minor optimization for -mbmi2 compilation flag

lib/compress/zstd_compress.c

@@ -112,7 +112,9 @@ static void ZSTD_initCCtx(ZSTD_CCtx* cctx, ZSTD_customMem memManager)
     assert(cctx != NULL);
     ZSTD_memset(cctx, 0, sizeof(*cctx));
     cctx->customMem = memManager;
+#if DYNAMIC_BMI2
     cctx->bmi2 = ZSTD_cpuSupportsBmi2();
+#endif
     {   size_t const err = ZSTD_CCtx_reset(cctx, ZSTD_reset_parameters);
         assert(!ZSTD_isError(err));
         (void)err;
@@ -152,7 +154,9 @@ ZSTD_CCtx* ZSTD_initStaticCCtx(void* workspace, size_t workspaceSize)
     cctx->blockState.nextCBlock = (ZSTD_compressedBlockState_t*)ZSTD_cwksp_reserve_object(&cctx->workspace, sizeof(ZSTD_compressedBlockState_t));
     cctx->tmpWorkspace = ZSTD_cwksp_reserve_object(&cctx->workspace, TMP_WORKSPACE_SIZE);
     cctx->tmpWkspSize = TMP_WORKSPACE_SIZE;
-    cctx->bmi2 = ZSTD_cpuid_bmi2(ZSTD_cpuid());
+#if DYNAMIC_BMI2
+    cctx->bmi2 = ZSTD_cpuSupportsBmi2();
+#endif
     return cctx;
 }
 
@@ -4152,7 +4156,7 @@ ZSTD_compressSeqStore_singleBlock(ZSTD_CCtx* zc,
                 op + ZSTD_blockHeaderSize, dstCapacity - ZSTD_blockHeaderSize,
                 srcSize,
                 zc->tmpWorkspace, zc->tmpWkspSize /* statically allocated in resetCCtx */,
-                zc->bmi2);
+                ZSTD_CCtx_get_bmi2(zc));
     FORWARD_IF_ERROR(cSeqsSize, "ZSTD_entropyCompressSeqStore failed!");
 
     if (!zc->isFirstBlock &&
@@ -4442,7 +4446,7 @@ ZSTD_compressBlock_internal(ZSTD_CCtx* zc,
             dst, dstCapacity,
             srcSize,
             zc->tmpWorkspace, zc->tmpWkspSize /* statically allocated in resetCCtx */,
-            zc->bmi2);
+            ZSTD_CCtx_get_bmi2(zc));
 
     if (frame &&
         /* We don't want to emit our first block as a RLE even if it qualifies because
@@ -7029,7 +7033,7 @@ ZSTD_compressSequences_internal(ZSTD_CCtx* cctx,
                                 op + ZSTD_blockHeaderSize /* Leave space for block header */, dstCapacity - ZSTD_blockHeaderSize,
                                 blockSize,
                                 cctx->tmpWorkspace, cctx->tmpWkspSize /* statically allocated in resetCCtx */,
-                                cctx->bmi2);
+                                ZSTD_CCtx_get_bmi2(cctx));
         FORWARD_IF_ERROR(compressedSeqsSize, "Compressing sequences of block failed");
         DEBUGLOG(5, "Compressed sequences size: %zu", compressedSeqsSize);
 
@@ -8057,7 +8061,7 @@ ZSTD_compressSequencesAndLiterals_internal(ZSTD_CCtx* cctx,
                                 &cctx->blockState.prevCBlock->entropy, &cctx->blockState.nextCBlock->entropy,
                                 &cctx->appliedParams,
                                 cctx->tmpWorkspace, cctx->tmpWkspSize /* statically allocated in resetCCtx */,
-                                cctx->bmi2);
+                                ZSTD_CCtx_get_bmi2(cctx));
         FORWARD_IF_ERROR(compressedSeqsSize, "Compressing sequences of block failed");
         /* note: the spec forbids for any compressed block to be larger than maximum block size */
         if (compressedSeqsSize > cctx->blockSizeMax) compressedSeqsSize = 0;

lib/compress/zstd_compress_internal.h

@@ -472,7 +472,11 @@ typedef struct {
 struct ZSTD_CCtx_s {
     ZSTD_compressionStage_e stage;
     int cParamsChanged;                  /* == 1 if cParams(except wlog) or compression level are changed in requestedParams. Triggers transmission of new params to ZSTDMT (if available) then reset to 0. */
-    int bmi2;                            /* == 1 if the CPU supports BMI2 and 0 otherwise. CPU support is determined dynamically once per context lifetime. */
+#if DYNAMIC_BMI2
+    int bmi2;                            /* == 1 if the CPU supports BMI1 & BMI2, determined once per context lifetime.
+                                          * Never read this directly: use ZSTD_CCtx_get_bmi2(), which also handles builds
+                                          * where BMI2 is enabled at compile time and this field does not exist. */
+#endif
     ZSTD_CCtx_params requestedParams;
     ZSTD_CCtx_params appliedParams;
     ZSTD_CCtx_params simpleApiParams;    /* Param storage used by the simple API - not sticky. Must only be used in top-level simple API functions for storage. */
@@ -545,6 +549,24 @@ struct ZSTD_CCtx_s {
     size_t extSeqBufCapacity;
 };
 
+/**
+ * @returns 1 if calls should be dispatched to the BMI2_TARGET_ATTRIBUTE variant
+ *          of a function, rather than to its default variant.
+ *
+ * Beware: this is *not* a "does the CPU support BMI2" test. When BMI2 is enabled
+ * at compile time, DYNAMIC_BMI2 is 0: the whole library is already compiled with
+ * BMI2, no separate variant exists, and this returns 0 even though the CPU does
+ * support BMI2. Only ever use it to select between the two variants of a function.
+ */
+MEM_STATIC int ZSTD_CCtx_get_bmi2(const struct ZSTD_CCtx_s* cctx) {
+#if DYNAMIC_BMI2
+    return cctx->bmi2;
+#else
+    (void)cctx;
+    return 0;
+#endif
+}
+
 typedef enum { ZSTD_dtlm_fast, ZSTD_dtlm_full } ZSTD_dictTableLoadMethod_e;
 typedef enum { ZSTD_tfp_forCCtx, ZSTD_tfp_forCDict } ZSTD_tableFillPurpose_e;
 

lib/compress/zstd_compress_superblock.c

@@ -683,6 +683,6 @@ size_t ZSTD_compressSuperBlock(ZSTD_CCtx* zc,
             &zc->appliedParams,
             dst, dstCapacity,
             src, srcSize,
-            zc->bmi2, lastBlock,
+            ZSTD_CCtx_get_bmi2(zc), lastBlock,
             zc->tmpWorkspace, zc->tmpWkspSize /* statically allocated in resetCCtx */);
 }

lib/decompress/zstd_decompress_internal.h

@@ -155,7 +155,9 @@ struct ZSTD_DCtx_s
     size_t staticSize;
     int isFrameDecompression;
 #if DYNAMIC_BMI2
-    int bmi2;                     /* == 1 if the CPU supports BMI2 and 0 otherwise. CPU support is determined dynamically once per context lifetime. */
+    int bmi2;                     /* == 1 if the CPU supports BMI1 & BMI2, determined once per context lifetime.
+                                   * Never read this directly: use ZSTD_DCtx_get_bmi2(), which also handles builds
+                                   * where BMI2 is enabled at compile time and this field does not exist. */
 #endif
 
     /* dictionary */
@@ -210,6 +212,15 @@ struct ZSTD_DCtx_s
 #endif
 };  /* typedef'd to ZSTD_DCtx within "zstd.h" */
 
+/**
+ * @returns 1 if calls should be dispatched to the BMI2_TARGET_ATTRIBUTE variant
+ *          of a function, rather than to its default variant.
+ *
+ * Beware: this is *not* a "does the CPU support BMI2" test. When BMI2 is enabled
+ * at compile time, DYNAMIC_BMI2 is 0: the whole library is already compiled with
+ * BMI2, no separate variant exists, and this returns 0 even though the CPU does
+ * support BMI2. Only ever use it to select between the two variants of a function.
+ */
 MEM_STATIC int ZSTD_DCtx_get_bmi2(const struct ZSTD_DCtx_s *dctx) {
 #if DYNAMIC_BMI2
     return dctx->bmi2;