1 ;*****************************************************************************
2 ;* x86inc.asm: x264asm abstraction layer
3 ;*****************************************************************************
4 ;* Copyright (C) 2005-2012 x264 project
6 ;* Authors: Loren Merritt <lorenm@u.washington.edu>
7 ;* Anton Mitrofanov <BugMaster@narod.ru>
8 ;* Jason Garrett-Glaser <darkshikari@gmail.com>
9 ;* Henrik Gramner <hengar-6@student.ltu.se>
11 ;* Permission to use, copy, modify, and/or distribute this software for any
12 ;* purpose with or without fee is hereby granted, provided that the above
13 ;* copyright notice and this permission notice appear in all copies.
15 ;* THE SOFTWARE IS PROVIDED "AS IS" AND THE AUTHOR DISCLAIMS ALL WARRANTIES
16 ;* WITH REGARD TO THIS SOFTWARE INCLUDING ALL IMPLIED WARRANTIES OF
17 ;* MERCHANTABILITY AND FITNESS. IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR
18 ;* ANY SPECIAL, DIRECT, INDIRECT, OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES
19 ;* WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN AN
20 ;* ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ARISING OUT OF
21 ;* OR IN CONNECTION WITH THE USE OR PERFORMANCE OF THIS SOFTWARE.
22 ;*****************************************************************************
24 ; This is a header file for the x264ASM assembly language, which uses
25 ; NASM/YASM syntax combined with a large number of macros to provide easy
26 ; abstraction between different calling conventions (x86_32, win64, linux64).
27 ; It also has various other useful features to simplify writing the kind of
28 ; DSP functions that are most often used in x264.
30 ; Unlike the rest of x264, this file is available under an ISC license, as it
31 ; has significant usefulness outside of x264 and we want it to be available
32 ; to the largest audience possible. Of course, if you modify it for your own
33 ; purposes to add a new feature, we strongly encourage contributing a patch
34 ; as this feature might be useful for others as well. Send patches or ideas
35 ; to x264-devel@videolan.org .
37 %define program_name ff
42 %ifidn __OUTPUT_FORMAT__,win32
44 %elifidn __OUTPUT_FORMAT__,win64
52 %define mangle(x) _ %+ x
57 ; Name of the .rodata section.
58 %macro SECTION_RODATA 0-1 16
59 ; Kludge: Something on OS X fails to align .rodata even given an align
60 ; attribute, so use a different read-only section. This has been fixed in
61 ; yasm 0.8.0 and nasm 2.6.
62 %ifdef __YASM_VERSION_ID__
63 %if __YASM_VERSION_ID__ < 00080000h
64 %define NEED_MACHO_RODATA_KLUDGE
66 %elifdef __NASM_VERSION_ID__
67 %if __NASM_VERSION_ID__ < 02060000h
68 %define NEED_MACHO_RODATA_KLUDGE
72 %ifidn __OUTPUT_FORMAT__,aout
75 %ifndef NEED_MACHO_RODATA_KLUDGE
76 SECTION .rodata align=%1
78 %ifidn __OUTPUT_FORMAT__,macho64
79 SECTION .text align=%1
80 %elifidn __OUTPUT_FORMAT__,macho
81 SECTION .text align=%1
84 SECTION .rodata align=%1
89 %undef NEED_MACHO_RODATA_KLUDGE
92 ; aout does not support align=
93 %macro SECTION_TEXT 0-1 16
94 %ifidn __OUTPUT_FORMAT__,aout
97 SECTION .text align=%1
103 %elif ARCH_X86_64 == 0
104 ; x86_32 doesn't require PIC.
105 ; Some distros prefer shared objects to be PIC, but nothing breaks if
106 ; the code contains a few textrels, so we'll skip that complexity.
113 ; Always use long nops (reduces 0x90 spam in disassembly on x86_32)
114 ; Not supported by NASM (except via smartalign package + ALIGNMODE k8,
115 ; however that fails when used together with the -M option)
120 ; Macros to eliminate most code duplication between x86_32 and x86_64:
121 ; Currently this works only for leaf functions which load all their arguments
122 ; into registers at the start, and make no other use of the stack. Luckily that
123 ; covers most of x264's asm.
126 ; %1 = number of arguments. loads them from stack if needed.
127 ; %2 = number of registers used. pushes callee-saved regs if needed.
128 ; %3 = number of xmm registers used. pushes callee-saved xmm regs if needed.
129 ; %4 = list of names to define to registers
130 ; PROLOGUE can also be invoked by adding the same options to cglobal
133 ; cglobal foo, 2,3,0, dst, src, tmp
134 ; declares a function (foo), taking two args (dst and src) and one local variable (tmp)
136 ; TODO Some functions can use some args directly from the stack. If they're the
137 ; last args then you can just not declare them, but if they're in the middle
138 ; we need more flexible macro.
141 ; Pops anything that was pushed by PROLOGUE, and returns.
144 ; Same, but if it doesn't pop anything it becomes a 2-byte ret, for athlons
145 ; which are slow when a normal ret follows a branch.
148 ; rN and rNq are the native-size register holding function argument N
149 ; rNd, rNw, rNb are dword, word, and byte size
150 ; rNh is the high 8 bits of the word size
151 ; rNm is the original location of arg N (a register or on the stack), dword
152 ; rNmp is native size
154 %macro DECLARE_REG 2-3
163 %elif ARCH_X86_64 ; memory
164 %define r%1m [rsp + stack_offset + %3]
165 %define r%1mp qword r %+ %1 %+ m
167 %define r%1m [esp + stack_offset + %3]
168 %define r%1mp dword r %+ %1 %+ m
173 %macro DECLARE_REG_SIZE 3
189 DECLARE_REG_SIZE ax, al, ah
190 DECLARE_REG_SIZE bx, bl, bh
191 DECLARE_REG_SIZE cx, cl, ch
192 DECLARE_REG_SIZE dx, dl, dh
193 DECLARE_REG_SIZE si, sil, null
194 DECLARE_REG_SIZE di, dil, null
195 DECLARE_REG_SIZE bp, bpl, null
197 ; t# defines for when per-arch register allocation is more complex than just function arguments
199 %macro DECLARE_REG_TMP 1-*
202 CAT_XDEFINE t, %%i, r%1
208 %macro DECLARE_REG_TMP_SIZE 0-*
210 %define t%1q t%1 %+ q
211 %define t%1d t%1 %+ d
212 %define t%1w t%1 %+ w
213 %define t%1h t%1 %+ h
214 %define t%1b t%1 %+ b
219 DECLARE_REG_TMP_SIZE 0,1,2,3,4,5,6,7,8,9,10,11,12,13,14
229 %assign stack_offset stack_offset+gprsize
234 %assign stack_offset stack_offset-gprsize
237 %macro PUSH_IF_USED 1-*
246 %macro POP_IF_USED 1-*
255 %macro LOAD_IF_USED 1-*
258 mov r%1, r %+ %1 %+ mp
267 %assign stack_offset stack_offset+(%2)
274 %assign stack_offset stack_offset-(%2)
284 %macro movsxdifnidn 2
296 %macro DEFINE_ARGS 0-*
300 CAT_UNDEF arg_name %+ %%i, q
301 CAT_UNDEF arg_name %+ %%i, d
302 CAT_UNDEF arg_name %+ %%i, w
303 CAT_UNDEF arg_name %+ %%i, h
304 CAT_UNDEF arg_name %+ %%i, b
305 CAT_UNDEF arg_name %+ %%i, m
306 CAT_UNDEF arg_name %+ %%i, mp
307 CAT_UNDEF arg_name, %%i
312 %xdefine %%stack_offset stack_offset
313 %undef stack_offset ; so that the current value of stack_offset doesn't get baked in by xdefine
316 %xdefine %1q r %+ %%i %+ q
317 %xdefine %1d r %+ %%i %+ d
318 %xdefine %1w r %+ %%i %+ w
319 %xdefine %1h r %+ %%i %+ h
320 %xdefine %1b r %+ %%i %+ b
321 %xdefine %1m r %+ %%i %+ m
322 %xdefine %1mp r %+ %%i %+ mp
323 CAT_XDEFINE arg_name, %%i, %1
327 %xdefine stack_offset %%stack_offset
328 %assign n_arg_names %0
331 %if WIN64 ; Windows x64 ;=================================================
337 DECLARE_REG 4, R10, 40
338 DECLARE_REG 5, R11, 48
339 DECLARE_REG 6, rax, 56
340 DECLARE_REG 7, rdi, 64
341 DECLARE_REG 8, rsi, 72
342 DECLARE_REG 9, rbx, 80
343 DECLARE_REG 10, rbp, 88
344 DECLARE_REG 11, R12, 96
345 DECLARE_REG 12, R13, 104
346 DECLARE_REG 13, R14, 112
347 DECLARE_REG 14, R15, 120
349 %macro PROLOGUE 2-4+ 0 ; #args, #regs, #xmm_regs, arg_names...
352 ASSERT regs_used >= num_args
353 ASSERT regs_used <= 15
354 PUSH_IF_USED 7, 8, 9, 10, 11, 12, 13, 14
356 %assign xmm_regs_used 0
360 LOAD_IF_USED 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14
364 %macro WIN64_SPILL_XMM 1
365 %assign xmm_regs_used %1
366 ASSERT xmm_regs_used <= 16
367 %if xmm_regs_used > 6
368 SUB rsp, (xmm_regs_used-6)*16+16
369 %assign %%i xmm_regs_used
370 %rep (xmm_regs_used-6)
372 movdqa [rsp + (%%i-6)*16+(~stack_offset&8)], xmm %+ %%i
377 %macro WIN64_RESTORE_XMM_INTERNAL 1
378 %if xmm_regs_used > 6
379 %assign %%i xmm_regs_used
380 %rep (xmm_regs_used-6)
382 movdqa xmm %+ %%i, [%1 + (%%i-6)*16+(~stack_offset&8)]
384 add %1, (xmm_regs_used-6)*16+16
388 %macro WIN64_RESTORE_XMM 1
389 WIN64_RESTORE_XMM_INTERNAL %1
390 %assign stack_offset stack_offset-(xmm_regs_used-6)*16+16
391 %assign xmm_regs_used 0
394 %define has_epilogue regs_used > 7 || xmm_regs_used > 6 || mmsize == 32
397 WIN64_RESTORE_XMM_INTERNAL rsp
398 POP_IF_USED 14, 13, 12, 11, 10, 9, 8, 7
405 %elif ARCH_X86_64 ; *nix x64 ;=============================================
413 DECLARE_REG 6, rax, 8
414 DECLARE_REG 7, R10, 16
415 DECLARE_REG 8, R11, 24
416 DECLARE_REG 9, rbx, 32
417 DECLARE_REG 10, rbp, 40
418 DECLARE_REG 11, R12, 48
419 DECLARE_REG 12, R13, 56
420 DECLARE_REG 13, R14, 64
421 DECLARE_REG 14, R15, 72
423 %macro PROLOGUE 2-4+ ; #args, #regs, #xmm_regs, arg_names...
426 ASSERT regs_used >= num_args
427 ASSERT regs_used <= 15
428 PUSH_IF_USED 9, 10, 11, 12, 13, 14
429 LOAD_IF_USED 6, 7, 8, 9, 10, 11, 12, 13, 14
433 %define has_epilogue regs_used > 9 || mmsize == 32
436 POP_IF_USED 14, 13, 12, 11, 10, 9
443 %else ; X86_32 ;==============================================================
445 DECLARE_REG 0, eax, 4
446 DECLARE_REG 1, ecx, 8
447 DECLARE_REG 2, edx, 12
448 DECLARE_REG 3, ebx, 16
449 DECLARE_REG 4, esi, 20
450 DECLARE_REG 5, edi, 24
451 DECLARE_REG 6, ebp, 28
454 %macro DECLARE_ARG 1-*
456 %define r%1m [esp + stack_offset + 4*%1 + 4]
457 %define r%1mp dword r%1m
462 DECLARE_ARG 7, 8, 9, 10, 11, 12, 13, 14
464 %macro PROLOGUE 2-4+ ; #args, #regs, #xmm_regs, arg_names...
473 ASSERT regs_used >= num_args
474 PUSH_IF_USED 3, 4, 5, 6
475 LOAD_IF_USED 0, 1, 2, 3, 4, 5, 6
479 %define has_epilogue regs_used > 3 || mmsize == 32
482 POP_IF_USED 6, 5, 4, 3
489 %endif ;======================================================================
492 %macro WIN64_SPILL_XMM 1
494 %macro WIN64_RESTORE_XMM 1
506 %macro TAIL_CALL 2 ; callee, is_nonadjacent
515 ;=============================================================================
516 ; arch-independent part
517 ;=============================================================================
519 %assign function_align 16
522 ; Applies any symbol mangling needed for C linkage, and sets up a define such that
523 ; subsequent uses of the function name automatically refer to the mangled version.
524 ; Appends cpuflags to the function name if cpuflags has been specified.
525 %macro cglobal 1-2+ ; name, [PROLOGUE args]
527 ; HACK: work around %+ broken with empty SUFFIX for nasm 2.09.10
531 cglobal_internal %1 %+ SUFFIX
534 ; HACK: work around %+ broken with empty SUFFIX for nasm 2.09.10
536 cglobal_internal %1, %2
538 cglobal_internal %1 %+ SUFFIX, %2
542 %macro cglobal_internal 1-2+
544 %xdefine %1 mangle(program_name %+ _ %+ %1)
545 %xdefine %1.skip_prologue %1 %+ .skip_prologue
546 CAT_XDEFINE cglobaled_, %1, 1
548 %xdefine current_function %1
549 %ifidn __OUTPUT_FORMAT__,elf
550 global %1:function hidden
556 RESET_MM_PERMUTATION ; not really needed, but makes disassembly somewhat nicer
557 %assign stack_offset 0
564 %xdefine %1 mangle(program_name %+ _ %+ %1)
565 CAT_XDEFINE cglobaled_, %1, 1
569 ; like cextern, but without the prefix
570 %macro cextern_naked 1
571 %xdefine %1 mangle(%1)
572 CAT_XDEFINE cglobaled_, %1, 1
577 %xdefine %1 mangle(program_name %+ _ %+ %1)
582 ; This is needed for ELF, otherwise the GNU linker assumes the stack is
583 ; executable by default.
584 %ifidn __OUTPUT_FORMAT__,elf
585 SECTION .note.GNU-stack noalloc noexec nowrite progbits
590 %assign cpuflags_mmx (1<<0)
591 %assign cpuflags_mmx2 (1<<1) | cpuflags_mmx
592 %assign cpuflags_3dnow (1<<2) | cpuflags_mmx
593 %assign cpuflags_3dnowext (1<<3) | cpuflags_3dnow
594 %assign cpuflags_sse (1<<4) | cpuflags_mmx2
595 %assign cpuflags_sse2 (1<<5) | cpuflags_sse
596 %assign cpuflags_sse2slow (1<<6) | cpuflags_sse2
597 %assign cpuflags_sse3 (1<<7) | cpuflags_sse2
598 %assign cpuflags_ssse3 (1<<8) | cpuflags_sse3
599 %assign cpuflags_sse4 (1<<9) | cpuflags_ssse3
600 %assign cpuflags_sse42 (1<<10)| cpuflags_sse4
601 %assign cpuflags_avx (1<<11)| cpuflags_sse42
602 %assign cpuflags_xop (1<<12)| cpuflags_avx
603 %assign cpuflags_fma4 (1<<13)| cpuflags_avx
604 %assign cpuflags_avx2 (1<<14)| cpuflags_avx
605 %assign cpuflags_fma3 (1<<15)| cpuflags_avx
607 %assign cpuflags_cache32 (1<<16)
608 %assign cpuflags_cache64 (1<<17)
609 %assign cpuflags_slowctz (1<<18)
610 %assign cpuflags_lzcnt (1<<19)
611 %assign cpuflags_misalign (1<<20)
612 %assign cpuflags_aligned (1<<21) ; not a cpu feature, but a function variant
613 %assign cpuflags_atom (1<<22)
614 %assign cpuflags_bmi1 (1<<23)
615 %assign cpuflags_bmi2 (1<<24)|cpuflags_bmi1
616 %assign cpuflags_tbm (1<<25)|cpuflags_bmi1
618 %define cpuflag(x) ((cpuflags & (cpuflags_ %+ x)) == (cpuflags_ %+ x))
619 %define notcpuflag(x) ((cpuflags & (cpuflags_ %+ x)) != (cpuflags_ %+ x))
621 ; Takes up to 2 cpuflags from the above list.
622 ; All subsequent functions (up to the next INIT_CPUFLAGS) is built for the specified cpu.
623 ; You shouldn't need to invoke this macro directly, it's a subroutine for INIT_MMX &co.
624 %macro INIT_CPUFLAGS 0-2
630 %assign cpuflags cpuflags_%1
632 %xdefine cpuname %1_%2
633 %assign cpuflags cpuflags | cpuflags_%2
635 %xdefine SUFFIX _ %+ cpuname
637 %assign avx_enabled 1
639 %if mmsize == 16 && notcpuflag(sse2)
642 %define movnta movntps
672 %assign avx_enabled 0
673 %define RESET_MM_PERMUTATION INIT_MMX %1
679 %define movnta movntq
682 CAT_XDEFINE m, %%i, mm %+ %%i
683 CAT_XDEFINE nmm, %%i, %%i
695 %assign avx_enabled 0
696 %define RESET_MM_PERMUTATION INIT_XMM %1
700 %define num_mmregs 16
705 %define movnta movntdq
708 CAT_XDEFINE m, %%i, xmm %+ %%i
709 CAT_XDEFINE nxmm, %%i, %%i
715 ; FIXME: INIT_AVX can be replaced by INIT_XMM avx
718 %assign avx_enabled 1
719 %define PALIGNR PALIGNR_SSSE3
720 %define RESET_MM_PERMUTATION INIT_AVX
724 %assign avx_enabled 1
725 %define RESET_MM_PERMUTATION INIT_YMM %1
729 %define num_mmregs 16
734 %define movnta vmovntps
737 CAT_XDEFINE m, %%i, ymm %+ %%i
738 CAT_XDEFINE nymm, %%i, %%i
746 ; I often want to use macros that permute their arguments. e.g. there's no
747 ; efficient way to implement butterfly or transpose or dct without swapping some
750 ; I would like to not have to manually keep track of the permutations:
751 ; If I insert a permutation in the middle of a function, it should automatically
752 ; change everything that follows. For more complex macros I may also have multiple
753 ; implementations, e.g. the SSE2 and SSSE3 versions may have different permutations.
755 ; Hence these macros. Insert a PERMUTE or some SWAPs at the end of a macro that
756 ; permutes its arguments. It's equivalent to exchanging the contents of the
757 ; registers, except that this way you exchange the register names instead, so it
758 ; doesn't cost any cycles.
760 %macro PERMUTE 2-* ; takes a list of pairs to swap
775 %macro SWAP 2-* ; swaps a single chain (sometimes more concise than pairs)
781 CAT_XDEFINE n, m%1, %1
782 CAT_XDEFINE n, m%2, %2
784 ; If we were called as "SWAP m0,m1" rather than "SWAP 0,1" infer the original numbers here.
785 ; Be careful using this mode in nested macros though, as in some cases there may be
786 ; other copies of m# that have already been dereferenced and don't get updated correctly.
787 %xdefine %%n1 n %+ %1
788 %xdefine %%n2 n %+ %2
789 %xdefine tmp m %+ %%n1
790 CAT_XDEFINE m, %%n1, m %+ %%n2
791 CAT_XDEFINE m, %%n2, tmp
792 CAT_XDEFINE n, m %+ %%n1, %%n1
793 CAT_XDEFINE n, m %+ %%n2, %%n2
800 ; If SAVE_MM_PERMUTATION is placed at the end of a function, then any later
801 ; calls to that function will automatically load the permutation, so values can
802 ; be returned in mmregs.
803 %macro SAVE_MM_PERMUTATION 0-1
807 %xdefine %%f current_function %+ _m
811 CAT_XDEFINE %%f, %%i, m %+ %%i
816 %macro LOAD_MM_PERMUTATION 1 ; name to load from
820 CAT_XDEFINE m, %%i, %1_m %+ %%i
821 CAT_XDEFINE n, m %+ %%i, %%i
827 ; Append cpuflags to the callee's name iff the appended name is known and the plain name isn't
829 ; HACK: work around %+ broken with empty SUFFIX for nasm 2.09.10
833 call_internal %1, %1 %+ SUFFIX
836 %macro call_internal 2
844 LOAD_MM_PERMUTATION %%i
847 ; Substitutions that reduce instruction size but are functionally equivalent
872 ;=============================================================================
873 ; AVX abstraction layer
874 ;=============================================================================
879 CAT_XDEFINE sizeofmm, i, 8
881 CAT_XDEFINE sizeofxmm, i, 16
882 CAT_XDEFINE sizeofymm, i, 32
887 %macro CHECK_AVX_INSTR_EMU 3-*
892 %error non-avx emulation of ``%%opcode'' is not supported
899 ;%2 == 1 if float, 0 if int
900 ;%3 == 1 if 4-operand (xmm, xmm, xmm, imm), 0 if 2- or 3-operand (xmm, xmm, xmm)
901 ;%4 == number of operands given
903 %macro RUN_AVX_INSTR 6-7+
905 %define %%sizeofreg sizeof%6
907 %define %%sizeofreg sizeof%5
909 %define %%sizeofreg mmsize
919 %define %%regmov movq
921 %define %%regmov movaps
923 %define %%regmov movdqa
928 %if avx_enabled && %%sizeofreg==16
931 CHECK_AVX_INSTR_EMU {%1 %5, %6, %7}, %5, %7
946 ; 3arg AVX ops with a memory arg can only have it in src2,
947 ; whereas SSE emulation of 3arg prefers to have it in src1 (i.e. the mov).
948 ; So, if the op is symmetric and the wrong one is memory, swap them.
949 %macro RUN_AVX_INSTR1 8
960 %if %%swap && %3 == 0 && %8 == 1
961 RUN_AVX_INSTR %1, %2, %3, %4, %5, %7, %6
963 RUN_AVX_INSTR %1, %2, %3, %4, %5, %6, %7
968 ;%2 == 1 if float, 0 if int
969 ;%3 == 1 if 4-operand (xmm, xmm, xmm, imm), 0 if 2- or 3-operand (xmm, xmm, xmm)
970 ;%4 == 1 if symmetric (i.e. doesn't matter which src arg is which), 0 if not
972 %macro %1 2-9 fnord, fnord, fnord, %1, %2, %3, %4
974 RUN_AVX_INSTR %6, %7, %8, 2, %1, %2
976 RUN_AVX_INSTR1 %6, %7, %8, 3, %1, %2, %3, %9
978 RUN_AVX_INSTR %6, %7, %8, 4, %1, %2, %3, %4
980 RUN_AVX_INSTR %6, %7, %8, 5, %1, %2, %3, %4, %5
985 AVX_INSTR addpd, 1, 0, 1
986 AVX_INSTR addps, 1, 0, 1
987 AVX_INSTR addsd, 1, 0, 1
988 AVX_INSTR addss, 1, 0, 1
989 AVX_INSTR addsubpd, 1, 0, 0
990 AVX_INSTR addsubps, 1, 0, 0
991 AVX_INSTR andpd, 1, 0, 1
992 AVX_INSTR andps, 1, 0, 1
993 AVX_INSTR andnpd, 1, 0, 0
994 AVX_INSTR andnps, 1, 0, 0
995 AVX_INSTR blendpd, 1, 0, 0
996 AVX_INSTR blendps, 1, 0, 0
997 AVX_INSTR blendvpd, 1, 0, 0
998 AVX_INSTR blendvps, 1, 0, 0
999 AVX_INSTR cmppd, 1, 0, 0
1000 AVX_INSTR cmpps, 1, 0, 0
1001 AVX_INSTR cmpsd, 1, 0, 0
1002 AVX_INSTR cmpss, 1, 0, 0
1003 AVX_INSTR cvtdq2ps, 1, 0, 0
1004 AVX_INSTR cvtps2dq, 1, 0, 0
1005 AVX_INSTR divpd, 1, 0, 0
1006 AVX_INSTR divps, 1, 0, 0
1007 AVX_INSTR divsd, 1, 0, 0
1008 AVX_INSTR divss, 1, 0, 0
1009 AVX_INSTR dppd, 1, 1, 0
1010 AVX_INSTR dpps, 1, 1, 0
1011 AVX_INSTR haddpd, 1, 0, 0
1012 AVX_INSTR haddps, 1, 0, 0
1013 AVX_INSTR hsubpd, 1, 0, 0
1014 AVX_INSTR hsubps, 1, 0, 0
1015 AVX_INSTR maxpd, 1, 0, 1
1016 AVX_INSTR maxps, 1, 0, 1
1017 AVX_INSTR maxsd, 1, 0, 1
1018 AVX_INSTR maxss, 1, 0, 1
1019 AVX_INSTR minpd, 1, 0, 1
1020 AVX_INSTR minps, 1, 0, 1
1021 AVX_INSTR minsd, 1, 0, 1
1022 AVX_INSTR minss, 1, 0, 1
1023 AVX_INSTR movhlps, 1, 0, 0
1024 AVX_INSTR movlhps, 1, 0, 0
1025 AVX_INSTR movsd, 1, 0, 0
1026 AVX_INSTR movss, 1, 0, 0
1027 AVX_INSTR mpsadbw, 0, 1, 0
1028 AVX_INSTR mulpd, 1, 0, 1
1029 AVX_INSTR mulps, 1, 0, 1
1030 AVX_INSTR mulsd, 1, 0, 1
1031 AVX_INSTR mulss, 1, 0, 1
1032 AVX_INSTR orpd, 1, 0, 1
1033 AVX_INSTR orps, 1, 0, 1
1034 AVX_INSTR pabsb, 0, 0, 0
1035 AVX_INSTR pabsw, 0, 0, 0
1036 AVX_INSTR pabsd, 0, 0, 0
1037 AVX_INSTR packsswb, 0, 0, 0
1038 AVX_INSTR packssdw, 0, 0, 0
1039 AVX_INSTR packuswb, 0, 0, 0
1040 AVX_INSTR packusdw, 0, 0, 0
1041 AVX_INSTR paddb, 0, 0, 1
1042 AVX_INSTR paddw, 0, 0, 1
1043 AVX_INSTR paddd, 0, 0, 1
1044 AVX_INSTR paddq, 0, 0, 1
1045 AVX_INSTR paddsb, 0, 0, 1
1046 AVX_INSTR paddsw, 0, 0, 1
1047 AVX_INSTR paddusb, 0, 0, 1
1048 AVX_INSTR paddusw, 0, 0, 1
1049 AVX_INSTR palignr, 0, 1, 0
1050 AVX_INSTR pand, 0, 0, 1
1051 AVX_INSTR pandn, 0, 0, 0
1052 AVX_INSTR pavgb, 0, 0, 1
1053 AVX_INSTR pavgw, 0, 0, 1
1054 AVX_INSTR pblendvb, 0, 0, 0
1055 AVX_INSTR pblendw, 0, 1, 0
1056 AVX_INSTR pcmpestri, 0, 0, 0
1057 AVX_INSTR pcmpestrm, 0, 0, 0
1058 AVX_INSTR pcmpistri, 0, 0, 0
1059 AVX_INSTR pcmpistrm, 0, 0, 0
1060 AVX_INSTR pcmpeqb, 0, 0, 1
1061 AVX_INSTR pcmpeqw, 0, 0, 1
1062 AVX_INSTR pcmpeqd, 0, 0, 1
1063 AVX_INSTR pcmpeqq, 0, 0, 1
1064 AVX_INSTR pcmpgtb, 0, 0, 0
1065 AVX_INSTR pcmpgtw, 0, 0, 0
1066 AVX_INSTR pcmpgtd, 0, 0, 0
1067 AVX_INSTR pcmpgtq, 0, 0, 0
1068 AVX_INSTR phaddw, 0, 0, 0
1069 AVX_INSTR phaddd, 0, 0, 0
1070 AVX_INSTR phaddsw, 0, 0, 0
1071 AVX_INSTR phsubw, 0, 0, 0
1072 AVX_INSTR phsubd, 0, 0, 0
1073 AVX_INSTR phsubsw, 0, 0, 0
1074 AVX_INSTR pmaddwd, 0, 0, 1
1075 AVX_INSTR pmaddubsw, 0, 0, 0
1076 AVX_INSTR pmaxsb, 0, 0, 1
1077 AVX_INSTR pmaxsw, 0, 0, 1
1078 AVX_INSTR pmaxsd, 0, 0, 1
1079 AVX_INSTR pmaxub, 0, 0, 1
1080 AVX_INSTR pmaxuw, 0, 0, 1
1081 AVX_INSTR pmaxud, 0, 0, 1
1082 AVX_INSTR pminsb, 0, 0, 1
1083 AVX_INSTR pminsw, 0, 0, 1
1084 AVX_INSTR pminsd, 0, 0, 1
1085 AVX_INSTR pminub, 0, 0, 1
1086 AVX_INSTR pminuw, 0, 0, 1
1087 AVX_INSTR pminud, 0, 0, 1
1088 AVX_INSTR pmovmskb, 0, 0, 0
1089 AVX_INSTR pmulhuw, 0, 0, 1
1090 AVX_INSTR pmulhrsw, 0, 0, 1
1091 AVX_INSTR pmulhw, 0, 0, 1
1092 AVX_INSTR pmullw, 0, 0, 1
1093 AVX_INSTR pmulld, 0, 0, 1
1094 AVX_INSTR pmuludq, 0, 0, 1
1095 AVX_INSTR pmuldq, 0, 0, 1
1096 AVX_INSTR por, 0, 0, 1
1097 AVX_INSTR psadbw, 0, 0, 1
1098 AVX_INSTR pshufb, 0, 0, 0
1099 AVX_INSTR pshufd, 0, 1, 0
1100 AVX_INSTR pshufhw, 0, 1, 0
1101 AVX_INSTR pshuflw, 0, 1, 0
1102 AVX_INSTR psignb, 0, 0, 0
1103 AVX_INSTR psignw, 0, 0, 0
1104 AVX_INSTR psignd, 0, 0, 0
1105 AVX_INSTR psllw, 0, 0, 0
1106 AVX_INSTR pslld, 0, 0, 0
1107 AVX_INSTR psllq, 0, 0, 0
1108 AVX_INSTR pslldq, 0, 0, 0
1109 AVX_INSTR psraw, 0, 0, 0
1110 AVX_INSTR psrad, 0, 0, 0
1111 AVX_INSTR psrlw, 0, 0, 0
1112 AVX_INSTR psrld, 0, 0, 0
1113 AVX_INSTR psrlq, 0, 0, 0
1114 AVX_INSTR psrldq, 0, 0, 0
1115 AVX_INSTR psubb, 0, 0, 0
1116 AVX_INSTR psubw, 0, 0, 0
1117 AVX_INSTR psubd, 0, 0, 0
1118 AVX_INSTR psubq, 0, 0, 0
1119 AVX_INSTR psubsb, 0, 0, 0
1120 AVX_INSTR psubsw, 0, 0, 0
1121 AVX_INSTR psubusb, 0, 0, 0
1122 AVX_INSTR psubusw, 0, 0, 0
1123 AVX_INSTR ptest, 0, 0, 0
1124 AVX_INSTR punpckhbw, 0, 0, 0
1125 AVX_INSTR punpckhwd, 0, 0, 0
1126 AVX_INSTR punpckhdq, 0, 0, 0
1127 AVX_INSTR punpckhqdq, 0, 0, 0
1128 AVX_INSTR punpcklbw, 0, 0, 0
1129 AVX_INSTR punpcklwd, 0, 0, 0
1130 AVX_INSTR punpckldq, 0, 0, 0
1131 AVX_INSTR punpcklqdq, 0, 0, 0
1132 AVX_INSTR pxor, 0, 0, 1
1133 AVX_INSTR shufps, 1, 1, 0
1134 AVX_INSTR subpd, 1, 0, 0
1135 AVX_INSTR subps, 1, 0, 0
1136 AVX_INSTR subsd, 1, 0, 0
1137 AVX_INSTR subss, 1, 0, 0
1138 AVX_INSTR unpckhpd, 1, 0, 0
1139 AVX_INSTR unpckhps, 1, 0, 0
1140 AVX_INSTR unpcklpd, 1, 0, 0
1141 AVX_INSTR unpcklps, 1, 0, 0
1142 AVX_INSTR xorpd, 1, 0, 1
1143 AVX_INSTR xorps, 1, 0, 1
1145 ; 3DNow instructions, for sharing code between AVX, SSE and 3DN
1146 AVX_INSTR pfadd, 1, 0, 1
1147 AVX_INSTR pfsub, 1, 0, 0
1148 AVX_INSTR pfmul, 1, 0, 1
1150 ; base-4 constants for shuffles
1153 %assign j ((i>>6)&3)*1000 + ((i>>4)&3)*100 + ((i>>2)&3)*10 + (i&3)
1155 CAT_XDEFINE q000, j, i
1157 CAT_XDEFINE q00, j, i
1159 CAT_XDEFINE q0, j, i
1169 %macro %1 5-8 %1, %2, %3
1170 %if cpuflag(xop) || cpuflag(fma4)
1184 FMA_INSTR fmaddps, mulps, addps
1185 FMA_INSTR pmacsdd, pmulld, paddd
1186 FMA_INSTR pmacsww, pmullw, paddw
1187 FMA_INSTR pmadcswd, pmaddwd, paddd
1189 ; tzcnt is equivalent to "rep bsf" and is backwards-compatible with bsf.
1190 ; This lets us use tzcnt without bumping the yasm version requirement yet.
1191 %define tzcnt rep bsf