diff --git a/src/memory.c b/src/memory.c
index f9c53782..cad2d774 100644
--- a/src/memory.c
+++ b/src/memory.c
@@ -247,24 +247,11 @@ static inline size_t mi_bsf(uintptr_t x) {
   #endif
   return idx;
 }
-static inline size_t mi_bsr(uintptr_t x) {
-  if (x==0) return 8*MI_INTPTR_SIZE;
-  DWORD idx;
-  #if (MI_INTPTR_SIZE==8)
-  _BitScanReverse64(&idx, x);
-  #else
-  _BitScanReverse(&idx, x);
-  #endif
-  return idx;
-}
 #elif defined(__GNUC__) || defined(__clang__)
 #define MI_HAVE_BITSCAN
 static inline size_t mi_bsf(uintptr_t x) {
   return (x==0 ? 8*MI_INTPTR_SIZE : __builtin_ctzl(x));
 }
-static inline size_t mi_bsr(uintptr_t x) {
-  return (x==0 ? 8*MI_INTPTR_SIZE : (8*MI_INTPTR_SIZE - 1) - __builtin_clzl(x));
-}
 #endif
 
 // Allocate `blocks` in a `region` at `idx` of a given `size`.
@@ -310,7 +297,7 @@ static bool mi_region_alloc_blocks(mem_region_t* region, size_t idx, size_t bloc
     else {
       // on to the next bit range
       #ifdef MI_HAVE_BITSCAN
-      size_t shift = (blocks == 1 ? 1 : mi_bsr(map & m) - bitidx + 1);
+      size_t shift = (blocks == 1 ? 1 : _mi_bsr(map & m) - bitidx + 1);
       mi_assert_internal(shift > 0 && shift <= blocks);
       #else
       size_t shift = 1;
diff --git a/src/page-queue.c b/src/page-queue.c
index 4af70b50..4e554047 100644
--- a/src/page-queue.c
+++ b/src/page-queue.c
@@ -59,10 +59,20 @@ static inline uint8_t mi_bsr32(uint32_t x) {
   _BitScanReverse((DWORD*)&idx, x);
   return (uint8_t)idx;
 }
+#define HAVE_MI_BSR64
+static inline size_t mi_bsr64(uint64_t x) {
+  uint64_t idx;
+  _BitScanReverse64((DWORD*)&idx, x);
+  return (uint8_t)idx;
+}
 #elif defined(__GNUC__) || defined(__clang__)
 static inline uint8_t mi_bsr32(uint32_t x) {
   return (31 - __builtin_clz(x));
 }
+#define HAVE_MI_BSR64
+static inline size_t mi_bsr64(uint64_t x) {
+  return (63 - __builtin_clzl(x));
+}
 #else
 static inline uint8_t mi_bsr32(uint32_t x) {
   // de Bruijn multiplication, see <http://supertech.csail.mit.edu/papers/debruijn.pdf>
@@ -84,8 +94,12 @@ static inline uint8_t mi_bsr32(uint32_t x) {
 uint8_t _mi_bsr(uintptr_t x) {
   if (x == 0) return 0;
 #if MI_INTPTR_SIZE==8
+  #ifdef HAVE_MI_BSR64
+  return mi_bsr64(x);
+  #else
   uint32_t hi = (x >> 32);
   return (hi == 0 ? mi_bsr32((uint32_t)x) : 32 + mi_bsr32(hi));
+  #endif
 #elif MI_INTPTR_SIZE==4
   return mi_bsr32(x);
 #else