valkey-io · Cuda-Chen · Oct 6, 2025
diff --git a/src/config.h b/src/config.h
@@ -383,6 +383,7 @@ void setcpuaffinity(const char *cpulist);
 
 #if HAVE_X86_SIMD
 #define ATTRIBUTE_TARGET_SSE2 __attribute__((target("sse2")))
+#define ATTRIBUTE_TARGET_PCLMUL __attribute__((target("pclmul")))
 #define ATTRIBUTE_TARGET_AVX2 __attribute__((target("avx2")))
 #define ATTRIBUTE_TARGET_AVX512 __attribute__((target("avx512f,avx512bw,avx512vl")))
 #else

diff --git a/src/crc16.c b/src/crc16.c
@@ -1,5 +1,8 @@
 #include "server.h"
 
+#include <smmintrin.h>
+#include <wmmintrin.h>
+
 /*
  * Copyright 2001-2010 Georges Menie (www.menie.org)
  * Copyright 2010-2012 Redis Ltd. (adapted to Redis coding style)
@@ -44,6 +47,54 @@
  * Output for "123456789"     : 31C3
  */
 
+#if HAVE_X86_SIMD
+ATTRIBUTE_TARGET_SSE2
+ATTRIBUTE_TARGET_PCLMUL
+static inline uint16_t crc16_base(uint16_t crc, uint8_t v) {
+	crc ^= v << 8;
+    __m128i orig = _mm_set_epi64x(0x0, (uint64_t)(crc) << (8));
+    __m128i tmp = orig;
+
+    uint64_t p = 0x11021;
+    //uint64_t p = 0x10811;
+
+    // 2^96 / p
+    // 0x111303471a041b343e569
+//#define mu 0x859b040b1c581911
+//#define mu 0x111303471a041b343e569
+
+    // 0x111303471a041
+    // 1 0001 0001 0011 0000 0011 0100 0111 0001 1010 0000 0100 0001
+    // 1000 0010 0000 0101 1000 1110 0010 1100 0000 1100 1000 1000 1
+    // 1[000 0|010 0|000 0|101 1|000 1|110 0|010 1|100 0|000 1|100 1|000 1|000 1]
+    // 1  0      4     0     b    1     c      5    8      1     9     1    1
+    //uint64_t mu = 0x1040b1c581911; // 2^64 / p
+    uint64_t mu = 0x111303471a041;
+
+    // 0x11130
+    // 1 0001 0001 0011 0000
+    // 0000 1100 1000 1000 1
+    // 0[000 1|100 1|000 1|000 1]
+    // 0    1    9    1     1
+    //uint64_t mu = 0x101911; // 2^32 / p
+    //uint64_t mu = 0x11130;
+
+    __m128i mul = _mm_set_epi64x(p, mu);
+
+    tmp = _mm_clmulepi64_si128(
+            tmp,
+            mul, 0x00);
+
+    tmp = _mm_clmulepi64_si128(tmp, 
+            mul, 0x11);
+
+    tmp = _mm_xor_si128(tmp, orig);
+
+    uint16_t ret = (uint16_t)(_mm_extract_epi16(tmp, 0x0));
+
+    return ret;
+}
+#else
 static const uint16_t crc16tab[256]= {
     0x0000,0x1021,0x2042,0x3063,0x4084,0x50a5,0x60c6,0x70e7,
     0x8108,0x9129,0xa14a,0xb16b,0xc18c,0xd1ad,0xe1ce,0xf1ef,
@@ -78,11 +129,22 @@ static const uint16_t crc16tab[256]= {
     0xef1f,0xff3e,0xcf5d,0xdf7c,0xaf9b,0xbfba,0x8fd9,0x9ff8,
     0x6e17,0x7e36,0x4e55,0x5e74,0x2e93,0x3eb2,0x0ed1,0x1ef0
 };
+#endif
 
 uint16_t crc16(const char *buf, int len) {
+#if HAVE_X86_SIMD
+    int counter;
+    uint16_t crc = 0x0; 
+    for(counter = 0; counter < len; counter++) {
+        uint8_t tmp = (uint8_t)buf[counter];
+        crc = crc16_base(crc, tmp);
+    }   
+    return crc;
+#else
     int counter;
     uint16_t crc = 0;
     for (counter = 0; counter < len; counter++)
             crc = (crc<<8) ^ crc16tab[((crc>>8) ^ *buf++)&0x00FF];
     return crc;
+#endif
 }
diff --git a/src/unit/test_crc16.c b/src/unit/test_crc16.c
@@ -0,0 +1,81 @@
+#include <stdio.h>
+#include <string.h>
+#include "../crc16.c"
+
+#include "test_help.h"
+
+/* CRC16 implementation according to CCITT standards.
+ *
+ * Note by @antirez: this is actually the XMODEM CRC 16 algorithm, using the
+ * following parameters:
+ *
+ * Name                       : "XMODEM", also known as "ZMODEM", "CRC-16/ACORN"
+ * Width                      : 16 bit
+ * Poly                       : 1021 (That is actually x^16 + x^12 + x^5 + 1)
+ * Initialization             : 0000
+ * Reflect Input byte         : False
+ * Reflect Output CRC         : False
+ * Xor constant to output CRC : 0000
+ * Output for "123456789"     : 31C3
+ */
+
+static const uint16_t crc16tab[256] = {
+    0x0000, 0x1021, 0x2042, 0x3063, 0x4084, 0x50a5, 0x60c6, 0x70e7,
+    0x8108, 0x9129, 0xa14a, 0xb16b, 0xc18c, 0xd1ad, 0xe1ce, 0xf1ef,
+    0x1231, 0x0210, 0x3273, 0x2252, 0x52b5, 0x4294, 0x72f7, 0x62d6,
+    0x9339, 0x8318, 0xb37b, 0xa35a, 0xd3bd, 0xc39c, 0xf3ff, 0xe3de,
+    0x2462, 0x3443, 0x0420, 0x1401, 0x64e6, 0x74c7, 0x44a4, 0x5485,
+    0xa56a, 0xb54b, 0x8528, 0x9509, 0xe5ee, 0xf5cf, 0xc5ac, 0xd58d,
+    0x3653, 0x2672, 0x1611, 0x0630, 0x76d7, 0x66f6, 0x5695, 0x46b4,
+    0xb75b, 0xa77a, 0x9719, 0x8738, 0xf7df, 0xe7fe, 0xd79d, 0xc7bc,
+    0x48c4, 0x58e5, 0x6886, 0x78a7, 0x0840, 0x1861, 0x2802, 0x3823,
+    0xc9cc, 0xd9ed, 0xe98e, 0xf9af, 0x8948, 0x9969, 0xa90a, 0xb92b,
+    0x5af5, 0x4ad4, 0x7ab7, 0x6a96, 0x1a71, 0x0a50, 0x3a33, 0x2a12,
+    0xdbfd, 0xcbdc, 0xfbbf, 0xeb9e, 0x9b79, 0x8b58, 0xbb3b, 0xab1a,
+    0x6ca6, 0x7c87, 0x4ce4, 0x5cc5, 0x2c22, 0x3c03, 0x0c60, 0x1c41,
+    0xedae, 0xfd8f, 0xcdec, 0xddcd, 0xad2a, 0xbd0b, 0x8d68, 0x9d49,
+    0x7e97, 0x6eb6, 0x5ed5, 0x4ef4, 0x3e13, 0x2e32, 0x1e51, 0x0e70,
+    0xff9f, 0xefbe, 0xdfdd, 0xcffc, 0xbf1b, 0xaf3a, 0x9f59, 0x8f78,
+    0x9188, 0x81a9, 0xb1ca, 0xa1eb, 0xd10c, 0xc12d, 0xf14e, 0xe16f,
+    0x1080, 0x00a1, 0x30c2, 0x20e3, 0x5004, 0x4025, 0x7046, 0x6067,
+    0x83b9, 0x9398, 0xa3fb, 0xb3da, 0xc33d, 0xd31c, 0xe37f, 0xf35e,
+    0x02b1, 0x1290, 0x22f3, 0x32d2, 0x4235, 0x5214, 0x6277, 0x7256,
+    0xb5ea, 0xa5cb, 0x95a8, 0x8589, 0xf56e, 0xe54f, 0xd52c, 0xc50d,
+    0x34e2, 0x24c3, 0x14a0, 0x0481, 0x7466, 0x6447, 0x5424, 0x4405,
+    0xa7db, 0xb7fa, 0x8799, 0x97b8, 0xe75f, 0xf77e, 0xc71d, 0xd73c,
+    0x26d3, 0x36f2, 0x0691, 0x16b0, 0x6657, 0x7676, 0x4615, 0x5634,
+    0xd94c, 0xc96d, 0xf90e, 0xe92f, 0x99c8, 0x89e9, 0xb98a, 0xa9ab,
+    0x5844, 0x4865, 0x7806, 0x6827, 0x18c0, 0x08e1, 0x3882, 0x28a3,
+    0xcb7d, 0xdb5c, 0xeb3f, 0xfb1e, 0x8bf9, 0x9bd8, 0xabbb, 0xbb9a,
+    0x4a75, 0x5a54, 0x6a37, 0x7a16, 0x0af1, 0x1ad0, 0x2ab3, 0x3a92,
+    0xfd2e, 0xed0f, 0xdd6c, 0xcd4d, 0xbdaa, 0xad8b, 0x9de8, 0x8dc9,
+    0x7c26, 0x6c07, 0x5c64, 0x4c45, 0x3ca2, 0x2c83, 0x1ce0, 0x0cc1,
+    0xef1f, 0xff3e, 0xcf5d, 0xdf7c, 0xaf9b, 0xbfba, 0x8fd9, 0x9ff8,
+    0x6e17, 0x7e36, 0x4e55, 0x5e74, 0x2e93, 0x3eb2, 0x0ed1, 0x1ef0};
+
+uint16_t crc16_ref(const char *buf, int len) {
+    int counter;
+    uint16_t crc = 0;
+    for (counter = 0; counter < len; counter++)
+        crc = (crc << 8) ^ crc16tab[((crc >> 8) ^ *buf++) & 0x00FF];
+    return crc;
+}
+
+int test_crc16(int argc, char **argv, int flags) {
+    UNUSED(argc);
+    UNUSED(argv);
+    UNUSED(flags);
+
+    const char *s = "123456789";
+    int len = strlen(s);
+    TEST_ASSERT(crc16(s, len) == crc16_ref(s, len));
+
+    // Test all permutation of bits in a byte
+    for (uint16_t i = 0; i <= 0xff; i++) {
+        char input[1];
+        input[0] = (char)i;
+        int sz = sizeof(input);
+        TEST_ASSERT(crc16(input, sz) == crc16_ref(input, sz));
+    }
+    return 0;
+}
diff --git a/src/unit/test_files.h b/src/unit/test_files.h
@@ -7,6 +7,7 @@ typedef struct unitTest {
 } unitTest;
 
 int test_popcount(int argc, char **argv, int flags);
+int test_crc16(int argc, char **argv, int flags);
 int test_crc64(int argc, char **argv, int flags);
 int test_crc64combine(int argc, char **argv, int flags);
 int test_dictCreate(int argc, char **argv, int flags);
@@ -256,6 +257,7 @@ int test_zmallocAllocReallocCallocAndFree(int argc, char **argv, int flags);
 int test_zmallocAllocZeroByteAndFree(int argc, char **argv, int flags);
 
 unitTest __test_bitops_c[] = {{"test_popcount", test_popcount}, {NULL, NULL}};
+unitTest __test_crc16_c[] = {{"test_crc16", test_crc16}, {NULL, NULL}};
 unitTest __test_crc64_c[] = {{"test_crc64", test_crc64}, {NULL, NULL}};
 unitTest __test_crc64combine_c[] = {{"test_crc64combine", test_crc64combine}, {NULL, NULL}};
 unitTest __test_dict_c[] = {{"test_dictCreate", test_dictCreate}, {"test_dictAdd16Keys", test_dictAdd16Keys}, {"test_dictDisableResize", test_dictDisableResize}, {"test_dictAddOneKeyTriggerResize", test_dictAddOneKeyTriggerResize}, {"test_dictDeleteKeys", test_dictDeleteKeys}, {"test_dictDeleteOneKeyTriggerResize", test_dictDeleteOneKeyTriggerResize}, {"test_dictEmptyDirAdd128Keys", test_dictEmptyDirAdd128Keys}, {"test_dictDisableResizeReduceTo3", test_dictDisableResizeReduceTo3}, {"test_dictDeleteOneKeyTriggerResizeAgain", test_dictDeleteOneKeyTriggerResizeAgain}, {"test_dictBenchmark", test_dictBenchmark}, {NULL, NULL}};
@@ -285,6 +287,7 @@ struct unitTestSuite {
     unitTest *tests;
 } unitTestSuite[] = {
     {"test_bitops.c", __test_bitops_c},
+    {"test_crc16.c", __test_crc16_c},
     {"test_crc64.c", __test_crc64_c},
     {"test_crc64combine.c", __test_crc64combine_c},
     {"test_dict.c", __test_dict_c},