Huffman ASM

build/meson/lib/meson.build

@@ -37,6 +37,7 @@ libzstd_sources = [join_paths(zstd_rootdir, 'lib/common/entropy_common.c'),
   join_paths(zstd_rootdir, 'lib/compress/zstd_opt.c'),
   join_paths(zstd_rootdir, 'lib/compress/zstd_ldm.c'),
   join_paths(zstd_rootdir, 'lib/decompress/huf_decompress.c'),
+  join_paths(zstd_rootdir, 'lib/decompress/huf_decompress_amd64.S'),
   join_paths(zstd_rootdir, 'lib/decompress/zstd_decompress.c'),
   join_paths(zstd_rootdir, 'lib/decompress/zstd_decompress_block.c'),
   join_paths(zstd_rootdir, 'lib/decompress/zstd_ddict.c'),

build/single_file_libs/zstd-in.c

@@ -43,6 +43,8 @@
 #define ZSTD_MULTITHREAD
 #endif
 #define ZSTD_TRACE 0
+/* TODO: Can't amalgamate ASM function */
+#define HUF_DISABLE_ASM 1
 
 /* Include zstd_deps.h first with all the options we need enabled. */
 #define ZSTD_DEPS_NEED_MALLOC

build/single_file_libs/zstddeclib-in.c

@@ -39,6 +39,8 @@
 #define ZSTD_LEGACY_SUPPORT 0
 #define ZSTD_STRIP_ERROR_STRINGS
 #define ZSTD_TRACE 0
+/* TODO: Can't amalgamate ASM function */
+#define HUF_DISABLE_ASM 1
 
 /* Include zstd_deps.h first with all the options we need enabled. */
 #define ZSTD_DEPS_NEED_MALLOC

contrib/linux-kernel/Makefile

@@ -35,6 +35,7 @@ libzstd:
 		-DXXH_STATIC_LINKING_ONLY \
 		-DMEM_FORCE_MEMORY_ACCESS=0 \
 		-D__GNUC__ \
+		-D__linux__=1 \
 		-DSTATIC_BMI2=0 \
 		-DZSTD_ADDRESS_SANITIZER=0 \
 		-DZSTD_MEMORY_SANITIZER=0 \

contrib/linux-kernel/decompress_sources.h

@@ -21,6 +21,11 @@
 #include "common/error_private.c"
 #include "common/fse_decompress.c"
 #include "common/zstd_common.c"
+/*
+ * Disable the ASM Huffman implementation because we need to
+ * include all the sources.
+ */
+#define HUF_DISABLE_ASM 1
 #include "decompress/huf_decompress.c"
 #include "decompress/zstd_ddict.c"
 #include "decompress/zstd_decompress.c"

contrib/linux-kernel/test/macro-test.sh

@@ -42,3 +42,4 @@ test_not_present "ZSTD_DLL_IMPORT"
 test_not_present "__ICCARM__"
 test_not_present "_MSC_VER"
 test_not_present "_WIN32"
+test_not_present "__linux__"

lib/common/compiler.h

@@ -108,7 +108,7 @@
   #if ((defined(__clang__) && __has_attribute(__target__)) \
       || (defined(__GNUC__) \
           && (__GNUC__ >= 5 || (__GNUC__ == 4 && __GNUC_MINOR__ >= 8)))) \
-      && (defined(__x86_64__) || defined(_M_X86)) \
+      && (defined(__x86_64__) || defined(_M_X64)) \
       && !defined(__BMI2__)
   #  define DYNAMIC_BMI2 1
   #else
@@ -212,7 +212,7 @@
 #  elif defined(ZSTD_ARCH_ARM_NEON)
 #    include <arm_neon.h>
 #  endif
-#endif  
+#endif
 
 /* compat. with non-clang compilers */
 #ifndef __has_builtin

lib/common/error_private.h

@@ -22,6 +22,8 @@ extern "C" {
 *  Dependencies
 ******************************************/
 #include "../zstd_errors.h"  /* enum list */
+#include "compiler.h"
+#include "debug.h"
 #include "zstd_deps.h"       /* size_t */
 
 
@@ -73,6 +75,83 @@ ERR_STATIC const char* ERR_getErrorName(size_t code)
     return ERR_getErrorString(ERR_getErrorCode(code));
 }
 
+/**
+ * Ignore: this is an internal helper.
+ *
+ * This is a helper function to help force C99-correctness during compilation.
+ * Under strict compilation modes, variadic macro arguments can't be empty.
+ * However, variadic function arguments can be. Using a function therefore lets
+ * us statically check that at least one (string) argument was passed,
+ * independent of the compilation flags.
+ */
+static INLINE_KEYWORD UNUSED_ATTR
+void _force_has_format_string(const char *format, ...) {
+  (void)format;
+}
+
+/**
+ * Ignore: this is an internal helper.
+ *
+ * We want to force this function invocation to be syntactically correct, but
+ * we don't want to force runtime evaluation of its arguments.
+ */
+#define _FORCE_HAS_FORMAT_STRING(...) \
+  if (0) { \
+    _force_has_format_string(__VA_ARGS__); \
+  }
+
+#define ERR_QUOTE(str) #str
+
+/**
+ * Return the specified error if the condition evaluates to true.
+ *
+ * In debug modes, prints additional information.
+ * In order to do that (particularly, printing the conditional that failed),
+ * this can't just wrap RETURN_ERROR().
+ */
+#define RETURN_ERROR_IF(cond, err, ...) \
+  if (cond) { \
+    RAWLOG(3, "%s:%d: ERROR!: check %s failed, returning %s", \
+           __FILE__, __LINE__, ERR_QUOTE(cond), ERR_QUOTE(ERROR(err))); \
+    _FORCE_HAS_FORMAT_STRING(__VA_ARGS__); \
+    RAWLOG(3, ": " __VA_ARGS__); \
+    RAWLOG(3, "\n"); \
+    return ERROR(err); \
+  }
+
+/**
+ * Unconditionally return the specified error.
+ *
+ * In debug modes, prints additional information.
+ */
+#define RETURN_ERROR(err, ...) \
+  do { \
+    RAWLOG(3, "%s:%d: ERROR!: unconditional check failed, returning %s", \
+           __FILE__, __LINE__, ERR_QUOTE(ERROR(err))); \
+    _FORCE_HAS_FORMAT_STRING(__VA_ARGS__); \
+    RAWLOG(3, ": " __VA_ARGS__); \
+    RAWLOG(3, "\n"); \
+    return ERROR(err); \
+  } while(0);
+
+/**
+ * If the provided expression evaluates to an error code, returns that error code.
+ *
+ * In debug modes, prints additional information.
+ */
+#define FORWARD_IF_ERROR(err, ...) \
+  do { \
+    size_t const err_code = (err); \
+    if (ERR_isError(err_code)) { \
+      RAWLOG(3, "%s:%d: ERROR!: forwarding error in %s: %s", \
+             __FILE__, __LINE__, ERR_QUOTE(err), ERR_getErrorName(err_code)); \
+      _FORCE_HAS_FORMAT_STRING(__VA_ARGS__); \
+      RAWLOG(3, ": " __VA_ARGS__); \
+      RAWLOG(3, "\n"); \
+      return err_code; \
+    } \
+  } while(0);
+
 #if defined (__cplusplus)
 }
 #endif

lib/common/huf.h

@@ -116,11 +116,11 @@ HUF_PUBLIC_API size_t HUF_compress4X_wksp (void* dst, size_t dstCapacity,
 
 
 /* *** Constants *** */
-#define HUF_TABLELOG_MAX      12      /* max runtime value of tableLog (due to static allocation); can be modified up to HUF_ABSOLUTEMAX_TABLELOG */
+#define HUF_TABLELOG_MAX      12      /* max runtime value of tableLog (due to static allocation); can be modified up to HUF_TABLELOG_ABSOLUTEMAX */
 #define HUF_TABLELOG_DEFAULT  11      /* default tableLog value when none specified */
 #define HUF_SYMBOLVALUE_MAX  255
 
-#define HUF_TABLELOG_ABSOLUTEMAX  15  /* absolute limit of HUF_MAX_TABLELOG. Beyond that value, code does not work */
+#define HUF_TABLELOG_ABSOLUTEMAX  12  /* absolute limit of HUF_MAX_TABLELOG. Beyond that value, code does not work */
 #if (HUF_TABLELOG_MAX > HUF_TABLELOG_ABSOLUTEMAX)
 #  error "HUF_TABLELOG_MAX is too large !"
 #endif
@@ -353,6 +353,9 @@ size_t HUF_decompress4X_hufOnly_wksp_bmi2(HUF_DTable* dctx, void* dst, size_t ds
 #ifndef HUF_FORCE_DECOMPRESS_X2
 size_t HUF_readDTableX1_wksp_bmi2(HUF_DTable* DTable, const void* src, size_t srcSize, void* workSpace, size_t wkspSize, int bmi2);
 #endif
+#ifndef HUF_FORCE_DECOMPRESS_X1
+size_t HUF_readDTableX2_wksp_bmi2(HUF_DTable* DTable, const void* src, size_t srcSize, void* workSpace, size_t wkspSize, int bmi2);
+#endif
 
 #endif /* HUF_STATIC_LINKING_ONLY */
 

lib/common/zstd_internal.h

@@ -58,81 +58,6 @@ extern "C" {
 #define MIN(a,b) ((a)<(b) ? (a) : (b))
 #define MAX(a,b) ((a)>(b) ? (a) : (b))
 
-/**
- * Ignore: this is an internal helper.
- *
- * This is a helper function to help force C99-correctness during compilation.
- * Under strict compilation modes, variadic macro arguments can't be empty.
- * However, variadic function arguments can be. Using a function therefore lets
- * us statically check that at least one (string) argument was passed,
- * independent of the compilation flags.
- */
-static INLINE_KEYWORD UNUSED_ATTR
-void _force_has_format_string(const char *format, ...) {
-  (void)format;
-}
-
-/**
- * Ignore: this is an internal helper.
- *
- * We want to force this function invocation to be syntactically correct, but
- * we don't want to force runtime evaluation of its arguments.
- */
-#define _FORCE_HAS_FORMAT_STRING(...) \
-  if (0) { \
-    _force_has_format_string(__VA_ARGS__); \
-  }
-
-/**
- * Return the specified error if the condition evaluates to true.
- *
- * In debug modes, prints additional information.
- * In order to do that (particularly, printing the conditional that failed),
- * this can't just wrap RETURN_ERROR().
- */
-#define RETURN_ERROR_IF(cond, err, ...) \
-  if (cond) { \
-    RAWLOG(3, "%s:%d: ERROR!: check %s failed, returning %s", \
-           __FILE__, __LINE__, ZSTD_QUOTE(cond), ZSTD_QUOTE(ERROR(err))); \
-    _FORCE_HAS_FORMAT_STRING(__VA_ARGS__); \
-    RAWLOG(3, ": " __VA_ARGS__); \
-    RAWLOG(3, "\n"); \
-    return ERROR(err); \
-  }
-
-/**
- * Unconditionally return the specified error.
- *
- * In debug modes, prints additional information.
- */
-#define RETURN_ERROR(err, ...) \
-  do { \
-    RAWLOG(3, "%s:%d: ERROR!: unconditional check failed, returning %s", \
-           __FILE__, __LINE__, ZSTD_QUOTE(ERROR(err))); \
-    _FORCE_HAS_FORMAT_STRING(__VA_ARGS__); \
-    RAWLOG(3, ": " __VA_ARGS__); \
-    RAWLOG(3, "\n"); \
-    return ERROR(err); \
-  } while(0);
-
-/**
- * If the provided expression evaluates to an error code, returns that error code.
- *
- * In debug modes, prints additional information.
- */
-#define FORWARD_IF_ERROR(err, ...) \
-  do { \
-    size_t const err_code = (err); \
-    if (ERR_isError(err_code)) { \
-      RAWLOG(3, "%s:%d: ERROR!: forwarding error in %s: %s", \
-             __FILE__, __LINE__, ZSTD_QUOTE(err), ERR_getErrorName(err_code)); \
-      _FORCE_HAS_FORMAT_STRING(__VA_ARGS__); \
-      RAWLOG(3, ": " __VA_ARGS__); \
-      RAWLOG(3, "\n"); \
-      return err_code; \
-    } \
-  } while(0);
-
 
 /*-*************************************
 *  Common constants
@@ -453,6 +378,51 @@ MEM_STATIC U32 ZSTD_highbit32(U32 val)   /* compress, dictBuilder, decodeCorpus
     }
 }
 
+/**
+ * Computes CTZ on a U64.
+ * This will be slow on 32-bit mode, and on unsupported compilers.
+ * If you need this function to be fast (because it is hot) expand
+ * support.
+ */
+MEM_STATIC unsigned ZSTD_countTrailingZeros(size_t val)
+{
+    if (MEM_64bits()) {
+#       if defined(_MSC_VER) && defined(_WIN64)
+#           if STATIC_BMI2
+                return _tzcnt_u64(val);
+#           else
+                unsigned long r = 0;
+                return _BitScanForward64( &r, (U64)val ) ? (unsigned)(r >> 3) : 0;
+#           endif
+#       elif defined(__GNUC__) && (__GNUC__ >= 4)
+            return __builtin_ctzll((U64)val);
+#       else
+            static const int DeBruijnBytePos[64] = {  0,  1,  2,  7,  3, 13,  8, 19,
+                                                      4, 25, 14, 28,  9, 34, 20, 56,
+                                                      5, 17, 26, 54, 15, 41, 29, 43,
+                                                      10, 31, 38, 35, 21, 45, 49, 57,
+                                                      63,  6, 12, 18, 24, 27, 33, 55,
+                                                      16, 53, 40, 42, 30, 37, 44, 48,
+                                                      62, 11, 23, 32, 52, 39, 36, 47,
+                                                      61, 22, 51, 46, 60, 50, 59, 58 };
+            return DeBruijnBytePos[((U64)((val & -(long long)val) * 0x0218A392CDABBD3FULL)) >> 58];
+#       endif
+    } else { /* 32 bits */
+#       if defined(_MSC_VER)
+            unsigned long r=0;
+            return _BitScanForward( &r, (U32)val ) ? (unsigned)(r >> 3) : 0;
+#       elif defined(__GNUC__) && (__GNUC__ >= 3)
+            return (__builtin_ctz((U32)val) >> 3);
+#       else
+            static const int DeBruijnBytePos[32] = {  0,  1, 28,  2, 29, 14, 24,  3,
+                                                     30, 22, 20, 15, 25, 17,  4,  8,
+                                                     31, 27, 13, 23, 21, 19, 16,  7,
+                                                     26, 12, 18,  6, 11,  5, 10,  9 };
+            return DeBruijnBytePos[((U32)((val & -(S32)val) * 0x077CB531U)) >> 27];
+#       endif
+    }
+}
+
 
 /* ZSTD_invalidateRepCodes() :
  * ensures next compression will not use repcodes from previous block.

lib/compress/huf_compress.c

@@ -809,6 +809,7 @@ FORCE_INLINE_TEMPLATE void HUF_addBits(HUF_CStream_t* bitC, HUF_CElt elt, int id
     {
         size_t const nbBits = HUF_getNbBits(elt);
         size_t const dirtyBits = nbBits == 0 ? 0 : BIT_highbit32((U32)nbBits) + 1;
+        (void)dirtyBits;
         /* Middle bits are 0. */
         assert(((elt >> dirtyBits) << (dirtyBits + nbBits)) == 0);
         /* We didn't overwrite any bits in the bit container. */