diff --git a/include/mimalloc-internal.h b/include/mimalloc-internal.h
index 16be12518..4ace8facc 100644
--- a/include/mimalloc-internal.h
+++ b/include/mimalloc-internal.h
@@ -131,6 +131,8 @@ void       _mi_heap_set_default_direct(mi_heap_t* heap);
 
 // "stats.c"
 void       _mi_stats_done(mi_stats_t* stats);
+void       _mi_histogram_log(size_t size);
+void       _mi_histogram_print(mi_output_fun* out);
 
 mi_msecs_t  _mi_clock_now(void);
 mi_msecs_t  _mi_clock_end(mi_msecs_t start);
diff --git a/include/mimalloc.h b/include/mimalloc.h
index 0b84f6c3a..eeeba6e5c 100644
--- a/include/mimalloc.h
+++ b/include/mimalloc.h
@@ -303,6 +303,7 @@ typedef enum mi_option_e {
   // stable options
   mi_option_show_errors,              // print error messages
   mi_option_show_stats,               // print statistics on termination
+  mi_option_show_histogram,           // print histogram
   mi_option_verbose,                  // print verbose messages
   // the following options are experimental (see src/options.h)
   mi_option_eager_commit,           
diff --git a/src/alloc.c b/src/alloc.c
index cd4afa1ec..be0fa5467 100644
--- a/src/alloc.c
+++ b/src/alloc.c
@@ -98,7 +98,11 @@ extern inline mi_decl_restrict void* mi_malloc_small(size_t size) mi_attr_noexce
 // The main allocation function
 extern inline mi_decl_restrict void* mi_heap_malloc(mi_heap_t* heap, size_t size) mi_attr_noexcept {
   if (mi_likely(size <= MI_SMALL_SIZE_MAX)) {
-    return mi_heap_malloc_small(heap, size);
+    void *p = mi_heap_malloc_small(heap, size);
+    #if MI_STAT>1
+    if (p) { _mi_histogram_log(size); }
+    #endif
+    return p;
   }
   else {
     mi_assert(heap!=NULL);
@@ -106,6 +110,9 @@ extern inline mi_decl_restrict void* mi_heap_malloc(mi_heap_t* heap, size_t size
     void* const p = _mi_malloc_generic(heap, size + MI_PADDING_SIZE);      // note: size can overflow but it is detected in malloc_generic
     mi_assert_internal(p == NULL || mi_usable_size(p) >= size);
     #if MI_STAT>1
+    if (p) { _mi_histogram_log(size); }
+    #endif
+    #if MI_STAT>1
     if (p != NULL) {
       if (!mi_heap_is_initialized(heap)) { heap = mi_get_default_heap(); }
       mi_heap_stat_increase(heap, malloc, mi_usable_size(p));
diff --git a/src/init.c b/src/init.c
index 90990cd4f..9ee0df19d 100644
--- a/src/init.c
+++ b/src/init.c
@@ -545,6 +545,11 @@ static void mi_process_done(void) {
   if (mi_option_is_enabled(mi_option_show_stats) || mi_option_is_enabled(mi_option_verbose)) {
     mi_stats_print(NULL);
   }
+  #if MI_STAT>1
+  if (mi_option_is_enabled(mi_option_show_histogram)) {
+    _mi_histogram_print(NULL);
+  }
+  #endif
   mi_allocator_done();  
   _mi_verbose_message("process done: 0x%zx\n", _mi_heap_main.thread_id);
   os_preloading = true; // don't call the C runtime anymore
diff --git a/src/options.c b/src/options.c
index c4ed91cba..5745a9002 100644
--- a/src/options.c
+++ b/src/options.c
@@ -63,6 +63,7 @@ static mi_option_desc_t options[_mi_option_last] =
   { 0, UNINIT, MI_OPTION(show_errors) },
 #endif
   { 0, UNINIT, MI_OPTION(show_stats) },
+  { 0, UNINIT, MI_OPTION(show_histogram) },
   { 0, UNINIT, MI_OPTION(verbose) },
 
   // the following options are experimental and not all combinations make sense.
diff --git a/src/stats.c b/src/stats.c
index 6d486f42b..1a5427e50 100644
--- a/src/stats.c
+++ b/src/stats.c
@@ -451,6 +451,85 @@ mi_msecs_t _mi_clock_end(mi_msecs_t start) {
 }
 
 
+/* -----------------------------------------------------------
+  Histogram of allocations sizes (power of 2)
+----------------------------------------------------------- */
+
+static _Atomic(size_t) mi_hist[MI_SIZE_BITS] = { 0 };
+
+void _mi_histogram_log(size_t size)
+{
+  if (mi_unlikely(size == 0)) return;
+  size_t bucket = mi_bsr(size);
+  mi_atomic_increment_relaxed(mi_hist + bucket);
+}
+
+static char* _mi_make_bar(char *buf, size_t buflen, size_t value, size_t max, size_t width)
+{
+  /* we can't dynamically detect if the terminal supports unicode block characters */
+  #if defined(_WIN32)
+  size_t v = value * width / max;
+  buf[0] = '\0';
+  while (v > 0) {
+    strncat(buf, "*", buflen--);
+    v--;
+  }
+  #else
+  static const char* a[] = { " ", "▏", "▎", "▍", "▌", "▋", "▊", "▉", "█" };
+  size_t v = value * width * 8 / max;
+  buf[0] = '\0';
+  while (v > 8) {
+    strncat(buf, a[8], buflen--);
+    v-=8;
+  }
+  strncat(buf, a[v], buflen--);
+  #endif
+  return buf;
+}
+
+static char* _mi_format_bytes(char *buf, size_t buflen, size_t count)
+{
+  #if MI_SIZE_BITS > 32
+  if (count & ~((((size_t)1) << 50) - (size_t)1)) {
+      snprintf(buf, buflen, "%zuPiB", (count+1) >> 50);
+  } else
+  if (count & ~((((size_t)1) << 40) - (size_t)1)) {
+      snprintf(buf, buflen, "%zuTiB", (count+1) >> 40);
+  } else
+  #endif
+  if (count & ~((((size_t)1) << 30) - (size_t)1)) {
+      snprintf(buf, buflen, "%zuGiB", (count+1) >> 30);
+  } else
+  if (count & ~((((size_t)1) << 20) - (size_t)1)) {
+      snprintf(buf, buflen, "%zuMiB", (count+1) >> 20);
+  } else
+  if (count & ~((((size_t)1) << 10) - (size_t)1)) {
+      snprintf(buf, buflen, "%zuKiB", (count+1) >> 10);
+  } else {
+      snprintf(buf, buflen, "%zuB", count);
+  }
+  return buf;
+}
+
+void _mi_histogram_print(mi_output_fun* out)
+{
+  #define _NCOLS 50
+  char bar[_NCOLS*3+1], num1[16], num2[16];
+  size_t max_allocs = 0;
+  for (size_t i = 0; i < MI_SIZE_BITS; i++) {
+    if (mi_hist[i] > max_allocs) { max_allocs = mi_hist[i]; }
+  }
+  _mi_fputs(out, NULL, NULL, "\nhistogram of allocation sizes\n");
+  for (size_t i = 0; i < MI_SIZE_BITS; i++) {
+    if (mi_hist[i]) {
+      _mi_fprintf(out, NULL, "%9s - %-9s [ %-9zu ] %s\n",
+        _mi_format_bytes(num1, sizeof(num1), ((size_t)1 << i)),
+        _mi_format_bytes(num2, sizeof(num2), ((size_t)1 << (i+1))-1),
+        mi_hist[i], _mi_make_bar(bar, sizeof(bar), mi_hist[i], max_allocs, _NCOLS));
+    }
+  }
+}
+
 // --------------------------------------------------------
 // Basic process statistics
 // --------------------------------------------------------