generic-library/vpx
1
0
Fork 0
Code Issues Pull Requests Releases Wiki Activity
bd5a5bb561
vpx/vpx_dsp/x86/sad_ssse3.asm

371 lines
13 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"
%macro PROCESS_16X2X3 1
%if %1
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
movdqa xmm0, XMMWORD PTR [rsi]
lddqu xmm5, XMMWORD PTR [rdi]
lddqu xmm6, XMMWORD PTR [rdi+1]
lddqu xmm7, XMMWORD PTR [rdi+2]
Initial WebM release
2010-05-18 17:58:33 +02:00
psadbw xmm5, xmm0
psadbw xmm6, xmm0
psadbw xmm7, xmm0
%else
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
movdqa xmm0, XMMWORD PTR [rsi]
lddqu xmm1, XMMWORD PTR [rdi]
lddqu xmm2, XMMWORD PTR [rdi+1]
lddqu xmm3, XMMWORD PTR [rdi+2]
Initial WebM release
2010-05-18 17:58:33 +02:00
psadbw xmm1, xmm0
psadbw xmm2, xmm0
psadbw xmm3, xmm0
paddw xmm5, xmm1
paddw xmm6, xmm2
paddw xmm7, xmm3
%endif
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
movdqa xmm0, XMMWORD PTR [rsi+rax]
lddqu xmm1, XMMWORD PTR [rdi+rdx]
lddqu xmm2, XMMWORD PTR [rdi+rdx+1]
lddqu xmm3, XMMWORD PTR [rdi+rdx+2]
Initial WebM release
2010-05-18 17:58:33 +02:00
lea rsi, [rsi+rax*2]
lea rdi, [rdi+rdx*2]
psadbw xmm1, xmm0
psadbw xmm2, xmm0
psadbw xmm3, xmm0
paddw xmm5, xmm1
paddw xmm6, xmm2
paddw xmm7, xmm3
%endmacro
%macro PROCESS_16X2X3_OFFSET 2
%if %1
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
movdqa xmm0, XMMWORD PTR [rsi]
movdqa xmm4, XMMWORD PTR [rdi]
movdqa xmm7, XMMWORD PTR [rdi+16]
Initial WebM release
2010-05-18 17:58:33 +02:00
movdqa xmm5, xmm7
palignr xmm5, xmm4, %2
movdqa xmm6, xmm7
palignr xmm6, xmm4, (%2+1)
palignr xmm7, xmm4, (%2+2)
psadbw xmm5, xmm0
psadbw xmm6, xmm0
psadbw xmm7, xmm0
%else
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
movdqa xmm0, XMMWORD PTR [rsi]
movdqa xmm4, XMMWORD PTR [rdi]
movdqa xmm3, XMMWORD PTR [rdi+16]
Initial WebM release
2010-05-18 17:58:33 +02:00
movdqa xmm1, xmm3
palignr xmm1, xmm4, %2
movdqa xmm2, xmm3
palignr xmm2, xmm4, (%2+1)
palignr xmm3, xmm4, (%2+2)
psadbw xmm1, xmm0
psadbw xmm2, xmm0
psadbw xmm3, xmm0
paddw xmm5, xmm1
paddw xmm6, xmm2
paddw xmm7, xmm3
%endif
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
movdqa xmm0, XMMWORD PTR [rsi+rax]
movdqa xmm4, XMMWORD PTR [rdi+rdx]
movdqa xmm3, XMMWORD PTR [rdi+rdx+16]
Initial WebM release
2010-05-18 17:58:33 +02:00
movdqa xmm1, xmm3
palignr xmm1, xmm4, %2
movdqa xmm2, xmm3
palignr xmm2, xmm4, (%2+1)
palignr xmm3, xmm4, (%2+2)
lea rsi, [rsi+rax*2]
lea rdi, [rdi+rdx*2]
psadbw xmm1, xmm0
psadbw xmm2, xmm0
psadbw xmm3, xmm0
paddw xmm5, xmm1
paddw xmm6, xmm2
paddw xmm7, xmm3
%endmacro
%macro PROCESS_16X16X3_OFFSET 2
%2_aligned_by_%1:
sub rdi, %1
PROCESS_16X2X3_OFFSET 1, %1
PROCESS_16X2X3_OFFSET 0, %1
PROCESS_16X2X3_OFFSET 0, %1
PROCESS_16X2X3_OFFSET 0, %1
PROCESS_16X2X3_OFFSET 0, %1
PROCESS_16X2X3_OFFSET 0, %1
PROCESS_16X2X3_OFFSET 0, %1
PROCESS_16X2X3_OFFSET 0, %1
jmp %2_store_off
%endmacro
%macro PROCESS_16X8X3_OFFSET 2
%2_aligned_by_%1:
sub rdi, %1
PROCESS_16X2X3_OFFSET 1, %1
PROCESS_16X2X3_OFFSET 0, %1
PROCESS_16X2X3_OFFSET 0, %1
PROCESS_16X2X3_OFFSET 0, %1
jmp %2_store_off
%endmacro
Move shared SAD code to vpx_dsp Create a new component, vpx_dsp, for code that can be shared between codecs. Move the SAD code into the component. This reduces the size of vpxenc/dec by 36k on x86_64 builds. Change-Id: I73f837ddaecac6b350bf757af0cfe19c4ab9327a
2015-04-17 22:11:38 +02:00
;void int vpx_sad16x16x3_ssse3(
Initial WebM release
2010-05-18 17:58:33 +02:00
; unsigned char *src_ptr,
; int src_stride,
; unsigned char *ref_ptr,
; int ref_stride,
; int *results)
Move shared SAD code to vpx_dsp Create a new component, vpx_dsp, for code that can be shared between codecs. Move the SAD code into the component. This reduces the size of vpxenc/dec by 36k on x86_64 builds. Change-Id: I73f837ddaecac6b350bf757af0cfe19c4ab9327a
2015-04-17 22:11:38 +02:00
global sym(vpx_sad16x16x3_ssse3) PRIVATE
sym(vpx_sad16x16x3_ssse3):
Initial WebM release
2010-05-18 17:58:33 +02:00
push rbp
mov rbp, rsp
SHADOW_ARGS_TO_STACK 5
modify SAVE_XMM for potential 64bit use the win64 abi requires saving and restoring xmm6:xmm15. currently SAVE_XMM and RESTORE XMM only allow for saving xmm6:xmm7. allow specifying the highest register used and if the stack is unaligned. Change-Id: Ica5699622ffe3346d3a486f48eef0206c51cf867
2011-04-15 16:05:20 +02:00
SAVE_XMM 7
Initial WebM release
2010-05-18 17:58:33 +02:00
push rsi
push rdi
push rcx
; end prolog
mov rsi, arg(0) ;src_ptr
mov rdi, arg(2) ;ref_ptr
mov rdx, 0xf
and rdx, rdi
Move shared SAD code to vpx_dsp Create a new component, vpx_dsp, for code that can be shared between codecs. Move the SAD code into the component. This reduces the size of vpxenc/dec by 36k on x86_64 builds. Change-Id: I73f837ddaecac6b350bf757af0cfe19c4ab9327a
2015-04-17 22:11:38 +02:00
jmp .vpx_sad16x16x3_ssse3_skiptable
.vpx_sad16x16x3_ssse3_jumptable:
dd .vpx_sad16x16x3_ssse3_aligned_by_0 - .vpx_sad16x16x3_ssse3_do_jump
dd .vpx_sad16x16x3_ssse3_aligned_by_1 - .vpx_sad16x16x3_ssse3_do_jump
dd .vpx_sad16x16x3_ssse3_aligned_by_2 - .vpx_sad16x16x3_ssse3_do_jump
dd .vpx_sad16x16x3_ssse3_aligned_by_3 - .vpx_sad16x16x3_ssse3_do_jump
dd .vpx_sad16x16x3_ssse3_aligned_by_4 - .vpx_sad16x16x3_ssse3_do_jump
dd .vpx_sad16x16x3_ssse3_aligned_by_5 - .vpx_sad16x16x3_ssse3_do_jump
dd .vpx_sad16x16x3_ssse3_aligned_by_6 - .vpx_sad16x16x3_ssse3_do_jump
dd .vpx_sad16x16x3_ssse3_aligned_by_7 - .vpx_sad16x16x3_ssse3_do_jump
dd .vpx_sad16x16x3_ssse3_aligned_by_8 - .vpx_sad16x16x3_ssse3_do_jump
dd .vpx_sad16x16x3_ssse3_aligned_by_9 - .vpx_sad16x16x3_ssse3_do_jump
dd .vpx_sad16x16x3_ssse3_aligned_by_10 - .vpx_sad16x16x3_ssse3_do_jump
dd .vpx_sad16x16x3_ssse3_aligned_by_11 - .vpx_sad16x16x3_ssse3_do_jump
dd .vpx_sad16x16x3_ssse3_aligned_by_12 - .vpx_sad16x16x3_ssse3_do_jump
dd .vpx_sad16x16x3_ssse3_aligned_by_13 - .vpx_sad16x16x3_ssse3_do_jump
dd .vpx_sad16x16x3_ssse3_aligned_by_14 - .vpx_sad16x16x3_ssse3_do_jump
dd .vpx_sad16x16x3_ssse3_aligned_by_15 - .vpx_sad16x16x3_ssse3_do_jump
.vpx_sad16x16x3_ssse3_skiptable:
call .vpx_sad16x16x3_ssse3_do_jump
.vpx_sad16x16x3_ssse3_do_jump:
Initial WebM release
2010-05-18 17:58:33 +02:00
pop rcx ; get the address of do_jump
Move shared SAD code to vpx_dsp Create a new component, vpx_dsp, for code that can be shared between codecs. Move the SAD code into the component. This reduces the size of vpxenc/dec by 36k on x86_64 builds. Change-Id: I73f837ddaecac6b350bf757af0cfe19c4ab9327a
2015-04-17 22:11:38 +02:00
mov rax, .vpx_sad16x16x3_ssse3_jumptable - .vpx_sad16x16x3_ssse3_do_jump
add rax, rcx ; get the absolute address of vpx_sad16x16x3_ssse3_jumptable
Initial WebM release
2010-05-18 17:58:33 +02:00
movsxd rax, dword [rax + 4*rdx] ; get the 32 bit offset from the jumptable
add rcx, rax
movsxd rax, dword ptr arg(1) ;src_stride
movsxd rdx, dword ptr arg(3) ;ref_stride
jmp rcx
Move shared SAD code to vpx_dsp Create a new component, vpx_dsp, for code that can be shared between codecs. Move the SAD code into the component. This reduces the size of vpxenc/dec by 36k on x86_64 builds. Change-Id: I73f837ddaecac6b350bf757af0cfe19c4ab9327a
2015-04-17 22:11:38 +02:00
PROCESS_16X16X3_OFFSET 0, .vpx_sad16x16x3_ssse3
PROCESS_16X16X3_OFFSET 1, .vpx_sad16x16x3_ssse3
PROCESS_16X16X3_OFFSET 2, .vpx_sad16x16x3_ssse3
PROCESS_16X16X3_OFFSET 3, .vpx_sad16x16x3_ssse3
PROCESS_16X16X3_OFFSET 4, .vpx_sad16x16x3_ssse3
PROCESS_16X16X3_OFFSET 5, .vpx_sad16x16x3_ssse3
PROCESS_16X16X3_OFFSET 6, .vpx_sad16x16x3_ssse3
PROCESS_16X16X3_OFFSET 7, .vpx_sad16x16x3_ssse3
PROCESS_16X16X3_OFFSET 8, .vpx_sad16x16x3_ssse3
PROCESS_16X16X3_OFFSET 9, .vpx_sad16x16x3_ssse3
PROCESS_16X16X3_OFFSET 10, .vpx_sad16x16x3_ssse3
PROCESS_16X16X3_OFFSET 11, .vpx_sad16x16x3_ssse3
PROCESS_16X16X3_OFFSET 12, .vpx_sad16x16x3_ssse3
PROCESS_16X16X3_OFFSET 13, .vpx_sad16x16x3_ssse3
PROCESS_16X16X3_OFFSET 14, .vpx_sad16x16x3_ssse3
.vpx_sad16x16x3_ssse3_aligned_by_15:
Initial WebM release
2010-05-18 17:58:33 +02:00
PROCESS_16X2X3 1
PROCESS_16X2X3 0
PROCESS_16X2X3 0
PROCESS_16X2X3 0
PROCESS_16X2X3 0
PROCESS_16X2X3 0
PROCESS_16X2X3 0
PROCESS_16X2X3 0
Move shared SAD code to vpx_dsp Create a new component, vpx_dsp, for code that can be shared between codecs. Move the SAD code into the component. This reduces the size of vpxenc/dec by 36k on x86_64 builds. Change-Id: I73f837ddaecac6b350bf757af0cfe19c4ab9327a
2015-04-17 22:11:38 +02:00
.vpx_sad16x16x3_ssse3_store_off:
Initial WebM release
2010-05-18 17:58:33 +02:00
mov rdi, arg(4) ;Results
movq xmm0, xmm5
psrldq xmm5, 8
paddw xmm0, xmm5
movd [rdi], xmm0
;-
movq xmm0, xmm6
psrldq xmm6, 8
paddw xmm0, xmm6
movd [rdi+4], xmm0
;-
movq xmm0, xmm7
psrldq xmm7, 8
paddw xmm0, xmm7
movd [rdi+8], xmm0
; begin epilog
pop rcx
pop rdi
pop rsi
Add save/restore xmm registers in x86 assembly code Went through the code and fixed it. Verified on Windows. Where possible, remove dependencies on xmm[67] Current code relies on pushing rbp to the stack to get 16 byte alignment. This broke when rbp wasn't pushed (vp8/encoder/x86/sad_sse3.asm). Work around this by using unaligned memory accesses. Revisit this and the offsets in vp8/encoder/x86/sad_sse3.asm in another change to SAVE_XMM. Change-Id: I5f940994d3ebfd977c3d68446cef20fd78b07877
2011-04-07 19:17:22 +02:00
RESTORE_XMM
Initial WebM release
2010-05-18 17:58:33 +02:00
UNSHADOW_ARGS
pop rbp
ret
Move shared SAD code to vpx_dsp Create a new component, vpx_dsp, for code that can be shared between codecs. Move the SAD code into the component. This reduces the size of vpxenc/dec by 36k on x86_64 builds. Change-Id: I73f837ddaecac6b350bf757af0cfe19c4ab9327a
2015-04-17 22:11:38 +02:00
;void int vpx_sad16x8x3_ssse3(
Initial WebM release
2010-05-18 17:58:33 +02:00
; unsigned char *src_ptr,
; int src_stride,
; unsigned char *ref_ptr,
; int ref_stride,
; int *results)
Move shared SAD code to vpx_dsp Create a new component, vpx_dsp, for code that can be shared between codecs. Move the SAD code into the component. This reduces the size of vpxenc/dec by 36k on x86_64 builds. Change-Id: I73f837ddaecac6b350bf757af0cfe19c4ab9327a
2015-04-17 22:11:38 +02:00
global sym(vpx_sad16x8x3_ssse3) PRIVATE
sym(vpx_sad16x8x3_ssse3):
Initial WebM release
2010-05-18 17:58:33 +02:00
push rbp
mov rbp, rsp
SHADOW_ARGS_TO_STACK 5
modify SAVE_XMM for potential 64bit use the win64 abi requires saving and restoring xmm6:xmm15. currently SAVE_XMM and RESTORE XMM only allow for saving xmm6:xmm7. allow specifying the highest register used and if the stack is unaligned. Change-Id: Ica5699622ffe3346d3a486f48eef0206c51cf867
2011-04-15 16:05:20 +02:00
SAVE_XMM 7
Initial WebM release
2010-05-18 17:58:33 +02:00
push rsi
push rdi
push rcx
; end prolog
mov rsi, arg(0) ;src_ptr
mov rdi, arg(2) ;ref_ptr
mov rdx, 0xf
and rdx, rdi
Move shared SAD code to vpx_dsp Create a new component, vpx_dsp, for code that can be shared between codecs. Move the SAD code into the component. This reduces the size of vpxenc/dec by 36k on x86_64 builds. Change-Id: I73f837ddaecac6b350bf757af0cfe19c4ab9327a
2015-04-17 22:11:38 +02:00
jmp .vpx_sad16x8x3_ssse3_skiptable
.vpx_sad16x8x3_ssse3_jumptable:
dd .vpx_sad16x8x3_ssse3_aligned_by_0 - .vpx_sad16x8x3_ssse3_do_jump
dd .vpx_sad16x8x3_ssse3_aligned_by_1 - .vpx_sad16x8x3_ssse3_do_jump
dd .vpx_sad16x8x3_ssse3_aligned_by_2 - .vpx_sad16x8x3_ssse3_do_jump
dd .vpx_sad16x8x3_ssse3_aligned_by_3 - .vpx_sad16x8x3_ssse3_do_jump
dd .vpx_sad16x8x3_ssse3_aligned_by_4 - .vpx_sad16x8x3_ssse3_do_jump
dd .vpx_sad16x8x3_ssse3_aligned_by_5 - .vpx_sad16x8x3_ssse3_do_jump
dd .vpx_sad16x8x3_ssse3_aligned_by_6 - .vpx_sad16x8x3_ssse3_do_jump
dd .vpx_sad16x8x3_ssse3_aligned_by_7 - .vpx_sad16x8x3_ssse3_do_jump
dd .vpx_sad16x8x3_ssse3_aligned_by_8 - .vpx_sad16x8x3_ssse3_do_jump
dd .vpx_sad16x8x3_ssse3_aligned_by_9 - .vpx_sad16x8x3_ssse3_do_jump
dd .vpx_sad16x8x3_ssse3_aligned_by_10 - .vpx_sad16x8x3_ssse3_do_jump
dd .vpx_sad16x8x3_ssse3_aligned_by_11 - .vpx_sad16x8x3_ssse3_do_jump
dd .vpx_sad16x8x3_ssse3_aligned_by_12 - .vpx_sad16x8x3_ssse3_do_jump
dd .vpx_sad16x8x3_ssse3_aligned_by_13 - .vpx_sad16x8x3_ssse3_do_jump
dd .vpx_sad16x8x3_ssse3_aligned_by_14 - .vpx_sad16x8x3_ssse3_do_jump
dd .vpx_sad16x8x3_ssse3_aligned_by_15 - .vpx_sad16x8x3_ssse3_do_jump
.vpx_sad16x8x3_ssse3_skiptable:
call .vpx_sad16x8x3_ssse3_do_jump
.vpx_sad16x8x3_ssse3_do_jump:
Initial WebM release
2010-05-18 17:58:33 +02:00
pop rcx ; get the address of do_jump
Move shared SAD code to vpx_dsp Create a new component, vpx_dsp, for code that can be shared between codecs. Move the SAD code into the component. This reduces the size of vpxenc/dec by 36k on x86_64 builds. Change-Id: I73f837ddaecac6b350bf757af0cfe19c4ab9327a
2015-04-17 22:11:38 +02:00
mov rax, .vpx_sad16x8x3_ssse3_jumptable - .vpx_sad16x8x3_ssse3_do_jump
add rax, rcx ; get the absolute address of vpx_sad16x8x3_ssse3_jumptable
Initial WebM release
2010-05-18 17:58:33 +02:00
movsxd rax, dword [rax + 4*rdx] ; get the 32 bit offset from the jumptable
add rcx, rax
movsxd rax, dword ptr arg(1) ;src_stride
movsxd rdx, dword ptr arg(3) ;ref_stride
jmp rcx
Move shared SAD code to vpx_dsp Create a new component, vpx_dsp, for code that can be shared between codecs. Move the SAD code into the component. This reduces the size of vpxenc/dec by 36k on x86_64 builds. Change-Id: I73f837ddaecac6b350bf757af0cfe19c4ab9327a
2015-04-17 22:11:38 +02:00
PROCESS_16X8X3_OFFSET 0, .vpx_sad16x8x3_ssse3
PROCESS_16X8X3_OFFSET 1, .vpx_sad16x8x3_ssse3
PROCESS_16X8X3_OFFSET 2, .vpx_sad16x8x3_ssse3
PROCESS_16X8X3_OFFSET 3, .vpx_sad16x8x3_ssse3
PROCESS_16X8X3_OFFSET 4, .vpx_sad16x8x3_ssse3
PROCESS_16X8X3_OFFSET 5, .vpx_sad16x8x3_ssse3
PROCESS_16X8X3_OFFSET 6, .vpx_sad16x8x3_ssse3
PROCESS_16X8X3_OFFSET 7, .vpx_sad16x8x3_ssse3
PROCESS_16X8X3_OFFSET 8, .vpx_sad16x8x3_ssse3
PROCESS_16X8X3_OFFSET 9, .vpx_sad16x8x3_ssse3
PROCESS_16X8X3_OFFSET 10, .vpx_sad16x8x3_ssse3
PROCESS_16X8X3_OFFSET 11, .vpx_sad16x8x3_ssse3
PROCESS_16X8X3_OFFSET 12, .vpx_sad16x8x3_ssse3
PROCESS_16X8X3_OFFSET 13, .vpx_sad16x8x3_ssse3
PROCESS_16X8X3_OFFSET 14, .vpx_sad16x8x3_ssse3
.vpx_sad16x8x3_ssse3_aligned_by_15:
Initial WebM release
2010-05-18 17:58:33 +02:00
PROCESS_16X2X3 1
PROCESS_16X2X3 0
PROCESS_16X2X3 0
PROCESS_16X2X3 0
Move shared SAD code to vpx_dsp Create a new component, vpx_dsp, for code that can be shared between codecs. Move the SAD code into the component. This reduces the size of vpxenc/dec by 36k on x86_64 builds. Change-Id: I73f837ddaecac6b350bf757af0cfe19c4ab9327a
2015-04-17 22:11:38 +02:00
.vpx_sad16x8x3_ssse3_store_off:
Initial WebM release
2010-05-18 17:58:33 +02:00
mov rdi, arg(4) ;Results
movq xmm0, xmm5
psrldq xmm5, 8
paddw xmm0, xmm5
movd [rdi], xmm0
;-
movq xmm0, xmm6
psrldq xmm6, 8
paddw xmm0, xmm6
movd [rdi+4], xmm0
;-
movq xmm0, xmm7
psrldq xmm7, 8
paddw xmm0, xmm7
movd [rdi+8], xmm0
; begin epilog
pop rcx
pop rdi
pop rsi
Add save/restore xmm registers in x86 assembly code Went through the code and fixed it. Verified on Windows. Where possible, remove dependencies on xmm[67] Current code relies on pushing rbp to the stack to get 16 byte alignment. This broke when rbp wasn't pushed (vp8/encoder/x86/sad_sse3.asm). Work around this by using unaligned memory accesses. Revisit this and the offsets in vp8/encoder/x86/sad_sse3.asm in another change to SAVE_XMM. Change-Id: I5f940994d3ebfd977c3d68446cef20fd78b07877
2011-04-07 19:17:22 +02:00
RESTORE_XMM
Initial WebM release
2010-05-18 17:58:33 +02:00
UNSHADOW_ARGS
pop rbp
ret
Reference in New Issue Copy Permalink