Add zfree_with_size to optimize sdsfree since we can get zmalloc_size from the header (#453)

### Description ###
zfree updates memory statistics. It gets the size of the buffer from
jemalloc by calling zmalloc_size. This operation is costly. We can avoid
it if we know the buffer size. For example, we can calculate size of sds
from the data we have in its header.

This commit introduces zfree_with_size function that accepts both
pointer to a buffer, and its size. zfree is refactored to call
zfree_with_size.

sdsfree uses the new interface for all but SDS_TYPE_5.

### Benchmark ###

Dataset is 3 million strings. Each benchmark run uses its own value size
(8192, 512, and 120). The benchmark is 100% write load for 5 minutes.

```
value size       new tps      old tps      %       new us/call    old us/call    %
8k               272088.53    269971.75    0.78    1.83           1.92           -4.69
512              356881.91    352856.72    1.14    1.27           1.35           -5.93
120              377523.81    368774.78    2.37    1.14           1.19           -4.20
```

---------

Signed-off-by: Vadym Khoptynets <vadymkh@amazon.com>
Signed-off-by: Madelyn Olson <madelyneolson@gmail.com>
Co-authored-by: Madelyn Olson <madelyneolson@gmail.com>

src/sds.c

@@ -195,7 +195,7 @@ sds sdsdup(const sds s) {
 /* Free an sds string. No operation is performed if 's' is NULL. */
 void sdsfree(sds s) {
     if (s == NULL) return;
-    s_free((char *)s - sdsHdrSize(s[-1]));
+    s_free_with_size(sdsAllocPtr(s), sdsAllocSize(s));
 }
 
 /* Set the sds string length to the length as obtained with strlen(), so
@@ -369,7 +369,7 @@ sds sdsResize(sds s, size_t size, int would_regrow) {
          * We aim to avoid calling realloc() when using Jemalloc if there is no
          * change in the allocation size, as it incurs a cost even if the
          * allocation size stays the same. */
-        bufsize = zmalloc_size(sh);
+        bufsize = sdsAllocSize(s);
         alloc_already_optimal = (je_nallocx(newlen, 0) == bufsize);
 #endif
         if (!alloc_already_optimal) {
@@ -412,8 +412,13 @@ sds sdsResize(sds s, size_t size, int would_regrow) {
  * 4) The implicit null term.
  */
 size_t sdsAllocSize(sds s) {
-    size_t alloc = sdsalloc(s);
-    return sdsHdrSize(s[-1]) + alloc + 1;
+    char type = s[-1] & SDS_TYPE_MASK;
+    /* SDS_TYPE_5 header doesn't contain the size of the allocation */
+    if (type == SDS_TYPE_5) {
+        return s_malloc_size(sdsAllocPtr(s));
+    } else {
+        return sdsHdrSize(type) + sdsalloc(s) + 1;
+    }
 }
 
 /* Return the pointer of the actual SDS allocation (normally SDS strings

src/sdsalloc.h

@@ -45,6 +45,7 @@
 #define s_trymalloc ztrymalloc
 #define s_tryrealloc ztryrealloc
 #define s_free zfree
+#define s_free_with_size zfree_with_size
 #define s_malloc_usable zmalloc_usable
 #define s_realloc_usable zrealloc_usable
 #define s_trymalloc_usable ztrymalloc_usable

src/zmalloc.c

@@ -31,6 +31,7 @@
 #include "fmacros.h"
 #include "config.h"
 #include "solarisfixes.h"
+#include "serverassert.h"
 
 #include <stdio.h>
 #include <stdlib.h>
@@ -361,22 +362,48 @@ size_t zmalloc_usable_size(void *ptr) {
 }
 #endif
 
+/* Frees the memory buffer pointed to by ptr and updates statistics. When using
+ * jemalloc it uses the fast track by specifying the buffer size.
+ *
+ * ptr must have been returned by a previous call to the system allocator which
+ * returned the usable size, such as zmalloc_usable. ptr must not be NULL. The
+ * caller is responsible to provide the actual allocation size, which may be
+ * different from the requested size. */
+static inline void zfree_internal(void *ptr, size_t size) {
+    assert(ptr != NULL);
+    update_zmalloc_stat_free(size);
+
+#ifdef USE_JEMALLOC
+    je_sdallocx(ptr, size, 0);
+#else
+    free(ptr);
+#endif
+}
+
 void zfree(void *ptr) {
-#ifndef HAVE_MALLOC_SIZE
-    void *realptr;
-    size_t oldsize;
+    if (ptr == NULL) return;
+
+#ifdef HAVE_MALLOC_SIZE
+    size_t size = zmalloc_size(ptr);
+#else
+    ptr = (char *)ptr - PREFIX_SIZE;
+    size_t data_size = *((size_t *)ptr);
+    size_t size = data_size + PREFIX_SIZE;
 #endif
 
+    zfree_internal(ptr, size);
+}
+
+/* Like zfree(), but doesn't call zmalloc_size(). */
+void zfree_with_size(void *ptr, size_t size) {
     if (ptr == NULL) return;
-#ifdef HAVE_MALLOC_SIZE
-    update_zmalloc_stat_free(zmalloc_size(ptr));
-    free(ptr);
-#else
-    realptr = (char *)ptr - PREFIX_SIZE;
-    oldsize = *((size_t *)realptr);
-    update_zmalloc_stat_free(oldsize + PREFIX_SIZE);
-    free(realptr);
+
+#ifndef HAVE_MALLOC_SIZE
+    ptr = (char *)ptr - PREFIX_SIZE;
+    size += PREFIX_SIZE;
 #endif
+
+    zfree_internal(ptr, size);
 }
 
 char *zstrdup(const char *s) {
@@ -604,8 +631,6 @@ size_t zmalloc_get_rss(void) {
 
 #if defined(USE_JEMALLOC)
 
-#include "serverassert.h"
-
 #define STRINGIFY_(x) #x
 #define STRINGIFY(x) STRINGIFY_(x)
 

src/zmalloc.h

@@ -118,6 +118,7 @@ __attribute__((malloc, alloc_size(1), noinline)) void *ztrymalloc(size_t size);
 __attribute__((malloc, alloc_size(1), noinline)) void *ztrycalloc(size_t size);
 __attribute__((alloc_size(2), noinline)) void *ztryrealloc(void *ptr, size_t size);
 void zfree(void *ptr);
+void zfree_with_size(void *ptr, size_t size);
 void *zmalloc_usable(size_t size, size_t *usable);
 void *zcalloc_usable(size_t size, size_t *usable);
 void *zrealloc_usable(void *ptr, size_t size, size_t *usable);