summaryrefslogtreecommitdiffstats
path: root/media/ffvpx/libavutil/x86
diff options
context:
space:
mode:
Diffstat (limited to 'media/ffvpx/libavutil/x86')
-rw-r--r--media/ffvpx/libavutil/x86/cpu.c14
-rw-r--r--media/ffvpx/libavutil/x86/cpu.h3
-rw-r--r--media/ffvpx/libavutil/x86/intmath.h3
-rw-r--r--media/ffvpx/libavutil/x86/x86inc.asm266
-rw-r--r--media/ffvpx/libavutil/x86/x86util.asm50
5 files changed, 82 insertions, 254 deletions
diff --git a/media/ffvpx/libavutil/x86/cpu.c b/media/ffvpx/libavutil/x86/cpu.c
index aca893174..f33088c8c 100644
--- a/media/ffvpx/libavutil/x86/cpu.c
+++ b/media/ffvpx/libavutil/x86/cpu.c
@@ -97,7 +97,6 @@ int ff_get_cpu_flags_x86(void)
int max_std_level, max_ext_level, std_caps = 0, ext_caps = 0;
int family = 0, model = 0;
union { int i[3]; char c[12]; } vendor;
- int xcr0_lo = 0, xcr0_hi = 0;
if (!cpuid_test())
return 0; /* CPUID not supported */
@@ -133,8 +132,8 @@ int ff_get_cpu_flags_x86(void)
/* Check OXSAVE and AVX bits */
if ((ecx & 0x18000000) == 0x18000000) {
/* Check for OS support */
- xgetbv(0, xcr0_lo, xcr0_hi);
- if ((xcr0_lo & 0x6) == 0x6) {
+ xgetbv(0, eax, edx);
+ if ((eax & 0x6) == 0x6) {
rval |= AV_CPU_FLAG_AVX;
if (ecx & 0x00001000)
rval |= AV_CPU_FLAG_FMA3;
@@ -148,13 +147,6 @@ int ff_get_cpu_flags_x86(void)
#if HAVE_AVX2
if ((rval & AV_CPU_FLAG_AVX) && (ebx & 0x00000020))
rval |= AV_CPU_FLAG_AVX2;
-#if HAVE_AVX512 /* F, CD, BW, DQ, VL */
- if ((xcr0_lo & 0xe0) == 0xe0) { /* OPMASK/ZMM state */
- if ((rval & AV_CPU_FLAG_AVX2) && (ebx & 0xd0030000) == 0xd0030000)
- rval |= AV_CPU_FLAG_AVX512;
-
- }
-#endif /* HAVE_AVX512 */
#endif /* HAVE_AVX2 */
/* BMI1/2 don't need OS support */
if (ebx & 0x00000008) {
@@ -246,8 +238,6 @@ size_t ff_get_cpu_max_align_x86(void)
{
int flags = av_get_cpu_flags();
- if (flags & AV_CPU_FLAG_AVX512)
- return 64;
if (flags & (AV_CPU_FLAG_AVX2 |
AV_CPU_FLAG_AVX |
AV_CPU_FLAG_XOP |
diff --git a/media/ffvpx/libavutil/x86/cpu.h b/media/ffvpx/libavutil/x86/cpu.h
index 937c697fa..309b8e746 100644
--- a/media/ffvpx/libavutil/x86/cpu.h
+++ b/media/ffvpx/libavutil/x86/cpu.h
@@ -19,6 +19,7 @@
#ifndef AVUTIL_X86_CPU_H
#define AVUTIL_X86_CPU_H
+#include "config.h"
#include "libavutil/cpu.h"
#include "libavutil/cpu_internal.h"
@@ -49,7 +50,6 @@
#define X86_FMA4(flags) CPUEXT(flags, FMA4)
#define X86_AVX2(flags) CPUEXT(flags, AVX2)
#define X86_AESNI(flags) CPUEXT(flags, AESNI)
-#define X86_AVX512(flags) CPUEXT(flags, AVX512)
#define EXTERNAL_AMD3DNOW(flags) CPUEXT_SUFFIX(flags, _EXTERNAL, AMD3DNOW)
#define EXTERNAL_AMD3DNOWEXT(flags) CPUEXT_SUFFIX(flags, _EXTERNAL, AMD3DNOWEXT)
@@ -79,7 +79,6 @@
#define EXTERNAL_AVX2_FAST(flags) CPUEXT_SUFFIX_FAST2(flags, _EXTERNAL, AVX2, AVX)
#define EXTERNAL_AVX2_SLOW(flags) CPUEXT_SUFFIX_SLOW2(flags, _EXTERNAL, AVX2, AVX)
#define EXTERNAL_AESNI(flags) CPUEXT_SUFFIX(flags, _EXTERNAL, AESNI)
-#define EXTERNAL_AVX512(flags) CPUEXT_SUFFIX(flags, _EXTERNAL, AVX512)
#define INLINE_AMD3DNOW(flags) CPUEXT_SUFFIX(flags, _INLINE, AMD3DNOW)
#define INLINE_AMD3DNOWEXT(flags) CPUEXT_SUFFIX(flags, _INLINE, AMD3DNOWEXT)
diff --git a/media/ffvpx/libavutil/x86/intmath.h b/media/ffvpx/libavutil/x86/intmath.h
index 40743fd13..e83971c08 100644
--- a/media/ffvpx/libavutil/x86/intmath.h
+++ b/media/ffvpx/libavutil/x86/intmath.h
@@ -47,8 +47,7 @@ static av_always_inline av_const int ff_log2_x86(unsigned int v)
# endif
# define ff_log2_16bit av_log2
-#if defined(__INTEL_COMPILER) || (defined(_MSC_VER) && (_MSC_VER >= 1700) && \
- (defined(__BMI__) || !defined(__clang__)))
+#if defined(__INTEL_COMPILER) || (defined(_MSC_VER) && (_MSC_VER >= 1700))
# define ff_ctz(v) _tzcnt_u32(v)
# if ARCH_X86_64
diff --git a/media/ffvpx/libavutil/x86/x86inc.asm b/media/ffvpx/libavutil/x86/x86inc.asm
index 5044ee86f..6a054a3e0 100644
--- a/media/ffvpx/libavutil/x86/x86inc.asm
+++ b/media/ffvpx/libavutil/x86/x86inc.asm
@@ -1,12 +1,12 @@
;*****************************************************************************
;* x86inc.asm: x264asm abstraction layer
;*****************************************************************************
-;* Copyright (C) 2005-2018 x264 project
+;* Copyright (C) 2005-2017 x264 project
;*
;* Authors: Loren Merritt <lorenm@u.washington.edu>
-;* Henrik Gramner <henrik@gramner.com>
;* Anton Mitrofanov <BugMaster@narod.ru>
;* Fiona Glaser <fiona@x264.com>
+;* Henrik Gramner <henrik@gramner.com>
;*
;* Permission to use, copy, modify, and/or distribute this software for any
;* purpose with or without fee is hereby granted, provided that the above
@@ -90,10 +90,6 @@
SECTION .text
%elifidn __OUTPUT_FORMAT__,coff
SECTION .text
- %elifidn __OUTPUT_FORMAT__,win32
- SECTION .rdata align=%1
- %elif WIN64
- SECTION .rdata align=%1
%else
SECTION .rodata align=%1
%endif
@@ -341,8 +337,6 @@ DECLARE_REG_TMP_SIZE 0,1,2,3,4,5,6,7,8,9,10,11,12,13,14
%endmacro
%define required_stack_alignment ((mmsize + 15) & ~15)
-%define vzeroupper_required (mmsize > 16 && (ARCH_X86_64 == 0 || xmm_regs_used > 16 || notcpuflag(avx512)))
-%define high_mm_regs (16*cpuflag(avx512))
%macro ALLOC_STACK 1-2 0 ; stack_size, n_xmm_regs (for win64 only)
%ifnum %1
@@ -456,16 +450,15 @@ DECLARE_REG 14, R13, 120
%macro WIN64_PUSH_XMM 0
; Use the shadow space to store XMM6 and XMM7, the rest needs stack space allocated.
- %if xmm_regs_used > 6 + high_mm_regs
+ %if xmm_regs_used > 6
movaps [rstk + stack_offset + 8], xmm6
%endif
- %if xmm_regs_used > 7 + high_mm_regs
+ %if xmm_regs_used > 7
movaps [rstk + stack_offset + 24], xmm7
%endif
- %assign %%xmm_regs_on_stack xmm_regs_used - high_mm_regs - 8
- %if %%xmm_regs_on_stack > 0
+ %if xmm_regs_used > 8
%assign %%i 8
- %rep %%xmm_regs_on_stack
+ %rep xmm_regs_used-8
movaps [rsp + (%%i-8)*16 + stack_size + 32], xmm %+ %%i
%assign %%i %%i+1
%endrep
@@ -474,11 +467,10 @@ DECLARE_REG 14, R13, 120
%macro WIN64_SPILL_XMM 1
%assign xmm_regs_used %1
- ASSERT xmm_regs_used <= 16 + high_mm_regs
- %assign %%xmm_regs_on_stack xmm_regs_used - high_mm_regs - 8
- %if %%xmm_regs_on_stack > 0
+ ASSERT xmm_regs_used <= 16
+ %if xmm_regs_used > 8
; Allocate stack space for callee-saved xmm registers plus shadow space and align the stack.
- %assign %%pad %%xmm_regs_on_stack*16 + 32
+ %assign %%pad (xmm_regs_used-8)*16 + 32
%assign stack_size_padded %%pad + ((-%%pad-stack_offset-gprsize) & (STACK_ALIGNMENT-1))
SUB rsp, stack_size_padded
%endif
@@ -487,10 +479,9 @@ DECLARE_REG 14, R13, 120
%macro WIN64_RESTORE_XMM_INTERNAL 0
%assign %%pad_size 0
- %assign %%xmm_regs_on_stack xmm_regs_used - high_mm_regs - 8
- %if %%xmm_regs_on_stack > 0
- %assign %%i xmm_regs_used - high_mm_regs
- %rep %%xmm_regs_on_stack
+ %if xmm_regs_used > 8
+ %assign %%i xmm_regs_used
+ %rep xmm_regs_used-8
%assign %%i %%i-1
movaps xmm %+ %%i, [rsp + (%%i-8)*16 + stack_size + 32]
%endrep
@@ -503,10 +494,10 @@ DECLARE_REG 14, R13, 120
%assign %%pad_size stack_size_padded
%endif
%endif
- %if xmm_regs_used > 7 + high_mm_regs
+ %if xmm_regs_used > 7
movaps xmm7, [rsp + stack_offset - %%pad_size + 24]
%endif
- %if xmm_regs_used > 6 + high_mm_regs
+ %if xmm_regs_used > 6
movaps xmm6, [rsp + stack_offset - %%pad_size + 8]
%endif
%endmacro
@@ -518,12 +509,12 @@ DECLARE_REG 14, R13, 120
%assign xmm_regs_used 0
%endmacro
-%define has_epilogue regs_used > 7 || stack_size > 0 || vzeroupper_required || xmm_regs_used > 6+high_mm_regs
+%define has_epilogue regs_used > 7 || xmm_regs_used > 6 || mmsize == 32 || stack_size > 0
%macro RET 0
WIN64_RESTORE_XMM_INTERNAL
POP_IF_USED 14, 13, 12, 11, 10, 9, 8, 7
- %if vzeroupper_required
+ %if mmsize == 32
vzeroupper
%endif
AUTO_REP_RET
@@ -547,10 +538,9 @@ DECLARE_REG 12, R15, 56
DECLARE_REG 13, R12, 64
DECLARE_REG 14, R13, 72
-%macro PROLOGUE 2-5+ 0 ; #args, #regs, #xmm_regs, [stack_size,] arg_names...
+%macro PROLOGUE 2-5+ ; #args, #regs, #xmm_regs, [stack_size,] arg_names...
%assign num_args %1
%assign regs_used %2
- %assign xmm_regs_used %3
ASSERT regs_used >= num_args
SETUP_STACK_POINTER %4
ASSERT regs_used <= 15
@@ -560,7 +550,7 @@ DECLARE_REG 14, R13, 72
DEFINE_ARGS_INTERNAL %0, %4, %5
%endmacro
-%define has_epilogue regs_used > 9 || stack_size > 0 || vzeroupper_required
+%define has_epilogue regs_used > 9 || mmsize == 32 || stack_size > 0
%macro RET 0
%if stack_size_padded > 0
@@ -571,7 +561,7 @@ DECLARE_REG 14, R13, 72
%endif
%endif
POP_IF_USED 14, 13, 12, 11, 10, 9
- %if vzeroupper_required
+ %if mmsize == 32
vzeroupper
%endif
AUTO_REP_RET
@@ -616,7 +606,7 @@ DECLARE_ARG 7, 8, 9, 10, 11, 12, 13, 14
DEFINE_ARGS_INTERNAL %0, %4, %5
%endmacro
-%define has_epilogue regs_used > 3 || stack_size > 0 || vzeroupper_required
+%define has_epilogue regs_used > 3 || mmsize == 32 || stack_size > 0
%macro RET 0
%if stack_size_padded > 0
@@ -627,7 +617,7 @@ DECLARE_ARG 7, 8, 9, 10, 11, 12, 13, 14
%endif
%endif
POP_IF_USED 6, 5, 4, 3
- %if vzeroupper_required
+ %if mmsize == 32
vzeroupper
%endif
AUTO_REP_RET
@@ -737,22 +727,12 @@ BRANCH_INSTR jz, je, jnz, jne, jl, jle, jnl, jnle, jg, jge, jng, jnge, ja, jae,
%assign stack_offset 0 ; stack pointer offset relative to the return address
%assign stack_size 0 ; amount of stack space that can be freely used inside a function
%assign stack_size_padded 0 ; total amount of allocated stack space, including space for callee-saved xmm registers on WIN64 and alignment padding
- %assign xmm_regs_used 0 ; number of XMM registers requested, used for dealing with callee-saved registers on WIN64 and vzeroupper
+ %assign xmm_regs_used 0 ; number of XMM registers requested, used for dealing with callee-saved registers on WIN64
%ifnidn %3, ""
PROLOGUE %3
%endif
%endmacro
-; Create a global symbol from a local label with the correct name mangling and type
-%macro cglobal_label 1
- %if FORMAT_ELF
- global current_function %+ %1:function hidden
- %else
- global current_function %+ %1
- %endif
- %1:
-%endmacro
-
%macro cextern 1
%xdefine %1 mangle(private_prefix %+ _ %+ %1)
CAT_XDEFINE cglobaled_, %1, 1
@@ -823,10 +803,10 @@ BRANCH_INSTR jz, je, jnz, jne, jl, jle, jnl, jnle, jg, jge, jng, jnge, ja, jae,
%assign cpuflags_bmi1 (1<<17)| cpuflags_avx|cpuflags_lzcnt
%assign cpuflags_bmi2 (1<<18)| cpuflags_bmi1
%assign cpuflags_avx2 (1<<19)| cpuflags_fma3|cpuflags_bmi2
-%assign cpuflags_avx512 (1<<20)| cpuflags_avx2 ; F, CD, BW, DQ, VL
-%assign cpuflags_cache32 (1<<21)
-%assign cpuflags_cache64 (1<<22)
+%assign cpuflags_cache32 (1<<20)
+%assign cpuflags_cache64 (1<<21)
+%assign cpuflags_slowctz (1<<22)
%assign cpuflags_aligned (1<<23) ; not a cpu feature, but a function variant
%assign cpuflags_atom (1<<24)
@@ -876,12 +856,11 @@ BRANCH_INSTR jz, je, jnz, jne, jl, jle, jnl, jnle, jg, jge, jng, jnge, ja, jae,
%endif
%endmacro
-; Merge mmx, sse*, and avx*
+; Merge mmx and sse*
; m# is a simd register of the currently selected size
; xm# is the corresponding xmm register if mmsize >= 16, otherwise the same as m#
; ym# is the corresponding ymm register if mmsize >= 32, otherwise the same as m#
-; zm# is the corresponding zmm register if mmsize >= 64, otherwise the same as m#
-; (All 4 remain in sync through SWAP.)
+; (All 3 remain in sync through SWAP.)
%macro CAT_XDEFINE 3
%xdefine %1%2 %3
@@ -891,99 +870,69 @@ BRANCH_INSTR jz, je, jnz, jne, jl, jle, jnl, jnle, jg, jge, jng, jnge, ja, jae,
%undef %1%2
%endmacro
-%macro DEFINE_MMREGS 1 ; mmtype
- %assign %%prev_mmregs 0
- %ifdef num_mmregs
- %assign %%prev_mmregs num_mmregs
- %endif
-
- %assign num_mmregs 8
- %if ARCH_X86_64 && mmsize >= 16
- %assign num_mmregs 16
- %if cpuflag(avx512) || mmsize == 64
- %assign num_mmregs 32
- %endif
- %endif
-
- %assign %%i 0
- %rep num_mmregs
- CAT_XDEFINE m, %%i, %1 %+ %%i
- CAT_XDEFINE nn%1, %%i, %%i
- %assign %%i %%i+1
- %endrep
- %if %%prev_mmregs > num_mmregs
- %rep %%prev_mmregs - num_mmregs
- CAT_UNDEF m, %%i
- CAT_UNDEF nn %+ mmtype, %%i
- %assign %%i %%i+1
- %endrep
- %endif
- %xdefine mmtype %1
-%endmacro
-
-; Prefer registers 16-31 over 0-15 to avoid having to use vzeroupper
-%macro AVX512_MM_PERMUTATION 0-1 0 ; start_reg
- %if ARCH_X86_64 && cpuflag(avx512)
- %assign %%i %1
- %rep 16-%1
- %assign %%i_high %%i+16
- SWAP %%i, %%i_high
- %assign %%i %%i+1
- %endrep
- %endif
-%endmacro
-
%macro INIT_MMX 0-1+
%assign avx_enabled 0
%define RESET_MM_PERMUTATION INIT_MMX %1
%define mmsize 8
+ %define num_mmregs 8
%define mova movq
%define movu movq
%define movh movd
%define movnta movntq
+ %assign %%i 0
+ %rep 8
+ CAT_XDEFINE m, %%i, mm %+ %%i
+ CAT_XDEFINE nnmm, %%i, %%i
+ %assign %%i %%i+1
+ %endrep
+ %rep 8
+ CAT_UNDEF m, %%i
+ CAT_UNDEF nnmm, %%i
+ %assign %%i %%i+1
+ %endrep
INIT_CPUFLAGS %1
- DEFINE_MMREGS mm
%endmacro
%macro INIT_XMM 0-1+
%assign avx_enabled 0
%define RESET_MM_PERMUTATION INIT_XMM %1
%define mmsize 16
+ %define num_mmregs 8
+ %if ARCH_X86_64
+ %define num_mmregs 16
+ %endif
%define mova movdqa
%define movu movdqu
%define movh movq
%define movnta movntdq
+ %assign %%i 0
+ %rep num_mmregs
+ CAT_XDEFINE m, %%i, xmm %+ %%i
+ CAT_XDEFINE nnxmm, %%i, %%i
+ %assign %%i %%i+1
+ %endrep
INIT_CPUFLAGS %1
- DEFINE_MMREGS xmm
- %if WIN64
- AVX512_MM_PERMUTATION 6 ; Swap callee-saved registers with volatile registers
- %endif
%endmacro
%macro INIT_YMM 0-1+
%assign avx_enabled 1
%define RESET_MM_PERMUTATION INIT_YMM %1
%define mmsize 32
+ %define num_mmregs 8
+ %if ARCH_X86_64
+ %define num_mmregs 16
+ %endif
%define mova movdqa
%define movu movdqu
%undef movh
%define movnta movntdq
+ %assign %%i 0
+ %rep num_mmregs
+ CAT_XDEFINE m, %%i, ymm %+ %%i
+ CAT_XDEFINE nnymm, %%i, %%i
+ %assign %%i %%i+1
+ %endrep
INIT_CPUFLAGS %1
- DEFINE_MMREGS ymm
- AVX512_MM_PERMUTATION
-%endmacro
-
-%macro INIT_ZMM 0-1+
- %assign avx_enabled 1
- %define RESET_MM_PERMUTATION INIT_ZMM %1
- %define mmsize 64
- %define mova movdqa
- %define movu movdqu
- %undef movh
- %define movnta movntdq
- INIT_CPUFLAGS %1
- DEFINE_MMREGS zmm
- AVX512_MM_PERMUTATION
%endmacro
INIT_XMM
@@ -992,26 +941,18 @@ INIT_XMM
%define mmmm%1 mm%1
%define mmxmm%1 mm%1
%define mmymm%1 mm%1
- %define mmzmm%1 mm%1
%define xmmmm%1 mm%1
%define xmmxmm%1 xmm%1
%define xmmymm%1 xmm%1
- %define xmmzmm%1 xmm%1
%define ymmmm%1 mm%1
%define ymmxmm%1 xmm%1
%define ymmymm%1 ymm%1
- %define ymmzmm%1 ymm%1
- %define zmmmm%1 mm%1
- %define zmmxmm%1 xmm%1
- %define zmmymm%1 ymm%1
- %define zmmzmm%1 zmm%1
%define xm%1 xmm %+ m%1
%define ym%1 ymm %+ m%1
- %define zm%1 zmm %+ m%1
%endmacro
%assign i 0
-%rep 32
+%rep 16
DECLARE_MMCAST i
%assign i i+1
%endrep
@@ -1146,17 +1087,12 @@ INIT_XMM
;=============================================================================
%assign i 0
-%rep 32
+%rep 16
%if i < 8
CAT_XDEFINE sizeofmm, i, 8
- CAT_XDEFINE regnumofmm, i, i
%endif
CAT_XDEFINE sizeofxmm, i, 16
CAT_XDEFINE sizeofymm, i, 32
- CAT_XDEFINE sizeofzmm, i, 64
- CAT_XDEFINE regnumofxmm, i, i
- CAT_XDEFINE regnumofymm, i, i
- CAT_XDEFINE regnumofzmm, i, i
%assign i i+1
%endrep
%undef i
@@ -1273,7 +1209,7 @@ INIT_XMM
%endmacro
%endmacro
-; Instructions with both VEX/EVEX and legacy encodings
+; Instructions with both VEX and non-VEX encodings
; Non-destructive instructions are written without parameters
AVX_INSTR addpd, sse2, 1, 0, 1
AVX_INSTR addps, sse, 1, 0, 1
@@ -1295,42 +1231,10 @@ AVX_INSTR blendpd, sse4, 1, 1, 0
AVX_INSTR blendps, sse4, 1, 1, 0
AVX_INSTR blendvpd, sse4 ; can't be emulated
AVX_INSTR blendvps, sse4 ; can't be emulated
-AVX_INSTR cmpeqpd, sse2, 1, 0, 1
-AVX_INSTR cmpeqps, sse, 1, 0, 1
-AVX_INSTR cmpeqsd, sse2, 1, 0, 0
-AVX_INSTR cmpeqss, sse, 1, 0, 0
-AVX_INSTR cmplepd, sse2, 1, 0, 0
-AVX_INSTR cmpleps, sse, 1, 0, 0
-AVX_INSTR cmplesd, sse2, 1, 0, 0
-AVX_INSTR cmpless, sse, 1, 0, 0
-AVX_INSTR cmpltpd, sse2, 1, 0, 0
-AVX_INSTR cmpltps, sse, 1, 0, 0
-AVX_INSTR cmpltsd, sse2, 1, 0, 0
-AVX_INSTR cmpltss, sse, 1, 0, 0
-AVX_INSTR cmpneqpd, sse2, 1, 0, 1
-AVX_INSTR cmpneqps, sse, 1, 0, 1
-AVX_INSTR cmpneqsd, sse2, 1, 0, 0
-AVX_INSTR cmpneqss, sse, 1, 0, 0
-AVX_INSTR cmpnlepd, sse2, 1, 0, 0
-AVX_INSTR cmpnleps, sse, 1, 0, 0
-AVX_INSTR cmpnlesd, sse2, 1, 0, 0
-AVX_INSTR cmpnless, sse, 1, 0, 0
-AVX_INSTR cmpnltpd, sse2, 1, 0, 0
-AVX_INSTR cmpnltps, sse, 1, 0, 0
-AVX_INSTR cmpnltsd, sse2, 1, 0, 0
-AVX_INSTR cmpnltss, sse, 1, 0, 0
-AVX_INSTR cmpordpd, sse2 1, 0, 1
-AVX_INSTR cmpordps, sse 1, 0, 1
-AVX_INSTR cmpordsd, sse2 1, 0, 0
-AVX_INSTR cmpordss, sse 1, 0, 0
AVX_INSTR cmppd, sse2, 1, 1, 0
AVX_INSTR cmpps, sse, 1, 1, 0
AVX_INSTR cmpsd, sse2, 1, 1, 0
AVX_INSTR cmpss, sse, 1, 1, 0
-AVX_INSTR cmpunordpd, sse2, 1, 0, 1
-AVX_INSTR cmpunordps, sse, 1, 0, 1
-AVX_INSTR cmpunordsd, sse2, 1, 0, 0
-AVX_INSTR cmpunordss, sse, 1, 0, 0
AVX_INSTR comisd, sse2
AVX_INSTR comiss, sse
AVX_INSTR cvtdq2pd, sse2
@@ -1641,52 +1545,6 @@ FMA4_INSTR fmsubadd, pd, ps
FMA4_INSTR fnmadd, pd, ps, sd, ss
FMA4_INSTR fnmsub, pd, ps, sd, ss
-; Macros for converting VEX instructions to equivalent EVEX ones.
-%macro EVEX_INSTR 2-3 0 ; vex, evex, prefer_evex
- %macro %1 2-7 fnord, fnord, %1, %2, %3
- %ifidn %3, fnord
- %define %%args %1, %2
- %elifidn %4, fnord
- %define %%args %1, %2, %3
- %else
- %define %%args %1, %2, %3, %4
- %endif
- %assign %%evex_required cpuflag(avx512) & %7
- %ifnum regnumof%1
- %if regnumof%1 >= 16 || sizeof%1 > 32
- %assign %%evex_required 1
- %endif
- %endif
- %ifnum regnumof%2
- %if regnumof%2 >= 16 || sizeof%2 > 32
- %assign %%evex_required 1
- %endif
- %endif
- %if %%evex_required
- %6 %%args
- %else
- %5 %%args ; Prefer VEX over EVEX due to shorter instruction length
- %endif
- %endmacro
-%endmacro
-
-EVEX_INSTR vbroadcastf128, vbroadcastf32x4
-EVEX_INSTR vbroadcasti128, vbroadcasti32x4
-EVEX_INSTR vextractf128, vextractf32x4
-EVEX_INSTR vextracti128, vextracti32x4
-EVEX_INSTR vinsertf128, vinsertf32x4
-EVEX_INSTR vinserti128, vinserti32x4
-EVEX_INSTR vmovdqa, vmovdqa32
-EVEX_INSTR vmovdqu, vmovdqu32
-EVEX_INSTR vpand, vpandd
-EVEX_INSTR vpandn, vpandnd
-EVEX_INSTR vpor, vpord
-EVEX_INSTR vpxor, vpxord
-EVEX_INSTR vrcpps, vrcp14ps, 1 ; EVEX versions have higher precision
-EVEX_INSTR vrcpss, vrcp14ss, 1
-EVEX_INSTR vrsqrtps, vrsqrt14ps, 1
-EVEX_INSTR vrsqrtss, vrsqrt14ss, 1
-
; workaround: vpbroadcastq is broken in x86_32 due to a yasm bug (fixed in 1.3.0)
%ifdef __YASM_VER__
%if __YASM_VERSION_ID__ < 0x01030000 && ARCH_X86_64 == 0
diff --git a/media/ffvpx/libavutil/x86/x86util.asm b/media/ffvpx/libavutil/x86/x86util.asm
index d7cd99684..e1220dfc1 100644
--- a/media/ffvpx/libavutil/x86/x86util.asm
+++ b/media/ffvpx/libavutil/x86/x86util.asm
@@ -357,7 +357,7 @@
%endif
%endmacro
-%macro ABSB 2 ; source mmreg, temp mmreg (unused for SSSE3)
+%macro ABSB 2 ; source mmreg, temp mmreg (unused for ssse3)
%if cpuflag(ssse3)
pabsb %1, %1
%else
@@ -381,7 +381,7 @@
%endif
%endmacro
-%macro ABSD2 4
+%macro ABSD2_MMX 4
pxor %3, %3
pxor %4, %4
pcmpgtd %3, %1
@@ -475,7 +475,7 @@
%else
palignr %1, %2, %3
%endif
-%else ; [dst,] src1, src2, imm, tmp
+%elif cpuflag(mmx) ; [dst,] src1, src2, imm, tmp
%define %%dst %1
%if %0==5
%ifnidn %1, %2
@@ -799,47 +799,37 @@
pminsw %1, %3
%endmacro
-%macro PMINSD 3 ; dst, src, tmp/unused
-%if cpuflag(sse4)
- pminsd %1, %2
-%elif cpuflag(sse2)
- cvtdq2ps %1, %1
- minps %1, %2
- cvtps2dq %1, %1
-%else
+%macro PMINSD_MMX 3 ; dst, src, tmp
mova %3, %2
pcmpgtd %3, %1
pxor %1, %2
pand %1, %3
pxor %1, %2
-%endif
%endmacro
-%macro PMAXSD 3 ; dst, src, tmp/unused
-%if cpuflag(sse4)
- pmaxsd %1, %2
-%else
+%macro PMAXSD_MMX 3 ; dst, src, tmp
mova %3, %1
pcmpgtd %3, %2
pand %1, %3
pandn %3, %2
por %1, %3
-%endif
%endmacro
-%macro CLIPD 3-4
-%if cpuflag(sse4); src/dst, min, max, unused
- pminsd %1, %3
- pmaxsd %1, %2
-%elif cpuflag(sse2) ; src/dst, min (float), max (float), unused
+%macro CLIPD_MMX 3-4 ; src/dst, min, max, tmp
+ PMINSD_MMX %1, %3, %4
+ PMAXSD_MMX %1, %2, %4
+%endmacro
+
+%macro CLIPD_SSE2 3-4 ; src/dst, min (float), max (float), unused
cvtdq2ps %1, %1
minps %1, %3
maxps %1, %2
cvtps2dq %1, %1
-%else ; src/dst, min, max, tmp
- PMINSD %1, %3, %4
- PMAXSD %1, %2, %4
-%endif
+%endmacro
+
+%macro CLIPD_SSE41 3-4 ; src/dst, min, max, unused
+ pminsd %1, %3
+ pmaxsd %1, %2
%endmacro
%macro VBROADCASTSS 2 ; dst xmm/ymm, src m32/xmm
@@ -890,14 +880,6 @@
%endif
%endmacro
-%macro VBROADCASTI128 2 ; dst xmm/ymm, src : 128bits val
-%if mmsize > 16
- vbroadcasti128 %1, %2
-%else
- mova %1, %2
-%endif
-%endmacro
-
%macro SHUFFLE_MASK_W 8
%rep 8
%if %1>=0x80