chromium/third_party/libvpx/source/libvpx/vpx_ports/x86.h

/*
 *  Copyright (c) 2010 The WebM project authors. All Rights Reserved.
 *
 *  Use of this source code is governed by a BSD-style license
 *  that can be found in the LICENSE file in the root of the source
 *  tree. An additional intellectual property rights grant can be found
 *  in the file PATENTS.  All contributing project authors may
 *  be found in the AUTHORS file in the root of the source tree.
 */

#ifndef VPX_VPX_PORTS_X86_H_
#define VPX_VPX_PORTS_X86_H_
#include <stdlib.h>

#if defined(_MSC_VER)
#include <intrin.h> /* For __cpuidex, __rdtsc */
#endif

#include "vpx_config.h"
#include "vpx/vpx_integer.h"

#ifdef __cplusplus
extern "C" {
#endif

vpx_cpu_t;

#if defined(__GNUC__) || defined(__ANDROID__)
#if VPX_ARCH_X86_64
#define cpuid(func, func2, ax, bx, cx, dx) …
#else
#define cpuid …
#endif
#elif defined(__SUNPRO_C) || \
    defined(__SUNPRO_CC) /* end __GNUC__ or __ANDROID__*/
#if VPX_ARCH_X86_64
#define cpuid …
#else
#define cpuid …
#endif
#else /* end __SUNPRO__ */
#if VPX_ARCH_X86_64
#if defined(_MSC_VER) && _MSC_VER > 1500
#define cpuid …
#else
#define cpuid …
#endif
#else
#define cpuid …
#endif
#endif /* end others */

// NaCl has no support for xgetbv or the raw opcode.
#if !defined(__native_client__) && (defined(__i386__) || defined(__x86_64__))
static INLINE uint64_t xgetbv(void) { … }
#elif (defined(_M_X64) || defined(_M_IX86)) && defined(_MSC_FULL_VER) && \
    _MSC_FULL_VER >= 160040219  // >= VS2010 SP1
#include <immintrin.h>
#define xgetbv …
#elif defined(_MSC_VER) && defined(_M_IX86)
static INLINE uint64_t xgetbv(void) {
  uint32_t eax_, edx_;
  __asm {
    xor ecx, ecx  // ecx = 0
    // Use the raw opcode for xgetbv for compatibility with older toolchains.
    __asm _emit 0x0f __asm _emit 0x01 __asm _emit 0xd0
    mov eax_, eax
    mov edx_, edx
  }
  return ((uint64_t)edx_ << 32) | eax_;
}
#else
#define xgetbv …
#endif

#if defined(_MSC_VER) && _MSC_VER >= 1700
#undef NOMINMAX
#define NOMINMAX
#ifndef WIN32_LEAN_AND_MEAN
#define WIN32_LEAN_AND_MEAN
#endif
#include <windows.h>
#if WINAPI_FAMILY_PARTITION(WINAPI_FAMILY_APP)
#define getenv …
#endif
#endif

#define HAS_MMX …
#define HAS_SSE …
#define HAS_SSE2 …
#define HAS_SSE3 …
#define HAS_SSSE3 …
#define HAS_SSE4_1 …
#define HAS_AVX …
#define HAS_AVX2 …
#define HAS_AVX512 …
#ifndef BIT
#define BIT(n) …
#endif

static INLINE int x86_simd_caps(void) { … }

// Fine-Grain Measurement Functions
//
// If you are timing a small region of code, access the timestamp counter
// (TSC) via:
//
// unsigned int start = x86_tsc_start();
//   ...
// unsigned int end = x86_tsc_end();
// unsigned int diff = end - start;
//
// The start/end functions introduce a few more instructions than using
// x86_readtsc directly, but prevent the CPU's out-of-order execution from
// affecting the measurement (by having earlier/later instructions be evaluated
// in the time interval). See the white paper, "How to Benchmark Code
// Execution Times on Intel(R) IA-32 and IA-64 Instruction Set Architectures" by
// Gabriele Paoloni for more information.
//
// If you are timing a large function (CPU time > a couple of seconds), use
// x86_readtsc64 to read the timestamp counter in a 64-bit integer. The
// out-of-order leakage that can occur is minimal compared to total runtime.
static INLINE unsigned int x86_readtsc(void) { … }
// 64-bit CPU cycle counter
static INLINE uint64_t x86_readtsc64(void) { … }

// 32-bit CPU cycle counter with a partial fence against out-of-order execution.
static INLINE unsigned int x86_readtscp(void) { … }

static INLINE unsigned int x86_tsc_start(void) { … }

static INLINE unsigned int x86_tsc_end(void) { … }

#if defined(__GNUC__)
#define x86_pause_hint() …
#elif defined(__SUNPRO_C) || defined(__SUNPRO_CC)
#define x86_pause_hint …
#else
#if VPX_ARCH_X86_64
#define x86_pause_hint …
#else
#define x86_pause_hint …
#endif
#endif

#if defined(__GNUC__)
static void x87_set_control_word(unsigned short mode) { … }
static unsigned short x87_get_control_word(void) { … }
#elif defined(__SUNPRO_C) || defined(__SUNPRO_CC)
static void x87_set_control_word(unsigned short mode) {
  asm volatile("fldcw %0" : : "m"(*&mode));
}
static unsigned short x87_get_control_word(void) {
  unsigned short mode;
  asm volatile("fstcw %0\n\t" : "=m"(*&mode) :);
  return mode;
}
#elif VPX_ARCH_X86_64
/* No fldcw intrinsics on Windows x64, punt to external asm */
extern void vpx_winx64_fldcw(unsigned short mode);
extern unsigned short vpx_winx64_fstcw(void);
#define x87_set_control_word …
#define x87_get_control_word …
#else
static void x87_set_control_word(unsigned short mode) {
  __asm { fldcw mode }
}
static unsigned short x87_get_control_word(void) {
  unsigned short mode;
  __asm { fstcw mode }
  return mode;
}
#endif

static INLINE unsigned int x87_set_double_precision(void) { … }

#ifdef __cplusplus
}  // extern "C"
#endif

#endif  // VPX_VPX_PORTS_X86_H_