Optimize deferred replies to use shared objects instead of sprintf (#10334)

Avoid sprintf/ll2string on setDeferredAggregateLen()/addReplyLongLongWithPrefix() when we can used shared objects.
In some pipelined workloads this achieves about 10% improvement.

Co-authored-by: Oran Agra <oran@redislabs.com>

src/networking.c

@@ -714,6 +714,24 @@ void setDeferredAggregateLen(client *c, void *node, long length, char prefix) {
      * we return NULL in addReplyDeferredLen() */
     if (node == NULL) return;
 
+    /* Things like *2\r\n, %3\r\n or ~4\r\n are emitted very often by the protocol
+     * so we have a few shared objects to use if the integer is small
+     * like it is most of the times. */
+    const size_t hdr_len = OBJ_SHARED_HDR_STRLEN(length);
+    const int opt_hdr = length < OBJ_SHARED_BULKHDR_LEN;
+    if (prefix == '*' && opt_hdr) {
+        setDeferredReply(c, node, shared.mbulkhdr[length]->ptr, hdr_len);
+        return;
+    }
+    if (prefix == '%' && opt_hdr) {
+        setDeferredReply(c, node, shared.maphdr[length]->ptr, hdr_len);
+        return;
+    }
+    if (prefix == '~' && opt_hdr) {
+        setDeferredReply(c, node, shared.sethdr[length]->ptr, hdr_len);
+        return;
+    }
+
     char lenstr[128];
     size_t lenstr_len = sprintf(lenstr, "%c%ld\r\n", prefix, length);
     setDeferredReply(c, node, lenstr, lenstr_len);
@@ -806,12 +824,19 @@ void addReplyLongLongWithPrefix(client *c, long long ll, char prefix) {
     /* Things like $3\r\n or *2\r\n are emitted very often by the protocol
      * so we have a few shared objects to use if the integer is small
      * like it is most of the times. */
-    if (prefix == '*' && ll < OBJ_SHARED_BULKHDR_LEN && ll >= 0) {
+    const int opt_hdr = ll < OBJ_SHARED_BULKHDR_LEN && ll >= 0;
+    if (prefix == '*' && opt_hdr) {
         addReply(c,shared.mbulkhdr[ll]);
         return;
-    } else if (prefix == '$' && ll < OBJ_SHARED_BULKHDR_LEN && ll >= 0) {
+    } else if (prefix == '$' && opt_hdr) {
         addReply(c,shared.bulkhdr[ll]);
         return;
+    } else if (prefix == '%' && opt_hdr) {
+        addReply(c,shared.maphdr[ll]);
+        return;
+    } else if (prefix == '~' && opt_hdr) {
+        addReply(c,shared.sethdr[ll]);
+        return;
     }
 
     buf[0] = prefix;

src/server.c

@@ -1790,6 +1790,10 @@ void createSharedObjects(void) {
             sdscatprintf(sdsempty(),"*%d\r\n",j));
         shared.bulkhdr[j] = createObject(OBJ_STRING,
             sdscatprintf(sdsempty(),"$%d\r\n",j));
+        shared.maphdr[j] = createObject(OBJ_STRING,
+            sdscatprintf(sdsempty(),"%%%d\r\n",j));
+        shared.sethdr[j] = createObject(OBJ_STRING,
+            sdscatprintf(sdsempty(),"~%d\r\n",j));
     }
     /* The following two shared objects, minstring and maxstring, are not
      * actually used for their value but as a special object meaning

src/server.h

@@ -109,6 +109,7 @@ typedef long long ustime_t; /* microsecond time type. */
 #define PROTO_SHARED_SELECT_CMDS 10
 #define OBJ_SHARED_INTEGERS 10000
 #define OBJ_SHARED_BULKHDR_LEN 32
+#define OBJ_SHARED_HDR_STRLEN(_len_) (((_len_) < 10) ? 4 : 5) /* see shared.mbulkhdr etc. */
 #define LOG_MAX_LEN    1024 /* Default maximum length of syslog messages.*/
 #define AOF_REWRITE_ITEMS_PER_CMD 64
 #define AOF_ANNOTATION_LINE_MAX_LEN 1024
@@ -1229,7 +1230,9 @@ struct sharedObjectsStruct {
     *select[PROTO_SHARED_SELECT_CMDS],
     *integers[OBJ_SHARED_INTEGERS],
     *mbulkhdr[OBJ_SHARED_BULKHDR_LEN], /* "*<value>\r\n" */
-    *bulkhdr[OBJ_SHARED_BULKHDR_LEN];  /* "$<value>\r\n" */
+    *bulkhdr[OBJ_SHARED_BULKHDR_LEN],  /* "$<value>\r\n" */
+    *maphdr[OBJ_SHARED_BULKHDR_LEN],   /* "%<value>\r\n" */
+    *sethdr[OBJ_SHARED_BULKHDR_LEN];   /* "~<value>\r\n" */
     sds minstring, maxstring;
 };