MDEV-22749 Implement portable PCLMUL accelerated crc32() with Intel intrinsics

Removed some inine assembly, replaced by code from
https://github.com/intel/soft-crc

Also,replace GCC inline assembly for cpuid in ut0crc32 with __cpuid,
to fix "PIC register clobbered by 'ebx' in 'asm'.
This enables fast CRC32C on 32bit Intel processors with GCC.

mysys/CMakeLists.txt

@@ -60,20 +60,21 @@ ENDIF()
 
 IF(MSVC)
   SET(HAVE_CPUID_INSTRUCTION 1 CACHE BOOL "")
-ELSEIF(CMAKE_SYSTEM_PROCESSOR MATCHES "x86_64|amd64|i[36]86")
-  #Check for CPUID and PCLMUL. GCC before version 5 would refuse to emit the
-  #CPUID instruction for -m32 -fPIC because it would clobber the EBX register.
-  CHECK_C_SOURCE_COMPILES("
-  int main()
-  {
-    asm volatile (\"cpuid\" : : \"a\"(1) : \"ebx\");
-    asm volatile (\"pclmulqdq \\$0x00, %%xmm1, %%xmm0\":::\"cc\");
-    return 0;
-  }" HAVE_CLMUL_INSTRUCTION)
-
-  IF(HAVE_CLMUL_INSTRUCTION)
+  SET(HAVE_CLMUL_INSTRUCTION 1 CACHE BOOL "")
+  SET(MYSYS_SOURCES ${MYSYS_SOURCES} crc32/crc32_x86.c)
+  IF(CLANG_CL)
+    SET_SOURCE_FILES_PROPERTIES(crc32/crc32_x86.c PROPERTIES COMPILE_FLAGS "-msse4.2 -mpclmul")
+  ENDIF()
+ELSEIF(CMAKE_SYSTEM_PROCESSOR MATCHES "x86_64|amd64|i386|i686")
   SET(HAVE_CPUID_INSTRUCTION 1 CACHE BOOL "")
+  MY_CHECK_C_COMPILER_FLAG(-msse4.2)
+  MY_CHECK_C_COMPILER_FLAG(-mpclmul)
+  CHECK_INCLUDE_FILE(cpuid.h HAVE_CPUID_H)
+  CHECK_INCLUDE_FILE(x86intrin.h HAVE_X86INTRIN_H)
+  IF(have_C__msse4.2 AND have_C__mpclmul AND HAVE_CPUID_H AND HAVE_X86INTRIN_H)
+    SET(HAVE_CLMUL_INSTRUCTION 1 CACHE BOOL "")
     SET(MYSYS_SOURCES ${MYSYS_SOURCES} crc32/crc32_x86.c)
+    SET_SOURCE_FILES_PROPERTIES(crc32/crc32_x86.c PROPERTIES COMPILE_FLAGS "-msse4.2 -mpclmul")
   ENDIF()
 ELSEIF(CMAKE_SYSTEM_PROCESSOR MATCHES "aarch64|AARCH64")
   IF(CMAKE_COMPILER_IS_GNUCC AND NOT CMAKE_CXX_COMPILER_VERSION VERSION_LESS 5.1)

storage/innobase/ut/ut0crc32.cc

@@ -89,7 +89,9 @@ mysys/my_perf.c, contributed by Facebook under the following license.
 #ifdef HAVE_CPUID_INSTRUCTION
 # ifdef _MSC_VER
 #  include <intrin.h>
-# elif defined __GNUC__ && !defined __clang__ && __GNUC__ < 5
+# else
+#  include <cpuid.h>
+#    if defined __GNUC__ && !defined __clang__ && __GNUC__ < 5
 /* <nmmintrin.h> does not really work in GCC before version 5 */
 #      define _mm_crc32_u8(crc,data) __builtin_ia32_crc32qi(crc,data)
 #      define _mm_crc32_u32(crc,data) __builtin_ia32_crc32si(crc,data)
@@ -97,6 +99,7 @@ mysys/my_perf.c, contributed by Facebook under the following license.
 #    else
 #      include <nmmintrin.h>
 #    endif
+#  endif
 #endif
 
 /* CRC32 hardware implementation. */
@@ -122,9 +125,9 @@ static inline bool has_sse4_2()
   __cpuid(data, 1);
   return !!(data[2] & 1 << 20);
 #  else
-  uint32_t eax, ecx;
-  asm("cpuid" : "=a"(eax), "=c"(ecx) : "a"(1) : "ebx", "edx");
-  return !!(ecx & 1 << 20);
+  uint32_t reax = 0, rebx = 0, recx = 0, redx = 0;
+  __cpuid(1, reax, rebx, recx, redx);
+  return !!(recx & 1 << 20);
 #  endif
 }