liblzma: Create crc_clmul.c.

Both crc32_clmul() and crc64_clmul() are now exported from
crc32_clmul.c as lzma_crc32_clmul() and lzma_crc64_clmul(). This
ensures that is_clmul_supported() (now lzma_is_clmul_supported()) is
not duplicated between crc32_fast.c and crc64_fast.c.

Also, it encapsulates the complexity of the CLMUL implementations into a
single file and reduces the complexity of crc32_fast.c and crc64_fast.c.
Before, CLMUL code was present in crc32_fast.c, crc64_fast.c, and
crc_common.h.

During the conversion, various cleanups were applied to code (thanks to
Lasse Collin) including:

- Require using semicolons with MASK_/L/H/LH macros.
- Variable typing and const handling improvements.
- Improvements to comments.
- Fixes to the pragmas used.
- Removed unneeded variables.
- Whitespace improvements.
- Fixed CRC_USE_GENERIC_FOR_SMALL_INPUTS handling.
- Silenced warnings and removed the need for some #pragmas

1 files changed, 4 insertions, 116 deletions

diff --git a/src/liblzma/check/crc32_fast.c b/src/liblzma/check/crc32_fast.c
index 3ac9b6d7..8849a476 100644
--- a/src/liblzma/check/crc32_fast.c
+++ b/src/liblzma/check/crc32_fast.c
@@ -34,11 +34,11 @@
 #include "check.h"
 #include "crc_common.h"
 
+#ifdef CRC_GENERIC
+
 ///////////////////
 // Generic CRC32 //
 ///////////////////
-#ifdef CRC_GENERIC
-
 
 static uint32_t
 crc32_generic(const uint8_t *buf, size_t size, uint32_t crc)
@@ -99,118 +99,6 @@ crc32_generic(const uint8_t *buf, size_t size, uint32_t crc)
 }
 #endif
 
-
-/////////////////////
-// x86 CLMUL CRC32 //
-/////////////////////
-
-#ifdef CRC_CLMUL
-
-#include <immintrin.h>
-
-
-/*
-// These functions were used to generate the constants
-// at the top of crc32_clmul().
-static uint64_t
-calc_lo(uint64_t p, uint64_t a, int n)
-{
-	uint64_t b = 0; int i;
-	for (i = 0; i < n; i++) {
-		b = b >> 1 | (a & 1) << (n - 1);
-		a = (a >> 1) ^ ((0 - (a & 1)) & p);
-	}
-	return b;
-}
-
-// same as ~crc(&a, sizeof(a), ~0)
-static uint64_t
-calc_hi(uint64_t p, uint64_t a, int n)
-{
-	int i;
-	for (i = 0; i < n; i++)
-		a = (a >> 1) ^ ((0 - (a & 1)) & p);
-	return a;
-}
-*/
-
-
-// MSVC (VS2015 - VS2022) produces bad 32-bit x86 code from the CLMUL CRC
-// code when optimizations are enabled (release build). According to the bug
-// report, the ebx register is corrupted and the calculated result is wrong.
-// Trying to workaround the problem with "__asm mov ebx, ebx" didn't help.
-// The following pragma works and performance is still good. x86-64 builds
-// aren't affected by this problem.
-//
-// NOTE: Another pragma after the function restores the optimizations.
-// If the #if condition here is updated, the other one must be updated too.
-#if defined(_MSC_VER) && !defined(__INTEL_COMPILER) && !defined(__clang__) \
-		&& defined(_M_IX86)
-#	pragma optimize("g", off)
-#endif
-
-// EDG-based compilers (Intel's classic compiler and compiler for E2K) can
-// define __GNUC__ but the attribute must not be used with them.
-// The new Clang-based ICX needs the attribute.
-//
-// NOTE: Build systems check for this too, keep them in sync with this.
-#if (defined(__GNUC__) || defined(__clang__)) && !defined(__EDG__)
-__attribute__((__target__("ssse3,sse4.1,pclmul")))
-#endif
-static uint32_t
-crc32_clmul(const uint8_t *buf, size_t size, uint32_t crc)
-{
-	// The prototypes of the intrinsics use signed types while most of
-	// the values are treated as unsigned here. These warnings in this
-	// function have been checked and found to be harmless so silence them.
-#if TUKLIB_GNUC_REQ(4, 6) || defined(__clang__)
-#	pragma GCC diagnostic push
-#	pragma GCC diagnostic ignored "-Wsign-conversion"
-#	pragma GCC diagnostic ignored "-Wconversion"
-#endif
-
-#ifndef CRC_USE_GENERIC_FOR_SMALL_INPUTS
-	// The code assumes that there is at least one byte of input.
-	if (size == 0)
-		return crc;
-#endif
-
-	// uint32_t poly = 0xedb88320;
-	uint64_t p = 0x1db710640; // p << 1
-	uint64_t mu = 0x1f7011641; // calc_lo(p, p, 32) << 1 | 1
-	uint64_t k5 = 0x163cd6124; // calc_hi(p, p, 32) << 1
-	uint64_t k4 = 0x0ccaa009e; // calc_hi(p, p, 64) << 1
-	uint64_t k3 = 0x1751997d0; // calc_hi(p, p, 128) << 1
-
-	__m128i vfold4 = _mm_set_epi64x(mu, p);
-	__m128i vfold8 = _mm_set_epi64x(0, k5);
-	__m128i vfold16 = _mm_set_epi64x(k4, k3);
-
-	__m128i v0, v1, v2;
-
-	crc_simd_body(buf,  size, &v0, &v1, vfold16, _mm_cvtsi32_si128(~crc));
-
-	v1 = _mm_xor_si128(
-			_mm_clmulepi64_si128(v0, vfold16, 0x10), v1); // xxx0
-	v2 = _mm_shuffle_epi32(v1, 0xe7); // 0xx0
-	v0 = _mm_slli_epi64(v1, 32);  // [0]
-	v0 = _mm_clmulepi64_si128(v0, vfold8, 0x00);
-	v0 = _mm_xor_si128(v0, v2);   // [1] [2]
-	v2 = _mm_clmulepi64_si128(v0, vfold4, 0x10);
-	v2 = _mm_clmulepi64_si128(v2, vfold4, 0x00);
-	v0 = _mm_xor_si128(v0, v2);   // [2]
-	return ~_mm_extract_epi32(v0, 2);
-
-#if TUKLIB_GNUC_REQ(4, 6) || defined(__clang__)
-#	pragma GCC diagnostic pop
-#endif
-}
-#if defined(_MSC_VER) && !defined(__INTEL_COMPILER) && !defined(__clang__) \
-		&& defined(_M_IX86)
-#	pragma optimize("", on)
-#endif
-#endif
-
 #if defined(CRC_GENERIC) && defined(CRC_CLMUL)
 typedef uint32_t (*crc32_func_type)(
 		const uint8_t *buf, size_t size, uint32_t crc);
@@ -226,7 +114,7 @@ typedef uint32_t (*crc32_func_type)(
 static crc32_func_type
 crc32_resolve(void)
 {
-	return is_clmul_supported() ? &crc32_clmul : &crc32_generic;
+	return lzma_is_clmul_supported() ? &lzma_crc32_clmul : &crc32_generic;
 }
 
 #if defined(HAVE_FUNC_ATTRIBUTE_IFUNC) && defined(__clang__)
@@ -305,7 +193,7 @@ lzma_crc32(const uint8_t *buf, size_t size, uint32_t crc)
 	return crc32_func(buf, size, crc);
 
 #elif defined(CRC_CLMUL)
-	return crc32_clmul(buf, size, crc);
+	return lzma_crc32_clmul(buf, size, crc);
 
 #else
 	return crc32_generic(buf, size, crc);