Enable Fast CRC32 for Win64

Summary:
Currently the fast crc32 path is not enabled on Windows. I am trying to enable it here, hopefully, with the minimum impact to the existing code structure.
Closes https://github.com/facebook/rocksdb/pull/2033

Differential Revision: D4770635

Pulled By: siying

fbshipit-source-id: 676f8b8

CMakeLists.txt

@@ -42,6 +42,11 @@ endif()
 list(APPEND CMAKE_MODULE_PATH "${CMAKE_SOURCE_DIR}/cmake/modules/")
 
 if(WIN32)
+  if (DEFINED AVX2)
+    set(USE_AVX2 ${AVX2})
+  else ()
+    set(USE_AVX2 1)
+  endif ()
   include(${CMAKE_CURRENT_SOURCE_DIR}/thirdparty.inc)
 else()
   option(WITH_JEMALLOC "build with JeMalloc" OFF)
@@ -111,7 +116,11 @@ add_library(build_version OBJECT ${BUILD_VERSION_CC})
 target_include_directories(build_version PRIVATE
   ${CMAKE_CURRENT_SOURCE_DIR}/util)
 if(WIN32)
-  set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} /Zi /nologo  /EHsc /GS /Gd /GR /GF /fp:precise /Zc:wchar_t /Zc:forScope /errorReport:queue")
+  if (${USE_AVX2} EQUAL 1)
+    set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} /Zi /nologo  /arch:AVX2 /EHsc /GS /Gd /GR /GF /fp:precise /Zc:wchar_t /Zc:forScope /errorReport:queue")
+  else ()
+    set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} /Zi /nologo  /EHsc /GS /Gd /GR /GF /fp:precise /Zc:wchar_t /Zc:forScope /errorReport:queue")
+  endif ()
 
   set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} /FC /d2Zi+ /W3 /wd4127 /wd4800 /wd4996 /wd4351")
 else()

util/crc32c.cc

@@ -16,6 +16,11 @@
 #ifdef __SSE4_2__
 #include <nmmintrin.h>
 #endif
+#if defined(_WIN64)
+#ifdef __AVX2__
+#include <nmmintrin.h>
+#endif
+#endif
 #include "util/coding.h"
 
 namespace rocksdb {
@@ -299,6 +304,13 @@ static inline uint64_t LE_LOAD64(const uint8_t *p) {
 #endif
 #endif
 
+#if defined(_WIN64)
+#ifdef __AVX2__
+static inline uint64_t LE_LOAD64(const uint8_t *p) {
+  return DecodeFixed64(reinterpret_cast<const char*>(p));
+}
+#endif
+#endif
 static inline void Slow_CRC32(uint64_t* l, uint8_t const **p) {
   uint32_t c = static_cast<uint32_t>(*l ^ LE_LOAD32(*p));
   *p += 4;
@@ -326,6 +338,13 @@ static inline void Fast_CRC32(uint64_t* l, uint8_t const **p) {
   *l = _mm_crc32_u32(static_cast<unsigned int>(*l), LE_LOAD32(*p));
   *p += 4;
 #endif
+#elif defined(_WIN64)
+#ifdef __AVX2__
+  *l = _mm_crc32_u64(*l, LE_LOAD64(*p));
+  *p += 8;
+#else
+  Slow_CRC32(l, p);
+#endif
 #else
   Slow_CRC32(l, p);
 #endif
@@ -381,6 +400,10 @@ static bool isSSE42() {
   uint32_t d_;
   __asm__("cpuid" : "=c"(c_), "=d"(d_) : "a"(1) : "ebx");
   return c_ & (1U << 20);  // copied from CpuId.h in Folly.
+#elif defined(_WIN64)
+  int info[4];
+  __cpuidex(info, 0x00000001, 0);
+  return (info[2] & ((int)1 << 20)) != 0;
 #else
   return false;
 #endif
@@ -395,6 +418,8 @@ static inline Function Choose_Extend() {
 bool IsFastCrc32Supported() {
 #ifdef __SSE4_2__
   return isSSE42();
+#elif defined(_WIN64)
+  return isSSE42();
 #else
   return false;
 #endif