diff --git a/Makefile.in b/Makefile.in
index 8a34928..b9917da 100644
--- a/Makefile.in
+++ b/Makefile.in
@@ -32,7 +32,7 @@
 O := @o@
 A := @a@
 EXE := @exe@
-LIB := @lib@
+LIBPREFIX := @libprefix@
 REV := @rev@
 install_suffix := @install_suffix@
 ABI := @abi@
@@ -51,12 +51,16 @@
 PIC_CFLAGS = @PIC_CFLAGS@
 
 ifeq (macho, $(ABI))
-TEST_LIBRARY_PATH := DYLD_FALLBACK_LIBRARY_PATH=$(objroot)lib
+TEST_LIBRARY_PATH := DYLD_FALLBACK_LIBRARY_PATH="$(objroot)lib"
+else
+ifeq (pecoff, $(ABI))
+TEST_LIBRARY_PATH := PATH="$(PATH):$(objroot)lib"
 else
 TEST_LIBRARY_PATH :=
 endif
+endif
 
-LIBJEMALLOC := $(LIB)jemalloc$(install_suffix)
+LIBJEMALLOC := $(LIBPREFIX)jemalloc$(install_suffix)
 
 # Lists of files.
 BINS := $(srcroot)bin/pprof $(objroot)bin/jemalloc.sh
diff --git a/configure.ac b/configure.ac
index 0ed0494..5bdddfa 100644
--- a/configure.ac
+++ b/configure.ac
@@ -198,7 +198,7 @@
 o="o"
 a="a"
 exe=
-lib="lib"
+libprefix="lib"
 DSO_LDFLAGS='-shared -Wl,-soname,$(@F)'
 RPATH='-Wl,-rpath,$(1)'
 SOREV='$(SO).$(REV)'
@@ -273,6 +273,19 @@
 	fi
 	abi="xcoff"
 	;;
+  *-*-mingw*)
+	abi="pecoff"
+	force_tls="0"
+	RPATH=""
+	so="dll"
+	DSO_LDFLAGS="-shared"
+	o="obj"
+	a="lib"
+	libprefix=""
+	exe=".exe"
+	SOREV='$(SO)'
+	PIC_CFLAGS=""
+	;;
   *)
 	AC_MSG_RESULT([Unsupported operating system: ${host}])
 	abi="elf"
@@ -285,7 +298,7 @@
 AC_SUBST([o])
 AC_SUBST([a])
 AC_SUBST([exe])
-AC_SUBST([lib])
+AC_SUBST([libprefix])
 AC_SUBST([DSO_LDFLAGS])
 AC_SUBST([SOREV])
 AC_SUBST([PIC_CFLAGS])
@@ -817,23 +830,36 @@
 AC_CACHE_CHECK([STATIC_PAGE_SHIFT],
                [je_cv_static_page_shift],
                AC_RUN_IFELSE([AC_LANG_PROGRAM(
-[[#include <stdio.h>
+[[
+#ifdef _WIN32
+#include <windows.h>
+#else
 #include <unistd.h>
 #include <strings.h>
+#endif
+#include <stdio.h>
 ]],
 [[
     long result;
     FILE *f;
 
+#ifdef _WIN32
+    SYSTEM_INFO si;
+    GetSystemInfo(&si);
+    result = si.dwPageSize;
+#else
     result = sysconf(_SC_PAGESIZE);
+#endif
     if (result == -1) {
 	return 1;
     }
+    result = ffsl(result) - 1;
+
     f = fopen("conftest.out", "w");
     if (f == NULL) {
 	return 1;
     }
-    fprintf(f, "%u\n", ffs((int)result) - 1);
+    fprintf(f, "%u\n", result);
     fclose(f);
 
     return 0;
@@ -871,12 +897,14 @@
 dnl ============================================================================
 dnl Configure pthreads.
 
-AC_CHECK_HEADERS([pthread.h], , [AC_MSG_ERROR([pthread.h is missing])])
-dnl Some systems may embed pthreads functionality in libc; check for libpthread
-dnl first, but try libc too before failing.
-AC_CHECK_LIB([pthread], [pthread_create], [LIBS="$LIBS -lpthread"],
-             [AC_SEARCH_LIBS([pthread_create], , ,
-                             AC_MSG_ERROR([libpthread is missing]))])
+if test "x$abi" != "xpecoff" ; then
+  AC_CHECK_HEADERS([pthread.h], , [AC_MSG_ERROR([pthread.h is missing])])
+  dnl Some systems may embed pthreads functionality in libc; check for libpthread
+  dnl first, but try libc too before failing.
+  AC_CHECK_LIB([pthread], [pthread_create], [LIBS="$LIBS -lpthread"],
+               [AC_SEARCH_LIBS([pthread_create], , ,
+                               AC_MSG_ERROR([libpthread is missing]))])
+fi
 
 CPPFLAGS="$CPPFLAGS -D_REENTRANT"
 
@@ -921,11 +949,13 @@
   enable_lazy_lock="1"
 fi
 if test "x$enable_lazy_lock" = "x1" ; then
-  AC_CHECK_HEADERS([dlfcn.h], , [AC_MSG_ERROR([dlfcn.h is missing])])
-  AC_CHECK_FUNC([dlsym], [],
-    [AC_CHECK_LIB([dl], [dlsym], [LIBS="$LIBS -ldl"],
-                  [AC_MSG_ERROR([libdl is missing])])
-    ])
+  if test "x$abi" != "xpecoff" ; then
+    AC_CHECK_HEADERS([dlfcn.h], , [AC_MSG_ERROR([dlfcn.h is missing])])
+    AC_CHECK_FUNC([dlsym], [],
+      [AC_CHECK_LIB([dl], [dlsym], [LIBS="$LIBS -ldl"],
+                    [AC_MSG_ERROR([libdl is missing])])
+      ])
+  fi
   AC_DEFINE([JEMALLOC_LAZY_LOCK], [ ])
 fi
 AC_SUBST([enable_lazy_lock])
diff --git a/include/jemalloc/internal/jemalloc_internal.h.in b/include/jemalloc/internal/jemalloc_internal.h.in
index b61abe8..fd5de72 100644
--- a/include/jemalloc/internal/jemalloc_internal.h.in
+++ b/include/jemalloc/internal/jemalloc_internal.h.in
@@ -1,21 +1,25 @@
 #ifndef JEMALLOC_INTERNAL_H
 #define JEMALLOC_INTERNAL_H
-#include <sys/mman.h>
 #include <sys/param.h>
-#include <sys/syscall.h>
-#if !defined(SYS_write) && defined(__NR_write)
-#define	SYS_write __NR_write
+#ifdef _WIN32
+#  include <windows.h>
+#else
+#  include <sys/mman.h>
+#  include <sys/syscall.h>
+#  if !defined(SYS_write) && defined(__NR_write)
+#    define SYS_write __NR_write
+#  endif
+#  include <sys/uio.h>
+#  include <pthread.h>
 #endif
 #include <sys/time.h>
 #include <sys/types.h>
-#include <sys/uio.h>
 
 #include <errno.h>
 #include <limits.h>
 #ifndef SIZE_T_MAX
 #  define SIZE_T_MAX	SIZE_MAX
 #endif
-#include <pthread.h>
 #include <sched.h>
 #include <stdarg.h>
 #include <stdbool.h>
@@ -32,7 +36,6 @@
 #include <ctype.h>
 #include <unistd.h>
 #include <fcntl.h>
-#include <pthread.h>
 #include <math.h>
 
 #define	JEMALLOC_NO_DEMANGLE
diff --git a/include/jemalloc/internal/mutex.h b/include/jemalloc/internal/mutex.h
index 8837ef5..de44e14 100644
--- a/include/jemalloc/internal/mutex.h
+++ b/include/jemalloc/internal/mutex.h
@@ -3,10 +3,12 @@
 
 typedef struct malloc_mutex_s malloc_mutex_t;
 
-#ifdef JEMALLOC_OSSPIN
-#define	MALLOC_MUTEX_INITIALIZER {0}
+#ifdef _WIN32
+#  define MALLOC_MUTEX_INITIALIZER
+#elif (defined(JEMALLOC_OSSPIN))
+#  define MALLOC_MUTEX_INITIALIZER {0}
 #elif (defined(JEMALLOC_MUTEX_INIT_CB))
-#define	MALLOC_MUTEX_INITIALIZER {PTHREAD_MUTEX_INITIALIZER, NULL}
+#  define MALLOC_MUTEX_INITIALIZER {PTHREAD_MUTEX_INITIALIZER, NULL}
 #else
 #  if (defined(PTHREAD_MUTEX_ADAPTIVE_NP) &&				\
        defined(PTHREAD_ADAPTIVE_MUTEX_INITIALIZER_NP))
@@ -23,7 +25,9 @@
 #ifdef JEMALLOC_H_STRUCTS
 
 struct malloc_mutex_s {
-#ifdef JEMALLOC_OSSPIN
+#ifdef _WIN32
+	CRITICAL_SECTION	lock;
+#elif (defined(JEMALLOC_OSSPIN))
 	OSSpinLock		lock;
 #elif (defined(JEMALLOC_MUTEX_INIT_CB))
 	pthread_mutex_t		lock;
@@ -65,7 +69,9 @@
 {
 
 	if (isthreaded) {
-#ifdef JEMALLOC_OSSPIN
+#ifdef _WIN32
+		EnterCriticalSection(&mutex->lock);
+#elif (defined(JEMALLOC_OSSPIN))
 		OSSpinLockLock(&mutex->lock);
 #else
 		pthread_mutex_lock(&mutex->lock);
@@ -78,7 +84,9 @@
 {
 
 	if (isthreaded) {
-#ifdef JEMALLOC_OSSPIN
+#ifdef _WIN32
+		LeaveCriticalSection(&mutex->lock);
+#elif (defined(JEMALLOC_OSSPIN))
 		OSSpinLockUnlock(&mutex->lock);
 #else
 		pthread_mutex_unlock(&mutex->lock);
diff --git a/include/jemalloc/internal/tsd.h b/include/jemalloc/internal/tsd.h
index 20491c8..0037cf3 100644
--- a/include/jemalloc/internal/tsd.h
+++ b/include/jemalloc/internal/tsd.h
@@ -74,6 +74,10 @@
 extern __thread a_type	a_name##_tls;					\
 extern pthread_key_t	a_name##_tsd;					\
 extern bool		a_name##_booted;
+#elif (defined(_WIN32))
+#define malloc_tsd_externs(a_name, a_type)				\
+extern DWORD		a_name##_tsd;					\
+extern bool		a_name##_booted;
 #else
 #define	malloc_tsd_externs(a_name, a_type)				\
 extern pthread_key_t	a_name##_tsd;					\
@@ -94,6 +98,10 @@
     a_name##_tls = a_initializer;					\
 a_attr pthread_key_t	a_name##_tsd;					\
 a_attr bool		a_name##_booted = false;
+#elif (defined(_WIN32))
+#define	malloc_tsd_data(a_attr, a_name, a_type, a_initializer)		\
+a_attr DWORD		a_name##_tsd;					\
+a_attr bool		a_name##_booted = false;
 #else
 #define	malloc_tsd_data(a_attr, a_name, a_type, a_initializer)		\
 a_attr pthread_key_t	a_name##_tsd;					\
@@ -182,6 +190,99 @@
 		}							\
 	}								\
 }
+#elif (defined(_WIN32))
+#define	malloc_tsd_funcs(a_attr, a_name, a_type, a_initializer,		\
+    a_cleanup)								\
+/* Data structure. */							\
+typedef struct {							\
+	bool	initialized;						\
+	a_type	val;							\
+} a_name##_tsd_wrapper_t;						\
+/* Initialization/cleanup. */						\
+a_attr bool								\
+a_name##_tsd_cleanup_wrapper(void)					\
+{									\
+	a_name##_tsd_wrapper_t *wrapper;				\
+									\
+	wrapper = (a_name##_tsd_wrapper_t *) TlsGetValue(a_name##_tsd);	\
+	if (wrapper == NULL)						\
+		return (false);						\
+	if (a_cleanup != malloc_tsd_no_cleanup &&			\
+	    wrapper->initialized) {					\
+		a_type val = wrapper->val;				\
+		a_type tsd_static_data = a_initializer;			\
+		wrapper->initialized = false;				\
+		wrapper->val = tsd_static_data;				\
+		a_cleanup(&val);					\
+		if (wrapper->initialized) {				\
+			/* Trigger another cleanup round. */		\
+			return (true);					\
+		}							\
+	}								\
+	malloc_tsd_dalloc(wrapper);					\
+	return (false);							\
+}									\
+a_attr bool								\
+a_name##_tsd_boot(void)							\
+{									\
+									\
+	a_name##_tsd = TlsAlloc();					\
+	if (a_name##_tsd == TLS_OUT_OF_INDEXES)				\
+		return (true);						\
+	if (a_cleanup != malloc_tsd_no_cleanup) {			\
+		malloc_tsd_cleanup_register(				\
+		    &a_name##_tsd_cleanup_wrapper);			\
+	}								\
+	a_name##_booted = true;						\
+	return (false);							\
+}									\
+/* Get/set. */								\
+a_attr a_name##_tsd_wrapper_t *						\
+a_name##_tsd_get_wrapper(void)						\
+{									\
+	a_name##_tsd_wrapper_t *wrapper = (a_name##_tsd_wrapper_t *)	\
+	    TlsGetValue(a_name##_tsd);					\
+									\
+	if (wrapper == NULL) {						\
+		wrapper = (a_name##_tsd_wrapper_t *)			\
+		    malloc_tsd_malloc(sizeof(a_name##_tsd_wrapper_t));	\
+		if (wrapper == NULL) {					\
+			malloc_write("<jemalloc>: Error allocating"	\
+			    " TSD for "#a_name"\n");			\
+			abort();					\
+		} else {						\
+			static a_type tsd_static_data = a_initializer;	\
+			wrapper->initialized = false;			\
+			wrapper->val = tsd_static_data;			\
+		}							\
+		if (!TlsSetValue(a_name##_tsd, (void *)wrapper)) {	\
+			malloc_write("<jemalloc>: Error setting"	\
+			    " TSD for "#a_name"\n");			\
+			abort();					\
+		}							\
+	}								\
+	return (wrapper);						\
+}									\
+a_attr a_type *								\
+a_name##_tsd_get(void)							\
+{									\
+	a_name##_tsd_wrapper_t *wrapper;				\
+									\
+	assert(a_name##_booted);					\
+	wrapper = a_name##_tsd_get_wrapper();				\
+	return (&wrapper->val);						\
+}									\
+a_attr void								\
+a_name##_tsd_set(a_type *val)						\
+{									\
+	a_name##_tsd_wrapper_t *wrapper;				\
+									\
+	assert(a_name##_booted);					\
+	wrapper = a_name##_tsd_get_wrapper();				\
+	wrapper->val = *(val);						\
+	if (a_cleanup != malloc_tsd_no_cleanup)				\
+		wrapper->initialized = true;				\
+}
 #else
 #define	malloc_tsd_funcs(a_attr, a_name, a_type, a_initializer,		\
     a_cleanup)								\
diff --git a/src/chunk_mmap.c b/src/chunk_mmap.c
index 9ff7480..0ad65a1 100644
--- a/src/chunk_mmap.c
+++ b/src/chunk_mmap.c
@@ -7,7 +7,7 @@
 static void	*pages_map(void *addr, size_t size);
 static void	pages_unmap(void *addr, size_t size);
 static void	*chunk_alloc_mmap_slow(size_t size, size_t alignment,
-    bool unaligned, bool *zero);
+    bool *zero);
 
 /******************************************************************************/
 
@@ -16,6 +16,14 @@
 {
 	void *ret;
 
+#ifdef _WIN32
+	/*
+	 * If VirtualAlloc can't allocate at the given address when one is
+	 * given, it fails and returns NULL.
+	 */
+	ret = VirtualAlloc(addr, size, MEM_COMMIT | MEM_RESERVE,
+	    PAGE_READWRITE);
+#else
 	/*
 	 * We don't use MAP_FIXED here, because it can cause the *replacement*
 	 * of existing mappings, and we only want to create new mappings.
@@ -41,7 +49,7 @@
 		}
 		ret = NULL;
 	}
-
+#endif
 	assert(ret == NULL || (addr == NULL && ret != addr)
 	    || (addr != NULL && ret == addr));
 	return (ret);
@@ -51,55 +59,94 @@
 pages_unmap(void *addr, size_t size)
 {
 
-	if (munmap(addr, size) == -1) {
+#ifdef _WIN32
+	if (VirtualFree(addr, 0, MEM_RELEASE) == 0)
+#else
+	if (munmap(addr, size) == -1)
+#endif
+	{
 		char buf[BUFERROR_BUF];
 
 		buferror(errno, buf, sizeof(buf));
-		malloc_printf("<jemalloc>: Error in munmap(): %s\n", buf);
+		malloc_printf("<jemalloc>: Error in "
+#ifdef _WIN32
+		              "VirtualFree"
+#else
+		              "munmap"
+#endif
+		              "(): %s\n", buf);
 		if (opt_abort)
 			abort();
 	}
 }
 
+static void *
+pages_trim(void *addr, size_t alloc_size, size_t leadsize, size_t size)
+{
+	void *ret = (void *)((uintptr_t)addr + leadsize);
+
+	assert(alloc_size >= leadsize + size);
+#ifdef _WIN32
+	{
+		void *new_addr;
+
+		pages_unmap(addr, alloc_size);
+		new_addr = pages_map(ret, size);
+		if (new_addr == ret)
+			return (ret);
+		if (new_addr)
+			pages_unmap(new_addr, size);
+		return (NULL);
+	}
+#else
+	{
+		size_t trailsize = alloc_size - leadsize - size;
+
+		if (leadsize != 0)
+			pages_unmap(addr, leadsize);
+		if (trailsize != 0)
+			pages_unmap((void *)((uintptr_t)ret + size), trailsize);
+		return (ret);
+	}
+#endif
+}
+
 void
 pages_purge(void *addr, size_t length)
 {
 
-#ifdef JEMALLOC_PURGE_MADVISE_DONTNEED
-#  define JEMALLOC_MADV_PURGE MADV_DONTNEED
-#elif defined(JEMALLOC_PURGE_MADVISE_FREE)
-#  define JEMALLOC_MADV_PURGE MADV_FREE
+#ifdef _WIN32
+	VirtualAlloc(addr, length, MEM_RESET, PAGE_READWRITE);
 #else
-#  error "No method defined for purging unused dirty pages."
-#endif
+#  ifdef JEMALLOC_PURGE_MADVISE_DONTNEED
+#    define JEMALLOC_MADV_PURGE MADV_DONTNEED
+#  elif defined(JEMALLOC_PURGE_MADVISE_FREE)
+#    define JEMALLOC_MADV_PURGE MADV_FREE
+#  else
+#    error "No method defined for purging unused dirty pages."
+#  endif
 	madvise(addr, length, JEMALLOC_MADV_PURGE);
+#endif
 }
 
 static void *
-chunk_alloc_mmap_slow(size_t size, size_t alignment, bool unaligned, bool *zero)
+chunk_alloc_mmap_slow(size_t size, size_t alignment, bool *zero)
 {
 	void *ret, *pages;
-	size_t alloc_size, leadsize, trailsize;
+	size_t alloc_size, leadsize;
 
 	alloc_size = size + alignment - PAGE;
 	/* Beware size_t wrap-around. */
 	if (alloc_size < size)
 		return (NULL);
-	pages = pages_map(NULL, alloc_size);
-	if (pages == NULL)
-		return (NULL);
-	leadsize = ALIGNMENT_CEILING((uintptr_t)pages, alignment) -
-	    (uintptr_t)pages;
-	assert(alloc_size >= leadsize + size);
-	trailsize = alloc_size - leadsize - size;
-	ret = (void *)((uintptr_t)pages + leadsize);
-	if (leadsize != 0) {
-		/* Note that mmap() returned an unaligned mapping. */
-		unaligned = true;
-		pages_unmap(pages, leadsize);
-	}
-	if (trailsize != 0)
-		pages_unmap((void *)((uintptr_t)ret + size), trailsize);
+	do {
+		pages = pages_map(NULL, alloc_size);
+		if (pages == NULL)
+			return (NULL);
+		leadsize = ALIGNMENT_CEILING((uintptr_t)pages, alignment) -
+		    (uintptr_t)pages;
+		ret = pages_trim(pages, alloc_size, leadsize, size);
+	} while (ret == NULL);
 
 	assert(ret != NULL);
 	*zero = true;
@@ -144,6 +191,9 @@
 
 	offset = ALIGNMENT_ADDR2OFFSET(ret, alignment);
 	if (offset != 0) {
+#ifdef _WIN32
+		return (chunk_alloc_mmap_slow(size, alignment, zero));
+#else
 		/* Try to extend chunk boundary. */
 		if (pages_map((void *)((uintptr_t)ret + size), chunksize -
 		    offset) == NULL) {
@@ -152,13 +202,13 @@
 			 * reliable-but-expensive method.
 			 */
 			pages_unmap(ret, size);
-			return (chunk_alloc_mmap_slow(size, alignment, true,
-			    zero));
+			return (chunk_alloc_mmap_slow(size, alignment, zero));
 		} else {
 			/* Clean up unneeded leading space. */
 			pages_unmap(ret, chunksize - offset);
 			ret = (void *)((uintptr_t)ret + (chunksize - offset));
 		}
+#endif
 	}
 
 	assert(ret != NULL);
diff --git a/src/jemalloc.c b/src/jemalloc.c
index f9c8916..67ac90b 100644
--- a/src/jemalloc.c
+++ b/src/jemalloc.c
@@ -52,7 +52,19 @@
 #endif
 
 /* Used to avoid initialization races. */
+#ifdef _WIN32
+static malloc_mutex_t	init_lock;
+
+JEMALLOC_ATTR(constructor)
+static void
+init_init_lock()
+{
+
+	malloc_mutex_init(&init_lock);
+}
+#else
 static malloc_mutex_t	init_lock = MALLOC_MUTEX_INITIALIZER;
+#endif
 
 typedef struct {
 	void	*p;	/* Input pointer (as in realloc(p, s)). */
@@ -229,11 +241,17 @@
 	unsigned ret;
 	long result;
 
+#ifdef _WIN32
+	SYSTEM_INFO si;
+	GetSystemInfo(&si);
+	result = si.dwNumberOfProcessors;
+#else
 	result = sysconf(_SC_NPROCESSORS_ONLN);
 	if (result == -1) {
 		/* Error. */
 		ret = 1;
 	}
+#endif
 	ret = (unsigned)result;
 
 	return (ret);
@@ -369,13 +387,14 @@
 			}
 			break;
 		case 1: {
+#ifndef _WIN32
 			int linklen;
 			const char *linkname =
-#ifdef JEMALLOC_PREFIX
+#  ifdef JEMALLOC_PREFIX
 			    "/etc/"JEMALLOC_PREFIX"malloc.conf"
-#else
+#  else
 			    "/etc/malloc.conf"
-#endif
+#  endif
 			    ;
 
 			if ((linklen = readlink(linkname, buf,
@@ -386,7 +405,9 @@
 				 */
 				buf[linklen] = '\0';
 				opts = buf;
-			} else {
+			} else
+#endif
+			{
 				/* No configuration specified. */
 				buf[0] = '\0';
 				opts = buf;
@@ -610,7 +631,8 @@
 
 	malloc_conf_init();
 
-#if (!defined(JEMALLOC_MUTEX_INIT_CB) && !defined(JEMALLOC_ZONE))
+#if (!defined(JEMALLOC_MUTEX_INIT_CB) && !defined(JEMALLOC_ZONE) \
+    && !defined(_WIN32))
 	/* Register fork handlers. */
 	if (pthread_atfork(jemalloc_prefork, jemalloc_postfork_parent,
 	    jemalloc_postfork_child) != 0) {
diff --git a/src/mutex.c b/src/mutex.c
index 4b8ce57..159d82a 100644
--- a/src/mutex.c
+++ b/src/mutex.c
@@ -1,10 +1,14 @@
 #define	JEMALLOC_MUTEX_C_
 #include "jemalloc/internal/jemalloc_internal.h"
 
-#ifdef JEMALLOC_LAZY_LOCK
+#if defined(JEMALLOC_LAZY_LOCK) && !defined(_WIN32)
 #include <dlfcn.h>
 #endif
 
+#ifndef _CRT_SPINCOUNT
+#define _CRT_SPINCOUNT 4000
+#endif
+
 /******************************************************************************/
 /* Data. */
 
@@ -16,7 +20,7 @@
 static malloc_mutex_t	*postponed_mutexes = NULL;
 #endif
 
-#ifdef JEMALLOC_LAZY_LOCK
+#if defined(JEMALLOC_LAZY_LOCK) && !defined(_WIN32)
 static void	pthread_create_once(void);
 #endif
 
@@ -26,7 +30,7 @@
  * process goes multi-threaded.
  */
 
-#ifdef JEMALLOC_LAZY_LOCK
+#if defined(JEMALLOC_LAZY_LOCK) && !defined(_WIN32)
 static int (*pthread_create_fptr)(pthread_t *__restrict, const pthread_attr_t *,
     void *(*)(void *), void *__restrict);
 
@@ -68,7 +72,11 @@
 bool
 malloc_mutex_init(malloc_mutex_t *mutex)
 {
-#ifdef JEMALLOC_OSSPIN
+#ifdef _WIN32
+	if (!InitializeCriticalSectionAndSpinCount(&mutex->lock,
+	    _CRT_SPINCOUNT))
+		return (true);
+#elif (defined(JEMALLOC_OSSPIN))
 	mutex->lock = 0;
 #elif (defined(JEMALLOC_MUTEX_INIT_CB))
 	if (postpone_init) {
diff --git a/src/tsd.c b/src/tsd.c
index 281a2e9..09f06e8 100644
--- a/src/tsd.c
+++ b/src/tsd.c
@@ -31,7 +31,7 @@
 	not_reached();
 }
 
-#ifdef JEMALLOC_MALLOC_THREAD_CLEANUP
+#if defined(JEMALLOC_MALLOC_THREAD_CLEANUP) || defined(_WIN32)
 JEMALLOC_ATTR(visibility("default"))
 void
 _malloc_thread_cleanup(void)
@@ -70,3 +70,28 @@
 
 	ncleanups = 0;
 }
+
+#ifdef _WIN32
+static BOOL WINAPI
+_tls_callback(HINSTANCE hinstDLL, DWORD fdwReason, LPVOID lpvReserved)
+{
+
+	switch (fdwReason) {
+#ifdef JEMALLOC_LAZY_LOCK
+	case DLL_THREAD_ATTACH:
+		isthreaded = true;
+		break;
+#endif
+	case DLL_THREAD_DETACH:
+		_malloc_thread_cleanup();
+		break;
+	default:
+		break;
+	}
+	return (true);
+}
+
+JEMALLOC_ATTR(section(".CRT$XLY")) JEMALLOC_ATTR(used)
+static const BOOL	(WINAPI *tls_callback)(HINSTANCE hinstDLL,
+    DWORD fdwReason, LPVOID lpvReserved) = _tls_callback;
+#endif
diff --git a/src/util.c b/src/util.c
index 99ae26d..ee9efdf 100644
--- a/src/util.c
+++ b/src/util.c
@@ -67,7 +67,12 @@
 int
 buferror(int errnum, char *buf, size_t buflen)
 {
-#ifdef _GNU_SOURCE
+
+#ifdef _WIN32
+	FormatMessageA(FORMAT_MESSAGE_FROM_SYSTEM, NULL, GetLastError(), 0,
+	    (LPSTR)buf, buflen, NULL);
+	return (0);
+#elif defined(_GNU_SOURCE)
 	char *b = strerror_r(errno, buf, buflen);
 	if (b != buf) {
 		strncpy(buf, b, buflen);
diff --git a/test/jemalloc_test.h.in b/test/jemalloc_test.h.in
index 8833a03..e38b48e 100644
--- a/test/jemalloc_test.h.in
+++ b/test/jemalloc_test.h.in
@@ -7,6 +7,29 @@
 #include "jemalloc/internal/jemalloc_internal.h"
 
 /* Abstraction layer for threading in tests */
+#ifdef _WIN32
+#include <windows.h>
+
+typedef HANDLE je_thread_t;
+
+void
+je_thread_create(je_thread_t *thread, void *(*proc)(void *), void *arg)
+{
+	LPTHREAD_START_ROUTINE routine = (LPTHREAD_START_ROUTINE)proc;
+	*thread = CreateThread(NULL, 0, routine, arg, 0, NULL);
+	if (*thread == NULL) {
+		malloc_printf("Error in CreateThread()\n");
+		exit(1);
+	}
+}
+
+void
+je_thread_join(je_thread_t thread, void **ret)
+{
+	WaitForSingleObject(thread, INFINITE);
+}
+
+#else
 #include <pthread.h>
 
 typedef pthread_t je_thread_t;
@@ -27,3 +50,4 @@
 
 	pthread_join(thread, ret);
 }
+#endif
diff --git a/test/rallocm.c b/test/rallocm.c
index 18db5ee..f2a4770 100644
--- a/test/rallocm.c
+++ b/test/rallocm.c
@@ -19,9 +19,15 @@
 
 	/* Get page size. */
 	{
+#ifdef _WIN32
+		SYSTEM_INFO si;
+		GetSystemInfo(&si);
+		pagesize = (size_t)si.dwPageSize;
+#else
 		long result = sysconf(_SC_PAGESIZE);
 		assert(result != -1);
 		pagesize = (size_t)result;
+#endif
 	}
 
 	r = allocm(&p, &sz, 42, 0);