[bmi2] Add lzcnt and bmi target attributes

* When dynamic dispatching to bmi2 add lzcnt and bmi to the
  TARGET_ATTRIBUTE.
* Centralize the bmi2 TARGET_ATTRIBUTE definition to
  BMI2_TARGET_ATTRIBUTE so we can change it in the future.
* Only enable bmi2 when both bmi1 & bmi2 are supported. There shouldn't
  be any cases where bmi2 is supported but bmi1 isn't. But, since we are
  using the instruction we should check bmi1 as well.

lib/common/compiler.h

@@ -101,6 +101,13 @@
 #  define TARGET_ATTRIBUTE(target)
 #endif
 
+/* Target attribute for BMI2 dynamic dispatch.
+ * Enable lzcnt, bmi, and bmi2.
+ * We test for bmi1 & bmi2. lzcnt is included in bmi1.
+ */
+#define BMI2_TARGET_ATTRIBUTE TARGET_ATTRIBUTE("lzcnt,bmi,bmi2")
+
+
 /* Enable runtime BMI2 dispatch based on the CPU.
  * Enabled for clang & gcc >=4.8 on x86 when BMI2 isn't enabled by default.
  */

lib/common/entropy_common.c

@@ -223,7 +223,7 @@ static size_t FSE_readNCount_body_default(
 }
 
 #if DYNAMIC_BMI2
-TARGET_ATTRIBUTE("bmi2") static size_t FSE_readNCount_body_bmi2(
+BMI2_TARGET_ATTRIBUTE static size_t FSE_readNCount_body_bmi2(
         short* normalizedCounter, unsigned* maxSVPtr, unsigned* tableLogPtr,
         const void* headerBuffer, size_t hbSize)
 {
@@ -343,7 +343,7 @@ static size_t HUF_readStats_body_default(BYTE* huffWeight, size_t hwSize, U32* r
 }
 
 #if DYNAMIC_BMI2
-static TARGET_ATTRIBUTE("bmi2") size_t HUF_readStats_body_bmi2(BYTE* huffWeight, size_t hwSize, U32* rankStats,
+static BMI2_TARGET_ATTRIBUTE size_t HUF_readStats_body_bmi2(BYTE* huffWeight, size_t hwSize, U32* rankStats,
                      U32* nbSymbolsPtr, U32* tableLogPtr,
                      const void* src, size_t srcSize,
                      void* workSpace, size_t wkspSize)

lib/common/fse_decompress.c

@@ -365,7 +365,7 @@ static size_t FSE_decompress_wksp_body_default(void* dst, size_t dstCapacity, co
 }
 
 #if DYNAMIC_BMI2
-TARGET_ATTRIBUTE("bmi2") static size_t FSE_decompress_wksp_body_bmi2(void* dst, size_t dstCapacity, const void* cSrc, size_t cSrcSize, unsigned maxLog, void* workSpace, size_t wkspSize)
+BMI2_TARGET_ATTRIBUTE static size_t FSE_decompress_wksp_body_bmi2(void* dst, size_t dstCapacity, const void* cSrc, size_t cSrcSize, unsigned maxLog, void* workSpace, size_t wkspSize)
 {
     return FSE_decompress_wksp_body(dst, dstCapacity, cSrc, cSrcSize, maxLog, workSpace, wkspSize, 1);
 }

lib/common/zstd_internal.h

@@ -20,6 +20,7 @@
 *  Dependencies
 ***************************************/
 #include "compiler.h"
+#include "cpu.h"
 #include "mem.h"
 #include "debug.h"                 /* assert, DEBUGLOG, RAWLOG, g_debuglevel */
 #include "error_private.h"
@@ -472,6 +473,14 @@ size_t ZSTD_getcBlockSize(const void* src, size_t srcSize,
 size_t ZSTD_decodeSeqHeaders(ZSTD_DCtx* dctx, int* nbSeqPtr,
                        const void* src, size_t srcSize);
 
+/**
+ * @returns true iff the CPU supports dynamic BMI2 dispatch.
+ */
+MEM_STATIC int ZSTD_cpuSupportsBmi2(void)
+{
+    ZSTD_cpuid_t cpuid = ZSTD_cpuid();
+    return ZSTD_cpuid_bmi1(cpuid) && ZSTD_cpuid_bmi2(cpuid);
+}
 
 #if defined (__cplusplus)
 }