;*****************************************************************************
;* x86inc.asm: x264asm abstraction layer
;*****************************************************************************
-;* Copyright (C) 2005-2013 x264 project
+;* Copyright (C) 2005-2015 x264 project
;*
;* Authors: Loren Merritt <lorenm@u.washington.edu>
;* Anton Mitrofanov <BugMaster@narod.ru>
%define public_prefix private_prefix
%endif
+%ifndef STACK_ALIGNMENT
+ %if ARCH_X86_64
+ %define STACK_ALIGNMENT 16
+ %else
+ %define STACK_ALIGNMENT 4
+ %endif
+%endif
+
%define WIN64 0
%define UNIX64 0
%if ARCH_X86_64
%define WIN64 1
%elifidn __OUTPUT_FORMAT__,win64
%define WIN64 1
+ %elifidn __OUTPUT_FORMAT__,x64
+ %define WIN64 1
%else
%define UNIX64 1
%endif
%endif
+%define FORMAT_ELF 0
+%ifidn __OUTPUT_FORMAT__,elf
+ %define FORMAT_ELF 1
+%elifidn __OUTPUT_FORMAT__,elf32
+ %define FORMAT_ELF 1
+%elifidn __OUTPUT_FORMAT__,elf64
+ %define FORMAT_ELF 1
+%endif
+
%ifdef PREFIX
%define mangle(x) _ %+ x
%else
SECTION .rodata align=%1
%endmacro
-%macro SECTION_TEXT 0-1 16
- SECTION .text align=%1
-%endmacro
-
%if WIN64
%define PIC
%elif ARCH_X86_64 == 0
default rel
%endif
-; Always use long nops (reduces 0x90 spam in disassembly on x86_32)
-CPU amdnop
+%ifdef __NASM_VER__
+ %use smartalign
+%endif
; Macros to eliminate most code duplication between x86_32 and x86_64:
; Currently this works only for leaf functions which load all their arguments
; %1 = number of arguments. loads them from stack if needed.
; %2 = number of registers used. pushes callee-saved regs if needed.
; %3 = number of xmm registers used. pushes callee-saved xmm regs if needed.
-; %4 = (optional) stack size to be allocated. If not aligned (x86-32 ICC 10.x,
-; MSVC or YMM), the stack will be manually aligned (to 16 or 32 bytes),
+; %4 = (optional) stack size to be allocated. The stack will be aligned before
+; allocating the specified stack size. If the required stack alignment is
+; larger than the known stack alignment the stack will be manually aligned
; and an extra register will be allocated to hold the original stack
; pointer (to not invalidate r0m etc.). To prevent the use of an extra
; register as stack pointer, request a negative stack size.
; PROLOGUE can also be invoked by adding the same options to cglobal
; e.g.
-; cglobal foo, 2,3,0, dst, src, tmp
-; declares a function (foo), taking two args (dst and src) and one local variable (tmp)
+; cglobal foo, 2,3,7,0x40, dst, src, tmp
+; declares a function (foo) that automatically loads two arguments (dst and
+; src) into registers, uses one additional register (tmp) plus 7 vector
+; registers (m0-m6) and allocates 0x40 bytes of stack space.
; TODO Some functions can use some args directly from the stack. If they're the
; last args then you can just not declare them, but if they're in the middle
%define r%1w %2w
%define r%1b %2b
%define r%1h %2h
+ %define %2q %2
%if %0 == 2
%define r%1m %2d
%define r%1mp %2
%assign n_arg_names %0
%endmacro
+%define required_stack_alignment ((mmsize + 15) & ~15)
+
%macro ALLOC_STACK 1-2 0 ; stack_size, n_xmm_regs (for win64 only)
%ifnum %1
%if %1 != 0
- %assign %%stack_alignment ((mmsize + 15) & ~15)
+ %assign %%pad 0
%assign stack_size %1
%if stack_size < 0
%assign stack_size -stack_size
%endif
- %assign stack_size_padded stack_size
%if WIN64
- %assign stack_size_padded stack_size_padded + 32 ; reserve 32 bytes for shadow space
+ %assign %%pad %%pad + 32 ; shadow space
%if mmsize != 8
%assign xmm_regs_used %2
%if xmm_regs_used > 8
- %assign stack_size_padded stack_size_padded + (xmm_regs_used-8)*16
+ %assign %%pad %%pad + (xmm_regs_used-8)*16 ; callee-saved xmm registers
%endif
%endif
%endif
- %if mmsize <= 16 && HAVE_ALIGNED_STACK
- %assign stack_size_padded stack_size_padded + %%stack_alignment - gprsize - (stack_offset & (%%stack_alignment - 1))
+ %if required_stack_alignment <= STACK_ALIGNMENT
+ ; maintain the current stack alignment
+ %assign stack_size_padded stack_size + %%pad + ((-%%pad-stack_offset-gprsize) & (STACK_ALIGNMENT-1))
SUB rsp, stack_size_padded
%else
%assign %%reg_num (regs_used - 1)
; it, i.e. in [rsp+stack_size_padded], so we can restore the
; stack in a single instruction (i.e. mov rsp, rstk or mov
; rsp, [rsp+stack_size_padded])
- mov rstk, rsp
%if %1 < 0 ; need to store rsp on stack
- sub rsp, gprsize+stack_size_padded
- and rsp, ~(%%stack_alignment-1)
- %xdefine rstkm [rsp+stack_size_padded]
- mov rstkm, rstk
+ %xdefine rstkm [rsp + stack_size + %%pad]
+ %assign %%pad %%pad + gprsize
%else ; can keep rsp in rstk during whole function
- sub rsp, stack_size_padded
- and rsp, ~(%%stack_alignment-1)
%xdefine rstkm rstk
%endif
+ %assign stack_size_padded stack_size + ((%%pad + required_stack_alignment-1) & ~(required_stack_alignment-1))
+ mov rstk, rsp
+ and rsp, ~(required_stack_alignment-1)
+ sub rsp, stack_size_padded
+ movifnidn rstkm, rstk
%endif
WIN64_PUSH_XMM
%endif
%macro SETUP_STACK_POINTER 1
%ifnum %1
- %if %1 != 0 && (HAVE_ALIGNED_STACK == 0 || mmsize == 32)
+ %if %1 != 0 && required_stack_alignment > STACK_ALIGNMENT
%if %1 > 0
%assign regs_used (regs_used + 1)
%elif ARCH_X86_64 && regs_used == num_args && num_args <= 4 + UNIX64 * 2
%assign xmm_regs_used %1
ASSERT xmm_regs_used <= 16
%if xmm_regs_used > 8
- %assign stack_size_padded (xmm_regs_used-8)*16 + (~stack_offset&8) + 32
+ ; Allocate stack space for callee-saved xmm registers plus shadow space and align the stack.
+ %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
WIN64_PUSH_XMM
%endrep
%endif
%if stack_size_padded > 0
- %if stack_size > 0 && (mmsize == 32 || HAVE_ALIGNED_STACK == 0)
+ %if stack_size > 0 && required_stack_alignment > STACK_ALIGNMENT
mov rsp, rstkm
%else
add %1, stack_size_padded
%macro RET 0
%if stack_size_padded > 0
-%if mmsize == 32 || HAVE_ALIGNED_STACK == 0
+%if required_stack_alignment > STACK_ALIGNMENT
mov rsp, rstkm
%else
add rsp, stack_size_padded
%macro RET 0
%if stack_size_padded > 0
-%if mmsize == 32 || HAVE_ALIGNED_STACK == 0
+%if required_stack_alignment > STACK_ALIGNMENT
mov rsp, rstkm
%else
add rsp, stack_size_padded
CAT_XDEFINE cglobaled_, %2, 1
%endif
%xdefine current_function %2
- %ifidn __OUTPUT_FORMAT__,elf
+ %if FORMAT_ELF
global %2:function %%VISIBILITY
%else
global %2
; like cextern, but without the prefix
%macro cextern_naked 1
- %xdefine %1 mangle(%1)
+ %ifdef PREFIX
+ %xdefine %1 mangle(%1)
+ %endif
CAT_XDEFINE cglobaled_, %1, 1
extern %1
%endmacro
%macro const 1-2+
%xdefine %1 mangle(private_prefix %+ _ %+ %1)
- %ifidn __OUTPUT_FORMAT__,elf
+ %if FORMAT_ELF
global %1:data hidden
%else
global %1
%1: %2
%endmacro
-; This is needed for ELF, otherwise the GNU linker assumes the stack is
-; executable by default.
-%ifidn __OUTPUT_FORMAT__,elf
-SECTION .note.GNU-stack noalloc noexec nowrite progbits
+; This is needed for ELF, otherwise the GNU linker assumes the stack is executable by default.
+%if FORMAT_ELF
+ [SECTION .note.GNU-stack noalloc noexec nowrite progbits]
%endif
; cpuflags
%assign cpuflags_avx (1<<11)| cpuflags_sse42
%assign cpuflags_xop (1<<12)| cpuflags_avx
%assign cpuflags_fma4 (1<<13)| cpuflags_avx
-%assign cpuflags_avx2 (1<<14)| cpuflags_avx
-%assign cpuflags_fma3 (1<<15)| cpuflags_avx
+%assign cpuflags_fma3 (1<<14)| cpuflags_avx
+%assign cpuflags_avx2 (1<<15)| cpuflags_fma3
%assign cpuflags_cache32 (1<<16)
%assign cpuflags_cache64 (1<<17)
%assign cpuflags_bmi1 (1<<22)|cpuflags_lzcnt
%assign cpuflags_bmi2 (1<<23)|cpuflags_bmi1
-%define cpuflag(x) ((cpuflags & (cpuflags_ %+ x)) == (cpuflags_ %+ x))
-%define notcpuflag(x) ((cpuflags & (cpuflags_ %+ x)) != (cpuflags_ %+ x))
+; Returns a boolean value expressing whether or not the specified cpuflag is enabled.
+%define cpuflag(x) (((((cpuflags & (cpuflags_ %+ x)) ^ (cpuflags_ %+ x)) - 1) >> 31) & 1)
+%define notcpuflag(x) (cpuflag(x) ^ 1)
-; Takes up to 2 cpuflags from the above list.
+; Takes an arbitrary number of cpuflags from the above list.
; All subsequent functions (up to the next INIT_CPUFLAGS) is built for the specified cpu.
; You shouldn't need to invoke this macro directly, it's a subroutine for INIT_MMX &co.
-%macro INIT_CPUFLAGS 0-2
- CPU amdnop
+%macro INIT_CPUFLAGS 0-*
+ %xdefine SUFFIX
+ %undef cpuname
+ %assign cpuflags 0
+
%if %0 >= 1
- %xdefine cpuname %1
- %assign cpuflags cpuflags_%1
- %if %0 >= 2
- %xdefine cpuname %1_%2
- %assign cpuflags cpuflags | cpuflags_%2
- %endif
+ %rep %0
+ %ifdef cpuname
+ %xdefine cpuname cpuname %+ _%1
+ %else
+ %xdefine cpuname %1
+ %endif
+ %assign cpuflags cpuflags | cpuflags_%1
+ %rotate 1
+ %endrep
%xdefine SUFFIX _ %+ cpuname
+
%if cpuflag(avx)
%assign avx_enabled 1
%endif
%endif
%if cpuflag(aligned)
%define movu mova
- %elifidn %1, sse3
+ %elif cpuflag(sse3) && notcpuflag(ssse3)
%define movu lddqu
%endif
- %if ARCH_X86_64 == 0 && notcpuflag(sse2)
- CPU basicnop
+ %endif
+
+ %if ARCH_X86_64 || cpuflag(sse2)
+ %ifdef __NASM_VER__
+ ALIGNMODE k8
+ %else
+ CPU amdnop
%endif
%else
- %xdefine SUFFIX
- %undef cpuname
- %undef cpuflags
+ %ifdef __NASM_VER__
+ ALIGNMODE nop
+ %else
+ CPU basicnop
+ %endif
%endif
%endmacro
; Merge mmx and sse*
-; m# is a simd regsiter of the currently selected size
-; xm# is the corresponding xmmreg (if selcted xmm or ymm size), or mmreg (if selected mmx)
-; ym# is the corresponding ymmreg (if selcted xmm or ymm size), or mmreg (if selected mmx)
+; 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#
; (All 3 remain in sync through SWAP.)
%macro CAT_XDEFINE 3
%assign %%i 0
%rep 8
CAT_XDEFINE m, %%i, mm %+ %%i
- CAT_XDEFINE nmm, %%i, %%i
+ CAT_XDEFINE nnmm, %%i, %%i
%assign %%i %%i+1
%endrep
%rep 8
CAT_UNDEF m, %%i
- CAT_UNDEF nmm, %%i
+ CAT_UNDEF nnmm, %%i
%assign %%i %%i+1
%endrep
INIT_CPUFLAGS %1
%assign %%i 0
%rep num_mmregs
CAT_XDEFINE m, %%i, xmm %+ %%i
- CAT_XDEFINE nxmm, %%i, %%i
+ CAT_XDEFINE nnxmm, %%i, %%i
%assign %%i %%i+1
%endrep
INIT_CPUFLAGS %1
%assign %%i 0
%rep num_mmregs
CAT_XDEFINE m, %%i, ymm %+ %%i
- CAT_XDEFINE nymm, %%i, %%i
+ CAT_XDEFINE nnymm, %%i, %%i
%assign %%i %%i+1
%endrep
INIT_CPUFLAGS %1
%define xmmxmm%1 xmm%1
%define xmmymm%1 xmm%1
%define ymmmm%1 mm%1
- %define ymmxmm%1 ymm%1
+ %define ymmxmm%1 xmm%1
%define ymmymm%1 ymm%1
%define xm%1 xmm %+ m%1
%define ym%1 ymm %+ m%1
%endrep
%rep %0/2
%xdefine m%1 %%tmp%2
- CAT_XDEFINE n, m%1, %1
+ CAT_XDEFINE nn, m%1, %1
%rotate 2
%endrep
%endmacro
%xdefine %%tmp m%1
%xdefine m%1 m%2
%xdefine m%2 %%tmp
- CAT_XDEFINE n, m%1, %1
- CAT_XDEFINE n, m%2, %2
+ CAT_XDEFINE nn, m%1, %1
+ CAT_XDEFINE nn, m%2, %2
%rotate 1
%endrep
%endmacro
%macro SWAP_INTERNAL_NAME 2-*
- %xdefine %%args n %+ %1
+ %xdefine %%args nn %+ %1
%rep %0-1
- %xdefine %%args %%args, n %+ %2
+ %xdefine %%args %%args, nn %+ %2
%rotate 1
%endrep
SWAP_INTERNAL_NUM %%args
%assign %%i 0
%rep num_mmregs
CAT_XDEFINE m, %%i, %1_m %+ %%i
- CAT_XDEFINE n, m %+ %%i, %%i
+ CAT_XDEFINE nn, m %+ %%i, %%i
%assign %%i %%i+1
%endrep
%endif
; Append cpuflags to the callee's name iff the appended name is known and the plain name isn't
%macro call 1
- call_internal %1, %1 %+ SUFFIX
+ call_internal %1 %+ SUFFIX, %1
%endmacro
%macro call_internal 2
- %xdefine %%i %1
- %ifndef cglobaled_%1
- %ifdef cglobaled_%2
- %xdefine %%i %2
+ %xdefine %%i %2
+ %ifndef cglobaled_%2
+ %ifdef cglobaled_%1
+ %xdefine %%i %1
%endif
%endif
call %%i
%endmacro
;%1 == instruction
-;%2 == 1 if float, 0 if int
-;%3 == 1 if non-destructive or 4-operand (xmm, xmm, xmm, imm), 0 otherwise
-;%4 == 1 if commutative (i.e. doesn't matter which src arg is which), 0 if not
-;%5+: operands
-%macro RUN_AVX_INSTR 5-8+
- %ifnum sizeof%6
- %assign %%sizeofreg sizeof%6
- %elifnum sizeof%5
- %assign %%sizeofreg sizeof%5
+;%2 == minimal instruction set
+;%3 == 1 if float, 0 if int
+;%4 == 1 if non-destructive or 4-operand (xmm, xmm, xmm, imm), 0 otherwise
+;%5 == 1 if commutative (i.e. doesn't matter which src arg is which), 0 if not
+;%6+: operands
+%macro RUN_AVX_INSTR 6-9+
+ %ifnum sizeof%7
+ %assign __sizeofreg sizeof%7
+ %elifnum sizeof%6
+ %assign __sizeofreg sizeof%6
%else
- %assign %%sizeofreg mmsize
+ %assign __sizeofreg mmsize
%endif
- %assign %%emulate_avx 0
- %if avx_enabled && %%sizeofreg >= 16
- %xdefine %%instr v%1
+ %assign __emulate_avx 0
+ %if avx_enabled && __sizeofreg >= 16
+ %xdefine __instr v%1
%else
- %xdefine %%instr %1
- %if %0 >= 7+%3
- %assign %%emulate_avx 1
+ %xdefine __instr %1
+ %if %0 >= 8+%4
+ %assign __emulate_avx 1
+ %endif
+ %endif
+ %ifnidn %2, fnord
+ %ifdef cpuname
+ %if notcpuflag(%2)
+ %error use of ``%1'' %2 instruction in cpuname function: current_function
+ %elif cpuflags_%2 < cpuflags_sse && notcpuflag(sse2) && __sizeofreg > 8
+ %error use of ``%1'' sse2 instruction in cpuname function: current_function
+ %endif
%endif
%endif
- %if %%emulate_avx
- %xdefine %%src1 %6
- %xdefine %%src2 %7
- %ifnidn %5, %6
- %if %0 >= 8
- CHECK_AVX_INSTR_EMU {%1 %5, %6, %7, %8}, %5, %7, %8
+ %if __emulate_avx
+ %xdefine __src1 %7
+ %xdefine __src2 %8
+ %ifnidn %6, %7
+ %if %0 >= 9
+ CHECK_AVX_INSTR_EMU {%1 %6, %7, %8, %9}, %6, %8, %9
%else
- CHECK_AVX_INSTR_EMU {%1 %5, %6, %7}, %5, %7
+ CHECK_AVX_INSTR_EMU {%1 %6, %7, %8}, %6, %8
%endif
- %if %4 && %3 == 0
- %ifnid %7
+ %if %5 && %4 == 0
+ %ifnid %8
; 3-operand AVX instructions with a memory arg can only have it in src2,
; whereas SSE emulation prefers to have it in src1 (i.e. the mov).
; So, if the instruction is commutative with a memory arg, swap them.
- %xdefine %%src1 %7
- %xdefine %%src2 %6
+ %xdefine __src1 %8
+ %xdefine __src2 %7
%endif
%endif
- %if %%sizeofreg == 8
- MOVQ %5, %%src1
- %elif %2
- MOVAPS %5, %%src1
+ %if __sizeofreg == 8
+ MOVQ %6, __src1
+ %elif %3
+ MOVAPS %6, __src1
%else
- MOVDQA %5, %%src1
+ MOVDQA %6, __src1
%endif
%endif
- %if %0 >= 8
- %1 %5, %%src2, %8
+ %if %0 >= 9
+ %1 %6, __src2, %9
%else
- %1 %5, %%src2
+ %1 %6, __src2
%endif
- %elif %0 >= 8
- %%instr %5, %6, %7, %8
+ %elif %0 >= 9
+ __instr %6, %7, %8, %9
+ %elif %0 == 8
+ __instr %6, %7, %8
%elif %0 == 7
- %%instr %5, %6, %7
- %elif %0 == 6
- %%instr %5, %6
+ __instr %6, %7
%else
- %%instr %5
+ __instr %6
%endif
%endmacro
;%1 == instruction
-;%2 == 1 if float, 0 if int
-;%3 == 1 if non-destructive or 4-operand (xmm, xmm, xmm, imm), 0 otherwise
-;%4 == 1 if commutative (i.e. doesn't matter which src arg is which), 0 if not
-%macro AVX_INSTR 1-4 0, 1, 0
- %macro %1 1-9 fnord, fnord, fnord, fnord, %1, %2, %3, %4
+;%2 == minimal instruction set
+;%3 == 1 if float, 0 if int
+;%4 == 1 if non-destructive or 4-operand (xmm, xmm, xmm, imm), 0 otherwise
+;%5 == 1 if commutative (i.e. doesn't matter which src arg is which), 0 if not
+%macro AVX_INSTR 1-5 fnord, 0, 1, 0
+ %macro %1 1-10 fnord, fnord, fnord, fnord, %1, %2, %3, %4, %5
%ifidn %2, fnord
- RUN_AVX_INSTR %6, %7, %8, %9, %1
+ RUN_AVX_INSTR %6, %7, %8, %9, %10, %1
%elifidn %3, fnord
- RUN_AVX_INSTR %6, %7, %8, %9, %1, %2
+ RUN_AVX_INSTR %6, %7, %8, %9, %10, %1, %2
%elifidn %4, fnord
- RUN_AVX_INSTR %6, %7, %8, %9, %1, %2, %3
+ RUN_AVX_INSTR %6, %7, %8, %9, %10, %1, %2, %3
%elifidn %5, fnord
- RUN_AVX_INSTR %6, %7, %8, %9, %1, %2, %3, %4
+ RUN_AVX_INSTR %6, %7, %8, %9, %10, %1, %2, %3, %4
%else
- RUN_AVX_INSTR %6, %7, %8, %9, %1, %2, %3, %4, %5
+ RUN_AVX_INSTR %6, %7, %8, %9, %10, %1, %2, %3, %4, %5
%endif
%endmacro
%endmacro
; Instructions with both VEX and non-VEX encodings
; Non-destructive instructions are written without parameters
-AVX_INSTR addpd, 1, 0, 1
-AVX_INSTR addps, 1, 0, 1
-AVX_INSTR addsd, 1, 0, 1
-AVX_INSTR addss, 1, 0, 1
-AVX_INSTR addsubpd, 1, 0, 0
-AVX_INSTR addsubps, 1, 0, 0
-AVX_INSTR aesdec, 0, 0, 0
-AVX_INSTR aesdeclast, 0, 0, 0
-AVX_INSTR aesenc, 0, 0, 0
-AVX_INSTR aesenclast, 0, 0, 0
+AVX_INSTR addpd, sse2, 1, 0, 1
+AVX_INSTR addps, sse, 1, 0, 1
+AVX_INSTR addsd, sse2, 1, 0, 1
+AVX_INSTR addss, sse, 1, 0, 1
+AVX_INSTR addsubpd, sse3, 1, 0, 0
+AVX_INSTR addsubps, sse3, 1, 0, 0
+AVX_INSTR aesdec, fnord, 0, 0, 0
+AVX_INSTR aesdeclast, fnord, 0, 0, 0
+AVX_INSTR aesenc, fnord, 0, 0, 0
+AVX_INSTR aesenclast, fnord, 0, 0, 0
AVX_INSTR aesimc
AVX_INSTR aeskeygenassist
-AVX_INSTR andnpd, 1, 0, 0
-AVX_INSTR andnps, 1, 0, 0
-AVX_INSTR andpd, 1, 0, 1
-AVX_INSTR andps, 1, 0, 1
-AVX_INSTR blendpd, 1, 0, 0
-AVX_INSTR blendps, 1, 0, 0
-AVX_INSTR blendvpd, 1, 0, 0
-AVX_INSTR blendvps, 1, 0, 0
-AVX_INSTR cmppd, 1, 1, 0
-AVX_INSTR cmpps, 1, 1, 0
-AVX_INSTR cmpsd, 1, 1, 0
-AVX_INSTR cmpss, 1, 1, 0
-AVX_INSTR comisd
-AVX_INSTR comiss
-AVX_INSTR cvtdq2pd
-AVX_INSTR cvtdq2ps
-AVX_INSTR cvtpd2dq
-AVX_INSTR cvtpd2ps
-AVX_INSTR cvtps2dq
-AVX_INSTR cvtps2pd
-AVX_INSTR cvtsd2si
-AVX_INSTR cvtsd2ss
-AVX_INSTR cvtsi2sd
-AVX_INSTR cvtsi2ss
-AVX_INSTR cvtss2sd
-AVX_INSTR cvtss2si
-AVX_INSTR cvttpd2dq
-AVX_INSTR cvttps2dq
-AVX_INSTR cvttsd2si
-AVX_INSTR cvttss2si
-AVX_INSTR divpd, 1, 0, 0
-AVX_INSTR divps, 1, 0, 0
-AVX_INSTR divsd, 1, 0, 0
-AVX_INSTR divss, 1, 0, 0
-AVX_INSTR dppd, 1, 1, 0
-AVX_INSTR dpps, 1, 1, 0
-AVX_INSTR extractps
-AVX_INSTR haddpd, 1, 0, 0
-AVX_INSTR haddps, 1, 0, 0
-AVX_INSTR hsubpd, 1, 0, 0
-AVX_INSTR hsubps, 1, 0, 0
-AVX_INSTR insertps, 1, 1, 0
-AVX_INSTR lddqu
-AVX_INSTR ldmxcsr
-AVX_INSTR maskmovdqu
-AVX_INSTR maxpd, 1, 0, 1
-AVX_INSTR maxps, 1, 0, 1
-AVX_INSTR maxsd, 1, 0, 1
-AVX_INSTR maxss, 1, 0, 1
-AVX_INSTR minpd, 1, 0, 1
-AVX_INSTR minps, 1, 0, 1
-AVX_INSTR minsd, 1, 0, 1
-AVX_INSTR minss, 1, 0, 1
-AVX_INSTR movapd
-AVX_INSTR movaps
-AVX_INSTR movd
-AVX_INSTR movddup
-AVX_INSTR movdqa
-AVX_INSTR movdqu
-AVX_INSTR movhlps, 1, 0, 0
-AVX_INSTR movhpd, 1, 0, 0
-AVX_INSTR movhps, 1, 0, 0
-AVX_INSTR movlhps, 1, 0, 0
-AVX_INSTR movlpd, 1, 0, 0
-AVX_INSTR movlps, 1, 0, 0
-AVX_INSTR movmskpd
-AVX_INSTR movmskps
-AVX_INSTR movntdq
-AVX_INSTR movntdqa
-AVX_INSTR movntpd
-AVX_INSTR movntps
-AVX_INSTR movq
-AVX_INSTR movsd, 1, 0, 0
-AVX_INSTR movshdup
-AVX_INSTR movsldup
-AVX_INSTR movss, 1, 0, 0
-AVX_INSTR movupd
-AVX_INSTR movups
-AVX_INSTR mpsadbw, 0, 1, 0
-AVX_INSTR mulpd, 1, 0, 1
-AVX_INSTR mulps, 1, 0, 1
-AVX_INSTR mulsd, 1, 0, 1
-AVX_INSTR mulss, 1, 0, 1
-AVX_INSTR orpd, 1, 0, 1
-AVX_INSTR orps, 1, 0, 1
-AVX_INSTR pabsb
-AVX_INSTR pabsd
-AVX_INSTR pabsw
-AVX_INSTR packsswb, 0, 0, 0
-AVX_INSTR packssdw, 0, 0, 0
-AVX_INSTR packuswb, 0, 0, 0
-AVX_INSTR packusdw, 0, 0, 0
-AVX_INSTR paddb, 0, 0, 1
-AVX_INSTR paddw, 0, 0, 1
-AVX_INSTR paddd, 0, 0, 1
-AVX_INSTR paddq, 0, 0, 1
-AVX_INSTR paddsb, 0, 0, 1
-AVX_INSTR paddsw, 0, 0, 1
-AVX_INSTR paddusb, 0, 0, 1
-AVX_INSTR paddusw, 0, 0, 1
-AVX_INSTR palignr, 0, 1, 0
-AVX_INSTR pand, 0, 0, 1
-AVX_INSTR pandn, 0, 0, 0
-AVX_INSTR pavgb, 0, 0, 1
-AVX_INSTR pavgw, 0, 0, 1
-AVX_INSTR pblendvb, 0, 0, 0
-AVX_INSTR pblendw, 0, 1, 0
-AVX_INSTR pclmulqdq, 0, 1, 0
-AVX_INSTR pcmpestri
-AVX_INSTR pcmpestrm
-AVX_INSTR pcmpistri
-AVX_INSTR pcmpistrm
-AVX_INSTR pcmpeqb, 0, 0, 1
-AVX_INSTR pcmpeqw, 0, 0, 1
-AVX_INSTR pcmpeqd, 0, 0, 1
-AVX_INSTR pcmpeqq, 0, 0, 1
-AVX_INSTR pcmpgtb, 0, 0, 0
-AVX_INSTR pcmpgtw, 0, 0, 0
-AVX_INSTR pcmpgtd, 0, 0, 0
-AVX_INSTR pcmpgtq, 0, 0, 0
-AVX_INSTR pextrb
-AVX_INSTR pextrd
-AVX_INSTR pextrq
-AVX_INSTR pextrw
-AVX_INSTR phaddw, 0, 0, 0
-AVX_INSTR phaddd, 0, 0, 0
-AVX_INSTR phaddsw, 0, 0, 0
-AVX_INSTR phminposuw
-AVX_INSTR phsubw, 0, 0, 0
-AVX_INSTR phsubd, 0, 0, 0
-AVX_INSTR phsubsw, 0, 0, 0
-AVX_INSTR pinsrb, 0, 1, 0
-AVX_INSTR pinsrd, 0, 1, 0
-AVX_INSTR pinsrq, 0, 1, 0
-AVX_INSTR pinsrw, 0, 1, 0
-AVX_INSTR pmaddwd, 0, 0, 1
-AVX_INSTR pmaddubsw, 0, 0, 0
-AVX_INSTR pmaxsb, 0, 0, 1
-AVX_INSTR pmaxsw, 0, 0, 1
-AVX_INSTR pmaxsd, 0, 0, 1
-AVX_INSTR pmaxub, 0, 0, 1
-AVX_INSTR pmaxuw, 0, 0, 1
-AVX_INSTR pmaxud, 0, 0, 1
-AVX_INSTR pminsb, 0, 0, 1
-AVX_INSTR pminsw, 0, 0, 1
-AVX_INSTR pminsd, 0, 0, 1
-AVX_INSTR pminub, 0, 0, 1
-AVX_INSTR pminuw, 0, 0, 1
-AVX_INSTR pminud, 0, 0, 1
-AVX_INSTR pmovmskb
-AVX_INSTR pmovsxbw
-AVX_INSTR pmovsxbd
-AVX_INSTR pmovsxbq
-AVX_INSTR pmovsxwd
-AVX_INSTR pmovsxwq
-AVX_INSTR pmovsxdq
-AVX_INSTR pmovzxbw
-AVX_INSTR pmovzxbd
-AVX_INSTR pmovzxbq
-AVX_INSTR pmovzxwd
-AVX_INSTR pmovzxwq
-AVX_INSTR pmovzxdq
-AVX_INSTR pmuldq, 0, 0, 1
-AVX_INSTR pmulhrsw, 0, 0, 1
-AVX_INSTR pmulhuw, 0, 0, 1
-AVX_INSTR pmulhw, 0, 0, 1
-AVX_INSTR pmullw, 0, 0, 1
-AVX_INSTR pmulld, 0, 0, 1
-AVX_INSTR pmuludq, 0, 0, 1
-AVX_INSTR por, 0, 0, 1
-AVX_INSTR psadbw, 0, 0, 1
-AVX_INSTR pshufb, 0, 0, 0
-AVX_INSTR pshufd
-AVX_INSTR pshufhw
-AVX_INSTR pshuflw
-AVX_INSTR psignb, 0, 0, 0
-AVX_INSTR psignw, 0, 0, 0
-AVX_INSTR psignd, 0, 0, 0
-AVX_INSTR psllw, 0, 0, 0
-AVX_INSTR pslld, 0, 0, 0
-AVX_INSTR psllq, 0, 0, 0
-AVX_INSTR pslldq, 0, 0, 0
-AVX_INSTR psraw, 0, 0, 0
-AVX_INSTR psrad, 0, 0, 0
-AVX_INSTR psrlw, 0, 0, 0
-AVX_INSTR psrld, 0, 0, 0
-AVX_INSTR psrlq, 0, 0, 0
-AVX_INSTR psrldq, 0, 0, 0
-AVX_INSTR psubb, 0, 0, 0
-AVX_INSTR psubw, 0, 0, 0
-AVX_INSTR psubd, 0, 0, 0
-AVX_INSTR psubq, 0, 0, 0
-AVX_INSTR psubsb, 0, 0, 0
-AVX_INSTR psubsw, 0, 0, 0
-AVX_INSTR psubusb, 0, 0, 0
-AVX_INSTR psubusw, 0, 0, 0
-AVX_INSTR ptest
-AVX_INSTR punpckhbw, 0, 0, 0
-AVX_INSTR punpckhwd, 0, 0, 0
-AVX_INSTR punpckhdq, 0, 0, 0
-AVX_INSTR punpckhqdq, 0, 0, 0
-AVX_INSTR punpcklbw, 0, 0, 0
-AVX_INSTR punpcklwd, 0, 0, 0
-AVX_INSTR punpckldq, 0, 0, 0
-AVX_INSTR punpcklqdq, 0, 0, 0
-AVX_INSTR pxor, 0, 0, 1
-AVX_INSTR rcpps, 1, 0, 0
-AVX_INSTR rcpss, 1, 0, 0
-AVX_INSTR roundpd
-AVX_INSTR roundps
-AVX_INSTR roundsd
-AVX_INSTR roundss
-AVX_INSTR rsqrtps, 1, 0, 0
-AVX_INSTR rsqrtss, 1, 0, 0
-AVX_INSTR shufpd, 1, 1, 0
-AVX_INSTR shufps, 1, 1, 0
-AVX_INSTR sqrtpd, 1, 0, 0
-AVX_INSTR sqrtps, 1, 0, 0
-AVX_INSTR sqrtsd, 1, 0, 0
-AVX_INSTR sqrtss, 1, 0, 0
-AVX_INSTR stmxcsr
-AVX_INSTR subpd, 1, 0, 0
-AVX_INSTR subps, 1, 0, 0
-AVX_INSTR subsd, 1, 0, 0
-AVX_INSTR subss, 1, 0, 0
-AVX_INSTR ucomisd
-AVX_INSTR ucomiss
-AVX_INSTR unpckhpd, 1, 0, 0
-AVX_INSTR unpckhps, 1, 0, 0
-AVX_INSTR unpcklpd, 1, 0, 0
-AVX_INSTR unpcklps, 1, 0, 0
-AVX_INSTR xorpd, 1, 0, 1
-AVX_INSTR xorps, 1, 0, 1
+AVX_INSTR andnpd, sse2, 1, 0, 0
+AVX_INSTR andnps, sse, 1, 0, 0
+AVX_INSTR andpd, sse2, 1, 0, 1
+AVX_INSTR andps, sse, 1, 0, 1
+AVX_INSTR blendpd, sse4, 1, 0, 0
+AVX_INSTR blendps, sse4, 1, 0, 0
+AVX_INSTR blendvpd, sse4, 1, 0, 0
+AVX_INSTR blendvps, sse4, 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 comisd, sse2
+AVX_INSTR comiss, sse
+AVX_INSTR cvtdq2pd, sse2
+AVX_INSTR cvtdq2ps, sse2
+AVX_INSTR cvtpd2dq, sse2
+AVX_INSTR cvtpd2ps, sse2
+AVX_INSTR cvtps2dq, sse2
+AVX_INSTR cvtps2pd, sse2
+AVX_INSTR cvtsd2si, sse2
+AVX_INSTR cvtsd2ss, sse2
+AVX_INSTR cvtsi2sd, sse2
+AVX_INSTR cvtsi2ss, sse
+AVX_INSTR cvtss2sd, sse2
+AVX_INSTR cvtss2si, sse
+AVX_INSTR cvttpd2dq, sse2
+AVX_INSTR cvttps2dq, sse2
+AVX_INSTR cvttsd2si, sse2
+AVX_INSTR cvttss2si, sse
+AVX_INSTR divpd, sse2, 1, 0, 0
+AVX_INSTR divps, sse, 1, 0, 0
+AVX_INSTR divsd, sse2, 1, 0, 0
+AVX_INSTR divss, sse, 1, 0, 0
+AVX_INSTR dppd, sse4, 1, 1, 0
+AVX_INSTR dpps, sse4, 1, 1, 0
+AVX_INSTR extractps, sse4
+AVX_INSTR haddpd, sse3, 1, 0, 0
+AVX_INSTR haddps, sse3, 1, 0, 0
+AVX_INSTR hsubpd, sse3, 1, 0, 0
+AVX_INSTR hsubps, sse3, 1, 0, 0
+AVX_INSTR insertps, sse4, 1, 1, 0
+AVX_INSTR lddqu, sse3
+AVX_INSTR ldmxcsr, sse
+AVX_INSTR maskmovdqu, sse2
+AVX_INSTR maxpd, sse2, 1, 0, 1
+AVX_INSTR maxps, sse, 1, 0, 1
+AVX_INSTR maxsd, sse2, 1, 0, 1
+AVX_INSTR maxss, sse, 1, 0, 1
+AVX_INSTR minpd, sse2, 1, 0, 1
+AVX_INSTR minps, sse, 1, 0, 1
+AVX_INSTR minsd, sse2, 1, 0, 1
+AVX_INSTR minss, sse, 1, 0, 1
+AVX_INSTR movapd, sse2
+AVX_INSTR movaps, sse
+AVX_INSTR movd, mmx
+AVX_INSTR movddup, sse3
+AVX_INSTR movdqa, sse2
+AVX_INSTR movdqu, sse2
+AVX_INSTR movhlps, sse, 1, 0, 0
+AVX_INSTR movhpd, sse2, 1, 0, 0
+AVX_INSTR movhps, sse, 1, 0, 0
+AVX_INSTR movlhps, sse, 1, 0, 0
+AVX_INSTR movlpd, sse2, 1, 0, 0
+AVX_INSTR movlps, sse, 1, 0, 0
+AVX_INSTR movmskpd, sse2
+AVX_INSTR movmskps, sse
+AVX_INSTR movntdq, sse2
+AVX_INSTR movntdqa, sse4
+AVX_INSTR movntpd, sse2
+AVX_INSTR movntps, sse
+AVX_INSTR movq, mmx
+AVX_INSTR movsd, sse2, 1, 0, 0
+AVX_INSTR movshdup, sse3
+AVX_INSTR movsldup, sse3
+AVX_INSTR movss, sse, 1, 0, 0
+AVX_INSTR movupd, sse2
+AVX_INSTR movups, sse
+AVX_INSTR mpsadbw, sse4
+AVX_INSTR mulpd, sse2, 1, 0, 1
+AVX_INSTR mulps, sse, 1, 0, 1
+AVX_INSTR mulsd, sse2, 1, 0, 1
+AVX_INSTR mulss, sse, 1, 0, 1
+AVX_INSTR orpd, sse2, 1, 0, 1
+AVX_INSTR orps, sse, 1, 0, 1
+AVX_INSTR pabsb, ssse3
+AVX_INSTR pabsd, ssse3
+AVX_INSTR pabsw, ssse3
+AVX_INSTR packsswb, mmx, 0, 0, 0
+AVX_INSTR packssdw, mmx, 0, 0, 0
+AVX_INSTR packuswb, mmx, 0, 0, 0
+AVX_INSTR packusdw, sse4, 0, 0, 0
+AVX_INSTR paddb, mmx, 0, 0, 1
+AVX_INSTR paddw, mmx, 0, 0, 1
+AVX_INSTR paddd, mmx, 0, 0, 1
+AVX_INSTR paddq, sse2, 0, 0, 1
+AVX_INSTR paddsb, mmx, 0, 0, 1
+AVX_INSTR paddsw, mmx, 0, 0, 1
+AVX_INSTR paddusb, mmx, 0, 0, 1
+AVX_INSTR paddusw, mmx, 0, 0, 1
+AVX_INSTR palignr, ssse3
+AVX_INSTR pand, mmx, 0, 0, 1
+AVX_INSTR pandn, mmx, 0, 0, 0
+AVX_INSTR pavgb, mmx2, 0, 0, 1
+AVX_INSTR pavgw, mmx2, 0, 0, 1
+AVX_INSTR pblendvb, sse4, 0, 0, 0
+AVX_INSTR pblendw, sse4
+AVX_INSTR pclmulqdq
+AVX_INSTR pcmpestri, sse42
+AVX_INSTR pcmpestrm, sse42
+AVX_INSTR pcmpistri, sse42
+AVX_INSTR pcmpistrm, sse42
+AVX_INSTR pcmpeqb, mmx, 0, 0, 1
+AVX_INSTR pcmpeqw, mmx, 0, 0, 1
+AVX_INSTR pcmpeqd, mmx, 0, 0, 1
+AVX_INSTR pcmpeqq, sse4, 0, 0, 1
+AVX_INSTR pcmpgtb, mmx, 0, 0, 0
+AVX_INSTR pcmpgtw, mmx, 0, 0, 0
+AVX_INSTR pcmpgtd, mmx, 0, 0, 0
+AVX_INSTR pcmpgtq, sse42, 0, 0, 0
+AVX_INSTR pextrb, sse4
+AVX_INSTR pextrd, sse4
+AVX_INSTR pextrq, sse4
+AVX_INSTR pextrw, mmx2
+AVX_INSTR phaddw, ssse3, 0, 0, 0
+AVX_INSTR phaddd, ssse3, 0, 0, 0
+AVX_INSTR phaddsw, ssse3, 0, 0, 0
+AVX_INSTR phminposuw, sse4
+AVX_INSTR phsubw, ssse3, 0, 0, 0
+AVX_INSTR phsubd, ssse3, 0, 0, 0
+AVX_INSTR phsubsw, ssse3, 0, 0, 0
+AVX_INSTR pinsrb, sse4
+AVX_INSTR pinsrd, sse4
+AVX_INSTR pinsrq, sse4
+AVX_INSTR pinsrw, mmx2
+AVX_INSTR pmaddwd, mmx, 0, 0, 1
+AVX_INSTR pmaddubsw, ssse3, 0, 0, 0
+AVX_INSTR pmaxsb, sse4, 0, 0, 1
+AVX_INSTR pmaxsw, mmx2, 0, 0, 1
+AVX_INSTR pmaxsd, sse4, 0, 0, 1
+AVX_INSTR pmaxub, mmx2, 0, 0, 1
+AVX_INSTR pmaxuw, sse4, 0, 0, 1
+AVX_INSTR pmaxud, sse4, 0, 0, 1
+AVX_INSTR pminsb, sse4, 0, 0, 1
+AVX_INSTR pminsw, mmx2, 0, 0, 1
+AVX_INSTR pminsd, sse4, 0, 0, 1
+AVX_INSTR pminub, mmx2, 0, 0, 1
+AVX_INSTR pminuw, sse4, 0, 0, 1
+AVX_INSTR pminud, sse4, 0, 0, 1
+AVX_INSTR pmovmskb, mmx2
+AVX_INSTR pmovsxbw, sse4
+AVX_INSTR pmovsxbd, sse4
+AVX_INSTR pmovsxbq, sse4
+AVX_INSTR pmovsxwd, sse4
+AVX_INSTR pmovsxwq, sse4
+AVX_INSTR pmovsxdq, sse4
+AVX_INSTR pmovzxbw, sse4
+AVX_INSTR pmovzxbd, sse4
+AVX_INSTR pmovzxbq, sse4
+AVX_INSTR pmovzxwd, sse4
+AVX_INSTR pmovzxwq, sse4
+AVX_INSTR pmovzxdq, sse4
+AVX_INSTR pmuldq, sse4, 0, 0, 1
+AVX_INSTR pmulhrsw, ssse3, 0, 0, 1
+AVX_INSTR pmulhuw, mmx2, 0, 0, 1
+AVX_INSTR pmulhw, mmx, 0, 0, 1
+AVX_INSTR pmullw, mmx, 0, 0, 1
+AVX_INSTR pmulld, sse4, 0, 0, 1
+AVX_INSTR pmuludq, sse2, 0, 0, 1
+AVX_INSTR por, mmx, 0, 0, 1
+AVX_INSTR psadbw, mmx2, 0, 0, 1
+AVX_INSTR pshufb, ssse3, 0, 0, 0
+AVX_INSTR pshufd, sse2
+AVX_INSTR pshufhw, sse2
+AVX_INSTR pshuflw, sse2
+AVX_INSTR psignb, ssse3, 0, 0, 0
+AVX_INSTR psignw, ssse3, 0, 0, 0
+AVX_INSTR psignd, ssse3, 0, 0, 0
+AVX_INSTR psllw, mmx, 0, 0, 0
+AVX_INSTR pslld, mmx, 0, 0, 0
+AVX_INSTR psllq, mmx, 0, 0, 0
+AVX_INSTR pslldq, sse2, 0, 0, 0
+AVX_INSTR psraw, mmx, 0, 0, 0
+AVX_INSTR psrad, mmx, 0, 0, 0
+AVX_INSTR psrlw, mmx, 0, 0, 0
+AVX_INSTR psrld, mmx, 0, 0, 0
+AVX_INSTR psrlq, mmx, 0, 0, 0
+AVX_INSTR psrldq, sse2, 0, 0, 0
+AVX_INSTR psubb, mmx, 0, 0, 0
+AVX_INSTR psubw, mmx, 0, 0, 0
+AVX_INSTR psubd, mmx, 0, 0, 0
+AVX_INSTR psubq, sse2, 0, 0, 0
+AVX_INSTR psubsb, mmx, 0, 0, 0
+AVX_INSTR psubsw, mmx, 0, 0, 0
+AVX_INSTR psubusb, mmx, 0, 0, 0
+AVX_INSTR psubusw, mmx, 0, 0, 0
+AVX_INSTR ptest, sse4
+AVX_INSTR punpckhbw, mmx, 0, 0, 0
+AVX_INSTR punpckhwd, mmx, 0, 0, 0
+AVX_INSTR punpckhdq, mmx, 0, 0, 0
+AVX_INSTR punpckhqdq, sse2, 0, 0, 0
+AVX_INSTR punpcklbw, mmx, 0, 0, 0
+AVX_INSTR punpcklwd, mmx, 0, 0, 0
+AVX_INSTR punpckldq, mmx, 0, 0, 0
+AVX_INSTR punpcklqdq, sse2, 0, 0, 0
+AVX_INSTR pxor, mmx, 0, 0, 1
+AVX_INSTR rcpps, sse, 1, 0, 0
+AVX_INSTR rcpss, sse, 1, 0, 0
+AVX_INSTR roundpd, sse4
+AVX_INSTR roundps, sse4
+AVX_INSTR roundsd, sse4
+AVX_INSTR roundss, sse4
+AVX_INSTR rsqrtps, sse, 1, 0, 0
+AVX_INSTR rsqrtss, sse, 1, 0, 0
+AVX_INSTR shufpd, sse2, 1, 1, 0
+AVX_INSTR shufps, sse, 1, 1, 0
+AVX_INSTR sqrtpd, sse2, 1, 0, 0
+AVX_INSTR sqrtps, sse, 1, 0, 0
+AVX_INSTR sqrtsd, sse2, 1, 0, 0
+AVX_INSTR sqrtss, sse, 1, 0, 0
+AVX_INSTR stmxcsr, sse
+AVX_INSTR subpd, sse2, 1, 0, 0
+AVX_INSTR subps, sse, 1, 0, 0
+AVX_INSTR subsd, sse2, 1, 0, 0
+AVX_INSTR subss, sse, 1, 0, 0
+AVX_INSTR ucomisd, sse2
+AVX_INSTR ucomiss, sse
+AVX_INSTR unpckhpd, sse2, 1, 0, 0
+AVX_INSTR unpckhps, sse, 1, 0, 0
+AVX_INSTR unpcklpd, sse2, 1, 0, 0
+AVX_INSTR unpcklps, sse, 1, 0, 0
+AVX_INSTR xorpd, sse2, 1, 0, 1
+AVX_INSTR xorps, sse, 1, 0, 1
; 3DNow instructions, for sharing code between AVX, SSE and 3DN
-AVX_INSTR pfadd, 1, 0, 1
-AVX_INSTR pfsub, 1, 0, 0
-AVX_INSTR pfmul, 1, 0, 1
+AVX_INSTR pfadd, 3dnow, 1, 0, 1
+AVX_INSTR pfsub, 3dnow, 1, 0, 0
+AVX_INSTR pfmul, 3dnow, 1, 0, 1
; base-4 constants for shuffles
%assign i 0
%macro %1 4-7 %1, %2, %3
%if cpuflag(xop)
v%5 %1, %2, %3, %4
- %else
+ %elifnidn %1, %4
%6 %1, %2, %3
%7 %1, %4
+ %else
+ %error non-xop emulation of ``%5 %1, %2, %3, %4'' is not supported
%endif
%endmacro
%endmacro
-FMA_INSTR pmacsdd, pmulld, paddd
FMA_INSTR pmacsww, pmullw, paddw
+FMA_INSTR pmacsdd, pmulld, paddd ; sse4 emulation
+FMA_INSTR pmacsdql, pmuldq, paddq ; sse4 emulation
FMA_INSTR pmadcswd, pmaddwd, paddd
; convert FMA4 to FMA3 if possible
FMA4_INSTR fnmsubsd, fnmsub132sd, fnmsub213sd, fnmsub231sd
FMA4_INSTR fnmsubss, fnmsub132ss, fnmsub213ss, fnmsub231ss
-; workaround: vpbroadcastq is broken in x86_32 due to a yasm bug
-%if ARCH_X86_64 == 0
-%macro vpbroadcastq 2
-%if sizeof%1 == 16
- movddup %1, %2
-%else
- vbroadcastsd %1, %2
-%endif
-%endmacro
+; 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
+ %macro vpbroadcastq 2
+ %if sizeof%1 == 16
+ movddup %1, %2
+ %else
+ vbroadcastsd %1, %2
+ %endif
+ %endmacro
+ %endif
%endif
projects / x264 / blobdiff