/sc.h"
+#include "jemalloc/internal/tsd.h"
+
+/*
+ * This radix tree implementation is tailored to the singular purpose of
+ * associating metadata with extents that are currently owned by jemalloc.
+ *
+ *******************************************************************************
+ */
+
+/* Number of high insignificant bits. */
+#define RTREE_NHIB ((1U << (LG_SIZEOF_PTR+3)) - LG_VADDR)
+/* Number of low insigificant bits. */
+#define RTREE_NLIB LG_PAGE
+/* Number of significant bits. */
+#define RTREE_NSB (LG_VADDR - RTREE_NLIB)
+/* Number of levels in radix tree. */
+#if RTREE_NSB <= 10
+#  define RTREE_HEIGHT 1
+#elif RTREE_NSB <= 36
+#  define RTREE_HEIGHT 2
+#elif RTREE_NSB <= 52
+#  define RTREE_HEIGHT 3
+#else
+#  error Unsupported number of significant virtual address bits
+#endif
+/* Use compact leaf representation if virtual address encoding allows. */
+#if RTREE_NHIB >= LG_CEIL(SC_NSIZES)
+#  define RTREE_LEAF_COMPACT
+#endif
+
+/* Needed for initialization only. */
+#define RTREE_LEAFKEY_INVALID ((uintptr_t)1)
+
+typedef struct rtree_node_elm_s rtree_node_elm_t;
+struct rtree_node_elm_s {
+	atomic_p_t	child; /* (rtree_{node,leaf}_elm_t *) */
+};
+
+struct rtree_leaf_elm_s {
+#ifdef RTREE_LEAF_COMPACT
+	/*
+	 * Single pointer-width field containing all three leaf element fields.
+	 * For example, on a 64-bit x64 system with 48 significant virtual
+	 * memory address bits, the index, extent, and slab fields are packed as
+	 * such:
+	 *
+	 * x: index
+	 * e: extent
+	 * b: slab
+	 *
+	 *   00000000 xxxxxxxx eeeeeeee [...] eeeeeeee eeee000b
+	 */
+	atomic_p_t	le_bits;
+#else
+	atomic_p_t	le_extent; /* (extent_t *) */
+	atomic_u_t	le_szind; /* (szind_t) */
+	atomic_b_t	le_slab; /* (bool) */
+#endif
+};
+
+typedef struct rtree_level_s rtree_level_t;
+struct rtree_level_s {
+	/* Number of key bits distinguished by this level. */
+	unsigned		bits;
+	/*
+	 * Cumulative number of key bits distinguished by traversing to
+	 * corresponding tree level.
+	 */
+	unsigned		cumbits;
+};
+
+typedef struct rtree_s rtree_t;
+struct rtree_s {
+	malloc_mutex_t		init_lock;
+	/* Number of elements based on rtree_levels[0].bits. */
+#if RTREE_HEIGHT > 1
+	rtree_node_elm_t	root[1U << (RTREE_NSB/RTREE_HEIGHT)];
+#else
+	rtree_leaf_elm_t	root[1U << (RTREE_NSB/RTREE_HEIGHT)];
+#endif
+};
+
+/*
+ * Split the bits into one to three partitions depending on number of
+ * significant bits.  It the number of bits does not divide evenly into the
+ * number of levels, place one remainder bit per level starting at the leaf
+ * level.
+ */
+static const rtree_level_t rtree_levels[] = {
+#if RTREE_HEIGHT == 1
+	{RTREE_NSB, RTREE_NHIB + RTREE_NSB}
+#elif RTREE_HEIGHT == 2
+	{RTREE_NSB/2, RTREE_NHIB + RTREE_NSB/2},
+	{RTREE_NSB/2 + RTREE_NSB%2, RTREE_NHIB + RTREE_NSB}
+#elif RTREE_HEIGHT == 3
+	{RTREE_NSB/3, RTREE_NHIB + RTREE_NSB/3},
+	{RTREE_NSB/3 + RTREE_NSB%3/2,
+	    RTREE_NHIB + RTREE_NSB/3*2 + RTREE_NSB%3/2},
+	{RTREE_NSB/3 + RTREE_NSB%3 - RTREE_NSB%3/2, RTREE_NHIB + RTREE_NSB}
+#else
+#  error Unsupported rtree height
+#endif
+};
+
+bool rtree_new(rtree_t *rtree, bool zeroed);
+
+typedef rtree_node_elm_t *(rtree_node_alloc_t)(tsdn_t *, rtree_t *, size_t);
+extern rtree_node_alloc_t *JET_MUTABLE rtree_node_alloc;
+
+typedef rtree_leaf_elm_t *(rtree_leaf_alloc_t)(tsdn_t *, rtree_t *, size_t);
+extern rtree_leaf_alloc_t *JET_MUTABLE rtree_leaf_alloc;
+
+typedef void (rtree_node_dalloc_t)(tsdn_t *, rtree_t *, rtree_node_elm_t *);
+extern rtree_node_dalloc_t *JET_MUTABLE rtree_node_dalloc;
+
+typedef void (rtree_leaf_dalloc_t)(tsdn_t *, rtree_t *, rtree_leaf_elm_t *);
+extern rtree_leaf_dalloc_t *JET_MUTABLE rtree_leaf_dalloc;
+#ifdef JEMALLOC_JET
+void rtree_delete(tsdn_t *tsdn, rtree_t *rtree);
+#endif
+rtree_leaf_elm_t *rtree_leaf_elm_lookup_hard(tsdn_t *tsdn, rtree_t *rtree,
+    rtree_ctx_t *rtree_ctx, uintptr_t key, bool dependent, bool init_missing);
+
+JEMALLOC_ALWAYS_INLINE uintptr_t
+rtree_leafkey(uintptr_t key) {
+	unsigned ptrbits = ZU(1) << (LG_SIZEOF_PTR+3);
+	unsigned cumbits = (rtree_levels[RTREE_HEIGHT-1].cumbits -
+	    rtree_levels[RTREE_HEIGHT-1].bits);
+	unsigned maskbits = ptrbits - cumbits;
+	uintptr_t mask = ~((ZU(1) << maskbits) - 1);
+	return (key & mask);
+}
+
+JEMALLOC_ALWAYS_INLINE size_t
+rtree_cache_direct_map(uintptr_t key) {
+	unsigned ptrbits = ZU(1) << (LG_SIZEOF_PTR+3);
+	unsigned cumbits = (rtree_levels[RTREE_HEIGHT-1].cumbits -
+	    rtree_levels[RTREE_HEIGHT-1].bits);
+	unsigned maskbits = ptrbits - cumbits;
+	return (size_t)((key >> maskbits) & (RTREE_CTX_NCACHE - 1));
+}
+
+JEMALLOC_ALWAYS_INLINE uintptr_t
+rtree_subkey(uintptr_t key, unsigned level) {
+	unsigned ptrbits = ZU(1) << (LG_SIZEOF_PTR+3);
+	unsigned cumbits = rtree_levels[level].cumbits;
+	unsigned shiftbits = ptrbits - cumbits;
+	unsigned maskbits = rtree_levels[level].bits;
+	uintptr_t mask = (ZU(1) << maskbits) - 1;
+	return ((key >> shiftbits) & mask);
+}
+
+/*
+ * Atomic getters.
+ *
+ * dependent: Reading a value on behalf of a pointer to a valid allocation
+ *            is guaranteed to be a clean read even without synchronization,
+ *            because the rtree update became visible in memory before the
+ *            pointer came into existence.
+ * !dependent: An arbitrary read, e.g. on behalf of ivsalloc(), may not be
+ *             dependent on a previous rtree write, which means a stale read
+ *             could result if synchronization were omitted here.
+ */
+#  ifdef RTREE_LEAF_COMPACT
+JEMALLOC_ALWAYS_INLINE uintptr_t
+rtree_leaf_elm_bits_read(tsdn_t *tsdn, rtree_t *rtree,
+    rtree_leaf_elm_t *elm, bool dependent) {
+	return (uintptr_t)atomic_load_p(&elm->le_bits, dependent
+	    ? ATOMIC_RELAXED : ATOMIC_ACQUIRE);
+}
+
+JEMALLOC_ALWAYS_INLINE extent_t *
+rtree_leaf_elm_bits_extent_get(uintptr_t bits) {
+#    ifdef __aarch64__
+	/*
+	 * aarch64 doesn't sign extend the highest virtual address bit to set
+	 * the higher ones.  Instead, the high bits gets zeroed.
+	 */
+	uintptr_t high_bit_mask = ((uintptr_t)1 << LG_VADDR) - 1;
+	/* Mask off the slab bit. */
+	uintptr_t low_bit_mask = ~(uintptr_t)1;
+	uintptr_t mask = high_bit_mask & low_bit_mask;
+	return (extent_t *)(bits & mask);
+#    else
+	/* Restore sign-extended high bits, mask slab bit. */
+	return (extent_t *)((uintptr_t)((intptr_t)(bits << RTREE_NHIB) >>
+	    RTREE_NHIB) & ~((uintptr_t)0x1));
+#    endif
+}
+
+JEMALLOC_ALWAYS_INLINE szind_t
+rtree_leaf_elm_bits_szind_get(uintptr_t bits) {
+	return (szind_t)(bits >> LG_VADDR);
+}
+
+JEMALLOC_ALWAYS_INLINE bool
+rtree_leaf_elm_bits_slab_get(uintptr_t bits) {
+	return (bool)(bits & (uintptr_t)0x1);
+}
+
+#  endif
+
+JEMALLOC_ALWAYS_INLINE extent_t *
+rtree_leaf_elm_extent_read(tsdn_t *tsdn, rtree_t *rtree,
+    rtree_leaf_elm_t *elm, bool dependent) {
+#ifdef RTREE_LEAF_COMPACT
+	uintptr_t bits = rtree_leaf_elm_bits_read(tsdn, rtree, elm, dependent);
+	return rtree_leaf_elm_bits_extent_get(bits);
+#else
+	extent_t *extent = (extent_t *)atomic_load_p(&elm->le_extent, dependent
+	    ? ATOMIC_RELAXED : ATOMIC_ACQUIRE);
+	return extent;
+#endif
+}
+
+JEMALLOC_ALWAYS_INLINE szind_t
+rtree_leaf_elm_szind_read(tsdn_t *tsdn, rtree_t *rtree,
+    rtree_leaf_elm_t *elm, bool dependent) {
+#ifdef RTREE_LEAF_COMPACT
+	uintptr_t bits = rtree_leaf_elm_bits_read(tsdn, rtree, elm, dependent);
+	return rtree_leaf_elm_bits_szind_get(bits);
+#else
+	return (szind_t)atomic_load_u(&elm->le_szind, dependent ? ATOMIC_RELAXED
+	    : ATOMIC_ACQUIRE);
+#endif
+}
+
+JEMALLOC_ALWAYS_INLINE bool
+rtree_leaf_elm_slab_read(tsdn_t *tsdn, rtree_t *rtree,
+    rtree_leaf_elm_t *elm, bool dependent) {
+#ifdef RTREE_LEAF_COMPACT
+	uintptr_t bits = rtree_leaf_elm_bits_read(tsdn, rtree, elm, dependent);
+	return rtree_leaf_elm_bits_slab_get(bits);
+#else
+	return atomic_load_b(&elm->le_slab, dependent ? ATOMIC_RELAXED :
+	    ATOMIC_ACQUIRE);
+#endif
+}
+
+static inline void
+rtree_leaf_elm_extent_write(tsdn_t *tsdn, rtree_t *rtree,
+    rtree_leaf_elm_t *elm, extent_t *extent) {
+#ifdef RTREE_LEAF_COMPACT
+	uintptr_t old_bits = rtree_leaf_elm_bits_read(tsdn, rtree, elm, true);
+	uintptr_t bits = ((uintptr_t)rtree_leaf_elm_bits_szind_get(old_bits) <<
+	    LG_VADDR) | ((uintptr_t)extent & (((uintptr_t)0x1 << LG_VADDR) - 1))
+	    | ((uintptr_t)rtree_leaf_elm_bits_slab_get(old_bits));
+	atomic_store_p(&elm->le_bits, (void *)bits, ATOMIC_RELEASE);
+#else
+	atomic_store_p(&elm->le_extent, extent, ATOMIC_RELEASE);
+#endif
+}
+
+static inline void
+rtree_leaf_elm_szind_write(tsdn_t *tsdn, rtree_t *rtree,
+    rtree_leaf_elm_t *elm, szind_t szind) {
+	assert(szind <= SC_NSIZES);
+
+#ifdef RTREE_LEAF_COMPACT
+	uintptr_t old_bits = rtree_leaf_elm_bits_read(tsdn, rtree, elm,
+	    true);
+	uintptr_t bits = ((uintptr_t)szind << LG_VADDR) |
+	    ((uintptr_t)rtree_leaf_elm_bits_extent_get(old_bits) &
+	    (((uintptr_t)0x1 << LG_VADDR) - 1)) |
+	    ((uintptr_t)rtree_leaf_elm_bits_slab_get(old_bits));
+	atomic_store_p(&elm->le_bits, (void *)bits, ATOMIC_RELEASE);
+#else
+	atomic_store_u(&elm->le_szind, szind, ATOMIC_RELEASE);
+#endif
+}
+
+static inline void
+rtree_leaf_elm_slab_write(tsdn_t *tsdn, rtree_t *rtree,
+    rtree_leaf_elm_t *elm, bool slab) {
+#ifdef RTREE_LEAF_COMPACT
+	uintptr_t old_bits = rtree_leaf_elm_bits_read(tsdn, rtree, elm,
+	    true);
+	uintptr_t bits = ((uintptr_t)rtree_leaf_elm_bits_szind_get(old_bits) <<
+	    LG_VADDR) | ((uintptr_t)rtree_leaf_elm_bits_extent_get(old_bits) &
+	    (((uintptr_t)0x1 << LG_VADDR) - 1)) | ((uintptr_t)slab);
+	atomic_store_p(&elm->le_bits, (void *)bits, ATOMIC_RELEASE);
+#else
+	atomic_store_b(&elm->le_slab, slab, ATOMIC_RELEASE);
+#endif
+}
+
+static inline void
+rtree_leaf_elm_write(tsdn_t *tsdn, rtree_t *rtree,
+    rtree_leaf_elm_t *elm, extent_t *extent, szind_t szind, bool slab) {
+#ifdef RTREE_LEAF_COMPACT
+	uintptr_t bits = ((uintptr_t)szind << LG_VADDR) |
+	    ((uintptr_t)extent & (((uintptr_t)0x1 << LG_VADDR) - 1)) |
+	    ((uintptr_t)slab);
+	atomic_store_p(&elm->le_bits, (void *)bits, ATOMIC_RELEASE);
+#else
+	rtree_leaf_elm_slab_write(tsdn, rtree, elm, slab);
+	rtree_leaf_elm_szind_write(tsdn, rtree, elm, szind);
+	/*
+	 * Write extent last, since the element is atomically considered valid
+	 * as soon as the extent field is non-NULL.
+	 */
+	rtree_leaf_elm_extent_write(tsdn, rtree, elm, extent);
+#endif
+}
+
+static inline void
+rtree_leaf_elm_szind_slab_update(tsdn_t *tsdn, rtree_t *rtree,
+    rtree_leaf_elm_t *elm, szind_t szind, bool slab) {
+	assert(!slab || szind < SC_NBINS);
+
+	/*
+	 * The caller implicitly assures that it is the only writer to the szind
+	 * and slab fields, and that the extent field cannot currently change.
+	 */
+	rtree_leaf_elm_slab_write(tsdn, rtree, elm, slab);
+	rtree_leaf_elm_szind_write(tsdn, rtree, elm, szind);
+}
+
+JEMALLOC_ALWAYS_INLINE rtree_leaf_elm_t *
+rtree_leaf_elm_lookup(tsdn_t *tsdn, rtree_t *rtree, rtree_ctx_t *rtree_ctx,
+    uintptr_t key, bool dependent, bool init_missing) {
+	assert(key != 0);
+	assert(!dependent || !init_missing);
+
+	size_t slot = rtree_cache_direct_map(key);
+	uintptr_t leafkey = rtree_leafkey(key);
+	assert(leafkey != RTREE_LEAFKEY_INVALID);
+
+	/* Fast path: L1 direct mapped cache. */
+	if (likely(rtree_ctx->cache[slot].leafkey == leafkey)) {
+		rtree_leaf_elm_t *leaf = rtree_ctx->cache[slot].leaf;
+		assert(leaf != NULL);
+		uintptr_t subkey = rtree_subkey(key, RTREE_HEIGHT-1);
+		return &leaf[subkey];
+	}
+	/*
+	 * Search the L2 LRU cache.  On hit, swap the matching element into the
+	 * slot in L1 cache, and move the position in L2 up by 1.
+	 */
+#define RTREE_CACHE_CHECK_L2(i) do {					\
+	if (likely(rtree_ctx->l2_cache[i].leafkey == leafkey)) {	\
+		rtree_leaf_elm_t *leaf = rtree_ctx->l2_cache[i].leaf;	\
+		assert(leaf != NULL);					\
+		if (i > 0) {						\
+			/* Bubble up by one. */				\
+			rtree_ctx->l2_cache[i].leafkey =		\
+				rtree_ctx->l2_cache[i - 1].leafkey;	\
+			rtree_ctx->l2_cache[i].leaf =			\
+				rtree_ctx->l2_cache[i - 1].leaf;	\
+			rtree_ctx->l2_cache[i - 1].leafkey =		\
+			    rtree_ctx->cache[slot].leafkey;		\
+			rtree_ctx->l2_cache[i - 1].leaf =		\
+			    rtree_ctx->cache[slot].leaf;		\
+		} else {						\
+			rtree_ctx->l2_cache[0].leafkey =		\
+			    rtree_ctx->cache[slot].leafkey;		\
+			rtree_ctx->l2_cache[0].leaf =			\
+			    rtree_ctx->cache[slot].leaf;		\
+		}							\
+		rtree_ctx->cache[slot].leafkey = leafkey;		\
+		rtree_ctx->cache[slot].leaf = leaf;			\
+		uintptr_t subkey = rtree_subkey(key, RTREE_HEIGHT-1);	\
+		return &leaf[subkey];					\
+	}								\
+} while (0)
+	/* Check the first cache entry. */
+	RTREE_CACHE_CHECK_L2(0);
+	/* Search the remaining cache elements. */
+	for (unsigned i = 1; i < RTREE_CTX_NCACHE_L2; i++) {
+		RTREE_CACHE_CHECK_L2(i);
+	}
+#undef RTREE_CACHE_CHECK_L2
+
+	return rtree_leaf_elm_lookup_hard(tsdn, rtree, rtree_ctx, key,
+	    dependent, init_missing);
+}
+
+static inline bool
+rtree_write(tsdn_t *tsdn, rtree_t *rtree, rtree_ctx_t *rtree_ctx, uintptr_t key,
+    extent_t *extent, szind_t szind, bool slab) {
+	/* Use rtree_clear() to set the extent to NULL. */
+	assert(extent != NULL);
+
+	rtree_leaf_elm_t *elm = rtree_leaf_elm_lookup(tsdn, rtree, rtree_ctx,
+	    key, false, true);
+	if (elm == NULL) {
+		return true;
+	}
+
+	assert(rtree_leaf_elm_extent_read(tsdn, rtree, elm, false) == NULL);
+	rtree_leaf_elm_write(tsdn, rtree, elm, extent, szind, slab);
+
+	return false;
+}
+
+JEMALLOC_ALWAYS_INLINE rtree_leaf_elm_t *
+rtree_read(tsdn_t *tsdn, rtree_t *rtree, rtree_ctx_t *rtree_ctx, uintptr_t key,
+    bool dependent) {
+	rtree_leaf_elm_t *elm = rtree_leaf_elm_lookup(tsdn, rtree, rtree_ctx,
+	    key, dependent, false);
+	if (!dependent && elm == NULL) {
+		return NULL;
+	}
+	assert(elm != NULL);
+	return elm;
+}
+
+JEMALLOC_ALWAYS_INLINE extent_t *
+rtree_extent_read(tsdn_t *tsdn, rtree_t *rtree, rtree_ctx_t *rtree_ctx,
+    uintptr_t key, bool dependent) {
+	rtree_leaf_elm_t *elm = rtree_read(tsdn, rtree, rtree_ctx, key,
+	    dependent);
+	if (!dependent && elm == NULL) {
+		return NULL;
+	}
+	return rtree_leaf_elm_extent_read(tsdn, rtree, elm, dependent);
+}
+
+JEMALLOC_ALWAYS_INLINE szind_t
+rtree_szind_read(tsdn_t *tsdn, rtree_t *rtree, rtree_ctx_t *rtree_ctx,
+    uintptr_t key, bool dependent) {
+	rtree_leaf_elm_t *elm = rtree_read(tsdn, rtree, rtree_ctx, key,
+	    dependent);
+	if (!dependent && elm == NULL) {
+		return SC_NSIZES;
+	}
+	return rtree_leaf_elm_szind_read(tsdn, rtree, elm, dependent);
+}
+
+/*
+ * rtree_slab_read() is intentionally omitted because slab is always read in
+ * conjunction with szind, which makes rtree_szind_slab_read() a better choice.
+ */
+
+JEMALLOC_ALWAYS_INLINE bool
+rtree_extent_szind_read(tsdn_t *tsdn, rtree_t *rtree, rtree_ctx_t *rtree_ctx,
+    uintptr_t key, bool dependent, extent_t **r_extent, szind_t *r_szind) {
+	rtree_leaf_elm_t *elm = rtree_read(tsdn, rtree, rtree_ctx, key,
+	    dependent);
+	if (!dependent && elm == NULL) {
+		return true;
+	}
+	*r_extent = rtree_leaf_elm_extent_read(tsdn, rtree, elm, dependent);
+	*r_szind = rtree_leaf_elm_szind_read(tsdn, rtree, elm, dependent);
+	return false;
+}
+
+/*
+ * Try to read szind_slab from the L1 cache.  Returns true on a hit,
+ * and fills in r_szind and r_slab.  Otherwise returns false.
+ *
+ * Key is allowed to be NULL in order to save an extra branch on the
+ * fastpath.  returns false in this case.
+ */
+JEMALLOC_ALWAYS_INLINE bool
+rtree_szind_slab_read_fast(tsdn_t *tsdn, rtree_t *rtree, rtree_ctx_t *rtree_ctx,
+			    uintptr_t key, szind_t *r_szind, bool *r_slab) {
+	rtree_leaf_elm_t *elm;
+
+	size_t slot = rtree_cache_direct_map(key);
+	uintptr_t leafkey = rtree_leafkey(key);
+	assert(leafkey != RTREE_LEAFKEY_INVALID);
+
+	if (likely(rtree_ctx->cache[slot].leafkey == leafkey)) {
+		rtree_leaf_elm_t *leaf = rtree_ctx->cache[slot].leaf;
+		assert(leaf != NULL);
+		uintptr_t subkey = rtree_subkey(key, RTREE_HEIGHT-1);
+		elm = &leaf[subkey];
+
+#ifdef RTREE_LEAF_COMPACT
+		uintptr_t bits = rtree_leaf_elm_bits_read(tsdn, rtree,
+							  elm, true);
+		*r_szind = rtree_leaf_elm_bits_szind_get(bits);
+		*r_slab = rtree_leaf_elm_bits_slab_get(bits);
+#else
+		*r_szind = rtree_leaf_elm_szind_read(tsdn, rtree, elm, true);
+		*r_slab = rtree_leaf_elm_slab_read(tsdn, rtree, elm, true);
+#endif
+		return true;
+	} else {
+		return false;
+	}
+}
+JEMALLOC_ALWAYS_INLINE bool
+rtree_szind_slab_read(tsdn_t *tsdn, rtree_t *rtree, rtree_ctx_t *rtree_ctx,
+    uintptr_t key, bool dependent, szind_t *r_szind, bool *r_slab) {
+	rtree_leaf_elm_t *elm = rtree_read(tsdn, rtree, rtree_ctx, key,
+	    dependent);
+	if (!dependent && elm == NULL) {
+		return true;
+	}
+#ifdef RTREE_LEAF_COMPACT
+	uintptr_t bits = rtree_leaf_elm_bits_read(tsdn, rtree, elm, dependent);
+	*r_szind = rtree_leaf_elm_bits_szind_get(bits);
+	*r_slab = rtree_leaf_elm_bits_slab_get(bits);
+#else
+	*r_szind = rtree_leaf_elm_szind_read(tsdn, rtree, elm, dependent);
+	*r_slab = rtree_leaf_elm_slab_read(tsdn, rtree, elm, dependent);
+#endif
+	return false;
+}
+
+static inline void
+rtree_szind_slab_update(tsdn_t *tsdn, rtree_t *rtree, rtree_ctx_t *rtree_ctx,
+    uintptr_t key, szind_t szind, bool slab) {
+	assert(!slab || szind < SC_NBINS);
+
+	rtree_leaf_elm_t *elm = rtree_read(tsdn, rtree, rtree_ctx, key, true);
+	rtree_leaf_elm_szind_slab_update(tsdn, rtree, elm, szind, slab);
+}
+
+static inline void
+rtree_clear(tsdn_t *tsdn, rtree_t *rtree, rtree_ctx_t *rtree_ctx,
+    uintptr_t key) {
+	rtree_leaf_elm_t *elm = rtree_read(tsdn, rtree, rtree_ctx, key, true);
+	assert(rtree_leaf_elm_extent_read(tsdn, rtree, elm, false) !=
+	    NULL);
+	rtree_leaf_elm_write(tsdn, rtree, elm, NULL, SC_NSIZES, false);
+}
+
+#endif /* JEMALLOC_INTERNAL_RTREE_H */

deps/jemalloc/include/jemalloc/internal/rtree_tsd.h

@@ -0,0 +1,50 @@
+#ifndef JEMALLOC_INTERNAL_RTREE_CTX_H
+#define JEMALLOC_INTERNAL_RTREE_CTX_H
+
+/*
+ * Number of leafkey/leaf pairs to cache in L1 and L2 level respectively.  Each
+ * entry supports an entire leaf, so the cache hit rate is typically high even
+ * with a small number of entries.  In rare cases extent activity will straddle
+ * the boundary between two leaf nodes.  Furthermore, an arena may use a
+ * combination of dss and mmap.  Note that as memory usage grows past the amount
+ * that this cache can directly cover, the cache will become less effective if
+ * locality of reference is low, but the consequence is merely cache misses
+ * while traversing the tree nodes.
+ *
+ * The L1 direct mapped cache offers consistent and low cost on cache hit.
+ * However collision could affect hit rate negatively.  This is resolved by
+ * combining with a L2 LRU cache, which requires linear search and re-ordering
+ * on access but suffers no collision.  Note that, the cache will itself suffer
+ * cache misses if made overly large, plus the cost of linear search in the LRU
+ * cache.
+ */
+#define RTREE_CTX_LG_NCACHE 4
+#define RTREE_CTX_NCACHE (1 << RTREE_CTX_LG_NCACHE)
+#define RTREE_CTX_NCACHE_L2 8
+
+/*
+ * Zero initializer required for tsd initialization only.  Proper initialization
+ * done via rtree_ctx_data_init().
+ */
+#define RTREE_CTX_ZERO_INITIALIZER {{{0, 0}}, {{0, 0}}}
+
+
+typedef struct rtree_leaf_elm_s rtree_leaf_elm_t;
+
+typedef struct rtree_ctx_cache_elm_s rtree_ctx_cache_elm_t;
+struct rtree_ctx_cache_elm_s {
+	uintptr_t		leafkey;
+	rtree_leaf_elm_t	*leaf;
+};
+
+typedef struct rtree_ctx_s rtree_ctx_t;
+struct rtree_ctx_s {
+	/* Direct mapped cache. */
+	rtree_ctx_cache_elm_t	cache[RTREE_CTX_NCACHE];
+	/* L2 LRU cache. */
+	rtree_ctx_cache_elm_t	l2_cache[RTREE_CTX_NCACHE_L2];
+};
+
+void rtree_ctx_data_init(rtree_ctx_t *ctx);
+
+#endif /* JEMALLOC_INTERNAL_RTREE_CTX_H */

deps/jemalloc/include/jemalloc/internal/safety_check.h

@@ -0,0 +1,26 @@
+#ifndef JEMALLOC_INTERNAL_SAFETY_CHECK_H
+#define JEMALLOC_INTERNAL_SAFETY_CHECK_H
+
+void safety_check_fail(const char *format, ...);
+/* Can set to NULL for a default. */
+void safety_check_set_abort(void (*abort_fn)());
+
+JEMALLOC_ALWAYS_INLINE void
+safety_check_set_redzone(void *ptr, size_t usize, size_t bumped_usize) {
+	assert(usize < bumped_usize);
+	for (size_t i = usize; i < bumped_usize && i < usize + 32; ++i) {
+		*((unsigned char *)ptr + i) = 0xBC;
+	}
+}
+
+JEMALLOC_ALWAYS_INLINE void
+safety_check_verify_redzone(const void *ptr, size_t usize, size_t bumped_usize)
+{
+	for (size_t i = usize; i < bumped_usize && i < usize + 32; ++i) {
+		if (unlikely(*((unsigned char *)ptr + i) != 0xBC)) {
+			safety_check_fail("Use after free error\n");
+		}
+	}
+}
+
+#endif /*JEMALLOC_INTERNAL_SAFETY_CHECK_H */

deps/jemalloc/include/jemalloc/internal/sc.h

@@ -0,0 +1,333 @@
+#ifndef JEMALLOC_INTERNAL_SC_H
+#define JEMALLOC_INTERNAL_SC_H
+
+#include "jemalloc/internal/jemalloc_internal_types.h"
+
+/*
+ * Size class computations:
+ *
+ * These are a little tricky; we'll first start by describing how things
+ * generally work, and then describe some of the details.
+ *
+ * Ignore the first few size classes for a moment. We can then split all the
+ * remaining size classes into groups. The size classes in a group are spaced
+ * such that they cover allocation request sizes in a power-of-2 range. The
+ * power of two is called the base of the group, and the size classes in it
+ * satisfy allocations in the half-open range (base, base * 2]. There are
+ * SC_NGROUP size classes in each group, equally spaced in the range, so that
+ * each one covers allocations for base / SC_NGROUP possible allocation sizes.
+ * We call that value (base / SC_NGROUP) the delta of the group. Each size class
+ * is delta larger than the one before it (including the initial size class in a
+ * group, which is delta larger than base, the largest size class in the
+ * previous group).
+ * To make the math all work out nicely, we require that SC_NGROUP is a power of
+ * two, and define it in terms of SC_LG_NGROUP. We'll often talk in terms of
+ * lg_base and lg_delta. For each of these groups then, we have that
+ * lg_delta == lg_base - SC_LG_NGROUP.
+ * The size classes in a group with a given lg_base and lg_delta (which, recall,
+ * can be computed from lg_base for these groups) are therefore:
+ *   base + 1 * delta
+ *     which covers allocations in (base, base + 1 * delta]
+ *   base + 2 * delta
+ *     which covers allocations in (base + 1 * delta, base + 2 * delta].
+ *   base + 3 * delta
+ *     which covers allocations in (base + 2 * delta, base + 3 * delta].
+ *   ...
+ *   base + SC_NGROUP * delta ( == 2 * base)
+ *     which covers allocations in (base + (SC_NGROUP - 1) * delta, 2 * base].
+ * (Note that currently SC_NGROUP is always 4, so the "..." is empty in
+ * practice.)
+ * Note that the last size class in the group is the next power of two (after
+ * base), so that we've set up the induction correctly for the next group's
+ * selection of delta.
+ *
+ * Now, let's start considering the first few size classes. Two extra constants
+ * come into play here: LG_QUANTUM and SC_LG_TINY_MIN. LG_QUANTUM ensures
+ * correct platform alignment; all objects of size (1 << LG_QUANTUM) or larger
+ * are at least (1 << LG_QUANTUM) aligned; this can be used to ensure that we
+ * never return improperly aligned memory, by making (1 << LG_QUANTUM) equal the
+ * highest required alignment of a platform. For allocation sizes smaller than
+ * (1 << LG_QUANTUM) though, we can be more relaxed (since we don't support
+ * platforms with types with alignment larger than their size). To allow such
+ * allocations (without wasting space unnecessarily), we introduce tiny size
+ * classes; one per power of two, up until we hit the quantum size. There are
+ * therefore LG_QUANTUM - SC_LG_TINY_MIN such size classes.
+ *
+ * Next, we have a size class of size (1 << LG_QUANTUM).  This can't be the
+ * start of a group in the sense we described above (covering a power of two
+ * range) since, if we divided into it to pick a value of delta, we'd get a
+ * delta smaller than (1 << LG_QUANTUM) for sizes >= (1 << LG_QUANTUM), which
+ * is against the rules.
+ *
+ * The first base we can divide by SC_NGROUP while still being at least
+ * (1 << LG_QUANTUM) is SC_NGROUP * (1 << LG_QUANTUM). We can get there by
+ * having SC_NGROUP size classes, spaced (1 << LG_QUANTUM) apart. These size
+ * classes are:
+ *   1 * (1 << LG_QUANTUM)
+ *   2 * (1 << LG_QUANTUM)
+ *   3 * (1 << LG_QUANTUM)
+ *   ... (although, as above, this "..." is empty in practice)
+ *   SC_NGROUP * (1 << LG_QUANTUM).
+ *
+ * There are SC_NGROUP of these size classes, so we can regard it as a sort of
+ * pseudo-group, even though it spans multiple powers of 2, is divided
+ * differently, and both starts and ends on a power of 2 (as opposed to just
+ * ending). SC_NGROUP is itself a power of two, so the first group after the
+ * pseudo-group has the power-of-two base SC_NGROUP * (1 << LG_QUANTUM), for a
+ * lg_base of LG_QUANTUM + SC_LG_NGROUP. We can divide this base into SC_NGROUP
+ * sizes without violating our LG_QUANTUM requirements, so we can safely set
+ * lg_delta = lg_base - SC_LG_GROUP (== LG_QUANTUM).
+ *
+ * So, in order, the size classes are:
+ *
+ * Tiny size classes:
+ * - Count: LG_QUANTUM - SC_LG_TINY_MIN.
+ * - Sizes:
+ *     1 << SC_LG_TINY_MIN
+ *     1 << (SC_LG_TINY_MIN + 1)
+ *     1 << (SC_LG_TINY_MIN + 2)
+ *     ...
+ *     1 << (LG_QUANTUM - 1)
+ *
+ * Initial pseudo-group:
+ * - Count: SC_NGROUP
+ * - Sizes:
+ *     1 * (1 << LG_QUANTUM)
+ *     2 * (1 << LG_QUANTUM)
+ *     3 * (1 << LG_QUANTUM)
+ *     ...
+ *     SC_NGROUP * (1 << LG_QUANTUM)
+ *
+ * Regular group 0:
+ * - Count: SC_NGROUP
+ * - Sizes:
+ *   (relative to lg_base of LG_QUANTUM + SC_LG_NGROUP and lg_delta of
+ *   lg_base - SC_LG_NGROUP)
+ *     (1 << lg_base) + 1 * (1 << lg_delta)
+ *     (1 << lg_base) + 2 * (1 << lg_delta)
+ *     (1 << lg_base) + 3 * (1 << lg_delta)
+ *     ...
+ *     (1 << lg_base) + SC_NGROUP * (1 << lg_delta) [ == (1 << (lg_base + 1)) ]
+ *
+ * Regular group 1:
+ * - Count: SC_NGROUP
+ * - Sizes:
+ *   (relative to lg_base of LG_QUANTUM + SC_LG_NGROUP + 1 and lg_delta of
+ *   lg_base - SC_LG_NGROUP)
+ *     (1 << lg_base) + 1 * (1 << lg_delta)
+ *     (1 << lg_base) + 2 * (1 << lg_delta)
+ *     (1 << lg_base) + 3 * (1 << lg_delta)
+ *     ...
+ *     (1 << lg_base) + SC_NGROUP * (1 << lg_delta) [ == (1 << (lg_base + 1)) ]
+ *
+ * ...
+ *
+ * Regular group N:
+ * - Count: SC_NGROUP
+ * - Sizes:
+ *   (relative to lg_base of LG_QUANTUM + SC_LG_NGROUP + N and lg_delta of
+ *   lg_base - SC_LG_NGROUP)
+ *     (1 << lg_base) + 1 * (1 << lg_delta)
+ *     (1 << lg_base) + 2 * (1 << lg_delta)
+ *     (1 << lg_base) + 3 * (1 << lg_delta)
+ *     ...
+ *     (1 << lg_base) + SC_NGROUP * (1 << lg_delta) [ == (1 << (lg_base + 1)) ]
+ *
+ *
+ * Representation of metadata:
+ * To make the math easy, we'll mostly work in lg quantities. We record lg_base,
+ * lg_delta, and ndelta (i.e. number of deltas above the base) on a
+ * per-size-class basis, and maintain the invariant that, across all size
+ * classes, size == (1 << lg_base) + ndelta * (1 << lg_delta).
+ *
+ * For regular groups (i.e. those with lg_base >= LG_QUANTUM + SC_LG_NGROUP),
+ * lg_delta is lg_base - SC_LG_NGROUP, and ndelta goes from 1 to SC_NGROUP.
+ *
+ * For the initial tiny size classes (if any), lg_base is lg(size class size).
+ * lg_delta is lg_base for the first size class, and lg_base - 1 for all
+ * subsequent ones. ndelta is always 0.
+ *
+ * For the pseudo-group, if there are no tiny size classes, then we set
+ * lg_base == LG_QUANTUM, lg_delta == LG_QUANTUM, and have ndelta range from 0
+ * to SC_NGROUP - 1. (Note that delta == base, so base + (SC_NGROUP - 1) * delta
+ * is just SC_NGROUP * base, or (1 << (SC_LG_NGROUP + LG_QUANTUM)), so we do
+ * indeed get a power of two that way). If there *are* tiny size classes, then
+ * the first size class needs to have lg_delta relative to the largest tiny size
+ * class. We therefore set lg_base == LG_QUANTUM - 1,
+ * lg_delta == LG_QUANTUM - 1, and ndelta == 1, keeping the rest of the
+ * pseudo-group the same.
+ *
+ *
+ * Other terminology:
+ * "Small" size classes mean those that are allocated out of bins, which is the
+ * same as those that are slab allocated.
+ * "Large" size classes are those that are not small. The cutoff for counting as
+ * large is page size * group size.
+ */
+
+/*
+ * Size class N + (1 << SC_LG_NGROUP) twice the size of size class N.
+ */
+#define SC_LG_NGROUP 2
+#define SC_LG_TINY_MIN 3
+
+#if SC_LG_TINY_MIN == 0
+/* The div module doesn't support division by 1, which this would require. */
+#error "Unsupported LG_TINY_MIN"
+#endif
+
+/*
+ * The definitions below are all determined by the above settings and system
+ * characteristics.
+ */
+#define SC_NGROUP (1ULL << SC_LG_NGROUP)
+#define SC_PTR_BITS ((1ULL << LG_SIZEOF_PTR) * 8)
+#define SC_NTINY (LG_QUANTUM - SC_LG_TINY_MIN)
+#define SC_LG_TINY_MAXCLASS (LG_QUANTUM > SC_LG_TINY_MIN ? LG_QUANTUM - 1 : -1)
+#define SC_NPSEUDO SC_NGROUP
+#define SC_LG_FIRST_REGULAR_BASE (LG_QUANTUM + SC_LG_NGROUP)
+/*
+ * We cap allocations to be less than 2 ** (ptr_bits - 1), so the highest base
+ * we need is 2 ** (ptr_bits - 2). (This also means that the last group is 1
+ * size class shorter than the others).
+ * We could probably save some space in arenas by capping this at LG_VADDR size.
+ */
+#define SC_LG_BASE_MAX (SC_PTR_BITS - 2)
+#define SC_NREGULAR (SC_NGROUP * 					\
+    (SC_LG_BASE_MAX - SC_LG_FIRST_REGULAR_BASE + 1) - 1)
+#define SC_NSIZES (SC_NTINY + SC_NPSEUDO + SC_NREGULAR)
+
+/* The number of size classes that are a multiple of the page size. */
+#define SC_NPSIZES (							\
+    /* Start with all the size classes. */				\
+    SC_NSIZES								\
+    /* Subtract out those groups with too small a base. */		\
+    - (LG_PAGE - 1 - SC_LG_FIRST_REGULAR_BASE) * SC_NGROUP		\
+    /* And the pseudo-group. */						\
+    - SC_NPSEUDO							\
+    /* And the tiny group. */						\
+    - SC_NTINY								\
+    /* Sizes where ndelta*delta is not a multiple of the page size. */	\
+    - (SC_LG_NGROUP * SC_NGROUP))
+/*
+ * Note that the last line is computed as the sum of the second column in the
+ * following table:
+ *                      lg(base) | count of sizes to exclude
+ * ------------------------------|-----------------------------
+ *                   LG_PAGE - 1 | SC_NGROUP - 1
+ *                       LG_PAGE | SC_NGROUP - 1
+ *                   LG_PAGE + 1 | SC_NGROUP - 2
+ *                   LG_PAGE + 2 | SC_NGROUP - 4
+ *                           ... | ...
+ *  LG_PAGE + (SC_LG_NGROUP - 1) | SC_NGROUP - (SC_NGROUP / 2)
+ */
+
+/*
+ * We declare a size class is binnable if size < page size * group. Or, in other
+ * words, lg(size) < lg(page size) + lg(group size).
+ */
+#define SC_NBINS (							\
+    /* Sub-regular size classes. */					\
+    SC_NTINY + SC_NPSEUDO						\
+    /* Groups with lg_regular_min_base <= lg_base <= lg_base_max */	\
+    + SC_NGROUP * (LG_PAGE + SC_LG_NGROUP - SC_LG_FIRST_REGULAR_BASE)	\
+    /* Last SC of the last group hits the bound exactly; exclude it. */	\
+    - 1)
+
+/*
+ * The size2index_tab lookup table uses uint8_t to encode each bin index, so we
+ * cannot support more than 256 small size classes.
+ */
+#if (SC_NBINS > 256)
+#  error "Too many small size classes"
+#endif
+
+/* The largest size class in the lookup table. */
+#define SC_LOOKUP_MAXCLASS ((size_t)1 << 12)
+
+/* Internal, only used for the definition of SC_SMALL_MAXCLASS. */
+#define SC_SMALL_MAX_BASE ((size_t)1 << (LG_PAGE + SC_LG_NGROUP - 1))
+#define SC_SMALL_MAX_DELTA ((size_t)1 << (LG_PAGE - 1))
+
+/* The largest size class allocated out of a slab. */
+#define SC_SMALL_MAXCLASS (SC_SMALL_MAX_BASE				\
+    + (SC_NGROUP - 1) * SC_SMALL_MAX_DELTA)
+
+/* The smallest size class not allocated out of a slab. */
+#define SC_LARGE_MINCLASS ((size_t)1ULL << (LG_PAGE + SC_LG_NGROUP))
+#define SC_LG_LARGE_MINCLASS (LG_PAGE + SC_LG_NGROUP)
+
+/* Internal; only used for the definition of SC_LARGE_MAXCLASS. */
+#define SC_MAX_BASE ((size_t)1 << (SC_PTR_BITS - 2))
+#define SC_MAX_DELTA ((size_t)1 << (SC_PTR_BITS - 2 - SC_LG_NGROUP))
+
+/* The largest size class supported. */
+#define SC_LARGE_MAXCLASS (SC_MAX_BASE + (SC_NGROUP - 1) * SC_MAX_DELTA)
+
+typedef struct sc_s sc_t;
+struct sc_s {
+	/* Size class index, or -1 if not a valid size class. */
+	int index;
+	/* Lg group base size (no deltas added). */
+	int lg_base;
+	/* Lg delta to previous size class. */
+	int lg_delta;
+	/* Delta multiplier.  size == 1<<lg_base + ndelta<<lg_delta */
+	int ndelta;
+	/*
+	 * True if the size class is a multiple of the page size, false
+	 * otherwise.
+	 */
+	bool psz;
+	/*
+	 * True if the size class is a small, bin, size class. False otherwise.
+	 */
+	bool bin;
+	/* The slab page count if a small bin size class, 0 otherwise. */
+	int pgs;
+	/* Same as lg_delta if a lookup table size class, 0 otherwise. */
+	int lg_delta_lookup;
+};
+
+typedef struct sc_data_s sc_data_t;
+struct sc_data_s {
+	/* Number of tiny size classes. */
+	unsigned ntiny;
+	/* Number of bins supported by the lookup table. */
+	int nlbins;
+	/* Number of small size class bins. */
+	int nbins;
+	/* Number of size classes. */
+	int nsizes;
+	/* Number of bits required to store NSIZES. */
+	int lg_ceil_nsizes;
+	/* Number of size classes that are a multiple of (1U << LG_PAGE). */
+	unsigned npsizes;
+	/* Lg of maximum tiny size class (or -1, if none). */
+	int lg_tiny_maxclass;
+	/* Maximum size class included in lookup table. */
+	size_t lookup_maxclass;
+	/* Maximum small size class. */
+	size_t small_maxclass;
+	/* Lg of minimum large size class. */
+	int lg_large_minclass;
+	/* The minimum large size class. */
+	size_t large_minclass;
+	/* Maximum (large) size class. */
+	size_t large_maxclass;
+	/* True if the sc_data_t has been initialized (for debugging only). */
+	bool initialized;
+
+	sc_t sc[SC_NSIZES];
+};
+
+void sc_data_init(sc_data_t *data);
+/*
+ * Updates slab sizes in [begin, end] to be pgs pages in length, if possible.
+ * Otherwise, does its best to accomodate the request.
+ */
+void sc_data_update_slab_size(sc_data_t *data, size_t begin, size_t end,
+    int pgs);
+void sc_boot(sc_data_t *data);
+
+#endif /* JEMALLOC_INTERNAL_SC_H */

deps/jemalloc/include/jemalloc/internal/seq.h

@@ -0,0 +1,55 @@
+#ifndef JEMALLOC_INTERNAL_SEQ_H
+#define JEMALLOC_INTERNAL_SEQ_H
+
+#include "jemalloc/internal/atomic.h"
+
+/*
+ * A simple seqlock implementation.
+ */
+
+#define seq_define(type, short_type)					\
+typedef struct {							\
+	atomic_zu_t seq;						\
+	atomic_zu_t data[						\
+	    (sizeof(type) + sizeof(size_t) - 1) / sizeof(size_t)];	\
+} seq_##short_type##_t;							\
+									\
+/*									\
+ * No internal synchronization -- the caller must ensure that there's	\
+ * only a single writer at a time.					\
+ */									\
+static inline void							\
+seq_store_##short_type(seq_##short_type##_t *dst, type *src) {		\
+	size_t buf[sizeof(dst->data) / sizeof(size_t)];			\
+	buf[sizeof(buf) / sizeof(size_t) - 1] = 0;			\
+	memcpy(buf, src, sizeof(type));					\
+	size_t old_seq = atomic_load_zu(&dst->seq, ATOMIC_RELAXED);	\
+	atomic_store_zu(&dst->seq, old_seq + 1, ATOMIC_RELAXED);	\
+	atomic_fence(ATOMIC_RELEASE);					\
+	for (size_t i = 0; i < sizeof(buf) / sizeof(size_t); i++) {	\
+		atomic_store_zu(&dst->data[i], buf[i], ATOMIC_RELAXED);	\
+	}								\
+	atomic_store_zu(&dst->seq, old_seq + 2, ATOMIC_RELEASE);	\
+}									\
+									\
+/* Returns whether or not the read was consistent. */			\
+static inline bool							\
+seq_try_load_##short_type(type *dst, seq_##short_type##_t *src) {	\
+	size_t buf[sizeof(src->data) / sizeof(size_t)];			\
+	size_t seq1 = atomic_load_zu(&src->seq, ATOMIC_ACQUIRE);	\
+	if (seq1 % 2 != 0) {						\
+		return false;						\
+	}								\
+	for (size_t i = 0; i < sizeof(buf) / sizeof(size_t); i++) {	\
+		buf[i] = atomic_load_zu(&src->data[i], ATOMIC_RELAXED);	\
+	}								\
+	atomic_fence(ATOMIC_ACQUIRE);					\
+	size_t seq2 = atomic_load_zu(&src->seq, ATOMIC_RELAXED);	\
+	if (seq1 != seq2) {						\
+		return false;						\
+	}								\
+	memcpy(dst, buf, sizeof(type));					\
+	return true;							\
+}
+
+#endif /* JEMALLOC_INTERNAL_SEQ_H */

deps/jemalloc/include/jemalloc/internal/smoothstep.h

@@ -0,0 +1,232 @@
+#ifndef JEMALLOC_INTERNAL_SMOOTHSTEP_H
+#define JEMALLOC_INTERNAL_SMOOTHSTEP_H
+
+/*
+ * This file was generated by the following command:
+ *   sh smoothstep.sh smoother 200 24 3 15
+ */
+/******************************************************************************/
+
+/*
+ * This header defines a precomputed table based on the smoothstep family of
+ * sigmoidal curves (https://en.wikipedia.org/wiki/Smoothstep) that grow from 0
+ * to 1 in 0 <= x <= 1.  The table is stored as integer fixed point values so
+ * that floating point math can be avoided.
+ *
+ *                      3     2
+ *   smoothstep(x) = -2x  + 3x
+ *
+ *                       5      4      3
+ *   smootherstep(x) = 6x  - 15x  + 10x
+ *
+ *                          7      6      5      4
+ *   smootheststep(x) = -20x  + 70x  - 84x  + 35x
+ */
+
+#define SMOOTHSTEP_VARIANT	"smoother"
+#define SMOOTHSTEP_NSTEPS	200
+#define SMOOTHSTEP_BFP		24
+#define SMOOTHSTEP \
+ /* STEP(step, h,                            x,     y) */ \
+    STEP(   1, UINT64_C(0x0000000000000014), 0.005, 0.000001240643750) \
+    STEP(   2, UINT64_C(0x00000000000000a5), 0.010, 0.000009850600000) \
+    STEP(   3, UINT64_C(0x0000000000000229), 0.015, 0.000032995181250) \
+    STEP(   4, UINT64_C(0x0000000000000516), 0.020, 0.000077619200000) \
+    STEP(   5, UINT64_C(0x00000000000009dc), 0.025, 0.000150449218750) \
+    STEP(   6, UINT64_C(0x00000000000010e8), 0.030, 0.000257995800000) \
+    STEP(   7, UINT64_C(0x0000000000001aa4), 0.035, 0.000406555756250) \
+    STEP(   8, UINT64_C(0x0000000000002777), 0.040, 0.000602214400000) \
+    STEP(   9, UINT64_C(0x00000000000037c2), 0.045, 0.000850847793750) \
+    STEP(  10, UINT64_C(0x0000000000004be6), 0.050, 0.001158125000000) \
+    STEP(  11, UINT64_C(0x000000000000643c), 0.055, 0.001529510331250) \
+    STEP(  12, UINT64_C(0x000000000000811f), 0.060, 0.001970265600000) \
+    STEP(  13, UINT64_C(0x000000000000a2e2), 0.065, 0.002485452368750) \
+    STEP(  14, UINT64_C(0x000000000000c9d8), 0.070, 0.003079934200000) \
+    STEP(  15, UINT64_C(0x000000000000f64f), 0.075, 0.003758378906250) \
+    STEP(  16, UINT64_C(0x0000000000012891), 0.080, 0.004525260800000) \
+    STEP(  17, UINT64_C(0x00000000000160e7), 0.085, 0.005384862943750) \
+    STEP(  18, UINT64_C(0x0000000000019f95), 0.090, 0.006341279400000) \
+    STEP(  19, UINT64_C(0x000000000001e4dc), 0.095, 0.007398417481250) \
+    STEP(  20, UINT64_C(0x00000000000230fc), 0.100, 0.008560000000000) \
+    STEP(  21, UINT64_C(0x0000000000028430), 0.105, 0.009829567518750) \
+    STEP(  22, UINT64_C(0x000000000002deb0), 0.110, 0.011210480600000) \
+    STEP(  23, UINT64_C(0x00000000000340b1), 0.115, 0.012705922056250) \
+    STEP(  24, UINT64_C(0x000000000003aa67), 0.120, 0.014318899200000) \
+    STEP(  25, UINT64_C(0x0000000000041c00), 0.125, 0.016052246093750) \
+    STEP(  26, UINT64_C(0x00000000000495a8), 0.130, 0.017908625800000) \
+    STEP(  27, UINT64_C(0x000000000005178b), 0.135, 0.019890532631250) \
+    STEP(  28, UINT64_C(0x000000000005a1cf), 0.140, 0.022000294400000) \
+    STEP(  29, UINT64_C(0x0000000000063498), 0.145, 0.024240074668750) \
+    STEP(  30, UINT64_C(0x000000000006d009), 0.150, 0.026611875000000) \
+    STEP(  31, UINT64_C(0x000000000007743f), 0.155, 0.029117537206250) \
+    STEP(  32, UINT64_C(0x0000000000082157), 0.160, 0.031758745600000) \
+    STEP(  33, UINT64_C(0x000000000008d76b), 0.165, 0.034537029243750) \
+    STEP(  34, UINT64_C(0x0000000000099691), 0.170, 0.037453764200000) \
+    STEP(  35, UINT64_C(0x00000000000a5edf), 0.175, 0.040510175781250) \
+    STEP(  36, UINT64_C(0x00000000000b3067), 0.180, 0.043707340800000) \
+    STEP(  37, UINT64_C(0x00000000000c0b38), 0.185, 0.047046189818750) \
+    STEP(  38, UINT64_C(0x00000000000cef5e), 0.190, 0.050527509400000) \
+    STEP(  39, UINT64_C(0x00000000000ddce6), 0.195, 0.054151944356250) \
+    STEP(  40, UINT64_C(0x00000000000ed3d8), 0.200, 0.057920000000000) \
+    STEP(  41, UINT64_C(0x00000000000fd439), 0.205, 0.061832044393750) \
+    STEP(  42, UINT64_C(0x000000000010de0e), 0.210, 0.065888310600000) \
+    STEP(  43, UINT64_C(0x000000000011f158), 0.215, 0.070088898931250) \
+    STEP(  44, UINT64_C(0x0000000000130e17), 0.220, 0.074433779200000) \
+    STEP(  45, UINT64_C(0x0000000000143448), 0.225, 0.078922792968750) \
+    STEP(  46, UINT64_C(0x00000000001563e7), 0.230, 0.083555655800000) \
+    STEP(  47, UINT64_C(0x0000000000169cec), 0.235, 0.088331959506250) \
+    STEP(  48, UINT64_C(0x000000000017df4f), 0.240, 0.093251174400000) \
+    STEP(  49, UINT64_C(0x0000000000192b04), 0.245, 0.098312651543750) \
+    STEP(  50, UINT64_C(0x00000000001a8000), 0.250, 0.103515625000000) \
+    STEP(  51, UINT64_C(0x00000000001bde32), 0.255, 0.108859214081250) \
+    STEP(  52, UINT64_C(0x00000000001d458b), 0.260, 0.114342425600000) \
+    STEP(  53, UINT64_C(0x00000000001eb5f8), 0.265, 0.119964156118750) \
+    STEP(  54, UINT64_C(0x0000000000202f65), 0.270, 0.125723194200000) \
+    STEP(  55, UINT64_C(0x000000000021b1bb), 0.275, 0.131618222656250) \
+    STEP(  56, UINT64_C(0x0000000000233ce3), 0.280, 0.137647820800000) \
+    STEP(  57, UINT64_C(0x000000000024d0c3), 0.285, 0.143810466693750) \
+    STEP(  58, UINT64_C(0x0000000000266d40), 0.290, 0.150104539400000) \
+    STEP(  59, UINT64_C(0x000000000028123d), 0.295, 0.156528321231250) \
+    STEP(  60, UINT64_C(0x000000000029bf9c), 0.300, 0.163080000000000) \
+    STEP(  61, UINT64_C(0x00000000002b753d), 0.305, 0.169757671268750) \
+    STEP(  62, UINT64_C(0x00000000002d32fe), 0.310, 0.176559340600000) \
+    STEP(  63, UINT64_C(0x00000000002ef8bc), 0.315, 0.183482925806250) \
+    STEP(  64, UINT64_C(0x000000000030c654), 0.320, 0.190526259200000) \
+    STEP(  65, UINT64_C(0x0000000000329b9f), 0.325, 0.197687089843750) \
+    STEP(  66, UINT64_C(0x0000000000347875), 0.330, 0.204963085800000) \
+    STEP(  67, UINT64_C(0x0000000000365cb0), 0.335, 0.212351836381250) \
+    STEP(  68, UINT64_C(0x0000000000384825), 0.340, 0.219850854400000) \
+    STEP(  69, UINT64_C(0x00000000003a3aa8), 0.345, 0.227457578418750) \
+    STEP(  70, UINT64_C(0x00000000003c340f), 0.350, 0.235169375000000) \
+    STEP(  71, UINT64_C(0x00000000003e342b), 0.355, 0.242983540956250) \
+    STEP(  72, UINT64_C(0x0000000000403ace), 0.360, 0.250897305600000) \
+    STEP(  73, UINT64_C(0x00000000004247c8), 0.365, 0.258907832993750) \
+    STEP(  74, UINT64_C(0x0000000000445ae9), 0.370, 0.267012224200000) \
+    STEP(  75, UINT64_C(0x0000000000467400), 0.375, 0.275207519531250) \
+    STEP(  76, UINT64_C(0x00000000004892d8), 0.380, 0.283490700800000) \
+    STEP(  77, UINT64_C(0x00000000004ab740), 0.385, 0.291858693568750) \
+    STEP(  78, UINT64_C(0x00000000004ce102), 0.390, 0.300308369400000) \
+    STEP(  79, UINT64_C(0x00000000004f0fe9), 0.395, 0.308836548106250) \
+    STEP(  80, UINT64_C(0x00000000005143bf), 0.400, 0.317440000000000) \
+    STEP(  81, UINT64_C(0x0000000000537c4d), 0.405, 0.326115448143750) \
+    STEP(  82, UINT64_C(0x000000000055b95b), 0.410, 0.334859570600000) \
+    STEP(  83, UINT64_C(0x000000000057fab1), 0.415, 0.343669002681250) \
+    STEP(  84, UINT64_C(0x00000000005a4015), 0.420, 0.352540339200000) \
+    STEP(  85, UINT64_C(0x00000000005c894e), 0.425, 0.361470136718750) \
+    STEP(  86, UINT64_C(0x00000000005ed622), 0.430, 0.370454915800000) \
+    STEP(  87, UINT64_C(0x0000000000612655), 0.435, 0.379491163256250) \
+    STEP(  88, UINT64_C(0x00000000006379ac), 0.440, 0.388575334400000) \
+    STEP(  89, UINT64_C(0x000000000065cfeb), 0.445, 0.397703855293750) \
+    STEP(  90, UINT64_C(0x00000000006828d6), 0.450, 0.406873125000000) \
+    STEP(  91, UINT64_C(0x00000000006a842f), 0.455, 0.416079517831250) \
+    STEP(  92, UINT64_C(0x00000000006ce1bb), 0.460, 0.425319385600000) \
+    STEP(  93, UINT64_C(0x00000000006f413a), 0.465, 0.434589059868750) \
+    STEP(  94, UINT64_C(0x000000000071a270), 0.470, 0.443884854200000) \
+    STEP(  95, UINT64_C(0x000000000074051d), 0.475, 0.453203066406250) \
+    STEP(  96, UINT64_C(0x0000000000766905), 0.480, 0.462539980800000) \
+    STEP(  97, UINT64_C(0x000000000078cde7), 0.485, 0.471891870443750) \
+    STEP(  98, UINT64_C(0x00000000007b3387), 0.490, 0.481254999400000) \
+    STEP(  99, UINT64_C(0x00000000007d99a4), 0.495, 0.490625624981250) \
+    STEP( 100, UINT64_C(0x0000000000800000), 0.500, 0.500000000000000) \
+    STEP( 101, UINT64_C(0x000000000082665b), 0.505, 0.509374375018750) \
+    STEP( 102, UINT64_C(0x000000000084cc78), 0.510, 0.518745000600000) \
+    STEP( 103, UINT64_C(0x0000000000873218), 0.515, 0.528108129556250) \
+    STEP( 104, UINT64_C(0x00000000008996fa), 0.520, 0.537460019200000) \
+    STEP( 105, UINT64_C(0x00000000008bfae2), 0.525, 0.546796933593750) \
+    STEP( 106, UINT64_C(0x00000000008e5d8f), 0.530, 0.556115145800000) \
+    STEP( 107, UINT64_C(0x000000000090bec5), 0.535, 0.565410940131250) \
+    STEP( 108, UINT64_C(0x0000000000931e44), 0.540, 0.574680614400000) \
+    STEP( 109, UINT64_C(0x0000000000957bd0), 0.545, 0.583920482168750) \
+    STEP( 110, UINT64_C(0x000000000097d729), 0.550, 0.593126875000000) \
+    STEP( 111, UINT64_C(0x00000000009a3014), 0.555, 0.602296144706250) \
+    STEP( 112, UINT64_C(0x00000000009c8653), 0.560, 0.611424665600000) \
+    STEP( 113, UINT64_C(0x00000000009ed9aa), 0.565, 0.620508836743750) \
+    STEP( 114, UINT64_C(0x0000000000a129dd), 0.570, 0.629545084200000) \
+    STEP( 115, UINT64_C(0x0000000000a376b1), 0.575, 0.638529863281250) \
+    STEP( 116, UINT64_C(0x0000000000a5bfea), 0.580, 0.647459660800000) \
+    STEP( 117, UINT64_C(0x0000000000a8054e), 0.585, 0.656330997318750) \
+    STEP( 118, UINT64_C(0x0000000000aa46a4), 0.590, 0.665140429400000) \
+    STEP( 119, UINT64_C(0x0000000000ac83b2), 0.595, 0.673884551856250) \
+    STEP( 120, UINT64_C(0x0000000000aebc40), 0.600, 0.682560000000000) \
+    STEP( 121, UINT64_C(0x0000000000b0f016), 0.605, 0.691163451893750) \
+    STEP( 122, UINT64_C(0x0000000000b31efd), 0.610, 0.699691630600000) \
+    STEP( 123, UINT64_C(0x0000000000b548bf), 0.615, 0.708141306431250) \
+    STEP( 124, UINT64_C(0x0000000000b76d27), 0.620, 0.716509299200000) \
+    STEP( 125, UINT64_C(0x0000000000b98c00), 0.625, 0.724792480468750) \
+    STEP( 126, UINT64_C(0x0000000000bba516), 0.630, 0.732987775800000) \
+    STEP( 127, UINT64_C(0x0000000000bdb837), 0.635, 0.741092167006250) \
+    STEP( 128, UINT64_C(0x0000000000bfc531), 0.640, 0.749102694400000) \
+    STEP( 129, UINT64_C(0x0000000000c1cbd4), 0.645, 0.757016459043750) \
+    STEP( 130, UINT64_C(0x0000000000c3cbf0), 0.650, 0.764830625000000) \
+    STEP( 131, UINT64_C(0x0000000000c5c557), 0.655, 0.772542421581250) \
+    STEP( 132, UINT64_C(0x0000000000c7b7da), 0.660, 0.780149145600000) \
+    STEP( 133, UINT64_C(0x0000000000c9a34f), 0.665, 0.787648163618750) \
+    STEP( 134, UINT64_C(0x0000000000cb878a), 0.670, 0.795036914200000) \
+    STEP( 135, UINT64_C(0x0000000000cd6460), 0.675, 0.802312910156250) \
+    STEP( 136, UINT64_C(0x0000000000cf39ab), 0.680, 0.809473740800000) \
+    STEP( 137, UINT64_C(0x0000000000d10743), 0.685, 0.816517074193750) \
+    STEP( 138, UINT64_C(0x0000000000d2cd01), 0.690, 0.823440659400000) \
+    STEP( 139, UINT64_C(0x0000000000d48ac2), 0.695, 0.830242328731250) \
+    STEP( 140, UINT64_C(0x0000000000d64063), 0.700, 0.836920000000000) \
+    STEP( 141, UINT64_C(0x0000000000d7edc2), 0.705, 0.843471678768750) \
+    STEP( 142, UINT64_C(0x0000000000d992bf), 0.710, 0.849895460600000) \
+    STEP( 143, UINT64_C(0x0000000000db2f3c), 0.715, 0.856189533306250) \
+    STEP( 144, UINT64_C(0x0000000000dcc31c), 0.720, 0.862352179200000) \
+    STEP( 145, UINT64_C(0x0000000000de4e44), 0.725, 0.868381777343750) \
+    STEP( 146, UINT64_C(0x0000000000dfd09a), 0.730, 0.874276805800000) \
+    STEP( 147, UINT64_C(0x0000000000e14a07), 0.735, 0.880035843881250) \
+    STEP( 148, UINT64_C(0x0000000000e2ba74), 0.740, 0.885657574400000) \
+    STEP( 149, UINT64_C(0x0000000000e421cd), 0.745, 0.891140785918750) \
+    STEP( 150, UINT64_C(0x0000000000e58000), 0.750, 0.896484375000000) \
+    STEP( 151, UINT64_C(0x0000000000e6d4fb), 0.755, 0.901687348456250) \
+    STEP( 152, UINT64_C(0x0000000000e820b0), 0.760, 0.906748825600000) \
+    STEP( 153, UINT64_C(0x0000000000e96313), 0.765, 0.911668040493750) \
+    STEP( 154, UINT64_C(0x0000000000ea9c18), 0.770, 0.916444344200000) \
+    STEP( 155, UINT64_C(0x0000000000ebcbb7), 0.775, 0.921077207031250) \
+    STEP( 156, UINT64_C(0x0000000000ecf1e8), 0.780, 0.925566220800000) \
+    STEP( 157, UINT64_C(0x0000000000ee0ea7), 0.785, 0.929911101068750) \
+    STEP( 158, UINT64_C(0x0000000000ef21f1), 0.790, 0.934111689400000) \
+    STEP( 159, UINT64_C(0x0000000000f02bc6), 0.795, 0.938167955606250) \
+    STEP( 160, UINT64_C(0x0000000000f12c27), 0.800, 0.942080000000000) \
+    STEP( 161, UINT64_C(0x0000000000f22319), 0.805, 0.945848055643750) \
+    STEP( 162, UINT64_C(0x0000000000f310a1), 0.810, 0.949472490600000) \
+    STEP( 163, UINT64_C(0x0000000000f3f4c7), 0.815, 0.952953810181250) \
+    STEP( 164, UINT64_C(0x0000000000f4cf98), 0.820, 0.956292659200000) \
+    STEP( 165, UINT64_C(0x0000000000f5a120), 0.825, 0.959489824218750) \
+    STEP( 166, UINT64_C(0x0000000000f6696e), 0.830, 0.962546235800000) \
+    STEP( 167, UINT64_C(0x0000000000f72894), 0.835, 0.965462970756250) \
+    STEP( 168, UINT64_C(0x0000000000f7dea8), 0.840, 0.968241254400000) \
+    STEP( 169, UINT64_C(0x0000000000f88bc0), 0.845, 0.970882462793750) \
+    STEP( 170, UINT64_C(0x0000000000f92ff6), 0.850, 0.973388125000000) \
+    STEP( 171, UINT64_C(0x0000000000f9cb67), 0.855, 0.975759925331250) \
+    STEP( 172, UINT64_C(0x0000000000fa5e30), 0.860, 0.977999705600000) \
+    STEP( 173, UINT64_C(0x0000000000fae874), 0.865, 0.980109467368750) \
+    STEP( 174, UINT64_C(0x0000000000fb6a57), 0.870, 0.982091374200000) \
+    STEP( 175, UINT64_C(0x0000000000fbe400), 0.875, 0.983947753906250) \
+    STEP( 176, UINT64_C(0x0000000000fc5598), 0.880, 0.985681100800000) \
+    STEP( 177, UINT64_C(0x0000000000fcbf4e), 0.885, 0.987294077943750) \
+    STEP( 178, UINT64_C(0x0000000000fd214f), 0.890, 0.988789519400000) \
+    STEP( 179, UINT64_C(0x0000000000fd7bcf), 0.895, 0.990170432481250) \
+    STEP( 180, UINT64_C(0x0000000000fdcf03), 0.900, 0.991440000000000) \
+    STEP( 181, UINT64_C(0x0000000000fe1b23), 0.905, 0.992601582518750) \
+    STEP( 182, UINT64_C(0x0000000000fe606a), 0.910, 0.993658720600000) \
+    STEP( 183, UINT64_C(0x0000000000fe9f18), 0.915, 0.994615137056250) \
+    STEP( 184, UINT64_C(0x0000000000fed76e), 0.920, 0.995474739200000) \
+    STEP( 185, UINT64_C(0x0000000000ff09b0), 0.925, 0.996241621093750) \
+    STEP( 186, UINT64_C(0x0000000000ff3627), 0.930, 0.996920065800000) \
+    STEP( 187, UINT64_C(0x0000000000ff5d1d), 0.935, 0.997514547631250) \
+    STEP( 188, UINT64_C(0x0000000000ff7ee0), 0.940, 0.998029734400000) \
+    STEP( 189, UINT64_C(0x0000000000ff9bc3), 0.945, 0.998470489668750) \
+    STEP( 190, UINT64_C(0x0000000000ffb419), 0.950, 0.998841875000000) \
+    STEP( 191, UINT64_C(0x0000000000ffc83d), 0.955, 0.999149152206250) \
+    STEP( 192, UINT64_C(0x0000000000ffd888), 0.960, 0.999397785600000) \
+    STEP( 193, UINT64_C(0x0000000000ffe55b), 0.965, 0.999593444243750) \
+    STEP( 194, UINT64_C(0x0000000000ffef17), 0.970, 0.999742004200000) \
+    STEP( 195, UINT64_C(0x0000000000fff623), 0.975, 0.999849550781250) \
+    STEP( 196, UINT64_C(0x0000000000fffae9), 0.980, 0.999922380800000) \
+    STEP( 197, UINT64_C(0x0000000000fffdd6), 0.985, 0.999967004818750) \
+    STEP( 198, UINT64_C(0x0000000000ffff5a), 0.990, 0.999990149400000) \
+    STEP( 199, UINT64_C(0x0000000000ffffeb), 0.995, 0.999998759356250) \
+    STEP( 200, UINT64_C(0x0000000001000000), 1.000, 1.000000000000000) \
+
+#endif /* JEMALLOC_INTERNAL_SMOOTHSTEP_H */

deps/jemalloc/include/jemalloc/internal/spin.h

@@ -0,0 +1,40 @@
+#ifndef JEMALLOC_INTERNAL_SPIN_H
+#define JEMALLOC_INTERNAL_SPIN_H
+
+#define SPIN_INITIALIZER {0U}
+
+typedef struct {
+	unsigned iteration;
+} spin_t;
+
+static inline void
+spin_cpu_spinwait() {
+#  if HAVE_CPU_SPINWAIT
+	CPU_SPINWAIT;
+#  else
+	volatile int x = 0;
+	x = x;
+#  endif
+}
+
+static inline void
+spin_adaptive(spin_t *spin) {
+	volatile uint32_t i;
+
+	if (spin->iteration < 5) {
+		for (i = 0; i < (1U << spin->iteration); i++) {
+			spin_cpu_spinwait();
+		}
+		spin->iteration++;
+	} else {
+#ifdef _WIN32
+		SwitchToThread();
+#else
+		sched_yield();
+#endif
+	}
+}
+
+#undef SPIN_INLINE
+
+#endif /* JEMALLOC_INTERNAL_SPIN_H */

deps/jemalloc/include/jemalloc/internal/stats.h

@@ -0,0 +1,31 @@
+#ifndef JEMALLOC_INTERNAL_STATS_H
+#define JEMALLOC_INTERNAL_STATS_H
+
+/*  OPTION(opt,		var_name,	default,	set_value_to) */
+#define STATS_PRINT_OPTIONS						\
+    OPTION('J',		json,		false,		true)		\
+    OPTION('g',		general,	true,		false)		\
+    OPTION('m',		merged,		config_stats,	false)		\
+    OPTION('d',		destroyed,	config_stats,	false)		\
+    OPTION('a',		unmerged,	config_stats,	false)		\
+    OPTION('b',		bins,		true,		false)		\
+    OPTION('l',		large,		true,		false)		\
+    OPTION('x',		mutex,		true,		false)		\
+    OPTION('e',		extents,	true,		false)
+
+enum {
+#define OPTION(o, v, d, s) stats_print_option_num_##v,
+    STATS_PRINT_OPTIONS
+#undef OPTION
+    stats_print_tot_num_options
+};
+
+/* Options for stats_print. */
+extern bool opt_stats_print;
+extern char opt_stats_print_opts[stats_print_tot_num_options+1];
+
+/* Implements je_malloc_stats_print. */
+void stats_print(void (*write_cb)(void *, const char *), void *cbopaque,
+    const char *opts);
+
+#endif /* JEMALLOC_INTERNAL_STATS_H */

deps/jemalloc/include/jemalloc/internal/stats_tsd.h

@@ -0,0 +1,12 @@
+#ifndef JEMALLOC_INTERNAL_STATS_TSD_H
+#define JEMALLOC_INTERNAL_STATS_TSD_H
+
+typedef struct tcache_bin_stats_s {
+	/*
+	 * Number of allocation requests that corresponded to the size of this
+	 * bin.
+	 */
+	uint64_t	nrequests;
+} tcache_bin_stats_t;
+
+#endif /* JEMALLOC_INTERNAL_STATS_TSD_H */

deps/jemalloc/include/jemalloc/internal/sz.h

@@ -0,0 +1,318 @@
+#ifndef JEMALLOC_INTERNAL_SIZE_H
+#define JEMALLOC_INTERNAL_SIZE_H
+
+#include "jemalloc/internal/bit_util.h"
+#include "jemalloc/internal/pages.h"
+#include "jemalloc/internal/sc.h"
+#include "jemalloc/internal/util.h"
+
+/*
+ * sz module: Size computations.
+ *
+ * Some abbreviations used here:
+ *   p: Page
+ *   ind: Index
+ *   s, sz: Size
+ *   u: Usable size
+ *   a: Aligned
+ *
+ * These are not always used completely consistently, but should be enough to
+ * interpret function names.  E.g. sz_psz2ind converts page size to page size
+ * index; sz_sa2u converts a (size, alignment) allocation request to the usable
+ * size that would result from such an allocation.
+ */
+
+/*
+ * sz_pind2sz_tab encodes the same information as could be computed by
+ * sz_pind2sz_compute().
+ */
+extern size_t sz_pind2sz_tab[SC_NPSIZES + 1];
+/*
+ * sz_index2size_tab encodes the same information as could be computed (at
+ * unacceptable cost in some code paths) by sz_index2size_compute().
+ */
+extern size_t sz_index2size_tab[SC_NSIZES];
+/*
+ * sz_size2index_tab is a compact lookup table that rounds request sizes up to
+ * size classes.  In order to reduce cache footprint, the table is compressed,
+ * and all accesses are via sz_size2index().
+ */
+extern uint8_t sz_size2index_tab[];
+
+static const size_t sz_large_pad =
+#ifdef JEMALLOC_CACHE_OBLIVIOUS
+    PAGE
+#else
+    0
+#endif
+    ;
+
+extern void sz_boot(const sc_data_t *sc_data);
+
+JEMALLOC_ALWAYS_INLINE pszind_t
+sz_psz2ind(size_t psz) {
+	if (unlikely(psz > SC_LARGE_MAXCLASS)) {
+		return SC_NPSIZES;
+	}
+	pszind_t x = lg_floor((psz<<1)-1);
+	pszind_t shift = (x < SC_LG_NGROUP + LG_PAGE) ?
+	    0 : x - (SC_LG_NGROUP + LG_PAGE);
+	pszind_t grp = shift << SC_LG_NGROUP;
+
+	pszind_t lg_delta = (x < SC_LG_NGROUP + LG_PAGE + 1) ?
+	    LG_PAGE : x - SC_LG_NGROUP - 1;
+
+	size_t delta_inverse_mask = ZU(-1) << lg_delta;
+	pszind_t mod = ((((psz-1) & delta_inverse_mask) >> lg_delta)) &
+	    ((ZU(1) << SC_LG_NGROUP) - 1);
+
+	pszind_t ind = grp + mod;
+	return ind;
+}
+
+static inline size_t
+sz_pind2sz_compute(pszind_t pind) {
+	if (unlikely(pind == SC_NPSIZES)) {
+		return SC_LARGE_MAXCLASS + PAGE;
+	}
+	size_t grp = pind >> SC_LG_NGROUP;
+	size_t mod = pind & ((ZU(1) << SC_LG_NGROUP) - 1);
+
+	size_t grp_size_mask = ~((!!grp)-1);
+	size_t grp_size = ((ZU(1) << (LG_PAGE + (SC_LG_NGROUP-1))) << grp)
+	    & grp_size_mask;
+
+	size_t shift = (grp == 0) ? 1 : grp;
+	size_t lg_delta = shift + (LG_PAGE-1);
+	size_t mod_size = (mod+1) << lg_delta;
+
+	size_t sz = grp_size + mod_size;
+	return sz;
+}
+
+static inline size_t
+sz_pind2sz_lookup(pszind_t pind) {
+	size_t ret = (size_t)sz_pind2sz_tab[pind];
+	assert(ret == sz_pind2sz_compute(pind));
+	return ret;
+}
+
+static inline size_t
+sz_pind2sz(pszind_t pind) {
+	assert(pind < SC_NPSIZES + 1);
+	return sz_pind2sz_lookup(pind);
+}
+
+static inline size_t
+sz_psz2u(size_t psz) {
+	if (unlikely(psz > SC_LARGE_MAXCLASS)) {
+		return SC_LARGE_MAXCLASS + PAGE;
+	}
+	size_t x = lg_floor((psz<<1)-1);
+	size_t lg_delta = (x < SC_LG_NGROUP + LG_PAGE + 1) ?
+	    LG_PAGE : x - SC_LG_NGROUP - 1;
+	size_t delta = ZU(1) << lg_delta;
+	size_t delta_mask = delta - 1;
+	size_t usize = (psz + delta_mask) & ~delta_mask;
+	return usize;
+}
+
+static inline szind_t
+sz_size2index_compute(size_t size) {
+	if (unlikely(size > SC_LARGE_MAXCLASS)) {
+		return SC_NSIZES;
+	}
+
+	if (size == 0) {
+		return 0;
+	}
+#if (SC_NTINY != 0)
+	if (size <= (ZU(1) << SC_LG_TINY_MAXCLASS)) {
+		szind_t lg_tmin = SC_LG_TINY_MAXCLASS - SC_NTINY + 1;
+		szind_t lg_ceil = lg_floor(pow2_ceil_zu(size));
+		return (lg_ceil < lg_tmin ? 0 : lg_ceil - lg_tmin);
+	}
+#endif
+	{
+		szind_t x = lg_floor((size<<1)-1);
+		szind_t shift = (x < SC_LG_NGROUP + LG_QUANTUM) ? 0 :
+		    x - (SC_LG_NGROUP + LG_QUANTUM);
+		szind_t grp = shift << SC_LG_NGROUP;
+
+		szind_t lg_delta = (x < SC_LG_NGROUP + LG_QUANTUM + 1)
+		    ? LG_QUANTUM : x - SC_LG_NGROUP - 1;
+
+		size_t delta_inverse_mask = ZU(-1) << lg_delta;
+		szind_t mod = ((((size-1) & delta_inverse_mask) >> lg_delta)) &
+		    ((ZU(1) << SC_LG_NGROUP) - 1);
+
+		szind_t index = SC_NTINY + grp + mod;
+		return index;
+	}
+}
+
+JEMALLOC_ALWAYS_INLINE szind_t
+sz_size2index_lookup(size_t size) {
+	assert(size <= SC_LOOKUP_MAXCLASS);
+	szind_t ret = (sz_size2index_tab[(size + (ZU(1) << SC_LG_TINY_MIN) - 1)
+					 >> SC_LG_TINY_MIN]);
+	assert(ret == sz_size2index_compute(size));
+	return ret;
+}
+
+JEMALLOC_ALWAYS_INLINE szind_t
+sz_size2index(size_t size) {
+	if (likely(size <= SC_LOOKUP_MAXCLASS)) {
+		return sz_size2index_lookup(size);
+	}
+	return sz_size2index_compute(size);
+}
+
+static inline size_t
+sz_index2size_compute(szind_t index) {
+#if (SC_NTINY > 0)
+	if (index < SC_NTINY) {
+		return (ZU(1) << (SC_LG_TINY_MAXCLASS - SC_NTINY + 1 + index));
+	}
+#endif
+	{
+		size_t reduced_index = index - SC_NTINY;
+		size_t grp = reduced_index >> SC_LG_NGROUP;
+		size_t mod = reduced_index & ((ZU(1) << SC_LG_NGROUP) -
+		    1);
+
+		size_t grp_size_mask = ~((!!grp)-1);
+		size_t grp_size = ((ZU(1) << (LG_QUANTUM +
+		    (SC_LG_NGROUP-1))) << grp) & grp_size_mask;
+
+		size_t shift = (grp == 0) ? 1 : grp;
+		size_t lg_delta = shift + (LG_QUANTUM-1);
+		size_t mod_size = (mod+1) << lg_delta;
+
+		size_t usize = grp_size + mod_size;
+		return usize;
+	}
+}
+
+JEMALLOC_ALWAYS_INLINE size_t
+sz_index2size_lookup(szind_t index) {
+	size_t ret = (size_t)sz_index2size_tab[index];
+	assert(ret == sz_index2size_compute(index));
+	return ret;
+}
+
+JEMALLOC_ALWAYS_INLINE size_t
+sz_index2size(szind_t index) {
+	assert(index < SC_NSIZES);
+	return sz_index2size_lookup(index);
+}
+
+JEMALLOC_ALWAYS_INLINE size_t
+sz_s2u_compute(size_t size) {
+	if (unlikely(size > SC_LARGE_MAXCLASS)) {
+		return 0;
+	}
+
+	if (size == 0) {
+		size++;
+	}
+#if (SC_NTINY > 0)
+	if (size <= (ZU(1) << SC_LG_TINY_MAXCLASS)) {
+		size_t lg_tmin = SC_LG_TINY_MAXCLASS - SC_NTINY + 1;
+		size_t lg_ceil = lg_floor(pow2_ceil_zu(size));
+		return (lg_ceil < lg_tmin ? (ZU(1) << lg_tmin) :
+		    (ZU(1) << lg_ceil));
+	}
+#endif
+	{
+		size_t x = lg_floor((size<<1)-1);
+		size_t lg_delta = (x < SC_LG_NGROUP + LG_QUANTUM + 1)
+		    ?  LG_QUANTUM : x - SC_LG_NGROUP - 1;
+		size_t delta = ZU(1) << lg_delta;
+		size_t delta_mask = delta - 1;
+		size_t usize = (size + delta_mask) & ~delta_mask;
+		return usize;
+	}
+}
+
+JEMALLOC_ALWAYS_INLINE size_t
+sz_s2u_lookup(size_t size) {
+	size_t ret = sz_index2size_lookup(sz_size2index_lookup(size));
+
+	assert(ret == sz_s2u_compute(size));
+	return ret;
+}
+
+/*
+ * Compute usable size that would result from allocating an object with the
+ * specified size.
+ */
+JEMALLOC_ALWAYS_INLINE size_t
+sz_s2u(size_t size) {
+	if (likely(size <= SC_LOOKUP_MAXCLASS)) {
+		return sz_s2u_lookup(size);
+	}
+	return sz_s2u_compute(size);
+}
+
+/*
+ * Compute usable size that would result from allocating an object with the
+ * specified size and alignment.
+ */
+JEMALLOC_ALWAYS_INLINE size_t
+sz_sa2u(size_t size, size_t alignment) {
+	size_t usize;
+
+	assert(alignment != 0 && ((alignment - 1) & alignment) == 0);
+
+	/* Try for a small size class. */
+	if (size <= SC_SMALL_MAXCLASS && alignment < PAGE) {
+		/*
+		 * Round size up to the nearest multiple of alignment.
+		 *
+		 * This done, we can take advantage of the fact that for each
+		 * small size class, every object is aligned at the smallest
+		 * power of two that is non-zero in the base two representation
+		 * of the size.  For example:
+		 *
+		 *   Size |   Base 2 | Minimum alignment
+		 *   -----+----------+------------------
+		 *     96 |  1100000 |  32
+		 *    144 | 10100000 |  32
+		 *    192 | 11000000 |  64
+		 */
+		usize = sz_s2u(ALIGNMENT_CEILING(size, alignment));
+		if (usize < SC_LARGE_MINCLASS) {
+			return usize;
+		}
+	}
+
+	/* Large size class.  Beware of overflow. */
+
+	if (unlikely(alignment > SC_LARGE_MAXCLASS)) {
+		return 0;
+	}
+
+	/* Make sure result is a large size class. */
+	if (size <= SC_LARGE_MINCLASS) {
+		usize = SC_LARGE_MINCLASS;
+	} else {
+		usize = sz_s2u(size);
+		if (usize < size) {
+			/* size_t overflow. */
+			return 0;
+		}
+	}
+
+	/*
+	 * Calculate the multi-page mapping that large_palloc() would need in
+	 * order to guarantee the alignment.
+	 */
+	if (usize + sz_large_pad + PAGE_CEILING(alignment) - PAGE < usize) {
+		/* size_t overflow. */
+		return 0;
+	}
+	return usize;
+}
+
+#endif /* JEMALLOC_INTERNAL_SIZE_H */

deps/jemalloc/include/jemalloc/internal/tcache.h

@@ -0,0 +1,443 @@
+/******************************************************************************/
+#ifdef JEMALLOC_H_TYPES
+
+typedef struct tcache_bin_info_s tcache_bin_info_t;
+typedef struct tcache_bin_s tcache_bin_t;
+typedef struct tcache_s tcache_t;
+
+/*
+ * tcache pointers close to NULL are used to encode state information that is
+ * used for two purposes: preventing thread caching on a per thread basis and
+ * cleaning up during thread shutdown.
+ */
+#define	TCACHE_STATE_DISABLED		((tcache_t *)(uintptr_t)1)
+#define	TCACHE_STATE_REINCARNATED	((tcache_t *)(uintptr_t)2)
+#define	TCACHE_STATE_PURGATORY		((tcache_t *)(uintptr_t)3)
+#define	TCACHE_STATE_MAX		TCACHE_STATE_PURGATORY
+
+/*
+ * Absolute maximum number of cache slots for each small bin in the thread
+ * cache.  This is an additional constraint beyond that imposed as: twice the
+ * number of regions per run for this size class.
+ *
+ * This constant must be an even number.
+ */
+#define	TCACHE_NSLOTS_SMALL_MAX		200
+
+/* Number of cache slots for large size classes. */
+#define	TCACHE_NSLOTS_LARGE		20
+
+/* (1U << opt_lg_tcache_max) is used to compute tcache_maxclass. */
+#define	LG_TCACHE_MAXCLASS_DEFAULT	15
+
+/*
+ * TCACHE_GC_SWEEP is the approximate number of allocation events between
+ * full GC sweeps.  Integer rounding may cause the actual number to be
+ * slightly higher, since GC is performed incrementally.
+ */
+#define	TCACHE_GC_SWEEP			8192
+
+/* Number of tcache allocation/deallocation events between incremental GCs. */
+#define	TCACHE_GC_INCR							\
+    ((TCACHE_GC_SWEEP / NBINS) + ((TCACHE_GC_SWEEP / NBINS == 0) ? 0 : 1))
+
+#endif /* JEMALLOC_H_TYPES */
+/******************************************************************************/
+#ifdef JEMALLOC_H_STRUCTS
+
+typedef enum {
+	tcache_enabled_false   = 0, /* Enable cast to/from bool. */
+	tcache_enabled_true    = 1,
+	tcache_enabled_default = 2
+} tcache_enabled_t;
+
+/*
+ * Read-only information associated with each element of tcache_t's tbins array
+ * is stored separately, mainly to reduce memory usage.
+ */
+struct tcache_bin_info_s {
+	unsigned	ncached_max;	/* Upper limit on ncached. */
+};
+
+struct tcache_bin_s {
+	tcache_bin_stats_t tstats;
+	int		low_water;	/* Min # cached since last GC. */
+	unsigned	lg_fill_div;	/* Fill (ncached_max >> lg_fill_div). */
+	unsigned	ncached;	/* # of cached objects. */
+	void		**avail;	/* Stack of available objects. */
+};
+
+struct tcache_s {
+	ql_elm(tcache_t) link;		/* Used for aggregating stats. */
+	uint64_t	prof_accumbytes;/* Cleared after arena_prof_accum() */
+	arena_t		*arena;		/* This thread's arena. */
+	unsigned	ev_cnt;		/* Event count since incremental GC. */
+	unsigned	next_gc_bin;	/* Next bin to GC. */
+	tcache_bin_t	tbins[1];	/* Dynamically sized. */
+	/*
+	 * The pointer stacks associated with tbins follow as a contiguous
+	 * array.  During tcache initialization, the avail pointer in each
+	 * element of tbins is initialized to point to the proper offset within
+	 * this array.
+	 */
+};
+
+#endif /* JEMALLOC_H_STRUCTS */
+/******************************************************************************/
+#ifdef JEMALLOC_H_EXTERNS
+
+extern bool	opt_tcache;
+extern ssize_t	opt_lg_tcache_max;
+
+extern tcache_bin_info_t	*tcache_bin_info;
+
+/*
+ * Number of tcache bins.  There are NBINS small-object bins, plus 0 or more
+ * large-object bins.
+ */
+extern size_t			nhbins;
+
+/* Maximum cached size class. */
+extern size_t			tcache_maxclass;
+
+size_t	tcache_salloc(const void *ptr);
+void	tcache_event_hard(tcache_t *tcache);
+void	*tcache_alloc_small_hard(tcache_t *tcache, tcache_bin_t *tbin,
+    size_t binind);
+void	tcache_bin_flush_small(tcache_bin_t *tbin, size_t binind, unsigned rem,
+    tcache_t *tcache);
+void	tcache_bin_flush_large(tcache_bin_t *tbin, size_t binind, unsigned rem,
+    tcache_t *tcache);
+void	tcache_arena_associate(tcache_t *tcache, arena_t *arena);
+void	tcache_arena_dissociate(tcache_t *tcache);
+tcache_t *tcache_create(arena_t *arena);
+void	tcache_destroy(tcache_t *tcache);
+void	tcache_thread_cleanup(void *arg);
+void	tcache_stats_merge(tcache_t *tcache, arena_t *arena);
+bool	tcache_boot0(void);
+bool	tcache_boot1(void);
+
+#endif /* JEMALLOC_H_EXTERNS */
+/******************************************************************************/
+#ifdef JEMALLOC_H_INLINES
+
+#ifndef JEMALLOC_ENABLE_INLINE
+malloc_tsd_protos(JEMALLOC_ATTR(unused), tcache, tcache_t *)
+malloc_tsd_protos(JEMALLOC_ATTR(unused), tcache_enabled, tcache_enabled_t)
+
+void	tcache_event(tcache_t *tcache);
+void	tcache_flush(void);
+bool	tcache_enabled_get(void);
+tcache_t *tcache_get(bool create);
+void	tcache_enabled_set(bool enabled);
+void	*tcache_alloc_easy(tcache_bin_t *tbin);
+void	*tcache_alloc_small(tcache_t *tcache, size_t size, bool zero);
+void	*tcache_alloc_large(tcache_t *tcache, size_t size, bool zero);
+void	tcache_dalloc_small(tcache_t *tcache, void *ptr, size_t binind);
+void	tcache_dalloc_large(tcache_t *tcache, void *ptr, size_t size);
+#endif
+
+#if (defined(JEMALLOC_ENABLE_INLINE) || defined(JEMALLOC_TCACHE_C_))
+/* Map of thread-specific caches. */
+malloc_tsd_externs(tcache, tcache_t *)
+malloc_tsd_funcs(JEMALLOC_ALWAYS_INLINE, tcache, tcache_t *, NULL,
+    tcache_thread_cleanup)
+/* Per thread flag that allows thread caches to be disabled. */
+malloc_tsd_externs(tcache_enabled, tcache_enabled_t)
+malloc_tsd_funcs(JEMALLOC_ALWAYS_INLINE, tcache_enabled, tcache_enabled_t,
+    tcache_enabled_default, malloc_tsd_no_cleanup)
+
+JEMALLOC_INLINE void
+tcache_flush(void)
+{
+	tcache_t *tcache;
+
+	cassert(config_tcache);
+
+	tcache = *tcache_tsd_get();
+	if ((uintptr_t)tcache <= (uintptr_t)TCACHE_STATE_MAX)
+		return;
+	tcache_destroy(tcache);
+	tcache = NULL;
+	tcache_tsd_set(&tcache);
+}
+
+JEMALLOC_INLINE bool
+tcache_enabled_get(void)
+{
+	tcache_enabled_t tcache_enabled;
+
+	cassert(config_tcache);
+
+	tcache_enabled = *tcache_enabled_tsd_get();
+	if (tcache_enabled == tcache_enabled_default) {
+		tcache_enabled = (tcache_enabled_t)opt_tcache;
+		tcache_enabled_tsd_set(&tcache_enabled);
+	}
+
+	return ((bool)tcache_enabled);
+}
+
+JEMALLOC_INLINE void
+tcache_enabled_set(bool enabled)
+{
+	tcache_enabled_t tcache_enabled;
+	tcache_t *tcache;
+
+	cassert(config_tcache);
+
+	tcache_enabled = (tcache_enabled_t)enabled;
+	tcache_enabled_tsd_set(&tcache_enabled);
+	tcache = *tcache_tsd_get();
+	if (enabled) {
+		if (tcache == TCACHE_STATE_DISABLED) {
+			tcache = NULL;
+			tcache_tsd_set(&tcache);
+		}
+	} else /* disabled */ {
+		if (tcache > TCACHE_STATE_MAX) {
+			tcache_destroy(tcache);
+			tcache = NULL;
+		}
+		if (tcache == NULL) {
+			tcache = TCACHE_STATE_DISABLED;
+			tcache_tsd_set(&tcache);
+		}
+	}
+}
+
+JEMALLOC_ALWAYS_INLINE tcache_t *
+tcache_get(bool create)
+{
+	tcache_t *tcache;
+
+	if (config_tcache == false)
+		return (NULL);
+	if (config_lazy_lock && isthreaded == false)
+		return (NULL);
+
+	tcache = *tcache_tsd_get();
+	if ((uintptr_t)tcache <= (uintptr_t)TCACHE_STATE_MAX) {
+		if (tcache == TCACHE_STATE_DISABLED)
+			return (NULL);
+		if (tcache == NULL) {
+			if (create == false) {
+				/*
+				 * Creating a tcache here would cause
+				 * allocation as a side effect of free().
+				 * Ordinarily that would be okay since
+				 * tcache_create() failure is a soft failure
+				 * that doesn't propagate.  However, if TLS
+				 * data are freed via free() as in glibc,
+				 * subtle corruption could result from setting
+				 * a TLS variable after its backing memory is
+				 * freed.
+				 */
+				return (NULL);
+			}
+			if (tcache_enabled_get() == false) {
+				tcache_enabled_set(false); /* Memoize. */
+				return (NULL);
+			}
+			return (tcache_create(choose_arena(NULL)));
+		}
+		if (tcache == TCACHE_STATE_PURGATORY) {
+			/*
+			 * Make a note that an allocator function was called
+			 * after tcache_thread_cleanup() was called.
+			 */
+			tcache = TCACHE_STATE_REINCARNATED;
+			tcache_tsd_set(&tcache);
+			return (NULL);
+		}
+		if (tcache == TCACHE_STATE_REINCARNATED)
+			return (NULL);
+		not_reached();
+	}
+
+	return (tcache);
+}
+
+JEMALLOC_ALWAYS_INLINE void
+tcache_event(tcache_t *tcache)
+{
+
+	if (TCACHE_GC_INCR == 0)
+		return;
+
+	tcache->ev_cnt++;
+	assert(tcache->ev_cnt <= TCACHE_GC_INCR);
+	if (tcache->ev_cnt == TCACHE_GC_INCR)
+		tcache_event_hard(tcache);
+}
+
+JEMALLOC_ALWAYS_INLINE void *
+tcache_alloc_easy(tcache_bin_t *tbin)
+{
+	void *ret;
+
+	if (tbin->ncached == 0) {
+		tbin->low_water = -1;
+		return (NULL);
+	}
+	tbin->ncached--;
+	if ((int)tbin->ncached < tbin->low_water)
+		tbin->low_water = tbin->ncached;
+	ret = tbin->avail[tbin->ncached];
+	return (ret);
+}
+
+JEMALLOC_ALWAYS_INLINE void *
+tcache_alloc_small(tcache_t *tcache, size_t size, bool zero)
+{
+	void *ret;
+	size_t binind;
+	tcache_bin_t *tbin;
+
+	binind = SMALL_SIZE2BIN(size);
+	assert(binind < NBINS);
+	tbin = &tcache->tbins[binind];
+	size = arena_bin_info[binind].reg_size;
+	ret = tcache_alloc_easy(tbin);
+	if (ret == NULL) {
+		ret = tcache_alloc_small_hard(tcache, tbin, binind);
+		if (ret == NULL)
+			return (NULL);
+	}
+	assert(tcache_salloc(ret) == arena_bin_info[binind].reg_size);
+
+	if (zero == false) {
+		if (config_fill) {
+			if (opt_junk) {
+				arena_alloc_junk_small(ret,
+				    &arena_bin_info[binind], false);
+			} else if (opt_zero)
+				memset(ret, 0, size);
+		}
+		VALGRIND_MAKE_MEM_UNDEFINED(ret, size);
+	} else {
+		if (config_fill && opt_junk) {
+			arena_alloc_junk_small(ret, &arena_bin_info[binind],
+			    true);
+		}
+		VALGRIND_MAKE_MEM_UNDEFINED(ret, size);
+		memset(ret, 0, size);
+	}
+
+	if (config_stats)
+		tbin->tstats.nrequests++;
+	if (config_prof)
+		tcache->prof_accumbytes += arena_bin_info[binind].reg_size;
+	tcache_event(tcache);
+	return (ret);
+}
+
+JEMALLOC_ALWAYS_INLINE void *
+tcache_alloc_large(tcache_t *tcache, size_t size, bool zero)
+{
+	void *ret;
+	size_t binind;
+	tcache_bin_t *tbin;
+
+	size = PAGE_CEILING(size);
+	assert(size <= tcache_maxclass);
+	binind = NBINS + (size >> LG_PAGE) - 1;
+	assert(binind < nhbins);
+	tbin = &tcache->tbins[binind];
+	ret = tcache_alloc_easy(tbin);
+	if (ret == NULL) {
+		/*
+		 * Only allocate one large object at a time, because it's quite
+		 * expensive to create one and not use it.
+		 */
+		ret = arena_malloc_large(tcache->arena, size, zero);
+		if (ret == NULL)
+			return (NULL);
+	} else {
+		if (config_prof && prof_promote && size == PAGE) {
+			arena_chunk_t *chunk =
+			    (arena_chunk_t *)CHUNK_ADDR2BASE(ret);
+			size_t pageind = (((uintptr_t)ret - (uintptr_t)chunk) >>
+			    LG_PAGE);
+			arena_mapbits_large_binind_set(chunk, pageind,
+			    BININD_INVALID);
+		}
+		if (zero == false) {
+			if (config_fill) {
+				if (opt_junk)
+					memset(ret, 0xa5, size);
+				else if (opt_zero)
+					memset(ret, 0, size);
+			}
+			VALGRIND_MAKE_MEM_UNDEFINED(ret, size);
+		} else {
+			VALGRIND_MAKE_MEM_UNDEFINED(ret, size);
+			memset(ret, 0, size);
+		}
+
+		if (config_stats)
+			tbin->tstats.nrequests++;
+		if (config_prof)
+			tcache->prof_accumbytes += size;
+	}
+
+	tcache_event(tcache);
+	return (ret);
+}
+
+JEMALLOC_ALWAYS_INLINE void
+tcache_dalloc_small(tcache_t *tcache, void *ptr, size_t binind)
+{
+	tcache_bin_t *tbin;
+	tcache_bin_info_t *tbin_info;
+
+	assert(tcache_salloc(ptr) <= SMALL_MAXCLASS);
+
+	if (config_fill && opt_junk)
+		arena_dalloc_junk_small(ptr, &arena_bin_info[binind]);
+
+	tbin = &tcache->tbins[binind];
+	tbin_info = &tcache_bin_info[binind];
+	if (tbin->ncached == tbin_info->ncached_max) {
+		tcache_bin_flush_small(tbin, binind, (tbin_info->ncached_max >>
+		    1), tcache);
+	}
+	assert(tbin->ncached < tbin_info->ncached_max);
+	tbin->avail[tbin->ncached] = ptr;
+	tbin->ncached++;
+
+	tcache_event(tcache);
+}
+
+JEMALLOC_ALWAYS_INLINE void
+tcache_dalloc_large(tcache_t *tcache, void *ptr, size_t size)
+{
+	size_t binind;
+	tcache_bin_t *tbin;
+	tcache_bin_info_t *tbin_info;
+
+	assert((size & PAGE_MASK) == 0);
+	assert(tcache_salloc(ptr) > SMALL_MAXCLASS);
+	assert(tcache_salloc(ptr) <= tcache_maxclass);
+
+	binind = NBINS + (size >> LG_PAGE) - 1;
+
+	if (config_fill && opt_junk)
+		memset(ptr, 0x5a, size);
+
+	tbin = &tcache->tbins[binind];
+	tbin_info = &tcache_bin_info[binind];
+	if (tbin->ncached == tbin_info->ncached_max) {
+		tcache_bin_flush_large(tbin, binind, (tbin_info->ncached_max >>
+		    1), tcache);
+	}
+	assert(tbin->ncached < tbin_info->ncached_max);
+	tbin->avail[tbin->ncached] = ptr;
+	tbin->ncached++;
+
+	tcache_event(tcache);
+}
+#endif
+
+#endif /* JEMALLOC_H_INLINES */
+/******************************************************************************/

deps/jemalloc/include/jemalloc/internal/tcache_externs.h

@@ -0,0 +1,53 @@
+#ifndef JEMALLOC_INTERNAL_TCACHE_EXTERNS_H
+#define JEMALLOC_INTERNAL_TCACHE_EXTERNS_H
+
+extern bool	opt_tcache;
+extern ssize_t	opt_lg_tcache_max;
+
+extern cache_bin_info_t	*tcache_bin_info;
+
+/*
+ * Number of tcache bins.  There are SC_NBINS small-object bins, plus 0 or more
+ * large-object bins.
+ */
+extern unsigned	nhbins;
+
+/* Maximum cached size class. */
+extern size_t	tcache_maxclass;
+
+/*
+ * Explicit tcaches, managed via the tcache.{create,flush,destroy} mallctls and
+ * usable via the MALLOCX_TCACHE() flag.  The automatic per thread tcaches are
+ * completely disjoint from this data structure.  tcaches starts off as a sparse
+ * array, so it has no physical memory footprint until individual pages are
+ * touched.  This allows the entire array to be allocated the first time an
+ * explicit tcache is created without a disproportionate impact on memory usage.
+ */
+extern tcaches_t	*tcaches;
+
+size_t	tcache_salloc(tsdn_t *tsdn, const void *ptr);
+void	tcache_event_hard(tsd_t *tsd, tcache_t *tcache);
+void	*tcache_alloc_small_hard(tsdn_t *tsdn, arena_t *arena, tcache_t *tcache,
+    cache_bin_t *tbin, szind_t binind, bool *tcache_success);
+void	tcache_bin_flush_small(tsd_t *tsd, tcache_t *tcache, cache_bin_t *tbin,
+    szind_t binind, unsigned rem);
+void	tcache_bin_flush_large(tsd_t *tsd, cache_bin_t *tbin, szind_t binind,
+    unsigned rem, tcache_t *tcache);
+void	tcache_arena_reassociate(tsdn_t *tsdn, tcache_t *tcache,
+    arena_t *arena);
+tcache_t *tcache_create_explicit(tsd_t *tsd);
+void	tcache_cleanup(tsd_t *tsd);
+void	tcache_stats_merge(tsdn_t *tsdn, tcache_t *tcache, arena_t *arena);
+bool	tcaches_create(tsd_t *tsd, unsigned *r_ind);
+void	tcaches_flush(tsd_t *tsd, unsigned ind);
+void	tcaches_destroy(tsd_t *tsd, unsigned ind);
+bool	tcache_boot(tsdn_t *tsdn);
+void tcache_arena_associate(tsdn_t *tsdn, tcache_t *tcache, arena_t *arena);
+void tcache_prefork(tsdn_t *tsdn);
+void tcache_postfork_parent(tsdn_t *tsdn);
+void tcache_postfork_child(tsdn_t *tsdn);
+void tcache_flush(tsd_t *tsd);
+bool tsd_tcache_data_init(tsd_t *tsd);
+bool tsd_tcache_enabled_data_init(tsd_t *tsd);
+
+#endif /* JEMALLOC_INTERNAL_TCACHE_EXTERNS_H */

deps/jemalloc/include/jemalloc/internal/tcache_inlines.h

@@ -0,0 +1,227 @@
+#ifndef JEMALLOC_INTERNAL_TCACHE_INLINES_H
+#define JEMALLOC_INTERNAL_TCACHE_INLINES_H
+
+#include "jemalloc/internal/bin.h"
+#include "jemalloc/internal/jemalloc_internal_types.h"
+#include "jemalloc/internal/sc.h"
+#include "jemalloc/internal/sz.h"
+#include "jemalloc/internal/ticker.h"
+#include "jemalloc/internal/util.h"
+
+static inline bool
+tcache_enabled_get(tsd_t *tsd) {
+	return tsd_tcache_enabled_get(tsd);
+}
+
+static inline void
+tcache_enabled_set(tsd_t *tsd, bool enabled) {
+	bool was_enabled = tsd_tcache_enabled_get(tsd);
+
+	if (!was_enabled && enabled) {
+		tsd_tcache_data_init(tsd);
+	} else if (was_enabled && !enabled) {
+		tcache_cleanup(tsd);
+	}
+	/* Commit the state last.  Above calls check current state. */
+	tsd_tcache_enabled_set(tsd, enabled);
+	tsd_slow_update(tsd);
+}
+
+JEMALLOC_ALWAYS_INLINE void
+tcache_event(tsd_t *tsd, tcache_t *tcache) {
+	if (TCACHE_GC_INCR == 0) {
+		return;
+	}
+
+	if (unlikely(ticker_tick(&tcache->gc_ticker))) {
+		tcache_event_hard(tsd, tcache);
+	}
+}
+
+JEMALLOC_ALWAYS_INLINE void *
+tcache_alloc_small(tsd_t *tsd, arena_t *arena, tcache_t *tcache,
+    size_t size, szind_t binind, bool zero, bool slow_path) {
+	void *ret;
+	cache_bin_t *bin;
+	bool tcache_success;
+	size_t usize JEMALLOC_CC_SILENCE_INIT(0);
+
+	assert(binind < SC_NBINS);
+	bin = tcache_small_bin_get(tcache, binind);
+	ret = cache_bin_alloc_easy(bin, &tcache_success);
+	assert(tcache_success == (ret != NULL));
+	if (unlikely(!tcache_success)) {
+		bool tcache_hard_success;
+		arena = arena_choose(tsd, arena);
+		if (unlikely(arena == NULL)) {
+			return NULL;
+		}
+
+		ret = tcache_alloc_small_hard(tsd_tsdn(tsd), arena, tcache,
+		    bin, binind, &tcache_hard_success);
+		if (tcache_hard_success == false) {
+			return NULL;
+		}
+	}
+
+	assert(ret);
+	/*
+	 * Only compute usize if required.  The checks in the following if
+	 * statement are all static.
+	 */
+	if (config_prof || (slow_path && config_fill) || unlikely(zero)) {
+		usize = sz_index2size(binind);
+		assert(tcache_salloc(tsd_tsdn(tsd), ret) == usize);
+	}
+
+	if (likely(!zero)) {
+		if (slow_path && config_fill) {
+			if (unlikely(opt_junk_alloc)) {
+				arena_alloc_junk_small(ret, &bin_infos[binind],
+				    false);
+			} else if (unlikely(opt_zero)) {
+				memset(ret, 0, usize);
+			}
+		}
+	} else {
+		if (slow_path && config_fill && unlikely(opt_junk_alloc)) {
+			arena_alloc_junk_small(ret, &bin_infos[binind], true);
+		}
+		memset(ret, 0, usize);
+	}
+
+	if (config_stats) {
+		bin->tstats.nrequests++;
+	}
+	if (config_prof) {
+		tcache->prof_accumbytes += usize;
+	}
+	tcache_event(tsd, tcache);
+	return ret;
+}
+
+JEMALLOC_ALWAYS_INLINE void *
+tcache_alloc_large(tsd_t *tsd, arena_t *arena, tcache_t *tcache, size_t size,
+    szind_t binind, bool zero, bool slow_path) {
+	void *ret;
+	cache_bin_t *bin;
+	bool tcache_success;
+
+	assert(binind >= SC_NBINS &&binind < nhbins);
+	bin = tcache_large_bin_get(tcache, binind);
+	ret = cache_bin_alloc_easy(bin, &tcache_success);
+	assert(tcache_success == (ret != NULL));
+	if (unlikely(!tcache_success)) {
+		/*
+		 * Only allocate one large object at a time, because it's quite
+		 * expensive to create one and not use it.
+		 */
+		arena = arena_choose(tsd, arena);
+		if (unlikely(arena == NULL)) {
+			return NULL;
+		}
+
+		ret = large_malloc(tsd_tsdn(tsd), arena, sz_s2u(size), zero);
+		if (ret == NULL) {
+			return NULL;
+		}
+	} else {
+		size_t usize JEMALLOC_CC_SILENCE_INIT(0);
+
+		/* Only compute usize on demand */
+		if (config_prof || (slow_path && config_fill) ||
+		    unlikely(zero)) {
+			usize = sz_index2size(binind);
+			assert(usize <= tcache_maxclass);
+		}
+
+		if (likely(!zero)) {
+			if (slow_path && config_fill) {
+				if (unlikely(opt_junk_alloc)) {
+					memset(ret, JEMALLOC_ALLOC_JUNK,
+					    usize);
+				} else if (unlikely(opt_zero)) {
+					memset(ret, 0, usize);
+				}
+			}
+		} else {
+			memset(ret, 0, usize);
+		}
+
+		if (config_stats) {
+			bin->tstats.nrequests++;
+		}
+		if (config_prof) {
+			tcache->prof_accumbytes += usize;
+		}
+	}
+
+	tcache_event(tsd, tcache);
+	return ret;
+}
+
+JEMALLOC_ALWAYS_INLINE void
+tcache_dalloc_small(tsd_t *tsd, tcache_t *tcache, void *ptr, szind_t binind,
+    bool slow_path) {
+	cache_bin_t *bin;
+	cache_bin_info_t *bin_info;
+
+	assert(tcache_salloc(tsd_tsdn(tsd), ptr)
+	    <= SC_SMALL_MAXCLASS);
+
+	if (slow_path && config_fill && unlikely(opt_junk_free)) {
+		arena_dalloc_junk_small(ptr, &bin_infos[binind]);
+	}
+
+	bin = tcache_small_bin_get(tcache, binind);
+	bin_info = &tcache_bin_info[binind];
+	if (unlikely(!cache_bin_dalloc_easy(bin, bin_info, ptr))) {
+		tcache_bin_flush_small(tsd, tcache, bin, binind,
+		    (bin_info->ncached_max >> 1));
+		bool ret = cache_bin_dalloc_easy(bin, bin_info, ptr);
+		assert(ret);
+	}
+
+	tcache_event(tsd, tcache);
+}
+
+JEMALLOC_ALWAYS_INLINE void
+tcache_dalloc_large(tsd_t *tsd, tcache_t *tcache, void *ptr, szind_t binind,
+    bool slow_path) {
+	cache_bin_t *bin;
+	cache_bin_info_t *bin_info;
+
+	assert(tcache_salloc(tsd_tsdn(tsd), ptr)
+	    > SC_SMALL_MAXCLASS);
+	assert(tcache_salloc(tsd_tsdn(tsd), ptr) <= tcache_maxclass);
+
+	if (slow_path && config_fill && unlikely(opt_junk_free)) {
+		large_dalloc_junk(ptr, sz_index2size(binind));
+	}
+
+	bin = tcache_large_bin_get(tcache, binind);
+	bin_info = &tcache_bin_info[binind];
+	if (unlikely(bin->ncached == bin_info->ncached_max)) {
+		tcache_bin_flush_large(tsd, bin, binind,
+		    (bin_info->ncached_max >> 1), tcache);
+	}
+	assert(bin->ncached < bin_info->ncached_max);
+	bin->ncached++;
+	*(bin->avail - bin->ncached) = ptr;
+
+	tcache_event(tsd, tcache);
+}
+
+JEMALLOC_ALWAYS_INLINE tcache_t *
+tcaches_get(tsd_t *tsd, unsigned ind) {
+	tcaches_t *elm = &tcaches[ind];
+	if (unlikely(elm->tcache == NULL)) {
+		malloc_printf("<jemalloc>: invalid tcache id (%u).\n", ind);
+		abort();
+	} else if (unlikely(elm->tcache == TCACHES_ELM_NEED_REINIT)) {
+		elm->tcache = tcache_create_explicit(tsd);
+	}
+	return elm->tcache;
+}
+
+#endif /* JEMALLOC_INTERNAL_TCACHE_INLINES_H */

deps/jemalloc/include/jemalloc/internal/tcache_structs.h

@@ -0,0 +1,70 @@
+#ifndef JEMALLOC_INTERNAL_TCACHE_STRUCTS_H
+#define JEMALLOC_INTERNAL_TCACHE_STRUCTS_H
+
+#include "jemalloc/internal/cache_bin.h"
+#include "jemalloc/internal/ql.h"
+#include "jemalloc/internal/sc.h"
+#include "jemalloc/internal/ticker.h"
+#include "jemalloc/internal/tsd_types.h"
+
+/* Various uses of this struct need it to be a named type. */
+typedef ql_elm(tsd_t) tsd_link_t;
+
+struct tcache_s {
+	/*
+	 * To minimize our cache-footprint, we put the frequently accessed data
+	 * together at the start of this struct.
+	 */
+
+	/* Cleared after arena_prof_accum(). */
+	uint64_t	prof_accumbytes;
+	/* Drives incremental GC. */
+	ticker_t	gc_ticker;
+	/*
+	 * The pointer stacks associated with bins follow as a contiguous array.
+	 * During tcache initialization, the avail pointer in each element of
+	 * tbins is initialized to point to the proper offset within this array.
+	 */
+	cache_bin_t	bins_small[SC_NBINS];
+
+	/*
+	 * This data is less hot; we can be a little less careful with our
+	 * footprint here.
+	 */
+	/* Lets us track all the tcaches in an arena. */
+	ql_elm(tcache_t) link;
+
+	/* Logically scoped to tsd, but put here for cache layout reasons. */
+	ql_elm(tsd_t) tsd_link;
+	bool in_hook;
+
+	/*
+	 * The descriptor lets the arena find our cache bins without seeing the
+	 * tcache definition.  This enables arenas to aggregate stats across
+	 * tcaches without having a tcache dependency.
+	 */
+	cache_bin_array_descriptor_t cache_bin_array_descriptor;
+
+	/* The arena this tcache is associated with. */
+	arena_t		*arena;
+	/* Next bin to GC. */
+	szind_t		next_gc_bin;
+	/* For small bins, fill (ncached_max >> lg_fill_div). */
+	uint8_t		lg_fill_div[SC_NBINS];
+	/*
+	 * We put the cache bins for large size classes at the end of the
+	 * struct, since some of them might not get used.  This might end up
+	 * letting us avoid touching an extra page if we don't have to.
+	 */
+	cache_bin_t	bins_large[SC_NSIZES-SC_NBINS];
+};
+
+/* Linkage for list of available (previously used) explicit tcache IDs. */
+struct tcaches_s {
+	union {
+		tcache_t	*tcache;
+		tcaches_t	*next;
+	};
+};
+
+#endif /* JEMALLOC_INTERNAL_TCACHE_STRUCTS_H */

deps/jemalloc/include/jemalloc/internal/tcache_types.h

@@ -0,0 +1,59 @@
+#ifndef JEMALLOC_INTERNAL_TCACHE_TYPES_H
+#define JEMALLOC_INTERNAL_TCACHE_TYPES_H
+
+#include "jemalloc/internal/sc.h"
+
+typedef struct tcache_s tcache_t;
+typedef struct tcaches_s tcaches_t;
+
+/*
+ * tcache pointers close to NULL are used to encode state information that is
+ * used for two purposes: preventing thread caching on a per thread basis and
+ * cleaning up during thread shutdown.
+ */
+#define TCACHE_STATE_DISABLED		((tcache_t *)(uintptr_t)1)
+#define TCACHE_STATE_REINCARNATED	((tcache_t *)(uintptr_t)2)
+#define TCACHE_STATE_PURGATORY		((tcache_t *)(uintptr_t)3)
+#define TCACHE_STATE_MAX		TCACHE_STATE_PURGATORY
+
+/*
+ * Absolute minimum number of cache slots for each small bin.
+ */
+#define TCACHE_NSLOTS_SMALL_MIN		20
+
+/*
+ * Absolute maximum number of cache slots for each small bin in the thread
+ * cache.  This is an additional constraint beyond that imposed as: twice the
+ * number of regions per slab for this size class.
+ *
+ * This constant must be an even number.
+ */
+#define TCACHE_NSLOTS_SMALL_MAX		200
+
+/* Number of cache slots for large size classes. */
+#define TCACHE_NSLOTS_LARGE		20
+
+/* (1U << opt_lg_tcache_max) is used to compute tcache_maxclass. */
+#define LG_TCACHE_MAXCLASS_DEFAULT	15
+
+/*
+ * TCACHE_GC_SWEEP is the approximate number of allocation events between
+ * full GC sweeps.  Integer rounding may cause the actual number to be
+ * slightly higher, since GC is performed incrementally.
+ */
+#define TCACHE_GC_SWEEP			8192
+
+/* Number of tcache allocation/deallocation events between incremental GCs. */
+#define TCACHE_GC_INCR							\
+    ((TCACHE_GC_SWEEP / SC_NBINS) + ((TCACHE_GC_SWEEP / SC_NBINS == 0) ? 0 : 1))
+
+/* Used in TSD static initializer only. Real init in tcache_data_init(). */
+#define TCACHE_ZERO_INITIALIZER {0}
+
+/* Used in TSD static initializer only. Will be initialized to opt_tcache. */
+#define TCACHE_ENABLED_ZERO_INITIALIZER false
+
+/* Used for explicit tcache only. Means flushed but not destroyed. */
+#define TCACHES_ELM_NEED_REINIT ((tcache_t *)(uintptr_t)1)
+
+#endif /* JEMALLOC_INTERNAL_TCACHE_TYPES_H */

deps/jemalloc/include/jemalloc/internal/test_hooks.h

@@ -0,0 +1,19 @@
+#ifndef JEMALLOC_INTERNAL_TEST_HOOKS_H
+#define JEMALLOC_INTERNAL_TEST_HOOKS_H
+
+extern JEMALLOC_EXPORT void (*test_hooks_arena_new_hook)();
+extern JEMALLOC_EXPORT void (*test_hooks_libc_hook)();
+
+#define JEMALLOC_HOOK(fn, hook) ((void)(hook != NULL && (hook(), 0)), fn)
+
+#define open JEMALLOC_HOOK(open, test_hooks_libc_hook)
+#define read JEMALLOC_HOOK(read, test_hooks_libc_hook)
+#define write JEMALLOC_HOOK(write, test_hooks_libc_hook)
+#define readlink JEMALLOC_HOOK(readlink, test_hooks_libc_hook)
+#define close JEMALLOC_HOOK(close, test_hooks_libc_hook)
+#define creat JEMALLOC_HOOK(creat, test_hooks_libc_hook)
+#define secure_getenv JEMALLOC_HOOK(secure_getenv, test_hooks_libc_hook)
+/* Note that this is undef'd and re-define'd in src/prof.c. */
+#define _Unwind_Backtrace JEMALLOC_HOOK(_Unwind_Backtrace, test_hooks_libc_hook)
+
+#endif /* JEMALLOC_INTERNAL_TEST_HOOKS_H */

deps/jemalloc/include/jemalloc/internal/ticker.h

@@ -0,0 +1,91 @@
+#ifndef JEMALLOC_INTERNAL_TICKER_H
+#define JEMALLOC_INTERNAL_TICKER_H
+
+#include "jemalloc/internal/util.h"
+
+/**
+ * A ticker makes it easy to count-down events until some limit.  You
+ * ticker_init the ticker to trigger every nticks events.  You then notify it
+ * that an event has occurred with calls to ticker_tick (or that nticks events
+ * have occurred with a call to ticker_ticks), which will return true (and reset
+ * the counter) if the countdown hit zero.
+ */
+
+typedef struct {
+	int32_t tick;
+	int32_t nticks;
+} ticker_t;
+
+static inline void
+ticker_init(ticker_t *ticker, int32_t nticks) {
+	ticker->tick = nticks;
+	ticker->nticks = nticks;
+}
+
+static inline void
+ticker_copy(ticker_t *ticker, const ticker_t *other) {
+	*ticker = *other;
+}
+
+static inline int32_t
+ticker_read(const ticker_t *ticker) {
+	return ticker->tick;
+}
+
+/*
+ * Not intended to be a public API.  Unfortunately, on x86, neither gcc nor
+ * clang seems smart enough to turn
+ *   ticker->tick -= nticks;
+ *   if (unlikely(ticker->tick < 0)) {
+ *     fixup ticker
+ *     return true;
+ *   }
+ *   return false;
+ * into
+ *   subq %nticks_reg, (%ticker_reg)
+ *   js fixup ticker
+ *
+ * unless we force "fixup ticker" out of line.  In that case, gcc gets it right,
+ * but clang now does worse than before.  So, on x86 with gcc, we force it out
+ * of line, but otherwise let the inlining occur.  Ordinarily this wouldn't be
+ * worth the hassle, but this is on the fast path of both malloc and free (via
+ * tcache_event).
+ */
+#if defined(__GNUC__) && !defined(__clang__)				\
+    && (defined(__x86_64__) || defined(__i386__))
+JEMALLOC_NOINLINE
+#endif
+static bool
+ticker_fixup(ticker_t *ticker) {
+	ticker->tick = ticker->nticks;
+	return true;
+}
+
+static inline bool
+ticker_ticks(ticker_t *ticker, int32_t nticks) {
+	ticker->tick -= nticks;
+	if (unlikely(ticker->tick < 0)) {
+		return ticker_fixup(ticker);
+	}
+	return false;
+}
+
+static inline bool
+ticker_tick(ticker_t *ticker) {
+	return ticker_ticks(ticker, 1);
+}
+
+/* 
+ * Try to tick.  If ticker would fire, return true, but rely on
+ * slowpath to reset ticker.
+ */
+static inline bool
+ticker_trytick(ticker_t *ticker) {
+	--ticker->tick;
+	if (unlikely(ticker->tick < 0)) {
+		return true;
+	}
+	return false;
+}
+
+#endif /* JEMALLOC_INTERNAL_TICKER_H */

deps/jemalloc/include/jemalloc/internal/tsd.h

@@ -0,0 +1,415 @@
+#ifndef JEMALLOC_INTERNAL_TSD_H
+#define JEMALLOC_INTERNAL_TSD_H
+
+#include "jemalloc/internal/arena_types.h"
+#include "jemalloc/internal/assert.h"
+#include "jemalloc/internal/bin_types.h"
+#include "jemalloc/internal/jemalloc_internal_externs.h"
+#include "jemalloc/internal/prof_types.h"
+#include "jemalloc/internal/ql.h"
+#include "jemalloc/internal/rtree_tsd.h"
+#include "jemalloc/internal/tcache_types.h"
+#include "jemalloc/internal/tcache_structs.h"
+#include "jemalloc/internal/util.h"
+#include "jemalloc/internal/witness.h"
+
+/*
+ * Thread-Specific-Data layout
+ * --- data accessed on tcache fast path: state, rtree_ctx, stats, prof ---
+ * s: state
+ * e: tcache_enabled
+ * m: thread_allocated (config_stats)
+ * f: thread_deallocated (config_stats)
+ * p: prof_tdata (config_prof)
+ * c: rtree_ctx (rtree cache accessed on deallocation)
+ * t: tcache
+ * --- data not accessed on tcache fast path: arena-related fields ---
+ * d: arenas_tdata_bypass
+ * r: reentrancy_level
+ * x: narenas_tdata
+ * i: iarena
+ * a: arena
+ * o: arenas_tdata
+ * Loading TSD data is on the critical path of basically all malloc operations.
+ * In particular, tcache and rtree_ctx rely on hot CPU cache to be effective.
+ * Use a compact layout to reduce cache footprint.
+ * +--- 64-bit and 64B cacheline; 1B each letter; First byte on the left. ---+
+ * |----------------------------  1st cacheline  ----------------------------|
+ * | sedrxxxx mmmmmmmm ffffffff pppppppp [c * 32  ........ ........ .......] |
+ * |----------------------------  2nd cacheline  ----------------------------|
+ * | [c * 64  ........ ........ ........ ........ ........ ........ .......] |
+ * |----------------------------  3nd cacheline  ----------------------------|
+ * | [c * 32  ........ ........ .......] iiiiiiii aaaaaaaa oooooooo [t...... |
+ * +-------------------------------------------------------------------------+
+ * Note: the entire tcache is embedded into TSD and spans multiple cachelines.
+ *
+ * The last 3 members (i, a and o) before tcache isn't really needed on tcache
+ * fast path.  However we have a number of unused tcache bins and witnesses
+ * (never touched unless config_debug) at the end of tcache, so we place them
+ * there to avoid breaking the cachelines and possibly paging in an extra page.
+ */
+#ifdef JEMALLOC_JET
+typedef void (*test_callback_t)(int *);
+#  define MALLOC_TSD_TEST_DATA_INIT 0x72b65c10
+#  define MALLOC_TEST_TSD \
+    O(test_data,		int,			int)		\
+    O(test_callback,		test_callback_t,	int)
+#  define MALLOC_TEST_TSD_INITIALIZER , MALLOC_TSD_TEST_DATA_INIT, NULL
+#else
+#  define MALLOC_TEST_TSD
+#  define MALLOC_TEST_TSD_INITIALIZER
+#endif
+
+/*  O(name,			type,			nullable type */
+#define MALLOC_TSD							\
+    O(tcache_enabled,		bool,			bool)		\
+    O(arenas_tdata_bypass,	bool,			bool)		\
+    O(reentrancy_level,		int8_t,			int8_t)		\
+    O(narenas_tdata,		uint32_t,		uint32_t)	\
+    O(offset_state,		uint64_t,		uint64_t)	\
+    O(thread_allocated,		uint64_t,		uint64_t)	\
+    O(thread_deallocated,	uint64_t,		uint64_t)	\
+    O(bytes_until_sample,	int64_t,		int64_t)	\
+    O(prof_tdata,		prof_tdata_t *,		prof_tdata_t *)	\
+    O(rtree_ctx,		rtree_ctx_t,		rtree_ctx_t)	\
+    O(iarena,			arena_t *,		arena_t *)	\
+    O(arena,			arena_t *,		arena_t *)	\
+    O(arenas_tdata,		arena_tdata_t *,	arena_tdata_t *)\
+    O(binshards,		tsd_binshards_t,	tsd_binshards_t)\
+    O(tcache,			tcache_t,		tcache_t)	\
+    O(witness_tsd,              witness_tsd_t,		witness_tsdn_t)	\
+    MALLOC_TEST_TSD
+
+#define TSD_INITIALIZER {						\
+    ATOMIC_INIT(tsd_state_uninitialized),				\
+    TCACHE_ENABLED_ZERO_INITIALIZER,					\
+    false,								\
+    0,									\
+    0,									\
+    0,									\
+    0,									\
+    0,									\
+    0,									\
+    NULL,								\
+    RTREE_CTX_ZERO_INITIALIZER,						\
+    NULL,								\
+    NULL,								\
+    NULL,								\
+    TSD_BINSHARDS_ZERO_INITIALIZER,					\
+    TCACHE_ZERO_INITIALIZER,						\
+    WITNESS_TSD_INITIALIZER						\
+    MALLOC_TEST_TSD_INITIALIZER						\
+}
+
+void *malloc_tsd_malloc(size_t size);
+void malloc_tsd_dalloc(void *wrapper);
+void malloc_tsd_cleanup_register(bool (*f)(void));
+tsd_t *malloc_tsd_boot0(void);
+void malloc_tsd_boot1(void);
+void tsd_cleanup(void *arg);
+tsd_t *tsd_fetch_slow(tsd_t *tsd, bool internal);
+void tsd_state_set(tsd_t *tsd, uint8_t new_state);
+void tsd_slow_update(tsd_t *tsd);
+void tsd_prefork(tsd_t *tsd);
+void tsd_postfork_parent(tsd_t *tsd);
+void tsd_postfork_child(tsd_t *tsd);
+
+/*
+ * Call ..._inc when your module wants to take all threads down the slow paths,
+ * and ..._dec when it no longer needs to.
+ */
+void tsd_global_slow_inc(tsdn_t *tsdn);
+void tsd_global_slow_dec(tsdn_t *tsdn);
+bool tsd_global_slow();
+
+enum {
+	/* Common case --> jnz. */
+	tsd_state_nominal = 0,
+	/* Initialized but on slow path. */
+	tsd_state_nominal_slow = 1,
+	/*
+	 * Some thread has changed global state in such a way that all nominal
+	 * threads need to recompute their fast / slow status the next time they
+	 * get a chance.
+	 *
+	 * Any thread can change another thread's status *to* recompute, but
+	 * threads are the only ones who can change their status *from*
+	 * recompute.
+	 */
+	tsd_state_nominal_recompute = 2,
+	/*
+	 * The above nominal states should be lower values.  We use
+	 * tsd_nominal_max to separate nominal states from threads in the
+	 * process of being born / dying.
+	 */
+	tsd_state_nominal_max = 2,
+
+	/*
+	 * A thread might free() during its death as its only allocator action;
+	 * in such scenarios, we need tsd, but set up in such a way that no
+	 * cleanup is necessary.
+	 */
+	tsd_state_minimal_initialized = 3,
+	/* States during which we know we're in thread death. */
+	tsd_state_purgatory = 4,
+	tsd_state_reincarnated = 5,
+	/*
+	 * What it says on the tin; tsd that hasn't been initialized.  Note
+	 * that even when the tsd struct lives in TLS, when need to keep track
+	 * of stuff like whether or not our pthread destructors have been
+	 * scheduled, so this really truly is different than the nominal state.
+	 */
+	tsd_state_uninitialized = 6
+};
+
+/*
+ * Some TSD accesses can only be done in a nominal state.  To enforce this, we
+ * wrap TSD member access in a function that asserts on TSD state, and mangle
+ * field names to prevent touching them accidentally.
+ */
+#define TSD_MANGLE(n) cant_access_tsd_items_directly_use_a_getter_or_setter_##n
+
+#ifdef JEMALLOC_U8_ATOMICS
+#  define tsd_state_t atomic_u8_t
+#  define tsd_atomic_load atomic_load_u8
+#  define tsd_atomic_store atomic_store_u8
+#  define tsd_atomic_exchange atomic_exchange_u8
+#else
+#  define tsd_state_t atomic_u32_t
+#  define tsd_atomic_load atomic_load_u32
+#  define tsd_atomic_store atomic_store_u32
+#  define tsd_atomic_exchange atomic_exchange_u32
+#endif
+
+/* The actual tsd. */
+struct tsd_s {
+	/*
+	 * The contents should be treated as totally opaque outside the tsd
+	 * module.  Access any thread-local state through the getters and
+	 * setters below.
+	 */
+
+	/*
+	 * We manually limit the state to just a single byte.  Unless the 8-bit
+	 * atomics are unavailable (which is rare).
+	 */
+	tsd_state_t state;
+#define O(n, t, nt)							\
+	t TSD_MANGLE(n);
+MALLOC_TSD
+#undef O
+};
+
+JEMALLOC_ALWAYS_INLINE uint8_t
+tsd_state_get(tsd_t *tsd) {
+	/*
+	 * This should be atomic.  Unfortunately, compilers right now can't tell
+	 * that this can be done as a memory comparison, and forces a load into
+	 * a register that hurts fast-path performance.
+	 */
+	/* return atomic_load_u8(&tsd->state, ATOMIC_RELAXED); */
+	return *(uint8_t *)&tsd->state;
+}
+
+/*
+ * Wrapper around tsd_t that makes it possible to avoid implicit conversion
+ * between tsd_t and tsdn_t, where tsdn_t is "nullable" and has to be
+ * explicitly converted to tsd_t, which is non-nullable.
+ */
+struct tsdn_s {
+	tsd_t tsd;
+};
+#define TSDN_NULL ((tsdn_t *)0)
+JEMALLOC_ALWAYS_INLINE tsdn_t *
+tsd_tsdn(tsd_t *tsd) {
+	return (tsdn_t *)tsd;
+}
+
+JEMALLOC_ALWAYS_INLINE bool
+tsdn_null(const tsdn_t *tsdn) {
+	return tsdn == NULL;
+}
+
+JEMALLOC_ALWAYS_INLINE tsd_t *
+tsdn_tsd(tsdn_t *tsdn) {
+	assert(!tsdn_null(tsdn));
+
+	return &tsdn->tsd;
+}
+
+/*
+ * We put the platform-specific data declarations and inlines into their own
+ * header files to avoid cluttering this file.  They define tsd_boot0,
+ * tsd_boot1, tsd_boot, tsd_booted_get, tsd_get_allocates, tsd_get, and tsd_set.
+ */
+#ifdef JEMALLOC_MALLOC_THREAD_CLEANUP
+#include "jemalloc/internal/tsd_malloc_thread_cleanup.h"
+#elif (defined(JEMALLOC_TLS))
+#include "jemalloc/internal/tsd_tls.h"
+#elif (defined(_WIN32))
+#include "jemalloc/internal/tsd_win.h"
+#else
+#include "jemalloc/internal/tsd_generic.h"
+#endif
+
+/*
+ * tsd_foop_get_unsafe(tsd) returns a pointer to the thread-local instance of
+ * foo.  This omits some safety checks, and so can be used during tsd
+ * initialization and cleanup.
+ */
+#define O(n, t, nt)							\
+JEMALLOC_ALWAYS_INLINE t *						\
+tsd_##n##p_get_unsafe(tsd_t *tsd) {					\
+	return &tsd->TSD_MANGLE(n);					\
+}
+MALLOC_TSD
+#undef O
+
+/* tsd_foop_get(tsd) returns a pointer to the thread-local instance of foo. */
+#define O(n, t, nt)							\
+JEMALLOC_ALWAYS_INLINE t *						\
+tsd_##n##p_get(tsd_t *tsd) {						\
+	/*								\
+	 * Because the state might change asynchronously if it's	\
+	 * nominal, we need to make sure that we only read it once.	\
+	 */								\
+	uint8_t state = tsd_state_get(tsd);				\
+	assert(state == tsd_state_nominal ||				\
+	    state == tsd_state_nominal_slow ||				\
+	    state == tsd_state_nominal_recompute ||			\
+	    state == tsd_state_reincarnated ||				\
+	    state == tsd_state_minimal_initialized);			\
+	return tsd_##n##p_get_unsafe(tsd);				\
+}
+MALLOC_TSD
+#undef O
+
+/*
+ * tsdn_foop_get(tsdn) returns either the thread-local instance of foo (if tsdn
+ * isn't NULL), or NULL (if tsdn is NULL), cast to the nullable pointer type.
+ */
+#define O(n, t, nt)							\
+JEMALLOC_ALWAYS_INLINE nt *						\
+tsdn_##n##p_get(tsdn_t *tsdn) {						\
+	if (tsdn_null(tsdn)) {						\
+		return NULL;						\
+	}								\
+	tsd_t *tsd = tsdn_tsd(tsdn);					\
+	return (nt *)tsd_##n##p_get(tsd);				\
+}
+MALLOC_TSD
+#undef O
+
+/* tsd_foo_get(tsd) returns the value of the thread-local instance of foo. */
+#define O(n, t, nt)							\
+JEMALLOC_ALWAYS_INLINE t						\
+tsd_##n##_get(tsd_t *tsd) {						\
+	return *tsd_##n##p_get(tsd);					\
+}
+MALLOC_TSD
+#undef O
+
+/* tsd_foo_set(tsd, val) updates the thread-local instance of foo to be val. */
+#define O(n, t, nt)							\
+JEMALLOC_ALWAYS_INLINE void						\
+tsd_##n##_set(tsd_t *tsd, t val) {					\
+	assert(tsd_state_get(tsd) != tsd_state_reincarnated &&		\
+	    tsd_state_get(tsd) != tsd_state_minimal_initialized);	\
+	*tsd_##n##p_get(tsd) = val;					\
+}
+MALLOC_TSD
+#undef O
+
+JEMALLOC_ALWAYS_INLINE void
+tsd_assert_fast(tsd_t *tsd) {
+	/*
+	 * Note that our fastness assertion does *not* include global slowness
+	 * counters; it's not in general possible to ensure that they won't
+	 * change asynchronously from underneath us.
+	 */
+	assert(!malloc_slow && tsd_tcache_enabled_get(tsd) &&
+	    tsd_reentrancy_level_get(tsd) == 0);
+}
+
+JEMALLOC_ALWAYS_INLINE bool
+tsd_fast(tsd_t *tsd) {
+	bool fast = (tsd_state_get(tsd) == tsd_state_nominal);
+	if (fast) {
+		tsd_assert_fast(tsd);
+	}
+
+	return fast;
+}
+
+JEMALLOC_ALWAYS_INLINE tsd_t *
+tsd_fetch_impl(bool init, bool minimal) {
+	tsd_t *tsd = tsd_get(init);
+
+	if (!init && tsd_get_allocates() && tsd == NULL) {
+		return NULL;
+	}
+	assert(tsd != NULL);
+
+	if (unlikely(tsd_state_get(tsd) != tsd_state_nominal)) {
+		return tsd_fetch_slow(tsd, minimal);
+	}
+	assert(tsd_fast(tsd));
+	tsd_assert_fast(tsd);
+
+	return tsd;
+}
+
+/* Get a minimal TSD that requires no cleanup.  See comments in free(). */
+JEMALLOC_ALWAYS_INLINE tsd_t *
+tsd_fetch_min(void) {
+	return tsd_fetch_impl(true, true);
+}
+
+/* For internal background threads use only. */
+JEMALLOC_ALWAYS_INLINE tsd_t *
+tsd_internal_fetch(void) {
+	tsd_t *tsd = tsd_fetch_min();
+	/* Use reincarnated state to prevent full initialization. */
+	tsd_state_set(tsd, tsd_state_reincarnated);
+
+	return tsd;
+}
+
+JEMALLOC_ALWAYS_INLINE tsd_t *
+tsd_fetch(void) {
+	return tsd_fetch_impl(true, false);
+}
+
+static inline bool
+tsd_nominal(tsd_t *tsd) {
+	return (tsd_state_get(tsd) <= tsd_state_nominal_max);
+}
+
+JEMALLOC_ALWAYS_INLINE tsdn_t *
+tsdn_fetch(void) {
+	if (!tsd_booted_get()) {
+		return NULL;
+	}
+
+	return tsd_tsdn(tsd_fetch_impl(false, false));
+}
+
+JEMALLOC_ALWAYS_INLINE rtree_ctx_t *
+tsd_rtree_ctx(tsd_t *tsd) {
+	return tsd_rtree_ctxp_get(tsd);
+}
+
+JEMALLOC_ALWAYS_INLINE rtree_ctx_t *
+tsdn_rtree_ctx(tsdn_t *tsdn, rtree_ctx_t *fallback) {
+	/*
+	 * If tsd cannot be accessed, initialize the fallback rtree_ctx and
+	 * return a pointer to it.
+	 */
+	if (unlikely(tsdn_null(tsdn))) {
+		rtree_ctx_data_init(fallback);
+		return fallback;
+	}
+	return tsd_rtree_ctx(tsdn_tsd(tsdn));
+}
+
+#endif /* JEMALLOC_INTERNAL_TSD_H */

deps/jemalloc/include/jemalloc/internal/tsd_generic.h

@@ -0,0 +1,163 @@
+#ifdef JEMALLOC_INTERNAL_TSD_GENERIC_H
+#error This file should be included only once, by tsd.h.
+#endif
+#define JEMALLOC_INTERNAL_TSD_GENERIC_H
+
+typedef struct tsd_init_block_s tsd_init_block_t;
+struct tsd_init_block_s {
+	ql_elm(tsd_init_block_t) link;
+	pthread_t thread;
+	void *data;
+};
+
+/* Defined in tsd.c, to allow the mutex headers to have tsd dependencies. */
+typedef struct tsd_init_head_s tsd_init_head_t;
+
+typedef struct {
+	bool initialized;
+	tsd_t val;
+} tsd_wrapper_t;
+
+void *tsd_init_check_recursion(tsd_init_head_t *head,
+    tsd_init_block_t *block);
+void tsd_init_finish(tsd_init_head_t *head, tsd_init_block_t *block);
+
+extern pthread_key_t tsd_tsd;
+extern tsd_init_head_t tsd_init_head;
+extern tsd_wrapper_t tsd_boot_wrapper;
+extern bool tsd_booted;
+
+/* Initialization/cleanup. */
+JEMALLOC_ALWAYS_INLINE void
+tsd_cleanup_wrapper(void *arg) {
+	tsd_wrapper_t *wrapper = (tsd_wrapper_t *)arg;
+
+	if (wrapper->initialized) {
+		wrapper->initialized = false;
+		tsd_cleanup(&wrapper->val);
+		if (wrapper->initialized) {
+			/* Trigger another cleanup round. */
+			if (pthread_setspecific(tsd_tsd, (void *)wrapper) != 0)
+			{
+				malloc_write("<jemalloc>: Error setting TSD\n");
+				if (opt_abort) {
+					abort();
+				}
+			}
+			return;
+		}
+	}
+	malloc_tsd_dalloc(wrapper);
+}
+
+JEMALLOC_ALWAYS_INLINE void
+tsd_wrapper_set(tsd_wrapper_t *wrapper) {
+	if (pthread_setspecific(tsd_tsd, (void *)wrapper) != 0) {
+		malloc_write("<jemalloc>: Error setting TSD\n");
+		abort();
+	}
+}
+
+JEMALLOC_ALWAYS_INLINE tsd_wrapper_t *
+tsd_wrapper_get(bool init) {
+	tsd_wrapper_t *wrapper = (tsd_wrapper_t *)pthread_getspecific(tsd_tsd);
+
+	if (init && unlikely(wrapper == NULL)) {
+		tsd_init_block_t block;
+		wrapper = (tsd_wrapper_t *)
+		    tsd_init_check_recursion(&tsd_init_head, &block);
+		if (wrapper) {
+			return wrapper;
+		}
+		wrapper = (tsd_wrapper_t *)
+		    malloc_tsd_malloc(sizeof(tsd_wrapper_t));
+		block.data = (void *)wrapper;
+		if (wrapper == NULL) {
+			malloc_write("<jemalloc>: Error allocating TSD\n");
+			abort();
+		} else {
+			wrapper->initialized = false;
+      JEMALLOC_DIAGNOSTIC_PUSH
+      JEMALLOC_DIAGNOSTIC_IGNORE_MISSING_STRUCT_FIELD_INITIALIZERS
+			tsd_t initializer = TSD_INITIALIZER;
+      JEMALLOC_DIAGNOSTIC_POP
+			wrapper->val = initializer;
+		}
+		tsd_wrapper_set(wrapper);
+		tsd_init_finish(&tsd_init_head, &block);
+	}
+	return wrapper;
+}
+
+JEMALLOC_ALWAYS_INLINE bool
+tsd_boot0(void) {
+	if (pthread_key_create(&tsd_tsd, tsd_cleanup_wrapper) != 0) {
+		return true;
+	}
+	tsd_wrapper_set(&tsd_boot_wrapper);
+	tsd_booted = true;
+	return false;
+}
+
+JEMALLOC_ALWAYS_INLINE void
+tsd_boot1(void) {
+	tsd_wrapper_t *wrapper;
+	wrapper = (tsd_wrapper_t *)malloc_tsd_malloc(sizeof(tsd_wrapper_t));
+	if (wrapper == NULL) {
+		malloc_write("<jemalloc>: Error allocating TSD\n");
+		abort();
+	}
+	tsd_boot_wrapper.initialized = false;
+	tsd_cleanup(&tsd_boot_wrapper.val);
+	wrapper->initialized = false;
+  JEMALLOC_DIAGNOSTIC_PUSH
+  JEMALLOC_DIAGNOSTIC_IGNORE_MISSING_STRUCT_FIELD_INITIALIZERS
+	tsd_t initializer = TSD_INITIALIZER;
+  JEMALLOC_DIAGNOSTIC_POP
+	wrapper->val = initializer;
+	tsd_wrapper_set(wrapper);
+}
+
+JEMALLOC_ALWAYS_INLINE bool
+tsd_boot(void) {
+	if (tsd_boot0()) {
+		return true;
+	}
+	tsd_boot1();
+	return false;
+}
+
+JEMALLOC_ALWAYS_INLINE bool
+tsd_booted_get(void) {
+	return tsd_booted;
+}
+
+JEMALLOC_ALWAYS_INLINE bool
+tsd_get_allocates(void) {
+	return true;
+}
+
+/* Get/set. */
+JEMALLOC_ALWAYS_INLINE tsd_t *
+tsd_get(bool init) {
+	tsd_wrapper_t *wrapper;
+
+	assert(tsd_booted);
+	wrapper = tsd_wrapper_get(init);
+	if (tsd_get_allocates() && !init && wrapper == NULL) {
+		return NULL;
+	}
+	return &wrapper->val;
+}
+
+JEMALLOC_ALWAYS_INLINE void
+tsd_set(tsd_t *val) {
+	tsd_wrapper_t *wrapper;
+
+	assert(tsd_booted);
+	wrapper = tsd_wrapper_get(true);
+	if (likely(&wrapper->val != val)) {
+		wrapper->val = *(val);
+	}
+	wrapper->initialized = true;
+}