generic-library/vpx
1
0
Fork 0
Code Issues Pull Requests Releases Wiki Activity
f86993f195
vpx/vp8/common/x86/sad_mmx.asm

428 lines
11 KiB
NASM
Raw Normal View History

Initial WebM release
2010-05-18 17:58:33 +02:00
;
Use WebM in copyright notice for consistency Changes 'The VP8 project' to 'The WebM project', for consistency with other webmproject.org repositories. Fixes issue #97. Change-Id: I37c13ed5fbdb9d334ceef71c6350e9febed9bbba
2010-09-09 14:16:39 +02:00
; Copyright (c) 2010 The WebM project authors. All Rights Reserved.
Initial WebM release
2010-05-18 17:58:33 +02:00
;
cosmetics: trim trailing whitespace When the license headers were updated, they accidentally contained trailing whitespace, so unfortunately we have to touch all the files again. Change-Id: I236c05fade06589e417179c0444cb39b09e4200d
2010-06-18 18:39:21 +02:00
; Use of this source code is governed by a BSD-style license
LICENSE: update with latest text Change-Id: Ieebea089095d9073b3a94932791099f614ce120c
2010-06-04 22:19:40 +02:00
; that can be found in the LICENSE file in the root of the source
; tree. An additional intellectual property rights grant can be found
cosmetics: trim trailing whitespace When the license headers were updated, they accidentally contained trailing whitespace, so unfortunately we have to touch all the files again. Change-Id: I236c05fade06589e417179c0444cb39b09e4200d
2010-06-18 18:39:21 +02:00
; in the file PATENTS. All contributing project authors may
LICENSE: update with latest text Change-Id: Ieebea089095d9073b3a94932791099f614ce120c
2010-06-04 22:19:40 +02:00
; be found in the AUTHORS file in the root of the source tree.
Initial WebM release
2010-05-18 17:58:33 +02:00
;
%include "vpx_ports/x86_abi_support.asm"
global sym(vp8_sad16x16_mmx)
global sym(vp8_sad8x16_mmx)
global sym(vp8_sad8x8_mmx)
global sym(vp8_sad4x4_mmx)
global sym(vp8_sad16x8_mmx)
;unsigned int vp8_sad16x16_mmx(
; unsigned char *src_ptr,
; int src_stride,
; unsigned char *ref_ptr,
; int ref_stride)
sym(vp8_sad16x16_mmx):
push rbp
mov rbp, rsp
SHADOW_ARGS_TO_STACK 4
push rsi
push rdi
; end prolog
mov rsi, arg(0) ;src_ptr
mov rdi, arg(2) ;ref_ptr
movsxd rax, dword ptr arg(1) ;src_stride
movsxd rdx, dword ptr arg(3) ;ref_stride
lea rcx, [rsi+rax*8]
lea rcx, [rcx+rax*8]
pxor mm7, mm7
pxor mm6, mm6
Use local labels for jumps/loops in x86 assembly. Prepend . to local labels in assembly code. This allows non unique labels within a file. Also makes profiling information more informative by keeping the function name with the loop name. Change-Id: I7a983cb3a5ba2413d5dafd0a37936b268fb9e37f
2011-08-23 00:29:41 +02:00
.x16x16sad_mmx_loop:
Initial WebM release
2010-05-18 17:58:33 +02:00
movq mm0, QWORD PTR [rsi]
movq mm2, QWORD PTR [rsi+8]
movq mm1, QWORD PTR [rdi]
movq mm3, QWORD PTR [rdi+8]
movq mm4, mm0
movq mm5, mm2
psubusb mm0, mm1
psubusb mm1, mm4
psubusb mm2, mm3
psubusb mm3, mm5
por mm0, mm1
por mm2, mm3
movq mm1, mm0
movq mm3, mm2
punpcklbw mm0, mm6
punpcklbw mm2, mm6
punpckhbw mm1, mm6
punpckhbw mm3, mm6
paddw mm0, mm2
paddw mm1, mm3
lea rsi, [rsi+rax]
add rdi, rdx
paddw mm7, mm0
paddw mm7, mm1
cmp rsi, rcx
Use local labels for jumps/loops in x86 assembly. Prepend . to local labels in assembly code. This allows non unique labels within a file. Also makes profiling information more informative by keeping the function name with the loop name. Change-Id: I7a983cb3a5ba2413d5dafd0a37936b268fb9e37f
2011-08-23 00:29:41 +02:00
jne .x16x16sad_mmx_loop
Initial WebM release
2010-05-18 17:58:33 +02:00
movq mm0, mm7
punpcklwd mm0, mm6
punpckhwd mm7, mm6
paddw mm0, mm7
movq mm7, mm0
psrlq mm0, 32
paddw mm7, mm0
nasm: match instruction length (movd/movq) to parameters nasm requires the instruction length (movd/movq) to match to its parameters. I find it more clear to really use 64bit instructions when we use 64bit registers in the assembly. Provide nasm compatibility. No binary change by this patch with yasm on {x86_64,i686}-fedora13-linux-gnu. Few longer opcodes with nasm on {x86_64,i686}-fedora13-linux-gnu have been checked as safe. Change-Id: Id9b1a5cdfb1bc05697e523c317a296df43d42a91
2010-10-04 23:19:33 +02:00
movq rax, mm7
Initial WebM release
2010-05-18 17:58:33 +02:00
pop rdi
pop rsi
mov rsp, rbp
; begin epilog
UNSHADOW_ARGS
pop rbp
ret
;unsigned int vp8_sad8x16_mmx(
; unsigned char *src_ptr,
; int src_stride,
; unsigned char *ref_ptr,
; int ref_stride)
sym(vp8_sad8x16_mmx):
push rbp
mov rbp, rsp
SHADOW_ARGS_TO_STACK 4
push rsi
push rdi
; end prolog
mov rsi, arg(0) ;src_ptr
mov rdi, arg(2) ;ref_ptr
movsxd rax, dword ptr arg(1) ;src_stride
movsxd rdx, dword ptr arg(3) ;ref_stride
lea rcx, [rsi+rax*8]
lea rcx, [rcx+rax*8]
pxor mm7, mm7
pxor mm6, mm6
Use local labels for jumps/loops in x86 assembly. Prepend . to local labels in assembly code. This allows non unique labels within a file. Also makes profiling information more informative by keeping the function name with the loop name. Change-Id: I7a983cb3a5ba2413d5dafd0a37936b268fb9e37f
2011-08-23 00:29:41 +02:00
.x8x16sad_mmx_loop:
Initial WebM release
2010-05-18 17:58:33 +02:00
movq mm0, QWORD PTR [rsi]
movq mm1, QWORD PTR [rdi]
movq mm2, mm0
psubusb mm0, mm1
psubusb mm1, mm2
por mm0, mm1
movq mm2, mm0
punpcklbw mm0, mm6
punpckhbw mm2, mm6
lea rsi, [rsi+rax]
add rdi, rdx
paddw mm7, mm0
paddw mm7, mm2
cmp rsi, rcx
Use local labels for jumps/loops in x86 assembly. Prepend . to local labels in assembly code. This allows non unique labels within a file. Also makes profiling information more informative by keeping the function name with the loop name. Change-Id: I7a983cb3a5ba2413d5dafd0a37936b268fb9e37f
2011-08-23 00:29:41 +02:00
jne .x8x16sad_mmx_loop
Initial WebM release
2010-05-18 17:58:33 +02:00
movq mm0, mm7
punpcklwd mm0, mm6
punpckhwd mm7, mm6
paddw mm0, mm7
movq mm7, mm0
psrlq mm0, 32
paddw mm7, mm0
nasm: match instruction length (movd/movq) to parameters nasm requires the instruction length (movd/movq) to match to its parameters. I find it more clear to really use 64bit instructions when we use 64bit registers in the assembly. Provide nasm compatibility. No binary change by this patch with yasm on {x86_64,i686}-fedora13-linux-gnu. Few longer opcodes with nasm on {x86_64,i686}-fedora13-linux-gnu have been checked as safe. Change-Id: Id9b1a5cdfb1bc05697e523c317a296df43d42a91
2010-10-04 23:19:33 +02:00
movq rax, mm7
Initial WebM release
2010-05-18 17:58:33 +02:00
pop rdi
pop rsi
mov rsp, rbp
; begin epilog
UNSHADOW_ARGS
pop rbp
ret
;unsigned int vp8_sad8x8_mmx(
; unsigned char *src_ptr,
; int src_stride,
; unsigned char *ref_ptr,
; int ref_stride)
sym(vp8_sad8x8_mmx):
push rbp
mov rbp, rsp
SHADOW_ARGS_TO_STACK 4
push rsi
push rdi
; end prolog
mov rsi, arg(0) ;src_ptr
mov rdi, arg(2) ;ref_ptr
movsxd rax, dword ptr arg(1) ;src_stride
movsxd rdx, dword ptr arg(3) ;ref_stride
lea rcx, [rsi+rax*8]
pxor mm7, mm7
pxor mm6, mm6
Use local labels for jumps/loops in x86 assembly. Prepend . to local labels in assembly code. This allows non unique labels within a file. Also makes profiling information more informative by keeping the function name with the loop name. Change-Id: I7a983cb3a5ba2413d5dafd0a37936b268fb9e37f
2011-08-23 00:29:41 +02:00
.x8x8sad_mmx_loop:
Initial WebM release
2010-05-18 17:58:33 +02:00
movq mm0, QWORD PTR [rsi]
movq mm1, QWORD PTR [rdi]
movq mm2, mm0
psubusb mm0, mm1
psubusb mm1, mm2
por mm0, mm1
movq mm2, mm0
punpcklbw mm0, mm6
punpckhbw mm2, mm6
paddw mm0, mm2
lea rsi, [rsi+rax]
add rdi, rdx
paddw mm7, mm0
cmp rsi, rcx
Use local labels for jumps/loops in x86 assembly. Prepend . to local labels in assembly code. This allows non unique labels within a file. Also makes profiling information more informative by keeping the function name with the loop name. Change-Id: I7a983cb3a5ba2413d5dafd0a37936b268fb9e37f
2011-08-23 00:29:41 +02:00
jne .x8x8sad_mmx_loop
Initial WebM release
2010-05-18 17:58:33 +02:00
movq mm0, mm7
punpcklwd mm0, mm6
punpckhwd mm7, mm6
paddw mm0, mm7
movq mm7, mm0
psrlq mm0, 32
paddw mm7, mm0
nasm: match instruction length (movd/movq) to parameters nasm requires the instruction length (movd/movq) to match to its parameters. I find it more clear to really use 64bit instructions when we use 64bit registers in the assembly. Provide nasm compatibility. No binary change by this patch with yasm on {x86_64,i686}-fedora13-linux-gnu. Few longer opcodes with nasm on {x86_64,i686}-fedora13-linux-gnu have been checked as safe. Change-Id: Id9b1a5cdfb1bc05697e523c317a296df43d42a91
2010-10-04 23:19:33 +02:00
movq rax, mm7
Initial WebM release
2010-05-18 17:58:33 +02:00
pop rdi
pop rsi
mov rsp, rbp
; begin epilog
UNSHADOW_ARGS
pop rbp
ret
;unsigned int vp8_sad4x4_mmx(
; unsigned char *src_ptr,
; int src_stride,
; unsigned char *ref_ptr,
; int ref_stride)
sym(vp8_sad4x4_mmx):
push rbp
mov rbp, rsp
SHADOW_ARGS_TO_STACK 4
push rsi
push rdi
; end prolog
mov rsi, arg(0) ;src_ptr
mov rdi, arg(2) ;ref_ptr
movsxd rax, dword ptr arg(1) ;src_stride
movsxd rdx, dword ptr arg(3) ;ref_stride
Correct QWORD usage in assembly files QWORD was being undefined because it was being used incorrectly. Change-Id: I3610cefa3d6f0da4054316760f78b9694cde3876
2010-10-14 01:57:57 +02:00
movd mm0, DWORD PTR [rsi]
movd mm1, DWORD PTR [rdi]
Initial WebM release
2010-05-18 17:58:33 +02:00
Correct QWORD usage in assembly files QWORD was being undefined because it was being used incorrectly. Change-Id: I3610cefa3d6f0da4054316760f78b9694cde3876
2010-10-14 01:57:57 +02:00
movd mm2, DWORD PTR [rsi+rax]
movd mm3, DWORD PTR [rdi+rdx]
Initial WebM release
2010-05-18 17:58:33 +02:00
punpcklbw mm0, mm2
punpcklbw mm1, mm3
movq mm2, mm0
psubusb mm0, mm1
psubusb mm1, mm2
por mm0, mm1
movq mm2, mm0
pxor mm3, mm3
punpcklbw mm0, mm3
punpckhbw mm2, mm3
paddw mm0, mm2
lea rsi, [rsi+rax*2]
lea rdi, [rdi+rdx*2]
Correct QWORD usage in assembly files QWORD was being undefined because it was being used incorrectly. Change-Id: I3610cefa3d6f0da4054316760f78b9694cde3876
2010-10-14 01:57:57 +02:00
movd mm4, DWORD PTR [rsi]
movd mm5, DWORD PTR [rdi]
Initial WebM release
2010-05-18 17:58:33 +02:00
Correct QWORD usage in assembly files QWORD was being undefined because it was being used incorrectly. Change-Id: I3610cefa3d6f0da4054316760f78b9694cde3876
2010-10-14 01:57:57 +02:00
movd mm6, DWORD PTR [rsi+rax]
movd mm7, DWORD PTR [rdi+rdx]
Initial WebM release
2010-05-18 17:58:33 +02:00
punpcklbw mm4, mm6
punpcklbw mm5, mm7
movq mm6, mm4
psubusb mm4, mm5
psubusb mm5, mm6
por mm4, mm5
movq mm5, mm4
punpcklbw mm4, mm3
punpckhbw mm5, mm3
paddw mm4, mm5
paddw mm0, mm4
movq mm1, mm0
punpcklwd mm0, mm3
punpckhwd mm1, mm3
paddw mm0, mm1
movq mm1, mm0
psrlq mm0, 32
paddw mm0, mm1
nasm: match instruction length (movd/movq) to parameters nasm requires the instruction length (movd/movq) to match to its parameters. I find it more clear to really use 64bit instructions when we use 64bit registers in the assembly. Provide nasm compatibility. No binary change by this patch with yasm on {x86_64,i686}-fedora13-linux-gnu. Few longer opcodes with nasm on {x86_64,i686}-fedora13-linux-gnu have been checked as safe. Change-Id: Id9b1a5cdfb1bc05697e523c317a296df43d42a91
2010-10-04 23:19:33 +02:00
movq rax, mm0
Initial WebM release
2010-05-18 17:58:33 +02:00
pop rdi
pop rsi
mov rsp, rbp
; begin epilog
UNSHADOW_ARGS
pop rbp
ret
;unsigned int vp8_sad16x8_mmx(
; unsigned char *src_ptr,
; int src_stride,
; unsigned char *ref_ptr,
; int ref_stride)
sym(vp8_sad16x8_mmx):
push rbp
mov rbp, rsp
SHADOW_ARGS_TO_STACK 4
push rsi
push rdi
; end prolog
mov rsi, arg(0) ;src_ptr
mov rdi, arg(2) ;ref_ptr
movsxd rax, dword ptr arg(1) ;src_stride
movsxd rdx, dword ptr arg(3) ;ref_stride
lea rcx, [rsi+rax*8]
pxor mm7, mm7
pxor mm6, mm6
Use local labels for jumps/loops in x86 assembly. Prepend . to local labels in assembly code. This allows non unique labels within a file. Also makes profiling information more informative by keeping the function name with the loop name. Change-Id: I7a983cb3a5ba2413d5dafd0a37936b268fb9e37f
2011-08-23 00:29:41 +02:00
.x16x8sad_mmx_loop:
Initial WebM release
2010-05-18 17:58:33 +02:00
movq mm0, [rsi]
movq mm1, [rdi]
movq mm2, [rsi+8]
movq mm3, [rdi+8]
movq mm4, mm0
movq mm5, mm2
psubusb mm0, mm1
psubusb mm1, mm4
psubusb mm2, mm3
psubusb mm3, mm5
por mm0, mm1
por mm2, mm3
movq mm1, mm0
movq mm3, mm2
punpcklbw mm0, mm6
punpckhbw mm1, mm6
punpcklbw mm2, mm6
punpckhbw mm3, mm6
paddw mm0, mm2
paddw mm1, mm3
paddw mm0, mm1
lea rsi, [rsi+rax]
add rdi, rdx
paddw mm7, mm0
cmp rsi, rcx
Use local labels for jumps/loops in x86 assembly. Prepend . to local labels in assembly code. This allows non unique labels within a file. Also makes profiling information more informative by keeping the function name with the loop name. Change-Id: I7a983cb3a5ba2413d5dafd0a37936b268fb9e37f
2011-08-23 00:29:41 +02:00
jne .x16x8sad_mmx_loop
Initial WebM release
2010-05-18 17:58:33 +02:00
movq mm0, mm7
punpcklwd mm0, mm6
punpckhwd mm7, mm6
paddw mm0, mm7
movq mm7, mm0
psrlq mm0, 32
paddw mm7, mm0
nasm: match instruction length (movd/movq) to parameters nasm requires the instruction length (movd/movq) to match to its parameters. I find it more clear to really use 64bit instructions when we use 64bit registers in the assembly. Provide nasm compatibility. No binary change by this patch with yasm on {x86_64,i686}-fedora13-linux-gnu. Few longer opcodes with nasm on {x86_64,i686}-fedora13-linux-gnu have been checked as safe. Change-Id: Id9b1a5cdfb1bc05697e523c317a296df43d42a91
2010-10-04 23:19:33 +02:00
movq rax, mm7
Initial WebM release
2010-05-18 17:58:33 +02:00
pop rdi
pop rsi
mov rsp, rbp
; begin epilog
UNSHADOW_ARGS
pop rbp
ret
Reference in New Issue Copy Permalink