deflate-ruby 0.1.0

data/ext/deflate_ruby/libdeflate/lib/utils.c

@@ -0,0 +1,141 @@
+/*
+ * utils.c - utility functions for libdeflate
+ *
+ * Copyright 2016 Eric Biggers
+ *
+ * Permission is hereby granted, free of charge, to any person
+ * obtaining a copy of this software and associated documentation
+ * files (the "Software"), to deal in the Software without
+ * restriction, including without limitation the rights to use,
+ * copy, modify, merge, publish, distribute, sublicense, and/or sell
+ * copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following
+ * conditions:
+ *
+ * The above copyright notice and this permission notice shall be
+ * included in all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+ * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES
+ * OF MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
+ * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT
+ * HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY,
+ * WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
+ * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ */
+
+#include "lib_common.h"
+
+#ifdef FREESTANDING
+#  define malloc NULL
+#  define free NULL
+#else
+#  include <stdlib.h>
+#endif
+
+malloc_func_t libdeflate_default_malloc_func = malloc;
+free_func_t libdeflate_default_free_func = free;
+
+void *
+libdeflate_aligned_malloc(malloc_func_t malloc_func,
+			  size_t alignment, size_t size)
+{
+	void *ptr = (*malloc_func)(sizeof(void *) + alignment - 1 + size);
+
+	if (ptr) {
+		void *orig_ptr = ptr;
+
+		ptr = (void *)ALIGN((uintptr_t)ptr + sizeof(void *), alignment);
+		((void **)ptr)[-1] = orig_ptr;
+	}
+	return ptr;
+}
+
+void
+libdeflate_aligned_free(free_func_t free_func, void *ptr)
+{
+	(*free_func)(((void **)ptr)[-1]);
+}
+
+LIBDEFLATEAPI void
+libdeflate_set_memory_allocator(malloc_func_t malloc_func,
+				free_func_t free_func)
+{
+	libdeflate_default_malloc_func = malloc_func;
+	libdeflate_default_free_func = free_func;
+}
+
+/*
+ * Implementations of libc functions for freestanding library builds.
+ * Normal library builds don't use these.  Not optimized yet; usually the
+ * compiler expands these functions and doesn't actually call them anyway.
+ */
+#ifdef FREESTANDING
+#undef memset
+void * __attribute__((weak))
+memset(void *s, int c, size_t n)
+{
+	u8 *p = s;
+	size_t i;
+
+	for (i = 0; i < n; i++)
+		p[i] = c;
+	return s;
+}
+
+#undef memcpy
+void * __attribute__((weak))
+memcpy(void *dest, const void *src, size_t n)
+{
+	u8 *d = dest;
+	const u8 *s = src;
+	size_t i;
+
+	for (i = 0; i < n; i++)
+		d[i] = s[i];
+	return dest;
+}
+
+#undef memmove
+void * __attribute__((weak))
+memmove(void *dest, const void *src, size_t n)
+{
+	u8 *d = dest;
+	const u8 *s = src;
+	size_t i;
+
+	if (d <= s)
+		return memcpy(d, s, n);
+
+	for (i = n; i > 0; i--)
+		d[i - 1] = s[i - 1];
+	return dest;
+}
+
+#undef memcmp
+int __attribute__((weak))
+memcmp(const void *s1, const void *s2, size_t n)
+{
+	const u8 *p1 = s1;
+	const u8 *p2 = s2;
+	size_t i;
+
+	for (i = 0; i < n; i++) {
+		if (p1[i] != p2[i])
+			return (int)p1[i] - (int)p2[i];
+	}
+	return 0;
+}
+#endif /* FREESTANDING */
+
+#ifdef LIBDEFLATE_ENABLE_ASSERTIONS
+#include <stdio.h>
+#include <stdlib.h>
+NORETURN void
+libdeflate_assertion_failed(const char *expr, const char *file, int line)
+{
+	fprintf(stderr, "Assertion failed: %s at %s:%d\n", expr, file, line);
+	abort();
+}
+#endif /* LIBDEFLATE_ENABLE_ASSERTIONS */

data/ext/deflate_ruby/libdeflate/lib/x86/adler32_impl.h

@@ -0,0 +1,134 @@
+/*
+ * x86/adler32_impl.h - x86 implementations of Adler-32 checksum algorithm
+ *
+ * Copyright 2016 Eric Biggers
+ *
+ * Permission is hereby granted, free of charge, to any person
+ * obtaining a copy of this software and associated documentation
+ * files (the "Software"), to deal in the Software without
+ * restriction, including without limitation the rights to use,
+ * copy, modify, merge, publish, distribute, sublicense, and/or sell
+ * copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following
+ * conditions:
+ *
+ * The above copyright notice and this permission notice shall be
+ * included in all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+ * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES
+ * OF MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
+ * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT
+ * HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY,
+ * WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
+ * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ */
+
+#ifndef LIB_X86_ADLER32_IMPL_H
+#define LIB_X86_ADLER32_IMPL_H
+
+#include "cpu_features.h"
+
+/* SSE2 and AVX2 implementations.  Used on older CPUs. */
+#if defined(__GNUC__) || defined(__clang__) || defined(_MSC_VER)
+#  define adler32_x86_sse2	adler32_x86_sse2
+#  define SUFFIX			   _sse2
+#  define ATTRIBUTES		_target_attribute("sse2")
+#  define VL			16
+#  define USE_VNNI		0
+#  define USE_AVX512		0
+#  include "adler32_template.h"
+
+#  define adler32_x86_avx2	adler32_x86_avx2
+#  define SUFFIX			   _avx2
+#  define ATTRIBUTES		_target_attribute("avx2")
+#  define VL			32
+#  define USE_VNNI		0
+#  define USE_AVX512		0
+#  include "adler32_template.h"
+#endif
+
+/*
+ * AVX-VNNI implementation.  This is used on CPUs that have AVX2 and AVX-VNNI
+ * but don't have AVX-512, for example Intel Alder Lake.
+ *
+ * Unusually for a new CPU feature, gcc added support for the AVX-VNNI
+ * intrinsics (in gcc 11.1) slightly before binutils added support for
+ * assembling AVX-VNNI instructions (in binutils 2.36).  Distros can reasonably
+ * have gcc 11 with binutils 2.35.  Because of this issue, we check for gcc 12
+ * instead of gcc 11.  (libdeflate supports direct compilation without a
+ * configure step, so checking the binutils version is not always an option.)
+ */
+#if GCC_PREREQ(12, 1) || CLANG_PREREQ(12, 0, 13000000) || MSVC_PREREQ(1930)
+#  define adler32_x86_avx2_vnni	adler32_x86_avx2_vnni
+#  define SUFFIX			   _avx2_vnni
+#  define ATTRIBUTES		_target_attribute("avx2,avxvnni")
+#  define VL			32
+#  define USE_VNNI		1
+#  define USE_AVX512		0
+#  include "adler32_template.h"
+#endif
+
+#if GCC_PREREQ(8, 1) || CLANG_PREREQ(6, 0, 10000000) || MSVC_PREREQ(1920)
+/*
+ * AVX512VNNI implementation using 256-bit vectors.  This is very similar to the
+ * AVX-VNNI implementation but takes advantage of masking and more registers.
+ * This is used on CPUs that support AVX-512 but where using 512-bit vectors
+ * causes downclocking.  This should also be the optimal implementation on CPUs
+ * that support AVX10/256 but not AVX10/512.
+ */
+#  define adler32_x86_avx512_vl256_vnni	adler32_x86_avx512_vl256_vnni
+#  define SUFFIX				   _avx512_vl256_vnni
+#  define ATTRIBUTES		_target_attribute("avx512bw,avx512vl,avx512vnni")
+#  define VL			32
+#  define USE_VNNI		1
+#  define USE_AVX512		1
+#  include "adler32_template.h"
+
+/*
+ * AVX512VNNI implementation using 512-bit vectors.  This is used on CPUs that
+ * have a good AVX-512 implementation including AVX512VNNI.  This should also be
+ * the optimal implementation on CPUs that support AVX10/512.
+ */
+#  define adler32_x86_avx512_vl512_vnni	adler32_x86_avx512_vl512_vnni
+#  define SUFFIX				   _avx512_vl512_vnni
+#  define ATTRIBUTES		_target_attribute("avx512bw,avx512vnni")
+#  define VL			64
+#  define USE_VNNI		1
+#  define USE_AVX512		1
+#  include "adler32_template.h"
+#endif
+
+static inline adler32_func_t
+arch_select_adler32_func(void)
+{
+	const u32 features MAYBE_UNUSED = get_x86_cpu_features();
+
+#ifdef adler32_x86_avx512_vl512_vnni
+	if ((features & X86_CPU_FEATURE_ZMM) &&
+	    HAVE_AVX512BW(features) && HAVE_AVX512VNNI(features))
+		return adler32_x86_avx512_vl512_vnni;
+#endif
+#ifdef adler32_x86_avx512_vl256_vnni
+	if (HAVE_AVX512BW(features) && HAVE_AVX512VL(features) &&
+	    HAVE_AVX512VNNI(features))
+		return adler32_x86_avx512_vl256_vnni;
+#endif
+#ifdef adler32_x86_avx2_vnni
+	if (HAVE_AVX2(features) && HAVE_AVXVNNI(features))
+		return adler32_x86_avx2_vnni;
+#endif
+#ifdef adler32_x86_avx2
+	if (HAVE_AVX2(features))
+		return adler32_x86_avx2;
+#endif
+#ifdef adler32_x86_sse2
+	if (HAVE_SSE2(features))
+		return adler32_x86_sse2;
+#endif
+	return NULL;
+}
+#define arch_select_adler32_func	arch_select_adler32_func
+
+#endif /* LIB_X86_ADLER32_IMPL_H */