Merge branch 'dev'

.gitignore

@@ -73,3 +73,19 @@ test/include/test/jemalloc_test_defs.h
 /test/unit/*.out
 
 /VERSION
+
+*.pdb
+*.sdf
+*.opendb
+*.opensdf
+*.cachefile
+*.suo
+*.user
+*.sln.docstates
+*.tmp
+/msvc/Win32/
+/msvc/x64/
+/msvc/projects/*/*/Debug*/
+/msvc/projects/*/*/Release*/
+/msvc/projects/*/*/Win32/
+/msvc/projects/*/*/x64/

ChangeLog

@@ -4,6 +4,79 @@ brevity.  Much more detail can be found in the git revision history:
 
     https://github.com/jemalloc/jemalloc
 
+* 4.1.0 (February 28, 2016)
+
+  This release is primarily about optimizations, but it also incorporates a lot
+  of portability-motivated refactoring and enhancements.  Many people worked on
+  this release, to an extent that even with the omission here of minor changes
+  (see git revision history), and of the people who reported and diagnosed
+  issues, so much of the work was contributed that starting with this release,
+  changes are annotated with author credits to help reflect the collaborative
+  effort involved.
+
+  New features:
+  - Implement decay-based unused dirty page purging, a major optimization with
+    mallctl API impact.  This is an alternative to the existing ratio-based
+    unused dirty page purging, and is intended to eventually become the sole
+    purging mechanism.  New mallctls:
+    + opt.purge
+    + opt.decay_time
+    + arena.<i>.decay
+    + arena.<i>.decay_time
+    + arenas.decay_time
+    + stats.arenas.<i>.decay_time
+    (@jasone, @cevans87)
+  - Add --with-malloc-conf, which makes it possible to embed a default
+    options string during configuration.  This was motivated by the desire to
+    specify --with-malloc-conf=purge:decay , since the default must remain
+    purge:ratio until the 5.0.0 release.  (@jasone)
+  - Add MS Visual Studio 2015 support.  (@rustyx, @yuslepukhin)
+  - Make *allocx() size class overflow behavior defined.  The maximum
+    size class is now less than PTRDIFF_MAX to protect applications against
+    numerical overflow, and all allocation functions are guaranteed to indicate
+    errors rather than potentially crashing if the request size exceeds the
+    maximum size class.  (@jasone)
+  - jeprof:
+    + Add raw heap profile support.  (@jasone)
+    + Add --retain and --exclude for backtrace symbol filtering.  (@jasone)
+
+  Optimizations:
+  - Optimize the fast path to combine various bootstrapping and configuration
+    checks and execute more streamlined code in the common case.  (@interwq)
+  - Use linear scan for small bitmaps (used for small object tracking).  In
+    addition to speeding up bitmap operations on 64-bit systems, this reduces
+    allocator metadata overhead by approximately 0.2%.  (@djwatson)
+  - Separate arena_avail trees, which substantially speeds up run tree
+    operations.  (@djwatson)
+  - Use memoization (boot-time-computed table) for run quantization.  Separate
+    arena_avail trees reduced the importance of this optimization.  (@jasone)
+  - Attempt mmap-based in-place huge reallocation.  This can dramatically speed
+    up incremental huge reallocation.  (@jasone)
+
+  Incompatible changes:
+  - Make opt.narenas unsigned rather than size_t.  (@jasone)
+
+  Bug fixes:
+  - Fix stats.cactive accounting regression.  (@rustyx, @jasone)
+  - Handle unaligned keys in hash().  This caused problems for some ARM systems.
+    (@jasone, Christopher Ferris)
+  - Refactor arenas array.  In addition to fixing a fork-related deadlock, this
+    makes arena lookups faster and simpler.  (@jasone)
+  - Move retained memory allocation out of the default chunk allocation
+    function, to a location that gets executed even if the application installs
+    a custom chunk allocation function.  This resolves a virtual memory leak.
+    (@buchgr)
+  - Fix a potential tsd cleanup leak.  (Christopher Ferris, @jasone)
+  - Fix run quantization.  In practice this bug had no impact unless
+    applications requested memory with alignment exceeding one page.
+    (@jasone, @djwatson)
+  - Fix LinuxThreads-specific bootstrapping deadlock.  (Cosmin Paraschiv)
+  - jeprof:
+    + Don't discard curl options if timeout is not defined.  (@djwatson)
+    + Detect failed profile fetches.  (@djwatson)
+  - Fix stats.arenas.<i>.{dss,lg_dirty_mult,decay_time,pactive,pdirty} for
+    --disable-stats case.  (@jasone)
+
 * 4.0.4 (October 24, 2015)
 
   This bugfix release fixes another xallocx() regression.  No other regressions

INSTALL

@@ -84,6 +84,14 @@ any of the following arguments (not a definitive list) to 'configure':
     versions of jemalloc can coexist in the same installation directory.  For
     example, libjemalloc.so.0 becomes libjemalloc<suffix>.so.0.
 
+--with-malloc-conf=<malloc_conf>
+    Embed <malloc_conf> as a run-time options string that is processed prior to
+    the malloc_conf global variable, the /etc/malloc.conf symlink, and the
+    MALLOC_CONF environment variable.  For example, to change the default chunk
+    size to 256 KiB:
+
+      --with-malloc-conf=lg_chunk:18
+
 --disable-cc-silence
     Disable code that silences non-useful compiler warnings.  This is mainly
     useful during development when auditing the set of warnings that are being

Makefile.in

@@ -78,15 +78,32 @@ LIBJEMALLOC := $(LIBPREFIX)jemalloc$(install_suffix)
 # Lists of files.
 BINS := $(objroot)bin/jemalloc-config $(objroot)bin/jemalloc.sh $(objroot)bin/jeprof
 C_HDRS := $(objroot)include/jemalloc/jemalloc$(install_suffix).h
-C_SRCS := $(srcroot)src/jemalloc.c $(srcroot)src/arena.c \
+C_SRCS := $(srcroot)src/jemalloc.c \
-	$(srcroot)src/atomic.c $(srcroot)src/base.c $(srcroot)src/bitmap.c \
+	$(srcroot)src/arena.c \
-	$(srcroot)src/chunk.c $(srcroot)src/chunk_dss.c \
+	$(srcroot)src/atomic.c \
-	$(srcroot)src/chunk_mmap.c $(srcroot)src/ckh.c $(srcroot)src/ctl.c \
+	$(srcroot)src/base.c \
-	$(srcroot)src/extent.c $(srcroot)src/hash.c $(srcroot)src/huge.c \
+	$(srcroot)src/bitmap.c \
-	$(srcroot)src/mb.c $(srcroot)src/mutex.c $(srcroot)src/pages.c \
+	$(srcroot)src/chunk.c \
-	$(srcroot)src/prof.c $(srcroot)src/quarantine.c $(srcroot)src/rtree.c \
+	$(srcroot)src/chunk_dss.c \
-	$(srcroot)src/stats.c $(srcroot)src/tcache.c $(srcroot)src/util.c \
+	$(srcroot)src/chunk_mmap.c \
-	$(srcroot)src/tsd.c
+	$(srcroot)src/ckh.c \
+	$(srcroot)src/ctl.c \
+	$(srcroot)src/extent.c \
+	$(srcroot)src/hash.c \
+	$(srcroot)src/huge.c \
+	$(srcroot)src/mb.c \
+	$(srcroot)src/mutex.c \
+	$(srcroot)src/nstime.c \
+	$(srcroot)src/pages.c \
+	$(srcroot)src/prng.c \
+	$(srcroot)src/prof.c \
+	$(srcroot)src/quarantine.c \
+	$(srcroot)src/rtree.c \
+	$(srcroot)src/stats.c \
+	$(srcroot)src/tcache.c \
+	$(srcroot)src/ticker.c \
+	$(srcroot)src/tsd.c \
+	$(srcroot)src/util.c
 ifeq ($(enable_valgrind), 1)
 C_SRCS += $(srcroot)src/valgrind.c
 endif
@@ -116,10 +133,11 @@ C_TESTLIB_SRCS := $(srcroot)test/src/btalloc.c $(srcroot)test/src/btalloc_0.c \
 	$(srcroot)test/src/mtx.c $(srcroot)test/src/mq.c \
 	$(srcroot)test/src/SFMT.c $(srcroot)test/src/test.c \
 	$(srcroot)test/src/thd.c $(srcroot)test/src/timer.c
-C_UTIL_INTEGRATION_SRCS := $(srcroot)src/util.c
+C_UTIL_INTEGRATION_SRCS := $(srcroot)src/nstime.c $(srcroot)src/util.c
 TESTS_UNIT := $(srcroot)test/unit/atomic.c \
 	$(srcroot)test/unit/bitmap.c \
 	$(srcroot)test/unit/ckh.c \
+	$(srcroot)test/unit/decay.c \
 	$(srcroot)test/unit/hash.c \
 	$(srcroot)test/unit/junk.c \
 	$(srcroot)test/unit/junk_alloc.c \
@@ -129,6 +147,7 @@ TESTS_UNIT := $(srcroot)test/unit/atomic.c \
 	$(srcroot)test/unit/math.c \
 	$(srcroot)test/unit/mq.c \
 	$(srcroot)test/unit/mtx.c \
+	$(srcroot)test/unit/prng.c \
 	$(srcroot)test/unit/prof_accum.c \
 	$(srcroot)test/unit/prof_active.c \
 	$(srcroot)test/unit/prof_gdump.c \
@@ -140,9 +159,13 @@ TESTS_UNIT := $(srcroot)test/unit/atomic.c \
 	$(srcroot)test/unit/quarantine.c \
 	$(srcroot)test/unit/rb.c \
 	$(srcroot)test/unit/rtree.c \
+	$(srcroot)test/unit/run_quantize.c \
 	$(srcroot)test/unit/SFMT.c \
 	$(srcroot)test/unit/size_classes.c \
+	$(srcroot)test/unit/smoothstep.c \
 	$(srcroot)test/unit/stats.c \
+	$(srcroot)test/unit/ticker.c \
+	$(srcroot)test/unit/nstime.c \
 	$(srcroot)test/unit/tsd.c \
 	$(srcroot)test/unit/util.c \
 	$(srcroot)test/unit/zero.c
@@ -349,18 +372,22 @@ stress_dir:
 check_dir: check_unit_dir check_integration_dir
 
 check_unit: tests_unit check_unit_dir
-	$(SHELL) $(objroot)test/test.sh $(TESTS_UNIT:$(srcroot)%.c=$(objroot)%)
+	$(MALLOC_CONF)="purge:ratio" $(SHELL) $(objroot)test/test.sh $(TESTS_UNIT:$(srcroot)%.c=$(objroot)%)
+	$(MALLOC_CONF)="purge:decay" $(SHELL) $(objroot)test/test.sh $(TESTS_UNIT:$(srcroot)%.c=$(objroot)%)
 check_integration_prof: tests_integration check_integration_dir
 ifeq ($(enable_prof), 1)
 	$(MALLOC_CONF)="prof:true" $(SHELL) $(objroot)test/test.sh $(TESTS_INTEGRATION:$(srcroot)%.c=$(objroot)%)
 	$(MALLOC_CONF)="prof:true,prof_active:false" $(SHELL) $(objroot)test/test.sh $(TESTS_INTEGRATION:$(srcroot)%.c=$(objroot)%)
 endif
+check_integration_decay: tests_integration check_integration_dir
+	$(MALLOC_CONF)="purge:decay,decay_time:-1" $(SHELL) $(objroot)test/test.sh $(TESTS_INTEGRATION:$(srcroot)%.c=$(objroot)%)
+	$(MALLOC_CONF)="purge:decay,decay_time:0" $(SHELL) $(objroot)test/test.sh $(TESTS_INTEGRATION:$(srcroot)%.c=$(objroot)%)
+	$(MALLOC_CONF)="purge:decay" $(SHELL) $(objroot)test/test.sh $(TESTS_INTEGRATION:$(srcroot)%.c=$(objroot)%)
 check_integration: tests_integration check_integration_dir
 	$(SHELL) $(objroot)test/test.sh $(TESTS_INTEGRATION:$(srcroot)%.c=$(objroot)%)
 stress: tests_stress stress_dir
 	$(SHELL) $(objroot)test/test.sh $(TESTS_STRESS:$(srcroot)%.c=$(objroot)%)
-check: tests check_dir check_integration_prof
+check: check_unit check_integration check_integration_decay check_integration_prof
-	$(SHELL) $(objroot)test/test.sh $(TESTS_UNIT:$(srcroot)%.c=$(objroot)%) $(TESTS_INTEGRATION:$(srcroot)%.c=$(objroot)%)
 
 ifeq ($(enable_code_coverage), 1)
 coverage_unit: check_unit

bin/jeprof.in

@@ -95,7 +95,7 @@ my @EVINCE = ("evince");    # could also be xpdf or perhaps acroread
 my @KCACHEGRIND = ("kcachegrind");
 my @PS2PDF = ("ps2pdf");
 # These are used for dynamic profiles
-my @URL_FETCHER = ("curl", "-s");
+my @URL_FETCHER = ("curl", "-s", "--fail");
 
 # These are the web pages that servers need to support for dynamic profiles
 my $HEAP_PAGE = "/pprof/heap";
@@ -223,12 +223,14 @@ Call-graph Options:
    --nodefraction=<f>  Hide nodes below <f>*total [default=.005]
    --edgefraction=<f>  Hide edges below <f>*total [default=.001]
    --maxdegree=<n>     Max incoming/outgoing edges per node [default=8]
-   --focus=<regexp>    Focus on nodes matching <regexp>
+   --focus=<regexp>    Focus on backtraces with nodes matching <regexp>
    --thread=<n>        Show profile for thread <n>
-   --ignore=<regexp>   Ignore nodes matching <regexp>
+   --ignore=<regexp>   Ignore backtraces with nodes matching <regexp>
    --scale=<n>         Set GV scaling [default=0]
    --heapcheck         Make nodes with non-0 object counts
                        (i.e. direct leak generators) more visible
+   --retain=<regexp>   Retain only nodes that match <regexp>
+   --exclude=<regexp>  Exclude all nodes that match <regexp>
 
 Miscellaneous:
    --tools=<prefix or binary:fullpath>[,...]   \$PATH for object tool pathnames
@@ -339,6 +341,8 @@ sub Init() {
   $main::opt_ignore = '';
   $main::opt_scale = 0;
   $main::opt_heapcheck = 0;
+  $main::opt_retain = '';
+  $main::opt_exclude = '';
   $main::opt_seconds = 30;
   $main::opt_lib = "";
 
@@ -410,6 +414,8 @@ sub Init() {
              "ignore=s"       => \$main::opt_ignore,
              "scale=i"        => \$main::opt_scale,
              "heapcheck"      => \$main::opt_heapcheck,
+             "retain=s"       => \$main::opt_retain,
+             "exclude=s"      => \$main::opt_exclude,
              "inuse_space!"   => \$main::opt_inuse_space,
              "inuse_objects!" => \$main::opt_inuse_objects,
              "alloc_space!"   => \$main::opt_alloc_space,
@@ -1160,8 +1166,21 @@ sub PrintSymbolizedProfile {
   }
   print '---', "\n";
 
+  my $profile_marker;
+  if ($main::profile_type eq 'heap') {
+    $HEAP_PAGE =~ m,[^/]+$,;    # matches everything after the last slash
+    $profile_marker = $&;
+  } elsif ($main::profile_type eq 'growth') {
+    $GROWTH_PAGE =~ m,[^/]+$,;    # matches everything after the last slash
+    $profile_marker = $&;
+  } elsif ($main::profile_type eq 'contention') {
+    $CONTENTION_PAGE =~ m,[^/]+$,;    # matches everything after the last slash
+    $profile_marker = $&;
+  } else { # elsif ($main::profile_type eq 'cpu')
     $PROFILE_PAGE =~ m,[^/]+$,;    # matches everything after the last slash
-  my $profile_marker = $&;
+    $profile_marker = $&;
+  }
+
   print '--- ', $profile_marker, "\n";
   if (defined($main::collected_profile)) {
     # if used with remote fetch, simply dump the collected profile to output.
@@ -1171,6 +1190,12 @@ sub PrintSymbolizedProfile {
     }
     close(SRC);
   } else {
+    # --raw/http: For everything to work correctly for non-remote profiles, we
+    # would need to extend PrintProfileData() to handle all possible profile
+    # types, re-enable the code that is currently disabled in ReadCPUProfile()
+    # and FixCallerAddresses(), and remove the remote profile dumping code in
+    # the block above.
+    die "--raw/http: jeprof can only dump remote profiles for --raw\n";
     # dump a cpu-format profile to standard out
     PrintProfileData($profile);
   }
@@ -2821,6 +2846,43 @@ sub ExtractCalls {
   return $calls;
 }
 
+sub FilterFrames {
+  my $symbols = shift;
+  my $profile = shift;
+
+  if ($main::opt_retain eq '' && $main::opt_exclude eq '') {
+    return $profile;
+  }
+
+  my $result = {};
+  foreach my $k (keys(%{$profile})) {
+    my $count = $profile->{$k};
+    my @addrs = split(/\n/, $k);
+    my @path = ();
+    foreach my $a (@addrs) {
+      my $sym;
+      if (exists($symbols->{$a})) {
+        $sym = $symbols->{$a}->[0];
+      } else {
+        $sym = $a;
+      }
+      if ($main::opt_retain ne '' && $sym !~ m/$main::opt_retain/) {
+        next;
+      }
+      if ($main::opt_exclude ne '' && $sym =~ m/$main::opt_exclude/) {
+        next;
+      }
+      push(@path, $a);
+    }
+    if (scalar(@path) > 0) {
+      my $reduced_path = join("\n", @path);
+      AddEntry($result, $reduced_path, $count);
+    }
+  }
+
+  return $result;
+}
+
 sub RemoveUninterestingFrames {
   my $symbols = shift;
   my $profile = shift;
@@ -2965,6 +3027,9 @@ sub RemoveUninterestingFrames {
     my $reduced_path = join("\n", @path);
     AddEntry($result, $reduced_path, $count);
   }
+
+  $result = FilterFrames($symbols, $result);
+
   return $result;
 }
 
@@ -3274,7 +3339,7 @@ sub ResolveRedirectionForCurl {
 # Add a timeout flat to URL_FETCHER.  Returns a new list.
 sub AddFetchTimeout {
   my $timeout = shift;
-  my @fetcher = shift;
+  my @fetcher = @_;
   if (defined($timeout)) {
     if (join(" ", @fetcher) =~ m/\bcurl -s/) {
       push(@fetcher, "--max-time", sprintf("%d", $timeout));
@@ -3320,6 +3385,27 @@ sub ReadSymbols {
   return $map;
 }
 
+sub URLEncode {
+  my $str = shift;
+  $str =~ s/([^A-Za-z0-9\-_.!~*'()])/ sprintf "%%%02x", ord $1 /eg;
+  return $str;
+}
+
+sub AppendSymbolFilterParams {
+  my $url = shift;
+  my @params = ();
+  if ($main::opt_retain ne '') {
+    push(@params, sprintf("retain=%s", URLEncode($main::opt_retain)));
+  }
+  if ($main::opt_exclude ne '') {
+    push(@params, sprintf("exclude=%s", URLEncode($main::opt_exclude)));
+  }
+  if (scalar @params > 0) {
+    $url = sprintf("%s?%s", $url, join("&", @params));
+  }
+  return $url;
+}
+
 # Fetches and processes symbols to prepare them for use in the profile output
 # code.  If the optional 'symbol_map' arg is not given, fetches symbols from
 # $SYMBOL_PAGE for all PC values found in profile.  Otherwise, the raw symbols
@@ -3344,9 +3430,11 @@ sub FetchSymbols {
     my $command_line;
     if (join(" ", @URL_FETCHER) =~ m/\bcurl -s/) {
       $url = ResolveRedirectionForCurl($url);
+      $url = AppendSymbolFilterParams($url);
       $command_line = ShellEscape(@URL_FETCHER, "-d", "\@$main::tmpfile_sym",
                                   $url);
     } else {
+      $url = AppendSymbolFilterParams($url);
       $command_line = (ShellEscape(@URL_FETCHER, "--post", $url)
                        . " < " . ShellEscape($main::tmpfile_sym));
     }
@@ -3427,12 +3515,22 @@ sub FetchDynamicProfile {
       }
       $url .= sprintf("seconds=%d", $main::opt_seconds);
       $fetch_timeout = $main::opt_seconds * 1.01 + 60;
+      # Set $profile_type for consumption by PrintSymbolizedProfile.
+      $main::profile_type = 'cpu';
     } else {
       # For non-CPU profiles, we add a type-extension to
       # the target profile file name.
       my $suffix = $path;
       $suffix =~ s,/,.,g;
       $profile_file .= $suffix;
+      # Set $profile_type for consumption by PrintSymbolizedProfile.
+      if ($path =~ m/$HEAP_PAGE/) {
+        $main::profile_type = 'heap';
+      } elsif ($path =~ m/$GROWTH_PAGE/) {
+        $main::profile_type = 'growth';
+      } elsif ($path =~ m/$CONTENTION_PAGE/) {
+        $main::profile_type = 'contention';
+      }
     }
 
     my $profile_dir = $ENV{"JEPROF_TMPDIR"} || ($ENV{HOME} . "/jeprof");
@@ -3730,6 +3828,8 @@ sub ReadProfile {
   my $symbol_marker = $&;
   $PROFILE_PAGE =~ m,[^/]+$,;    # matches everything after the last slash
   my $profile_marker = $&;
+  $HEAP_PAGE =~ m,[^/]+$,;    # matches everything after the last slash
+  my $heap_marker = $&;
 
   # Look at first line to see if it is a heap or a CPU profile.
   # CPU profile may start with no header at all, and just binary data
@@ -3756,7 +3856,13 @@ sub ReadProfile {
     $header = ReadProfileHeader(*PROFILE) || "";
   }
 
+  if ($header =~ m/^--- *($heap_marker|$growth_marker)/o) {
+    # Skip "--- ..." line for profile types that have their own headers.
+    $header = ReadProfileHeader(*PROFILE) || "";
+  }
+
   $main::profile_type = '';
+
   if ($header =~ m/^heap profile:.*$growth_marker/o) {
     $main::profile_type = 'growth';
     $result =  ReadHeapProfile($prog, *PROFILE, $header);
@@ -3808,9 +3914,9 @@ sub ReadProfile {
 # independent implementation.
 sub FixCallerAddresses {
   my $stack = shift;
-  if ($main::use_symbolized_profile) {
+  # --raw/http: Always subtract one from pc's, because PrintSymbolizedProfile()
-    return $stack;
+  # dumps unadjusted profiles.
-  } else {
+  {
     $stack =~ /(\s)/;
     my $delimiter = $1;
     my @addrs = split(' ', $stack);
@@ -3878,12 +3984,7 @@ sub ReadCPUProfile {
     for (my $j = 0; $j < $d; $j++) {
       my $pc = $slots->get($i+$j);
       # Subtract one from caller pc so we map back to call instr.
-      # However, don't do this if we're reading a symbolized profile
-      # file, in which case the subtract-one was done when the file
-      # was written.
-      if ($j > 0 && !$main::use_symbolized_profile) {
       $pc--;
-      }
       $pc = sprintf("%0*x", $address_length, $pc);
       $pcs->{$pc} = 1;
       push @k, $pc;

configure.ac

@@ -1,6 +1,8 @@
 dnl Process this file with autoconf to produce a configure script.
 AC_INIT([Makefile.in])
 
+AC_CONFIG_AUX_DIR([build-aux])
+
 dnl ============================================================================
 dnl Custom macro definitions.
 
@@ -138,6 +140,7 @@ if test "x$CFLAGS" = "x" ; then
     fi
     JE_CFLAGS_APPEND([-Wall])
     JE_CFLAGS_APPEND([-Werror=declaration-after-statement])
+    JE_CFLAGS_APPEND([-Wshorten-64-to-32])
     JE_CFLAGS_APPEND([-pipe])
     JE_CFLAGS_APPEND([-g3])
   elif test "x$je_cv_msvc" = "xyes" ; then
@@ -164,13 +167,18 @@ if test "x${je_cv_msvc}" = "xyes" -a "x${ac_cv_header_inttypes_h}" = "xno"; then
   CPPFLAGS="$CPPFLAGS -I${srcdir}/include/msvc_compat/C99"
 fi
 
-AC_CHECK_SIZEOF([void *])
+if test "x${je_cv_msvc}" = "xyes" ; then
-if test "x${ac_cv_sizeof_void_p}" = "x8" ; then
+  LG_SIZEOF_PTR=LG_SIZEOF_PTR_WIN
-  LG_SIZEOF_PTR=3
+  AC_MSG_RESULT([Using a predefined value for sizeof(void *): 4 for 32-bit, 8 for 64-bit])
-elif test "x${ac_cv_sizeof_void_p}" = "x4" ; then
-  LG_SIZEOF_PTR=2
 else
+  AC_CHECK_SIZEOF([void *])
+  if test "x${ac_cv_sizeof_void_p}" = "x8" ; then
+    LG_SIZEOF_PTR=3
+  elif test "x${ac_cv_sizeof_void_p}" = "x4" ; then
+    LG_SIZEOF_PTR=2
+  else
     AC_MSG_ERROR([Unsupported pointer size: ${ac_cv_sizeof_void_p}])
+  fi
 fi
 AC_DEFINE_UNQUOTED([LG_SIZEOF_PTR], [$LG_SIZEOF_PTR])
 
@@ -194,6 +202,16 @@ else
 fi
 AC_DEFINE_UNQUOTED([LG_SIZEOF_LONG], [$LG_SIZEOF_LONG])
 
+AC_CHECK_SIZEOF([long long])
+if test "x${ac_cv_sizeof_long_long}" = "x8" ; then
+  LG_SIZEOF_LONG_LONG=3
+elif test "x${ac_cv_sizeof_long_long}" = "x4" ; then
+  LG_SIZEOF_LONG_LONG=2
+else
+  AC_MSG_ERROR([Unsupported long long size: ${ac_cv_sizeof_long_long}])
+fi
+AC_DEFINE_UNQUOTED([LG_SIZEOF_LONG_LONG], [$LG_SIZEOF_LONG_LONG])
+
 AC_CHECK_SIZEOF([intmax_t])
 if test "x${ac_cv_sizeof_intmax_t}" = "x16" ; then
   LG_SIZEOF_INTMAX_T=4
@@ -211,6 +229,15 @@ dnl CPU-specific settings.
 CPU_SPINWAIT=""
 case "${host_cpu}" in
   i686|x86_64)
+	if test "x${je_cv_msvc}" = "xyes" ; then
+	    AC_CACHE_VAL([je_cv_pause_msvc],
+	      [JE_COMPILABLE([pause instruction MSVC], [],
+					[[_mm_pause(); return 0;]],
+					[je_cv_pause_msvc])])
+	    if test "x${je_cv_pause_msvc}" = "xyes" ; then
+		CPU_SPINWAIT='_mm_pause()'
+	    fi
+	else
 	    AC_CACHE_VAL([je_cv_pause],
 	      [JE_COMPILABLE([pause instruction], [],
 					[[__asm__ volatile("pause"); return 0;]],
@@ -218,6 +245,7 @@ case "${host_cpu}" in
 	    if test "x${je_cv_pause}" = "xyes" ; then
 		CPU_SPINWAIT='__asm__ volatile("pause")'
 	    fi
+	fi
 	;;
   powerpc)
 	AC_DEFINE_UNQUOTED([HAVE_ALTIVEC], [ ])
@@ -575,6 +603,15 @@ AC_ARG_WITH([install_suffix],
 install_suffix="$INSTALL_SUFFIX"
 AC_SUBST([install_suffix])
 
+dnl Specify default malloc_conf.
+AC_ARG_WITH([malloc_conf],
+  [AS_HELP_STRING([--with-malloc-conf=<malloc_conf>], [config.malloc_conf options string])],
+  [JEMALLOC_CONFIG_MALLOC_CONF="$with_malloc_conf"],
+  [JEMALLOC_CONFIG_MALLOC_CONF=""]
+)
+config_malloc_conf="$JEMALLOC_CONFIG_MALLOC_CONF"
+AC_DEFINE_UNQUOTED([JEMALLOC_CONFIG_MALLOC_CONF], ["$config_malloc_conf"])
+
 dnl Substitute @je_@ in jemalloc_protos.h.in, primarily to make generation of
 dnl jemalloc_protos_jet.h easy.
 je_="je_"
@@ -1014,7 +1051,7 @@ dnl ============================================================================
 dnl Check for  __builtin_ffsl(), then ffsl(3), and fail if neither are found.
 dnl One of those two functions should (theoretically) exist on all platforms
 dnl that jemalloc currently has a chance of functioning on without modification.
-dnl We additionally assume ffs() or __builtin_ffs() are defined if
+dnl We additionally assume ffs[ll]() or __builtin_ffs[ll]() are defined if
 dnl ffsl() or __builtin_ffsl() are defined, respectively.
 JE_COMPILABLE([a program using __builtin_ffsl], [
 #include <stdio.h>
@@ -1027,6 +1064,7 @@ JE_COMPILABLE([a program using __builtin_ffsl], [
 	}
 ], [je_cv_gcc_builtin_ffsl])
 if test "x${je_cv_gcc_builtin_ffsl}" = "xyes" ; then
+  AC_DEFINE([JEMALLOC_INTERNAL_FFSLL], [__builtin_ffsll])
   AC_DEFINE([JEMALLOC_INTERNAL_FFSL], [__builtin_ffsl])
   AC_DEFINE([JEMALLOC_INTERNAL_FFS], [__builtin_ffs])
 else
@@ -1041,6 +1079,7 @@ else
 	}
   ], [je_cv_function_ffsl])
   if test "x${je_cv_function_ffsl}" = "xyes" ; then
+    AC_DEFINE([JEMALLOC_INTERNAL_FFSLL], [ffsll])
     AC_DEFINE([JEMALLOC_INTERNAL_FFSL], [ffsl])
     AC_DEFINE([JEMALLOC_INTERNAL_FFS], [ffs])
   else
@@ -1100,7 +1139,7 @@ if test "x$LG_PAGE" = "xdetect"; then
     if (f == NULL) {
 	return 1;
     }
-    fprintf(f, "%d\n", result);
+    fprintf(f, "%d", result);
     fclose(f);
 
     return 0;
@@ -1724,6 +1763,7 @@ AC_MSG_RESULT([JEMALLOC_PREFIX    : ${JEMALLOC_PREFIX}])
 AC_MSG_RESULT([JEMALLOC_PRIVATE_NAMESPACE])
 AC_MSG_RESULT([                   : ${JEMALLOC_PRIVATE_NAMESPACE}])
 AC_MSG_RESULT([install_suffix     : ${install_suffix}])
+AC_MSG_RESULT([malloc_conf        : ${config_malloc_conf}])
 AC_MSG_RESULT([autogen            : ${enable_autogen}])
 AC_MSG_RESULT([cc-silence         : ${enable_cc_silence}])
 AC_MSG_RESULT([debug              : ${enable_debug}])

doc/jemalloc.xml.in

@@ -310,16 +310,14 @@
       <para>The <function>mallocx<parameter/></function> function allocates at
       least <parameter>size</parameter> bytes of memory, and returns a pointer
       to the base address of the allocation.  Behavior is undefined if
-      <parameter>size</parameter> is <constant>0</constant>, or if request size
+      <parameter>size</parameter> is <constant>0</constant>.</para>
-      overflows due to size class and/or alignment constraints.</para>
 
       <para>The <function>rallocx<parameter/></function> function resizes the
       allocation at <parameter>ptr</parameter> to be at least
       <parameter>size</parameter> bytes, and returns a pointer to the base
       address of the resulting allocation, which may or may not have moved from
       its original location.  Behavior is undefined if
-      <parameter>size</parameter> is <constant>0</constant>, or if request size
+      <parameter>size</parameter> is <constant>0</constant>.</para>
-      overflows due to size class and/or alignment constraints.</para>
 
       <para>The <function>xallocx<parameter/></function> function resizes the
       allocation at <parameter>ptr</parameter> in place to be at least
@@ -354,10 +352,10 @@
       memory, but it performs the same size computation as the
       <function>mallocx<parameter/></function> function, and returns the real
       size of the allocation that would result from the equivalent
-      <function>mallocx<parameter/></function> function call.  Behavior is
+      <function>mallocx<parameter/></function> function call, or
-      undefined if <parameter>size</parameter> is <constant>0</constant>, or if
+      <constant>0</constant> if the inputs exceed the maximum supported size
-      request size overflows due to size class and/or alignment
+      class and/or alignment.  Behavior is undefined if
-      constraints.</para>
+      <parameter>size</parameter> is <constant>0</constant>.</para>
 
       <para>The <function>mallctl<parameter/></function> function provides a
       general interface for introspecting the memory allocator, as well as
@@ -455,19 +453,20 @@ for (i = 0; i < nbins; i++) {
     routines, the allocator initializes its internals based in part on various
     options that can be specified at compile- or run-time.</para>
 
-    <para>The string pointed to by the global variable
+    <para>The string specified via <option>--with-malloc-conf</option>, the
-    <varname>malloc_conf</varname>, the &ldquo;name&rdquo; of the file
+    string pointed to by the global variable <varname>malloc_conf</varname>, the
-    referenced by the symbolic link named <filename
+    &ldquo;name&rdquo; of the file referenced by the symbolic link named
-    class="symlink">/etc/malloc.conf</filename>, and the value of the
+    <filename class="symlink">/etc/malloc.conf</filename>, and the value of the
     environment variable <envar>MALLOC_CONF</envar>, will be interpreted, in
     that order, from left to right as options.  Note that
     <varname>malloc_conf</varname> may be read before
     <function>main<parameter/></function> is entered, so the declaration of
     <varname>malloc_conf</varname> should specify an initializer that contains
-    the final value to be read by jemalloc.  <varname>malloc_conf</varname> is
+    the final value to be read by jemalloc.  <option>--with-malloc-conf</option>
-    a compile-time setting, whereas <filename
+    and <varname>malloc_conf</varname> are compile-time mechanisms, whereas
-    class="symlink">/etc/malloc.conf</filename> and <envar>MALLOC_CONF</envar>
+    <filename class="symlink">/etc/malloc.conf</filename> and
-    can be safely set any time prior to program invocation.</para>
+    <envar>MALLOC_CONF</envar> can be safely set any time prior to program
+    invocation.</para>
 
     <para>An options string is a comma-separated list of option:value pairs.
     There is one key corresponding to each <link
@@ -517,23 +516,18 @@ for (i = 0; i < nbins; i++) {
     common case, but it increases memory usage and fragmentation, since a
     bounded number of objects can remain allocated in each thread cache.</para>
 
-    <para>Memory is conceptually broken into equal-sized chunks, where the
+    <para>Memory is conceptually broken into equal-sized chunks, where the chunk
-    chunk size is a power of two that is greater than the page size.  Chunks
+    size is a power of two that is greater than the page size.  Chunks are
-    are always aligned to multiples of the chunk size.  This alignment makes it
+    always aligned to multiples of the chunk size.  This alignment makes it
-    possible to find metadata for user objects very quickly.</para>
+    possible to find metadata for user objects very quickly.  User objects are
-
+    broken into three categories according to size: small, large, and huge.
-    <para>User objects are broken into three categories according to size:
+    Multiple small and large objects can reside within a single chunk, whereas
-    small, large, and huge.  Small and large objects are managed entirely by
+    huge objects each have one or more chunks backing them.  Each chunk that
-    arenas; huge objects are additionally aggregated in a single data structure
+    contains small and/or large objects tracks its contents as runs of
-    that is shared by all threads.  Huge objects are typically used by
-    applications infrequently enough that this single data structure is not a
-    scalability issue.</para>
-
-    <para>Each chunk that is managed by an arena tracks its contents as runs of
     contiguous pages (unused, backing a set of small objects, or backing one
-    large object).  The combination of chunk alignment and chunk page maps
+    large object).  The combination of chunk alignment and chunk page maps makes
-    makes it possible to determine all metadata regarding small and large
+    it possible to determine all metadata regarding small and large allocations
-    allocations in constant time.</para>
+    in constant time.</para>
 
     <para>Small objects are managed in groups by page runs.  Each run maintains
     a bitmap to track which regions are in use.  Allocation requests that are no
@@ -776,6 +770,17 @@ for (i = 0; i < nbins; i++) {
         during build configuration.</para></listitem>
       </varlistentry>
 
+      <varlistentry id="config.malloc_conf">
+        <term>
+          <mallctl>config.malloc_conf</mallctl>
+          (<type>const char *</type>)
+          <literal>r-</literal>
+        </term>
+        <listitem><para>Embedded configure-time-specified run-time options
+        string, empty unless <option>--with-malloc-conf</option> was specified
+        during build configuration.</para></listitem>
+      </varlistentry>
+
       <varlistentry id="config.munmap">
         <term>
           <mallctl>config.munmap</mallctl>
@@ -929,7 +934,7 @@ for (i = 0; i < nbins; i++) {
       <varlistentry id="opt.narenas">
         <term>
           <mallctl>opt.narenas</mallctl>
-          (<type>size_t</type>)
+          (<type>unsigned</type>)
           <literal>r-</literal>
         </term>
         <listitem><para>Maximum number of arenas to use for automatic
@@ -937,6 +942,20 @@ for (i = 0; i < nbins; i++) {
         number of CPUs, or one if there is a single CPU.</para></listitem>
       </varlistentry>
 
+      <varlistentry id="opt.purge">
+        <term>
+          <mallctl>opt.purge</mallctl>
+          (<type>const char *</type>)
+          <literal>r-</literal>
+        </term>
+        <listitem><para>Purge mode is &ldquo;ratio&rdquo; (default) or
+        &ldquo;decay&rdquo;.  See <link
+        linkend="opt.lg_dirty_mult"><mallctl>opt.lg_dirty_mult</mallctl></link>
+        for details of the ratio mode.  See <link
+        linkend="opt.decay_time"><mallctl>opt.decay_time</mallctl></link> for
+        details of the decay mode.</para></listitem>
+      </varlistentry>
+
       <varlistentry id="opt.lg_dirty_mult">
         <term>
           <mallctl>opt.lg_dirty_mult</mallctl>
@@ -959,6 +978,26 @@ for (i = 0; i < nbins; i++) {
         for related dynamic control options.</para></listitem>
       </varlistentry>
 
+      <varlistentry id="opt.decay_time">
+        <term>
+          <mallctl>opt.decay_time</mallctl>
+          (<type>ssize_t</type>)
+          <literal>r-</literal>
+        </term>
+        <listitem><para>Approximate time in seconds from the creation of a set
+        of unused dirty pages until an equivalent set of unused dirty pages is
+        purged and/or reused.  The pages are incrementally purged according to a
+        sigmoidal decay curve that starts and ends with zero purge rate.  A
+        decay time of 0 causes all unused dirty pages to be purged immediately
+        upon creation.  A decay time of -1 disables purging.  The default decay
+        time is 10 seconds.  See <link
+        linkend="arenas.decay_time"><mallctl>arenas.decay_time</mallctl></link>
+        and <link
+        linkend="arena.i.decay_time"><mallctl>arena.&lt;i&gt;.decay_time</mallctl></link>
+        for related dynamic control options.
+        </para></listitem>
+      </varlistentry>
+
       <varlistentry id="opt.stats_print">
         <term>
           <mallctl>opt.stats_print</mallctl>
@@ -1150,7 +1189,8 @@ malloc_conf = "xmalloc:true";]]></programlisting>
         the <command>jeprof</command> command, which is based on the
         <command>pprof</command> that is developed as part of the <ulink
         url="http://code.google.com/p/gperftools/">gperftools
-        package</ulink>.</para></listitem>
+        package</ulink>.  See <link linkend="heap_profile_format">HEAP PROFILE
+        FORMAT</link> for heap profile format documentation.</para></listitem>
       </varlistentry>
 
       <varlistentry id="opt.prof_prefix">
@@ -1467,7 +1507,7 @@ malloc_conf = "xmalloc:true";]]></programlisting>
         <listitem><para>Flush the specified thread-specific cache (tcache).  The
         same considerations apply to this interface as to <link
         linkend="thread.tcache.flush"><mallctl>thread.tcache.flush</mallctl></link>,
-        except that the tcache will never be automatically be discarded.
+        except that the tcache will never be automatically discarded.
         </para></listitem>
       </varlistentry>
 
@@ -1489,12 +1529,27 @@ malloc_conf = "xmalloc:true";]]></programlisting>
           (<type>void</type>)
           <literal>--</literal>
         </term>
-        <listitem><para>Purge unused dirty pages for arena &lt;i&gt;, or for
+        <listitem><para>Purge all unused dirty pages for arena &lt;i&gt;, or for
         all arenas if &lt;i&gt; equals <link
         linkend="arenas.narenas"><mallctl>arenas.narenas</mallctl></link>.
         </para></listitem>
       </varlistentry>
 
+      <varlistentry id="arena.i.decay">
+        <term>
+          <mallctl>arena.&lt;i&gt;.decay</mallctl>
+          (<type>void</type>)
+          <literal>--</literal>
+        </term>
+        <listitem><para>Trigger decay-based purging of unused dirty pages for
+        arena &lt;i&gt;, or for all arenas if &lt;i&gt; equals <link
+        linkend="arenas.narenas"><mallctl>arenas.narenas</mallctl></link>.
+        The proportion of unused dirty pages to be purged depends on the current
+        time; see <link
+        linkend="opt.decay_time"><mallctl>opt.decay_time</mallctl></link> for
+        details.</para></listitem>
+      </varlistentry>
+
       <varlistentry id="arena.i.dss">
         <term>
           <mallctl>arena.&lt;i&gt;.dss</mallctl>
@@ -1523,6 +1578,22 @@ malloc_conf = "xmalloc:true";]]></programlisting>
         for additional information.</para></listitem>
       </varlistentry>
 
+      <varlistentry id="arena.i.decay_time">
+        <term>
+          <mallctl>arena.&lt;i&gt;.decay_time</mallctl>
+          (<type>ssize_t</type>)
+          <literal>rw</literal>
+        </term>
+        <listitem><para>Current per-arena approximate time in seconds from the
+        creation of a set of unused dirty pages until an equivalent set of
+        unused dirty pages is purged and/or reused.  Each time this interface is
+        set, all currently unused dirty pages are considered to have fully
+        decayed, which causes immediate purging of all unused dirty pages unless
+        the decay time is set to -1 (i.e. purging disabled).  See <link
+        linkend="opt.decay_time"><mallctl>opt.decay_time</mallctl></link> for
+        additional information.</para></listitem>
+      </varlistentry>
+
       <varlistentry id="arena.i.chunk_hooks">
         <term>
           <mallctl>arena.&lt;i&gt;.chunk_hooks</mallctl>
@@ -1757,6 +1828,21 @@ typedef struct {
         for additional information.</para></listitem>
       </varlistentry>
 
+      <varlistentry id="arenas.decay_time">
+        <term>
+          <mallctl>arenas.decay_time</mallctl>
+          (<type>ssize_t</type>)
+          <literal>rw</literal>
+        </term>
+        <listitem><para>Current default per-arena approximate time in seconds
+        from the creation of a set of unused dirty pages until an equivalent set
+        of unused dirty pages is purged and/or reused, used to initialize <link
+        linkend="arena.i.decay_time"><mallctl>arena.&lt;i&gt;.decay_time</mallctl></link>
+        during arena creation.  See <link
+        linkend="opt.decay_time"><mallctl>opt.decay_time</mallctl></link> for
+        additional information.</para></listitem>
+      </varlistentry>
+
       <varlistentry id="arenas.quantum">
         <term>
           <mallctl>arenas.quantum</mallctl>
@@ -2101,6 +2187,19 @@ typedef struct {
         for details.</para></listitem>
       </varlistentry>
 
+      <varlistentry id="stats.arenas.i.decay_time">
+        <term>
+          <mallctl>stats.arenas.&lt;i&gt;.decay_time</mallctl>
+          (<type>ssize_t</type>)
+          <literal>r-</literal>
+        </term>
+        <listitem><para>Approximate time in seconds from the creation of a set
+        of unused dirty pages until an equivalent set of unused dirty pages is
+        purged and/or reused.  See <link
+        linkend="opt.decay_time"><mallctl>opt.decay_time</mallctl></link>
+        for details.</para></listitem>
+      </varlistentry>
+
       <varlistentry id="stats.arenas.i.nthreads">
         <term>
           <mallctl>stats.arenas.&lt;i&gt;.nthreads</mallctl>
@@ -2523,6 +2622,53 @@ typedef struct {
       </varlistentry>
     </variablelist>
   </refsect1>
+  <refsect1 id="heap_profile_format">
+    <title>HEAP PROFILE FORMAT</title>
+    <para>Although the heap profiling functionality was originally designed to
+    be compatible with the
+    <command>pprof</command> command that is developed as part of the <ulink
+    url="http://code.google.com/p/gperftools/">gperftools
+    package</ulink>, the addition of per thread heap profiling functionality
+    required a different heap profile format.  The <command>jeprof</command>
+    command is derived from <command>pprof</command>, with enhancements to
+    support the heap profile format described here.</para>
+
+    <para>In the following hypothetical heap profile, <constant>[...]</constant>
+    indicates elision for the sake of compactness.  <programlisting><![CDATA[
+heap_v2/524288
+  t*: 28106: 56637512 [0: 0]
+  [...]
+  t3: 352: 16777344 [0: 0]
+  [...]
+  t99: 17754: 29341640 [0: 0]
+  [...]
+@ 0x5f86da8 0x5f5a1dc [...] 0x29e4d4e 0xa200316 0xabb2988 [...]
+  t*: 13: 6688 [0: 0]
+  t3: 12: 6496 [0: ]
+  t99: 1: 192 [0: 0]
+[...]
+
+MAPPED_LIBRARIES:
+[...]]]></programlisting> The following matches the above heap profile, but most
+tokens are replaced with <constant>&lt;description&gt;</constant> to indicate
+descriptions of the corresponding fields.  <programlisting><![CDATA[
+<heap_profile_format_version>/<mean_sample_interval>
+  <aggregate>: <curobjs>: <curbytes> [<cumobjs>: <cumbytes>]
+  [...]
+  <thread_3_aggregate>: <curobjs>: <curbytes>[<cumobjs>: <cumbytes>]
+  [...]
+  <thread_99_aggregate>: <curobjs>: <curbytes>[<cumobjs>: <cumbytes>]
+  [...]
+@ <top_frame> <frame> [...] <frame> <frame> <frame> [...]
+  <backtrace_aggregate>: <curobjs>: <curbytes> [<cumobjs>: <cumbytes>]
+  <backtrace_thread_3>: <curobjs>: <curbytes> [<cumobjs>: <cumbytes>]
+  <backtrace_thread_99>: <curobjs>: <curbytes> [<cumobjs>: <cumbytes>]
+[...]
+
+MAPPED_LIBRARIES:
+</proc/<pid>/maps>]]></programlisting></para>
+  </refsect1>
+
   <refsect1 id="debugging_malloc_problems">
     <title>DEBUGGING MALLOC PROBLEMS</title>
     <para>When debugging, it is a good idea to configure/build jemalloc with

include/jemalloc/internal/arena.h

@@ -23,6 +23,18 @@
  */
 #define	LG_DIRTY_MULT_DEFAULT	3
 
+typedef enum {
+	purge_mode_ratio = 0,
+	purge_mode_decay = 1,
+
+	purge_mode_limit = 2
+} purge_mode_t;
+#define	PURGE_DEFAULT		purge_mode_ratio
+/* Default decay time in seconds. */
+#define	DECAY_TIME_DEFAULT	10
+/* Number of event ticks between time checks. */
+#define	DECAY_NTICKS_PER_UPDATE	1000
+
 typedef struct arena_runs_dirty_link_s arena_runs_dirty_link_t;
 typedef struct arena_run_s arena_run_t;
 typedef struct arena_chunk_map_bits_s arena_chunk_map_bits_t;
@@ -31,6 +43,7 @@ typedef struct arena_chunk_s arena_chunk_t;
 typedef struct arena_bin_info_s arena_bin_info_t;
 typedef struct arena_bin_s arena_bin_t;
 typedef struct arena_s arena_t;
+typedef struct arena_tdata_s arena_tdata_t;
 
 #endif /* JEMALLOC_H_TYPES */
 /******************************************************************************/
@@ -162,7 +175,6 @@ struct arena_chunk_map_misc_s {
 		arena_run_t			run;
 	};
 };
-typedef rb_tree(arena_chunk_map_misc_t) arena_avail_tree_t;
 typedef rb_tree(arena_chunk_map_misc_t) arena_run_tree_t;
 #endif /* JEMALLOC_ARENA_STRUCTS_A */
 
@@ -278,14 +290,14 @@ struct arena_s {
 
 	/*
 	 * Number of threads currently assigned to this arena.  This field is
-	 * protected by arenas_lock.
+	 * synchronized via atomic operations.
 	 */
 	unsigned		nthreads;
 
 	/*
 	 * There are three classes of arena operations from a locking
 	 * perspective:
-	 * 1) Thread assignment (modifies nthreads) is protected by arenas_lock.
+	 * 1) Thread assignment (modifies nthreads) is synchronized via atomics.
 	 * 2) Bin-related operations are protected by bin locks.
 	 * 3) Chunk- and run-related operations are protected by this mutex.
 	 */
@@ -324,7 +336,7 @@ struct arena_s {
 	/* Minimum ratio (log base 2) of nactive:ndirty. */
 	ssize_t			lg_dirty_mult;
 
-	/* True if a thread is currently executing arena_purge(). */
+	/* True if a thread is currently executing arena_purge_to_limit(). */
 	bool			purging;
 
 	/* Number of pages in active runs and huge regions. */
@@ -338,12 +350,6 @@ struct arena_s {
 	 */
 	size_t			ndirty;
 
-	/*
-	 * Size/address-ordered tree of this arena's available runs.  The tree
-	 * is used for first-best-fit run allocation.
-	 */
-	arena_avail_tree_t	runs_avail;
-
 	/*
 	 * Unused dirty memory this arena manages.  Dirty memory is conceptually
 	 * tracked as an arbitrarily interleaved LRU of dirty runs and cached
@@ -375,6 +381,53 @@ struct arena_s {
 	arena_runs_dirty_link_t	runs_dirty;
 	extent_node_t		chunks_cache;
 
+	/*
+	 * Approximate time in seconds from the creation of a set of unused
+	 * dirty pages until an equivalent set of unused dirty pages is purged
+	 * and/or reused.
+	 */
+	ssize_t			decay_time;
+	/* decay_time / SMOOTHSTEP_NSTEPS. */
+	nstime_t		decay_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		decay_epoch;
+	/* decay_deadline randomness generator. */
+	uint64_t		decay_jitter_state;
+	/*
+	 * Deadline for current epoch.  This is the sum of decay_interval and
+	 * per epoch jitter which is a uniform random variable in
+	 * [0..decay_interval).  Epochs always advance by precise multiples of
+	 * decay_interval, but we randomize the deadline to reduce the
+	 * likelihood of arenas purging in lockstep.
+	 */
+	nstime_t		decay_deadline;
+	/*
+	 * Number of dirty pages at beginning of current epoch.  During epoch
+	 * advancement we use the delta between decay_ndirty and ndirty to
+	 * determine how many dirty pages, if any, were generated, and record
+	 * the result in decay_backlog.
+	 */
+	size_t			decay_ndirty;
+	/*
+	 * Memoized result of arena_decay_backlog_npages_limit() corresponding
+	 * to the current contents of decay_backlog, i.e. the limit on how many
+	 * pages are allowed to exist for the decay epochs.
+	 */
+	size_t			decay_backlog_npages_limit;
+	/*
+	 * 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 decay_epoch.
+	 */
+	size_t			decay_backlog[SMOOTHSTEP_NSTEPS];
+
 	/* Extant huge allocations. */
 	ql_head(extent_node_t)	huge;
 	/* Synchronizes all huge allocation/update/deallocation. */
@@ -402,6 +455,17 @@ struct arena_s {
 
 	/* bins is used to store trees of free regions. */
 	arena_bin_t		bins[NBINS];
+
+	/*
+	 * Quantized address-ordered trees of this arena's available runs.  The
+	 * trees are used for first-best-fit run allocation.
+	 */
+	arena_run_tree_t	runs_avail[1]; /* Dynamically sized. */
+};
+
+/* Used in conjunction with tsd for fast arena-related context lookup. */
+struct arena_tdata_s {
+	ticker_t		decay_ticker;
 };
 #endif /* JEMALLOC_ARENA_STRUCTS_B */
 
@@ -417,7 +481,10 @@ static const size_t	large_pad =
 #endif
     ;
 
+extern purge_mode_t	opt_purge;
+extern const char	*purge_mode_names[];
 extern ssize_t		opt_lg_dirty_mult;
+extern ssize_t		opt_decay_time;
 
 extern arena_bin_info_t	arena_bin_info[NBINS];
 
@@ -425,9 +492,15 @@ extern size_t		map_bias; /* Number of arena chunk header pages. */
 extern size_t		map_misc_offset;
 extern size_t		arena_maxrun; /* Max run size for arenas. */
 extern size_t		large_maxclass; /* Max large size class. */
+extern size_t		run_quantize_max; /* Max run_quantize_*() input. */
 extern unsigned		nlclasses; /* Number of large size classes. */
 extern unsigned		nhclasses; /* Number of huge size classes. */
 
+#ifdef JEMALLOC_JET
+typedef size_t (run_quantize_t)(size_t);
+extern run_quantize_t *run_quantize_floor;
+extern run_quantize_t *run_quantize_ceil;
+#endif
 void	arena_chunk_cache_maybe_insert(arena_t *arena, extent_node_t *node,
     bool cache);
 void	arena_chunk_cache_maybe_remove(arena_t *arena, extent_node_t *node,
@@ -445,9 +518,11 @@ bool	arena_chunk_ralloc_huge_expand(arena_t *arena, void *chunk,
     size_t oldsize, size_t usize, bool *zero);
 ssize_t	arena_lg_dirty_mult_get(arena_t *arena);
 bool	arena_lg_dirty_mult_set(arena_t *arena, ssize_t lg_dirty_mult);
+ssize_t	arena_decay_time_get(arena_t *arena);
+bool	arena_decay_time_set(arena_t *arena, ssize_t decay_time);
 void	arena_maybe_purge(arena_t *arena);
-void	arena_purge_all(arena_t *arena);
+void	arena_purge(arena_t *arena, bool all);
-void	arena_tcache_fill_small(arena_t *arena, tcache_bin_t *tbin,
+void	arena_tcache_fill_small(tsd_t *tsd, arena_t *arena, tcache_bin_t *tbin,
     szind_t binind, uint64_t prof_accumbytes);
 void	arena_alloc_junk_small(void *ptr, arena_bin_info_t *bin_info,
     bool zero);
@@ -461,8 +536,9 @@ extern arena_dalloc_junk_small_t *arena_dalloc_junk_small;
 void	arena_dalloc_junk_small(void *ptr, arena_bin_info_t *bin_info);
 #endif
 void	arena_quarantine_junk_small(void *ptr, size_t usize);
-void	*arena_malloc_small(arena_t *arena, size_t size, bool zero);
+void	*arena_malloc_large(tsd_t *tsd, arena_t *arena, szind_t ind, bool zero);
-void	*arena_malloc_large(arena_t *arena, size_t size, bool zero);
+void	*arena_malloc_hard(tsd_t *tsd, arena_t *arena, size_t size, szind_t ind,
+    bool zero, tcache_t *tcache);
 void	*arena_palloc(tsd_t *tsd, arena_t *arena, size_t usize,
     size_t alignment, bool zero, tcache_t *tcache);
 void	arena_prof_promoted(const void *ptr, size_t size);
@@ -470,8 +546,8 @@ void	arena_dalloc_bin_junked_locked(arena_t *arena, arena_chunk_t *chunk,
     void *ptr, arena_chunk_map_bits_t *bitselm);
 void	arena_dalloc_bin(arena_t *arena, arena_chunk_t *chunk, void *ptr,
     size_t pageind, arena_chunk_map_bits_t *bitselm);
-void	arena_dalloc_small(arena_t *arena, arena_chunk_t *chunk, void *ptr,
+void	arena_dalloc_small(tsd_t *tsd, arena_t *arena, arena_chunk_t *chunk,
-    size_t pageind);
+    void *ptr, size_t pageind);
 #ifdef JEMALLOC_JET
 typedef void (arena_dalloc_junk_large_t)(void *, size_t);
 extern arena_dalloc_junk_large_t *arena_dalloc_junk_large;
@@ -480,12 +556,13 @@ void	arena_dalloc_junk_large(void *ptr, size_t usize);
 #endif
 void	arena_dalloc_large_junked_locked(arena_t *arena, arena_chunk_t *chunk,
     void *ptr);
-void	arena_dalloc_large(arena_t *arena, arena_chunk_t *chunk, void *ptr);
+void	arena_dalloc_large(tsd_t *tsd, arena_t *arena, arena_chunk_t *chunk,
+    void *ptr);
 #ifdef JEMALLOC_JET
 typedef void (arena_ralloc_junk_large_t)(void *, size_t, size_t);
 extern arena_ralloc_junk_large_t *arena_ralloc_junk_large;
 #endif
-bool	arena_ralloc_no_move(void *ptr, size_t oldsize, size_t size,
+bool	arena_ralloc_no_move(tsd_t *tsd, void *ptr, size_t oldsize, size_t size,
     size_t extra, bool zero);
 void	*arena_ralloc(tsd_t *tsd, arena_t *arena, void *ptr, size_t oldsize,
     size_t size, size_t alignment, bool zero, tcache_t *tcache);
@@ -493,10 +570,18 @@ 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_lg_dirty_mult_default_get(void);
 bool	arena_lg_dirty_mult_default_set(ssize_t lg_dirty_mult);
-void	arena_stats_merge(arena_t *arena, const char **dss,
+ssize_t	arena_decay_time_default_get(void);
-    ssize_t *lg_dirty_mult, size_t *nactive, size_t *ndirty,
+bool	arena_decay_time_default_set(ssize_t decay_time);
-    arena_stats_t *astats, malloc_bin_stats_t *bstats,
+void	arena_basic_stats_merge(arena_t *arena, unsigned *nthreads,
+    const char **dss, ssize_t *lg_dirty_mult, ssize_t *decay_time,
+    size_t *nactive, size_t *ndirty);
+void	arena_stats_merge(arena_t *arena, unsigned *nthreads, const char **dss,
+    ssize_t *lg_dirty_mult, ssize_t *decay_time, size_t *nactive,
+    size_t *ndirty, arena_stats_t *astats, malloc_bin_stats_t *bstats,
     malloc_large_stats_t *lstats, malloc_huge_stats_t *hstats);
+unsigned	arena_nthreads_get(arena_t *arena);
+void	arena_nthreads_inc(arena_t *arena);
+void	arena_nthreads_dec(arena_t *arena);
 arena_t	*arena_new(unsigned ind);
 bool	arena_boot(void);
 void	arena_prefork(arena_t *arena);
@@ -512,7 +597,7 @@ arena_chunk_map_bits_t	*arena_bitselm_get(arena_chunk_t *chunk,
     size_t pageind);
 arena_chunk_map_misc_t	*arena_miscelm_get(arena_chunk_t *chunk,
     size_t pageind);
-size_t	arena_miscelm_to_pageind(arena_chunk_map_misc_t *miscelm);
+size_t	arena_miscelm_to_pageind(const arena_chunk_map_misc_t *miscelm);
 void	*arena_miscelm_to_rpages(arena_chunk_map_misc_t *miscelm);
 arena_chunk_map_misc_t	*arena_rd_to_miscelm(arena_runs_dirty_link_t *rd);
 arena_chunk_map_misc_t	*arena_run_to_miscelm(arena_run_t *run);
@@ -552,17 +637,19 @@ bool	arena_prof_accum_locked(arena_t *arena, uint64_t accumbytes);
 bool	arena_prof_accum(arena_t *arena, uint64_t accumbytes);
 szind_t	arena_ptr_small_binind_get(const void *ptr, size_t mapbits);
 szind_t	arena_bin_index(arena_t *arena, arena_bin_t *bin);
-unsigned	arena_run_regind(arena_run_t *run, arena_bin_info_t *bin_info,
+size_t	arena_run_regind(arena_run_t *run, arena_bin_info_t *bin_info,
     const void *ptr);
 prof_tctx_t	*arena_prof_tctx_get(const void *ptr);
 void	arena_prof_tctx_set(const void *ptr, size_t usize, prof_tctx_t *tctx);
 void	arena_prof_tctx_reset(const void *ptr, size_t usize,
     const void *old_ptr, prof_tctx_t *old_tctx);
-void	*arena_malloc(tsd_t *tsd, arena_t *arena, size_t size, bool zero,
+void	arena_decay_ticks(tsd_t *tsd, arena_t *arena, unsigned nticks);
-    tcache_t *tcache);
+void	arena_decay_tick(tsd_t *tsd, arena_t *arena);
+void	*arena_malloc(tsd_t *tsd, arena_t *arena, size_t size, szind_t ind,
+    bool zero, tcache_t *tcache, bool slow_path);
 arena_t	*arena_aalloc(const void *ptr);
 size_t	arena_salloc(const void *ptr, bool demote);
-void	arena_dalloc(tsd_t *tsd, void *ptr, tcache_t *tcache);
+void	arena_dalloc(tsd_t *tsd, void *ptr, tcache_t *tcache, bool slow_path);
 void	arena_sdalloc(tsd_t *tsd, void *ptr, size_t size, tcache_t *tcache);
 #endif
 
@@ -590,7 +677,7 @@ arena_miscelm_get(arena_chunk_t *chunk, size_t pageind)
 }
 
 JEMALLOC_ALWAYS_INLINE size_t
-arena_miscelm_to_pageind(arena_chunk_map_misc_t *miscelm)
+arena_miscelm_to_pageind(const arena_chunk_map_misc_t *miscelm)
 {
 	arena_chunk_t *chunk = (arena_chunk_t *)CHUNK_ADDR2BASE(miscelm);
 	size_t pageind = ((uintptr_t)miscelm - ((uintptr_t)chunk +
@@ -970,7 +1057,7 @@ arena_ptr_small_binind_get(const void *ptr, size_t mapbits)
 		run = &miscelm->run;
 		run_binind = run->binind;
 		bin = &arena->bins[run_binind];
-		actual_binind = bin - arena->bins;
+		actual_binind = (szind_t)(bin - arena->bins);
 		assert(run_binind == actual_binind);
 		bin_info = &arena_bin_info[actual_binind];
 		rpages = arena_miscelm_to_rpages(miscelm);
@@ -987,16 +1074,15 @@ arena_ptr_small_binind_get(const void *ptr, size_t mapbits)
 JEMALLOC_INLINE szind_t
 arena_bin_index(arena_t *arena, arena_bin_t *bin)
 {
-	szind_t binind = bin - arena->bins;
+	szind_t binind = (szind_t)(bin - arena->bins);
 	assert(binind < NBINS);
 	return (binind);
 }
 
-JEMALLOC_INLINE unsigned
+JEMALLOC_INLINE size_t
 arena_run_regind(arena_run_t *run, arena_bin_info_t *bin_info, const void *ptr)
 {
-	unsigned shift, diff, regind;
+	size_t diff, interval, shift, regind;
-	size_t interval;
 	arena_chunk_map_misc_t *miscelm = arena_run_to_miscelm(run);
 	void *rpages = arena_miscelm_to_rpages(miscelm);
 
@@ -1011,12 +1097,12 @@ arena_run_regind(arena_run_t *run, arena_bin_info_t *bin_info, const void *ptr)
 	 * Avoid doing division with a variable divisor if possible.  Using
 	 * actual division here can reduce allocator throughput by over 20%!
 	 */
-	diff = (unsigned)((uintptr_t)ptr - (uintptr_t)rpages -
+	diff = (size_t)((uintptr_t)ptr - (uintptr_t)rpages -
 	    bin_info->reg0_offset);
 
 	/* Rescale (factor powers of 2 out of the numerator and denominator). */
 	interval = bin_info->reg_interval;
-	shift = jemalloc_ffs(interval) - 1;
+	shift = ffs_zu(interval) - 1;
 	diff >>= shift;
 	interval >>= shift;
 
@@ -1038,9 +1124,9 @@ arena_run_regind(arena_run_t *run, arena_bin_info_t *bin_info, const void *ptr)
 		 * divide by 0, and 1 and 2 are both powers of two, which are
 		 * handled above.
 		 */
-#define	SIZE_INV_SHIFT	((sizeof(unsigned) << 3) - LG_RUN_MAXREGS)
+#define	SIZE_INV_SHIFT	((sizeof(size_t) << 3) - LG_RUN_MAXREGS)
-#define	SIZE_INV(s)	(((1U << SIZE_INV_SHIFT) / (s)) + 1)
+#define	SIZE_INV(s)	(((ZU(1) << SIZE_INV_SHIFT) / (s)) + 1)
-		static const unsigned interval_invs[] = {
+		static const size_t interval_invs[] = {
 		    SIZE_INV(3),
 		    SIZE_INV(4), SIZE_INV(5), SIZE_INV(6), SIZE_INV(7),
 		    SIZE_INV(8), SIZE_INV(9), SIZE_INV(10), SIZE_INV(11),
@@ -1051,8 +1137,8 @@ arena_run_regind(arena_run_t *run, arena_bin_info_t *bin_info, const void *ptr)
 		    SIZE_INV(28), SIZE_INV(29), SIZE_INV(30), SIZE_INV(31)
 		};
 
-		if (likely(interval <= ((sizeof(interval_invs) /
+		if (likely(interval <= ((sizeof(interval_invs) / sizeof(size_t))
-		    sizeof(unsigned)) + 2))) {
+		    + 2))) {
 			regind = (diff * interval_invs[interval - 3]) >>
 			    SIZE_INV_SHIFT;
 		} else
@@ -1157,35 +1243,48 @@ arena_prof_tctx_reset(const void *ptr, size_t usize, const void *old_ptr,
 	}
 }
 
+JEMALLOC_ALWAYS_INLINE void
+arena_decay_ticks(tsd_t *tsd, arena_t *arena, unsigned nticks)
+{
+	ticker_t *decay_ticker;
+
+	if (unlikely(tsd == NULL))
+		return;
+	decay_ticker = decay_ticker_get(tsd, arena->ind);
+	if (unlikely(decay_ticker == NULL))
+		return;
+	if (unlikely(ticker_ticks(decay_ticker, nticks)))
+		arena_purge(arena, false);
+}
+
+JEMALLOC_ALWAYS_INLINE void
+arena_decay_tick(tsd_t *tsd, arena_t *arena)
+{
+
+	arena_decay_ticks(tsd, arena, 1);
+}
+
 JEMALLOC_ALWAYS_INLINE void *
-arena_malloc(tsd_t *tsd, arena_t *arena, size_t size, bool zero,
+arena_malloc(tsd_t *tsd, arena_t *arena, size_t size, szind_t ind, bool zero,
-    tcache_t *tcache)
+    tcache_t *tcache, bool slow_path)
 {
 
 	assert(size != 0);
 
-	arena = arena_choose(tsd, arena);
-	if (unlikely(arena == NULL))
-		return (NULL);
-
-	if (likely(size <= SMALL_MAXCLASS)) {
 	if (likely(tcache != NULL)) {
+		if (likely(size <= SMALL_MAXCLASS)) {
 			return (tcache_alloc_small(tsd, arena, tcache, size,
-			    zero));
+			    ind, zero, slow_path));
-		} else
+		}
-			return (arena_malloc_small(arena, size, zero));
+		if (likely(size <= tcache_maxclass)) {
-	} else if (likely(size <= large_maxclass)) {
-		/*
-		 * Initialize tcache after checking size in order to avoid
-		 * infinite recursion during tcache initialization.
-		 */
-		if (likely(tcache != NULL) && size <= tcache_maxclass) {
 			return (tcache_alloc_large(tsd, arena, tcache, size,
-			    zero));
+			    ind, zero, slow_path));
-		} else
+		}
-			return (arena_malloc_large(arena, size, zero));
+		/* (size > tcache_maxclass) case falls through. */
-	} else
+		assert(size > tcache_maxclass);
-		return (huge_malloc(tsd, arena, size, zero, tcache));
+	}
+
+	return (arena_malloc_hard(tsd, arena, size, ind, zero, tcache));
 }
 
 JEMALLOC_ALWAYS_INLINE arena_t *
@@ -1251,7 +1350,7 @@ arena_salloc(const void *ptr, bool demote)
 }
 
 JEMALLOC_ALWAYS_INLINE void
-arena_dalloc(tsd_t *tsd, void *ptr, tcache_t *tcache)
+arena_dalloc(tsd_t *tsd, void *ptr, tcache_t *tcache, bool slow_path)
 {
 	arena_chunk_t *chunk;
 	size_t pageind, mapbits;
@@ -1268,9 +1367,10 @@ arena_dalloc(tsd_t *tsd, void *ptr, tcache_t *tcache)
 			if (likely(tcache != NULL)) {
 				szind_t binind = arena_ptr_small_binind_get(ptr,
 				    mapbits);
-				tcache_dalloc_small(tsd, tcache, ptr, binind);
+				tcache_dalloc_small(tsd, tcache, ptr, binind,
+				    slow_path);
 			} else {
-				arena_dalloc_small(extent_node_arena_get(
+				arena_dalloc_small(tsd, extent_node_arena_get(
 				    &chunk->node), chunk, ptr, pageind);
 			}
 		} else {
@@ -1283,9 +1383,9 @@ arena_dalloc(tsd_t *tsd, void *ptr, tcache_t *tcache)
 			if (likely(tcache != NULL) && size - large_pad <=
 			    tcache_maxclass) {
 				tcache_dalloc_large(tsd, tcache, ptr, size -
-				    large_pad);
+				    large_pad, slow_path);
 			} else {
-				arena_dalloc_large(extent_node_arena_get(
+				arena_dalloc_large(tsd, extent_node_arena_get(
 				    &chunk->node), chunk, ptr);
 			}
 		}
@@ -1303,7 +1403,8 @@ arena_sdalloc(tsd_t *tsd, void *ptr, size_t size, tcache_t *tcache)
 		if (config_prof && opt_prof) {
 			size_t pageind = ((uintptr_t)ptr - (uintptr_t)chunk) >>
 			    LG_PAGE;
-			assert(arena_mapbits_allocated_get(chunk, pageind) != 0);
+			assert(arena_mapbits_allocated_get(chunk, pageind) !=
+			    0);
 			if (arena_mapbits_large_get(chunk, pageind) != 0) {
 				/*
 				 * Make sure to use promoted size, not request
@@ -1319,21 +1420,23 @@ arena_sdalloc(tsd_t *tsd, void *ptr, size_t size, tcache_t *tcache)
 			/* Small allocation. */
 			if (likely(tcache != NULL)) {
 				szind_t binind = size2index(size);
-				tcache_dalloc_small(tsd, tcache, ptr, binind);
+				tcache_dalloc_small(tsd, tcache, ptr, binind,
+				    true);
 			} else {
 				size_t pageind = ((uintptr_t)ptr -
 				    (uintptr_t)chunk) >> LG_PAGE;
-				arena_dalloc_small(extent_node_arena_get(
+				arena_dalloc_small(tsd, extent_node_arena_get(
 				    &chunk->node), chunk, ptr, pageind);
 			}
 		} else {
 			assert(config_cache_oblivious || ((uintptr_t)ptr &
 			    PAGE_MASK) == 0);
 
-			if (likely(tcache != NULL) && size <= tcache_maxclass)
+			if (likely(tcache != NULL) && size <= tcache_maxclass) {
-				tcache_dalloc_large(tsd, tcache, ptr, size);
+				tcache_dalloc_large(tsd, tcache, ptr, size,
-			else {
+				    true);
-				arena_dalloc_large(extent_node_arena_get(
+			} else {
+				arena_dalloc_large(tsd, extent_node_arena_get(
 				    &chunk->node), chunk, ptr);
 			}
 		}

include/jemalloc/internal/assert.h

@@ -0,0 +1,45 @@
+/*
+ * 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
+
+

include/jemalloc/internal/atomic.h

@@ -28,8 +28,8 @@
  * callers.
  *
  *   <t> atomic_read_<t>(<t> *p) { return (*p); }
- *   <t> atomic_add_<t>(<t> *p, <t> x) { return (*p + x); }
+ *   <t> atomic_add_<t>(<t> *p, <t> x) { return (*p += x); }
- *   <t> atomic_sub_<t>(<t> *p, <t> x) { return (*p - x); }
+ *   <t> atomic_sub_<t>(<t> *p, <t> x) { return (*p -= x); }
  *   bool atomic_cas_<t>(<t> *p, <t> c, <t> s)
  *   {
  *     if (*p != c)

include/jemalloc/internal/bitmap.h

@@ -15,6 +15,15 @@ typedef unsigned long bitmap_t;
 #define	BITMAP_GROUP_NBITS		(ZU(1) << 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 ffsl more than 2^3 times, use a
+ * tree instead.
+ */
+#if LG_BITMAP_MAXBITS - LG_BITMAP_GROUP_NBITS > 3
+#  define 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)
@@ -48,6 +57,8 @@ typedef unsigned long bitmap_t;
 /*
  * Maximum number of groups required to support LG_BITMAP_MAXBITS.
  */
+#ifdef USE_TREE
+
 #if LG_BITMAP_MAXBITS <= LG_BITMAP_GROUP_NBITS
 #  define BITMAP_GROUPS_MAX	BITMAP_GROUPS_1_LEVEL(BITMAP_MAXBITS)
 #elif LG_BITMAP_MAXBITS <= LG_BITMAP_GROUP_NBITS * 2
@@ -65,6 +76,12 @@ typedef unsigned long bitmap_t;
     (LG_BITMAP_MAXBITS / LG_SIZEOF_BITMAP)				\
     + !!(LG_BITMAP_MAXBITS % LG_SIZEOF_BITMAP)
 
+#else /* USE_TREE */
+
+#define	BITMAP_GROUPS_MAX BITMAP_BITS2GROUPS(BITMAP_MAXBITS)
+
+#endif /* USE_TREE */
+
 #endif /* JEMALLOC_H_TYPES */
 /******************************************************************************/
 #ifdef JEMALLOC_H_STRUCTS
@@ -78,6 +95,7 @@ struct bitmap_info_s {
 	/* Logical number of bits in bitmap (stored at bottom level). */
 	size_t nbits;
 
+#ifdef USE_TREE
 	/* Number of levels necessary for nbits. */
 	unsigned nlevels;
 
@@ -86,6 +104,10 @@ struct bitmap_info_s {
 	 * bottom to top (e.g. the bottom level is stored in levels[0]).
 	 */
 	bitmap_level_t levels[BITMAP_MAX_LEVELS+1];
+#else /* USE_TREE */
+	/* Number of groups necessary for nbits. */
+	size_t ngroups;
+#endif /* USE_TREE */
 };
 
 #endif /* JEMALLOC_H_STRUCTS */
@@ -93,9 +115,8 @@ struct bitmap_info_s {
 #ifdef JEMALLOC_H_EXTERNS
 
 void	bitmap_info_init(bitmap_info_t *binfo, size_t nbits);
-size_t	bitmap_info_ngroups(const bitmap_info_t *binfo);
-size_t	bitmap_size(size_t nbits);
 void	bitmap_init(bitmap_t *bitmap, const bitmap_info_t *binfo);
+size_t	bitmap_size(const bitmap_info_t *binfo);
 
 #endif /* JEMALLOC_H_EXTERNS */
 /******************************************************************************/
@@ -113,10 +134,20 @@ void	bitmap_unset(bitmap_t *bitmap, const bitmap_info_t *binfo, size_t bit);
 JEMALLOC_INLINE bool
 bitmap_full(bitmap_t *bitmap, const bitmap_info_t *binfo)
 {
-	unsigned rgoff = binfo->levels[binfo->nlevels].group_offset - 1;
+#ifdef 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
 }
 
 JEMALLOC_INLINE bool
@@ -128,7 +159,7 @@ bitmap_get(bitmap_t *bitmap, const bitmap_info_t *binfo, size_t bit)
 	assert(bit < binfo->nbits);
 	goff = bit >> LG_BITMAP_GROUP_NBITS;
 	g = bitmap[goff];
-	return (!(g & (1LU << (bit & BITMAP_GROUP_NBITS_MASK))));
+	return (!(g & (ZU(1) << (bit & BITMAP_GROUP_NBITS_MASK))));
 }
 
 JEMALLOC_INLINE void
@@ -143,10 +174,11 @@ bitmap_set(bitmap_t *bitmap, const bitmap_info_t *binfo, size_t bit)
 	goff = bit >> LG_BITMAP_GROUP_NBITS;
 	gp = &bitmap[goff];
 	g = *gp;
-	assert(g & (1LU << (bit & BITMAP_GROUP_NBITS_MASK)));
+	assert(g & (ZU(1) << (bit & BITMAP_GROUP_NBITS_MASK)));
-	g ^= 1LU << (bit & BITMAP_GROUP_NBITS_MASK);
+	g ^= ZU(1) << (bit & BITMAP_GROUP_NBITS_MASK);
 	*gp = g;
 	assert(bitmap_get(bitmap, binfo, bit));
+#ifdef USE_TREE
 	/* Propagate group state transitions up the tree. */
 	if (g == 0) {
 		unsigned i;
@@ -155,13 +187,14 @@ bitmap_set(bitmap_t *bitmap, const bitmap_info_t *binfo, size_t bit)
 			goff = bit >> LG_BITMAP_GROUP_NBITS;
 			gp = &bitmap[binfo->levels[i].group_offset + goff];
 			g = *gp;
-			assert(g & (1LU << (bit & BITMAP_GROUP_NBITS_MASK)));
+			assert(g & (ZU(1) << (bit & BITMAP_GROUP_NBITS_MASK)));
-			g ^= 1LU << (bit & BITMAP_GROUP_NBITS_MASK);
+			g ^= ZU(1) << (bit & BITMAP_GROUP_NBITS_MASK);
 			*gp = g;
 			if (g != 0)
 				break;
 		}
 	}
+#endif
 }
 
 /* sfu: set first unset. */
@@ -174,15 +207,24 @@ bitmap_sfu(bitmap_t *bitmap, const bitmap_info_t *binfo)
 
 	assert(!bitmap_full(bitmap, binfo));
 
+#ifdef USE_TREE
 	i = binfo->nlevels - 1;
 	g = bitmap[binfo->levels[i].group_offset];
-	bit = jemalloc_ffsl(g) - 1;
+	bit = ffs_lu(g) - 1;
 	while (i > 0) {
 		i--;
 		g = bitmap[binfo->levels[i].group_offset + bit];
-		bit = (bit << LG_BITMAP_GROUP_NBITS) + (jemalloc_ffsl(g) - 1);
+		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 = (bit - 1) + (i << 6);
+#endif
 	bitmap_set(bitmap, binfo, bit);
 	return (bit);
 }
@@ -193,7 +235,7 @@ bitmap_unset(bitmap_t *bitmap, const bitmap_info_t *binfo, size_t bit)
 	size_t goff;
 	bitmap_t *gp;
 	bitmap_t g;
-	bool propagate;
+	UNUSED bool propagate;
 
 	assert(bit < binfo->nbits);
 	assert(bitmap_get(bitmap, binfo, bit));
@@ -201,10 +243,11 @@ bitmap_unset(bitmap_t *bitmap, const bitmap_info_t *binfo, size_t bit)
 	gp = &bitmap[goff];
 	g = *gp;
 	propagate = (g == 0);
-	assert((g & (1LU << (bit & BITMAP_GROUP_NBITS_MASK))) == 0);
+	assert((g & (ZU(1) << (bit & BITMAP_GROUP_NBITS_MASK))) == 0);
-	g ^= 1LU << (bit & BITMAP_GROUP_NBITS_MASK);
+	g ^= ZU(1) << (bit & BITMAP_GROUP_NBITS_MASK);
 	*gp = g;
 	assert(!bitmap_get(bitmap, binfo, bit));
+#ifdef USE_TREE
 	/* Propagate group state transitions up the tree. */
 	if (propagate) {
 		unsigned i;
@@ -214,14 +257,15 @@ bitmap_unset(bitmap_t *bitmap, const bitmap_info_t *binfo, size_t bit)
 			gp = &bitmap[binfo->levels[i].group_offset + goff];
 			g = *gp;
 			propagate = (g == 0);
-			assert((g & (1LU << (bit & BITMAP_GROUP_NBITS_MASK)))
+			assert((g & (ZU(1) << (bit & BITMAP_GROUP_NBITS_MASK)))
 			    == 0);
-			g ^= 1LU << (bit & BITMAP_GROUP_NBITS_MASK);
+			g ^= ZU(1) << (bit & BITMAP_GROUP_NBITS_MASK);
 			*gp = g;
 			if (!propagate)
 				break;
 		}
 	}
+#endif /* USE_TREE */
 }
 
 #endif

include/jemalloc/internal/chunk_mmap.h

@@ -9,8 +9,8 @@
 /******************************************************************************/
 #ifdef JEMALLOC_H_EXTERNS
 
-void	*chunk_alloc_mmap(size_t size, size_t alignment, bool *zero,
+void	*chunk_alloc_mmap(void *new_addr, size_t size, size_t alignment,
-    bool *commit);
+    bool *zero, bool *commit);
 bool	chunk_dalloc_mmap(void *chunk, size_t size);
 
 #endif /* JEMALLOC_H_EXTERNS */

include/jemalloc/internal/ckh.h

@@ -40,9 +40,7 @@ struct ckh_s {
 #endif
 
 	/* Used for pseudo-random number generation. */
-#define	CKH_A		1103515241
+	uint64_t	prng_state;
-#define	CKH_C		12347
-	uint32_t	prng_state;
 
 	/* Total number of items. */
 	size_t		count;
@@ -74,7 +72,7 @@ bool	ckh_iter(ckh_t *ckh, size_t *tabind, void **key, void **data);
 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 *seachkey, void **key, void **data);
+bool	ckh_search(ckh_t *ckh, const void *searchkey, void **key, void **data);
 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]);

include/jemalloc/internal/ctl.h

@@ -35,8 +35,12 @@ struct ctl_arena_stats_s {
 	unsigned		nthreads;
 	const char		*dss;
 	ssize_t			lg_dirty_mult;
+	ssize_t			decay_time;
 	size_t			pactive;
 	size_t			pdirty;
+
+	/* The remainder are only populated if config_stats is true. */
+
 	arena_stats_t		astats;
 
 	/* Aggregate stats for small size classes, based on bin stats. */

include/jemalloc/internal/hash.h

@@ -1,6 +1,6 @@
 /*
  * The following hash function is based on MurmurHash3, placed into the public
- * domain by Austin Appleby.  See http://code.google.com/p/smhasher/ for
+ * domain by Austin Appleby.  See https://github.com/aappleby/smhasher for
  * details.
  */
 /******************************************************************************/
@@ -49,6 +49,14 @@ JEMALLOC_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, &p[i], sizeof(uint32_t));
+		return (ret);
+	}
+
 	return (p[i]);
 }
 
@@ -56,6 +64,14 @@ JEMALLOC_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, &p[i], sizeof(uint64_t));
+		return (ret);
+	}
+
 	return (p[i]);
 }
 
@@ -321,13 +337,18 @@ hash_x64_128(const void *key, const int len, const uint32_t seed,
 JEMALLOC_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, len, seed, (uint64_t *)r_hash);
+	hash_x64_128(key, (int)len, seed, (uint64_t *)r_hash);
 #else
+	{
 		uint64_t hashes[2];
-	hash_x86_128(key, len, seed, hashes);
+		hash_x86_128(key, (int)len, seed, hashes);
 		r_hash[0] = (size_t)hashes[0];
 		r_hash[1] = (size_t)hashes[1];
+	}
 #endif
 }
 #endif

include/jemalloc/internal/huge.h

@@ -9,12 +9,12 @@
 /******************************************************************************/
 #ifdef JEMALLOC_H_EXTERNS
 
-void	*huge_malloc(tsd_t *tsd, arena_t *arena, size_t size, bool zero,
+void	*huge_malloc(tsd_t *tsd, arena_t *arena, size_t usize, bool zero,
     tcache_t *tcache);
-void	*huge_palloc(tsd_t *tsd, arena_t *arena, size_t size, size_t alignment,
+void	*huge_palloc(tsd_t *tsd, arena_t *arena, size_t usize, size_t alignment,
     bool zero, tcache_t *tcache);
-bool	huge_ralloc_no_move(void *ptr, size_t oldsize, size_t usize_min,
+bool	huge_ralloc_no_move(tsd_t *tsd, void *ptr, size_t oldsize,
-    size_t usize_max, bool zero);
+    size_t usize_min, size_t usize_max, bool zero);
 void	*huge_ralloc(tsd_t *tsd, arena_t *arena, void *ptr, size_t oldsize,
     size_t usize, size_t alignment, bool zero, tcache_t *tcache);
 #ifdef JEMALLOC_JET

include/jemalloc/internal/jemalloc_internal.h.in

@@ -49,6 +49,7 @@ static const bool config_lazy_lock =
     false
 #endif
     ;
+static const char * const config_malloc_conf = JEMALLOC_CONFIG_MALLOC_CONF;
 static const bool config_prof =
 #ifdef JEMALLOC_PROF
     true
@@ -355,12 +356,15 @@ typedef unsigned szind_t;
 #  define VARIABLE_ARRAY(type, name, count) type name[(count)]
 #endif
 
+#include "jemalloc/internal/nstime.h"
 #include "jemalloc/internal/valgrind.h"
 #include "jemalloc/internal/util.h"
 #include "jemalloc/internal/atomic.h"
 #include "jemalloc/internal/prng.h"
+#include "jemalloc/internal/ticker.h"
 #include "jemalloc/internal/ckh.h"
 #include "jemalloc/internal/size_classes.h"
+#include "jemalloc/internal/smoothstep.h"
 #include "jemalloc/internal/stats.h"
 #include "jemalloc/internal/ctl.h"
 #include "jemalloc/internal/mutex.h"
@@ -383,12 +387,15 @@ typedef unsigned szind_t;
 /******************************************************************************/
 #define	JEMALLOC_H_STRUCTS
 
+#include "jemalloc/internal/nstime.h"
 #include "jemalloc/internal/valgrind.h"
 #include "jemalloc/internal/util.h"
 #include "jemalloc/internal/atomic.h"
 #include "jemalloc/internal/prng.h"
+#include "jemalloc/internal/ticker.h"
 #include "jemalloc/internal/ckh.h"
 #include "jemalloc/internal/size_classes.h"
+#include "jemalloc/internal/smoothstep.h"
 #include "jemalloc/internal/stats.h"
 #include "jemalloc/internal/ctl.h"
 #include "jemalloc/internal/mutex.h"
@@ -426,18 +433,24 @@ extern bool	opt_redzone;
 extern bool	opt_utrace;
 extern bool	opt_xmalloc;
 extern bool	opt_zero;
-extern size_t	opt_narenas;
+extern unsigned	opt_narenas;
 
 extern bool	in_valgrind;
 
 /* Number of CPUs. */
 extern unsigned	ncpus;
 
+/*
+ * 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 arena_t	**arenas;
+
 /*
  * index2size_tab encodes the same information as could be computed (at
  * unacceptable cost in some code paths) by index2size_compute().
  */
-extern size_t const	index2size_tab[NSIZES];
+extern size_t const	index2size_tab[NSIZES+1];
 /*
  * 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,
@@ -445,35 +458,36 @@ extern size_t const	index2size_tab[NSIZES];
  */
 extern uint8_t const	size2index_tab[];
 
-arena_t	*a0get(void);
 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);
 arena_t	*arenas_extend(unsigned ind);
-arena_t	*arena_init(unsigned ind);
 unsigned	narenas_total_get(void);
-arena_t	*arena_get_hard(tsd_t *tsd, unsigned ind, bool init_if_missing);
+arena_t	*arena_init(unsigned ind);
+arena_tdata_t	*arena_tdata_get_hard(tsd_t *tsd, unsigned ind);
 arena_t	*arena_choose_hard(tsd_t *tsd);
 void	arena_migrate(tsd_t *tsd, unsigned oldind, unsigned newind);
-unsigned	arena_nbound(unsigned ind);
 void	thread_allocated_cleanup(tsd_t *tsd);
 void	thread_deallocated_cleanup(tsd_t *tsd);
 void	arena_cleanup(tsd_t *tsd);
-void	arenas_cache_cleanup(tsd_t *tsd);
+void	arenas_tdata_cleanup(tsd_t *tsd);
-void	narenas_cache_cleanup(tsd_t *tsd);
+void	narenas_tdata_cleanup(tsd_t *tsd);
-void	arenas_cache_bypass_cleanup(tsd_t *tsd);
+void	arenas_tdata_bypass_cleanup(tsd_t *tsd);
 void	jemalloc_prefork(void);
 void	jemalloc_postfork_parent(void);
 void	jemalloc_postfork_child(void);
 
+#include "jemalloc/internal/nstime.h"
 #include "jemalloc/internal/valgrind.h"
 #include "jemalloc/internal/util.h"
 #include "jemalloc/internal/atomic.h"
 #include "jemalloc/internal/prng.h"
+#include "jemalloc/internal/ticker.h"
 #include "jemalloc/internal/ckh.h"
 #include "jemalloc/internal/size_classes.h"
+#include "jemalloc/internal/smoothstep.h"
 #include "jemalloc/internal/stats.h"
 #include "jemalloc/internal/ctl.h"
 #include "jemalloc/internal/mutex.h"
@@ -496,12 +510,15 @@ void	jemalloc_postfork_child(void);
 /******************************************************************************/
 #define	JEMALLOC_H_INLINES
 
+#include "jemalloc/internal/nstime.h"
 #include "jemalloc/internal/valgrind.h"
 #include "jemalloc/internal/util.h"
 #include "jemalloc/internal/atomic.h"
 #include "jemalloc/internal/prng.h"
+#include "jemalloc/internal/ticker.h"
 #include "jemalloc/internal/ckh.h"
 #include "jemalloc/internal/size_classes.h"
+#include "jemalloc/internal/smoothstep.h"
 #include "jemalloc/internal/stats.h"
 #include "jemalloc/internal/ctl.h"
 #include "jemalloc/internal/mutex.h"
@@ -526,8 +543,10 @@ size_t	s2u_lookup(size_t size);
 size_t	s2u(size_t size);
 size_t	sa2u(size_t size, size_t alignment);
 arena_t	*arena_choose(tsd_t *tsd, arena_t *arena);
-arena_t	*arena_get(tsd_t *tsd, unsigned ind, bool init_if_missing,
+arena_tdata_t	*arena_tdata_get(tsd_t *tsd, unsigned ind,
     bool refresh_if_missing);
+arena_t	*arena_get(unsigned ind, bool init_if_missing);
+ticker_t	*decay_ticker_get(tsd_t *tsd, unsigned ind);
 #endif
 
 #if (defined(JEMALLOC_ENABLE_INLINE) || defined(JEMALLOC_C_))
@@ -537,27 +556,27 @@ size2index_compute(size_t size)
 
 #if (NTBINS != 0)
 	if (size <= (ZU(1) << LG_TINY_MAXCLASS)) {
-		size_t lg_tmin = LG_TINY_MAXCLASS - NTBINS + 1;
+		szind_t lg_tmin = LG_TINY_MAXCLASS - NTBINS + 1;
-		size_t lg_ceil = lg_floor(pow2_ceil(size));
+		szind_t lg_ceil = lg_floor(pow2_ceil_zu(size));
 		return (lg_ceil < lg_tmin ? 0 : lg_ceil - lg_tmin);
 	}
 #endif
 	{
-		size_t x = unlikely(ZI(size) < 0) ? ((size<<1) ?
+		szind_t x = unlikely(ZI(size) < 0) ? ((size<<1) ?
 		    (ZU(1)<<(LG_SIZEOF_PTR+3)) : ((ZU(1)<<(LG_SIZEOF_PTR+3))-1))
 		    : lg_floor((size<<1)-1);
-		size_t shift = (x < LG_SIZE_CLASS_GROUP + LG_QUANTUM) ? 0 :
+		szind_t shift = (x < LG_SIZE_CLASS_GROUP + LG_QUANTUM) ? 0 :
 		    x - (LG_SIZE_CLASS_GROUP + LG_QUANTUM);
-		size_t grp = shift << LG_SIZE_CLASS_GROUP;
+		szind_t grp = shift << LG_SIZE_CLASS_GROUP;
 
-		size_t lg_delta = (x < LG_SIZE_CLASS_GROUP + LG_QUANTUM + 1)
+		szind_t lg_delta = (x < LG_SIZE_CLASS_GROUP + LG_QUANTUM + 1)
 		    ? LG_QUANTUM : x - LG_SIZE_CLASS_GROUP - 1;
 
 		size_t delta_inverse_mask = ZI(-1) << lg_delta;
-		size_t mod = ((((size-1) & delta_inverse_mask) >> lg_delta)) &
+		szind_t mod = ((((size-1) & delta_inverse_mask) >> lg_delta)) &
 		    ((ZU(1) << LG_SIZE_CLASS_GROUP) - 1);
 
-		size_t index = NTBINS + grp + mod;
+		szind_t index = NTBINS + grp + mod;
 		return (index);
 	}
 }
@@ -568,8 +587,7 @@ size2index_lookup(size_t size)
 
 	assert(size <= LOOKUP_MAXCLASS);
 	{
-		size_t ret = ((size_t)(size2index_tab[(size-1) >>
+		szind_t ret = (size2index_tab[(size-1) >> LG_TINY_MIN]);
-		    LG_TINY_MIN]));
 		assert(ret == size2index_compute(size));
 		return (ret);
 	}
@@ -635,7 +653,7 @@ s2u_compute(size_t size)
 #if (NTBINS > 0)
 	if (size <= (ZU(1) << LG_TINY_MAXCLASS)) {
 		size_t lg_tmin = LG_TINY_MAXCLASS - NTBINS + 1;
-		size_t lg_ceil = lg_floor(pow2_ceil(size));
+		size_t lg_ceil = lg_floor(pow2_ceil_zu(size));
 		return (lg_ceil < lg_tmin ? (ZU(1) << lg_tmin) :
 		    (ZU(1) << lg_ceil));
 	}
@@ -727,17 +745,16 @@ sa2u(size_t size, size_t alignment)
 			return (usize);
 	}
 
-	/* Huge size class.  Beware of size_t overflow. */
+	/* Huge size class.  Beware of overflow. */
+
+	if (unlikely(alignment > HUGE_MAXCLASS))
+		return (0);
 
 	/*
 	 * We can't achieve subchunk alignment, so round up alignment to the
 	 * minimum that can actually be supported.
 	 */
 	alignment = CHUNK_CEILING(alignment);
-	if (alignment == 0) {
-		/* size_t overflow. */
-		return (0);
-	}
 
 	/* Make sure result is a huge size class. */
 	if (size <= chunksize)
@@ -776,32 +793,56 @@ arena_choose(tsd_t *tsd, arena_t *arena)
 	return (ret);
 }
 
-JEMALLOC_INLINE arena_t *
+JEMALLOC_INLINE arena_tdata_t *
-arena_get(tsd_t *tsd, unsigned ind, bool init_if_missing,
+arena_tdata_get(tsd_t *tsd, unsigned ind, bool refresh_if_missing)
-    bool refresh_if_missing)
 {
-	arena_t *arena;
+	arena_tdata_t *tdata;
-	arena_t **arenas_cache = tsd_arenas_cache_get(tsd);
+	arena_tdata_t *arenas_tdata = tsd_arenas_tdata_get(tsd);
 
-	/* init_if_missing requires refresh_if_missing. */
+	if (unlikely(arenas_tdata == NULL)) {
-	assert(!init_if_missing || refresh_if_missing);
+		/* arenas_tdata hasn't been initialized yet. */
-
+		return (arena_tdata_get_hard(tsd, ind));
-	if (unlikely(arenas_cache == NULL)) {
-		/* arenas_cache hasn't been initialized yet. */
-		return (arena_get_hard(tsd, ind, init_if_missing));
 	}
-	if (unlikely(ind >= tsd_narenas_cache_get(tsd))) {
+	if (unlikely(ind >= tsd_narenas_tdata_get(tsd))) {
 		/*
-		 * ind is invalid, cache is old (too small), or arena to be
+		 * ind is invalid, cache is old (too small), or tdata to be
 		 * initialized.
 		 */
-		return (refresh_if_missing ? arena_get_hard(tsd, ind,
+		return (refresh_if_missing ? arena_tdata_get_hard(tsd, ind) :
-		    init_if_missing) : NULL);
+		    NULL);
 	}
-	arena = arenas_cache[ind];
+
-	if (likely(arena != NULL) || !refresh_if_missing)
+	tdata = &arenas_tdata[ind];
-		return (arena);
+	if (likely(tdata != NULL) || !refresh_if_missing)
-	return (arena_get_hard(tsd, ind, init_if_missing));
+		return (tdata);
+	return (arena_tdata_get_hard(tsd, ind));
+}
+
+JEMALLOC_INLINE arena_t *
+arena_get(unsigned ind, bool init_if_missing)
+{
+	arena_t *ret;
+
+	assert(ind <= MALLOCX_ARENA_MAX);
+
+	ret = arenas[ind];
+	if (unlikely(ret == NULL)) {
+		ret = atomic_read_p((void *)&arenas[ind]);
+		if (init_if_missing && unlikely(ret == NULL))
+			ret = arena_init(ind);
+	}
+	return (ret);
+}
+
+JEMALLOC_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);
 }
 #endif
 
@@ -823,12 +864,14 @@ arena_get(tsd_t *tsd, unsigned ind, bool init_if_missing,
 #ifndef JEMALLOC_ENABLE_INLINE
 arena_t	*iaalloc(const void *ptr);
 size_t	isalloc(const void *ptr, bool demote);
-void	*iallocztm(tsd_t *tsd, size_t size, bool zero, tcache_t *tcache,
+void	*iallocztm(tsd_t *tsd, size_t size, szind_t ind, bool zero,
-    bool is_metadata, arena_t *arena);
+    tcache_t *tcache, bool is_metadata, arena_t *arena, bool slow_path);
-void	*imalloct(tsd_t *tsd, size_t size, tcache_t *tcache, arena_t *arena);
+void	*imalloct(tsd_t *tsd, size_t size, szind_t ind, tcache_t *tcache,
-void	*imalloc(tsd_t *tsd, size_t size);
+    arena_t *arena);
-void	*icalloct(tsd_t *tsd, size_t size, tcache_t *tcache, arena_t *arena);
+void	*imalloc(tsd_t *tsd, size_t size, szind_t ind, bool slow_path);
-void	*icalloc(tsd_t *tsd, size_t size);
+void	*icalloct(tsd_t *tsd, size_t size, szind_t ind, tcache_t *tcache,
+    arena_t *arena);
+void	*icalloc(tsd_t *tsd, size_t size, szind_t ind);
 void	*ipallocztm(tsd_t *tsd, size_t usize, size_t alignment, bool zero,
     tcache_t *tcache, bool is_metadata, arena_t *arena);
 void	*ipalloct(tsd_t *tsd, size_t usize, size_t alignment, bool zero,
@@ -837,10 +880,11 @@ void	*ipalloc(tsd_t *tsd, size_t usize, size_t alignment, bool zero);
 size_t	ivsalloc(const void *ptr, bool demote);
 size_t	u2rz(size_t usize);
 size_t	p2rz(const void *ptr);
-void	idalloctm(tsd_t *tsd, void *ptr, tcache_t *tcache, bool is_metadata);
+void	idalloctm(tsd_t *tsd, void *ptr, tcache_t *tcache, bool is_metadata,
+    bool slow_path);
 void	idalloct(tsd_t *tsd, void *ptr, tcache_t *tcache);
 void	idalloc(tsd_t *tsd, void *ptr);
-void	iqalloc(tsd_t *tsd, void *ptr, tcache_t *tcache);
+void	iqalloc(tsd_t *tsd, void *ptr, tcache_t *tcache, bool slow_path);
 void	isdalloct(tsd_t *tsd, void *ptr, size_t size, tcache_t *tcache);
 void	isqalloc(tsd_t *tsd, void *ptr, size_t size, tcache_t *tcache);
 void	*iralloct_realign(tsd_t *tsd, void *ptr, size_t oldsize, size_t size,
@@ -850,8 +894,8 @@ void	*iralloct(tsd_t *tsd, void *ptr, size_t oldsize, size_t size,
     size_t alignment, bool zero, tcache_t *tcache, arena_t *arena);
 void	*iralloc(tsd_t *tsd, void *ptr, size_t oldsize, size_t size,
     size_t alignment, bool zero);
-bool	ixalloc(void *ptr, size_t oldsize, size_t size, size_t extra,
+bool	ixalloc(tsd_t *tsd, void *ptr, size_t oldsize, size_t size,
-    size_t alignment, bool zero);
+    size_t extra, size_t alignment, bool zero);
 #endif
 
 #if (defined(JEMALLOC_ENABLE_INLINE) || defined(JEMALLOC_C_))
@@ -881,14 +925,14 @@ isalloc(const void *ptr, bool demote)
 }
 
 JEMALLOC_ALWAYS_INLINE void *
-iallocztm(tsd_t *tsd, size_t size, bool zero, tcache_t *tcache, bool is_metadata,
+iallocztm(tsd_t *tsd, size_t size, szind_t ind, bool zero, tcache_t *tcache,
-    arena_t *arena)
+    bool is_metadata, arena_t *arena, bool slow_path)
 {
 	void *ret;
 
 	assert(size != 0);
 
-	ret = arena_malloc(tsd, arena, size, zero, tcache);
+	ret = arena_malloc(tsd, arena, size, ind, zero, tcache, slow_path);
 	if (config_stats && is_metadata && likely(ret != NULL)) {
 		arena_metadata_allocated_add(iaalloc(ret), isalloc(ret,
 		    config_prof));
@@ -897,31 +941,33 @@ iallocztm(tsd_t *tsd, size_t size, bool zero, tcache_t *tcache, bool is_metadata
 }
 
 JEMALLOC_ALWAYS_INLINE void *
-imalloct(tsd_t *tsd, size_t size, tcache_t *tcache, arena_t *arena)
+imalloct(tsd_t *tsd, size_t size, szind_t ind, tcache_t *tcache, arena_t *arena)
 {
 
-	return (iallocztm(tsd, size, false, tcache, false, arena));
+	return (iallocztm(tsd, size, ind, false, tcache, false, arena, true));
 }
 
 JEMALLOC_ALWAYS_INLINE void *
-imalloc(tsd_t *tsd, size_t size)
+imalloc(tsd_t *tsd, size_t size, szind_t ind, bool slow_path)
 {
 
-	return (iallocztm(tsd, size, false, tcache_get(tsd, true), false, NULL));
+	return (iallocztm(tsd, size, ind, false, tcache_get(tsd, true), false,
+	    NULL, slow_path));
 }
 
 JEMALLOC_ALWAYS_INLINE void *
-icalloct(tsd_t *tsd, size_t size, tcache_t *tcache, arena_t *arena)
+icalloct(tsd_t *tsd, size_t size, szind_t ind, tcache_t *tcache, arena_t *arena)
 {
 
-	return (iallocztm(tsd, size, true, tcache, false, arena));
+	return (iallocztm(tsd, size, ind, true, tcache, false, arena, true));
 }
 
 JEMALLOC_ALWAYS_INLINE void *
-icalloc(tsd_t *tsd, size_t size)
+icalloc(tsd_t *tsd, size_t size, szind_t ind)
 {
 
-	return (iallocztm(tsd, size, true, tcache_get(tsd, true), false, NULL));
+	return (iallocztm(tsd, size, ind, true, tcache_get(tsd, true), false,
+	    NULL, true));
 }
 
 JEMALLOC_ALWAYS_INLINE void *
@@ -954,8 +1000,8 @@ JEMALLOC_ALWAYS_INLINE void *
 ipalloc(tsd_t *tsd, size_t usize, size_t alignment, bool zero)
 {
 
-	return (ipallocztm(tsd, usize, alignment, zero, tcache_get(tsd,
+	return (ipallocztm(tsd, usize, alignment, zero, tcache_get(tsd, true),
-	    NULL), false, NULL));
+	    false, NULL));
 }
 
 JEMALLOC_ALWAYS_INLINE size_t
@@ -997,7 +1043,8 @@ p2rz(const void *ptr)
 }
 
 JEMALLOC_ALWAYS_INLINE void
-idalloctm(tsd_t *tsd, void *ptr, tcache_t *tcache, bool is_metadata)
+idalloctm(tsd_t *tsd, void *ptr, tcache_t *tcache, bool is_metadata,
+    bool slow_path)
 {
 
 	assert(ptr != NULL);
@@ -1006,31 +1053,31 @@ idalloctm(tsd_t *tsd, void *ptr, tcache_t *tcache, bool is_metadata)
 		    config_prof));
 	}
 
-	arena_dalloc(tsd, ptr, tcache);
+	arena_dalloc(tsd, ptr, tcache, slow_path);
 }
 
 JEMALLOC_ALWAYS_INLINE void
 idalloct(tsd_t *tsd, void *ptr, tcache_t *tcache)
 {
 
-	idalloctm(tsd, ptr, tcache, false);
+	idalloctm(tsd, ptr, tcache, false, true);
 }
 
 JEMALLOC_ALWAYS_INLINE void
 idalloc(tsd_t *tsd, void *ptr)
 {
 
-	idalloctm(tsd, ptr, tcache_get(tsd, false), false);
+	idalloctm(tsd, ptr, tcache_get(tsd, false), false, true);
 }
 
 JEMALLOC_ALWAYS_INLINE void
-iqalloc(tsd_t *tsd, void *ptr, tcache_t *tcache)
+iqalloc(tsd_t *tsd, void *ptr, tcache_t *tcache, bool slow_path)
 {
 
-	if (config_fill && unlikely(opt_quarantine))
+	if (slow_path && config_fill && unlikely(opt_quarantine))
 		quarantine(tsd, ptr);
 	else
-		idalloctm(tsd, ptr, tcache, false);
+		idalloctm(tsd, ptr, tcache, false, slow_path);
 }
 
 JEMALLOC_ALWAYS_INLINE void
@@ -1058,7 +1105,7 @@ iralloct_realign(tsd_t *tsd, void *ptr, size_t oldsize, size_t size,
 	size_t usize, copysize;
 
 	usize = sa2u(size + extra, alignment);
-	if (usize == 0)
+	if (unlikely(usize == 0 || usize > HUGE_MAXCLASS))
 		return (NULL);
 	p = ipalloct(tsd, usize, alignment, zero, tcache, arena);
 	if (p == NULL) {
@@ -1066,7 +1113,7 @@ iralloct_realign(tsd_t *tsd, void *ptr, size_t oldsize, size_t size,
 			return (NULL);
 		/* Try again, without extra this time. */
 		usize = sa2u(size, alignment);
-		if (usize == 0)
+		if (unlikely(usize == 0 || usize > HUGE_MAXCLASS))
 			return (NULL);
 		p = ipalloct(tsd, usize, alignment, zero, tcache, arena);
 		if (p == NULL)
@@ -1114,8 +1161,8 @@ iralloc(tsd_t *tsd, void *ptr, size_t oldsize, size_t size, size_t alignment,
 }
 
 JEMALLOC_ALWAYS_INLINE bool
-ixalloc(void *ptr, size_t oldsize, size_t size, size_t extra, size_t alignment,
+ixalloc(tsd_t *tsd, void *ptr, size_t oldsize, size_t size, size_t extra,
-    bool zero)
+    size_t alignment, bool zero)
 {
 
 	assert(ptr != NULL);
@@ -1127,7 +1174,7 @@ ixalloc(void *ptr, size_t oldsize, size_t size, size_t extra, size_t alignment,
 		return (true);
 	}
 
-	return (arena_ralloc_no_move(ptr, oldsize, size, extra, zero));
+	return (arena_ralloc_no_move(tsd, ptr, oldsize, size, extra, zero));
 }
 #endif
 

include/jemalloc/internal/jemalloc_internal_decls.h

@@ -18,6 +18,7 @@
 #  endif
 #  include <pthread.h>
 #  include <errno.h>
+#  include <sys/time.h>
 #endif
 #include <sys/types.h>
 

include/jemalloc/internal/jemalloc_internal_defs.h.in

@@ -189,9 +189,10 @@
 #undef JEMALLOC_TLS
 
 /*
- * ffs()/ffsl() functions to use for bitmapping.  Don't use these directly;
+ * ffs*() functions to use for bitmapping.  Don't use these directly; instead,
- * instead, use jemalloc_ffs() or jemalloc_ffsl() from util.h.
+ * use ffs_*() from util.h.
  */
+#undef JEMALLOC_INTERNAL_FFSLL
 #undef JEMALLOC_INTERNAL_FFSL
 #undef JEMALLOC_INTERNAL_FFS
 
@@ -241,6 +242,9 @@
 /* sizeof(long) == 2^LG_SIZEOF_LONG. */
 #undef LG_SIZEOF_LONG
 
+/* sizeof(long long) == 2^LG_SIZEOF_LONG_LONG. */
+#undef LG_SIZEOF_LONG_LONG
+
 /* sizeof(intmax_t) == 2^LG_SIZEOF_INTMAX_T. */
 #undef LG_SIZEOF_INTMAX_T
 
@@ -259,4 +263,7 @@
  */
 #undef JEMALLOC_EXPORT
 
+/* config.malloc_conf options string. */
+#undef JEMALLOC_CONFIG_MALLOC_CONF
+
 #endif /* JEMALLOC_INTERNAL_DEFS_H_ */

include/jemalloc/internal/nstime.h

@@ -0,0 +1,48 @@
+/******************************************************************************/
+#ifdef JEMALLOC_H_TYPES
+
+#define JEMALLOC_CLOCK_GETTIME defined(_POSIX_MONOTONIC_CLOCK) \
+    && _POSIX_MONOTONIC_CLOCK >= 0
+
+typedef struct nstime_s nstime_t;
+
+/* Maximum supported number of seconds (~584 years). */
+#define	NSTIME_SEC_MAX	18446744072
+
+#endif /* JEMALLOC_H_TYPES */
+/******************************************************************************/
+#ifdef JEMALLOC_H_STRUCTS
+
+struct nstime_s {
+	uint64_t	ns;
+};
+
+#endif /* JEMALLOC_H_STRUCTS */
+/******************************************************************************/
+#ifdef JEMALLOC_H_EXTERNS
+
+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_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_subtract(nstime_t *time, const nstime_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);
+#ifdef JEMALLOC_JET
+typedef bool (nstime_update_t)(nstime_t *);
+extern nstime_update_t *nstime_update;
+#else
+bool	nstime_update(nstime_t *time);
+#endif
+
+#endif /* JEMALLOC_H_EXTERNS */
+/******************************************************************************/
+#ifdef JEMALLOC_H_INLINES
+
+#endif /* JEMALLOC_H_INLINES */
+/******************************************************************************/

include/jemalloc/internal/private_symbols.txt

@@ -1,8 +1,8 @@
 a0dalloc
-a0get
 a0malloc
 arena_aalloc
 arena_alloc_junk_small
+arena_basic_stats_merge
 arena_bin_index
 arena_bin_info
 arena_bitselm_get
@@ -25,18 +25,23 @@ arena_dalloc_junk_small
 arena_dalloc_large
 arena_dalloc_large_junked_locked
 arena_dalloc_small
+arena_decay_tick
+arena_decay_ticks
+arena_decay_time_default_get
+arena_decay_time_default_set
+arena_decay_time_get
+arena_decay_time_set
 arena_dss_prec_get
 arena_dss_prec_set
 arena_get
-arena_get_hard
 arena_init
 arena_lg_dirty_mult_default_get
 arena_lg_dirty_mult_default_set
 arena_lg_dirty_mult_get
 arena_lg_dirty_mult_set
 arena_malloc
+arena_malloc_hard
 arena_malloc_large
-arena_malloc_small
 arena_mapbits_allocated_get
 arena_mapbits_binind_get
 arena_mapbits_decommitted_get
@@ -47,9 +52,6 @@ arena_mapbits_large_binind_set
 arena_mapbits_large_get
 arena_mapbits_large_set
 arena_mapbits_large_size_get
-arena_mapbitsp_get
-arena_mapbitsp_read
-arena_mapbitsp_write
 arena_mapbits_size_decode
 arena_mapbits_size_encode
 arena_mapbits_small_runind_get
@@ -58,6 +60,9 @@ arena_mapbits_unallocated_set
 arena_mapbits_unallocated_size_get
 arena_mapbits_unallocated_size_set
 arena_mapbits_unzeroed_get
+arena_mapbitsp_get
+arena_mapbitsp_read
+arena_mapbitsp_write
 arena_maxrun
 arena_maybe_purge
 arena_metadata_allocated_add
@@ -67,10 +72,12 @@ arena_migrate
 arena_miscelm_get
 arena_miscelm_to_pageind
 arena_miscelm_to_rpages
-arena_nbound
 arena_new
 arena_node_alloc
 arena_node_dalloc
+arena_nthreads_dec
+arena_nthreads_get
+arena_nthreads_inc
 arena_palloc
 arena_postfork_child
 arena_postfork_parent
@@ -83,7 +90,7 @@ arena_prof_tctx_get
 arena_prof_tctx_reset
 arena_prof_tctx_set
 arena_ptr_small_binind_get
-arena_purge_all
+arena_purge
 arena_quarantine_junk_small
 arena_ralloc
 arena_ralloc_junk_large
@@ -93,11 +100,14 @@ arena_redzone_corruption
 arena_run_regind
 arena_run_to_miscelm
 arena_salloc
-arenas_cache_bypass_cleanup
-arenas_cache_cleanup
 arena_sdalloc
 arena_stats_merge
 arena_tcache_fill_small
+arena_tdata_get
+arena_tdata_get_hard
+arenas
+arenas_tdata_bypass_cleanup
+arenas_tdata_cleanup
 atomic_add_p
 atomic_add_u
 atomic_add_uint32
@@ -122,7 +132,6 @@ base_stats_get
 bitmap_full
 bitmap_get
 bitmap_info_init
-bitmap_info_ngroups
 bitmap_init
 bitmap_set
 bitmap_sfu
@@ -162,9 +171,9 @@ chunk_prefork
 chunk_purge_arena
 chunk_purge_wrapper
 chunk_register
+chunks_rtree
 chunksize
 chunksize_mask
-chunks_rtree
 ckh_count
 ckh_delete
 ckh_insert
@@ -183,6 +192,7 @@ ctl_nametomib
 ctl_postfork_child
 ctl_postfork_parent
 ctl_prefork
+decay_ticker_get
 dss_prec_names
 extent_node_achunk_get
 extent_node_achunk_set
@@ -234,6 +244,12 @@ extent_tree_szad_reverse_iter
 extent_tree_szad_reverse_iter_recurse
 extent_tree_szad_reverse_iter_start
 extent_tree_szad_search
+ffs_llu
+ffs_lu
+ffs_u
+ffs_u32
+ffs_u64
+ffs_zu
 get_errno
 hash
 hash_fmix_32
@@ -265,11 +281,11 @@ idalloct
 idalloctm
 imalloc
 imalloct
+in_valgrind
 index2size
 index2size_compute
 index2size_lookup
 index2size_tab
-in_valgrind
 ipalloc
 ipalloct
 ipallocztm
@@ -311,11 +327,25 @@ map_bias
 map_misc_offset
 mb_write
 mutex_boot
-narenas_cache_cleanup
+narenas_tdata_cleanup
 narenas_total_get
 ncpus
 nhbins
+nstime_add
+nstime_compare
+nstime_copy
+nstime_divide
+nstime_idivide
+nstime_imultiply
+nstime_init
+nstime_init2
+nstime_ns
+nstime_nsec
+nstime_sec
+nstime_subtract
+nstime_update
 opt_abort
+opt_decay_time
 opt_dss
 opt_junk
 opt_junk_alloc
@@ -334,6 +364,7 @@ opt_prof_gdump
 opt_prof_leak
 opt_prof_prefix
 opt_prof_thread_active_init
+opt_purge
 opt_quarantine
 opt_redzone
 opt_stats_print
@@ -348,7 +379,11 @@ pages_map
 pages_purge
 pages_trim
 pages_unmap
-pow2_ceil
+pow2_ceil_u32
+pow2_ceil_u64
+pow2_ceil_zu
+prng_lg_range
+prng_range
 prof_active_get
 prof_active_get_unlocked
 prof_active_set
@@ -393,6 +428,7 @@ prof_thread_active_init_set
 prof_thread_active_set
 prof_thread_name_get
 prof_thread_name_set
+purge_mode_names
 quarantine
 quarantine_alloc_hook
 quarantine_alloc_hook_work
@@ -413,6 +449,9 @@ rtree_subtree_read_hard
 rtree_subtree_tryread
 rtree_val_read
 rtree_val_write
+run_quantize_ceil
+run_quantize_floor
+run_quantize_max
 s2u
 s2u_compute
 s2u_lookup
@@ -451,15 +490,20 @@ tcache_flush
 tcache_get
 tcache_get_hard
 tcache_maxclass
-tcaches
 tcache_salloc
+tcache_stats_merge
+tcaches
 tcaches_create
 tcaches_destroy
 tcaches_flush
 tcaches_get
-tcache_stats_merge
 thread_allocated_cleanup
 thread_deallocated_cleanup
+ticker_copy
+ticker_init
+ticker_read
+ticker_tick
+ticker_ticks
 tsd_arena_get
 tsd_arena_set
 tsd_boot
@@ -477,6 +521,8 @@ tsd_init_check_recursion
 tsd_init_finish
 tsd_init_head
 tsd_nominal
+tsd_prof_tdata_get
+tsd_prof_tdata_set
 tsd_quarantine_get
 tsd_quarantine_set
 tsd_set
@@ -484,14 +530,12 @@ tsd_tcache_enabled_get
 tsd_tcache_enabled_set
 tsd_tcache_get
 tsd_tcache_set
-tsd_tls
-tsd_tsd
-tsd_prof_tdata_get
-tsd_prof_tdata_set
 tsd_thread_allocated_get
 tsd_thread_allocated_set
 tsd_thread_deallocated_get
 tsd_thread_deallocated_set
+tsd_tls
+tsd_tsd
 u2rz
 valgrind_freelike_block
 valgrind_make_mem_defined

include/jemalloc/internal/prng.h

@@ -18,31 +18,9 @@
  * 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.
- *
- * Macro parameters:
- *   uint32_t r          : Result.
- *   unsigned lg_range   : (0..32], number of least significant bits to return.
- *   uint32_t state      : Seed value.
- *   const uint32_t a, c : See above discussion.
  */
-#define	prng32(r, lg_range, state, a, c) do {				\
+#define	PRNG_A	UINT64_C(6364136223846793005)
-	assert((lg_range) > 0);						\
+#define	PRNG_C	UINT64_C(1442695040888963407)
-	assert((lg_range) <= 32);					\
-									\
-	r = (state * (a)) + (c);					\
-	state = r;							\
-	r >>= (32 - (lg_range));					\
-} while (false)
-
-/* Same as prng32(), but 64 bits of pseudo-randomness, using uint64_t. */
-#define	prng64(r, lg_range, state, a, c) do {				\
-	assert((lg_range) > 0);						\
-	assert((lg_range) <= 64);					\
-									\
-	r = (state * (a)) + (c);					\
-	state = r;							\
-	r >>= (64 - (lg_range));					\
-} while (false)
 
 #endif /* JEMALLOC_H_TYPES */
 /******************************************************************************/
@@ -56,5 +34,46 @@
 /******************************************************************************/
 #ifdef JEMALLOC_H_INLINES
 
+#ifndef JEMALLOC_ENABLE_INLINE
+uint64_t	prng_lg_range(uint64_t *state, unsigned lg_range);
+uint64_t	prng_range(uint64_t *state, uint64_t range);
+#endif
+
+#if (defined(JEMALLOC_ENABLE_INLINE) || defined(JEMALLOC_PRNG_C_))
+JEMALLOC_ALWAYS_INLINE uint64_t
+prng_lg_range(uint64_t *state, unsigned lg_range)
+{
+	uint64_t ret;
+
+	assert(lg_range > 0);
+	assert(lg_range <= 64);
+
+	ret = (*state * PRNG_A) + PRNG_C;
+	*state = ret;
+	ret >>= (64 - lg_range);
+
+	return (ret);
+}
+
+JEMALLOC_ALWAYS_INLINE uint64_t
+prng_range(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(state, lg_range);
+	} while (ret >= range);
+
+	return (ret);
+}
+#endif
+
 #endif /* JEMALLOC_H_INLINES */
 /******************************************************************************/

include/jemalloc/internal/prof.h

@@ -436,16 +436,16 @@ prof_sample_accum_update(tsd_t *tsd, size_t usize, bool update,
 	cassert(config_prof);
 
 	tdata = prof_tdata_get(tsd, true);
-	if ((uintptr_t)tdata <= (uintptr_t)PROF_TDATA_STATE_MAX)
+	if (unlikely((uintptr_t)tdata <= (uintptr_t)PROF_TDATA_STATE_MAX))
 		tdata = NULL;
 
 	if (tdata_out != NULL)
 		*tdata_out = tdata;
 
-	if (tdata == NULL)
+	if (unlikely(tdata == NULL))
 		return (true);
 
-	if (tdata->bytes_until_sample >= usize) {
+	if (likely(tdata->bytes_until_sample >= usize)) {
 		if (update)
 			tdata->bytes_until_sample -= usize;
 		return (true);

include/jemalloc/internal/rb.h

@@ -42,7 +42,6 @@ struct {								\
 #define	rb_tree(a_type)							\
 struct {								\
     a_type *rbt_root;							\
-    a_type rbt_nil;							\
 }
 
 /* Left accessors. */
@@ -79,6 +78,15 @@ struct {								\
     (a_node)->a_field.rbn_right_red = (a_type *) (((intptr_t)		\
       (a_node)->a_field.rbn_right_red) & ((ssize_t)-2));		\
 } while (0)
+
+/* Node initializer. */
+#define	rbt_node_new(a_type, a_field, a_rbt, a_node) do {		\
+    /* Bookkeeping bit cannot be used by node pointer. */		\
+    assert(((uintptr_t)(a_node) & 0x1) == 0);				\
+    rbtn_left_set(a_type, a_field, (a_node), NULL);	\
+    rbtn_right_set(a_type, a_field, (a_node), NULL);	\
+    rbtn_red_set(a_type, a_field, (a_node));				\
+} while (0)
 #else
 /* Right accessors. */
 #define	rbtn_right_get(a_type, a_field, a_node)				\
@@ -99,28 +107,26 @@ struct {								\
 #define	rbtn_black_set(a_type, a_field, a_node) do {			\
     (a_node)->a_field.rbn_red = false;					\
 } while (0)
-#endif
 
 /* Node initializer. */
 #define	rbt_node_new(a_type, a_field, a_rbt, a_node) do {		\
-    rbtn_left_set(a_type, a_field, (a_node), &(a_rbt)->rbt_nil);	\
+    rbtn_left_set(a_type, a_field, (a_node), NULL);	\
-    rbtn_right_set(a_type, a_field, (a_node), &(a_rbt)->rbt_nil);	\
+    rbtn_right_set(a_type, a_field, (a_node), NULL);	\
     rbtn_red_set(a_type, a_field, (a_node));				\
 } while (0)
+#endif
 
 /* Tree initializer. */
 #define	rb_new(a_type, a_field, a_rbt) do {				\
-    (a_rbt)->rbt_root = &(a_rbt)->rbt_nil;				\
+    (a_rbt)->rbt_root = NULL;						\
-    rbt_node_new(a_type, a_field, a_rbt, &(a_rbt)->rbt_nil);		\
-    rbtn_black_set(a_type, a_field, &(a_rbt)->rbt_nil);			\
 } while (0)
 
 /* Internal utility macros. */
 #define	rbtn_first(a_type, a_field, a_rbt, a_root, r_node) do {		\
     (r_node) = (a_root);						\
-    if ((r_node) != &(a_rbt)->rbt_nil) {				\
+    if ((r_node) != NULL) {						\
 	for (;								\
-	  rbtn_left_get(a_type, a_field, (r_node)) != &(a_rbt)->rbt_nil;\
+	  rbtn_left_get(a_type, a_field, (r_node)) != NULL;		\
 	  (r_node) = rbtn_left_get(a_type, a_field, (r_node))) {	\
 	}								\
     }									\
@@ -128,10 +134,9 @@ struct {								\
 
 #define	rbtn_last(a_type, a_field, a_rbt, a_root, r_node) do {		\
     (r_node) = (a_root);						\
-    if ((r_node) != &(a_rbt)->rbt_nil) {				\
+    if ((r_node) != NULL) {						\
-	for (; rbtn_right_get(a_type, a_field, (r_node)) !=		\
+	for (; rbtn_right_get(a_type, a_field, (r_node)) != NULL;	\
-	  &(a_rbt)->rbt_nil; (r_node) = rbtn_right_get(a_type, a_field,	\
+	  (r_node) = rbtn_right_get(a_type, a_field, (r_node))) {	\
-	  (r_node))) {							\
 	}								\
     }									\
 } while (0)
@@ -169,11 +174,11 @@ a_prefix##next(a_rbt_type *rbtree, a_type *node);			\
 a_attr a_type *								\
 a_prefix##prev(a_rbt_type *rbtree, a_type *node);			\
 a_attr a_type *								\
-a_prefix##search(a_rbt_type *rbtree, a_type *key);			\
+a_prefix##search(a_rbt_type *rbtree, const a_type *key);		\
 a_attr a_type *								\
-a_prefix##nsearch(a_rbt_type *rbtree, a_type *key);			\
+a_prefix##nsearch(a_rbt_type *rbtree, const a_type *key);		\
 a_attr a_type *								\
-a_prefix##psearch(a_rbt_type *rbtree, a_type *key);			\
+a_prefix##psearch(a_rbt_type *rbtree, const a_type *key);		\
 a_attr void								\
 a_prefix##insert(a_rbt_type *rbtree, a_type *node);			\
 a_attr void								\
@@ -183,7 +188,10 @@ a_prefix##iter(a_rbt_type *rbtree, a_type *start, a_type *(*cb)(	\
   a_rbt_type *, a_type *, void *), void *arg);				\
 a_attr a_type *								\
 a_prefix##reverse_iter(a_rbt_type *rbtree, a_type *start,		\
-  a_type *(*cb)(a_rbt_type *, a_type *, void *), void *arg);
+  a_type *(*cb)(a_rbt_type *, a_type *, void *), void *arg);		\
+a_attr void								\
+a_prefix##destroy(a_rbt_type *rbtree, void (*cb)(a_type *, void *),	\
+  void *arg);
 
 /*
  * The rb_gen() macro generates a type-specific red-black tree implementation,
@@ -254,7 +262,7 @@ a_prefix##reverse_iter(a_rbt_type *rbtree, a_type *start,		\
  *            last/first.
  *
  *   static ex_node_t *
- *   ex_search(ex_t *tree, ex_node_t *key);
+ *   ex_search(ex_t *tree, const ex_node_t *key);
  *       Description: Search for node that matches key.
  *       Args:
  *         tree: Pointer to an initialized red-black tree object.
@@ -262,9 +270,9 @@ a_prefix##reverse_iter(a_rbt_type *rbtree, a_type *start,		\
  *       Ret: Node in tree that matches key, or NULL if no match.
  *
  *   static ex_node_t *
- *   ex_nsearch(ex_t *tree, ex_node_t *key);
+ *   ex_nsearch(ex_t *tree, const ex_node_t *key);
  *   static ex_node_t *
- *   ex_psearch(ex_t *tree, ex_node_t *key);
+ *   ex_psearch(ex_t *tree, const ex_node_t *key);
  *       Description: Search for node that matches key.  If no match is found,
  *                    return what would be key's successor/predecessor, were
  *                    key in tree.
@@ -312,6 +320,20 @@ a_prefix##reverse_iter(a_rbt_type *rbtree, a_type *start,		\
  *         arg  : Opaque pointer passed to cb().
  *       Ret: NULL if iteration completed, or the non-NULL callback return value
  *            that caused termination of the iteration.
+ *
+ *   static void
+ *   ex_destroy(ex_t *tree, void (*cb)(ex_node_t *, void *), void *arg);
+ *       Description: Iterate over the tree with post-order traversal, remove
+ *                    each node, and run the callback if non-null.  This is
+ *                    used for destroying a tree without paying the cost to
+ *                    rebalance it.  The tree must not be otherwise altered
+ *                    during traversal.
+ *       Args:
+ *         tree: Pointer to an initialized red-black tree object.
+ *         cb  : Callback function, which, if non-null, is called for each node
+ *               during iteration.  There is no way to stop iteration once it
+ *               has begun.
+ *         arg : Opaque pointer passed to cb().
  */
 #define	rb_gen(a_attr, a_prefix, a_rbt_type, a_type, a_field, a_cmp)	\
 a_attr void								\
@@ -320,36 +342,30 @@ a_prefix##new(a_rbt_type *rbtree) {					\
 }									\
 a_attr bool								\
 a_prefix##empty(a_rbt_type *rbtree) {					\
-    return (rbtree->rbt_root == &rbtree->rbt_nil);			\
+    return (rbtree->rbt_root == NULL);					\
 }									\
 a_attr a_type *								\
 a_prefix##first(a_rbt_type *rbtree) {					\
     a_type *ret;							\
     rbtn_first(a_type, a_field, rbtree, rbtree->rbt_root, ret);		\
-    if (ret == &rbtree->rbt_nil) {					\
-	ret = NULL;							\
-    }									\
     return (ret);							\
 }									\
 a_attr a_type *								\
 a_prefix##last(a_rbt_type *rbtree) {					\
     a_type *ret;							\
     rbtn_last(a_type, a_field, rbtree, rbtree->rbt_root, ret);		\
-    if (ret == &rbtree->rbt_nil) {					\
-	ret = NULL;							\
-    }									\
     return (ret);							\
 }									\
 a_attr a_type *								\
 a_prefix##next(a_rbt_type *rbtree, a_type *node) {			\
     a_type *ret;							\
-    if (rbtn_right_get(a_type, a_field, node) != &rbtree->rbt_nil) {	\
+    if (rbtn_right_get(a_type, a_field, node) != NULL) {		\
 	rbtn_first(a_type, a_field, rbtree, rbtn_right_get(a_type,	\
 	  a_field, node), ret);						\
     } else {								\
 	a_type *tnode = rbtree->rbt_root;				\
-	assert(tnode != &rbtree->rbt_nil);				\
+	assert(tnode != NULL);						\
-	ret = &rbtree->rbt_nil;						\
+	ret = NULL;							\
 	while (true) {							\
 	    int cmp = (a_cmp)(node, tnode);				\
 	    if (cmp < 0) {						\
@@ -360,24 +376,21 @@ a_prefix##next(a_rbt_type *rbtree, a_type *node) {			\
 	    } else {							\
 		break;							\
 	    }								\
-	    assert(tnode != &rbtree->rbt_nil);				\
+	    assert(tnode != NULL);					\
 	}								\
     }									\
-    if (ret == &rbtree->rbt_nil) {					\
-	ret = (NULL);							\
-    }									\
     return (ret);							\
 }									\
 a_attr a_type *								\
 a_prefix##prev(a_rbt_type *rbtree, a_type *node) {			\
     a_type *ret;							\
-    if (rbtn_left_get(a_type, a_field, node) != &rbtree->rbt_nil) {	\
+    if (rbtn_left_get(a_type, a_field, node) != NULL) {			\
 	rbtn_last(a_type, a_field, rbtree, rbtn_left_get(a_type,	\
 	  a_field, node), ret);						\
     } else {								\
 	a_type *tnode = rbtree->rbt_root;				\
-	assert(tnode != &rbtree->rbt_nil);				\
+	assert(tnode != NULL);						\
-	ret = &rbtree->rbt_nil;						\
+	ret = NULL;							\
 	while (true) {							\
 	    int cmp = (a_cmp)(node, tnode);				\
 	    if (cmp < 0) {						\
@@ -388,20 +401,17 @@ a_prefix##prev(a_rbt_type *rbtree, a_type *node) {			\
 	    } else {							\
 		break;							\
 	    }								\
-	    assert(tnode != &rbtree->rbt_nil);				\
+	    assert(tnode != NULL);					\
 	}								\
     }									\
-    if (ret == &rbtree->rbt_nil) {					\
-	ret = (NULL);							\
-    }									\
     return (ret);							\
 }									\
 a_attr a_type *								\
-a_prefix##search(a_rbt_type *rbtree, a_type *key) {			\
+a_prefix##search(a_rbt_type *rbtree, const a_type *key) {		\
     a_type *ret;							\
     int cmp;								\
     ret = rbtree->rbt_root;						\
-    while (ret != &rbtree->rbt_nil					\
+    while (ret != NULL							\
       && (cmp = (a_cmp)(key, ret)) != 0) {				\
 	if (cmp < 0) {							\
 	    ret = rbtn_left_get(a_type, a_field, ret);			\
@@ -409,17 +419,14 @@ a_prefix##search(a_rbt_type *rbtree, a_type *key) {			\
 	    ret = rbtn_right_get(a_type, a_field, ret);			\
 	}								\
     }									\
-    if (ret == &rbtree->rbt_nil) {					\
-	ret = (NULL);							\
-    }									\
     return (ret);							\
 }									\
 a_attr a_type *								\
-a_prefix##nsearch(a_rbt_type *rbtree, a_type *key) {			\
+a_prefix##nsearch(a_rbt_type *rbtree, const a_type *key) {		\
     a_type *ret;							\
     a_type *tnode = rbtree->rbt_root;					\
-    ret = &rbtree->rbt_nil;						\
+    ret = NULL;								\
-    while (tnode != &rbtree->rbt_nil) {					\
+    while (tnode != NULL) {						\
 	int cmp = (a_cmp)(key, tnode);					\
 	if (cmp < 0) {							\
 	    ret = tnode;						\
@@ -431,17 +438,14 @@ a_prefix##nsearch(a_rbt_type *rbtree, a_type *key) {			\
 	    break;							\
 	}								\
     }									\
-    if (ret == &rbtree->rbt_nil) {					\
-	ret = (NULL);							\
-    }									\
     return (ret);							\
 }									\
 a_attr a_type *								\
-a_prefix##psearch(a_rbt_type *rbtree, a_type *key) {			\
+a_prefix##psearch(a_rbt_type *rbtree, const a_type *key) {		\
     a_type *ret;							\
     a_type *tnode = rbtree->rbt_root;					\
-    ret = &rbtree->rbt_nil;						\
+    ret = NULL;								\
-    while (tnode != &rbtree->rbt_nil) {					\
+    while (tnode != NULL) {						\
 	int cmp = (a_cmp)(key, tnode);					\
 	if (cmp < 0) {							\
 	    tnode = rbtn_left_get(a_type, a_field, tnode);		\
@@ -453,9 +457,6 @@ a_prefix##psearch(a_rbt_type *rbtree, a_type *key) {			\
 	    break;							\
 	}								\
     }									\
-    if (ret == &rbtree->rbt_nil) {					\
-	ret = (NULL);							\
-    }									\
     return (ret);							\
 }									\
 a_attr void								\
@@ -467,7 +468,7 @@ a_prefix##insert(a_rbt_type *rbtree, a_type *node) {			\
     rbt_node_new(a_type, a_field, rbtree, node);			\
     /* Wind. */								\
     path->node = rbtree->rbt_root;					\
-    for (pathp = path; pathp->node != &rbtree->rbt_nil; pathp++) {	\
+    for (pathp = path; pathp->node != NULL; pathp++) {			\
 	int cmp = pathp->cmp = a_cmp(node, pathp->node);		\
 	assert(cmp != 0);						\
 	if (cmp < 0) {							\
@@ -487,7 +488,8 @@ a_prefix##insert(a_rbt_type *rbtree, a_type *node) {			\
 	    rbtn_left_set(a_type, a_field, cnode, left);		\
 	    if (rbtn_red_get(a_type, a_field, left)) {			\
 		a_type *leftleft = rbtn_left_get(a_type, a_field, left);\
-		if (rbtn_red_get(a_type, a_field, leftleft)) {		\
+		if (leftleft != NULL && rbtn_red_get(a_type, a_field,	\
+		  leftleft)) {						\
 		    /* Fix up 4-node. */				\
 		    a_type *tnode;					\
 		    rbtn_black_set(a_type, a_field, leftleft);		\
@@ -502,7 +504,8 @@ a_prefix##insert(a_rbt_type *rbtree, a_type *node) {			\
 	    rbtn_right_set(a_type, a_field, cnode, right);		\
 	    if (rbtn_red_get(a_type, a_field, right)) {			\
 		a_type *left = rbtn_left_get(a_type, a_field, cnode);	\
-		if (rbtn_red_get(a_type, a_field, left)) {		\
+		if (left != NULL && rbtn_red_get(a_type, a_field,	\
+		  left)) {						\
 		    /* Split 4-node. */					\
 		    rbtn_black_set(a_type, a_field, left);		\
 		    rbtn_black_set(a_type, a_field, right);		\
@@ -535,7 +538,7 @@ a_prefix##remove(a_rbt_type *rbtree, a_type *node) {			\
     /* Wind. */								\
     nodep = NULL; /* Silence compiler warning. */			\
     path->node = rbtree->rbt_root;					\
-    for (pathp = path; pathp->node != &rbtree->rbt_nil; pathp++) {	\
+    for (pathp = path; pathp->node != NULL; pathp++) {			\
 	int cmp = pathp->cmp = a_cmp(node, pathp->node);		\
 	if (cmp < 0) {							\
 	    pathp[1].node = rbtn_left_get(a_type, a_field,		\
@@ -547,7 +550,7 @@ a_prefix##remove(a_rbt_type *rbtree, a_type *node) {			\
 	        /* Find node's successor, in preparation for swap. */	\
 		pathp->cmp = 1;						\
 		nodep = pathp;						\
-		for (pathp++; pathp->node != &rbtree->rbt_nil;		\
+		for (pathp++; pathp->node != NULL;			\
 		  pathp++) {						\
 		    pathp->cmp = -1;					\
 		    pathp[1].node = rbtn_left_get(a_type, a_field,	\
@@ -590,7 +593,7 @@ a_prefix##remove(a_rbt_type *rbtree, a_type *node) {			\
 	}								\
     } else {								\
 	a_type *left = rbtn_left_get(a_type, a_field, node);		\
-	if (left != &rbtree->rbt_nil) {					\
+	if (left != NULL) {						\
 	    /* node has no successor, but it has a left child.        */\
 	    /* Splice node out, without losing the left child.        */\
 	    assert(!rbtn_red_get(a_type, a_field, node));		\
@@ -610,33 +613,32 @@ a_prefix##remove(a_rbt_type *rbtree, a_type *node) {			\
 	    return;							\
 	} else if (pathp == path) {					\
 	    /* The tree only contained one node. */			\
-	    rbtree->rbt_root = &rbtree->rbt_nil;			\
+	    rbtree->rbt_root = NULL;					\
 	    return;							\
 	}								\
     }									\
     if (rbtn_red_get(a_type, a_field, pathp->node)) {			\
 	/* Prune red node, which requires no fixup. */			\
 	assert(pathp[-1].cmp < 0);					\
-	rbtn_left_set(a_type, a_field, pathp[-1].node,			\
+	rbtn_left_set(a_type, a_field, pathp[-1].node, NULL);		\
-	  &rbtree->rbt_nil);						\
 	return;								\
     }									\
     /* The node to be pruned is black, so unwind until balance is     */\
     /* restored.                                                      */\
-    pathp->node = &rbtree->rbt_nil;					\
+    pathp->node = NULL;							\
     for (pathp--; (uintptr_t)pathp >= (uintptr_t)path; pathp--) {	\
 	assert(pathp->cmp != 0);					\
 	if (pathp->cmp < 0) {						\
 	    rbtn_left_set(a_type, a_field, pathp->node,			\
 	      pathp[1].node);						\
-	    assert(!rbtn_red_get(a_type, a_field, pathp[1].node));	\
 	    if (rbtn_red_get(a_type, a_field, pathp->node)) {		\
 		a_type *right = rbtn_right_get(a_type, a_field,		\
 		  pathp->node);						\
 		a_type *rightleft = rbtn_left_get(a_type, a_field,	\
 		  right);						\
 		a_type *tnode;						\
-		if (rbtn_red_get(a_type, a_field, rightleft)) {		\
+		if (rightleft != NULL && rbtn_red_get(a_type, a_field,	\
+		  rightleft)) {						\
 		    /* In the following diagrams, ||, //, and \\      */\
 		    /* indicate the path to the removed node.         */\
 		    /*                                                */\
@@ -679,7 +681,8 @@ a_prefix##remove(a_rbt_type *rbtree, a_type *node) {			\
 		  pathp->node);						\
 		a_type *rightleft = rbtn_left_get(a_type, a_field,	\
 		  right);						\
-		if (rbtn_red_get(a_type, a_field, rightleft)) {		\
+		if (rightleft != NULL && rbtn_red_get(a_type, a_field,	\
+		  rightleft)) {						\
 		    /*      ||                                        */\
 		    /*    pathp(b)                                    */\
 		    /*  //        \                                   */\
@@ -733,7 +736,8 @@ a_prefix##remove(a_rbt_type *rbtree, a_type *node) {			\
 		  left);						\
 		a_type *leftrightleft = rbtn_left_get(a_type, a_field,	\
 		  leftright);						\
-		if (rbtn_red_get(a_type, a_field, leftrightleft)) {	\
+		if (leftrightleft != NULL && rbtn_red_get(a_type,	\
+		  a_field, leftrightleft)) {				\
 		    /*      ||                                        */\
 		    /*    pathp(b)                                    */\
 		    /*   /        \\                                  */\
@@ -759,7 +763,7 @@ a_prefix##remove(a_rbt_type *rbtree, a_type *node) {			\
 		    /*   (b)                                          */\
 		    /*   /                                            */\
 		    /* (b)                                            */\
-		    assert(leftright != &rbtree->rbt_nil);		\
+		    assert(leftright != NULL);				\
 		    rbtn_red_set(a_type, a_field, leftright);		\
 		    rbtn_rotate_right(a_type, a_field, pathp->node,	\
 		      tnode);						\
@@ -782,7 +786,8 @@ a_prefix##remove(a_rbt_type *rbtree, a_type *node) {			\
 		return;							\
 	    } else if (rbtn_red_get(a_type, a_field, pathp->node)) {	\
 		a_type *leftleft = rbtn_left_get(a_type, a_field, left);\
-		if (rbtn_red_get(a_type, a_field, leftleft)) {		\
+		if (leftleft != NULL && rbtn_red_get(a_type, a_field,	\
+		  leftleft)) {						\
 		    /*        ||                                      */\
 		    /*      pathp(r)                                  */\
 		    /*     /        \\                                */\
@@ -820,7 +825,8 @@ a_prefix##remove(a_rbt_type *rbtree, a_type *node) {			\
 		}							\
 	    } else {							\
 		a_type *leftleft = rbtn_left_get(a_type, a_field, left);\
-		if (rbtn_red_get(a_type, a_field, leftleft)) {		\
+		if (leftleft != NULL && rbtn_red_get(a_type, a_field,	\
+		  leftleft)) {						\
 		    /*               ||                               */\
 		    /*             pathp(b)                           */\
 		    /*            /        \\                         */\
@@ -866,13 +872,13 @@ a_prefix##remove(a_rbt_type *rbtree, a_type *node) {			\
 a_attr a_type *								\
 a_prefix##iter_recurse(a_rbt_type *rbtree, a_type *node,		\
   a_type *(*cb)(a_rbt_type *, a_type *, void *), void *arg) {		\
-    if (node == &rbtree->rbt_nil) {					\
+    if (node == NULL) {							\
-	return (&rbtree->rbt_nil);					\
+	return (NULL);							\
     } else {								\
 	a_type *ret;							\
 	if ((ret = a_prefix##iter_recurse(rbtree, rbtn_left_get(a_type,	\
-	  a_field, node), cb, arg)) != &rbtree->rbt_nil			\
+	  a_field, node), cb, arg)) != NULL || (ret = cb(rbtree, node,	\
-	  || (ret = cb(rbtree, node, arg)) != NULL) {			\
+	  arg)) != NULL) {						\
 	    return (ret);						\
 	}								\
 	return (a_prefix##iter_recurse(rbtree, rbtn_right_get(a_type,	\
@@ -886,8 +892,8 @@ a_prefix##iter_start(a_rbt_type *rbtree, a_type *start, a_type *node,	\
     if (cmp < 0) {							\
 	a_type *ret;							\
 	if ((ret = a_prefix##iter_start(rbtree, start,			\
-	  rbtn_left_get(a_type, a_field, node), cb, arg)) !=		\
+	  rbtn_left_get(a_type, a_field, node), cb, arg)) != NULL ||	\
-	  &rbtree->rbt_nil || (ret = cb(rbtree, node, arg)) != NULL) {	\
+	  (ret = cb(rbtree, node, arg)) != NULL) {			\
 	    return (ret);						\
 	}								\
 	return (a_prefix##iter_recurse(rbtree, rbtn_right_get(a_type,	\
@@ -914,21 +920,18 @@ a_prefix##iter(a_rbt_type *rbtree, a_type *start, a_type *(*cb)(	\
     } else {								\
 	ret = a_prefix##iter_recurse(rbtree, rbtree->rbt_root, cb, arg);\
     }									\
-    if (ret == &rbtree->rbt_nil) {					\
-	ret = NULL;							\
-    }									\
     return (ret);							\
 }									\
 a_attr a_type *								\
 a_prefix##reverse_iter_recurse(a_rbt_type *rbtree, a_type *node,	\
   a_type *(*cb)(a_rbt_type *, a_type *, void *), void *arg) {		\
-    if (node == &rbtree->rbt_nil) {					\
+    if (node == NULL) {							\
-	return (&rbtree->rbt_nil);					\
+	return (NULL);							\
     } else {								\
 	a_type *ret;							\
 	if ((ret = a_prefix##reverse_iter_recurse(rbtree,		\
-	  rbtn_right_get(a_type, a_field, node), cb, arg)) !=		\
+	  rbtn_right_get(a_type, a_field, node), cb, arg)) != NULL ||	\
-	  &rbtree->rbt_nil || (ret = cb(rbtree, node, arg)) != NULL) {	\
+	  (ret = cb(rbtree, node, arg)) != NULL) {			\
 	    return (ret);						\
 	}								\
 	return (a_prefix##reverse_iter_recurse(rbtree,			\
@@ -943,8 +946,8 @@ a_prefix##reverse_iter_start(a_rbt_type *rbtree, a_type *start,		\
     if (cmp > 0) {							\
 	a_type *ret;							\
 	if ((ret = a_prefix##reverse_iter_start(rbtree, start,		\
-	  rbtn_right_get(a_type, a_field, node), cb, arg)) !=		\
+	  rbtn_right_get(a_type, a_field, node), cb, arg)) != NULL ||	\
-	  &rbtree->rbt_nil || (ret = cb(rbtree, node, arg)) != NULL) {	\
+	  (ret = cb(rbtree, node, arg)) != NULL) {			\
 	    return (ret);						\
 	}								\
 	return (a_prefix##reverse_iter_recurse(rbtree,			\
@@ -972,10 +975,29 @@ a_prefix##reverse_iter(a_rbt_type *rbtree, a_type *start,		\
 	ret = a_prefix##reverse_iter_recurse(rbtree, rbtree->rbt_root,	\
 	  cb, arg);							\
     }									\
-    if (ret == &rbtree->rbt_nil) {					\
-	ret = NULL;							\
-    }									\
     return (ret);							\
+}									\
+a_attr void								\
+a_prefix##destroy_recurse(a_rbt_type *rbtree, a_type *node, void (*cb)(	\
+  a_type *, void *), void *arg) {					\
+    if (node == NULL) {							\
+	return;								\
+    }									\
+    a_prefix##destroy_recurse(rbtree, rbtn_left_get(a_type, a_field,	\
+      node), cb, arg);							\
+    rbtn_left_set(a_type, a_field, (node), NULL);			\
+    a_prefix##destroy_recurse(rbtree, rbtn_right_get(a_type, a_field,	\
+      node), cb, arg);							\
+    rbtn_right_set(a_type, a_field, (node), NULL);			\
+    if (cb) {								\
+	cb(node, arg);							\
+    }									\
+}									\
+a_attr void								\
+a_prefix##destroy(a_rbt_type *rbtree, void (*cb)(a_type *, void *),	\
+  void *arg) {								\
+    a_prefix##destroy_recurse(rbtree, rbtree->rbt_root, cb, arg);	\
+    rbtree->rbt_root = NULL;						\
 }
 
 #endif /* RB_H_ */

include/jemalloc/internal/size_classes.sh

@@ -142,10 +142,10 @@ size_classes() {
 
   # All remaining groups.
   lg_grp=$((${lg_grp} + ${lg_g}))
-  while [ ${lg_grp} -lt ${ptr_bits} ] ; do
+  while [ ${lg_grp} -lt $((${ptr_bits} - 1)) ] ; do
     sep_line
     ndelta=1
-    if [ ${lg_grp} -eq $((${ptr_bits} - 1)) ] ; then
+    if [ ${lg_grp} -eq $((${ptr_bits} - 2)) ] ; then
       ndelta_limit=$((${g} - 1))
     else
       ndelta_limit=${g}

include/jemalloc/internal/smoothstep.h

@@ -0,0 +1,246 @@
+/*
+ * This file was generated by the following command:
+ *   sh smoothstep.sh smoother 200 24 3 15
+ */
+/******************************************************************************/
+#ifdef JEMALLOC_H_TYPES
+
+/*
+ * 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_H_TYPES */
+/******************************************************************************/
+#ifdef JEMALLOC_H_STRUCTS
+
+
+#endif /* JEMALLOC_H_STRUCTS */
+/******************************************************************************/
+#ifdef JEMALLOC_H_EXTERNS
+
+
+#endif /* JEMALLOC_H_EXTERNS */
+/******************************************************************************/
+#ifdef JEMALLOC_H_INLINES
+
+
+#endif /* JEMALLOC_H_INLINES */
+/******************************************************************************/

include/jemalloc/internal/smoothstep.sh

@@ -0,0 +1,115 @@
+#!/bin/sh
+#
+# Generate a discrete lookup table for a sigmoid function in the smoothstep
+# family (https://en.wikipedia.org/wiki/Smoothstep), where the lookup table
+# entries correspond to x in [1/nsteps, 2/nsteps, ..., nsteps/nsteps].  Encode
+# the entries using a binary fixed point representation.
+#
+# Usage: smoothstep.sh <variant> <nsteps> <bfp> <xprec> <yprec>
+#
+#        <variant> is in {smooth, smoother, smoothest}.
+#        <nsteps> must be greater than zero.
+#        <bfp> must be in [0..62]; reasonable values are roughly [10..30].
+#        <xprec> is x decimal precision.
+#        <yprec> is y decimal precision.
+
+#set -x
+
+cmd="sh smoothstep.sh $*"
+variant=$1
+nsteps=$2
+bfp=$3
+xprec=$4
+yprec=$5
+
+case "${variant}" in
+  smooth)
+    ;;
+  smoother)
+    ;;
+  smoothest)
+    ;;
+  *)
+    echo "Unsupported variant"
+    exit 1
+    ;;
+esac
+
+smooth() {
+  step=$1
+  y=`echo ${yprec} k ${step} ${nsteps} / sx _2 lx 3 ^ '*' 3 lx 2 ^ '*' + p | dc | tr -d '\\\\\n' | sed -e 's#^\.#0.#g'`
+  h=`echo ${yprec} k 2 ${bfp} ^ ${y} '*' p | dc | tr -d '\\\\\n' | sed -e 's#^\.#0.#g' | tr '.' ' ' | awk '{print $1}' `
+}
+
+smoother() {
+  step=$1
+  y=`echo ${yprec} k ${step} ${nsteps} / sx 6 lx 5 ^ '*' _15 lx 4 ^ '*' + 10 lx 3 ^ '*' + p | dc | tr -d '\\\\\n' | sed -e 's#^\.#0.#g'`
+  h=`echo ${yprec} k 2 ${bfp} ^ ${y} '*' p | dc | tr -d '\\\\\n' | sed -e 's#^\.#0.#g' | tr '.' ' ' | awk '{print $1}' `
+}
+
+smoothest() {
+  step=$1
+  y=`echo ${yprec} k ${step} ${nsteps} / sx _20 lx 7 ^ '*' 70 lx 6 ^ '*' + _84 lx 5 ^ '*' + 35 lx 4 ^ '*' + p | dc | tr -d '\\\\\n' | sed -e 's#^\.#0.#g'`
+  h=`echo ${yprec} k 2 ${bfp} ^ ${y} '*' p | dc | tr -d '\\\\\n' | sed -e 's#^\.#0.#g' | tr '.' ' ' | awk '{print $1}' `
+}
+
+cat <<EOF
+/*
+ * This file was generated by the following command:
+ *   $cmd
+ */
+/******************************************************************************/
+#ifdef JEMALLOC_H_TYPES
+
+/*
+ * This header defines a precomputed table based on the smoothstep family of
+ * sigmoidal curves (https://en.wikipedia.org/wiki/Smoothstep) that grow from 0
+ * to 1 in 0 <= x <= 1.  The table is stored as integer fixed point values so
+ * that floating point math can be avoided.
+ *
+ *                      3     2
+ *   smoothstep(x) = -2x  + 3x
+ *
+ *                       5      4      3
+ *   smootherstep(x) = 6x  - 15x  + 10x
+ *
+ *                          7      6      5      4
+ *   smootheststep(x) = -20x  + 70x  - 84x  + 35x
+ */
+
+#define	SMOOTHSTEP_VARIANT	"${variant}"
+#define	SMOOTHSTEP_NSTEPS	${nsteps}
+#define	SMOOTHSTEP_BFP		${bfp}
+#define	SMOOTHSTEP \\
+ /* STEP(step, h,                            x,     y) */ \\
+EOF
+
+s=1
+while [ $s -le $nsteps ] ; do
+  $variant ${s}
+  x=`echo ${xprec} k ${s} ${nsteps} / p | dc | tr -d '\\\\\n' | sed -e 's#^\.#0.#g'`
+  printf '    STEP(%4d, UINT64_C(0x%016x), %s, %s) \\\n' ${s} ${h} ${x} ${y}
+
+  s=$((s+1))
+done
+echo
+
+cat <<EOF
+#endif /* JEMALLOC_H_TYPES */
+/******************************************************************************/
+#ifdef JEMALLOC_H_STRUCTS
+
+
+#endif /* JEMALLOC_H_STRUCTS */
+/******************************************************************************/
+#ifdef JEMALLOC_H_EXTERNS
+
+
+#endif /* JEMALLOC_H_EXTERNS */
+/******************************************************************************/
+#ifdef JEMALLOC_H_INLINES
+
+
+#endif /* JEMALLOC_H_INLINES */
+/******************************************************************************/
+EOF

include/jemalloc/internal/stats.h

@@ -167,15 +167,25 @@ stats_cactive_get(void)
 JEMALLOC_INLINE void
 stats_cactive_add(size_t size)
 {
+	UNUSED size_t cactive;
 
-	atomic_add_z(&stats_cactive, size);
+	assert(size > 0);
+	assert((size & chunksize_mask) == 0);
+
+	cactive = atomic_add_z(&stats_cactive, size);
+	assert(cactive - size < cactive);
 }
 
 JEMALLOC_INLINE void
 stats_cactive_sub(size_t size)
 {
+	UNUSED size_t cactive;
 
-	atomic_sub_z(&stats_cactive, size);
+	assert(size > 0);
+	assert((size & chunksize_mask) == 0);
+
+	cactive = atomic_sub_z(&stats_cactive, size);
+	assert(cactive + size > cactive);
 }
 #endif
 

include/jemalloc/internal/tcache.h

@@ -70,13 +70,20 @@ struct tcache_bin_s {
 	int		low_water;	/* Min # cached since last GC. */
 	unsigned	lg_fill_div;	/* Fill (ncached_max >> lg_fill_div). */
 	unsigned	ncached;	/* # of cached 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;	/* 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(). */
-	unsigned	ev_cnt;		/* Event count since incremental GC. */
+	ticker_t	gc_ticker;	/* Drives incremental GC. */
 	szind_t		next_gc_bin;	/* Next bin to GC. */
 	tcache_bin_t	tbins[1];	/* Dynamically sized. */
 	/*
@@ -108,7 +115,7 @@ 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;
+extern unsigned	nhbins;
 
 /* Maximum cached size class. */
 extern size_t	tcache_maxclass;
@@ -126,7 +133,7 @@ extern tcaches_t	*tcaches;
 size_t	tcache_salloc(const void *ptr);
 void	tcache_event_hard(tsd_t *tsd, tcache_t *tcache);
 void	*tcache_alloc_small_hard(tsd_t *tsd, arena_t *arena, tcache_t *tcache,
-    tcache_bin_t *tbin, szind_t binind);
+    tcache_bin_t *tbin, szind_t binind, bool *tcache_success);
 void	tcache_bin_flush_small(tsd_t *tsd, tcache_t *tcache, tcache_bin_t *tbin,
     szind_t binind, unsigned rem);
 void	tcache_bin_flush_large(tsd_t *tsd, tcache_bin_t *tbin, szind_t binind,
@@ -155,15 +162,15 @@ void	tcache_flush(void);
 bool	tcache_enabled_get(void);
 tcache_t *tcache_get(tsd_t *tsd, bool create);
 void	tcache_enabled_set(bool enabled);
-void	*tcache_alloc_easy(tcache_bin_t *tbin);
+void	*tcache_alloc_easy(tcache_bin_t *tbin, bool *tcache_success);
 void	*tcache_alloc_small(tsd_t *tsd, arena_t *arena, tcache_t *tcache,
-    size_t size, bool zero);
+    size_t size, szind_t ind, bool zero, bool slow_path);
 void	*tcache_alloc_large(tsd_t *tsd, arena_t *arena, tcache_t *tcache,
-    size_t size, bool zero);
+    size_t size, szind_t ind, bool zero, bool slow_path);
 void	tcache_dalloc_small(tsd_t *tsd, tcache_t *tcache, void *ptr,
-    szind_t binind);
+    szind_t binind, bool slow_path);
 void	tcache_dalloc_large(tsd_t *tsd, tcache_t *tcache, void *ptr,
-    size_t size);
+    size_t size, bool slow_path);
 tcache_t	*tcaches_get(tsd_t *tsd, unsigned ind);
 #endif
 
@@ -240,51 +247,74 @@ tcache_event(tsd_t *tsd, tcache_t *tcache)
 	if (TCACHE_GC_INCR == 0)
 		return;
 
-	tcache->ev_cnt++;
+	if (unlikely(ticker_tick(&tcache->gc_ticker)))
-	assert(tcache->ev_cnt <= TCACHE_GC_INCR);
-	if (unlikely(tcache->ev_cnt == TCACHE_GC_INCR))
 		tcache_event_hard(tsd, tcache);
 }
 
 JEMALLOC_ALWAYS_INLINE void *
-tcache_alloc_easy(tcache_bin_t *tbin)
+tcache_alloc_easy(tcache_bin_t *tbin, bool *tcache_success)
 {
 	void *ret;
 
 	if (unlikely(tbin->ncached == 0)) {
 		tbin->low_water = -1;
+		*tcache_success = false;
 		return (NULL);
 	}
+	/*
+	 * tcache_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).
+	 */
+	*tcache_success = true;
+	ret = *(tbin->avail - tbin->ncached);
 	tbin->ncached--;
+
 	if (unlikely((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(tsd_t *tsd, arena_t *arena, tcache_t *tcache, size_t size,
-    bool zero)
+    szind_t binind, bool zero, bool slow_path)
 {
 	void *ret;
-	szind_t binind;
-	size_t usize;
 	tcache_bin_t *tbin;
+	bool tcache_success;
+	size_t usize JEMALLOC_CC_SILENCE_INIT(0);
 
-	binind = size2index(size);
 	assert(binind < NBINS);
 	tbin = &tcache->tbins[binind];
-	usize = index2size(binind);
+	ret = tcache_alloc_easy(tbin, &tcache_success);
-	ret = tcache_alloc_easy(tbin);
+	assert(tcache_success == (ret != NULL));
-	if (unlikely(ret == NULL)) {
+	if (unlikely(!tcache_success)) {
-		ret = tcache_alloc_small_hard(tsd, arena, tcache, tbin, binind);
+		bool tcache_hard_success;
-		if (ret == NULL)
+		arena = arena_choose(tsd, arena);
+		if (unlikely(arena == NULL))
+			return (NULL);
+
+		ret = tcache_alloc_small_hard(tsd, arena, tcache, tbin, 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 = index2size(binind);
 		assert(tcache_salloc(ret) == usize);
+	}
 
 	if (likely(!zero)) {
-		if (config_fill) {
+		if (slow_path && config_fill) {
 			if (unlikely(opt_junk_alloc)) {
 				arena_alloc_junk_small(ret,
 				    &arena_bin_info[binind], false);
@@ -292,7 +322,7 @@ tcache_alloc_small(tsd_t *tsd, arena_t *arena, tcache_t *tcache, size_t size,
 				memset(ret, 0, usize);
 		}
 	} else {
-		if (config_fill && unlikely(opt_junk_alloc)) {
+		if (slow_path && config_fill && unlikely(opt_junk_alloc)) {
 			arena_alloc_junk_small(ret, &arena_bin_info[binind],
 			    true);
 		}
@@ -309,28 +339,38 @@ tcache_alloc_small(tsd_t *tsd, arena_t *arena, tcache_t *tcache, size_t size,
 
 JEMALLOC_ALWAYS_INLINE void *
 tcache_alloc_large(tsd_t *tsd, arena_t *arena, tcache_t *tcache, size_t size,
-    bool zero)
+    szind_t binind, bool zero, bool slow_path)
 {
 	void *ret;
-	szind_t binind;
-	size_t usize;
 	tcache_bin_t *tbin;
+	bool tcache_success;
 
-	binind = size2index(size);
-	usize = index2size(binind);
-	assert(usize <= tcache_maxclass);
 	assert(binind < nhbins);
 	tbin = &tcache->tbins[binind];
-	ret = tcache_alloc_easy(tbin);
+	ret = tcache_alloc_easy(tbin, &tcache_success);
-	if (unlikely(ret == NULL)) {
+	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.
 		 */
-		ret = arena_malloc_large(arena, usize, zero);
+		arena = arena_choose(tsd, arena);
+		if (unlikely(arena == NULL))
+			return (NULL);
+
+		ret = arena_malloc_large(tsd, arena, binind, 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 = index2size(binind);
+			assert(usize <= tcache_maxclass);
+		}
+
 		if (config_prof && usize == LARGE_MINCLASS) {
 			arena_chunk_t *chunk =
 			    (arena_chunk_t *)CHUNK_ADDR2BASE(ret);
@@ -340,7 +380,7 @@ tcache_alloc_large(tsd_t *tsd, arena_t *arena, tcache_t *tcache, size_t size,
 			    BININD_INVALID);
 		}
 		if (likely(!zero)) {
-			if (config_fill) {
+			if (slow_path && config_fill) {
 				if (unlikely(opt_junk_alloc))
 					memset(ret, 0xa5, usize);
 				else if (unlikely(opt_zero))
@@ -360,14 +400,15 @@ tcache_alloc_large(tsd_t *tsd, arena_t *arena, tcache_t *tcache, size_t size,
 }
 
 JEMALLOC_ALWAYS_INLINE void
-tcache_dalloc_small(tsd_t *tsd, tcache_t *tcache, void *ptr, szind_t binind)
+tcache_dalloc_small(tsd_t *tsd, tcache_t *tcache, void *ptr, szind_t binind,
+    bool slow_path)
 {
 	tcache_bin_t *tbin;
 	tcache_bin_info_t *tbin_info;
 
 	assert(tcache_salloc(ptr) <= SMALL_MAXCLASS);
 
-	if (config_fill && unlikely(opt_junk_free))
+	if (slow_path && config_fill && unlikely(opt_junk_free))
 		arena_dalloc_junk_small(ptr, &arena_bin_info[binind]);
 
 	tbin = &tcache->tbins[binind];
@@ -377,14 +418,15 @@ tcache_dalloc_small(tsd_t *tsd, tcache_t *tcache, void *ptr, szind_t binind)
 		    (tbin_info->ncached_max >> 1));
 	}
 	assert(tbin->ncached < tbin_info->ncached_max);
-	tbin->avail[tbin->ncached] = ptr;
 	tbin->ncached++;
+	*(tbin->avail - tbin->ncached) = ptr;
 
 	tcache_event(tsd, tcache);
 }
 
 JEMALLOC_ALWAYS_INLINE void
-tcache_dalloc_large(tsd_t *tsd, tcache_t *tcache, void *ptr, size_t size)
+tcache_dalloc_large(tsd_t *tsd, tcache_t *tcache, void *ptr, size_t size,
+    bool slow_path)
 {
 	szind_t binind;
 	tcache_bin_t *tbin;
@@ -396,7 +438,7 @@ tcache_dalloc_large(tsd_t *tsd, tcache_t *tcache, void *ptr, size_t size)
 
 	binind = size2index(size);
 
-	if (config_fill && unlikely(opt_junk_free))
+	if (slow_path && config_fill && unlikely(opt_junk_free))
 		arena_dalloc_junk_large(ptr, size);
 
 	tbin = &tcache->tbins[binind];
@@ -406,8 +448,8 @@ tcache_dalloc_large(tsd_t *tsd, tcache_t *tcache, void *ptr, size_t size)
 		    (tbin_info->ncached_max >> 1), tcache);
 	}
 	assert(tbin->ncached < tbin_info->ncached_max);
-	tbin->avail[tbin->ncached] = ptr;
 	tbin->ncached++;
+	*(tbin->avail - tbin->ncached) = ptr;
 
 	tcache_event(tsd, tcache);
 }

include/jemalloc/internal/ticker.h

@@ -0,0 +1,75 @@
+/******************************************************************************/
+#ifdef JEMALLOC_H_TYPES
+
+typedef struct ticker_s ticker_t;
+
+#endif /* JEMALLOC_H_TYPES */
+/******************************************************************************/
+#ifdef JEMALLOC_H_STRUCTS
+
+struct ticker_s {
+	int32_t	tick;
+	int32_t	nticks;
+};
+
+#endif /* JEMALLOC_H_STRUCTS */
+/******************************************************************************/
+#ifdef JEMALLOC_H_EXTERNS
+
+#endif /* JEMALLOC_H_EXTERNS */
+/******************************************************************************/
+#ifdef JEMALLOC_H_INLINES
+
+#ifndef JEMALLOC_ENABLE_INLINE
+void	ticker_init(ticker_t *ticker, int32_t nticks);
+void	ticker_copy(ticker_t *ticker, const ticker_t *other);
+int32_t	ticker_read(const ticker_t *ticker);
+bool	ticker_ticks(ticker_t *ticker, int32_t nticks);
+bool	ticker_tick(ticker_t *ticker);
+#endif
+
+#if (defined(JEMALLOC_ENABLE_INLINE) || defined(JEMALLOC_TICKER_C_))
+JEMALLOC_INLINE void
+ticker_init(ticker_t *ticker, int32_t nticks)
+{
+
+	ticker->tick = nticks;
+	ticker->nticks = nticks;
+}
+
+JEMALLOC_INLINE void
+ticker_copy(ticker_t *ticker, const ticker_t *other)
+{
+
+	*ticker = *other;
+}
+
+JEMALLOC_INLINE int32_t
+ticker_read(const ticker_t *ticker)
+{
+
+	return (ticker->tick);
+}
+
+JEMALLOC_INLINE bool
+ticker_ticks(ticker_t *ticker, int32_t nticks)
+{
+
+	if (unlikely(ticker->tick < nticks)) {
+		ticker->tick = ticker->nticks;
+		return (true);
+	}
+	ticker->tick -= nticks;
+	return(false);
+}
+
+JEMALLOC_INLINE bool
+ticker_tick(ticker_t *ticker)
+{
+
+	return (ticker_ticks(ticker, 1));
+}
+#endif
+
+#endif /* JEMALLOC_H_INLINES */
+/******************************************************************************/

include/jemalloc/internal/tsd.h

@@ -537,9 +537,9 @@ struct tsd_init_head_s {
     O(thread_deallocated,	uint64_t)				\
     O(prof_tdata,		prof_tdata_t *)				\
     O(arena,			arena_t *)				\
-    O(arenas_cache,		arena_t **)				\
+    O(arenas_tdata,		arena_tdata_t *)			\
-    O(narenas_cache,		unsigned)				\
+    O(narenas_tdata,		unsigned)				\
-    O(arenas_cache_bypass,	bool)					\
+    O(arenas_tdata_bypass,	bool)					\
     O(tcache_enabled,		tcache_enabled_t)			\
     O(quarantine,		quarantine_t *)				\
 

include/jemalloc/internal/util.h

@@ -81,49 +81,7 @@
 #	define unreachable()
 #endif
 
-/*
+#include "jemalloc/internal/assert.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). */
 #define	cassert(c) do {							\
@@ -163,10 +121,16 @@ void	malloc_printf(const char *format, ...) JEMALLOC_FORMAT_PRINTF(1, 2);
 #ifdef JEMALLOC_H_INLINES
 
 #ifndef JEMALLOC_ENABLE_INLINE
-int	jemalloc_ffsl(long bitmap);
+unsigned	ffs_llu(unsigned long long bitmap);
-int	jemalloc_ffs(int bitmap);
+unsigned	ffs_lu(unsigned long bitmap);
-size_t	pow2_ceil(size_t x);
+unsigned	ffs_u(unsigned bitmap);
-size_t	lg_floor(size_t x);
+unsigned	ffs_zu(size_t bitmap);
+unsigned	ffs_u64(uint64_t bitmap);
+unsigned	ffs_u32(uint32_t bitmap);
+uint64_t	pow2_ceil_u64(uint64_t x);
+uint32_t	pow2_ceil_u32(uint32_t x);
+size_t	pow2_ceil_zu(size_t x);
+unsigned	lg_floor(size_t x);
 void	set_errno(int errnum);
 int	get_errno(void);
 #endif
@@ -174,27 +138,74 @@ int	get_errno(void);
 #if (defined(JEMALLOC_ENABLE_INLINE) || defined(JEMALLOC_UTIL_C_))
 
 /* Sanity check. */
-#if !defined(JEMALLOC_INTERNAL_FFSL) || !defined(JEMALLOC_INTERNAL_FFS)
+#if !defined(JEMALLOC_INTERNAL_FFSLL) || !defined(JEMALLOC_INTERNAL_FFSL) \
-#  error Both JEMALLOC_INTERNAL_FFSL && JEMALLOC_INTERNAL_FFS should have been defined by configure
+    || !defined(JEMALLOC_INTERNAL_FFS)
+#  error JEMALLOC_INTERNAL_FFS{,L,LL} should have been defined by configure
 #endif
 
-JEMALLOC_ALWAYS_INLINE int
+JEMALLOC_ALWAYS_INLINE unsigned
-jemalloc_ffsl(long bitmap)
+ffs_llu(unsigned long long bitmap)
+{
+
+	return (JEMALLOC_INTERNAL_FFSLL(bitmap));
+}
+
+JEMALLOC_ALWAYS_INLINE unsigned
+ffs_lu(unsigned long bitmap)
 {
 
 	return (JEMALLOC_INTERNAL_FFSL(bitmap));
 }
 
-JEMALLOC_ALWAYS_INLINE int
+JEMALLOC_ALWAYS_INLINE unsigned
-jemalloc_ffs(int bitmap)
+ffs_u(unsigned bitmap)
 {
 
 	return (JEMALLOC_INTERNAL_FFS(bitmap));
 }
 
-/* Compute the smallest power of 2 that is >= x. */
+JEMALLOC_ALWAYS_INLINE unsigned
-JEMALLOC_INLINE size_t
+ffs_zu(size_t bitmap)
-pow2_ceil(size_t x)
+{
+
+#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
+}
+
+JEMALLOC_ALWAYS_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
+}
+
+JEMALLOC_ALWAYS_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));
+}
+
+JEMALLOC_INLINE uint64_t
+pow2_ceil_u64(uint64_t x)
 {
 
 	x--;
@@ -203,15 +214,39 @@ pow2_ceil(size_t x)
 	x |= x >> 4;
 	x |= x >> 8;
 	x |= x >> 16;
-#if (LG_SIZEOF_PTR == 3)
 	x |= x >> 32;
-#endif
 	x++;
 	return (x);
 }
 
-#if (defined(__i386__) || defined(__amd64__) || defined(__x86_64__))
+JEMALLOC_INLINE uint32_t
+pow2_ceil_u32(uint32_t x)
+{
+
+	x--;
+	x |= x >> 1;
+	x |= x >> 2;
+	x |= x >> 4;
+	x |= x >> 8;
+	x |= x >> 16;
+	x++;
+	return (x);
+}
+
+/* Compute the smallest power of 2 that is >= x. */
 JEMALLOC_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__))
+JEMALLOC_INLINE unsigned
 lg_floor(size_t x)
 {
 	size_t ret;
@@ -222,10 +257,11 @@ lg_floor(size_t x)
 	    : "=r"(ret) // Outputs.
 	    : "r"(x)    // Inputs.
 	    );
-	return (ret);
+	assert(ret < UINT_MAX);
+	return ((unsigned)ret);
 }
 #elif (defined(_MSC_VER))
-JEMALLOC_INLINE size_t
+JEMALLOC_INLINE unsigned
 lg_floor(size_t x)
 {
 	unsigned long ret;
@@ -237,12 +273,13 @@ lg_floor(size_t x)
 #elif (LG_SIZEOF_PTR == 2)
 	_BitScanReverse(&ret, x);
 #else
-#  error "Unsupported type sizes for lg_floor()"
+#  error "Unsupported type size for lg_floor()"
 #endif
-	return (ret);
+	assert(ret < UINT_MAX);
+	return ((unsigned)ret);
 }
 #elif (defined(JEMALLOC_HAVE_BUILTIN_CLZ))
-JEMALLOC_INLINE size_t
+JEMALLOC_INLINE unsigned
 lg_floor(size_t x)
 {
 
@@ -253,11 +290,11 @@ lg_floor(size_t x)
 #elif (LG_SIZEOF_PTR == LG_SIZEOF_LONG)
 	return (((8 << LG_SIZEOF_PTR) - 1) - __builtin_clzl(x));
 #else
-#  error "Unsupported type sizes for lg_floor()"
+#  error "Unsupported type size for lg_floor()"
 #endif
 }
 #else
-JEMALLOC_INLINE size_t
+JEMALLOC_INLINE unsigned
 lg_floor(size_t x)
 {
 
@@ -268,20 +305,13 @@ lg_floor(size_t x)
 	x |= (x >> 4);
 	x |= (x >> 8);
 	x |= (x >> 16);
-#if (LG_SIZEOF_PTR == 3 && LG_SIZEOF_PTR == LG_SIZEOF_LONG)
+#if (LG_SIZEOF_PTR == 3)
 	x |= (x >> 32);
-	if (x == KZU(0xffffffffffffffff))
-		return (63);
-	x++;
-	return (jemalloc_ffsl(x) - 2);
-#elif (LG_SIZEOF_PTR == 2)
-	if (x == KZU(0xffffffff))
-		return (31);
-	x++;
-	return (jemalloc_ffs(x) - 2);
-#else
-#  error "Unsupported type sizes for lg_floor()"
 #endif
+	if (x == SIZE_T_MAX)
+		return ((8 << LG_SIZEOF_PTR) - 1);
+	x++;
+	return (ffs_zu(x) - 2);
 }
 #endif
 

include/jemalloc/jemalloc_defs.h.in

@@ -33,5 +33,13 @@
  */
 #undef JEMALLOC_USE_CXX_THROW
 
+#ifdef _MSC_VER
+#  ifdef _WIN64
+#    define LG_SIZEOF_PTR_WIN 3
+#  else
+#    define LG_SIZEOF_PTR_WIN 2
+#  endif
+#endif
+
 /* sizeof(void *) == 2^LG_SIZEOF_PTR. */
 #undef LG_SIZEOF_PTR

include/jemalloc/jemalloc_macros.h.in

@@ -11,12 +11,13 @@
 #define	JEMALLOC_VERSION_NREV @jemalloc_version_nrev@
 #define	JEMALLOC_VERSION_GID "@jemalloc_version_gid@"
 
-#  define MALLOCX_LG_ALIGN(la)	(la)
+#  define MALLOCX_LG_ALIGN(la)	((int)(la))
 #  if LG_SIZEOF_PTR == 2
-#    define MALLOCX_ALIGN(a)	(ffs(a)-1)
+#    define MALLOCX_ALIGN(a)	((int)(ffs(a)-1))
 #  else
 #    define MALLOCX_ALIGN(a)						\
-	 ((a < (size_t)INT_MAX) ? ffs(a)-1 : ffs(a>>32)+31)
+       ((int)(((a) < (size_t)INT_MAX) ? ffs((int)(a))-1 :		\
+       ffs((int)((a)>>32))+31))
 #  endif
 #  define MALLOCX_ZERO	((int)0x40)
 /*
@@ -36,32 +37,7 @@
 #  define JEMALLOC_CXX_THROW
 #endif
 
-#ifdef JEMALLOC_HAVE_ATTR
+#if _MSC_VER
-#  define JEMALLOC_ATTR(s) __attribute__((s))
-#  define JEMALLOC_ALIGNED(s) JEMALLOC_ATTR(aligned(s))
-#  ifdef JEMALLOC_HAVE_ATTR_ALLOC_SIZE
-#    define JEMALLOC_ALLOC_SIZE(s) JEMALLOC_ATTR(alloc_size(s))
-#    define JEMALLOC_ALLOC_SIZE2(s1, s2) JEMALLOC_ATTR(alloc_size(s1, s2))
-#  else
-#    define JEMALLOC_ALLOC_SIZE(s)
-#    define JEMALLOC_ALLOC_SIZE2(s1, s2)
-#  endif
-#  ifndef JEMALLOC_EXPORT
-#    define JEMALLOC_EXPORT JEMALLOC_ATTR(visibility("default"))
-#  endif
-#  ifdef JEMALLOC_HAVE_ATTR_FORMAT_GNU_PRINTF
-#    define JEMALLOC_FORMAT_PRINTF(s, i) JEMALLOC_ATTR(format(gnu_printf, s, i))
-#  elif defined(JEMALLOC_HAVE_ATTR_FORMAT_PRINTF)
-#    define JEMALLOC_FORMAT_PRINTF(s, i) JEMALLOC_ATTR(format(printf, s, i))
-#  else
-#    define JEMALLOC_FORMAT_PRINTF(s, i)
-#  endif
-#  define JEMALLOC_NOINLINE JEMALLOC_ATTR(noinline)
-#  define JEMALLOC_NOTHROW JEMALLOC_ATTR(nothrow)
-#  define JEMALLOC_SECTION(s) JEMALLOC_ATTR(section(s))
-#  define JEMALLOC_RESTRICT_RETURN
-#  define JEMALLOC_ALLOCATOR
-#elif _MSC_VER
 #  define JEMALLOC_ATTR(s)
 #  define JEMALLOC_ALIGNED(s) __declspec(align(s))
 #  define JEMALLOC_ALLOC_SIZE(s)
@@ -87,6 +63,31 @@
 #  else
 #    define JEMALLOC_ALLOCATOR
 #  endif
+#elif defined(JEMALLOC_HAVE_ATTR)
+#  define JEMALLOC_ATTR(s) __attribute__((s))
+#  define JEMALLOC_ALIGNED(s) JEMALLOC_ATTR(aligned(s))
+#  ifdef JEMALLOC_HAVE_ATTR_ALLOC_SIZE
+#    define JEMALLOC_ALLOC_SIZE(s) JEMALLOC_ATTR(alloc_size(s))
+#    define JEMALLOC_ALLOC_SIZE2(s1, s2) JEMALLOC_ATTR(alloc_size(s1, s2))
+#  else
+#    define JEMALLOC_ALLOC_SIZE(s)
+#    define JEMALLOC_ALLOC_SIZE2(s1, s2)
+#  endif
+#  ifndef JEMALLOC_EXPORT
+#    define JEMALLOC_EXPORT JEMALLOC_ATTR(visibility("default"))
+#  endif
+#  ifdef JEMALLOC_HAVE_ATTR_FORMAT_GNU_PRINTF
+#    define JEMALLOC_FORMAT_PRINTF(s, i) JEMALLOC_ATTR(format(gnu_printf, s, i))
+#  elif defined(JEMALLOC_HAVE_ATTR_FORMAT_PRINTF)
+#    define JEMALLOC_FORMAT_PRINTF(s, i) JEMALLOC_ATTR(format(printf, s, i))
+#  else
+#    define JEMALLOC_FORMAT_PRINTF(s, i)
+#  endif
+#  define JEMALLOC_NOINLINE JEMALLOC_ATTR(noinline)
+#  define JEMALLOC_NOTHROW JEMALLOC_ATTR(nothrow)
+#  define JEMALLOC_SECTION(s) JEMALLOC_ATTR(section(s))
+#  define JEMALLOC_RESTRICT_RETURN
+#  define JEMALLOC_ALLOCATOR
 #else
 #  define JEMALLOC_ATTR(s)
 #  define JEMALLOC_ALIGNED(s)

include/msvc_compat/strings.h

@@ -21,7 +21,37 @@ static __forceinline int ffs(int x)
 	return (ffsl(x));
 }
 
+#  ifdef  _M_X64
+#    pragma intrinsic(_BitScanForward64)
+#  endif
+
+static __forceinline int ffsll(unsigned __int64 x)
+{
+	unsigned long i;
+#ifdef  _M_X64
+	if (_BitScanForward64(&i, x))
+		return (i + 1);
+	return (0);
 #else
+// Fallback for 32-bit build where 64-bit version not available
+// assuming little endian
+	union {
+		unsigned __int64 ll;
+		unsigned   long l[2];
+	} s;
+
+	s.ll = x;
+
+	if (_BitScanForward(&i, s.l[0]))
+		return (i + 1);
+	else if(_BitScanForward(&i, s.l[1]))
+		return (i + 33);
+	return (0);
+#endif
+}
+
+#else
+#  define ffsll(x) __builtin_ffsll(x)
 #  define ffsl(x) __builtin_ffsl(x)
 #  define ffs(x) __builtin_ffs(x)
 #endif

msvc/ReadMe.txt

@@ -0,0 +1,24 @@
+
+How to build jemalloc for Windows
+=================================
+
+1. Install Cygwin with at least the following packages:
+   * autoconf
+   * autogen
+   * gawk
+   * grep
+   * sed
+
+2. Install Visual Studio 2015 with Visual C++
+
+3. Add Cygwin\bin to the PATH environment variable
+
+4. Open "VS2015 x86 Native Tools Command Prompt"
+   (note: x86/x64 doesn't matter at this point)
+
+5. Generate header files:
+   sh -c "./autogen.sh CC=cl --enable-lazy-lock=no"
+
+6. Now the project can be opened and built in Visual Studio:
+   msvc\jemalloc_vc2015.sln
+

msvc/jemalloc_vc2015.sln

@@ -0,0 +1,63 @@
+
+Microsoft Visual Studio Solution File, Format Version 12.00
+# Visual Studio 14
+VisualStudioVersion = 14.0.24720.0
+MinimumVisualStudioVersion = 10.0.40219.1
+Project("{2150E333-8FDC-42A3-9474-1A3956D46DE8}") = "Solution Items", "Solution Items", "{70A99006-6DE9-472B-8F83-4CEE6C616DF3}"
+	ProjectSection(SolutionItems) = preProject
+		ReadMe.txt = ReadMe.txt
+	EndProjectSection
+EndProject
+Project("{8BC9CEB8-8B4A-11D0-8D11-00A0C91BC942}") = "jemalloc", "projects\vc2015\jemalloc\jemalloc.vcxproj", "{8D6BB292-9E1C-413D-9F98-4864BDC1514A}"
+EndProject
+Project("{8BC9CEB8-8B4A-11D0-8D11-00A0C91BC942}") = "test_threads", "projects\vc2015\test_threads\test_threads.vcxproj", "{09028CFD-4EB7-491D-869C-0708DB97ED44}"
+EndProject
+Global
+	GlobalSection(SolutionConfigurationPlatforms) = preSolution
+		Debug|x64 = Debug|x64
+		Debug|x86 = Debug|x86
+		Debug-static|x64 = Debug-static|x64
+		Debug-static|x86 = Debug-static|x86
+		Release|x64 = Release|x64
+		Release|x86 = Release|x86
+		Release-static|x64 = Release-static|x64
+		Release-static|x86 = Release-static|x86
+	EndGlobalSection
+	GlobalSection(ProjectConfigurationPlatforms) = postSolution
+		{8D6BB292-9E1C-413D-9F98-4864BDC1514A}.Debug|x64.ActiveCfg = Debug|x64
+		{8D6BB292-9E1C-413D-9F98-4864BDC1514A}.Debug|x64.Build.0 = Debug|x64
+		{8D6BB292-9E1C-413D-9F98-4864BDC1514A}.Debug|x86.ActiveCfg = Debug|Win32
+		{8D6BB292-9E1C-413D-9F98-4864BDC1514A}.Debug|x86.Build.0 = Debug|Win32
+		{8D6BB292-9E1C-413D-9F98-4864BDC1514A}.Debug-static|x64.ActiveCfg = Debug-static|x64
+		{8D6BB292-9E1C-413D-9F98-4864BDC1514A}.Debug-static|x64.Build.0 = Debug-static|x64
+		{8D6BB292-9E1C-413D-9F98-4864BDC1514A}.Debug-static|x86.ActiveCfg = Debug-static|Win32
+		{8D6BB292-9E1C-413D-9F98-4864BDC1514A}.Debug-static|x86.Build.0 = Debug-static|Win32
+		{8D6BB292-9E1C-413D-9F98-4864BDC1514A}.Release|x64.ActiveCfg = Release|x64
+		{8D6BB292-9E1C-413D-9F98-4864BDC1514A}.Release|x64.Build.0 = Release|x64
+		{8D6BB292-9E1C-413D-9F98-4864BDC1514A}.Release|x86.ActiveCfg = Release|Win32
+		{8D6BB292-9E1C-413D-9F98-4864BDC1514A}.Release|x86.Build.0 = Release|Win32
+		{8D6BB292-9E1C-413D-9F98-4864BDC1514A}.Release-static|x64.ActiveCfg = Release-static|x64
+		{8D6BB292-9E1C-413D-9F98-4864BDC1514A}.Release-static|x64.Build.0 = Release-static|x64
+		{8D6BB292-9E1C-413D-9F98-4864BDC1514A}.Release-static|x86.ActiveCfg = Release-static|Win32
+		{8D6BB292-9E1C-413D-9F98-4864BDC1514A}.Release-static|x86.Build.0 = Release-static|Win32
+		{09028CFD-4EB7-491D-869C-0708DB97ED44}.Debug|x64.ActiveCfg = Debug|x64
+		{09028CFD-4EB7-491D-869C-0708DB97ED44}.Debug|x64.Build.0 = Debug|x64
+		{09028CFD-4EB7-491D-869C-0708DB97ED44}.Debug|x86.ActiveCfg = Debug|Win32
+		{09028CFD-4EB7-491D-869C-0708DB97ED44}.Debug|x86.Build.0 = Debug|Win32
+		{09028CFD-4EB7-491D-869C-0708DB97ED44}.Debug-static|x64.ActiveCfg = Debug-static|x64
+		{09028CFD-4EB7-491D-869C-0708DB97ED44}.Debug-static|x64.Build.0 = Debug-static|x64
+		{09028CFD-4EB7-491D-869C-0708DB97ED44}.Debug-static|x86.ActiveCfg = Debug-static|Win32
+		{09028CFD-4EB7-491D-869C-0708DB97ED44}.Debug-static|x86.Build.0 = Debug-static|Win32
+		{09028CFD-4EB7-491D-869C-0708DB97ED44}.Release|x64.ActiveCfg = Release|x64
+		{09028CFD-4EB7-491D-869C-0708DB97ED44}.Release|x64.Build.0 = Release|x64
+		{09028CFD-4EB7-491D-869C-0708DB97ED44}.Release|x86.ActiveCfg = Release|Win32
+		{09028CFD-4EB7-491D-869C-0708DB97ED44}.Release|x86.Build.0 = Release|Win32
+		{09028CFD-4EB7-491D-869C-0708DB97ED44}.Release-static|x64.ActiveCfg = Release-static|x64
+		{09028CFD-4EB7-491D-869C-0708DB97ED44}.Release-static|x64.Build.0 = Release-static|x64
+		{09028CFD-4EB7-491D-869C-0708DB97ED44}.Release-static|x86.ActiveCfg = Release-static|Win32
+		{09028CFD-4EB7-491D-869C-0708DB97ED44}.Release-static|x86.Build.0 = Release-static|Win32
+	EndGlobalSection
+	GlobalSection(SolutionProperties) = preSolution
+		HideSolutionNode = FALSE
+	EndGlobalSection
+EndGlobal

msvc/projects/vc2015/jemalloc/jemalloc.vcxproj

@@ -0,0 +1,392 @@
+<?xml version="1.0" encoding="utf-8"?>
+<Project DefaultTargets="Build" ToolsVersion="14.0" xmlns="http://schemas.microsoft.com/developer/msbuild/2003">
+  <ItemGroup Label="ProjectConfigurations">
+    <ProjectConfiguration Include="Debug-static|Win32">
+      <Configuration>Debug-static</Configuration>
+      <Platform>Win32</Platform>
+    </ProjectConfiguration>
+    <ProjectConfiguration Include="Debug-static|x64">
+      <Configuration>Debug-static</Configuration>
+      <Platform>x64</Platform>
+    </ProjectConfiguration>
+    <ProjectConfiguration Include="Debug|Win32">
+      <Configuration>Debug</Configuration>
+      <Platform>Win32</Platform>
+    </ProjectConfiguration>
+    <ProjectConfiguration Include="Release-static|Win32">
+      <Configuration>Release-static</Configuration>
+      <Platform>Win32</Platform>
+    </ProjectConfiguration>
+    <ProjectConfiguration Include="Release-static|x64">
+      <Configuration>Release-static</Configuration>
+      <Platform>x64</Platform>
+    </ProjectConfiguration>
+    <ProjectConfiguration Include="Release|Win32">
+      <Configuration>Release</Configuration>
+      <Platform>Win32</Platform>
+    </ProjectConfiguration>
+    <ProjectConfiguration Include="Debug|x64">
+      <Configuration>Debug</Configuration>
+      <Platform>x64</Platform>
+    </ProjectConfiguration>
+    <ProjectConfiguration Include="Release|x64">
+      <Configuration>Release</Configuration>
+      <Platform>x64</Platform>
+    </ProjectConfiguration>
+  </ItemGroup>
+  <ItemGroup>
+    <ClInclude Include="..\..\..\..\include\jemalloc\internal\arena.h" />
+    <ClInclude Include="..\..\..\..\include\jemalloc\internal\assert.h" />
+    <ClInclude Include="..\..\..\..\include\jemalloc\internal\atomic.h" />
+    <ClInclude Include="..\..\..\..\include\jemalloc\internal\base.h" />
+    <ClInclude Include="..\..\..\..\include\jemalloc\internal\bitmap.h" />
+    <ClInclude Include="..\..\..\..\include\jemalloc\internal\chunk.h" />
+    <ClInclude Include="..\..\..\..\include\jemalloc\internal\chunk_dss.h" />
+    <ClInclude Include="..\..\..\..\include\jemalloc\internal\chunk_mmap.h" />
+    <ClInclude Include="..\..\..\..\include\jemalloc\internal\ckh.h" />
+    <ClInclude Include="..\..\..\..\include\jemalloc\internal\ctl.h" />
+    <ClInclude Include="..\..\..\..\include\jemalloc\internal\extent.h" />
+    <ClInclude Include="..\..\..\..\include\jemalloc\internal\hash.h" />
+    <ClInclude Include="..\..\..\..\include\jemalloc\internal\huge.h" />
+    <ClInclude Include="..\..\..\..\include\jemalloc\internal\jemalloc_internal.h" />
+    <ClInclude Include="..\..\..\..\include\jemalloc\internal\jemalloc_internal_decls.h" />
+    <ClInclude Include="..\..\..\..\include\jemalloc\internal\jemalloc_internal_defs.h" />
+    <ClInclude Include="..\..\..\..\include\jemalloc\internal\jemalloc_internal_macros.h" />
+    <ClInclude Include="..\..\..\..\include\jemalloc\internal\mb.h" />
+    <ClInclude Include="..\..\..\..\include\jemalloc\internal\mutex.h" />
+    <ClInclude Include="..\..\..\..\include\jemalloc\internal\pages.h" />
+    <ClInclude Include="..\..\..\..\include\jemalloc\internal\private_namespace.h" />
+    <ClInclude Include="..\..\..\..\include\jemalloc\internal\private_unnamespace.h" />
+    <ClInclude Include="..\..\..\..\include\jemalloc\internal\prng.h" />
+    <ClInclude Include="..\..\..\..\include\jemalloc\internal\prof.h" />
+    <ClInclude Include="..\..\..\..\include\jemalloc\internal\public_namespace.h" />
+    <ClInclude Include="..\..\..\..\include\jemalloc\internal\public_unnamespace.h" />
+    <ClInclude Include="..\..\..\..\include\jemalloc\internal\ql.h" />
+    <ClInclude Include="..\..\..\..\include\jemalloc\internal\qr.h" />
+    <ClInclude Include="..\..\..\..\include\jemalloc\internal\quarantine.h" />
+    <ClInclude Include="..\..\..\..\include\jemalloc\internal\rb.h" />
+    <ClInclude Include="..\..\..\..\include\jemalloc\internal\rtree.h" />
+    <ClInclude Include="..\..\..\..\include\jemalloc\internal\size_classes.h" />
+    <ClInclude Include="..\..\..\..\include\jemalloc\internal\stats.h" />
+    <ClInclude Include="..\..\..\..\include\jemalloc\internal\tcache.h" />
+    <ClInclude Include="..\..\..\..\include\jemalloc\internal\tsd.h" />
+    <ClInclude Include="..\..\..\..\include\jemalloc\internal\util.h" />
+    <ClInclude Include="..\..\..\..\include\jemalloc\internal\valgrind.h" />
+    <ClInclude Include="..\..\..\..\include\jemalloc\jemalloc.h" />
+    <ClInclude Include="..\..\..\..\include\jemalloc\jemalloc_defs.h" />
+    <ClInclude Include="..\..\..\..\include\jemalloc\jemalloc_macros.h" />
+    <ClInclude Include="..\..\..\..\include\jemalloc\jemalloc_mangle.h" />
+    <ClInclude Include="..\..\..\..\include\jemalloc\jemalloc_protos.h" />
+    <ClInclude Include="..\..\..\..\include\jemalloc\jemalloc_protos_jet.h" />
+    <ClInclude Include="..\..\..\..\include\jemalloc\jemalloc_rename.h" />
+    <ClInclude Include="..\..\..\..\include\jemalloc\jemalloc_typedefs.h" />
+    <ClInclude Include="..\..\..\..\include\msvc_compat\C99\stdbool.h" />
+    <ClInclude Include="..\..\..\..\include\msvc_compat\C99\stdint.h" />
+    <ClInclude Include="..\..\..\..\include\msvc_compat\strings.h" />
+    <ClInclude Include="..\..\..\..\include\msvc_compat\windows_extra.h" />
+  </ItemGroup>
+  <ItemGroup>
+    <ClCompile Include="..\..\..\..\src\arena.c" />
+    <ClCompile Include="..\..\..\..\src\atomic.c" />
+    <ClCompile Include="..\..\..\..\src\base.c" />
+    <ClCompile Include="..\..\..\..\src\bitmap.c" />
+    <ClCompile Include="..\..\..\..\src\chunk.c" />
+    <ClCompile Include="..\..\..\..\src\chunk_dss.c" />
+    <ClCompile Include="..\..\..\..\src\chunk_mmap.c" />
+    <ClCompile Include="..\..\..\..\src\ckh.c" />
+    <ClCompile Include="..\..\..\..\src\ctl.c" />
+    <ClCompile Include="..\..\..\..\src\extent.c" />
+    <ClCompile Include="..\..\..\..\src\hash.c" />
+    <ClCompile Include="..\..\..\..\src\huge.c" />
+    <ClCompile Include="..\..\..\..\src\jemalloc.c" />
+    <ClCompile Include="..\..\..\..\src\mb.c" />
+    <ClCompile Include="..\..\..\..\src\mutex.c" />
+    <ClCompile Include="..\..\..\..\src\nstime.c" />
+    <ClCompile Include="..\..\..\..\src\pages.c" />
+    <ClCompile Include="..\..\..\..\src\prof.c" />
+    <ClCompile Include="..\..\..\..\src\quarantine.c" />
+    <ClCompile Include="..\..\..\..\src\rtree.c" />
+    <ClCompile Include="..\..\..\..\src\stats.c" />
+    <ClCompile Include="..\..\..\..\src\tcache.c" />
+    <ClCompile Include="..\..\..\..\src\tsd.c" />
+    <ClCompile Include="..\..\..\..\src\util.c" />
+  </ItemGroup>
+  <PropertyGroup Label="Globals">
+    <ProjectGuid>{8D6BB292-9E1C-413D-9F98-4864BDC1514A}</ProjectGuid>
+    <Keyword>Win32Proj</Keyword>
+    <RootNamespace>jemalloc</RootNamespace>
+    <WindowsTargetPlatformVersion>8.1</WindowsTargetPlatformVersion>
+  </PropertyGroup>
+  <Import Project="$(VCTargetsPath)\Microsoft.Cpp.Default.props" />
+  <PropertyGroup Condition="'$(Configuration)|$(Platform)'=='Debug|Win32'" Label="Configuration">
+    <ConfigurationType>DynamicLibrary</ConfigurationType>
+    <UseDebugLibraries>true</UseDebugLibraries>
+    <PlatformToolset>v140</PlatformToolset>
+    <CharacterSet>MultiByte</CharacterSet>
+  </PropertyGroup>
+  <PropertyGroup Condition="'$(Configuration)|$(Platform)'=='Debug-static|Win32'" Label="Configuration">
+    <ConfigurationType>StaticLibrary</ConfigurationType>
+    <UseDebugLibraries>true</UseDebugLibraries>
+    <PlatformToolset>v140</PlatformToolset>
+    <CharacterSet>MultiByte</CharacterSet>
+  </PropertyGroup>
+  <PropertyGroup Condition="'$(Configuration)|$(Platform)'=='Release|Win32'" Label="Configuration">
+    <ConfigurationType>DynamicLibrary</ConfigurationType>
+    <UseDebugLibraries>false</UseDebugLibraries>
+    <PlatformToolset>v140</PlatformToolset>
+    <WholeProgramOptimization>true</WholeProgramOptimization>
+    <CharacterSet>MultiByte</CharacterSet>
+  </PropertyGroup>
+  <PropertyGroup Condition="'$(Configuration)|$(Platform)'=='Release-static|Win32'" Label="Configuration">
+    <ConfigurationType>StaticLibrary</ConfigurationType>
+    <UseDebugLibraries>false</UseDebugLibraries>
+    <PlatformToolset>v140</PlatformToolset>
+    <WholeProgramOptimization>true</WholeProgramOptimization>
+    <CharacterSet>MultiByte</CharacterSet>
+  </PropertyGroup>
+  <PropertyGroup Condition="'$(Configuration)|$(Platform)'=='Debug|x64'" Label="Configuration">
+    <ConfigurationType>DynamicLibrary</ConfigurationType>
+    <UseDebugLibraries>true</UseDebugLibraries>
+    <PlatformToolset>v140</PlatformToolset>
+    <CharacterSet>MultiByte</CharacterSet>
+  </PropertyGroup>
+  <PropertyGroup Condition="'$(Configuration)|$(Platform)'=='Debug-static|x64'" Label="Configuration">
+    <ConfigurationType>StaticLibrary</ConfigurationType>
+    <UseDebugLibraries>true</UseDebugLibraries>
+    <PlatformToolset>v140</PlatformToolset>
+    <CharacterSet>MultiByte</CharacterSet>
+  </PropertyGroup>
+  <PropertyGroup Condition="'$(Configuration)|$(Platform)'=='Release|x64'" Label="Configuration">
+    <ConfigurationType>DynamicLibrary</ConfigurationType>
+    <UseDebugLibraries>false</UseDebugLibraries>
+    <PlatformToolset>v140</PlatformToolset>
+    <WholeProgramOptimization>true</WholeProgramOptimization>
+    <CharacterSet>MultiByte</CharacterSet>
+  </PropertyGroup>
+  <PropertyGroup Condition="'$(Configuration)|$(Platform)'=='Release-static|x64'" Label="Configuration">
+    <ConfigurationType>StaticLibrary</ConfigurationType>
+    <UseDebugLibraries>false</UseDebugLibraries>
+    <PlatformToolset>v140</PlatformToolset>
+    <WholeProgramOptimization>true</WholeProgramOptimization>
+    <CharacterSet>MultiByte</CharacterSet>
+  </PropertyGroup>
+  <Import Project="$(VCTargetsPath)\Microsoft.Cpp.props" />
+  <ImportGroup Label="ExtensionSettings">
+  </ImportGroup>
+  <ImportGroup Label="Shared">
+  </ImportGroup>
+  <ImportGroup Label="PropertySheets" Condition="'$(Configuration)|$(Platform)'=='Debug|Win32'">
+    <Import Project="$(UserRootDir)\Microsoft.Cpp.$(Platform).user.props" Condition="exists('$(UserRootDir)\Microsoft.Cpp.$(Platform).user.props')" Label="LocalAppDataPlatform" />
+  </ImportGroup>
+  <ImportGroup Condition="'$(Configuration)|$(Platform)'=='Debug-static|Win32'" Label="PropertySheets">
+    <Import Project="$(UserRootDir)\Microsoft.Cpp.$(Platform).user.props" Condition="exists('$(UserRootDir)\Microsoft.Cpp.$(Platform).user.props')" Label="LocalAppDataPlatform" />
+  </ImportGroup>
+  <ImportGroup Label="PropertySheets" Condition="'$(Configuration)|$(Platform)'=='Release|Win32'">
+    <Import Project="$(UserRootDir)\Microsoft.Cpp.$(Platform).user.props" Condition="exists('$(UserRootDir)\Microsoft.Cpp.$(Platform).user.props')" Label="LocalAppDataPlatform" />
+  </ImportGroup>
+  <ImportGroup Condition="'$(Configuration)|$(Platform)'=='Release-static|Win32'" Label="PropertySheets">
+    <Import Project="$(UserRootDir)\Microsoft.Cpp.$(Platform).user.props" Condition="exists('$(UserRootDir)\Microsoft.Cpp.$(Platform).user.props')" Label="LocalAppDataPlatform" />
+  </ImportGroup>
+  <ImportGroup Label="PropertySheets" Condition="'$(Configuration)|$(Platform)'=='Debug|x64'">
+    <Import Project="$(UserRootDir)\Microsoft.Cpp.$(Platform).user.props" Condition="exists('$(UserRootDir)\Microsoft.Cpp.$(Platform).user.props')" Label="LocalAppDataPlatform" />
+  </ImportGroup>
+  <ImportGroup Condition="'$(Configuration)|$(Platform)'=='Debug-static|x64'" Label="PropertySheets">
+    <Import Project="$(UserRootDir)\Microsoft.Cpp.$(Platform).user.props" Condition="exists('$(UserRootDir)\Microsoft.Cpp.$(Platform).user.props')" Label="LocalAppDataPlatform" />
+  </ImportGroup>
+  <ImportGroup Label="PropertySheets" Condition="'$(Configuration)|$(Platform)'=='Release|x64'">
+    <Import Project="$(UserRootDir)\Microsoft.Cpp.$(Platform).user.props" Condition="exists('$(UserRootDir)\Microsoft.Cpp.$(Platform).user.props')" Label="LocalAppDataPlatform" />
+  </ImportGroup>
+  <ImportGroup Condition="'$(Configuration)|$(Platform)'=='Release-static|x64'" Label="PropertySheets">
+    <Import Project="$(UserRootDir)\Microsoft.Cpp.$(Platform).user.props" Condition="exists('$(UserRootDir)\Microsoft.Cpp.$(Platform).user.props')" Label="LocalAppDataPlatform" />
+  </ImportGroup>
+  <PropertyGroup Label="UserMacros" />
+  <PropertyGroup Condition="'$(Configuration)|$(Platform)'=='Debug|Win32'">
+    <OutDir>$(SolutionDir)$(Platform)\$(Configuration)\</OutDir>
+    <IntDir>$(Platform)\$(Configuration)\</IntDir>
+    <TargetName>$(ProjectName)d</TargetName>
+  </PropertyGroup>
+  <PropertyGroup Condition="'$(Configuration)|$(Platform)'=='Debug-static|Win32'">
+    <OutDir>$(SolutionDir)$(Platform)\$(Configuration)\</OutDir>
+    <IntDir>$(Platform)\$(Configuration)\</IntDir>
+    <TargetName>$(ProjectName)-$(PlatformToolset)-$(Configuration)</TargetName>
+  </PropertyGroup>
+  <PropertyGroup Condition="'$(Configuration)|$(Platform)'=='Release|Win32'">
+    <OutDir>$(SolutionDir)$(Platform)\$(Configuration)\</OutDir>
+    <IntDir>$(Platform)\$(Configuration)\</IntDir>
+  </PropertyGroup>
+  <PropertyGroup Condition="'$(Configuration)|$(Platform)'=='Release-static|Win32'">
+    <OutDir>$(SolutionDir)$(Platform)\$(Configuration)\</OutDir>
+    <IntDir>$(Platform)\$(Configuration)\</IntDir>
+    <TargetName>$(ProjectName)-$(PlatformToolset)-$(Configuration)</TargetName>
+  </PropertyGroup>
+  <PropertyGroup Condition="'$(Configuration)|$(Platform)'=='Debug|x64'">
+    <OutDir>$(SolutionDir)$(Platform)\$(Configuration)\</OutDir>
+    <IntDir>$(Platform)\$(Configuration)\</IntDir>
+    <TargetName>$(ProjectName)d</TargetName>
+  </PropertyGroup>
+  <PropertyGroup Condition="'$(Configuration)|$(Platform)'=='Debug-static|x64'">
+    <OutDir>$(SolutionDir)$(Platform)\$(Configuration)\</OutDir>
+    <IntDir>$(Platform)\$(Configuration)\</IntDir>
+    <TargetName>$(ProjectName)-$(PlatformToolset)-$(Configuration)</TargetName>
+  </PropertyGroup>
+  <PropertyGroup Condition="'$(Configuration)|$(Platform)'=='Release|x64'">
+    <OutDir>$(SolutionDir)$(Platform)\$(Configuration)\</OutDir>
+    <IntDir>$(Platform)\$(Configuration)\</IntDir>
+  </PropertyGroup>
+  <PropertyGroup Condition="'$(Configuration)|$(Platform)'=='Release-static|x64'">
+    <OutDir>$(SolutionDir)$(Platform)\$(Configuration)\</OutDir>
+    <IntDir>$(Platform)\$(Configuration)\</IntDir>
+    <TargetName>$(ProjectName)-$(PlatformToolset)-$(Configuration)</TargetName>
+  </PropertyGroup>
+  <ItemDefinitionGroup Condition="'$(Configuration)|$(Platform)'=='Debug|Win32'">
+    <ClCompile>
+      <PrecompiledHeader>
+      </PrecompiledHeader>
+      <WarningLevel>Level3</WarningLevel>
+      <Optimization>Disabled</Optimization>
+      <PreprocessorDefinitions>_REENTRANT;_WINDLL;DLLEXPORT;JEMALLOC_DEBUG;_DEBUG;%(PreprocessorDefinitions)</PreprocessorDefinitions>
+      <AdditionalIncludeDirectories>..\..\..\..\include;..\..\..\..\include\msvc_compat;%(AdditionalIncludeDirectories)</AdditionalIncludeDirectories>
+      <DisableSpecificWarnings>4090;4146;4244;4267;4334</DisableSpecificWarnings>
+      <ProgramDataBaseFileName>$(OutputPath)$(TargetName).pdb</ProgramDataBaseFileName>
+    </ClCompile>
+    <Link>
+      <SubSystem>Windows</SubSystem>
+      <GenerateDebugInformation>true</GenerateDebugInformation>
+    </Link>
+  </ItemDefinitionGroup>
+  <ItemDefinitionGroup Condition="'$(Configuration)|$(Platform)'=='Debug-static|Win32'">
+    <ClCompile>
+      <PrecompiledHeader>
+      </PrecompiledHeader>
+      <WarningLevel>Level3</WarningLevel>
+      <Optimization>Disabled</Optimization>
+      <PreprocessorDefinitions>JEMALLOC_DEBUG;_REENTRANT;JEMALLOC_EXPORT=;_DEBUG;_LIB;%(PreprocessorDefinitions)</PreprocessorDefinitions>
+      <AdditionalIncludeDirectories>..\..\..\..\include;..\..\..\..\include\msvc_compat;%(AdditionalIncludeDirectories)</AdditionalIncludeDirectories>
+      <RuntimeLibrary>MultiThreadedDebug</RuntimeLibrary>
+      <DisableSpecificWarnings>4090;4146;4244;4267;4334</DisableSpecificWarnings>
+      <ProgramDataBaseFileName>$(OutputPath)$(TargetName).pdb</ProgramDataBaseFileName>
+    </ClCompile>
+    <Link>
+      <SubSystem>Windows</SubSystem>
+      <GenerateDebugInformation>true</GenerateDebugInformation>
+    </Link>
+  </ItemDefinitionGroup>
+  <ItemDefinitionGroup Condition="'$(Configuration)|$(Platform)'=='Debug|x64'">
+    <ClCompile>
+      <PrecompiledHeader>
+      </PrecompiledHeader>
+      <WarningLevel>Level3</WarningLevel>
+      <Optimization>Disabled</Optimization>
+      <PreprocessorDefinitions>_REENTRANT;_WINDLL;DLLEXPORT;JEMALLOC_DEBUG;_DEBUG;%(PreprocessorDefinitions)</PreprocessorDefinitions>
+      <AdditionalIncludeDirectories>..\..\..\..\include;..\..\..\..\include\msvc_compat;%(AdditionalIncludeDirectories)</AdditionalIncludeDirectories>
+      <DisableSpecificWarnings>4090;4146;4244;4267;4334</DisableSpecificWarnings>
+      <ProgramDataBaseFileName>$(OutputPath)$(TargetName).pdb</ProgramDataBaseFileName>
+    </ClCompile>
+    <Link>
+      <SubSystem>Windows</SubSystem>
+      <GenerateDebugInformation>true</GenerateDebugInformation>
+    </Link>
+  </ItemDefinitionGroup>
+  <ItemDefinitionGroup Condition="'$(Configuration)|$(Platform)'=='Debug-static|x64'">
+    <ClCompile>
+      <PrecompiledHeader>
+      </PrecompiledHeader>
+      <WarningLevel>Level3</WarningLevel>
+      <Optimization>Disabled</Optimization>
+      <PreprocessorDefinitions>JEMALLOC_DEBUG;_REENTRANT;JEMALLOC_EXPORT=;_DEBUG;_LIB;%(PreprocessorDefinitions)</PreprocessorDefinitions>
+      <AdditionalIncludeDirectories>..\..\..\..\include;..\..\..\..\include\msvc_compat;%(AdditionalIncludeDirectories)</AdditionalIncludeDirectories>
+      <RuntimeLibrary>MultiThreadedDebug</RuntimeLibrary>
+      <DisableSpecificWarnings>4090;4146;4244;4267;4334</DisableSpecificWarnings>
+      <ProgramDataBaseFileName>$(OutputPath)$(TargetName).pdb</ProgramDataBaseFileName>
+    </ClCompile>
+    <Link>
+      <SubSystem>Windows</SubSystem>
+      <GenerateDebugInformation>true</GenerateDebugInformation>
+    </Link>
+  </ItemDefinitionGroup>
+  <ItemDefinitionGroup Condition="'$(Configuration)|$(Platform)'=='Release|Win32'">
+    <ClCompile>
+      <WarningLevel>Level3</WarningLevel>
+      <PrecompiledHeader>
+      </PrecompiledHeader>
+      <Optimization>MaxSpeed</Optimization>
+      <FunctionLevelLinking>true</FunctionLevelLinking>
+      <IntrinsicFunctions>true</IntrinsicFunctions>
+      <PreprocessorDefinitions>_REENTRANT;_WINDLL;DLLEXPORT;NDEBUG;%(PreprocessorDefinitions)</PreprocessorDefinitions>
+      <AdditionalIncludeDirectories>..\..\..\..\include;..\..\..\..\include\msvc_compat;%(AdditionalIncludeDirectories)</AdditionalIncludeDirectories>
+      <DisableSpecificWarnings>4090;4146;4244;4267;4334</DisableSpecificWarnings>
+      <ProgramDataBaseFileName>$(OutputPath)$(TargetName).pdb</ProgramDataBaseFileName>
+    </ClCompile>
+    <Link>
+      <SubSystem>Windows</SubSystem>
+      <GenerateDebugInformation>true</GenerateDebugInformation>
+      <EnableCOMDATFolding>true</EnableCOMDATFolding>
+      <OptimizeReferences>true</OptimizeReferences>
+    </Link>
+  </ItemDefinitionGroup>
+  <ItemDefinitionGroup Condition="'$(Configuration)|$(Platform)'=='Release-static|Win32'">
+    <ClCompile>
+      <WarningLevel>Level3</WarningLevel>
+      <PrecompiledHeader>
+      </PrecompiledHeader>
+      <Optimization>MaxSpeed</Optimization>
+      <FunctionLevelLinking>true</FunctionLevelLinking>
+      <IntrinsicFunctions>true</IntrinsicFunctions>
+      <PreprocessorDefinitions>_REENTRANT;JEMALLOC_EXPORT=;NDEBUG;_LIB;%(PreprocessorDefinitions)</PreprocessorDefinitions>
+      <AdditionalIncludeDirectories>..\..\..\..\include;..\..\..\..\include\msvc_compat;%(AdditionalIncludeDirectories)</AdditionalIncludeDirectories>
+      <RuntimeLibrary>MultiThreaded</RuntimeLibrary>
+      <DisableSpecificWarnings>4090;4146;4244;4267;4334</DisableSpecificWarnings>
+      <ProgramDataBaseFileName>$(OutputPath)$(TargetName).pdb</ProgramDataBaseFileName>
+    </ClCompile>
+    <Link>
+      <SubSystem>Windows</SubSystem>
+      <GenerateDebugInformation>true</GenerateDebugInformation>
+      <EnableCOMDATFolding>true</EnableCOMDATFolding>
+      <OptimizeReferences>true</OptimizeReferences>
+    </Link>
+  </ItemDefinitionGroup>
+  <ItemDefinitionGroup Condition="'$(Configuration)|$(Platform)'=='Release|x64'">
+    <ClCompile>
+      <WarningLevel>Level3</WarningLevel>
+      <PrecompiledHeader>
+      </PrecompiledHeader>
+      <Optimization>MaxSpeed</Optimization>
+      <FunctionLevelLinking>true</FunctionLevelLinking>
+      <IntrinsicFunctions>true</IntrinsicFunctions>
+      <AdditionalIncludeDirectories>..\..\..\..\include;..\..\..\..\include\msvc_compat;%(AdditionalIncludeDirectories)</AdditionalIncludeDirectories>
+      <PreprocessorDefinitions>_REENTRANT;_WINDLL;DLLEXPORT;NDEBUG;%(PreprocessorDefinitions)</PreprocessorDefinitions>
+      <DisableSpecificWarnings>4090;4146;4244;4267;4334</DisableSpecificWarnings>
+      <ProgramDataBaseFileName>$(OutputPath)$(TargetName).pdb</ProgramDataBaseFileName>
+    </ClCompile>
+    <Link>
+      <SubSystem>Windows</SubSystem>
+      <GenerateDebugInformation>true</GenerateDebugInformation>
+      <EnableCOMDATFolding>true</EnableCOMDATFolding>
+      <OptimizeReferences>true</OptimizeReferences>
+    </Link>
+  </ItemDefinitionGroup>
+  <ItemDefinitionGroup Condition="'$(Configuration)|$(Platform)'=='Release-static|x64'">
+    <ClCompile>
+      <WarningLevel>Level3</WarningLevel>
+      <PrecompiledHeader>
+      </PrecompiledHeader>
+      <Optimization>MaxSpeed</Optimization>
+      <FunctionLevelLinking>true</FunctionLevelLinking>
+      <IntrinsicFunctions>true</IntrinsicFunctions>
+      <PreprocessorDefinitions>_REENTRANT;JEMALLOC_EXPORT=;NDEBUG;_LIB;%(PreprocessorDefinitions)</PreprocessorDefinitions>
+      <AdditionalIncludeDirectories>..\..\..\..\include;..\..\..\..\include\msvc_compat;%(AdditionalIncludeDirectories)</AdditionalIncludeDirectories>
+      <RuntimeLibrary>MultiThreaded</RuntimeLibrary>
+      <DisableSpecificWarnings>4090;4146;4244;4267;4334</DisableSpecificWarnings>
+      <ProgramDataBaseFileName>$(OutputPath)$(TargetName).pdb</ProgramDataBaseFileName>
+    </ClCompile>
+    <Link>
+      <SubSystem>Windows</SubSystem>
+      <GenerateDebugInformation>true</GenerateDebugInformation>
+      <EnableCOMDATFolding>true</EnableCOMDATFolding>
+      <OptimizeReferences>true</OptimizeReferences>
+    </Link>
+  </ItemDefinitionGroup>
+  <Import Project="$(VCTargetsPath)\Microsoft.Cpp.targets" />
+  <ImportGroup Label="ExtensionTargets">
+  </ImportGroup>
+</Project>

msvc/projects/vc2015/jemalloc/jemalloc.vcxproj.filters

@@ -0,0 +1,245 @@
+<?xml version="1.0" encoding="utf-8"?>
+<Project ToolsVersion="4.0" xmlns="http://schemas.microsoft.com/developer/msbuild/2003">
+  <ItemGroup>
+    <Filter Include="Source Files">
+      <UniqueIdentifier>{4FC737F1-C7A5-4376-A066-2A32D752A2FF}</UniqueIdentifier>
+      <Extensions>cpp;c;cc;cxx;def;odl;idl;hpj;bat;asm;asmx</Extensions>
+    </Filter>
+    <Filter Include="Header Files">
+      <UniqueIdentifier>{93995380-89BD-4b04-88EB-625FBE52EBFB}</UniqueIdentifier>
+      <Extensions>h;hh;hpp;hxx;hm;inl;inc;xsd</Extensions>
+    </Filter>
+    <Filter Include="Header Files\internal">
+      <UniqueIdentifier>{5697dfa3-16cf-4932-b428-6e0ec6e9f98e}</UniqueIdentifier>
+    </Filter>
+    <Filter Include="Header Files\msvc_compat">
+      <UniqueIdentifier>{0cbd2ca6-42a7-4f82-8517-d7e7a14fd986}</UniqueIdentifier>
+    </Filter>
+    <Filter Include="Header Files\msvc_compat\C99">
+      <UniqueIdentifier>{0abe6f30-49b5-46dd-8aca-6e33363fa52c}</UniqueIdentifier>
+    </Filter>
+  </ItemGroup>
+  <ItemGroup>
+    <ClInclude Include="..\..\..\..\include\jemalloc\jemalloc.h">
+      <Filter>Header Files</Filter>
+    </ClInclude>
+    <ClInclude Include="..\..\..\..\include\jemalloc\jemalloc_defs.h">
+      <Filter>Header Files</Filter>
+    </ClInclude>
+    <ClInclude Include="..\..\..\..\include\jemalloc\jemalloc_macros.h">
+      <Filter>Header Files</Filter>
+    </ClInclude>
+    <ClInclude Include="..\..\..\..\include\jemalloc\jemalloc_mangle.h">
+      <Filter>Header Files</Filter>
+    </ClInclude>
+    <ClInclude Include="..\..\..\..\include\jemalloc\jemalloc_protos.h">
+      <Filter>Header Files</Filter>
+    </ClInclude>
+    <ClInclude Include="..\..\..\..\include\jemalloc\jemalloc_protos_jet.h">
+      <Filter>Header Files</Filter>
+    </ClInclude>
+    <ClInclude Include="..\..\..\..\include\jemalloc\jemalloc_rename.h">
+      <Filter>Header Files</Filter>
+    </ClInclude>
+    <ClInclude Include="..\..\..\..\include\jemalloc\jemalloc_typedefs.h">
+      <Filter>Header Files</Filter>
+    </ClInclude>
+    <ClInclude Include="..\..\..\..\include\jemalloc\internal\arena.h">
+      <Filter>Header Files\internal</Filter>
+    </ClInclude>
+    <ClInclude Include="..\..\..\..\include\jemalloc\internal\assert.h">
+      <Filter>Header Files\internal</Filter>
+    </ClInclude>
+    <ClInclude Include="..\..\..\..\include\jemalloc\internal\atomic.h">
+      <Filter>Header Files\internal</Filter>
+    </ClInclude>
+    <ClInclude Include="..\..\..\..\include\jemalloc\internal\base.h">
+      <Filter>Header Files\internal</Filter>
+    </ClInclude>
+    <ClInclude Include="..\..\..\..\include\jemalloc\internal\bitmap.h">
+      <Filter>Header Files\internal</Filter>
+    </ClInclude>
+    <ClInclude Include="..\..\..\..\include\jemalloc\internal\chunk.h">
+      <Filter>Header Files\internal</Filter>
+    </ClInclude>
+    <ClInclude Include="..\..\..\..\include\jemalloc\internal\chunk_dss.h">
+      <Filter>Header Files\internal</Filter>
+    </ClInclude>
+    <ClInclude Include="..\..\..\..\include\jemalloc\internal\chunk_mmap.h">
+      <Filter>Header Files\internal</Filter>
+    </ClInclude>
+    <ClInclude Include="..\..\..\..\include\jemalloc\internal\ckh.h">
+      <Filter>Header Files\internal</Filter>
+    </ClInclude>
+    <ClInclude Include="..\..\..\..\include\jemalloc\internal\ctl.h">
+      <Filter>Header Files\internal</Filter>
+    </ClInclude>
+    <ClInclude Include="..\..\..\..\include\jemalloc\internal\extent.h">
+      <Filter>Header Files\internal</Filter>
+    </ClInclude>
+    <ClInclude Include="..\..\..\..\include\jemalloc\internal\hash.h">
+      <Filter>Header Files\internal</Filter>
+    </ClInclude>
+    <ClInclude Include="..\..\..\..\include\jemalloc\internal\huge.h">
+      <Filter>Header Files\internal</Filter>
+    </ClInclude>
+    <ClInclude Include="..\..\..\..\include\jemalloc\internal\jemalloc_internal.h">
+      <Filter>Header Files\internal</Filter>
+    </ClInclude>
+    <ClInclude Include="..\..\..\..\include\jemalloc\internal\jemalloc_internal_decls.h">
+      <Filter>Header Files\internal</Filter>
+    </ClInclude>
+    <ClInclude Include="..\..\..\..\include\jemalloc\internal\jemalloc_internal_defs.h">
+      <Filter>Header Files\internal</Filter>
+    </ClInclude>
+    <ClInclude Include="..\..\..\..\include\jemalloc\internal\jemalloc_internal_macros.h">
+      <Filter>Header Files\internal</Filter>
+    </ClInclude>
+    <ClInclude Include="..\..\..\..\include\jemalloc\internal\mb.h">
+      <Filter>Header Files\internal</Filter>
+    </ClInclude>
+    <ClInclude Include="..\..\..\..\include\jemalloc\internal\mutex.h">
+      <Filter>Header Files\internal</Filter>
+    </ClInclude>
+    <ClInclude Include="..\..\..\..\include\jemalloc\internal\pages.h">
+      <Filter>Header Files\internal</Filter>
+    </ClInclude>
+    <ClInclude Include="..\..\..\..\include\jemalloc\internal\private_namespace.h">
+      <Filter>Header Files\internal</Filter>
+    </ClInclude>
+    <ClInclude Include="..\..\..\..\include\jemalloc\internal\private_unnamespace.h">
+      <Filter>Header Files\internal</Filter>
+    </ClInclude>
+    <ClInclude Include="..\..\..\..\include\jemalloc\internal\prng.h">
+      <Filter>Header Files\internal</Filter>
+    </ClInclude>
+    <ClInclude Include="..\..\..\..\include\jemalloc\internal\prof.h">
+      <Filter>Header Files\internal</Filter>
+    </ClInclude>
+    <ClInclude Include="..\..\..\..\include\jemalloc\internal\public_namespace.h">
+      <Filter>Header Files\internal</Filter>
+    </ClInclude>
+    <ClInclude Include="..\..\..\..\include\jemalloc\internal\public_unnamespace.h">
+      <Filter>Header Files\internal</Filter>
+    </ClInclude>
+    <ClInclude Include="..\..\..\..\include\jemalloc\internal\ql.h">
+      <Filter>Header Files\internal</Filter>
+    </ClInclude>
+    <ClInclude Include="..\..\..\..\include\jemalloc\internal\qr.h">
+      <Filter>Header Files\internal</Filter>
+    </ClInclude>
+    <ClInclude Include="..\..\..\..\include\jemalloc\internal\quarantine.h">
+      <Filter>Header Files\internal</Filter>
+    </ClInclude>
+    <ClInclude Include="..\..\..\..\include\jemalloc\internal\rb.h">
+      <Filter>Header Files\internal</Filter>
+    </ClInclude>
+    <ClInclude Include="..\..\..\..\include\jemalloc\internal\rtree.h">
+      <Filter>Header Files\internal</Filter>
+    </ClInclude>
+    <ClInclude Include="..\..\..\..\include\jemalloc\internal\size_classes.h">
+      <Filter>Header Files\internal</Filter>
+    </ClInclude>
+    <ClInclude Include="..\..\..\..\include\jemalloc\internal\stats.h">
+      <Filter>Header Files\internal</Filter>
+    </ClInclude>
+    <ClInclude Include="..\..\..\..\include\jemalloc\internal\tcache.h">
+      <Filter>Header Files\internal</Filter>
+    </ClInclude>
+    <ClInclude Include="..\..\..\..\include\jemalloc\internal\tsd.h">
+      <Filter>Header Files\internal</Filter>
+    </ClInclude>
+    <ClInclude Include="..\..\..\..\include\jemalloc\internal\util.h">
+      <Filter>Header Files\internal</Filter>
+    </ClInclude>
+    <ClInclude Include="..\..\..\..\include\jemalloc\internal\valgrind.h">
+      <Filter>Header Files\internal</Filter>
+    </ClInclude>
+    <ClInclude Include="..\..\..\..\include\msvc_compat\strings.h">
+      <Filter>Header Files\msvc_compat</Filter>
+    </ClInclude>
+    <ClInclude Include="..\..\..\..\include\msvc_compat\windows_extra.h">
+      <Filter>Header Files\msvc_compat</Filter>
+    </ClInclude>
+    <ClInclude Include="..\..\..\..\include\msvc_compat\C99\stdbool.h">
+      <Filter>Header Files\msvc_compat\C99</Filter>
+    </ClInclude>
+    <ClInclude Include="..\..\..\..\include\msvc_compat\C99\stdint.h">
+      <Filter>Header Files\msvc_compat\C99</Filter>
+    </ClInclude>
+  </ItemGroup>
+  <ItemGroup>
+    <ClCompile Include="..\..\..\..\src\arena.c">
+      <Filter>Source Files</Filter>
+    </ClCompile>
+    <ClCompile Include="..\..\..\..\src\atomic.c">
+      <Filter>Source Files</Filter>
+    </ClCompile>
+    <ClCompile Include="..\..\..\..\src\base.c">
+      <Filter>Source Files</Filter>
+    </ClCompile>
+    <ClCompile Include="..\..\..\..\src\bitmap.c">
+      <Filter>Source Files</Filter>
+    </ClCompile>
+    <ClCompile Include="..\..\..\..\src\chunk.c">
+      <Filter>Source Files</Filter>
+    </ClCompile>
+    <ClCompile Include="..\..\..\..\src\chunk_dss.c">
+      <Filter>Source Files</Filter>
+    </ClCompile>
+    <ClCompile Include="..\..\..\..\src\chunk_mmap.c">
+      <Filter>Source Files</Filter>
+    </ClCompile>
+    <ClCompile Include="..\..\..\..\src\ckh.c">
+      <Filter>Source Files</Filter>
+    </ClCompile>
+    <ClCompile Include="..\..\..\..\src\ctl.c">
+      <Filter>Source Files</Filter>
+    </ClCompile>
+    <ClCompile Include="..\..\..\..\src\extent.c">
+      <Filter>Source Files</Filter>
+    </ClCompile>
+    <ClCompile Include="..\..\..\..\src\hash.c">
+      <Filter>Source Files</Filter>
+    </ClCompile>
+    <ClCompile Include="..\..\..\..\src\huge.c">
+      <Filter>Source Files</Filter>
+    </ClCompile>
+    <ClCompile Include="..\..\..\..\src\jemalloc.c">
+      <Filter>Source Files</Filter>
+    </ClCompile>
+    <ClCompile Include="..\..\..\..\src\mb.c">
+      <Filter>Source Files</Filter>
+    </ClCompile>
+    <ClCompile Include="..\..\..\..\src\mutex.c">
+      <Filter>Source Files</Filter>
+    </ClCompile>
+    <ClCompile Include="..\..\..\..\src\pages.c">
+      <Filter>Source Files</Filter>
+    </ClCompile>
+    <ClCompile Include="..\..\..\..\src\prof.c">
+      <Filter>Source Files</Filter>
+    </ClCompile>
+    <ClCompile Include="..\..\..\..\src\quarantine.c">
+      <Filter>Source Files</Filter>
+    </ClCompile>
+    <ClCompile Include="..\..\..\..\src\rtree.c">
+      <Filter>Source Files</Filter>
+    </ClCompile>
+    <ClCompile Include="..\..\..\..\src\stats.c">
+      <Filter>Source Files</Filter>
+    </ClCompile>
+    <ClCompile Include="..\..\..\..\src\tcache.c">
+      <Filter>Source Files</Filter>
+    </ClCompile>
+    <ClCompile Include="..\..\..\..\src\tsd.c">
+      <Filter>Source Files</Filter>
+    </ClCompile>
+    <ClCompile Include="..\..\..\..\src\util.c">
+      <Filter>Source Files</Filter>
+    </ClCompile>
+    <ClCompile Include="..\..\..\..\src\nstime.c">
+      <Filter>Source Files</Filter>
+    </ClCompile>
+  </ItemGroup>
+</Project>

msvc/projects/vc2015/test_threads/test_threads.cpp

@@ -0,0 +1,89 @@
+// jemalloc C++ threaded test
+// Author: Rustam Abdullaev
+// Public Domain
+
+#include <atomic>
+#include <functional>
+#include <future>
+#include <random>
+#include <thread>
+#include <vector>
+#include <stdio.h>
+#include <jemalloc/jemalloc.h>
+
+using std::vector;
+using std::thread;
+using std::uniform_int_distribution;
+using std::minstd_rand;
+
+int test_threads()
+{
+  je_malloc_conf = "narenas:3";
+  int narenas = 0;
+  size_t sz = sizeof(narenas);
+  je_mallctl("opt.narenas", &narenas, &sz, NULL, 0);
+  if (narenas != 3) {
+    printf("Error: unexpected number of arenas: %d\n", narenas);
+    return 1;
+  }
+  static const int sizes[] = { 7, 16, 32, 60, 91, 100, 120, 144, 169, 199, 255, 400, 670, 900, 917, 1025, 3333, 5190, 13131, 49192, 99999, 123123, 255265, 2333111 };
+  static const int numSizes = (int)(sizeof(sizes) / sizeof(sizes[0]));
+  vector<thread> workers;
+  static const int numThreads = narenas + 1, numAllocsMax = 25, numIter1 = 50, numIter2 = 50;
+  je_malloc_stats_print(NULL, NULL, NULL);
+  size_t allocated1;
+  size_t sz1 = sizeof(allocated1);
+  je_mallctl("stats.active", &allocated1, &sz1, NULL, 0);
+  printf("\nPress Enter to start threads...\n");
+  getchar();
+  printf("Starting %d threads x %d x %d iterations...\n", numThreads, numIter1, numIter2);
+  for (int i = 0; i < numThreads; i++) {
+    workers.emplace_back([tid=i]() {
+      uniform_int_distribution<int> sizeDist(0, numSizes - 1);
+      minstd_rand rnd(tid * 17);
+      uint8_t* ptrs[numAllocsMax];
+      int ptrsz[numAllocsMax];
+      for (int i = 0; i < numIter1; ++i) {
+        thread t([&]() {
+          for (int i = 0; i < numIter2; ++i) {
+            const int numAllocs = numAllocsMax - sizeDist(rnd);
+            for (int j = 0; j < numAllocs; j += 64) {
+              const int x = sizeDist(rnd);
+              const int sz = sizes[x];
+              ptrsz[j] = sz;
+              ptrs[j] = (uint8_t*)je_malloc(sz);
+              if (!ptrs[j]) {
+                printf("Unable to allocate %d bytes in thread %d, iter %d, alloc %d. %d\n", sz, tid, i, j, x);
+                exit(1);
+              }
+              for (int k = 0; k < sz; k++)
+                ptrs[j][k] = tid + k;
+            }
+            for (int j = 0; j < numAllocs; j += 64) {
+              for (int k = 0, sz = ptrsz[j]; k < sz; k++)
+                if (ptrs[j][k] != (uint8_t)(tid + k)) {
+                  printf("Memory error in thread %d, iter %d, alloc %d @ %d : %02X!=%02X\n", tid, i, j, k, ptrs[j][k], (uint8_t)(tid + k));
+                  exit(1);
+                }
+              je_free(ptrs[j]);
+            }
+          }
+        });
+        t.join();
+      }
+    });
+  }
+  for (thread& t : workers) {
+    t.join();
+  }
+  je_malloc_stats_print(NULL, NULL, NULL);
+  size_t allocated2;
+  je_mallctl("stats.active", &allocated2, &sz1, NULL, 0);
+  size_t leaked = allocated2 - allocated1;
+  printf("\nDone. Leaked: %zd bytes\n", leaked);
+  bool failed = leaked > 65536; // in case C++ runtime allocated something (e.g. iostream locale or facet)
+  printf("\nTest %s!\n", (failed ? "FAILED" : "successful"));
+  printf("\nPress Enter to continue...\n");
+  getchar();
+  return failed ? 1 : 0;
+}

msvc/projects/vc2015/test_threads/test_threads.h

@@ -0,0 +1,3 @@
+#pragma once
+
+int test_threads();

msvc/projects/vc2015/test_threads/test_threads.vcxproj

@@ -0,0 +1,327 @@
+<?xml version="1.0" encoding="utf-8"?>
+<Project DefaultTargets="Build" ToolsVersion="14.0" xmlns="http://schemas.microsoft.com/developer/msbuild/2003">
+  <ItemGroup Label="ProjectConfigurations">
+    <ProjectConfiguration Include="Debug-static|Win32">
+      <Configuration>Debug-static</Configuration>
+      <Platform>Win32</Platform>
+    </ProjectConfiguration>
+    <ProjectConfiguration Include="Debug-static|x64">
+      <Configuration>Debug-static</Configuration>
+      <Platform>x64</Platform>
+    </ProjectConfiguration>
+    <ProjectConfiguration Include="Debug|Win32">
+      <Configuration>Debug</Configuration>
+      <Platform>Win32</Platform>
+    </ProjectConfiguration>
+    <ProjectConfiguration Include="Release-static|Win32">
+      <Configuration>Release-static</Configuration>
+      <Platform>Win32</Platform>
+    </ProjectConfiguration>
+    <ProjectConfiguration Include="Release-static|x64">
+      <Configuration>Release-static</Configuration>
+      <Platform>x64</Platform>
+    </ProjectConfiguration>
+    <ProjectConfiguration Include="Release|Win32">
+      <Configuration>Release</Configuration>
+      <Platform>Win32</Platform>
+    </ProjectConfiguration>
+    <ProjectConfiguration Include="Debug|x64">
+      <Configuration>Debug</Configuration>
+      <Platform>x64</Platform>
+    </ProjectConfiguration>
+    <ProjectConfiguration Include="Release|x64">
+      <Configuration>Release</Configuration>
+      <Platform>x64</Platform>
+    </ProjectConfiguration>
+  </ItemGroup>
+  <PropertyGroup Label="Globals">
+    <ProjectGuid>{09028CFD-4EB7-491D-869C-0708DB97ED44}</ProjectGuid>
+    <Keyword>Win32Proj</Keyword>
+    <RootNamespace>test_threads</RootNamespace>
+    <WindowsTargetPlatformVersion>8.1</WindowsTargetPlatformVersion>
+  </PropertyGroup>
+  <Import Project="$(VCTargetsPath)\Microsoft.Cpp.Default.props" />
+  <PropertyGroup Condition="'$(Configuration)|$(Platform)'=='Debug|Win32'" Label="Configuration">
+    <ConfigurationType>Application</ConfigurationType>
+    <UseDebugLibraries>true</UseDebugLibraries>
+    <PlatformToolset>v140</PlatformToolset>
+    <CharacterSet>MultiByte</CharacterSet>
+  </PropertyGroup>
+  <PropertyGroup Condition="'$(Configuration)|$(Platform)'=='Debug-static|Win32'" Label="Configuration">
+    <ConfigurationType>Application</ConfigurationType>
+    <UseDebugLibraries>true</UseDebugLibraries>
+    <PlatformToolset>v140</PlatformToolset>
+    <CharacterSet>MultiByte</CharacterSet>
+  </PropertyGroup>
+  <PropertyGroup Condition="'$(Configuration)|$(Platform)'=='Release|Win32'" Label="Configuration">
+    <ConfigurationType>Application</ConfigurationType>
+    <UseDebugLibraries>false</UseDebugLibraries>
+    <PlatformToolset>v140</PlatformToolset>
+    <WholeProgramOptimization>true</WholeProgramOptimization>
+    <CharacterSet>MultiByte</CharacterSet>
+  </PropertyGroup>
+  <PropertyGroup Condition="'$(Configuration)|$(Platform)'=='Release-static|Win32'" Label="Configuration">
+    <ConfigurationType>Application</ConfigurationType>
+    <UseDebugLibraries>false</UseDebugLibraries>
+    <PlatformToolset>v140</PlatformToolset>
+    <WholeProgramOptimization>true</WholeProgramOptimization>
+    <CharacterSet>MultiByte</CharacterSet>
+  </PropertyGroup>
+  <PropertyGroup Condition="'$(Configuration)|$(Platform)'=='Debug|x64'" Label="Configuration">
+    <ConfigurationType>Application</ConfigurationType>
+    <UseDebugLibraries>true</UseDebugLibraries>
+    <PlatformToolset>v140</PlatformToolset>
+    <CharacterSet>MultiByte</CharacterSet>
+  </PropertyGroup>
+  <PropertyGroup Condition="'$(Configuration)|$(Platform)'=='Debug-static|x64'" Label="Configuration">
+    <ConfigurationType>Application</ConfigurationType>
+    <UseDebugLibraries>true</UseDebugLibraries>
+    <PlatformToolset>v140</PlatformToolset>
+    <CharacterSet>MultiByte</CharacterSet>
+  </PropertyGroup>
+  <PropertyGroup Condition="'$(Configuration)|$(Platform)'=='Release|x64'" Label="Configuration">
+    <ConfigurationType>Application</ConfigurationType>
+    <UseDebugLibraries>false</UseDebugLibraries>
+    <PlatformToolset>v140</PlatformToolset>
+    <WholeProgramOptimization>true</WholeProgramOptimization>
+    <CharacterSet>MultiByte</CharacterSet>
+  </PropertyGroup>
+  <PropertyGroup Condition="'$(Configuration)|$(Platform)'=='Release-static|x64'" Label="Configuration">
+    <ConfigurationType>Application</ConfigurationType>
+    <UseDebugLibraries>false</UseDebugLibraries>
+    <PlatformToolset>v140</PlatformToolset>
+    <WholeProgramOptimization>true</WholeProgramOptimization>
+    <CharacterSet>MultiByte</CharacterSet>
+  </PropertyGroup>
+  <Import Project="$(VCTargetsPath)\Microsoft.Cpp.props" />
+  <ImportGroup Label="ExtensionSettings">
+  </ImportGroup>
+  <ImportGroup Label="Shared">
+  </ImportGroup>
+  <ImportGroup Label="PropertySheets" Condition="'$(Configuration)|$(Platform)'=='Debug|Win32'">
+    <Import Project="$(UserRootDir)\Microsoft.Cpp.$(Platform).user.props" Condition="exists('$(UserRootDir)\Microsoft.Cpp.$(Platform).user.props')" Label="LocalAppDataPlatform" />
+  </ImportGroup>
+  <ImportGroup Condition="'$(Configuration)|$(Platform)'=='Debug-static|Win32'" Label="PropertySheets">
+    <Import Project="$(UserRootDir)\Microsoft.Cpp.$(Platform).user.props" Condition="exists('$(UserRootDir)\Microsoft.Cpp.$(Platform).user.props')" Label="LocalAppDataPlatform" />
+  </ImportGroup>
+  <ImportGroup Label="PropertySheets" Condition="'$(Configuration)|$(Platform)'=='Release|Win32'">
+    <Import Project="$(UserRootDir)\Microsoft.Cpp.$(Platform).user.props" Condition="exists('$(UserRootDir)\Microsoft.Cpp.$(Platform).user.props')" Label="LocalAppDataPlatform" />
+  </ImportGroup>
+  <ImportGroup Condition="'$(Configuration)|$(Platform)'=='Release-static|Win32'" Label="PropertySheets">
+    <Import Project="$(UserRootDir)\Microsoft.Cpp.$(Platform).user.props" Condition="exists('$(UserRootDir)\Microsoft.Cpp.$(Platform).user.props')" Label="LocalAppDataPlatform" />
+  </ImportGroup>
+  <ImportGroup Label="PropertySheets" Condition="'$(Configuration)|$(Platform)'=='Debug|x64'">
+    <Import Project="$(UserRootDir)\Microsoft.Cpp.$(Platform).user.props" Condition="exists('$(UserRootDir)\Microsoft.Cpp.$(Platform).user.props')" Label="LocalAppDataPlatform" />
+  </ImportGroup>
+  <ImportGroup Condition="'$(Configuration)|$(Platform)'=='Debug-static|x64'" Label="PropertySheets">
+    <Import Project="$(UserRootDir)\Microsoft.Cpp.$(Platform).user.props" Condition="exists('$(UserRootDir)\Microsoft.Cpp.$(Platform).user.props')" Label="LocalAppDataPlatform" />
+  </ImportGroup>
+  <ImportGroup Label="PropertySheets" Condition="'$(Configuration)|$(Platform)'=='Release|x64'">
+    <Import Project="$(UserRootDir)\Microsoft.Cpp.$(Platform).user.props" Condition="exists('$(UserRootDir)\Microsoft.Cpp.$(Platform).user.props')" Label="LocalAppDataPlatform" />
+  </ImportGroup>
+  <ImportGroup Condition="'$(Configuration)|$(Platform)'=='Release-static|x64'" Label="PropertySheets">
+    <Import Project="$(UserRootDir)\Microsoft.Cpp.$(Platform).user.props" Condition="exists('$(UserRootDir)\Microsoft.Cpp.$(Platform).user.props')" Label="LocalAppDataPlatform" />
+  </ImportGroup>
+  <PropertyGroup Label="UserMacros" />
+  <PropertyGroup Condition="'$(Configuration)|$(Platform)'=='Debug|Win32'">
+    <OutDir>$(SolutionDir)$(Platform)\$(Configuration)\</OutDir>
+    <IntDir>$(Platform)\$(Configuration)\</IntDir>
+    <LinkIncremental>true</LinkIncremental>
+  </PropertyGroup>
+  <PropertyGroup Condition="'$(Configuration)|$(Platform)'=='Debug-static|Win32'">
+    <OutDir>$(SolutionDir)$(Platform)\$(Configuration)\</OutDir>
+    <IntDir>$(Platform)\$(Configuration)\</IntDir>
+    <LinkIncremental>true</LinkIncremental>
+  </PropertyGroup>
+  <PropertyGroup Condition="'$(Configuration)|$(Platform)'=='Debug|x64'">
+    <LinkIncremental>true</LinkIncremental>
+    <OutDir>$(SolutionDir)$(Platform)\$(Configuration)\</OutDir>
+  </PropertyGroup>
+  <PropertyGroup Condition="'$(Configuration)|$(Platform)'=='Debug-static|x64'">
+    <LinkIncremental>true</LinkIncremental>
+    <OutDir>$(SolutionDir)$(Platform)\$(Configuration)\</OutDir>
+  </PropertyGroup>
+  <PropertyGroup Condition="'$(Configuration)|$(Platform)'=='Release|Win32'">
+    <OutDir>$(SolutionDir)$(Platform)\$(Configuration)\</OutDir>
+    <IntDir>$(Platform)\$(Configuration)\</IntDir>
+    <LinkIncremental>false</LinkIncremental>
+  </PropertyGroup>
+  <PropertyGroup Condition="'$(Configuration)|$(Platform)'=='Release-static|Win32'">
+    <OutDir>$(SolutionDir)$(Platform)\$(Configuration)\</OutDir>
+    <IntDir>$(Platform)\$(Configuration)\</IntDir>
+    <LinkIncremental>false</LinkIncremental>
+  </PropertyGroup>
+  <PropertyGroup Condition="'$(Configuration)|$(Platform)'=='Release|x64'">
+    <OutDir>$(SolutionDir)$(Platform)\$(Configuration)\</OutDir>
+    <IntDir>$(Platform)\$(Configuration)\</IntDir>
+    <LinkIncremental>false</LinkIncremental>
+  </PropertyGroup>
+  <PropertyGroup Condition="'$(Configuration)|$(Platform)'=='Release-static|x64'">
+    <OutDir>$(SolutionDir)$(Platform)\$(Configuration)\</OutDir>
+    <IntDir>$(Platform)\$(Configuration)\</IntDir>
+    <LinkIncremental>false</LinkIncremental>
+  </PropertyGroup>
+  <ItemDefinitionGroup Condition="'$(Configuration)|$(Platform)'=='Debug|Win32'">
+    <ClCompile>
+      <PrecompiledHeader>
+      </PrecompiledHeader>
+      <WarningLevel>Level3</WarningLevel>
+      <Optimization>Disabled</Optimization>
+      <PreprocessorDefinitions>WIN32;_DEBUG;_CONSOLE;%(PreprocessorDefinitions)</PreprocessorDefinitions>
+      <AdditionalIncludeDirectories>..\..\..\..\test\include;..\..\..\..\include;..\..\..\..\include\msvc_compat;%(AdditionalIncludeDirectories)</AdditionalIncludeDirectories>
+    </ClCompile>
+    <Link>
+      <SubSystem>Console</SubSystem>
+      <GenerateDebugInformation>true</GenerateDebugInformation>
+      <AdditionalLibraryDirectories>$(SolutionDir)$(Platform)\$(Configuration)</AdditionalLibraryDirectories>
+      <AdditionalDependencies>jemallocd.lib;kernel32.lib;user32.lib;gdi32.lib;winspool.lib;comdlg32.lib;advapi32.lib;shell32.lib;ole32.lib;oleaut32.lib;uuid.lib;odbc32.lib;odbccp32.lib;%(AdditionalDependencies)</AdditionalDependencies>
+    </Link>
+  </ItemDefinitionGroup>
+  <ItemDefinitionGroup Condition="'$(Configuration)|$(Platform)'=='Debug-static|Win32'">
+    <ClCompile>
+      <PrecompiledHeader>
+      </PrecompiledHeader>
+      <WarningLevel>Level3</WarningLevel>
+      <Optimization>Disabled</Optimization>
+      <PreprocessorDefinitions>JEMALLOC_EXPORT=;JEMALLOC_STATIC;_DEBUG;_CONSOLE;%(PreprocessorDefinitions)</PreprocessorDefinitions>
+      <AdditionalIncludeDirectories>..\..\..\..\test\include;..\..\..\..\include;..\..\..\..\include\msvc_compat;%(AdditionalIncludeDirectories)</AdditionalIncludeDirectories>
+      <RuntimeLibrary>MultiThreadedDebug</RuntimeLibrary>
+    </ClCompile>
+    <Link>
+      <SubSystem>Console</SubSystem>
+      <GenerateDebugInformation>true</GenerateDebugInformation>
+      <AdditionalLibraryDirectories>$(SolutionDir)$(Platform)\$(Configuration)</AdditionalLibraryDirectories>
+      <AdditionalDependencies>jemalloc-$(PlatformToolset)-$(Configuration).lib;kernel32.lib;user32.lib;gdi32.lib;winspool.lib;comdlg32.lib;advapi32.lib;shell32.lib;ole32.lib;oleaut32.lib;uuid.lib;odbc32.lib;odbccp32.lib;%(AdditionalDependencies)</AdditionalDependencies>
+    </Link>
+  </ItemDefinitionGroup>
+  <ItemDefinitionGroup Condition="'$(Configuration)|$(Platform)'=='Debug|x64'">
+    <ClCompile>
+      <PrecompiledHeader>
+      </PrecompiledHeader>
+      <WarningLevel>Level3</WarningLevel>
+      <Optimization>Disabled</Optimization>
+      <PreprocessorDefinitions>_DEBUG;%(PreprocessorDefinitions)</PreprocessorDefinitions>
+      <AdditionalIncludeDirectories>..\..\..\..\test\include;..\..\..\..\include;..\..\..\..\include\msvc_compat;%(AdditionalIncludeDirectories)</AdditionalIncludeDirectories>
+    </ClCompile>
+    <Link>
+      <SubSystem>Console</SubSystem>
+      <GenerateDebugInformation>true</GenerateDebugInformation>
+      <AdditionalDependencies>jemallocd.lib;kernel32.lib;user32.lib;gdi32.lib;winspool.lib;comdlg32.lib;advapi32.lib;shell32.lib;ole32.lib;oleaut32.lib;uuid.lib;odbc32.lib;odbccp32.lib;%(AdditionalDependencies)</AdditionalDependencies>
+      <AdditionalLibraryDirectories>$(SolutionDir)$(Platform)\$(Configuration)</AdditionalLibraryDirectories>
+    </Link>
+  </ItemDefinitionGroup>
+  <ItemDefinitionGroup Condition="'$(Configuration)|$(Platform)'=='Debug-static|x64'">
+    <ClCompile>
+      <PrecompiledHeader>
+      </PrecompiledHeader>
+      <WarningLevel>Level3</WarningLevel>
+      <Optimization>Disabled</Optimization>
+      <PreprocessorDefinitions>JEMALLOC_EXPORT=;JEMALLOC_STATIC;_DEBUG;%(PreprocessorDefinitions)</PreprocessorDefinitions>
+      <AdditionalIncludeDirectories>..\..\..\..\test\include;..\..\..\..\include;..\..\..\..\include\msvc_compat;%(AdditionalIncludeDirectories)</AdditionalIncludeDirectories>
+      <RuntimeLibrary>MultiThreadedDebug</RuntimeLibrary>
+    </ClCompile>
+    <Link>
+      <SubSystem>Console</SubSystem>
+      <GenerateDebugInformation>true</GenerateDebugInformation>
+      <AdditionalDependencies>jemalloc-$(PlatformToolset)-$(Configuration).lib;kernel32.lib;user32.lib;gdi32.lib;winspool.lib;comdlg32.lib;advapi32.lib;shell32.lib;ole32.lib;oleaut32.lib;uuid.lib;odbc32.lib;odbccp32.lib;%(AdditionalDependencies)</AdditionalDependencies>
+      <AdditionalLibraryDirectories>$(SolutionDir)$(Platform)\$(Configuration)</AdditionalLibraryDirectories>
+    </Link>
+  </ItemDefinitionGroup>
+  <ItemDefinitionGroup Condition="'$(Configuration)|$(Platform)'=='Release|Win32'">
+    <ClCompile>
+      <WarningLevel>Level3</WarningLevel>
+      <PrecompiledHeader>
+      </PrecompiledHeader>
+      <Optimization>MaxSpeed</Optimization>
+      <FunctionLevelLinking>true</FunctionLevelLinking>
+      <IntrinsicFunctions>true</IntrinsicFunctions>
+      <PreprocessorDefinitions>WIN32;NDEBUG;_CONSOLE;%(PreprocessorDefinitions)</PreprocessorDefinitions>
+      <AdditionalIncludeDirectories>..\..\..\..\test\include;..\..\..\..\include;..\..\..\..\include\msvc_compat;%(AdditionalIncludeDirectories)</AdditionalIncludeDirectories>
+    </ClCompile>
+    <Link>
+      <SubSystem>Console</SubSystem>
+      <GenerateDebugInformation>true</GenerateDebugInformation>
+      <EnableCOMDATFolding>true</EnableCOMDATFolding>
+      <OptimizeReferences>true</OptimizeReferences>
+      <AdditionalLibraryDirectories>$(SolutionDir)$(Platform)\$(Configuration)</AdditionalLibraryDirectories>
+      <AdditionalDependencies>jemalloc.lib;kernel32.lib;user32.lib;gdi32.lib;winspool.lib;comdlg32.lib;advapi32.lib;shell32.lib;ole32.lib;oleaut32.lib;uuid.lib;odbc32.lib;odbccp32.lib;%(AdditionalDependencies)</AdditionalDependencies>
+    </Link>
+  </ItemDefinitionGroup>
+  <ItemDefinitionGroup Condition="'$(Configuration)|$(Platform)'=='Release-static|Win32'">
+    <ClCompile>
+      <WarningLevel>Level3</WarningLevel>
+      <PrecompiledHeader>
+      </PrecompiledHeader>
+      <Optimization>MaxSpeed</Optimization>
+      <FunctionLevelLinking>true</FunctionLevelLinking>
+      <IntrinsicFunctions>true</IntrinsicFunctions>
+      <PreprocessorDefinitions>JEMALLOC_EXPORT=;JEMALLOC_STATIC;NDEBUG;_CONSOLE;%(PreprocessorDefinitions)</PreprocessorDefinitions>
+      <AdditionalIncludeDirectories>..\..\..\..\test\include;..\..\..\..\include;..\..\..\..\include\msvc_compat;%(AdditionalIncludeDirectories)</AdditionalIncludeDirectories>
+      <RuntimeLibrary>MultiThreaded</RuntimeLibrary>
+    </ClCompile>
+    <Link>
+      <SubSystem>Console</SubSystem>
+      <GenerateDebugInformation>true</GenerateDebugInformation>
+      <EnableCOMDATFolding>true</EnableCOMDATFolding>
+      <OptimizeReferences>true</OptimizeReferences>
+      <AdditionalLibraryDirectories>$(SolutionDir)$(Platform)\$(Configuration)</AdditionalLibraryDirectories>
+      <AdditionalDependencies>jemalloc-$(PlatformToolset)-$(Configuration).lib;kernel32.lib;user32.lib;gdi32.lib;winspool.lib;comdlg32.lib;advapi32.lib;shell32.lib;ole32.lib;oleaut32.lib;uuid.lib;odbc32.lib;odbccp32.lib;%(AdditionalDependencies)</AdditionalDependencies>
+    </Link>
+  </ItemDefinitionGroup>
+  <ItemDefinitionGroup Condition="'$(Configuration)|$(Platform)'=='Release|x64'">
+    <ClCompile>
+      <WarningLevel>Level3</WarningLevel>
+      <PrecompiledHeader>
+      </PrecompiledHeader>
+      <Optimization>MaxSpeed</Optimization>
+      <FunctionLevelLinking>true</FunctionLevelLinking>
+      <IntrinsicFunctions>true</IntrinsicFunctions>
+      <PreprocessorDefinitions>NDEBUG;_CONSOLE;%(PreprocessorDefinitions)</PreprocessorDefinitions>
+      <AdditionalIncludeDirectories>..\..\..\..\test\include;..\..\..\..\include;..\..\..\..\include\msvc_compat;%(AdditionalIncludeDirectories)</AdditionalIncludeDirectories>
+    </ClCompile>
+    <Link>
+      <SubSystem>Console</SubSystem>
+      <GenerateDebugInformation>true</GenerateDebugInformation>
+      <EnableCOMDATFolding>true</EnableCOMDATFolding>
+      <OptimizeReferences>true</OptimizeReferences>
+      <AdditionalLibraryDirectories>$(SolutionDir)$(Platform)\$(Configuration)</AdditionalLibraryDirectories>
+      <AdditionalDependencies>jemalloc.lib;kernel32.lib;user32.lib;gdi32.lib;winspool.lib;comdlg32.lib;advapi32.lib;shell32.lib;ole32.lib;oleaut32.lib;uuid.lib;odbc32.lib;odbccp32.lib;%(AdditionalDependencies)</AdditionalDependencies>
+    </Link>
+  </ItemDefinitionGroup>
+  <ItemDefinitionGroup Condition="'$(Configuration)|$(Platform)'=='Release-static|x64'">
+    <ClCompile>
+      <WarningLevel>Level3</WarningLevel>
+      <PrecompiledHeader>
+      </PrecompiledHeader>
+      <Optimization>MaxSpeed</Optimization>
+      <FunctionLevelLinking>true</FunctionLevelLinking>
+      <IntrinsicFunctions>true</IntrinsicFunctions>
+      <PreprocessorDefinitions>JEMALLOC_EXPORT=;JEMALLOC_STATIC;NDEBUG;_CONSOLE;%(PreprocessorDefinitions)</PreprocessorDefinitions>
+      <AdditionalIncludeDirectories>..\..\..\..\test\include;..\..\..\..\include;..\..\..\..\include\msvc_compat;%(AdditionalIncludeDirectories)</AdditionalIncludeDirectories>
+      <RuntimeLibrary>MultiThreaded</RuntimeLibrary>
+    </ClCompile>
+    <Link>
+      <SubSystem>Console</SubSystem>
+      <GenerateDebugInformation>true</GenerateDebugInformation>
+      <EnableCOMDATFolding>true</EnableCOMDATFolding>
+      <OptimizeReferences>true</OptimizeReferences>
+      <AdditionalLibraryDirectories>$(SolutionDir)$(Platform)\$(Configuration)</AdditionalLibraryDirectories>
+      <AdditionalDependencies>jemalloc-$(PlatformToolset)-$(Configuration).lib;kernel32.lib;user32.lib;gdi32.lib;winspool.lib;comdlg32.lib;advapi32.lib;shell32.lib;ole32.lib;oleaut32.lib;uuid.lib;odbc32.lib;odbccp32.lib;%(AdditionalDependencies)</AdditionalDependencies>
+    </Link>
+  </ItemDefinitionGroup>
+  <ItemGroup>
+    <ClCompile Include="test_threads.cpp" />
+    <ClCompile Include="test_threads_main.cpp" />
+  </ItemGroup>
+  <ItemGroup>
+    <ProjectReference Include="..\jemalloc\jemalloc.vcxproj">
+      <Project>{8d6bb292-9e1c-413d-9f98-4864bdc1514a}</Project>
+    </ProjectReference>
+  </ItemGroup>
+  <ItemGroup>
+    <ClInclude Include="test_threads.h" />
+  </ItemGroup>
+  <Import Project="$(VCTargetsPath)\Microsoft.Cpp.targets" />
+  <ImportGroup Label="ExtensionTargets">
+  </ImportGroup>
+</Project>

msvc/projects/vc2015/test_threads/test_threads.vcxproj.filters

@@ -0,0 +1,26 @@
+<?xml version="1.0" encoding="utf-8"?>
+<Project ToolsVersion="4.0" xmlns="http://schemas.microsoft.com/developer/msbuild/2003">
+  <ItemGroup>
+    <Filter Include="Source Files">
+      <UniqueIdentifier>{4FC737F1-C7A5-4376-A066-2A32D752A2FF}</UniqueIdentifier>
+      <Extensions>cpp;c;cc;cxx;def;odl;idl;hpj;bat;asm;asmx</Extensions>
+    </Filter>
+    <Filter Include="Header Files">
+      <UniqueIdentifier>{93995380-89BD-4b04-88EB-625FBE52EBFB}</UniqueIdentifier>
+      <Extensions>h;hh;hpp;hxx;hm;inl;inc;xsd</Extensions>
+    </Filter>
+  </ItemGroup>
+  <ItemGroup>
+    <ClCompile Include="test_threads.cpp">
+      <Filter>Source Files</Filter>
+    </ClCompile>
+    <ClCompile Include="test_threads_main.cpp">
+      <Filter>Source Files</Filter>
+    </ClCompile>
+  </ItemGroup>
+  <ItemGroup>
+    <ClInclude Include="test_threads.h">
+      <Filter>Header Files</Filter>
+    </ClInclude>
+  </ItemGroup>
+</Project>

msvc/projects/vc2015/test_threads/test_threads_main.cpp

@@ -0,0 +1,12 @@
+#include "test_threads.h"
+#include <future>
+#include <functional>
+#include <chrono>
+
+using namespace std::chrono_literals;
+
+int main(int argc, char** argv)
+{
+  int rc = test_threads();
+  return rc;
+}

src/bitmap.c

@@ -3,6 +3,8 @@
 
 /******************************************************************************/
 
+#ifdef USE_TREE
+
 void
 bitmap_info_init(bitmap_info_t *binfo, size_t nbits)
 {
@@ -32,20 +34,11 @@ bitmap_info_init(bitmap_info_t *binfo, size_t nbits)
 	binfo->nbits = nbits;
 }
 
-size_t
+static size_t
 bitmap_info_ngroups(const bitmap_info_t *binfo)
 {
 
-	return (binfo->levels[binfo->nlevels].group_offset << LG_SIZEOF_BITMAP);
+	return (binfo->levels[binfo->nlevels].group_offset);
-}
-
-size_t
-bitmap_size(size_t nbits)
-{
-	bitmap_info_t binfo;
-
-	bitmap_info_init(&binfo, nbits);
-	return (bitmap_info_ngroups(&binfo));
 }
 
 void
@@ -61,8 +54,7 @@ bitmap_init(bitmap_t *bitmap, const bitmap_info_t *binfo)
 	 * correspond to the first logical bit in the group, so extra bits
 	 * are the most significant bits of the last group.
 	 */
-	memset(bitmap, 0xffU, binfo->levels[binfo->nlevels].group_offset <<
+	memset(bitmap, 0xffU, bitmap_size(binfo));
-	    LG_SIZEOF_BITMAP);
 	extra = (BITMAP_GROUP_NBITS - (binfo->nbits & BITMAP_GROUP_NBITS_MASK))
 	    & BITMAP_GROUP_NBITS_MASK;
 	if (extra != 0)
@@ -76,3 +68,47 @@ bitmap_init(bitmap_t *bitmap, const bitmap_info_t *binfo)
 			bitmap[binfo->levels[i+1].group_offset - 1] >>= extra;
 	}
 }
+
+#else /* USE_TREE */
+
+void
+bitmap_info_init(bitmap_info_t *binfo, size_t nbits)
+{
+	size_t i;
+
+	assert(nbits > 0);
+	assert(nbits <= (ZU(1) << LG_BITMAP_MAXBITS));
+
+	i = nbits >> LG_BITMAP_GROUP_NBITS;
+	if (nbits % BITMAP_GROUP_NBITS != 0)
+		i++;
+	binfo->ngroups = i;
+	binfo->nbits = nbits;
+}
+
+static size_t
+bitmap_info_ngroups(const bitmap_info_t *binfo)
+{
+
+	return (binfo->ngroups);
+}
+
+void
+bitmap_init(bitmap_t *bitmap, const bitmap_info_t *binfo)
+{
+	size_t extra;
+
+	memset(bitmap, 0xffU, bitmap_size(binfo));
+	extra = (binfo->nbits % (binfo->ngroups * BITMAP_GROUP_NBITS));
+	if (extra != 0)
+		bitmap[binfo->ngroups - 1] >>= (BITMAP_GROUP_NBITS - extra);
+}
+
+#endif /* USE_TREE */
+
+size_t
+bitmap_size(const bitmap_info_t *binfo)
+{
+
+	return (bitmap_info_ngroups(binfo) << LG_SIZEOF_BITMAP);
+}

src/chunk.c

@@ -332,30 +332,20 @@ chunk_alloc_core(arena_t *arena, void *new_addr, size_t size, size_t alignment,
     bool *zero, bool *commit, dss_prec_t dss_prec)
 {
 	void *ret;
-	chunk_hooks_t chunk_hooks = CHUNK_HOOKS_INITIALIZER;
 
 	assert(size != 0);
 	assert((size & chunksize_mask) == 0);
 	assert(alignment != 0);
 	assert((alignment & chunksize_mask) == 0);
 
-	/* Retained. */
-	if ((ret = chunk_recycle(arena, &chunk_hooks,
-	    &arena->chunks_szad_retained, &arena->chunks_ad_retained, false,
-	    new_addr, size, alignment, zero, commit, true)) != NULL)
-		return (ret);
-
 	/* "primary" dss. */
 	if (have_dss && dss_prec == dss_prec_primary && (ret =
 	    chunk_alloc_dss(arena, new_addr, size, alignment, zero, commit)) !=
 	    NULL)
 		return (ret);
-	/*
+	/* mmap. */
-	 * mmap.  Requesting an address is not implemented for
+	if ((ret = chunk_alloc_mmap(new_addr, size, alignment, zero, commit)) !=
-	 * chunk_alloc_mmap(), so only call it if (new_addr == NULL).
+	    NULL)
-	 */
-	if (new_addr == NULL && (ret = chunk_alloc_mmap(size, alignment, zero,
-	    commit)) != NULL)
 		return (ret);
 	/* "secondary" dss. */
 	if (have_dss && dss_prec == dss_prec_secondary && (ret =
@@ -380,7 +370,7 @@ chunk_alloc_base(size_t size)
 	 */
 	zero = true;
 	commit = true;
-	ret = chunk_alloc_mmap(size, chunksize, &zero, &commit);
+	ret = chunk_alloc_mmap(NULL, size, chunksize, &zero, &commit);
 	if (ret == NULL)
 		return (NULL);
 	if (config_valgrind)
@@ -418,9 +408,7 @@ chunk_arena_get(unsigned arena_ind)
 {
 	arena_t *arena;
 
-	/* Dodge tsd for a0 in order to avoid bootstrapping issues. */
+	arena = arena_get(arena_ind, false);
-	arena = (arena_ind == 0) ? a0get() : arena_get(tsd_fetch(), arena_ind,
-	     false, true);
 	/*
 	 * The arena we're allocating on behalf of must have been initialized
 	 * already.
@@ -447,6 +435,21 @@ chunk_alloc_default(void *new_addr, size_t size, size_t alignment, bool *zero,
 	return (ret);
 }
 
+static void *
+chunk_alloc_retained(arena_t *arena, chunk_hooks_t *chunk_hooks, void *new_addr,
+    size_t size, size_t alignment, bool *zero, bool *commit)
+{
+
+	assert(size != 0);
+	assert((size & chunksize_mask) == 0);
+	assert(alignment != 0);
+	assert((alignment & chunksize_mask) == 0);
+
+	return (chunk_recycle(arena, chunk_hooks, &arena->chunks_szad_retained,
+	    &arena->chunks_ad_retained, false, new_addr, size, alignment, zero,
+	    commit, true));
+}
+
 void *
 chunk_alloc_wrapper(arena_t *arena, chunk_hooks_t *chunk_hooks, void *new_addr,
     size_t size, size_t alignment, bool *zero, bool *commit)
@@ -454,10 +457,16 @@ chunk_alloc_wrapper(arena_t *arena, chunk_hooks_t *chunk_hooks, void *new_addr,
 	void *ret;
 
 	chunk_hooks_assure_initialized(arena, chunk_hooks);
-	ret = chunk_hooks->alloc(new_addr, size, alignment, zero, commit,
+
-	    arena->ind);
+	ret = chunk_alloc_retained(arena, chunk_hooks, new_addr, size,
+	    alignment, zero, commit);
+	if (ret == NULL) {
+		ret = chunk_hooks->alloc(new_addr, size, alignment, zero,
+		    commit, arena->ind);
 		if (ret == NULL)
 			return (NULL);
+	}
+
 	if (config_valgrind && chunk_hooks->alloc != chunk_alloc_default)
 		JEMALLOC_VALGRIND_MAKE_MEM_UNDEFINED(ret, chunksize);
 	return (ret);
@@ -716,7 +725,7 @@ chunk_boot(void)
 	 * so pages_map will always take fast path.
 	 */
 	if (!opt_lg_chunk) {
-		opt_lg_chunk = jemalloc_ffs((int)info.dwAllocationGranularity)
+		opt_lg_chunk = ffs_u((unsigned)info.dwAllocationGranularity)
 		    - 1;
 	}
 #else
@@ -732,8 +741,8 @@ chunk_boot(void)
 
 	if (have_dss && chunk_dss_boot())
 		return (true);
-	if (rtree_new(&chunks_rtree, (ZU(1) << (LG_SIZEOF_PTR+3)) -
+	if (rtree_new(&chunks_rtree, (unsigned)((ZU(1) << (LG_SIZEOF_PTR+3)) -
-	    opt_lg_chunk, chunks_rtree_node_alloc, NULL))
+	    opt_lg_chunk), chunks_rtree_node_alloc, NULL))
 		return (true);
 
 	return (false);

src/chunk_mmap.c

@@ -32,7 +32,8 @@ chunk_alloc_mmap_slow(size_t size, size_t alignment, bool *zero, bool *commit)
 }
 
 void *
-chunk_alloc_mmap(size_t size, size_t alignment, bool *zero, bool *commit)
+chunk_alloc_mmap(void *new_addr, size_t size, size_t alignment, bool *zero,
+    bool *commit)
 {
 	void *ret;
 	size_t offset;
@@ -53,9 +54,10 @@ chunk_alloc_mmap(size_t size, size_t alignment, bool *zero, bool *commit)
 	assert(alignment != 0);
 	assert((alignment & chunksize_mask) == 0);
 
-	ret = pages_map(NULL, size);
+	ret = pages_map(new_addr, size);
-	if (ret == NULL)
+	if (ret == NULL || ret == new_addr)
-		return (NULL);
+		return (ret);
+	assert(new_addr == NULL);
 	offset = ALIGNMENT_ADDR2OFFSET(ret, alignment);
 	if (offset != 0) {
 		pages_unmap(ret, size);

src/ckh.c

@@ -99,7 +99,7 @@ ckh_try_bucket_insert(ckh_t *ckh, size_t bucket, const void *key,
 	 * Cycle through the cells in the bucket, starting at a random position.
 	 * The randomness avoids worst-case search overhead as buckets fill up.
 	 */
-	prng32(offset, LG_CKH_BUCKET_CELLS, ckh->prng_state, CKH_A, CKH_C);
+	offset = (unsigned)prng_lg_range(&ckh->prng_state, LG_CKH_BUCKET_CELLS);
 	for (i = 0; i < (ZU(1) << LG_CKH_BUCKET_CELLS); i++) {
 		cell = &ckh->tab[(bucket << LG_CKH_BUCKET_CELLS) +
 		    ((i + offset) & ((ZU(1) << LG_CKH_BUCKET_CELLS) - 1))];
@@ -141,7 +141,8 @@ ckh_evict_reloc_insert(ckh_t *ckh, size_t argbucket, void const **argkey,
 		 * were an item for which both hashes indicated the same
 		 * bucket.
 		 */
-		prng32(i, LG_CKH_BUCKET_CELLS, ckh->prng_state, CKH_A, CKH_C);
+		i = (unsigned)prng_lg_range(&ckh->prng_state,
+		    LG_CKH_BUCKET_CELLS);
 		cell = &ckh->tab[(bucket << LG_CKH_BUCKET_CELLS) + i];
 		assert(cell->key != NULL);
 
@@ -247,8 +248,7 @@ ckh_grow(tsd_t *tsd, ckh_t *ckh)
 {
 	bool ret;
 	ckhc_t *tab, *ttab;
-	size_t lg_curcells;
+	unsigned lg_prevbuckets, lg_curcells;
-	unsigned lg_prevbuckets;
 
 #ifdef CKH_COUNT
 	ckh->ngrows++;
@@ -266,7 +266,7 @@ ckh_grow(tsd_t *tsd, ckh_t *ckh)
 
 		lg_curcells++;
 		usize = sa2u(sizeof(ckhc_t) << lg_curcells, CACHELINE);
-		if (usize == 0) {
+		if (unlikely(usize == 0 || usize > HUGE_MAXCLASS)) {
 			ret = true;
 			goto label_return;
 		}
@@ -283,12 +283,12 @@ ckh_grow(tsd_t *tsd, ckh_t *ckh)
 		ckh->lg_curbuckets = lg_curcells - LG_CKH_BUCKET_CELLS;
 
 		if (!ckh_rebuild(ckh, tab)) {
-			idalloctm(tsd, tab, tcache_get(tsd, false), true);
+			idalloctm(tsd, tab, tcache_get(tsd, false), true, true);
 			break;
 		}
 
 		/* Rebuilding failed, so back out partially rebuilt table. */
-		idalloctm(tsd, ckh->tab, tcache_get(tsd, false), true);
+		idalloctm(tsd, ckh->tab, tcache_get(tsd, false), true, true);
 		ckh->tab = tab;
 		ckh->lg_curbuckets = lg_prevbuckets;
 	}
@@ -302,8 +302,8 @@ static void
 ckh_shrink(tsd_t *tsd, ckh_t *ckh)
 {
 	ckhc_t *tab, *ttab;
-	size_t lg_curcells, usize;
+	size_t usize;
-	unsigned lg_prevbuckets;
+	unsigned lg_prevbuckets, lg_curcells;
 
 	/*
 	 * It is possible (though unlikely, given well behaved hashes) that the
@@ -312,7 +312,7 @@ ckh_shrink(tsd_t *tsd, ckh_t *ckh)
 	lg_prevbuckets = ckh->lg_curbuckets;
 	lg_curcells = ckh->lg_curbuckets + LG_CKH_BUCKET_CELLS - 1;
 	usize = sa2u(sizeof(ckhc_t) << lg_curcells, CACHELINE);
-	if (usize == 0)
+	if (unlikely(usize == 0 || usize > HUGE_MAXCLASS))
 		return;
 	tab = (ckhc_t *)ipallocztm(tsd, usize, CACHELINE, true, NULL, true,
 	    NULL);
@@ -330,7 +330,7 @@ ckh_shrink(tsd_t *tsd, ckh_t *ckh)
 	ckh->lg_curbuckets = lg_curcells - LG_CKH_BUCKET_CELLS;
 
 	if (!ckh_rebuild(ckh, tab)) {
-		idalloctm(tsd, tab, tcache_get(tsd, false), true);
+		idalloctm(tsd, tab, tcache_get(tsd, false), true, true);
 #ifdef CKH_COUNT
 		ckh->nshrinks++;
 #endif
@@ -338,7 +338,7 @@ ckh_shrink(tsd_t *tsd, ckh_t *ckh)
 	}
 
 	/* Rebuilding failed, so back out partially rebuilt table. */
-	idalloctm(tsd, ckh->tab, tcache_get(tsd, false), true);
+	idalloctm(tsd, ckh->tab, tcache_get(tsd, false), true, true);
 	ckh->tab = tab;
 	ckh->lg_curbuckets = lg_prevbuckets;
 #ifdef CKH_COUNT
@@ -387,7 +387,7 @@ ckh_new(tsd_t *tsd, ckh_t *ckh, size_t minitems, ckh_hash_t *hash,
 	ckh->keycomp = keycomp;
 
 	usize = sa2u(sizeof(ckhc_t) << lg_mincells, CACHELINE);
-	if (usize == 0) {
+	if (unlikely(usize == 0 || usize > HUGE_MAXCLASS)) {
 		ret = true;
 		goto label_return;
 	}
@@ -421,7 +421,7 @@ ckh_delete(tsd_t *tsd, ckh_t *ckh)
 	    (unsigned long long)ckh->nrelocs);
 #endif
 
-	idalloctm(tsd, ckh->tab, tcache_get(tsd, false), true);
+	idalloctm(tsd, ckh->tab, tcache_get(tsd, false), true, true);
 	if (config_debug)
 		memset(ckh, 0x5a, sizeof(ckh_t));
 }

src/ctl.c

@@ -24,7 +24,7 @@ ctl_named_node(const ctl_node_t *node)
 }
 
 JEMALLOC_INLINE_C const ctl_named_node_t *
-ctl_named_children(const ctl_named_node_t *node, int index)
+ctl_named_children(const ctl_named_node_t *node, size_t index)
 {
 	const ctl_named_node_t *children = ctl_named_node(node->children);
 
@@ -77,6 +77,7 @@ CTL_PROTO(config_cache_oblivious)
 CTL_PROTO(config_debug)
 CTL_PROTO(config_fill)
 CTL_PROTO(config_lazy_lock)
+CTL_PROTO(config_malloc_conf)
 CTL_PROTO(config_munmap)
 CTL_PROTO(config_prof)
 CTL_PROTO(config_prof_libgcc)
@@ -91,7 +92,9 @@ CTL_PROTO(opt_abort)
 CTL_PROTO(opt_dss)
 CTL_PROTO(opt_lg_chunk)
 CTL_PROTO(opt_narenas)
+CTL_PROTO(opt_purge)
 CTL_PROTO(opt_lg_dirty_mult)
+CTL_PROTO(opt_decay_time)
 CTL_PROTO(opt_stats_print)
 CTL_PROTO(opt_junk)
 CTL_PROTO(opt_zero)
@@ -114,10 +117,12 @@ CTL_PROTO(opt_prof_accum)
 CTL_PROTO(tcache_create)
 CTL_PROTO(tcache_flush)
 CTL_PROTO(tcache_destroy)
+static void	arena_i_purge(unsigned arena_ind, bool all);
 CTL_PROTO(arena_i_purge)
-static void	arena_purge(unsigned arena_ind);
+CTL_PROTO(arena_i_decay)
 CTL_PROTO(arena_i_dss)
 CTL_PROTO(arena_i_lg_dirty_mult)
+CTL_PROTO(arena_i_decay_time)
 CTL_PROTO(arena_i_chunk_hooks)
 INDEX_PROTO(arena_i)
 CTL_PROTO(arenas_bin_i_size)
@@ -131,6 +136,7 @@ INDEX_PROTO(arenas_hchunk_i)
 CTL_PROTO(arenas_narenas)
 CTL_PROTO(arenas_initialized)
 CTL_PROTO(arenas_lg_dirty_mult)
+CTL_PROTO(arenas_decay_time)
 CTL_PROTO(arenas_quantum)
 CTL_PROTO(arenas_page)
 CTL_PROTO(arenas_tcache_max)
@@ -181,6 +187,7 @@ INDEX_PROTO(stats_arenas_i_hchunks_j)
 CTL_PROTO(stats_arenas_i_nthreads)
 CTL_PROTO(stats_arenas_i_dss)
 CTL_PROTO(stats_arenas_i_lg_dirty_mult)
+CTL_PROTO(stats_arenas_i_decay_time)
 CTL_PROTO(stats_arenas_i_pactive)
 CTL_PROTO(stats_arenas_i_pdirty)
 CTL_PROTO(stats_arenas_i_mapped)
@@ -241,6 +248,7 @@ static const ctl_named_node_t	config_node[] = {
 	{NAME("debug"),		CTL(config_debug)},
 	{NAME("fill"),		CTL(config_fill)},
 	{NAME("lazy_lock"),	CTL(config_lazy_lock)},
+	{NAME("malloc_conf"),	CTL(config_malloc_conf)},
 	{NAME("munmap"),	CTL(config_munmap)},
 	{NAME("prof"),		CTL(config_prof)},
 	{NAME("prof_libgcc"),	CTL(config_prof_libgcc)},
@@ -258,7 +266,9 @@ static const ctl_named_node_t opt_node[] = {
 	{NAME("dss"),		CTL(opt_dss)},
 	{NAME("lg_chunk"),	CTL(opt_lg_chunk)},
 	{NAME("narenas"),	CTL(opt_narenas)},
+	{NAME("purge"),		CTL(opt_purge)},
 	{NAME("lg_dirty_mult"),	CTL(opt_lg_dirty_mult)},
+	{NAME("decay_time"),	CTL(opt_decay_time)},
 	{NAME("stats_print"),	CTL(opt_stats_print)},
 	{NAME("junk"),		CTL(opt_junk)},
 	{NAME("zero"),		CTL(opt_zero)},
@@ -288,8 +298,10 @@ static const ctl_named_node_t	tcache_node[] = {
 
 static const ctl_named_node_t arena_i_node[] = {
 	{NAME("purge"),		CTL(arena_i_purge)},
+	{NAME("decay"),		CTL(arena_i_decay)},
 	{NAME("dss"),		CTL(arena_i_dss)},
 	{NAME("lg_dirty_mult"),	CTL(arena_i_lg_dirty_mult)},
+	{NAME("decay_time"),	CTL(arena_i_decay_time)},
 	{NAME("chunk_hooks"),	CTL(arena_i_chunk_hooks)}
 };
 static const ctl_named_node_t super_arena_i_node[] = {
@@ -339,6 +351,7 @@ static const ctl_named_node_t arenas_node[] = {
 	{NAME("narenas"),	CTL(arenas_narenas)},
 	{NAME("initialized"),	CTL(arenas_initialized)},
 	{NAME("lg_dirty_mult"),	CTL(arenas_lg_dirty_mult)},
+	{NAME("decay_time"),	CTL(arenas_decay_time)},
 	{NAME("quantum"),	CTL(arenas_quantum)},
 	{NAME("page"),		CTL(arenas_page)},
 	{NAME("tcache_max"),	CTL(arenas_tcache_max)},
@@ -439,6 +452,7 @@ static const ctl_named_node_t stats_arenas_i_node[] = {
 	{NAME("nthreads"),	CTL(stats_arenas_i_nthreads)},
 	{NAME("dss"),		CTL(stats_arenas_i_dss)},
 	{NAME("lg_dirty_mult"),	CTL(stats_arenas_i_lg_dirty_mult)},
+	{NAME("decay_time"),	CTL(stats_arenas_i_decay_time)},
 	{NAME("pactive"),	CTL(stats_arenas_i_pactive)},
 	{NAME("pdirty"),	CTL(stats_arenas_i_pdirty)},
 	{NAME("mapped"),	CTL(stats_arenas_i_mapped)},
@@ -519,8 +533,10 @@ static void
 ctl_arena_clear(ctl_arena_stats_t *astats)
 {
 
+	astats->nthreads = 0;
 	astats->dss = dss_prec_names[dss_prec_limit];
 	astats->lg_dirty_mult = -1;
+	astats->decay_time = -1;
 	astats->pactive = 0;
 	astats->pdirty = 0;
 	if (config_stats) {
@@ -542,9 +558,11 @@ ctl_arena_stats_amerge(ctl_arena_stats_t *cstats, arena_t *arena)
 {
 	unsigned i;
 
-	arena_stats_merge(arena, &cstats->dss, &cstats->lg_dirty_mult,
+	if (config_stats) {
-	    &cstats->pactive, &cstats->pdirty, &cstats->astats, cstats->bstats,
+		arena_stats_merge(arena, &cstats->nthreads, &cstats->dss,
-	    cstats->lstats, cstats->hstats);
+		    &cstats->lg_dirty_mult, &cstats->decay_time,
+		    &cstats->pactive, &cstats->pdirty, &cstats->astats,
+		    cstats->bstats, cstats->lstats, cstats->hstats);
 
 		for (i = 0; i < NBINS; i++) {
 			cstats->allocated_small += cstats->bstats[i].curregs *
@@ -553,6 +571,11 @@ ctl_arena_stats_amerge(ctl_arena_stats_t *cstats, arena_t *arena)
 			cstats->ndalloc_small += cstats->bstats[i].ndalloc;
 			cstats->nrequests_small += cstats->bstats[i].nrequests;
 		}
+	} else {
+		arena_basic_stats_merge(arena, &cstats->nthreads, &cstats->dss,
+		    &cstats->lg_dirty_mult, &cstats->decay_time,
+		    &cstats->pactive, &cstats->pdirty);
+	}
 }
 
 static void
@@ -560,26 +583,32 @@ ctl_arena_stats_smerge(ctl_arena_stats_t *sstats, ctl_arena_stats_t *astats)
 {
 	unsigned i;
 
+	sstats->nthreads += astats->nthreads;
 	sstats->pactive += astats->pactive;
 	sstats->pdirty += astats->pdirty;
 
+	if (config_stats) {
 		sstats->astats.mapped += astats->astats.mapped;
 		sstats->astats.npurge += astats->astats.npurge;
 		sstats->astats.nmadvise += astats->astats.nmadvise;
 		sstats->astats.purged += astats->astats.purged;
 
-	sstats->astats.metadata_mapped += astats->astats.metadata_mapped;
+		sstats->astats.metadata_mapped +=
-	sstats->astats.metadata_allocated += astats->astats.metadata_allocated;
+		    astats->astats.metadata_mapped;
+		sstats->astats.metadata_allocated +=
+		    astats->astats.metadata_allocated;
 
 		sstats->allocated_small += astats->allocated_small;
 		sstats->nmalloc_small += astats->nmalloc_small;
 		sstats->ndalloc_small += astats->ndalloc_small;
 		sstats->nrequests_small += astats->nrequests_small;
 
-	sstats->astats.allocated_large += astats->astats.allocated_large;
+		sstats->astats.allocated_large +=
+		    astats->astats.allocated_large;
 		sstats->astats.nmalloc_large += astats->astats.nmalloc_large;
 		sstats->astats.ndalloc_large += astats->astats.ndalloc_large;
-	sstats->astats.nrequests_large += astats->astats.nrequests_large;
+		sstats->astats.nrequests_large +=
+		    astats->astats.nrequests_large;
 
 		sstats->astats.allocated_huge += astats->astats.allocated_huge;
 		sstats->astats.nmalloc_huge += astats->astats.nmalloc_huge;
@@ -588,10 +617,12 @@ ctl_arena_stats_smerge(ctl_arena_stats_t *sstats, ctl_arena_stats_t *astats)
 		for (i = 0; i < NBINS; i++) {
 			sstats->bstats[i].nmalloc += astats->bstats[i].nmalloc;
 			sstats->bstats[i].ndalloc += astats->bstats[i].ndalloc;
-		sstats->bstats[i].nrequests += astats->bstats[i].nrequests;
+			sstats->bstats[i].nrequests +=
+			    astats->bstats[i].nrequests;
 			sstats->bstats[i].curregs += astats->bstats[i].curregs;
 			if (config_tcache) {
-			sstats->bstats[i].nfills += astats->bstats[i].nfills;
+				sstats->bstats[i].nfills +=
+				    astats->bstats[i].nfills;
 				sstats->bstats[i].nflushes +=
 				    astats->bstats[i].nflushes;
 			}
@@ -603,14 +634,17 @@ ctl_arena_stats_smerge(ctl_arena_stats_t *sstats, ctl_arena_stats_t *astats)
 		for (i = 0; i < nlclasses; i++) {
 			sstats->lstats[i].nmalloc += astats->lstats[i].nmalloc;
 			sstats->lstats[i].ndalloc += astats->lstats[i].ndalloc;
-		sstats->lstats[i].nrequests += astats->lstats[i].nrequests;
+			sstats->lstats[i].nrequests +=
+			    astats->lstats[i].nrequests;
 			sstats->lstats[i].curruns += astats->lstats[i].curruns;
 		}
 
 		for (i = 0; i < nhclasses; i++) {
 			sstats->hstats[i].nmalloc += astats->hstats[i].nmalloc;
 			sstats->hstats[i].ndalloc += astats->hstats[i].ndalloc;
-		sstats->hstats[i].curhchunks += astats->hstats[i].curhchunks;
+			sstats->hstats[i].curhchunks +=
+			    astats->hstats[i].curhchunks;
+		}
 	}
 }
 
@@ -621,19 +655,9 @@ ctl_arena_refresh(arena_t *arena, unsigned i)
 	ctl_arena_stats_t *sstats = &ctl_stats.arenas[ctl_stats.narenas];
 
 	ctl_arena_clear(astats);
-
-	sstats->nthreads += astats->nthreads;
-	if (config_stats) {
 	ctl_arena_stats_amerge(astats, arena);
 	/* Merge into sum stats as well. */
 	ctl_arena_stats_smerge(sstats, astats);
-	} else {
-		astats->pactive += arena->nactive;
-		astats->pdirty += arena->ndirty;
-		/* Merge into sum stats as well. */
-		sstats->pactive += arena->nactive;
-		sstats->pdirty += arena->ndirty;
-	}
 }
 
 static bool
@@ -679,33 +703,17 @@ ctl_grow(void)
 static void
 ctl_refresh(void)
 {
-	tsd_t *tsd;
 	unsigned i;
-	bool refreshed;
 	VARIABLE_ARRAY(arena_t *, tarenas, ctl_stats.narenas);
 
 	/*
 	 * Clear sum stats, since they will be merged into by
 	 * ctl_arena_refresh().
 	 */
-	ctl_stats.arenas[ctl_stats.narenas].nthreads = 0;
 	ctl_arena_clear(&ctl_stats.arenas[ctl_stats.narenas]);
 
-	tsd = tsd_fetch();
+	for (i = 0; i < ctl_stats.narenas; i++)
-	for (i = 0, refreshed = false; i < ctl_stats.narenas; i++) {
+		tarenas[i] = arena_get(i, false);
-		tarenas[i] = arena_get(tsd, i, false, false);
-		if (tarenas[i] == NULL && !refreshed) {
-			tarenas[i] = arena_get(tsd, i, false, true);
-			refreshed = true;
-		}
-	}
-
-	for (i = 0; i < ctl_stats.narenas; i++) {
-		if (tarenas[i] != NULL)
-			ctl_stats.arenas[i].nthreads = arena_nbound(i);
-		else
-			ctl_stats.arenas[i].nthreads = 0;
-	}
 
 	for (i = 0; i < ctl_stats.narenas; i++) {
 		bool initialized = (tarenas[i] != NULL);
@@ -960,7 +968,7 @@ ctl_bymib(const size_t *mib, size_t miblen, void *oldp, size_t *oldlenp,
 		assert(node->nchildren > 0);
 		if (ctl_named_node(node->children) != NULL) {
 			/* Children are named. */
-			if (node->nchildren <= mib[i]) {
+			if (node->nchildren <= (unsigned)mib[i]) {
 				ret = ENOENT;
 				goto label_return;
 			}
@@ -1199,17 +1207,17 @@ label_return:								\
 	return (ret);							\
 }
 
-#define	CTL_RO_BOOL_CONFIG_GEN(n)					\
+#define	CTL_RO_CONFIG_GEN(n, t)						\
 static int								\
 n##_ctl(const size_t *mib, size_t miblen, void *oldp, size_t *oldlenp,	\
     void *newp, size_t newlen)						\
 {									\
 	int ret;							\
-	bool oldval;							\
+	t oldval;							\
 									\
 	READONLY();							\
 	oldval = n;							\
-	READ(oldval, bool);						\
+	READ(oldval, t);						\
 									\
 	ret = 0;							\
 label_return:								\
@@ -1241,28 +1249,31 @@ label_return:
 
 /******************************************************************************/
 
-CTL_RO_BOOL_CONFIG_GEN(config_cache_oblivious)
+CTL_RO_CONFIG_GEN(config_cache_oblivious, bool)
-CTL_RO_BOOL_CONFIG_GEN(config_debug)
+CTL_RO_CONFIG_GEN(config_debug, bool)
-CTL_RO_BOOL_CONFIG_GEN(config_fill)
+CTL_RO_CONFIG_GEN(config_fill, bool)
-CTL_RO_BOOL_CONFIG_GEN(config_lazy_lock)
+CTL_RO_CONFIG_GEN(config_lazy_lock, bool)
-CTL_RO_BOOL_CONFIG_GEN(config_munmap)
+CTL_RO_CONFIG_GEN(config_malloc_conf, const char *)
-CTL_RO_BOOL_CONFIG_GEN(config_prof)
+CTL_RO_CONFIG_GEN(config_munmap, bool)
-CTL_RO_BOOL_CONFIG_GEN(config_prof_libgcc)
+CTL_RO_CONFIG_GEN(config_prof, bool)
-CTL_RO_BOOL_CONFIG_GEN(config_prof_libunwind)
+CTL_RO_CONFIG_GEN(config_prof_libgcc, bool)
-CTL_RO_BOOL_CONFIG_GEN(config_stats)
+CTL_RO_CONFIG_GEN(config_prof_libunwind, bool)
-CTL_RO_BOOL_CONFIG_GEN(config_tcache)
+CTL_RO_CONFIG_GEN(config_stats, bool)
-CTL_RO_BOOL_CONFIG_GEN(config_tls)
+CTL_RO_CONFIG_GEN(config_tcache, bool)
-CTL_RO_BOOL_CONFIG_GEN(config_utrace)
+CTL_RO_CONFIG_GEN(config_tls, bool)
-CTL_RO_BOOL_CONFIG_GEN(config_valgrind)
+CTL_RO_CONFIG_GEN(config_utrace, bool)
-CTL_RO_BOOL_CONFIG_GEN(config_xmalloc)
+CTL_RO_CONFIG_GEN(config_valgrind, bool)
+CTL_RO_CONFIG_GEN(config_xmalloc, bool)
 
 /******************************************************************************/
 
 CTL_RO_NL_GEN(opt_abort, opt_abort, bool)
 CTL_RO_NL_GEN(opt_dss, opt_dss, const char *)
 CTL_RO_NL_GEN(opt_lg_chunk, opt_lg_chunk, size_t)
-CTL_RO_NL_GEN(opt_narenas, opt_narenas, size_t)
+CTL_RO_NL_GEN(opt_narenas, opt_narenas, unsigned)
+CTL_RO_NL_GEN(opt_purge, purge_mode_names[opt_purge], const char *)
 CTL_RO_NL_GEN(opt_lg_dirty_mult, opt_lg_dirty_mult, ssize_t)
+CTL_RO_NL_GEN(opt_decay_time, opt_decay_time, ssize_t)
 CTL_RO_NL_GEN(opt_stats_print, opt_stats_print, bool)
 CTL_RO_NL_CGEN(config_fill, opt_junk, opt_junk, const char *)
 CTL_RO_NL_CGEN(config_fill, opt_quarantine, opt_quarantine, size_t)
@@ -1314,7 +1325,7 @@ thread_arena_ctl(const size_t *mib, size_t miblen, void *oldp, size_t *oldlenp,
 		}
 
 		/* Initialize arena if necessary. */
-		newarena = arena_get(tsd, newind, true, true);
+		newarena = arena_get(newind, true);
 		if (newarena == NULL) {
 			ret = EAGAIN;
 			goto label_return;
@@ -1536,34 +1547,44 @@ label_return:
 
 /******************************************************************************/
 
-/* ctl_mutex must be held during execution of this function. */
 static void
-arena_purge(unsigned arena_ind)
+arena_i_purge(unsigned arena_ind, bool all)
 {
-	tsd_t *tsd;
-	unsigned i;
-	bool refreshed;
-	VARIABLE_ARRAY(arena_t *, tarenas, ctl_stats.narenas);
 
-	tsd = tsd_fetch();
+	malloc_mutex_lock(&ctl_mtx);
-	for (i = 0, refreshed = false; i < ctl_stats.narenas; i++) {
+	{
-		tarenas[i] = arena_get(tsd, i, false, false);
+		unsigned narenas = ctl_stats.narenas;
-		if (tarenas[i] == NULL && !refreshed) {
-			tarenas[i] = arena_get(tsd, i, false, true);
-			refreshed = true;
-		}
-	}
 
-	if (arena_ind == ctl_stats.narenas) {
+		if (arena_ind == narenas) {
 			unsigned i;
-		for (i = 0; i < ctl_stats.narenas; i++) {
+			VARIABLE_ARRAY(arena_t *, tarenas, narenas);
+
+			for (i = 0; i < narenas; i++)
+				tarenas[i] = arena_get(i, false);
+
+			/*
+			 * No further need to hold ctl_mtx, since narenas and
+			 * tarenas contain everything needed below.
+			 */
+			malloc_mutex_unlock(&ctl_mtx);
+
+			for (i = 0; i < narenas; i++) {
 				if (tarenas[i] != NULL)
-				arena_purge_all(tarenas[i]);
+					arena_purge(tarenas[i], all);
 			}
 		} else {
-		assert(arena_ind < ctl_stats.narenas);
+			arena_t *tarena;
-		if (tarenas[arena_ind] != NULL)
+
-			arena_purge_all(tarenas[arena_ind]);
+			assert(arena_ind < narenas);
+
+			tarena = arena_get(arena_ind, false);
+
+			/* No further need to hold ctl_mtx. */
+			malloc_mutex_unlock(&ctl_mtx);
+
+			if (tarena != NULL)
+				arena_purge(tarena, all);
+		}
 	}
 }
 
@@ -1575,9 +1596,22 @@ arena_i_purge_ctl(const size_t *mib, size_t miblen, void *oldp, size_t *oldlenp,
 
 	READONLY();
 	WRITEONLY();
-	malloc_mutex_lock(&ctl_mtx);
+	arena_i_purge((unsigned)mib[1], true);
-	arena_purge(mib[1]);
+
-	malloc_mutex_unlock(&ctl_mtx);
+	ret = 0;
+label_return:
+	return (ret);
+}
+
+static int
+arena_i_decay_ctl(const size_t *mib, size_t miblen, void *oldp, size_t *oldlenp,
+    void *newp, size_t newlen)
+{
+	int ret;
+
+	READONLY();
+	WRITEONLY();
+	arena_i_purge((unsigned)mib[1], false);
 
 	ret = 0;
 label_return:
@@ -1590,7 +1624,7 @@ arena_i_dss_ctl(const size_t *mib, size_t miblen, void *oldp, size_t *oldlenp,
 {
 	int ret;
 	const char *dss = NULL;
-	unsigned arena_ind = mib[1];
+	unsigned arena_ind = (unsigned)mib[1];
 	dss_prec_t dss_prec_old = dss_prec_limit;
 	dss_prec_t dss_prec = dss_prec_limit;
 
@@ -1615,7 +1649,7 @@ arena_i_dss_ctl(const size_t *mib, size_t miblen, void *oldp, size_t *oldlenp,
 	}
 
 	if (arena_ind < ctl_stats.narenas) {
-		arena_t *arena = arena_get(tsd_fetch(), arena_ind, false, true);
+		arena_t *arena = arena_get(arena_ind, false);
 		if (arena == NULL || (dss_prec != dss_prec_limit &&
 		    arena_dss_prec_set(arena, dss_prec))) {
 			ret = EFAULT;
@@ -1645,10 +1679,10 @@ arena_i_lg_dirty_mult_ctl(const size_t *mib, size_t miblen, void *oldp,
     size_t *oldlenp, void *newp, size_t newlen)
 {
 	int ret;
-	unsigned arena_ind = mib[1];
+	unsigned arena_ind = (unsigned)mib[1];
 	arena_t *arena;
 
-	arena = arena_get(tsd_fetch(), arena_ind, false, true);
+	arena = arena_get(arena_ind, false);
 	if (arena == NULL) {
 		ret = EFAULT;
 		goto label_return;
@@ -1674,17 +1708,51 @@ label_return:
 	return (ret);
 }
 
+static int
+arena_i_decay_time_ctl(const size_t *mib, size_t miblen, void *oldp,
+    size_t *oldlenp, void *newp, size_t newlen)
+{
+	int ret;
+	unsigned arena_ind = (unsigned)mib[1];
+	arena_t *arena;
+
+	arena = arena_get(arena_ind, false);
+	if (arena == NULL) {
+		ret = EFAULT;
+		goto label_return;
+	}
+
+	if (oldp != NULL && oldlenp != NULL) {
+		size_t oldval = arena_decay_time_get(arena);
+		READ(oldval, ssize_t);
+	}
+	if (newp != NULL) {
+		if (newlen != sizeof(ssize_t)) {
+			ret = EINVAL;
+			goto label_return;
+		}
+		if (arena_decay_time_set(arena, *(ssize_t *)newp)) {
+			ret = EFAULT;
+			goto label_return;
+		}
+	}
+
+	ret = 0;
+label_return:
+	return (ret);
+}
+
 static int
 arena_i_chunk_hooks_ctl(const size_t *mib, size_t miblen, void *oldp,
     size_t *oldlenp, void *newp, size_t newlen)
 {
 	int ret;
-	unsigned arena_ind = mib[1];
+	unsigned arena_ind = (unsigned)mib[1];
 	arena_t *arena;
 
 	malloc_mutex_lock(&ctl_mtx);
 	if (arena_ind < narenas_total_get() && (arena =
-	    arena_get(tsd_fetch(), arena_ind, false, true)) != NULL) {
+	    arena_get(arena_ind, false)) != NULL) {
 		if (newp != NULL) {
 			chunk_hooks_t old_chunk_hooks, new_chunk_hooks;
 			WRITE(new_chunk_hooks, chunk_hooks_t);
@@ -1758,7 +1826,7 @@ arenas_initialized_ctl(const size_t *mib, size_t miblen, void *oldp,
 	if (*oldlenp != ctl_stats.narenas * sizeof(bool)) {
 		ret = EINVAL;
 		nread = (*oldlenp < ctl_stats.narenas * sizeof(bool))
-		    ? (*oldlenp / sizeof(bool)) : ctl_stats.narenas;
+		    ? (unsigned)(*oldlenp / sizeof(bool)) : ctl_stats.narenas;
 	} else {
 		ret = 0;
 		nread = ctl_stats.narenas;
@@ -1798,6 +1866,32 @@ label_return:
 	return (ret);
 }
 
+static int
+arenas_decay_time_ctl(const size_t *mib, size_t miblen, void *oldp,
+    size_t *oldlenp, void *newp, size_t newlen)
+{
+	int ret;
+
+	if (oldp != NULL && oldlenp != NULL) {
+		size_t oldval = arena_decay_time_default_get();
+		READ(oldval, ssize_t);
+	}
+	if (newp != NULL) {
+		if (newlen != sizeof(ssize_t)) {
+			ret = EINVAL;
+			goto label_return;
+		}
+		if (arena_decay_time_default_set(*(ssize_t *)newp)) {
+			ret = EFAULT;
+			goto label_return;
+		}
+	}
+
+	ret = 0;
+label_return:
+	return (ret);
+}
+
 CTL_RO_NL_GEN(arenas_quantum, QUANTUM, size_t)
 CTL_RO_NL_GEN(arenas_page, PAGE, size_t)
 CTL_RO_NL_CGEN(config_tcache, arenas_tcache_max, tcache_maxclass, size_t)
@@ -1816,7 +1910,7 @@ arenas_bin_i_index(const size_t *mib, size_t miblen, size_t i)
 }
 
 CTL_RO_NL_GEN(arenas_nlruns, nlclasses, unsigned)
-CTL_RO_NL_GEN(arenas_lrun_i_size, index2size(NBINS+mib[2]), size_t)
+CTL_RO_NL_GEN(arenas_lrun_i_size, index2size(NBINS+(szind_t)mib[2]), size_t)
 static const ctl_named_node_t *
 arenas_lrun_i_index(const size_t *mib, size_t miblen, size_t i)
 {
@@ -1827,7 +1921,8 @@ arenas_lrun_i_index(const size_t *mib, size_t miblen, size_t i)
 }
 
 CTL_RO_NL_GEN(arenas_nhchunks, nhclasses, unsigned)
-CTL_RO_NL_GEN(arenas_hchunk_i_size, index2size(NBINS+nlclasses+mib[2]), size_t)
+CTL_RO_NL_GEN(arenas_hchunk_i_size, index2size(NBINS+nlclasses+(szind_t)mib[2]),
+    size_t)
 static const ctl_named_node_t *
 arenas_hchunk_i_index(const size_t *mib, size_t miblen, size_t i)
 {
@@ -1999,6 +2094,8 @@ CTL_RO_CGEN(config_stats, stats_mapped, ctl_stats.mapped, size_t)
 CTL_RO_GEN(stats_arenas_i_dss, ctl_stats.arenas[mib[2]].dss, const char *)
 CTL_RO_GEN(stats_arenas_i_lg_dirty_mult, ctl_stats.arenas[mib[2]].lg_dirty_mult,
     ssize_t)
+CTL_RO_GEN(stats_arenas_i_decay_time, ctl_stats.arenas[mib[2]].decay_time,
+    ssize_t)
 CTL_RO_GEN(stats_arenas_i_nthreads, ctl_stats.arenas[mib[2]].nthreads, unsigned)
 CTL_RO_GEN(stats_arenas_i_pactive, ctl_stats.arenas[mib[2]].pactive, size_t)
 CTL_RO_GEN(stats_arenas_i_pdirty, ctl_stats.arenas[mib[2]].pdirty, size_t)

src/extent.c

@@ -15,7 +15,7 @@ extent_quantize(size_t size)
 }
 
 JEMALLOC_INLINE_C int
-extent_szad_comp(extent_node_t *a, extent_node_t *b)
+extent_szad_comp(const extent_node_t *a, const extent_node_t *b)
 {
 	int ret;
 	size_t a_qsize = extent_quantize(extent_node_size_get(a));
@@ -41,7 +41,7 @@ rb_gen(, extent_tree_szad_, extent_tree_t, extent_node_t, szad_link,
     extent_szad_comp)
 
 JEMALLOC_INLINE_C int
-extent_ad_comp(extent_node_t *a, extent_node_t *b)
+extent_ad_comp(const extent_node_t *a, const extent_node_t *b)
 {
 	uintptr_t a_addr = (uintptr_t)extent_node_addr_get(a);
 	uintptr_t b_addr = (uintptr_t)extent_node_addr_get(b);

src/huge.c

@@ -31,35 +31,30 @@ huge_node_unset(const void *ptr, const extent_node_t *node)
 }
 
 void *
-huge_malloc(tsd_t *tsd, arena_t *arena, size_t size, bool zero,
+huge_malloc(tsd_t *tsd, arena_t *arena, size_t usize, bool zero,
     tcache_t *tcache)
 {
-	size_t usize;
 
-	usize = s2u(size);
+	assert(usize == s2u(usize));
-	if (usize == 0) {
-		/* size_t overflow. */
-		return (NULL);
-	}
 
 	return (huge_palloc(tsd, arena, usize, chunksize, zero, tcache));
 }
 
 void *
-huge_palloc(tsd_t *tsd, arena_t *arena, size_t size, size_t alignment,
+huge_palloc(tsd_t *tsd, arena_t *arena, size_t usize, size_t alignment,
     bool zero, tcache_t *tcache)
 {
 	void *ret;
-	size_t usize;
+	size_t ausize;
 	extent_node_t *node;
 	bool is_zeroed;
 
 	/* Allocate one or more contiguous chunks for this request. */
 
-	usize = sa2u(size, alignment);
+	ausize = sa2u(usize, alignment);
-	if (unlikely(usize == 0))
+	if (unlikely(ausize == 0 || ausize > HUGE_MAXCLASS))
 		return (NULL);
-	assert(usize >= chunksize);
+	assert(ausize >= chunksize);
 
 	/* Allocate an extent node with which to track the chunk. */
 	node = ipallocztm(tsd, CACHELINE_CEILING(sizeof(extent_node_t)),
@@ -74,16 +69,16 @@ huge_palloc(tsd_t *tsd, arena_t *arena, size_t size, size_t alignment,
 	is_zeroed = zero;
 	arena = arena_choose(tsd, arena);
 	if (unlikely(arena == NULL) || (ret = arena_chunk_alloc_huge(arena,
-	    size, alignment, &is_zeroed)) == NULL) {
+	    usize, alignment, &is_zeroed)) == NULL) {
-		idalloctm(tsd, node, tcache, true);
+		idalloctm(tsd, node, tcache, true, true);
 		return (NULL);
 	}
 
-	extent_node_init(node, arena, ret, size, is_zeroed, true);
+	extent_node_init(node, arena, ret, usize, is_zeroed, true);
 
 	if (huge_node_set(ret, node)) {
-		arena_chunk_dalloc_huge(arena, ret, size);
+		arena_chunk_dalloc_huge(arena, ret, usize);
-		idalloctm(tsd, node, tcache, true);
+		idalloctm(tsd, node, tcache, true, true);
 		return (NULL);
 	}
 
@@ -95,10 +90,11 @@ huge_palloc(tsd_t *tsd, arena_t *arena, size_t size, size_t alignment,
 
 	if (zero || (config_fill && unlikely(opt_zero))) {
 		if (!is_zeroed)
-			memset(ret, 0, size);
+			memset(ret, 0, usize);
 	} else if (config_fill && unlikely(opt_junk_alloc))
-		memset(ret, 0xa5, size);
+		memset(ret, 0xa5, usize);
 
+	arena_decay_tick(tsd, arena);
 	return (ret);
 }
 
@@ -280,11 +276,13 @@ huge_ralloc_no_move_expand(void *ptr, size_t oldsize, size_t usize, bool zero) {
 }
 
 bool
-huge_ralloc_no_move(void *ptr, size_t oldsize, size_t usize_min,
+huge_ralloc_no_move(tsd_t *tsd, void *ptr, size_t oldsize, size_t usize_min,
     size_t usize_max, bool zero)
 {
 
 	assert(s2u(oldsize) == oldsize);
+	/* The following should have been caught by callers. */
+	assert(usize_min > 0 && usize_max <= HUGE_MAXCLASS);
 
 	/* Both allocations must be huge to avoid a move. */
 	if (oldsize < chunksize || usize_max < chunksize)
@@ -292,14 +290,19 @@ huge_ralloc_no_move(void *ptr, size_t oldsize, size_t usize_min,
 
 	if (CHUNK_CEILING(usize_max) > CHUNK_CEILING(oldsize)) {
 		/* Attempt to expand the allocation in-place. */
-		if (!huge_ralloc_no_move_expand(ptr, oldsize, usize_max, zero))
+		if (!huge_ralloc_no_move_expand(ptr, oldsize, usize_max,
+		    zero)) {
+			arena_decay_tick(tsd, huge_aalloc(ptr));
 			return (false);
+		}
 		/* Try again, this time with usize_min. */
 		if (usize_min < usize_max && CHUNK_CEILING(usize_min) >
 		    CHUNK_CEILING(oldsize) && huge_ralloc_no_move_expand(ptr,
-		    oldsize, usize_min, zero))
+		    oldsize, usize_min, zero)) {
+			arena_decay_tick(tsd, huge_aalloc(ptr));
 			return (false);
 		}
+	}
 
 	/*
 	 * Avoid moving the allocation if the existing chunk size accommodates
@@ -309,12 +312,17 @@ huge_ralloc_no_move(void *ptr, size_t oldsize, size_t usize_min,
 	    && CHUNK_CEILING(oldsize) <= CHUNK_CEILING(usize_max)) {
 		huge_ralloc_no_move_similar(ptr, oldsize, usize_min, usize_max,
 		    zero);
+		arena_decay_tick(tsd, huge_aalloc(ptr));
 		return (false);
 	}
 
 	/* Attempt to shrink the allocation in-place. */
-	if (CHUNK_CEILING(oldsize) > CHUNK_CEILING(usize_max))
+	if (CHUNK_CEILING(oldsize) > CHUNK_CEILING(usize_max)) {
-		return (huge_ralloc_no_move_shrink(ptr, oldsize, usize_max));
+		if (!huge_ralloc_no_move_shrink(ptr, oldsize, usize_max)) {
+			arena_decay_tick(tsd, huge_aalloc(ptr));
+			return (false);
+		}
+	}
 	return (true);
 }
 
@@ -335,8 +343,11 @@ huge_ralloc(tsd_t *tsd, arena_t *arena, void *ptr, size_t oldsize, size_t usize,
 	void *ret;
 	size_t copysize;
 
+	/* The following should have been caught by callers. */
+	assert(usize > 0 && usize <= HUGE_MAXCLASS);
+
 	/* Try to avoid moving the allocation. */
-	if (!huge_ralloc_no_move(ptr, oldsize, usize, usize, zero))
+	if (!huge_ralloc_no_move(tsd, ptr, oldsize, usize, usize, zero))
 		return (ptr);
 
 	/*
@@ -372,7 +383,9 @@ huge_dalloc(tsd_t *tsd, void *ptr, tcache_t *tcache)
 	    extent_node_size_get(node));
 	arena_chunk_dalloc_huge(extent_node_arena_get(node),
 	    extent_node_addr_get(node), extent_node_size_get(node));
-	idalloctm(tsd, node, tcache, true);
+	idalloctm(tsd, node, tcache, true, true);
+
+	arena_decay_tick(tsd, arena);
 }
 
 arena_t *

src/nstime.c

@@ -0,0 +1,148 @@
+#include "jemalloc/internal/jemalloc_internal.h"
+
+#define	BILLION	UINT64_C(1000000000)
+
+void
+nstime_init(nstime_t *time, uint64_t ns)
+{
+
+	time->ns = ns;
+}
+
+void
+nstime_init2(nstime_t *time, uint64_t sec, uint64_t nsec)
+{
+
+	time->ns = sec * BILLION + nsec;
+}
+
+uint64_t
+nstime_ns(const nstime_t *time)
+{
+
+	return (time->ns);
+}
+
+uint64_t
+nstime_sec(const nstime_t *time)
+{
+
+	return (time->ns / BILLION);
+}
+
+uint64_t
+nstime_nsec(const nstime_t *time)
+{
+
+	return (time->ns % BILLION);
+}
+
+void
+nstime_copy(nstime_t *time, const nstime_t *source)
+{
+
+	*time = *source;
+}
+
+int
+nstime_compare(const nstime_t *a, const nstime_t *b)
+{
+
+	return ((a->ns > b->ns) - (a->ns < b->ns));
+}
+
+void
+nstime_add(nstime_t *time, const nstime_t *addend)
+{
+
+	assert(UINT64_MAX - time->ns >= addend->ns);
+
+	time->ns += addend->ns;
+}
+
+void
+nstime_subtract(nstime_t *time, const nstime_t *subtrahend)
+{
+
+	assert(nstime_compare(time, subtrahend) >= 0);
+
+	time->ns -= subtrahend->ns;
+}
+
+void
+nstime_imultiply(nstime_t *time, uint64_t multiplier)
+{
+
+	assert((((time->ns | multiplier) & (UINT64_MAX << (sizeof(uint64_t) <<
+	    2))) == 0) || ((time->ns * multiplier) / multiplier == time->ns));
+
+	time->ns *= multiplier;
+}
+
+void
+nstime_idivide(nstime_t *time, uint64_t divisor)
+{
+
+	assert(divisor != 0);
+
+	time->ns /= divisor;
+}
+
+uint64_t
+nstime_divide(const nstime_t *time, const nstime_t *divisor)
+{
+
+	assert(divisor->ns != 0);
+
+	return (time->ns / divisor->ns);
+}
+
+#ifdef JEMALLOC_JET
+#undef nstime_update
+#define	nstime_update JEMALLOC_N(nstime_update_impl)
+#endif
+bool
+nstime_update(nstime_t *time)
+{
+	nstime_t old_time;
+
+	nstime_copy(&old_time, time);
+
+#ifdef _WIN32
+	{
+		FILETIME ft;
+		uint64_t ticks;
+		GetSystemTimeAsFileTime(&ft);
+		ticks = (((uint64_t)ft.dwHighDateTime) << 32) |
+		    ft.dwLowDateTime;
+		time->ns = ticks * 100;
+	}
+#elif JEMALLOC_CLOCK_GETTIME
+	{
+		struct timespec ts;
+
+		if (sysconf(_SC_MONOTONIC_CLOCK) > 0)
+			clock_gettime(CLOCK_MONOTONIC, &ts);
+		else
+			clock_gettime(CLOCK_REALTIME, &ts);
+		time->ns = ts.tv_sec * BILLION + ts.tv_nsec;
+	}
+#else
+	struct timeval tv;
+	gettimeofday(&tv, NULL);
+	time->ns = tv.tv_sec * BILLION + tv.tv_usec * 1000;
+#endif
+
+	/* Handle non-monotonic clocks. */
+	if (unlikely(nstime_compare(&old_time, time) > 0)) {
+		nstime_copy(time, &old_time);
+		return (true);
+	}
+
+	return (false);
+}
+#ifdef JEMALLOC_JET
+#undef nstime_update
+#define	nstime_update JEMALLOC_N(nstime_update)
+nstime_update_t *nstime_update = JEMALLOC_N(nstime_update_impl);
+#endif

src/prng.c

@@ -0,0 +1,2 @@
+#define	JEMALLOC_PRNG_C_
+#include "jemalloc/internal/jemalloc_internal.h"

src/prof.c

@@ -109,7 +109,7 @@ static char		prof_dump_buf[
     1
 #endif
 ];
-static unsigned		prof_dump_buf_end;
+static size_t		prof_dump_buf_end;
 static int		prof_dump_fd;
 
 /* Do not dump any profiles until bootstrapping is complete. */
@@ -551,9 +551,9 @@ prof_gctx_create(tsd_t *tsd, prof_bt_t *bt)
 	/*
 	 * Create a single allocation that has space for vec of length bt->len.
 	 */
-	prof_gctx_t *gctx = (prof_gctx_t *)iallocztm(tsd, offsetof(prof_gctx_t,
+	size_t size = offsetof(prof_gctx_t, vec) + (bt->len * sizeof(void *));
-	    vec) + (bt->len * sizeof(void *)), false, tcache_get(tsd, true),
+	prof_gctx_t *gctx = (prof_gctx_t *)iallocztm(tsd, size,
-	    true, NULL);
+	    size2index(size), false, tcache_get(tsd, true), true, NULL, true);
 	if (gctx == NULL)
 		return (NULL);
 	gctx->lock = prof_gctx_mutex_choose();
@@ -594,7 +594,7 @@ prof_gctx_try_destroy(tsd_t *tsd, prof_tdata_t *tdata_self, prof_gctx_t *gctx,
 		prof_leave(tsd, tdata_self);
 		/* Destroy gctx. */
 		malloc_mutex_unlock(gctx->lock);
-		idalloctm(tsd, gctx, tcache_get(tsd, false), true);
+		idalloctm(tsd, gctx, tcache_get(tsd, false), true, true);
 	} else {
 		/*
 		 * Compensate for increment in prof_tctx_destroy() or
@@ -701,7 +701,7 @@ prof_tctx_destroy(tsd_t *tsd, prof_tctx_t *tctx)
 		prof_tdata_destroy(tsd, tdata, false);
 
 	if (destroy_tctx)
-		idalloctm(tsd, tctx, tcache_get(tsd, false), true);
+		idalloctm(tsd, tctx, tcache_get(tsd, false), true, true);
 }
 
 static bool
@@ -730,7 +730,8 @@ prof_lookup_global(tsd_t *tsd, prof_bt_t *bt, prof_tdata_t *tdata,
 		if (ckh_insert(tsd, &bt2gctx, btkey.v, gctx.v)) {
 			/* OOM. */
 			prof_leave(tsd, tdata);
-			idalloctm(tsd, gctx.v, tcache_get(tsd, false), true);
+			idalloctm(tsd, gctx.v, tcache_get(tsd, false), true,
+			    true);
 			return (true);
 		}
 		new_gctx = true;
@@ -789,8 +790,9 @@ prof_lookup(tsd_t *tsd, prof_bt_t *bt)
 
 		/* Link a prof_tctx_t into gctx for this thread. */
 		tcache = tcache_get(tsd, true);
-		ret.v = iallocztm(tsd, sizeof(prof_tctx_t), false, tcache, true,
+		ret.v = iallocztm(tsd, sizeof(prof_tctx_t),
-		    NULL);
+		    size2index(sizeof(prof_tctx_t)), false, tcache, true, NULL,
+		    true);
 		if (ret.p == NULL) {
 			if (new_gctx)
 				prof_gctx_try_destroy(tsd, tdata, gctx, tdata);
@@ -810,7 +812,7 @@ prof_lookup(tsd_t *tsd, prof_bt_t *bt)
 		if (error) {
 			if (new_gctx)
 				prof_gctx_try_destroy(tsd, tdata, gctx, tdata);
-			idalloctm(tsd, ret.v, tcache, true);
+			idalloctm(tsd, ret.v, tcache, true, true);
 			return (NULL);
 		}
 		malloc_mutex_lock(gctx->lock);
@@ -869,8 +871,7 @@ prof_sample_threshold_update(prof_tdata_t *tdata)
 	 *   pp 500
 	 *   (http://luc.devroye.org/rnbookindex.html)
 	 */
-	prng64(r, 53, tdata->prng_state, UINT64_C(6364136223846793005),
+	r = prng_lg_range(&tdata->prng_state, 53);
-	    UINT64_C(1442695040888963407));
 	u = (double)r * (1.0/9007199254740992.0L);
 	tdata->bytes_until_sample = (uint64_t)(log(u) /
 	    log(1.0 - (1.0 / (double)((uint64_t)1U << lg_prof_sample))))
@@ -988,7 +989,7 @@ prof_dump_close(bool propagate_err)
 static bool
 prof_dump_write(bool propagate_err, const char *s)
 {
-	unsigned i, slen, n;
+	size_t i, slen, n;
 
 	cassert(config_prof);
 
@@ -1211,7 +1212,7 @@ prof_gctx_finish(tsd_t *tsd, prof_gctx_tree_t *gctxs)
 					tctx_tree_remove(&gctx->tctxs,
 					    to_destroy);
 					idalloctm(tsd, to_destroy,
-					    tcache_get(tsd, false), true);
+					    tcache_get(tsd, false), true, true);
 				} else
 					next = NULL;
 			} while (next != NULL);
@@ -1358,6 +1359,7 @@ label_return:
 	return (ret);
 }
 
+#ifndef _WIN32
 JEMALLOC_FORMAT_PRINTF(1, 2)
 static int
 prof_open_maps(const char *format, ...)
@@ -1373,6 +1375,18 @@ prof_open_maps(const char *format, ...)
 
 	return (mfd);
 }
+#endif
+
+static int
+prof_getpid(void)
+{
+
+#ifdef _WIN32
+	return (GetCurrentProcessId());
+#else
+	return (getpid());
+#endif
+}
 
 static bool
 prof_dump_maps(bool propagate_err)
@@ -1383,9 +1397,11 @@ prof_dump_maps(bool propagate_err)
 	cassert(config_prof);
 #ifdef __FreeBSD__
 	mfd = prof_open_maps("/proc/curproc/map");
+#elif defined(_WIN32)
+	mfd = -1; // Not implemented
 #else
 	{
-		int pid = getpid();
+		int pid = prof_getpid();
 
 		mfd = prof_open_maps("/proc/%d/task/%d/maps", pid, pid);
 		if (mfd == -1)
@@ -1554,12 +1570,12 @@ prof_dump_filename(char *filename, char v, uint64_t vseq)
 	        /* "<prefix>.<pid>.<seq>.v<vseq>.heap" */
 		malloc_snprintf(filename, DUMP_FILENAME_BUFSIZE,
 		    "%s.%d.%"FMTu64".%c%"FMTu64".heap",
-		    opt_prof_prefix, (int)getpid(), prof_dump_seq, v, vseq);
+		    opt_prof_prefix, prof_getpid(), prof_dump_seq, v, vseq);
 	} else {
 	        /* "<prefix>.<pid>.<seq>.<v>.heap" */
 		malloc_snprintf(filename, DUMP_FILENAME_BUFSIZE,
 		    "%s.%d.%"FMTu64".%c.heap",
-		    opt_prof_prefix, (int)getpid(), prof_dump_seq, v);
+		    opt_prof_prefix, prof_getpid(), prof_dump_seq, v);
 	}
 	prof_dump_seq++;
 }
@@ -1714,8 +1730,8 @@ prof_tdata_init_impl(tsd_t *tsd, uint64_t thr_uid, uint64_t thr_discrim,
 
 	/* Initialize an empty cache for this thread. */
 	tcache = tcache_get(tsd, true);
-	tdata = (prof_tdata_t *)iallocztm(tsd, sizeof(prof_tdata_t), false,
+	tdata = (prof_tdata_t *)iallocztm(tsd, sizeof(prof_tdata_t),
-	    tcache, true, NULL);
+	    size2index(sizeof(prof_tdata_t)), false, tcache, true, NULL, true);
 	if (tdata == NULL)
 		return (NULL);
 
@@ -1729,7 +1745,7 @@ prof_tdata_init_impl(tsd_t *tsd, uint64_t thr_uid, uint64_t thr_discrim,
 
 	if (ckh_new(tsd, &tdata->bt2tctx, PROF_CKH_MINITEMS,
 	    prof_bt_hash, prof_bt_keycomp)) {
-		idalloctm(tsd, tdata, tcache, true);
+		idalloctm(tsd, tdata, tcache, true, true);
 		return (NULL);
 	}
 
@@ -1784,9 +1800,9 @@ prof_tdata_destroy_locked(tsd_t *tsd, prof_tdata_t *tdata,
 
 	tcache = tcache_get(tsd, false);
 	if (tdata->thread_name != NULL)
-		idalloctm(tsd, tdata->thread_name, tcache, true);
+		idalloctm(tsd, tdata->thread_name, tcache, true, true);
 	ckh_delete(tsd, &tdata->bt2tctx);
-	idalloctm(tsd, tdata, tcache, true);
+	idalloctm(tsd, tdata, tcache, true, true);
 }
 
 static void
@@ -1947,7 +1963,8 @@ prof_thread_name_alloc(tsd_t *tsd, const char *thread_name)
 	if (size == 1)
 		return ("");
 
-	ret = iallocztm(tsd, size, false, tcache_get(tsd, true), true, NULL);
+	ret = iallocztm(tsd, size, size2index(size), false, tcache_get(tsd,
+	    true), true, NULL, true);
 	if (ret == NULL)
 		return (NULL);
 	memcpy(ret, thread_name, size);
@@ -1980,7 +1997,7 @@ prof_thread_name_set(tsd_t *tsd, const char *thread_name)
 
 	if (tdata->thread_name != NULL) {
 		idalloctm(tsd, tdata->thread_name, tcache_get(tsd, false),
-		    true);
+		    true, true);
 		tdata->thread_name = NULL;
 	}
 	if (strlen(s) > 0)

src/quarantine.c

@@ -23,12 +23,14 @@ static quarantine_t *
 quarantine_init(tsd_t *tsd, size_t lg_maxobjs)
 {
 	quarantine_t *quarantine;
+	size_t size;
 
 	assert(tsd_nominal(tsd));
 
-	quarantine = (quarantine_t *)iallocztm(tsd, offsetof(quarantine_t, objs)
+	size = offsetof(quarantine_t, objs) + ((ZU(1) << lg_maxobjs) *
-	    + ((ZU(1) << lg_maxobjs) * sizeof(quarantine_obj_t)), false,
+	    sizeof(quarantine_obj_t));
-	    tcache_get(tsd, true), true, NULL);
+	quarantine = (quarantine_t *)iallocztm(tsd, size, size2index(size),
+	    false, tcache_get(tsd, true), true, NULL, true);
 	if (quarantine == NULL)
 		return (NULL);
 	quarantine->curbytes = 0;
@@ -55,7 +57,7 @@ quarantine_alloc_hook_work(tsd_t *tsd)
 	if (tsd_quarantine_get(tsd) == NULL)
 		tsd_quarantine_set(tsd, quarantine);
 	else
-		idalloctm(tsd, quarantine, tcache_get(tsd, false), true);
+		idalloctm(tsd, quarantine, tcache_get(tsd, false), true, true);
 }
 
 static quarantine_t *
@@ -87,7 +89,7 @@ quarantine_grow(tsd_t *tsd, quarantine_t *quarantine)
 		memcpy(&ret->objs[ncopy_a], quarantine->objs, ncopy_b *
 		    sizeof(quarantine_obj_t));
 	}
-	idalloctm(tsd, quarantine, tcache_get(tsd, false), true);
+	idalloctm(tsd, quarantine, tcache_get(tsd, false), true, true);
 
 	tsd_quarantine_set(tsd, ret);
 	return (ret);
@@ -98,7 +100,7 @@ quarantine_drain_one(tsd_t *tsd, quarantine_t *quarantine)
 {
 	quarantine_obj_t *obj = &quarantine->objs[quarantine->first];
 	assert(obj->usize == isalloc(obj->ptr, config_prof));
-	idalloctm(tsd, obj->ptr, NULL, false);
+	idalloctm(tsd, obj->ptr, NULL, false, true);
 	quarantine->curbytes -= obj->usize;
 	quarantine->curobjs--;
 	quarantine->first = (quarantine->first + 1) & ((ZU(1) <<
@@ -123,7 +125,7 @@ quarantine(tsd_t *tsd, void *ptr)
 	assert(opt_quarantine);
 
 	if ((quarantine = tsd_quarantine_get(tsd)) == NULL) {
-		idalloctm(tsd, ptr, NULL, false);
+		idalloctm(tsd, ptr, NULL, false, true);
 		return;
 	}
 	/*
@@ -162,7 +164,7 @@ quarantine(tsd_t *tsd, void *ptr)
 		}
 	} else {
 		assert(quarantine->curbytes == 0);
-		idalloctm(tsd, ptr, NULL, false);
+		idalloctm(tsd, ptr, NULL, false, true);
 	}
 }
 
@@ -177,7 +179,7 @@ quarantine_cleanup(tsd_t *tsd)
 	quarantine = tsd_quarantine_get(tsd);
 	if (quarantine != NULL) {
 		quarantine_drain(tsd, quarantine, 0);
-		idalloctm(tsd, quarantine, tcache_get(tsd, false), true);
+		idalloctm(tsd, quarantine, tcache_get(tsd, false), true, true);
 		tsd_quarantine_set(tsd, NULL);
 	}
 }

src/stats.c

@@ -258,7 +258,7 @@ stats_arena_print(void (*write_cb)(void *, const char *), void *cbopaque,
 {
 	unsigned nthreads;
 	const char *dss;
-	ssize_t lg_dirty_mult;
+	ssize_t lg_dirty_mult, decay_time;
 	size_t page, pactive, pdirty, mapped;
 	size_t metadata_mapped, metadata_allocated;
 	uint64_t npurge, nmadvise, purged;
@@ -278,6 +278,7 @@ stats_arena_print(void (*write_cb)(void *, const char *), void *cbopaque,
 	malloc_cprintf(write_cb, cbopaque, "dss allocation precedence: %s\n",
 	    dss);
 	CTL_M2_GET("stats.arenas.0.lg_dirty_mult", i, &lg_dirty_mult, ssize_t);
+	if (opt_purge == purge_mode_ratio) {
 		if (lg_dirty_mult >= 0) {
 			malloc_cprintf(write_cb, cbopaque,
 			    "min active:dirty page ratio: %u:1\n",
@@ -286,15 +287,23 @@ stats_arena_print(void (*write_cb)(void *, const char *), void *cbopaque,
 			malloc_cprintf(write_cb, cbopaque,
 			    "min active:dirty page ratio: N/A\n");
 		}
+	}
+	CTL_M2_GET("stats.arenas.0.decay_time", i, &decay_time, ssize_t);
+	if (opt_purge == purge_mode_decay) {
+		if (decay_time >= 0) {
+			malloc_cprintf(write_cb, cbopaque, "decay time: %zd\n",
+			    decay_time);
+		} else
+			malloc_cprintf(write_cb, cbopaque, "decay time: N/A\n");
+	}
 	CTL_M2_GET("stats.arenas.0.pactive", i, &pactive, size_t);
 	CTL_M2_GET("stats.arenas.0.pdirty", i, &pdirty, size_t);
 	CTL_M2_GET("stats.arenas.0.npurge", i, &npurge, uint64_t);
 	CTL_M2_GET("stats.arenas.0.nmadvise", i, &nmadvise, uint64_t);
 	CTL_M2_GET("stats.arenas.0.purged", i, &purged, uint64_t);
 	malloc_cprintf(write_cb, cbopaque,
-	    "dirty pages: %zu:%zu active:dirty, %"FMTu64" sweep%s, %"FMTu64
+	    "purging: dirty: %zu, sweeps: %"FMTu64", madvises: %"FMTu64", "
-	    " madvise%s, %"FMTu64" purged\n", pactive, pdirty, npurge, npurge ==
+	    "purged: %"FMTu64"\n", pdirty, npurge, nmadvise, purged);
-	    1 ? "" : "s", nmadvise, nmadvise == 1 ? "" : "s", purged);
 
 	malloc_cprintf(write_cb, cbopaque,
 	    "                            allocated      nmalloc      ndalloc"
@@ -426,9 +435,10 @@ stats_print(void (*write_cb)(void *, const char *), void *cbopaque,
 		bool bv;
 		unsigned uv;
 		ssize_t ssv;
-		size_t sv, bsz, ssz, sssz, cpsz;
+		size_t sv, bsz, usz, ssz, sssz, cpsz;
 
 		bsz = sizeof(bool);
+		usz = sizeof(unsigned);
 		ssz = sizeof(size_t);
 		sssz = sizeof(ssize_t);
 		cpsz = sizeof(const char *);
@@ -438,6 +448,8 @@ stats_print(void (*write_cb)(void *, const char *), void *cbopaque,
 		CTL_GET("config.debug", &bv, bool);
 		malloc_cprintf(write_cb, cbopaque, "Assertions %s\n",
 		    bv ? "enabled" : "disabled");
+		malloc_cprintf(write_cb, cbopaque,
+		    "config.malloc_conf: \"%s\"\n", config_malloc_conf);
 
 #define	OPT_WRITE_BOOL(n)						\
 		if (je_mallctl("opt."#n, &bv, &bsz, NULL, 0) == 0) {	\
@@ -453,6 +465,11 @@ stats_print(void (*write_cb)(void *, const char *), void *cbopaque,
 			    : "false", bv2 ? "true" : "false");		\
 		}							\
 }
+#define	OPT_WRITE_UNSIGNED(n)						\
+		if (je_mallctl("opt."#n, &uv, &usz, NULL, 0) == 0) {	\
+			malloc_cprintf(write_cb, cbopaque,		\
+			"  opt."#n": %zu\n", sv);			\
+		}
 #define	OPT_WRITE_SIZE_T(n)						\
 		if (je_mallctl("opt."#n, &sv, &ssz, NULL, 0) == 0) {	\
 			malloc_cprintf(write_cb, cbopaque,		\
@@ -483,8 +500,14 @@ stats_print(void (*write_cb)(void *, const char *), void *cbopaque,
 		OPT_WRITE_BOOL(abort)
 		OPT_WRITE_SIZE_T(lg_chunk)
 		OPT_WRITE_CHAR_P(dss)
-		OPT_WRITE_SIZE_T(narenas)
+		OPT_WRITE_UNSIGNED(narenas)
-		OPT_WRITE_SSIZE_T_MUTABLE(lg_dirty_mult, arenas.lg_dirty_mult)
+		OPT_WRITE_CHAR_P(purge)
+		if (opt_purge == purge_mode_ratio) {
+			OPT_WRITE_SSIZE_T_MUTABLE(lg_dirty_mult,
+			    arenas.lg_dirty_mult)
+		}
+		if (opt_purge == purge_mode_decay)
+			OPT_WRITE_SSIZE_T_MUTABLE(decay_time, arenas.decay_time)
 		OPT_WRITE_BOOL(stats_print)
 		OPT_WRITE_CHAR_P(junk)
 		OPT_WRITE_SIZE_T(quarantine)
@@ -529,13 +552,22 @@ stats_print(void (*write_cb)(void *, const char *), void *cbopaque,
 		malloc_cprintf(write_cb, cbopaque, "Page size: %zu\n", sv);
 
 		CTL_GET("arenas.lg_dirty_mult", &ssv, ssize_t);
+		if (opt_purge == purge_mode_ratio) {
 			if (ssv >= 0) {
 				malloc_cprintf(write_cb, cbopaque,
-			    "Min active:dirty page ratio per arena: %u:1\n",
+				    "Min active:dirty page ratio per arena: "
-			    (1U << ssv));
+				    "%u:1\n", (1U << ssv));
 			} else {
 				malloc_cprintf(write_cb, cbopaque,
-			    "Min active:dirty page ratio per arena: N/A\n");
+				    "Min active:dirty page ratio per arena: "
+				    "N/A\n");
+			}
+		}
+		CTL_GET("arenas.decay_time", &ssv, ssize_t);
+		if (opt_purge == purge_mode_decay) {
+			malloc_cprintf(write_cb, cbopaque,
+			    "Unused dirty page decay time: %zd%s\n",
+			    ssv, (ssv < 0) ? " (no decay)" : "");
 		}
 		if (je_mallctl("arenas.tcache_max", &sv, &ssz, NULL, 0) == 0) {
 			malloc_cprintf(write_cb, cbopaque,

src/tcache.c

@@ -10,7 +10,7 @@ ssize_t	opt_lg_tcache_max = LG_TCACHE_MAXCLASS_DEFAULT;
 tcache_bin_info_t	*tcache_bin_info;
 static unsigned		stack_nelms; /* Total stack elms per tcache. */
 
-size_t			nhbins;
+unsigned		nhbins;
 size_t			tcache_maxclass;
 
 tcaches_t		*tcaches;
@@ -67,20 +67,19 @@ tcache_event_hard(tsd_t *tsd, tcache_t *tcache)
 	tcache->next_gc_bin++;
 	if (tcache->next_gc_bin == nhbins)
 		tcache->next_gc_bin = 0;
-	tcache->ev_cnt = 0;
 }
 
 void *
 tcache_alloc_small_hard(tsd_t *tsd, arena_t *arena, tcache_t *tcache,
-    tcache_bin_t *tbin, szind_t binind)
+    tcache_bin_t *tbin, szind_t binind, bool *tcache_success)
 {
 	void *ret;
 
-	arena_tcache_fill_small(arena, tbin, binind, config_prof ?
+	arena_tcache_fill_small(tsd, arena, tbin, binind, config_prof ?
 	    tcache->prof_accumbytes : 0);
 	if (config_prof)
 		tcache->prof_accumbytes = 0;
-	ret = tcache_alloc_easy(tbin);
+	ret = tcache_alloc_easy(tbin, tcache_success);
 
 	return (ret);
 }
@@ -102,7 +101,7 @@ tcache_bin_flush_small(tsd_t *tsd, tcache_t *tcache, tcache_bin_t *tbin,
 	for (nflush = tbin->ncached - rem; nflush > 0; nflush = ndeferred) {
 		/* Lock the arena bin associated with the first object. */
 		arena_chunk_t *chunk = (arena_chunk_t *)CHUNK_ADDR2BASE(
-		    tbin->avail[0]);
+		    *(tbin->avail - 1));
 		arena_t *bin_arena = extent_node_arena_get(&chunk->node);
 		arena_bin_t *bin = &bin_arena->bins[binind];
 
@@ -122,7 +121,7 @@ tcache_bin_flush_small(tsd_t *tsd, tcache_t *tcache, tcache_bin_t *tbin,
 		}
 		ndeferred = 0;
 		for (i = 0; i < nflush; i++) {
-			ptr = tbin->avail[i];
+			ptr = *(tbin->avail - 1 - i);
 			assert(ptr != NULL);
 			chunk = (arena_chunk_t *)CHUNK_ADDR2BASE(ptr);
 			if (extent_node_arena_get(&chunk->node) == bin_arena) {
@@ -139,11 +138,12 @@ tcache_bin_flush_small(tsd_t *tsd, tcache_t *tcache, tcache_bin_t *tbin,
 				 * locked.  Stash the object, so that it can be
 				 * handled in a future pass.
 				 */
-				tbin->avail[ndeferred] = ptr;
+				*(tbin->avail - 1 - ndeferred) = ptr;
 				ndeferred++;
 			}
 		}
 		malloc_mutex_unlock(&bin->lock);
+		arena_decay_ticks(tsd, bin_arena, nflush - ndeferred);
 	}
 	if (config_stats && !merged_stats) {
 		/*
@@ -158,8 +158,8 @@ tcache_bin_flush_small(tsd_t *tsd, tcache_t *tcache, tcache_bin_t *tbin,
 		malloc_mutex_unlock(&bin->lock);
 	}
 
-	memmove(tbin->avail, &tbin->avail[tbin->ncached - rem],
+	memmove(tbin->avail - rem, tbin->avail - tbin->ncached, rem *
-	    rem * sizeof(void *));
+	    sizeof(void *));
 	tbin->ncached = rem;
 	if ((int)tbin->ncached < tbin->low_water)
 		tbin->low_water = tbin->ncached;
@@ -182,7 +182,7 @@ tcache_bin_flush_large(tsd_t *tsd, tcache_bin_t *tbin, szind_t binind,
 	for (nflush = tbin->ncached - rem; nflush > 0; nflush = ndeferred) {
 		/* Lock the arena associated with the first object. */
 		arena_chunk_t *chunk = (arena_chunk_t *)CHUNK_ADDR2BASE(
-		    tbin->avail[0]);
+		    *(tbin->avail - 1));
 		arena_t *locked_arena = extent_node_arena_get(&chunk->node);
 		UNUSED bool idump;
 
@@ -206,7 +206,7 @@ tcache_bin_flush_large(tsd_t *tsd, tcache_bin_t *tbin, szind_t binind,
 		}
 		ndeferred = 0;
 		for (i = 0; i < nflush; i++) {
-			ptr = tbin->avail[i];
+			ptr = *(tbin->avail - 1 - i);
 			assert(ptr != NULL);
 			chunk = (arena_chunk_t *)CHUNK_ADDR2BASE(ptr);
 			if (extent_node_arena_get(&chunk->node) ==
@@ -220,13 +220,14 @@ tcache_bin_flush_large(tsd_t *tsd, tcache_bin_t *tbin, szind_t binind,
 				 * Stash the object, so that it can be handled
 				 * in a future pass.
 				 */
-				tbin->avail[ndeferred] = ptr;
+				*(tbin->avail - 1 - ndeferred) = ptr;
 				ndeferred++;
 			}
 		}
 		malloc_mutex_unlock(&locked_arena->lock);
 		if (config_prof && idump)
 			prof_idump();
+		arena_decay_ticks(tsd, locked_arena, nflush - ndeferred);
 	}
 	if (config_stats && !merged_stats) {
 		/*
@@ -241,8 +242,8 @@ tcache_bin_flush_large(tsd_t *tsd, tcache_bin_t *tbin, szind_t binind,
 		malloc_mutex_unlock(&arena->lock);
 	}
 
-	memmove(tbin->avail, &tbin->avail[tbin->ncached - rem],
+	memmove(tbin->avail - rem, tbin->avail - tbin->ncached, rem *
-	    rem * sizeof(void *));
+	    sizeof(void *));
 	tbin->ncached = rem;
 	if ((int)tbin->ncached < tbin->low_water)
 		tbin->low_water = tbin->ncached;
@@ -324,18 +325,26 @@ tcache_create(tsd_t *tsd, arena_t *arena)
 	/* Avoid false cacheline sharing. */
 	size = sa2u(size, CACHELINE);
 
-	tcache = ipallocztm(tsd, size, CACHELINE, true, false, true, a0get());
+	tcache = ipallocztm(tsd, size, CACHELINE, true, false, true,
+	    arena_get(0, false));
 	if (tcache == NULL)
 		return (NULL);
 
 	tcache_arena_associate(tcache, arena);
 
+	ticker_init(&tcache->gc_ticker, TCACHE_GC_INCR);
+
 	assert((TCACHE_NSLOTS_SMALL_MAX & 1U) == 0);
 	for (i = 0; i < nhbins; i++) {
 		tcache->tbins[i].lg_fill_div = 1;
+		stack_offset += tcache_bin_info[i].ncached_max * sizeof(void *);
+		/*
+		 * avail points past the available space.  Allocations will
+		 * access the slots toward higher addresses (for the benefit of
+		 * prefetch).
+		 */
 		tcache->tbins[i].avail = (void **)((uintptr_t)tcache +
 		    (uintptr_t)stack_offset);
-		stack_offset += tcache_bin_info[i].ncached_max * sizeof(void *);
 	}
 
 	return (tcache);
@@ -379,7 +388,7 @@ tcache_destroy(tsd_t *tsd, tcache_t *tcache)
 	    arena_prof_accum(arena, tcache->prof_accumbytes))
 		prof_idump();
 
-	idalloctm(tsd, tcache, false, true);
+	idalloctm(tsd, tcache, false, true, true);
 }
 
 void
@@ -445,7 +454,7 @@ tcaches_create(tsd_t *tsd, unsigned *r_ind)
 
 	if (tcaches_avail == NULL && tcaches_past > MALLOCX_TCACHE_MAX)
 		return (true);
-	tcache = tcache_create(tsd, a0get());
+	tcache = tcache_create(tsd, arena_get(0, false));
 	if (tcache == NULL)
 		return (true);
 
@@ -453,7 +462,7 @@ tcaches_create(tsd_t *tsd, unsigned *r_ind)
 		elm = tcaches_avail;
 		tcaches_avail = tcaches_avail->next;
 		elm->tcache = tcache;
-		*r_ind = elm - tcaches;
+		*r_ind = (unsigned)(elm - tcaches);
 	} else {
 		elm = &tcaches[tcaches_past];
 		elm->tcache = tcache;

src/ticker.c

@@ -0,0 +1,2 @@
+#define	JEMALLOC_TICKER_C_
+#include "jemalloc/internal/jemalloc_internal.h"

src/tsd.c

@@ -113,7 +113,7 @@ malloc_tsd_boot0(void)
 	ncleanups = 0;
 	if (tsd_boot0())
 		return (true);
-	*tsd_arenas_cache_bypassp_get(tsd_fetch()) = true;
+	*tsd_arenas_tdata_bypassp_get(tsd_fetch()) = true;
 	return (false);
 }
 
@@ -122,7 +122,7 @@ malloc_tsd_boot1(void)
 {
 
 	tsd_boot1();
-	*tsd_arenas_cache_bypassp_get(tsd_fetch()) = false;
+	*tsd_arenas_tdata_bypassp_get(tsd_fetch()) = false;
 }
 
 #ifdef _WIN32
@@ -148,13 +148,15 @@ _tls_callback(HINSTANCE hinstDLL, DWORD fdwReason, LPVOID lpvReserved)
 #ifdef _MSC_VER
 #  ifdef _M_IX86
 #    pragma comment(linker, "/INCLUDE:__tls_used")
+#    pragma comment(linker, "/INCLUDE:_tls_callback")
 #  else
 #    pragma comment(linker, "/INCLUDE:_tls_used")
+#    pragma comment(linker, "/INCLUDE:tls_callback")
 #  endif
 #  pragma section(".CRT$XLY",long,read)
 #endif
 JEMALLOC_SECTION(".CRT$XLY") JEMALLOC_ATTR(used)
-static BOOL	(WINAPI *const tls_callback)(HINSTANCE hinstDLL,
+BOOL	(WINAPI *const tls_callback)(HINSTANCE hinstDLL,
     DWORD fdwReason, LPVOID lpvReserved) = _tls_callback;
 #endif
 

src/util.c

@@ -1,3 +1,7 @@
+/*
+ * Define simple versions of assertion macros that won't recurse in case
+ * of assertion failures in malloc_*printf().
+ */
 #define	assert(e) do {							\
 	if (config_debug && !(e)) {					\
 		malloc_write("<jemalloc>: Failed assertion\n");		\
@@ -49,10 +53,14 @@ wrtmessage(void *cbopaque, const char *s)
 	 * 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
+	 * unused result in the widest plausible type to avoid compiler
+	 * warnings.
 	 */
-	UNUSED int result = syscall(SYS_write, STDERR_FILENO, s, strlen(s));
+	UNUSED long result = syscall(SYS_write, STDERR_FILENO, s, strlen(s));
 #else
-	UNUSED int result = write(STDERR_FILENO, s, strlen(s));
+	UNUSED ssize_t result = write(STDERR_FILENO, s, strlen(s));
 #endif
 }
 
@@ -82,7 +90,7 @@ buferror(int err, char *buf, size_t buflen)
 
 #ifdef _WIN32
 	FormatMessageA(FORMAT_MESSAGE_FROM_SYSTEM, NULL, err, 0,
-	    (LPSTR)buf, buflen, NULL);
+	    (LPSTR)buf, (DWORD)buflen, NULL);
 	return (0);
 #elif defined(__GLIBC__) && defined(_GNU_SOURCE)
 	char *b = strerror_r(err, buf, buflen);
@@ -577,7 +585,8 @@ malloc_vsnprintf(char *str, size_t size, const char *format, va_list ap)
 		str[i] = '\0';
 	else
 		str[size - 1] = '\0';
-	ret = i;
+	assert(i < INT_MAX);
+	ret = (int)i;
 
 #undef APPEND_C
 #undef APPEND_S
@@ -648,3 +657,12 @@ malloc_printf(const char *format, ...)
 	malloc_vcprintf(NULL, NULL, format, ap);
 	va_end(ap);
 }
+
+/*
+ * Restore normal assertion macros, in order to make it possible to compile all
+ * C files as a single concatenation.
+ */
+#undef assert
+#undef not_reached
+#undef not_implemented
+#include "jemalloc/internal/assert.h"

src/zone.c

@@ -121,9 +121,11 @@ zone_memalign(malloc_zone_t *zone, size_t alignment, size_t size)
 static void
 zone_free_definite_size(malloc_zone_t *zone, void *ptr, size_t size)
 {
+	size_t alloc_size;
 
-	if (ivsalloc(ptr, config_prof) != 0) {
+	alloc_size = ivsalloc(ptr, config_prof);
-		assert(ivsalloc(ptr, config_prof) == size);
+	if (alloc_size != 0) {
+		assert(alloc_size == size);
 		je_free(ptr);
 		return;
 	}

test/include/test/jemalloc_test.h.in

@@ -11,7 +11,6 @@
 #ifdef _WIN32
 #  include "msvc_compat/strings.h"
 #endif
-#include <sys/time.h>
 
 #ifdef _WIN32
 #  include <windows.h>
@@ -94,6 +93,7 @@
 #  define JEMALLOC_H_STRUCTS
 #  define JEMALLOC_H_EXTERNS
 #  define JEMALLOC_H_INLINES
+#  include "jemalloc/internal/nstime.h"
 #  include "jemalloc/internal/util.h"
 #  include "jemalloc/internal/qr.h"
 #  include "jemalloc/internal/ql.h"

test/include/test/timer.h

@@ -1,23 +1,8 @@
 /* Simple timer, for use in benchmark reporting. */
 
-#include <unistd.h>
-#include <sys/time.h>
-
-#define JEMALLOC_CLOCK_GETTIME defined(_POSIX_MONOTONIC_CLOCK) \
-    && _POSIX_MONOTONIC_CLOCK >= 0
-
 typedef struct {
-#ifdef _WIN32
+	nstime_t t0;
-	FILETIME ft0;
+	nstime_t t1;
-	FILETIME ft1;
-#elif JEMALLOC_CLOCK_GETTIME
-	struct timespec ts0;
-	struct timespec ts1;
-	int clock_id;
-#else
-	struct timeval tv0;
-	struct timeval tv1;
-#endif
 } timedelta_t;
 
 void	timer_start(timedelta_t *timer);

test/integration/mallocx.c

@@ -46,6 +46,27 @@ get_huge_size(size_t ind)
 	return (get_size_impl("arenas.hchunk.0.size", ind));
 }
 
+TEST_BEGIN(test_overflow)
+{
+	size_t hugemax;
+
+	hugemax = get_huge_size(get_nhuge()-1);
+
+	assert_ptr_null(mallocx(hugemax+1, 0),
+	    "Expected OOM for mallocx(size=%#zx, 0)", hugemax+1);
+
+	assert_ptr_null(mallocx(ZU(PTRDIFF_MAX)+1, 0),
+	    "Expected OOM for mallocx(size=%#zx, 0)", ZU(PTRDIFF_MAX)+1);
+
+	assert_ptr_null(mallocx(SIZE_T_MAX, 0),
+	    "Expected OOM for mallocx(size=%#zx, 0)", SIZE_T_MAX);
+
+	assert_ptr_null(mallocx(1, MALLOCX_ALIGN(ZU(PTRDIFF_MAX)+1)),
+	    "Expected OOM for mallocx(size=1, MALLOCX_ALIGN(%#zx))",
+	    ZU(PTRDIFF_MAX)+1);
+}
+TEST_END
+
 TEST_BEGIN(test_oom)
 {
 	size_t hugemax, size, alignment;
@@ -176,6 +197,7 @@ main(void)
 {
 
 	return (test(
+	    test_overflow,
 	    test_oom,
 	    test_basic,
 	    test_alignment_and_size));

test/integration/rallocx.c

@@ -1,5 +1,51 @@
 #include "test/jemalloc_test.h"
 
+static unsigned
+get_nsizes_impl(const char *cmd)
+{
+	unsigned ret;
+	size_t z;
+
+	z = sizeof(unsigned);
+	assert_d_eq(mallctl(cmd, &ret, &z, NULL, 0), 0,
+	    "Unexpected mallctl(\"%s\", ...) failure", cmd);
+
+	return (ret);
+}
+
+static unsigned
+get_nhuge(void)
+{
+
+	return (get_nsizes_impl("arenas.nhchunks"));
+}
+
+static size_t
+get_size_impl(const char *cmd, size_t ind)
+{
+	size_t ret;
+	size_t z;
+	size_t mib[4];
+	size_t miblen = 4;
+
+	z = sizeof(size_t);
+	assert_d_eq(mallctlnametomib(cmd, mib, &miblen),
+	    0, "Unexpected mallctlnametomib(\"%s\", ...) failure", cmd);
+	mib[2] = ind;
+	z = sizeof(size_t);
+	assert_d_eq(mallctlbymib(mib, miblen, &ret, &z, NULL, 0),
+	    0, "Unexpected mallctlbymib([\"%s\", %zu], ...) failure", cmd, ind);
+
+	return (ret);
+}
+
+static size_t
+get_huge_size(size_t ind)
+{
+
+	return (get_size_impl("arenas.hchunk.0.size", ind));
+}
+
 TEST_BEGIN(test_grow_and_shrink)
 {
 	void *p, *q;
@@ -138,22 +184,22 @@ TEST_END
 TEST_BEGIN(test_lg_align_and_zero)
 {
 	void *p, *q;
-	size_t lg_align, sz;
+	unsigned lg_align;
+	size_t sz;
 #define	MAX_LG_ALIGN 25
 #define	MAX_VALIDATE (ZU(1) << 22)
 
-	lg_align = ZU(0);
+	lg_align = 0;
 	p = mallocx(1, MALLOCX_LG_ALIGN(lg_align)|MALLOCX_ZERO);
 	assert_ptr_not_null(p, "Unexpected mallocx() error");
 
 	for (lg_align++; lg_align <= MAX_LG_ALIGN; lg_align++) {
 		q = rallocx(p, 1, MALLOCX_LG_ALIGN(lg_align)|MALLOCX_ZERO);
 		assert_ptr_not_null(q,
-		    "Unexpected rallocx() error for lg_align=%zu", lg_align);
+		    "Unexpected rallocx() error for lg_align=%u", lg_align);
 		assert_ptr_null(
 		    (void *)((uintptr_t)q & ((ZU(1) << lg_align)-1)),
-		    "%p inadequately aligned for lg_align=%zu",
+		    "%p inadequately aligned for lg_align=%u", q, lg_align);
-		    q, lg_align);
 		sz = sallocx(q, 0);
 		if ((sz << 1) <= MAX_VALIDATE) {
 			assert_false(validate_fill(q, 0, 0, sz),
@@ -173,6 +219,33 @@ TEST_BEGIN(test_lg_align_and_zero)
 }
 TEST_END
 
+TEST_BEGIN(test_overflow)
+{
+	size_t hugemax;
+	void *p;
+
+	hugemax = get_huge_size(get_nhuge()-1);
+
+	p = mallocx(1, 0);
+	assert_ptr_not_null(p, "Unexpected mallocx() failure");
+
+	assert_ptr_null(rallocx(p, hugemax+1, 0),
+	    "Expected OOM for rallocx(p, size=%#zx, 0)", hugemax+1);
+
+	assert_ptr_null(rallocx(p, ZU(PTRDIFF_MAX)+1, 0),
+	    "Expected OOM for rallocx(p, size=%#zx, 0)", ZU(PTRDIFF_MAX)+1);
+
+	assert_ptr_null(rallocx(p, SIZE_T_MAX, 0),
+	    "Expected OOM for rallocx(p, size=%#zx, 0)", SIZE_T_MAX);
+
+	assert_ptr_null(rallocx(p, 1, MALLOCX_ALIGN(ZU(PTRDIFF_MAX)+1)),
+	    "Expected OOM for rallocx(p, size=1, MALLOCX_ALIGN(%#zx))",
+	    ZU(PTRDIFF_MAX)+1);
+
+	dallocx(p, 0);
+}
+TEST_END
+
 int
 main(void)
 {
@@ -181,5 +254,6 @@ main(void)
 	    test_grow_and_shrink,
 	    test_zero,
 	    test_align,
-	    test_lg_align_and_zero));
+	    test_lg_align_and_zero,
+	    test_overflow));
 }

test/integration/xallocx.c

@@ -305,63 +305,63 @@ TEST_END
 TEST_BEGIN(test_extra_huge)
 {
 	int flags = MALLOCX_ARENA(arena_ind());
-	size_t largemax, huge0, huge1, huge2, hugemax;
+	size_t largemax, huge1, huge2, huge3, hugemax;
 	void *p;
 
 	/* Get size classes. */
 	largemax = get_large_size(get_nlarge()-1);
-	huge0 = get_huge_size(0);
 	huge1 = get_huge_size(1);
 	huge2 = get_huge_size(2);
+	huge3 = get_huge_size(3);
 	hugemax = get_huge_size(get_nhuge()-1);
 
-	p = mallocx(huge2, flags);
+	p = mallocx(huge3, flags);
 	assert_ptr_not_null(p, "Unexpected mallocx() error");
 
-	assert_zu_eq(xallocx(p, huge2, 0, flags), huge2,
+	assert_zu_eq(xallocx(p, huge3, 0, flags), huge3,
 	    "Unexpected xallocx() behavior");
 	/* Test size decrease with zero extra. */
-	assert_zu_ge(xallocx(p, huge0, 0, flags), huge0,
+	assert_zu_ge(xallocx(p, huge1, 0, flags), huge1,
 	    "Unexpected xallocx() behavior");
-	assert_zu_ge(xallocx(p, largemax, 0, flags), huge0,
+	assert_zu_ge(xallocx(p, largemax, 0, flags), huge1,
 	    "Unexpected xallocx() behavior");
 
-	assert_zu_eq(xallocx(p, huge2, 0, flags), huge2,
+	assert_zu_eq(xallocx(p, huge3, 0, flags), huge3,
 	    "Unexpected xallocx() behavior");
 	/* Test size decrease with non-zero extra. */
-	assert_zu_eq(xallocx(p, huge0, huge2 - huge0, flags), huge2,
+	assert_zu_eq(xallocx(p, huge1, huge3 - huge1, flags), huge3,
+	    "Unexpected xallocx() behavior");
+	assert_zu_eq(xallocx(p, huge2, huge3 - huge2, flags), huge3,
 	    "Unexpected xallocx() behavior");
 	assert_zu_eq(xallocx(p, huge1, huge2 - huge1, flags), huge2,
 	    "Unexpected xallocx() behavior");
-	assert_zu_eq(xallocx(p, huge0, huge1 - huge0, flags), huge1,
+	assert_zu_ge(xallocx(p, largemax, huge1 - largemax, flags), huge1,
-	    "Unexpected xallocx() behavior");
-	assert_zu_ge(xallocx(p, largemax, huge0 - largemax, flags), huge0,
 	    "Unexpected xallocx() behavior");
 
-	assert_zu_ge(xallocx(p, huge0, 0, flags), huge0,
+	assert_zu_ge(xallocx(p, huge1, 0, flags), huge1,
 	    "Unexpected xallocx() behavior");
 	/* Test size increase with zero extra. */
-	assert_zu_le(xallocx(p, huge2, 0, flags), huge2,
+	assert_zu_le(xallocx(p, huge3, 0, flags), huge3,
 	    "Unexpected xallocx() behavior");
-	assert_zu_le(xallocx(p, hugemax+1, 0, flags), huge2,
+	assert_zu_le(xallocx(p, hugemax+1, 0, flags), huge3,
 	    "Unexpected xallocx() behavior");
 
-	assert_zu_ge(xallocx(p, huge0, 0, flags), huge0,
+	assert_zu_ge(xallocx(p, huge1, 0, flags), huge1,
 	    "Unexpected xallocx() behavior");
 	/* Test size increase with non-zero extra. */
-	assert_zu_le(xallocx(p, huge0, SIZE_T_MAX - huge0, flags), hugemax,
+	assert_zu_le(xallocx(p, huge1, SIZE_T_MAX - huge1, flags), hugemax,
 	    "Unexpected xallocx() behavior");
 
-	assert_zu_ge(xallocx(p, huge0, 0, flags), huge0,
+	assert_zu_ge(xallocx(p, huge1, 0, flags), huge1,
 	    "Unexpected xallocx() behavior");
 	/* Test size increase with non-zero extra. */
-	assert_zu_le(xallocx(p, huge0, huge2 - huge0, flags), huge2,
+	assert_zu_le(xallocx(p, huge1, huge3 - huge1, flags), huge3,
 	    "Unexpected xallocx() behavior");
 
-	assert_zu_eq(xallocx(p, huge2, 0, flags), huge2,
+	assert_zu_eq(xallocx(p, huge3, 0, flags), huge3,
 	    "Unexpected xallocx() behavior");
 	/* Test size+extra overflow. */
-	assert_zu_le(xallocx(p, huge2, hugemax - huge2 + 1, flags), hugemax,
+	assert_zu_le(xallocx(p, huge3, hugemax - huge3 + 1, flags), hugemax,
 	    "Unexpected xallocx() behavior");
 
 	dallocx(p, flags);

test/src/timer.c

@@ -4,50 +4,26 @@ void
 timer_start(timedelta_t *timer)
 {
 
-#ifdef _WIN32
+	nstime_init(&timer->t0, 0);
-	GetSystemTimeAsFileTime(&timer->ft0);
+	nstime_update(&timer->t0);
-#elif JEMALLOC_CLOCK_GETTIME
-	if (sysconf(_SC_MONOTONIC_CLOCK) <= 0)
-		timer->clock_id = CLOCK_REALTIME;
-	else
-		timer->clock_id = CLOCK_MONOTONIC;
-	clock_gettime(timer->clock_id, &timer->ts0);
-#else
-	gettimeofday(&timer->tv0, NULL);
-#endif
 }
 
 void
 timer_stop(timedelta_t *timer)
 {
 
-#ifdef _WIN32
+	nstime_copy(&timer->t1, &timer->t0);
-	GetSystemTimeAsFileTime(&timer->ft0);
+	nstime_update(&timer->t1);
-#elif JEMALLOC_CLOCK_GETTIME
-	clock_gettime(timer->clock_id, &timer->ts1);
-#else
-	gettimeofday(&timer->tv1, NULL);
-#endif
 }
 
 uint64_t
 timer_usec(const timedelta_t *timer)
 {
+	nstime_t delta;
 
-#ifdef _WIN32
+	nstime_copy(&delta, &timer->t1);
-	uint64_t t0, t1;
+	nstime_subtract(&delta, &timer->t0);
-	t0 = (((uint64_t)timer->ft0.dwHighDateTime) << 32) |
+	return (nstime_ns(&delta) / 1000);
-	    timer->ft0.dwLowDateTime;
-	t1 = (((uint64_t)timer->ft1.dwHighDateTime) << 32) |
-	    timer->ft1.dwLowDateTime;
-	return ((t1 - t0) / 10);
-#elif JEMALLOC_CLOCK_GETTIME
-	return (((timer->ts1.tv_sec - timer->ts0.tv_sec) * 1000000) +
-	    (timer->ts1.tv_nsec - timer->ts0.tv_nsec) / 1000);
-#else
-	return (((timer->tv1.tv_sec - timer->tv0.tv_sec) * 1000000) +
-	    timer->tv1.tv_usec - timer->tv0.tv_usec);
-#endif
 }
 
 void

test/unit/bitmap.c

@@ -6,7 +6,11 @@ TEST_BEGIN(test_bitmap_size)
 
 	prev_size = 0;
 	for (i = 1; i <= BITMAP_MAXBITS; i++) {
-		size_t size = bitmap_size(i);
+		bitmap_info_t binfo;
+		size_t size;
+
+		bitmap_info_init(&binfo, i);
+		size = bitmap_size(&binfo);
 		assert_true(size >= prev_size,
 		    "Bitmap size is smaller than expected");
 		prev_size = size;
@@ -23,8 +27,8 @@ TEST_BEGIN(test_bitmap_init)
 		bitmap_info_init(&binfo, i);
 		{
 			size_t j;
-			bitmap_t *bitmap = (bitmap_t *)malloc(sizeof(bitmap_t) *
+			bitmap_t *bitmap = (bitmap_t *)malloc(
-				bitmap_info_ngroups(&binfo));
+			    bitmap_size(&binfo));
 			bitmap_init(bitmap, &binfo);
 
 			for (j = 0; j < i; j++) {
@@ -46,8 +50,8 @@ TEST_BEGIN(test_bitmap_set)
 		bitmap_info_init(&binfo, i);
 		{
 			size_t j;
-			bitmap_t *bitmap = (bitmap_t *)malloc(sizeof(bitmap_t) *
+			bitmap_t *bitmap = (bitmap_t *)malloc(
-				bitmap_info_ngroups(&binfo));
+			    bitmap_size(&binfo));
 			bitmap_init(bitmap, &binfo);
 
 			for (j = 0; j < i; j++)
@@ -69,8 +73,8 @@ TEST_BEGIN(test_bitmap_unset)
 		bitmap_info_init(&binfo, i);
 		{
 			size_t j;
-			bitmap_t *bitmap = (bitmap_t *)malloc(sizeof(bitmap_t) *
+			bitmap_t *bitmap = (bitmap_t *)malloc(
-				bitmap_info_ngroups(&binfo));
+			    bitmap_size(&binfo));
 			bitmap_init(bitmap, &binfo);
 
 			for (j = 0; j < i; j++)
@@ -98,8 +102,8 @@ TEST_BEGIN(test_bitmap_sfu)
 		bitmap_info_init(&binfo, i);
 		{
 			ssize_t j;
-			bitmap_t *bitmap = (bitmap_t *)malloc(sizeof(bitmap_t) *
+			bitmap_t *bitmap = (bitmap_t *)malloc(
-				bitmap_info_ngroups(&binfo));
+			    bitmap_size(&binfo));
 			bitmap_init(bitmap, &binfo);
 
 			/* Iteratively set bits starting at the beginning. */

test/unit/decay.c

@@ -0,0 +1,360 @@
+#include "test/jemalloc_test.h"
+
+const char *malloc_conf = "purge:decay,decay_time:1";
+
+static nstime_update_t *nstime_update_orig;
+
+static unsigned nupdates_mock;
+static nstime_t time_mock;
+static bool nonmonotonic_mock;
+
+static bool
+nstime_update_mock(nstime_t *time)
+{
+
+	nupdates_mock++;
+	if (!nonmonotonic_mock)
+		nstime_copy(time, &time_mock);
+	return (nonmonotonic_mock);
+}
+
+TEST_BEGIN(test_decay_ticks)
+{
+	ticker_t *decay_ticker;
+	unsigned tick0, tick1;
+	size_t sz, huge0, large0;
+	void *p;
+
+	test_skip_if(opt_purge != purge_mode_decay);
+
+	decay_ticker = decay_ticker_get(tsd_fetch(), 0);
+	assert_ptr_not_null(decay_ticker,
+	    "Unexpected failure getting decay ticker");
+
+	sz = sizeof(size_t);
+	assert_d_eq(mallctl("arenas.hchunk.0.size", &huge0, &sz, NULL, 0), 0,
+	    "Unexpected mallctl failure");
+	assert_d_eq(mallctl("arenas.lrun.0.size", &large0, &sz, NULL, 0), 0,
+	    "Unexpected mallctl failure");
+
+	/*
+	 * Test the standard APIs using a huge size class, since we can't
+	 * control tcache interactions (except by completely disabling tcache
+	 * for the entire test program).
+	 */
+
+	/* malloc(). */
+	tick0 = ticker_read(decay_ticker);
+	p = malloc(huge0);
+	assert_ptr_not_null(p, "Unexpected malloc() failure");
+	tick1 = ticker_read(decay_ticker);
+	assert_u32_ne(tick1, tick0, "Expected ticker to tick during malloc()");
+	/* free(). */
+	tick0 = ticker_read(decay_ticker);
+	free(p);
+	tick1 = ticker_read(decay_ticker);
+	assert_u32_ne(tick1, tick0, "Expected ticker to tick during free()");
+
+	/* calloc(). */
+	tick0 = ticker_read(decay_ticker);
+	p = calloc(1, huge0);
+	assert_ptr_not_null(p, "Unexpected calloc() failure");
+	tick1 = ticker_read(decay_ticker);
+	assert_u32_ne(tick1, tick0, "Expected ticker to tick during calloc()");
+	free(p);
+
+	/* posix_memalign(). */
+	tick0 = ticker_read(decay_ticker);
+	assert_d_eq(posix_memalign(&p, sizeof(size_t), huge0), 0,
+	    "Unexpected posix_memalign() failure");
+	tick1 = ticker_read(decay_ticker);
+	assert_u32_ne(tick1, tick0,
+	    "Expected ticker to tick during posix_memalign()");
+	free(p);
+
+	/* aligned_alloc(). */
+	tick0 = ticker_read(decay_ticker);
+	p = aligned_alloc(sizeof(size_t), huge0);
+	assert_ptr_not_null(p, "Unexpected aligned_alloc() failure");
+	tick1 = ticker_read(decay_ticker);
+	assert_u32_ne(tick1, tick0,
+	    "Expected ticker to tick during aligned_alloc()");
+	free(p);
+
+	/* realloc(). */
+	/* Allocate. */
+	tick0 = ticker_read(decay_ticker);
+	p = realloc(NULL, huge0);
+	assert_ptr_not_null(p, "Unexpected realloc() failure");
+	tick1 = ticker_read(decay_ticker);
+	assert_u32_ne(tick1, tick0, "Expected ticker to tick during realloc()");
+	/* Reallocate. */
+	tick0 = ticker_read(decay_ticker);
+	p = realloc(p, huge0);
+	assert_ptr_not_null(p, "Unexpected realloc() failure");
+	tick1 = ticker_read(decay_ticker);
+	assert_u32_ne(tick1, tick0, "Expected ticker to tick during realloc()");
+	/* Deallocate. */
+	tick0 = ticker_read(decay_ticker);
+	realloc(p, 0);
+	tick1 = ticker_read(decay_ticker);
+	assert_u32_ne(tick1, tick0, "Expected ticker to tick during realloc()");
+
+	/*
+	 * Test the *allocx() APIs using huge, large, and small size classes,
+	 * with tcache explicitly disabled.
+	 */
+	{
+		unsigned i;
+		size_t allocx_sizes[3];
+		allocx_sizes[0] = huge0;
+		allocx_sizes[1] = large0;
+		allocx_sizes[2] = 1;
+
+		for (i = 0; i < sizeof(allocx_sizes) / sizeof(size_t); i++) {
+			sz = allocx_sizes[i];
+
+			/* mallocx(). */
+			tick0 = ticker_read(decay_ticker);
+			p = mallocx(sz, MALLOCX_TCACHE_NONE);
+			assert_ptr_not_null(p, "Unexpected mallocx() failure");
+			tick1 = ticker_read(decay_ticker);
+			assert_u32_ne(tick1, tick0,
+			    "Expected ticker to tick during mallocx() (sz=%zu)",
+			    sz);
+			/* rallocx(). */
+			tick0 = ticker_read(decay_ticker);
+			p = rallocx(p, sz, MALLOCX_TCACHE_NONE);
+			assert_ptr_not_null(p, "Unexpected rallocx() failure");
+			tick1 = ticker_read(decay_ticker);
+			assert_u32_ne(tick1, tick0,
+			    "Expected ticker to tick during rallocx() (sz=%zu)",
+			    sz);
+			/* xallocx(). */
+			tick0 = ticker_read(decay_ticker);
+			xallocx(p, sz, 0, MALLOCX_TCACHE_NONE);
+			tick1 = ticker_read(decay_ticker);
+			assert_u32_ne(tick1, tick0,
+			    "Expected ticker to tick during xallocx() (sz=%zu)",
+			    sz);
+			/* dallocx(). */
+			tick0 = ticker_read(decay_ticker);
+			dallocx(p, MALLOCX_TCACHE_NONE);
+			tick1 = ticker_read(decay_ticker);
+			assert_u32_ne(tick1, tick0,
+			    "Expected ticker to tick during dallocx() (sz=%zu)",
+			    sz);
+			/* sdallocx(). */
+			p = mallocx(sz, MALLOCX_TCACHE_NONE);
+			assert_ptr_not_null(p, "Unexpected mallocx() failure");
+			tick0 = ticker_read(decay_ticker);
+			sdallocx(p, sz, MALLOCX_TCACHE_NONE);
+			tick1 = ticker_read(decay_ticker);
+			assert_u32_ne(tick1, tick0,
+			    "Expected ticker to tick during sdallocx() "
+			    "(sz=%zu)", sz);
+		}
+	}
+
+	/*
+	 * Test tcache fill/flush interactions for large and small size classes,
+	 * using an explicit tcache.
+	 */
+	if (config_tcache) {
+		unsigned tcache_ind, i;
+		size_t tcache_sizes[2];
+		tcache_sizes[0] = large0;
+		tcache_sizes[1] = 1;
+
+		sz = sizeof(unsigned);
+		assert_d_eq(mallctl("tcache.create", &tcache_ind, &sz, NULL, 0),
+		    0, "Unexpected mallctl failure");
+
+		for (i = 0; i < sizeof(tcache_sizes) / sizeof(size_t); i++) {
+			sz = tcache_sizes[i];
+
+			/* tcache fill. */
+			tick0 = ticker_read(decay_ticker);
+			p = mallocx(sz, MALLOCX_TCACHE(tcache_ind));
+			assert_ptr_not_null(p, "Unexpected mallocx() failure");
+			tick1 = ticker_read(decay_ticker);
+			assert_u32_ne(tick1, tick0,
+			    "Expected ticker to tick during tcache fill "
+			    "(sz=%zu)", sz);
+			/* tcache flush. */
+			dallocx(p, MALLOCX_TCACHE(tcache_ind));
+			tick0 = ticker_read(decay_ticker);
+			assert_d_eq(mallctl("tcache.flush", NULL, NULL,
+			    &tcache_ind, sizeof(unsigned)), 0,
+			    "Unexpected mallctl failure");
+			tick1 = ticker_read(decay_ticker);
+			assert_u32_ne(tick1, tick0,
+			    "Expected ticker to tick during tcache flush "
+			    "(sz=%zu)", sz);
+		}
+	}
+}
+TEST_END
+
+TEST_BEGIN(test_decay_ticker)
+{
+#define	NPS 1024
+	int flags = (MALLOCX_ARENA(0) | MALLOCX_TCACHE_NONE);
+	void *ps[NPS];
+	uint64_t epoch;
+	uint64_t npurge0 = 0;
+	uint64_t npurge1 = 0;
+	size_t sz, large;
+	unsigned i, nupdates0;
+	nstime_t time, decay_time, deadline;
+
+	test_skip_if(opt_purge != purge_mode_decay);
+
+	/*
+	 * Allocate a bunch of large objects, pause the clock, deallocate the
+	 * objects, restore the clock, then [md]allocx() in a tight loop to
+	 * verify the ticker triggers purging.
+	 */
+
+	if (config_tcache) {
+		size_t tcache_max;
+
+		sz = sizeof(size_t);
+		assert_d_eq(mallctl("arenas.tcache_max", &tcache_max, &sz, NULL,
+		    0), 0, "Unexpected mallctl failure");
+		large = nallocx(tcache_max + 1, flags);
+	}  else {
+		sz = sizeof(size_t);
+		assert_d_eq(mallctl("arenas.lrun.0.size", &large, &sz, NULL, 0),
+		    0, "Unexpected mallctl failure");
+	}
+
+	assert_d_eq(mallctl("arena.0.purge", NULL, NULL, NULL, 0), 0,
+	    "Unexpected mallctl failure");
+	assert_d_eq(mallctl("epoch", NULL, NULL, &epoch, sizeof(uint64_t)), 0,
+	    "Unexpected mallctl failure");
+	sz = sizeof(uint64_t);
+	assert_d_eq(mallctl("stats.arenas.0.npurge", &npurge0, &sz, NULL, 0),
+	    config_stats ? 0 : ENOENT, "Unexpected mallctl result");
+
+	for (i = 0; i < NPS; i++) {
+		ps[i] = mallocx(large, flags);
+		assert_ptr_not_null(ps[i], "Unexpected mallocx() failure");
+	}
+
+	nupdates_mock = 0;
+	nstime_init(&time_mock, 0);
+	nstime_update(&time_mock);
+	nonmonotonic_mock = false;
+
+	nstime_update_orig = nstime_update;
+	nstime_update = nstime_update_mock;
+
+	for (i = 0; i < NPS; i++) {
+		dallocx(ps[i], flags);
+		nupdates0 = nupdates_mock;
+		assert_d_eq(mallctl("arena.0.decay", NULL, NULL, NULL, 0), 0,
+		    "Unexpected arena.0.decay failure");
+		assert_u_gt(nupdates_mock, nupdates0,
+		    "Expected nstime_update() to be called");
+	}
+
+	nstime_update = nstime_update_orig;
+
+	nstime_init(&time, 0);
+	nstime_update(&time);
+	nstime_init2(&decay_time, opt_decay_time, 0);
+	nstime_copy(&deadline, &time);
+	nstime_add(&deadline, &decay_time);
+	do {
+		for (i = 0; i < DECAY_NTICKS_PER_UPDATE / 2; i++) {
+			void *p = mallocx(1, flags);
+			assert_ptr_not_null(p, "Unexpected mallocx() failure");
+			dallocx(p, flags);
+		}
+		assert_d_eq(mallctl("epoch", NULL, NULL, &epoch,
+		    sizeof(uint64_t)), 0, "Unexpected mallctl failure");
+		sz = sizeof(uint64_t);
+		assert_d_eq(mallctl("stats.arenas.0.npurge", &npurge1, &sz,
+		    NULL, 0), config_stats ? 0 : ENOENT,
+		    "Unexpected mallctl result");
+
+		nstime_update(&time);
+	} while (nstime_compare(&time, &deadline) <= 0 && npurge1 == npurge0);
+
+	if (config_stats)
+		assert_u64_gt(npurge1, npurge0, "Expected purging to occur");
+#undef NPS
+}
+TEST_END
+
+TEST_BEGIN(test_decay_nonmonotonic)
+{
+#define	NPS (SMOOTHSTEP_NSTEPS + 1)
+	int flags = (MALLOCX_ARENA(0) | MALLOCX_TCACHE_NONE);
+	void *ps[NPS];
+	uint64_t epoch;
+	uint64_t npurge0 = 0;
+	uint64_t npurge1 = 0;
+	size_t sz, large0;
+	unsigned i, nupdates0;
+
+	test_skip_if(opt_purge != purge_mode_decay);
+
+	sz = sizeof(size_t);
+	assert_d_eq(mallctl("arenas.lrun.0.size", &large0, &sz, NULL, 0), 0,
+	    "Unexpected mallctl failure");
+
+	assert_d_eq(mallctl("arena.0.purge", NULL, NULL, NULL, 0), 0,
+	    "Unexpected mallctl failure");
+	assert_d_eq(mallctl("epoch", NULL, NULL, &epoch, sizeof(uint64_t)), 0,
+	    "Unexpected mallctl failure");
+	sz = sizeof(uint64_t);
+	assert_d_eq(mallctl("stats.arenas.0.npurge", &npurge0, &sz, NULL, 0),
+	    config_stats ? 0 : ENOENT, "Unexpected mallctl result");
+
+	nupdates_mock = 0;
+	nstime_init(&time_mock, 0);
+	nstime_update(&time_mock);
+	nonmonotonic_mock = true;
+
+	nstime_update_orig = nstime_update;
+	nstime_update = nstime_update_mock;
+
+	for (i = 0; i < NPS; i++) {
+		ps[i] = mallocx(large0, flags);
+		assert_ptr_not_null(ps[i], "Unexpected mallocx() failure");
+	}
+
+	for (i = 0; i < NPS; i++) {
+		dallocx(ps[i], flags);
+		nupdates0 = nupdates_mock;
+		assert_d_eq(mallctl("arena.0.decay", NULL, NULL, NULL, 0), 0,
+		    "Unexpected arena.0.decay failure");
+		assert_u_gt(nupdates_mock, nupdates0,
+		    "Expected nstime_update() to be called");
+	}
+
+	assert_d_eq(mallctl("epoch", NULL, NULL, &epoch, sizeof(uint64_t)), 0,
+	    "Unexpected mallctl failure");
+	sz = sizeof(uint64_t);
+	assert_d_eq(mallctl("stats.arenas.0.npurge", &npurge1, &sz, NULL, 0),
+	    config_stats ? 0 : ENOENT, "Unexpected mallctl result");
+
+	if (config_stats)
+		assert_u64_gt(npurge1, npurge0, "Expected purging to occur");
+
+	nstime_update = nstime_update_orig;
+#undef NPS
+}
+TEST_END
+
+int
+main(void)
+{
+
+	return (test(
+	    test_decay_ticks,
+	    test_decay_ticker,
+	    test_decay_nonmonotonic));
+}

test/unit/hash.c

@@ -35,7 +35,7 @@ typedef enum {
 	hash_variant_x64_128
 } hash_variant_t;
 
-static size_t
+static int
 hash_variant_bits(hash_variant_t variant)
 {
 
@@ -59,17 +59,17 @@ hash_variant_string(hash_variant_t variant)
 	}
 }
 
+#define	KEY_SIZE	256
 static void
-hash_variant_verify(hash_variant_t variant)
+hash_variant_verify_key(hash_variant_t variant, uint8_t *key)
 {
-	const size_t hashbytes = hash_variant_bits(variant) / 8;
+	const int hashbytes = hash_variant_bits(variant) / 8;
-	uint8_t key[256];
 	VARIABLE_ARRAY(uint8_t, hashes, hashbytes * 256);
 	VARIABLE_ARRAY(uint8_t, final, hashbytes);
 	unsigned i;
 	uint32_t computed, expected;
 
-	memset(key, 0, sizeof(key));
+	memset(key, 0, KEY_SIZE);
 	memset(hashes, 0, sizeof(hashes));
 	memset(final, 0, sizeof(final));
 
@@ -139,6 +139,19 @@ hash_variant_verify(hash_variant_t variant)
 	    hash_variant_string(variant), expected, computed);
 }
 
+static void
+hash_variant_verify(hash_variant_t variant)
+{
+#define	MAX_ALIGN	16
+	uint8_t key[KEY_SIZE + (MAX_ALIGN - 1)];
+	unsigned i;
+
+	for (i = 0; i < MAX_ALIGN; i++)
+		hash_variant_verify_key(variant, &key[i]);
+#undef MAX_ALIGN
+}
+#undef KEY_SIZE
+
 TEST_BEGIN(test_hash_x86_32)
 {
 

test/unit/mallctl.c

@@ -117,8 +117,8 @@ TEST_END
 TEST_BEGIN(test_mallctl_config)
 {
 
-#define	TEST_MALLCTL_CONFIG(config) do {				\
+#define	TEST_MALLCTL_CONFIG(config, t) do {				\
-	bool oldval;							\
+	t oldval;							\
 	size_t sz = sizeof(oldval);					\
 	assert_d_eq(mallctl("config."#config, &oldval, &sz, NULL, 0),	\
 	    0, "Unexpected mallctl() failure");				\
@@ -126,20 +126,21 @@ TEST_BEGIN(test_mallctl_config)
 	assert_zu_eq(sz, sizeof(oldval), "Unexpected output size");	\
 } while (0)
 
-	TEST_MALLCTL_CONFIG(cache_oblivious);
+	TEST_MALLCTL_CONFIG(cache_oblivious, bool);
-	TEST_MALLCTL_CONFIG(debug);
+	TEST_MALLCTL_CONFIG(debug, bool);
-	TEST_MALLCTL_CONFIG(fill);
+	TEST_MALLCTL_CONFIG(fill, bool);
-	TEST_MALLCTL_CONFIG(lazy_lock);
+	TEST_MALLCTL_CONFIG(lazy_lock, bool);
-	TEST_MALLCTL_CONFIG(munmap);
+	TEST_MALLCTL_CONFIG(malloc_conf, const char *);
-	TEST_MALLCTL_CONFIG(prof);
+	TEST_MALLCTL_CONFIG(munmap, bool);
-	TEST_MALLCTL_CONFIG(prof_libgcc);
+	TEST_MALLCTL_CONFIG(prof, bool);
-	TEST_MALLCTL_CONFIG(prof_libunwind);
+	TEST_MALLCTL_CONFIG(prof_libgcc, bool);
-	TEST_MALLCTL_CONFIG(stats);
+	TEST_MALLCTL_CONFIG(prof_libunwind, bool);
-	TEST_MALLCTL_CONFIG(tcache);
+	TEST_MALLCTL_CONFIG(stats, bool);
-	TEST_MALLCTL_CONFIG(tls);
+	TEST_MALLCTL_CONFIG(tcache, bool);
-	TEST_MALLCTL_CONFIG(utrace);
+	TEST_MALLCTL_CONFIG(tls, bool);
-	TEST_MALLCTL_CONFIG(valgrind);
+	TEST_MALLCTL_CONFIG(utrace, bool);
-	TEST_MALLCTL_CONFIG(xmalloc);
+	TEST_MALLCTL_CONFIG(valgrind, bool);
+	TEST_MALLCTL_CONFIG(xmalloc, bool);
 
 #undef TEST_MALLCTL_CONFIG
 }
@@ -162,8 +163,10 @@ TEST_BEGIN(test_mallctl_opt)
 	TEST_MALLCTL_OPT(bool, abort, always);
 	TEST_MALLCTL_OPT(size_t, lg_chunk, always);
 	TEST_MALLCTL_OPT(const char *, dss, always);
-	TEST_MALLCTL_OPT(size_t, narenas, always);
+	TEST_MALLCTL_OPT(unsigned, narenas, always);
+	TEST_MALLCTL_OPT(const char *, purge, always);
 	TEST_MALLCTL_OPT(ssize_t, lg_dirty_mult, always);
+	TEST_MALLCTL_OPT(ssize_t, decay_time, always);
 	TEST_MALLCTL_OPT(bool, stats_print, always);
 	TEST_MALLCTL_OPT(const char *, junk, fill);
 	TEST_MALLCTL_OPT(size_t, quarantine, fill);
@@ -354,6 +357,8 @@ TEST_BEGIN(test_arena_i_lg_dirty_mult)
 	ssize_t lg_dirty_mult, orig_lg_dirty_mult, prev_lg_dirty_mult;
 	size_t sz = sizeof(ssize_t);
 
+	test_skip_if(opt_purge != purge_mode_ratio);
+
 	assert_d_eq(mallctl("arena.0.lg_dirty_mult", &orig_lg_dirty_mult, &sz,
 	    NULL, 0), 0, "Unexpected mallctl() failure");
 
@@ -381,6 +386,39 @@ TEST_BEGIN(test_arena_i_lg_dirty_mult)
 }
 TEST_END
 
+TEST_BEGIN(test_arena_i_decay_time)
+{
+	ssize_t decay_time, orig_decay_time, prev_decay_time;
+	size_t sz = sizeof(ssize_t);
+
+	test_skip_if(opt_purge != purge_mode_decay);
+
+	assert_d_eq(mallctl("arena.0.decay_time", &orig_decay_time, &sz,
+	    NULL, 0), 0, "Unexpected mallctl() failure");
+
+	decay_time = -2;
+	assert_d_eq(mallctl("arena.0.decay_time", NULL, NULL,
+	    &decay_time, sizeof(ssize_t)), EFAULT,
+	    "Unexpected mallctl() success");
+
+	decay_time = 0x7fffffff;
+	assert_d_eq(mallctl("arena.0.decay_time", NULL, NULL,
+	    &decay_time, sizeof(ssize_t)), 0,
+	    "Unexpected mallctl() failure");
+
+	for (prev_decay_time = decay_time, decay_time = -1;
+	    decay_time < 20; prev_decay_time = decay_time, decay_time++) {
+		ssize_t old_decay_time;
+
+		assert_d_eq(mallctl("arena.0.decay_time", &old_decay_time,
+		    &sz, &decay_time, sizeof(ssize_t)), 0,
+		    "Unexpected mallctl() failure");
+		assert_zd_eq(old_decay_time, prev_decay_time,
+		    "Unexpected old arena.0.decay_time");
+	}
+}
+TEST_END
+
 TEST_BEGIN(test_arena_i_purge)
 {
 	unsigned narenas;
@@ -401,6 +439,26 @@ TEST_BEGIN(test_arena_i_purge)
 }
 TEST_END
 
+TEST_BEGIN(test_arena_i_decay)
+{
+	unsigned narenas;
+	size_t sz = sizeof(unsigned);
+	size_t mib[3];
+	size_t miblen = 3;
+
+	assert_d_eq(mallctl("arena.0.decay", NULL, NULL, NULL, 0), 0,
+	    "Unexpected mallctl() failure");
+
+	assert_d_eq(mallctl("arenas.narenas", &narenas, &sz, NULL, 0), 0,
+	    "Unexpected mallctl() failure");
+	assert_d_eq(mallctlnametomib("arena.0.decay", mib, &miblen), 0,
+	    "Unexpected mallctlnametomib() failure");
+	mib[1] = narenas;
+	assert_d_eq(mallctlbymib(mib, miblen, NULL, NULL, NULL, 0), 0,
+	    "Unexpected mallctlbymib() failure");
+}
+TEST_END
+
 TEST_BEGIN(test_arena_i_dss)
 {
 	const char *dss_prec_old, *dss_prec_new;
@@ -465,6 +523,8 @@ TEST_BEGIN(test_arenas_lg_dirty_mult)
 	ssize_t lg_dirty_mult, orig_lg_dirty_mult, prev_lg_dirty_mult;
 	size_t sz = sizeof(ssize_t);
 
+	test_skip_if(opt_purge != purge_mode_ratio);
+
 	assert_d_eq(mallctl("arenas.lg_dirty_mult", &orig_lg_dirty_mult, &sz,
 	    NULL, 0), 0, "Unexpected mallctl() failure");
 
@@ -492,6 +552,39 @@ TEST_BEGIN(test_arenas_lg_dirty_mult)
 }
 TEST_END
 
+TEST_BEGIN(test_arenas_decay_time)
+{
+	ssize_t decay_time, orig_decay_time, prev_decay_time;
+	size_t sz = sizeof(ssize_t);
+
+	test_skip_if(opt_purge != purge_mode_decay);
+
+	assert_d_eq(mallctl("arenas.decay_time", &orig_decay_time, &sz,
+	    NULL, 0), 0, "Unexpected mallctl() failure");
+
+	decay_time = -2;
+	assert_d_eq(mallctl("arenas.decay_time", NULL, NULL,
+	    &decay_time, sizeof(ssize_t)), EFAULT,
+	    "Unexpected mallctl() success");
+
+	decay_time = 0x7fffffff;
+	assert_d_eq(mallctl("arenas.decay_time", NULL, NULL,
+	    &decay_time, sizeof(ssize_t)), 0,
+	    "Expected mallctl() failure");
+
+	for (prev_decay_time = decay_time, decay_time = -1;
+	    decay_time < 20; prev_decay_time = decay_time, decay_time++) {
+		ssize_t old_decay_time;
+
+		assert_d_eq(mallctl("arenas.decay_time", &old_decay_time,
+		    &sz, &decay_time, sizeof(ssize_t)), 0,
+		    "Unexpected mallctl() failure");
+		assert_zd_eq(old_decay_time, prev_decay_time,
+		    "Unexpected old arenas.decay_time");
+	}
+}
+TEST_END
+
 TEST_BEGIN(test_arenas_constants)
 {
 
@@ -594,8 +687,10 @@ TEST_BEGIN(test_stats_arenas)
 	    0), 0, "Unexpected mallctl() failure");			\
 } while (0)
 
-	TEST_STATS_ARENAS(const char *, dss);
 	TEST_STATS_ARENAS(unsigned, nthreads);
+	TEST_STATS_ARENAS(const char *, dss);
+	TEST_STATS_ARENAS(ssize_t, lg_dirty_mult);
+	TEST_STATS_ARENAS(ssize_t, decay_time);
 	TEST_STATS_ARENAS(size_t, pactive);
 	TEST_STATS_ARENAS(size_t, pdirty);
 
@@ -620,10 +715,13 @@ main(void)
 	    test_tcache,
 	    test_thread_arena,
 	    test_arena_i_lg_dirty_mult,
+	    test_arena_i_decay_time,
 	    test_arena_i_purge,
+	    test_arena_i_decay,
 	    test_arena_i_dss,
 	    test_arenas_initialized,
 	    test_arenas_lg_dirty_mult,
+	    test_arenas_decay_time,
 	    test_arenas_constants,
 	    test_arenas_bin_constants,
 	    test_arenas_lrun_constants,

test/unit/nstime.c

@@ -0,0 +1,220 @@
+#include "test/jemalloc_test.h"
+
+#define	BILLION	UINT64_C(1000000000)
+
+TEST_BEGIN(test_nstime_init)
+{
+	nstime_t nst;
+
+	nstime_init(&nst, 42000000043);
+	assert_u64_eq(nstime_ns(&nst), 42000000043, "ns incorrectly read");
+	assert_u64_eq(nstime_sec(&nst), 42, "sec incorrectly read");
+	assert_u64_eq(nstime_nsec(&nst), 43, "nsec incorrectly read");
+}
+TEST_END
+
+TEST_BEGIN(test_nstime_init2)
+{
+	nstime_t nst;
+
+	nstime_init2(&nst, 42, 43);
+	assert_u64_eq(nstime_sec(&nst), 42, "sec incorrectly read");
+	assert_u64_eq(nstime_nsec(&nst), 43, "nsec incorrectly read");
+}
+TEST_END
+
+TEST_BEGIN(test_nstime_copy)
+{
+	nstime_t nsta, nstb;
+
+	nstime_init2(&nsta, 42, 43);
+	nstime_init(&nstb, 0);
+	nstime_copy(&nstb, &nsta);
+	assert_u64_eq(nstime_sec(&nstb), 42, "sec incorrectly copied");
+	assert_u64_eq(nstime_nsec(&nstb), 43, "nsec incorrectly copied");
+}
+TEST_END
+
+TEST_BEGIN(test_nstime_compare)
+{
+	nstime_t nsta, nstb;
+
+	nstime_init2(&nsta, 42, 43);
+	nstime_copy(&nstb, &nsta);
+	assert_d_eq(nstime_compare(&nsta, &nstb), 0, "Times should be equal");
+	assert_d_eq(nstime_compare(&nstb, &nsta), 0, "Times should be equal");
+
+	nstime_init2(&nstb, 42, 42);
+	assert_d_eq(nstime_compare(&nsta, &nstb), 1,
+	    "nsta should be greater than nstb");
+	assert_d_eq(nstime_compare(&nstb, &nsta), -1,
+	    "nstb should be less than nsta");
+
+	nstime_init2(&nstb, 42, 44);
+	assert_d_eq(nstime_compare(&nsta, &nstb), -1,
+	    "nsta should be less than nstb");
+	assert_d_eq(nstime_compare(&nstb, &nsta), 1,
+	    "nstb should be greater than nsta");
+
+	nstime_init2(&nstb, 41, BILLION - 1);
+	assert_d_eq(nstime_compare(&nsta, &nstb), 1,
+	    "nsta should be greater than nstb");
+	assert_d_eq(nstime_compare(&nstb, &nsta), -1,
+	    "nstb should be less than nsta");
+
+	nstime_init2(&nstb, 43, 0);
+	assert_d_eq(nstime_compare(&nsta, &nstb), -1,
+	    "nsta should be less than nstb");
+	assert_d_eq(nstime_compare(&nstb, &nsta), 1,
+	    "nstb should be greater than nsta");
+}
+TEST_END
+
+TEST_BEGIN(test_nstime_add)
+{
+	nstime_t nsta, nstb;
+
+	nstime_init2(&nsta, 42, 43);
+	nstime_copy(&nstb, &nsta);
+	nstime_add(&nsta, &nstb);
+	nstime_init2(&nstb, 84, 86);
+	assert_d_eq(nstime_compare(&nsta, &nstb), 0,
+	    "Incorrect addition result");
+
+	nstime_init2(&nsta, 42, BILLION - 1);
+	nstime_copy(&nstb, &nsta);
+	nstime_add(&nsta, &nstb);
+	nstime_init2(&nstb, 85, BILLION - 2);
+	assert_d_eq(nstime_compare(&nsta, &nstb), 0,
+	    "Incorrect addition result");
+}
+TEST_END
+
+TEST_BEGIN(test_nstime_subtract)
+{
+	nstime_t nsta, nstb;
+
+	nstime_init2(&nsta, 42, 43);
+	nstime_copy(&nstb, &nsta);
+	nstime_subtract(&nsta, &nstb);
+	nstime_init(&nstb, 0);
+	assert_d_eq(nstime_compare(&nsta, &nstb), 0,
+	    "Incorrect subtraction result");
+
+	nstime_init2(&nsta, 42, 43);
+	nstime_init2(&nstb, 41, 44);
+	nstime_subtract(&nsta, &nstb);
+	nstime_init2(&nstb, 0, BILLION - 1);
+	assert_d_eq(nstime_compare(&nsta, &nstb), 0,
+	    "Incorrect subtraction result");
+}
+TEST_END
+
+TEST_BEGIN(test_nstime_imultiply)
+{
+	nstime_t nsta, nstb;
+
+	nstime_init2(&nsta, 42, 43);
+	nstime_imultiply(&nsta, 10);
+	nstime_init2(&nstb, 420, 430);
+	assert_d_eq(nstime_compare(&nsta, &nstb), 0,
+	    "Incorrect multiplication result");
+
+	nstime_init2(&nsta, 42, 666666666);
+	nstime_imultiply(&nsta, 3);
+	nstime_init2(&nstb, 127, 999999998);
+	assert_d_eq(nstime_compare(&nsta, &nstb), 0,
+	    "Incorrect multiplication result");
+}
+TEST_END
+
+TEST_BEGIN(test_nstime_idivide)
+{
+	nstime_t nsta, nstb;
+
+	nstime_init2(&nsta, 42, 43);
+	nstime_copy(&nstb, &nsta);
+	nstime_imultiply(&nsta, 10);
+	nstime_idivide(&nsta, 10);
+	assert_d_eq(nstime_compare(&nsta, &nstb), 0,
+	    "Incorrect division result");
+
+	nstime_init2(&nsta, 42, 666666666);
+	nstime_copy(&nstb, &nsta);
+	nstime_imultiply(&nsta, 3);
+	nstime_idivide(&nsta, 3);
+	assert_d_eq(nstime_compare(&nsta, &nstb), 0,
+	    "Incorrect division result");
+}
+TEST_END
+
+TEST_BEGIN(test_nstime_divide)
+{
+	nstime_t nsta, nstb, nstc;
+
+	nstime_init2(&nsta, 42, 43);
+	nstime_copy(&nstb, &nsta);
+	nstime_imultiply(&nsta, 10);
+	assert_u64_eq(nstime_divide(&nsta, &nstb), 10,
+	    "Incorrect division result");
+
+	nstime_init2(&nsta, 42, 43);
+	nstime_copy(&nstb, &nsta);
+	nstime_imultiply(&nsta, 10);
+	nstime_init(&nstc, 1);
+	nstime_add(&nsta, &nstc);
+	assert_u64_eq(nstime_divide(&nsta, &nstb), 10,
+	    "Incorrect division result");
+
+	nstime_init2(&nsta, 42, 43);
+	nstime_copy(&nstb, &nsta);
+	nstime_imultiply(&nsta, 10);
+	nstime_init(&nstc, 1);
+	nstime_subtract(&nsta, &nstc);
+	assert_u64_eq(nstime_divide(&nsta, &nstb), 9,
+	    "Incorrect division result");
+}
+TEST_END
+
+TEST_BEGIN(test_nstime_update)
+{
+	nstime_t nst;
+
+	nstime_init(&nst, 0);
+
+	assert_false(nstime_update(&nst), "Basic time update failed.");
+
+	/* Only Rip Van Winkle sleeps this long. */
+	{
+		nstime_t addend;
+		nstime_init2(&addend, 631152000, 0);
+		nstime_add(&nst, &addend);
+	}
+	{
+		nstime_t nst0;
+		nstime_copy(&nst0, &nst);
+		assert_true(nstime_update(&nst),
+		    "Update should detect time roll-back.");
+		assert_d_eq(nstime_compare(&nst, &nst0), 0,
+		    "Time should not have been modified");
+	}
+
+}
+TEST_END
+
+int
+main(void)
+{
+
+	return (test(
+	    test_nstime_init,
+	    test_nstime_init2,
+	    test_nstime_copy,
+	    test_nstime_compare,
+	    test_nstime_add,
+	    test_nstime_subtract,
+	    test_nstime_imultiply,
+	    test_nstime_idivide,
+	    test_nstime_divide,
+	    test_nstime_update));
+}

test/unit/prng.c

@@ -0,0 +1,68 @@
+#include "test/jemalloc_test.h"
+
+TEST_BEGIN(test_prng_lg_range)
+{
+	uint64_t sa, sb, ra, rb;
+	unsigned lg_range;
+
+	sa = 42;
+	ra = prng_lg_range(&sa, 64);
+	sa = 42;
+	rb = prng_lg_range(&sa, 64);
+	assert_u64_eq(ra, rb,
+	    "Repeated generation should produce repeated results");
+
+	sb = 42;
+	rb = prng_lg_range(&sb, 64);
+	assert_u64_eq(ra, rb,
+	    "Equivalent generation should produce equivalent results");
+
+	sa = 42;
+	ra = prng_lg_range(&sa, 64);
+	rb = prng_lg_range(&sa, 64);
+	assert_u64_ne(ra, rb,
+	    "Full-width results must not immediately repeat");
+
+	sa = 42;
+	ra = prng_lg_range(&sa, 64);
+	for (lg_range = 63; lg_range > 0; lg_range--) {
+		sb = 42;
+		rb = prng_lg_range(&sb, lg_range);
+		assert_u64_eq((rb & (UINT64_C(0xffffffffffffffff) << lg_range)),
+		    0, "High order bits should be 0, lg_range=%u", lg_range);
+		assert_u64_eq(rb, (ra >> (64 - lg_range)),
+		    "Expected high order bits of full-width result, "
+		    "lg_range=%u", lg_range);
+	}
+}
+TEST_END
+
+TEST_BEGIN(test_prng_range)
+{
+	uint64_t range;
+#define	MAX_RANGE	10000000
+#define	RANGE_STEP	97
+#define	NREPS		10
+
+	for (range = 2; range < MAX_RANGE; range += RANGE_STEP) {
+		uint64_t s;
+		unsigned rep;
+
+		s = range;
+		for (rep = 0; rep < NREPS; rep++) {
+			uint64_t r = prng_range(&s, range);
+
+			assert_u64_lt(r, range, "Out of range");
+		}
+	}
+}
+TEST_END
+
+int
+main(void)
+{
+
+	return (test(
+	    test_prng_lg_range,
+	    test_prng_range));
+}

test/unit/rb.c

@@ -3,7 +3,7 @@
 #define	rbtn_black_height(a_type, a_field, a_rbt, r_height) do {	\
     a_type *rbp_bh_t;							\
     for (rbp_bh_t = (a_rbt)->rbt_root, (r_height) = 0;			\
-      rbp_bh_t != &(a_rbt)->rbt_nil;					\
+	 rbp_bh_t != NULL;						\
       rbp_bh_t = rbtn_left_get(a_type, a_field, rbp_bh_t)) {		\
 	if (!rbtn_red_get(a_type, a_field, rbp_bh_t)) {			\
 	    (r_height)++;						\
@@ -21,7 +21,7 @@ struct node_s {
 };
 
 static int
-node_cmp(node_t *a, node_t *b) {
+node_cmp(const node_t *a, const node_t *b) {
 	int ret;
 
 	assert_u32_eq(a->magic, NODE_MAGIC, "Bad magic");
@@ -68,38 +68,43 @@ TEST_BEGIN(test_rb_empty)
 TEST_END
 
 static unsigned
-tree_recurse(node_t *node, unsigned black_height, unsigned black_depth,
+tree_recurse(node_t *node, unsigned black_height, unsigned black_depth)
-    node_t *nil)
 {
 	unsigned ret = 0;
-	node_t *left_node = rbtn_left_get(node_t, link, node);
+	node_t *left_node;
-	node_t *right_node = rbtn_right_get(node_t, link, node);
+	node_t *right_node;
+
+	if (node == NULL)
+		return (ret);
+
+	left_node = rbtn_left_get(node_t, link, node);
+	right_node = rbtn_right_get(node_t, link, node);
 
 	if (!rbtn_red_get(node_t, link, node))
 		black_depth++;
 
 	/* Red nodes must be interleaved with black nodes. */
 	if (rbtn_red_get(node_t, link, node)) {
+		if (left_node != NULL)
 			assert_false(rbtn_red_get(node_t, link, left_node),
 				"Node should be black");
+		if (right_node != NULL)
 			assert_false(rbtn_red_get(node_t, link, right_node),
 			    "Node should be black");
 	}
 
-	if (node == nil)
-		return (ret);
 	/* Self. */
 	assert_u32_eq(node->magic, NODE_MAGIC, "Bad magic");
 
 	/* Left subtree. */
-	if (left_node != nil)
+	if (left_node != NULL)
-		ret += tree_recurse(left_node, black_height, black_depth, nil);
+		ret += tree_recurse(left_node, black_height, black_depth);
 	else
 		ret += (black_depth != black_height);
 
 	/* Right subtree. */
-	if (right_node != nil)
+	if (right_node != NULL)
-		ret += tree_recurse(right_node, black_height, black_depth, nil);
+		ret += tree_recurse(right_node, black_height, black_depth);
 	else
 		ret += (black_depth != black_height);
 
@@ -181,8 +186,7 @@ node_remove(tree_t *tree, node_t *node, unsigned nnodes)
 	node->magic = 0;
 
 	rbtn_black_height(node_t, link, tree, black_height);
-	imbalances = tree_recurse(tree->rbt_root, black_height, 0,
+	imbalances = tree_recurse(tree->rbt_root, black_height, 0);
-	    &(tree->rbt_nil));
 	assert_u_eq(imbalances, 0, "Tree is unbalanced");
 	assert_u_eq(tree_iterate(tree), nnodes-1,
 	    "Unexpected node iteration count");
@@ -212,6 +216,15 @@ remove_reverse_iterate_cb(tree_t *tree, node_t *node, void *data)
 	return (ret);
 }
 
+static void
+destroy_cb(node_t *node, void *data)
+{
+	unsigned *nnodes = (unsigned *)data;
+
+	assert_u_gt(*nnodes, 0, "Destruction removed too many nodes");
+	(*nnodes)--;
+}
+
 TEST_BEGIN(test_rb_random)
 {
 #define	NNODES 25
@@ -244,7 +257,6 @@ TEST_BEGIN(test_rb_random)
 		for (j = 1; j <= NNODES; j++) {
 			/* Initialize tree and nodes. */
 			tree_new(&tree);
-			tree.rbt_nil.magic = 0;
 			for (k = 0; k < j; k++) {
 				nodes[k].magic = NODE_MAGIC;
 				nodes[k].key = bag[k];
@@ -257,7 +269,7 @@ TEST_BEGIN(test_rb_random)
 				rbtn_black_height(node_t, link, &tree,
 				    black_height);
 				imbalances = tree_recurse(tree.rbt_root,
-				    black_height, 0, &(tree.rbt_nil));
+				    black_height, 0);
 				assert_u_eq(imbalances, 0,
 				    "Tree is unbalanced");
 
@@ -278,7 +290,7 @@ TEST_BEGIN(test_rb_random)
 			}
 
 			/* Remove nodes. */
-			switch (i % 4) {
+			switch (i % 5) {
 			case 0:
 				for (k = 0; k < j; k++)
 					node_remove(&tree, &nodes[k], j - k);
@@ -314,6 +326,12 @@ TEST_BEGIN(test_rb_random)
 				assert_u_eq(nnodes, 0,
 				    "Removal terminated early");
 				break;
+			} case 4: {
+				unsigned nnodes = j;
+				tree_destroy(&tree, destroy_cb, &nnodes);
+				assert_u_eq(nnodes, 0,
+				    "Destruction terminated early");
+				break;
 			} default:
 				not_reached();
 			}

test/unit/run_quantize.c

@@ -0,0 +1,149 @@
+#include "test/jemalloc_test.h"
+
+TEST_BEGIN(test_small_run_size)
+{
+	unsigned nbins, i;
+	size_t sz, run_size;
+	size_t mib[4];
+	size_t miblen = sizeof(mib) / sizeof(size_t);
+
+	/*
+	 * Iterate over all small size classes, get their run sizes, and verify
+	 * that the quantized size is the same as the run size.
+	 */
+
+	sz = sizeof(unsigned);
+	assert_d_eq(mallctl("arenas.nbins", &nbins, &sz, NULL, 0), 0,
+	    "Unexpected mallctl failure");
+
+	assert_d_eq(mallctlnametomib("arenas.bin.0.run_size", mib, &miblen), 0,
+	    "Unexpected mallctlnametomib failure");
+	for (i = 0; i < nbins; i++) {
+		mib[2] = i;
+		sz = sizeof(size_t);
+		assert_d_eq(mallctlbymib(mib, miblen, &run_size, &sz, NULL, 0),
+		    0, "Unexpected mallctlbymib failure");
+		assert_zu_eq(run_size, run_quantize_floor(run_size),
+		    "Small run quantization should be a no-op (run_size=%zu)",
+		    run_size);
+		assert_zu_eq(run_size, run_quantize_ceil(run_size),
+		    "Small run quantization should be a no-op (run_size=%zu)",
+		    run_size);
+	}
+}
+TEST_END
+
+TEST_BEGIN(test_large_run_size)
+{
+	bool cache_oblivious;
+	unsigned nlruns, i;
+	size_t sz, run_size_prev, ceil_prev;
+	size_t mib[4];
+	size_t miblen = sizeof(mib) / sizeof(size_t);
+
+	/*
+	 * Iterate over all large size classes, get their run sizes, and verify
+	 * that the quantized size is the same as the run size.
+	 */
+
+	sz = sizeof(bool);
+	assert_d_eq(mallctl("config.cache_oblivious", &cache_oblivious, &sz,
+	    NULL, 0), 0, "Unexpected mallctl failure");
+
+	sz = sizeof(unsigned);
+	assert_d_eq(mallctl("arenas.nlruns", &nlruns, &sz, NULL, 0), 0,
+	    "Unexpected mallctl failure");
+
+	assert_d_eq(mallctlnametomib("arenas.lrun.0.size", mib, &miblen), 0,
+	    "Unexpected mallctlnametomib failure");
+	for (i = 0; i < nlruns; i++) {
+		size_t lrun_size, run_size, floor, ceil;
+
+		mib[2] = i;
+		sz = sizeof(size_t);
+		assert_d_eq(mallctlbymib(mib, miblen, &lrun_size, &sz, NULL, 0),
+		    0, "Unexpected mallctlbymib failure");
+		run_size = cache_oblivious ? lrun_size + PAGE : lrun_size;
+		floor = run_quantize_floor(run_size);
+		ceil = run_quantize_ceil(run_size);
+
+		assert_zu_eq(run_size, floor,
+		    "Large run quantization should be a no-op for precise "
+		    "size (lrun_size=%zu, run_size=%zu)", lrun_size, run_size);
+		assert_zu_eq(run_size, ceil,
+		    "Large run quantization should be a no-op for precise "
+		    "size (lrun_size=%zu, run_size=%zu)", lrun_size, run_size);
+
+		if (i > 0) {
+			assert_zu_eq(run_size_prev, run_quantize_floor(run_size
+			    - PAGE), "Floor should be a precise size");
+			if (run_size_prev < ceil_prev) {
+				assert_zu_eq(ceil_prev, run_size,
+				    "Ceiling should be a precise size "
+				    "(run_size_prev=%zu, ceil_prev=%zu, "
+				    "run_size=%zu)", run_size_prev, ceil_prev,
+				    run_size);
+			}
+		}
+		run_size_prev = floor;
+		ceil_prev = run_quantize_ceil(run_size + PAGE);
+	}
+}
+TEST_END
+
+TEST_BEGIN(test_monotonic)
+{
+	unsigned nbins, nlruns, i;
+	size_t sz, floor_prev, ceil_prev;
+
+	/*
+	 * Iterate over all run sizes and verify that
+	 * run_quantize_{floor,ceil}() are monotonic.
+	 */
+
+	sz = sizeof(unsigned);
+	assert_d_eq(mallctl("arenas.nbins", &nbins, &sz, NULL, 0), 0,
+	    "Unexpected mallctl failure");
+
+	sz = sizeof(unsigned);
+	assert_d_eq(mallctl("arenas.nlruns", &nlruns, &sz, NULL, 0), 0,
+	    "Unexpected mallctl failure");
+
+	floor_prev = 0;
+	ceil_prev = 0;
+	for (i = 1; i < run_quantize_max >> LG_PAGE; i++) {
+		size_t run_size, floor, ceil;
+
+		run_size = i << LG_PAGE;
+		floor = run_quantize_floor(run_size);
+		ceil = run_quantize_ceil(run_size);
+
+		assert_zu_le(floor, run_size,
+		    "Floor should be <= (floor=%zu, run_size=%zu, ceil=%zu)",
+		    floor, run_size, ceil);
+		assert_zu_ge(ceil, run_size,
+		    "Ceiling should be >= (floor=%zu, run_size=%zu, ceil=%zu)",
+		    floor, run_size, ceil);
+
+		assert_zu_le(floor_prev, floor, "Floor should be monotonic "
+		    "(floor_prev=%zu, floor=%zu, run_size=%zu, ceil=%zu)",
+		    floor_prev, floor, run_size, ceil);
+		assert_zu_le(ceil_prev, ceil, "Ceiling should be monotonic "
+		    "(floor=%zu, run_size=%zu, ceil_prev=%zu, ceil=%zu)",
+		    floor, run_size, ceil_prev, ceil);
+
+		floor_prev = floor;
+		ceil_prev = ceil;
+	}
+}
+TEST_END
+
+int
+main(void)
+{
+
+	return (test(
+	    test_small_run_size,
+	    test_large_run_size,
+	    test_monotonic));
+}

test/unit/size_classes.c

@@ -80,10 +80,33 @@ TEST_BEGIN(test_size_classes)
 }
 TEST_END
 
+TEST_BEGIN(test_overflow)
+{
+	size_t max_size_class;
+
+	max_size_class = get_max_size_class();
+
+	assert_u_ge(size2index(max_size_class+1), NSIZES,
+	    "size2index() should return >= NSIZES on overflow");
+	assert_u_ge(size2index(ZU(PTRDIFF_MAX)+1), NSIZES,
+	    "size2index() should return >= NSIZES on overflow");
+	assert_u_ge(size2index(SIZE_T_MAX), NSIZES,
+	    "size2index() should return >= NSIZES on overflow");
+
+	assert_zu_gt(s2u(max_size_class+1), HUGE_MAXCLASS,
+	    "s2u() should return > HUGE_MAXCLASS for unsupported size");
+	assert_zu_gt(s2u(ZU(PTRDIFF_MAX)+1), HUGE_MAXCLASS,
+	    "s2u() should return > HUGE_MAXCLASS for unsupported size");
+	assert_zu_eq(s2u(SIZE_T_MAX), 0,
+	    "s2u() should return 0 on overflow");
+}
+TEST_END
+
 int
 main(void)
 {
 
 	return (test(
-	    test_size_classes));
+	    test_size_classes,
+	    test_overflow));
 }

test/unit/smoothstep.c

@@ -0,0 +1,106 @@
+#include "test/jemalloc_test.h"
+
+static const uint64_t smoothstep_tab[] = {
+#define	STEP(step, h, x, y) \
+	h,
+	SMOOTHSTEP
+#undef STEP
+};
+
+TEST_BEGIN(test_smoothstep_integral)
+{
+	uint64_t sum, min, max;
+	unsigned i;
+
+	/*
+	 * The integral of smoothstep in the [0..1] range equals 1/2.  Verify
+	 * that the fixed point representation's integral is no more than
+	 * rounding error distant from 1/2.  Regarding rounding, each table
+	 * element is rounded down to the nearest fixed point value, so the
+	 * integral may be off by as much as SMOOTHSTEP_NSTEPS ulps.
+	 */
+	sum = 0;
+	for (i = 0; i < SMOOTHSTEP_NSTEPS; i++)
+		sum += smoothstep_tab[i];
+
+	max = (KQU(1) << (SMOOTHSTEP_BFP-1)) * (SMOOTHSTEP_NSTEPS+1);
+	min = max - SMOOTHSTEP_NSTEPS;
+
+	assert_u64_ge(sum, min,
+	    "Integral too small, even accounting for truncation");
+	assert_u64_le(sum, max, "Integral exceeds 1/2");
+	if (false) {
+		malloc_printf("%"FMTu64" ulps under 1/2 (limit %d)\n",
+		    max - sum, SMOOTHSTEP_NSTEPS);
+	}
+}
+TEST_END
+
+TEST_BEGIN(test_smoothstep_monotonic)
+{
+	uint64_t prev_h;
+	unsigned i;
+
+	/*
+	 * The smoothstep function is monotonic in [0..1], i.e. its slope is
+	 * non-negative.  In practice we want to parametrize table generation
+	 * such that piecewise slope is greater than zero, but do not require
+	 * that here.
+	 */
+	prev_h = 0;
+	for (i = 0; i < SMOOTHSTEP_NSTEPS; i++) {
+		uint64_t h = smoothstep_tab[i];
+		assert_u64_ge(h, prev_h, "Piecewise non-monotonic, i=%u", i);
+		prev_h = h;
+	}
+	assert_u64_eq(smoothstep_tab[SMOOTHSTEP_NSTEPS-1],
+	    (KQU(1) << SMOOTHSTEP_BFP), "Last step must equal 1");
+}
+TEST_END
+
+TEST_BEGIN(test_smoothstep_slope)
+{
+	uint64_t prev_h, prev_delta;
+	unsigned i;
+
+	/*
+	 * The smoothstep slope strictly increases until x=0.5, and then
+	 * strictly decreases until x=1.0.  Verify the slightly weaker
+	 * requirement of monotonicity, so that inadequate table precision does
+	 * not cause false test failures.
+	 */
+	prev_h = 0;
+	prev_delta = 0;
+	for (i = 0; i < SMOOTHSTEP_NSTEPS / 2 + SMOOTHSTEP_NSTEPS % 2; i++) {
+		uint64_t h = smoothstep_tab[i];
+		uint64_t delta = h - prev_h;
+		assert_u64_ge(delta, prev_delta,
+		    "Slope must monotonically increase in 0.0 <= x <= 0.5, "
+		    "i=%u", i);
+		prev_h = h;
+		prev_delta = delta;
+	}
+
+	prev_h = KQU(1) << SMOOTHSTEP_BFP;
+	prev_delta = 0;
+	for (i = SMOOTHSTEP_NSTEPS-1; i >= SMOOTHSTEP_NSTEPS / 2; i--) {
+		uint64_t h = smoothstep_tab[i];
+		uint64_t delta = prev_h - h;
+		assert_u64_ge(delta, prev_delta,
+		    "Slope must monotonically decrease in 0.5 <= x <= 1.0, "
+		    "i=%u", i);
+		prev_h = h;
+		prev_delta = delta;
+	}
+}
+TEST_END
+
+int
+main(void)
+{
+
+	return (test(
+	    test_smoothstep_integral,
+	    test_smoothstep_monotonic,
+	    test_smoothstep_slope));
+}

test/unit/stats.c

@@ -93,6 +93,10 @@ TEST_BEGIN(test_stats_arenas_summary)
 	huge = mallocx(chunksize, 0);
 	assert_ptr_not_null(huge, "Unexpected mallocx() failure");
 
+	dallocx(little, 0);
+	dallocx(large, 0);
+	dallocx(huge, 0);
+
 	assert_d_eq(mallctl("arena.0.purge", NULL, NULL, NULL, 0), 0,
 	    "Unexpected mallctl() failure");
 
@@ -116,10 +120,6 @@ TEST_BEGIN(test_stats_arenas_summary)
 		assert_u64_le(nmadvise, purged,
 		    "nmadvise should be no greater than purged");
 	}
-
-	dallocx(little, 0);
-	dallocx(large, 0);
-	dallocx(huge, 0);
 }
 TEST_END
 

test/unit/ticker.c

@@ -0,0 +1,76 @@
+#include "test/jemalloc_test.h"
+
+TEST_BEGIN(test_ticker_tick)
+{
+#define	NREPS 2
+#define	NTICKS 3
+	ticker_t ticker;
+	int32_t i, j;
+
+	ticker_init(&ticker, NTICKS);
+	for (i = 0; i < NREPS; i++) {
+		for (j = 0; j < NTICKS; j++) {
+			assert_u_eq(ticker_read(&ticker), NTICKS - j,
+			    "Unexpected ticker value (i=%d, j=%d)", i, j);
+			assert_false(ticker_tick(&ticker),
+			    "Unexpected ticker fire (i=%d, j=%d)", i, j);
+		}
+		assert_u32_eq(ticker_read(&ticker), 0,
+		    "Expected ticker depletion");
+		assert_true(ticker_tick(&ticker),
+		    "Expected ticker fire (i=%d)", i);
+		assert_u32_eq(ticker_read(&ticker), NTICKS,
+		    "Expected ticker reset");
+	}
+#undef NTICKS
+}
+TEST_END
+
+TEST_BEGIN(test_ticker_ticks)
+{
+#define	NTICKS 3
+	ticker_t ticker;
+
+	ticker_init(&ticker, NTICKS);
+
+	assert_u_eq(ticker_read(&ticker), NTICKS, "Unexpected ticker value");
+	assert_false(ticker_ticks(&ticker, NTICKS), "Unexpected ticker fire");
+	assert_u_eq(ticker_read(&ticker), 0, "Unexpected ticker value");
+	assert_true(ticker_ticks(&ticker, NTICKS), "Expected ticker fire");
+	assert_u_eq(ticker_read(&ticker), NTICKS, "Unexpected ticker value");
+
+	assert_true(ticker_ticks(&ticker, NTICKS + 1), "Expected ticker fire");
+	assert_u_eq(ticker_read(&ticker), NTICKS, "Unexpected ticker value");
+#undef NTICKS
+}
+TEST_END
+
+TEST_BEGIN(test_ticker_copy)
+{
+#define	NTICKS 3
+	ticker_t ta, tb;
+
+	ticker_init(&ta, NTICKS);
+	ticker_copy(&tb, &ta);
+	assert_u_eq(ticker_read(&tb), NTICKS, "Unexpected ticker value");
+	assert_true(ticker_ticks(&tb, NTICKS + 1), "Expected ticker fire");
+	assert_u_eq(ticker_read(&tb), NTICKS, "Unexpected ticker value");
+
+	ticker_tick(&ta);
+	ticker_copy(&tb, &ta);
+	assert_u_eq(ticker_read(&tb), NTICKS - 1, "Unexpected ticker value");
+	assert_true(ticker_ticks(&tb, NTICKS), "Expected ticker fire");
+	assert_u_eq(ticker_read(&tb), NTICKS, "Unexpected ticker value");
+#undef NTICKS
+}
+TEST_END
+
+int
+main(void)
+{
+
+	return (test(
+	    test_ticker_tick,
+	    test_ticker_ticks,
+	    test_ticker_copy));
+}

test/unit/util.c

@@ -1,33 +1,54 @@
 #include "test/jemalloc_test.h"
 
-TEST_BEGIN(test_pow2_ceil)
+#define	TEST_POW2_CEIL(t, suf, pri) do {				\
+	unsigned i, pow2;						\
+	t x;								\
+									\
+	assert_zu_eq(pow2_ceil_##suf(0), 0, "Unexpected result");	\
+									\
+	for (i = 0; i < sizeof(t) * 8; i++) {				\
+		assert_zu_eq(pow2_ceil_##suf(((t)1) << i), ((t)1) << i,	\
+		    "Unexpected result");				\
+	}								\
+									\
+	for (i = 2; i < sizeof(t) * 8; i++) {				\
+		assert_zu_eq(pow2_ceil_##suf((((t)1) << i) - 1),	\
+		    ((t)1) << i, "Unexpected result");			\
+	}								\
+									\
+	for (i = 0; i < sizeof(t) * 8 - 1; i++) {			\
+		assert_zu_eq(pow2_ceil_##suf((((t)1) << i) + 1),	\
+		    ((t)1) << (i+1), "Unexpected result");		\
+	}								\
+									\
+	for (pow2 = 1; pow2 < 25; pow2++) {				\
+		for (x = (((t)1) << (pow2-1)) + 1; x <= ((t)1) << pow2;	\
+		    x++) {						\
+			assert_zu_eq(pow2_ceil_##suf(x),		\
+			    ((t)1) << pow2,				\
+			    "Unexpected result, x=%"pri, x);		\
+		}							\
+	}								\
+} while (0)
+
+TEST_BEGIN(test_pow2_ceil_u64)
 {
-	unsigned i, pow2;
-	size_t x;
 
-	assert_zu_eq(pow2_ceil(0), 0, "Unexpected result");
+	TEST_POW2_CEIL(uint64_t, u64, FMTu64);
+}
+TEST_END
 
-	for (i = 0; i < sizeof(size_t) * 8; i++) {
+TEST_BEGIN(test_pow2_ceil_u32)
-		assert_zu_eq(pow2_ceil(ZU(1) << i), ZU(1) << i,
+{
-		    "Unexpected result");
-	}
 
-	for (i = 2; i < sizeof(size_t) * 8; i++) {
+	TEST_POW2_CEIL(uint32_t, u32, FMTu32);
-		assert_zu_eq(pow2_ceil((ZU(1) << i) - 1), ZU(1) << i,
+}
-		    "Unexpected result");
+TEST_END
-	}
 
-	for (i = 0; i < sizeof(size_t) * 8 - 1; i++) {
+TEST_BEGIN(test_pow2_ceil_zu)
-		assert_zu_eq(pow2_ceil((ZU(1) << i) + 1), ZU(1) << (i+1),
+{
-		    "Unexpected result");
-	}
 
-	for (pow2 = 1; pow2 < 25; pow2++) {
+	TEST_POW2_CEIL(size_t, zu, "zu");
-		for (x = (ZU(1) << (pow2-1)) + 1; x <= ZU(1) << pow2; x++) {
-			assert_zu_eq(pow2_ceil(x), ZU(1) << pow2,
-			    "Unexpected result, x=%zu", x);
-		}
-	}
 }
 TEST_END
 
@@ -286,7 +307,9 @@ main(void)
 {
 
 	return (test(
-	    test_pow2_ceil,
+	    test_pow2_ceil_u64,
+	    test_pow2_ceil_u32,
+	    test_pow2_ceil_zu,
 	    test_malloc_strtoumax_no_endptr,
 	    test_malloc_strtoumax,
 	    test_malloc_snprintf_truncated,