More simplifications to the websocket unmasking code

lpereira · Apr 28, 2024 · 00a4c53 · 00a4c53
1 parent f3e4036
commit 00a4c53
Showing 1 changed file with 20 additions and 32 deletions.
diff --git a/src/lib/lwan-websocket.c b/src/lib/lwan-websocket.c
@@ -157,14 +157,9 @@ static void unmask(char *msg, size_t msg_len, char mask[static 4])
 #if defined(__AVX2__)
     const __m256i mask256 =
         _mm256_castps_si256(_mm256_broadcast_ss((const float *)mask));
-    if (msg_len >= 32) {
-        do {
-            __m256i v = _mm256_lddqu_si256((const __m256i *)msg);
-            _mm256_storeu_si256((__m256i *)msg, _mm256_xor_si256(v, mask256));
-
-            msg += 32;
-            msg_len -= 32;
-        } while (msg_len >= 32);
+    for (; msg_len >= 32; msg_len -= 32, msg += 32) {
+        __m256i v = _mm256_lddqu_si256((const __m256i *)msg);
+        _mm256_storeu_si256((__m256i *)msg, _mm256_xor_si256(v, mask256));
     }
 #endif
 
@@ -176,40 +171,33 @@ static void unmask(char *msg, size_t msg_len, char mask[static 4])
 #else
     const __m128i mask128 = _mm_loadu_si128((const __m128i *)mask);
 #endif
-    if (msg_len >= 16) {
-        do {
+    for (; msg_len >= 16; msg_len -= 16, msg += 16) {
 #if defined(__SSE3__)
-            __m128i v = _mm_lddqu_si128((const __m128i *)msg);
+        __m128i v = _mm_lddqu_si128((const __m128i *)msg);
 #else
-            __m128i v = _mm_loadu_si128((const __m128i *)msg);
+        __m128i v = _mm_loadu_si128((const __m128i *)msg);
 #endif
 
-            _mm_storeu_si128((__m128i *)msg, _mm_xor_si128(v, mask128));
-
-            msg += 16;
-            msg_len -= 16;
-        } while (msg_len >= 16);
+        _mm_storeu_si128((__m128i *)msg, _mm_xor_si128(v, mask128));
     }
 #endif
 
-    if (sizeof(void *) == 8) {
-        if (msg_len >= 8) {
+    if (sizeof(void *) == 8 && msg_len >= 8) {
 #if defined(__SSE_4_1__)
-            /* We're far away enough from the AVX2 path that it's
-             * probably better to use mask128 instead of mask256
-             * here. */
-            const __int64 mask64 = _mm_extract_epi64(mask128, 0);
+        /* We're far away enough from the AVX2 path that it's
+         * probably better to use mask128 instead of mask256
+         * here. */
+        const __int64 mask64 = _mm_extract_epi64(mask128, 0);
 #else
-            const uint32_t mask32 = string_as_uint32(mask);
-            const uint64_t mask64 = (uint64_t)mask32 << 32 | (uint64_t)mask32;
+        const uint32_t mask32 = string_as_uint32(mask);
+        const uint64_t mask64 = (uint64_t)mask32 << 32 | (uint64_t)mask32;
 #endif
-            do {
-                uint64_t v = string_as_uint64(msg);
-                v ^= (uint64_t)mask64;
-                msg = mempcpy(msg, &v, sizeof(v));
-                msg_len -= 8;
-            } while (msg_len >= 8);
-        }
+        do {
+            uint64_t v = string_as_uint64(msg);
+            v ^= (uint64_t)mask64;
+            msg = mempcpy(msg, &v, sizeof(v));
+            msg_len -= 8;
+        } while (msg_len >= 8);
     }
 
     if (msg_len >= 4) {