generic-library/vpx
1
0
Fork 0
Code Issues Pull Requests Releases Wiki Activity
1b70862916
vpx/vp8/common/x86/subpixel_ssse3.asm

1555 lines
43 KiB
NASM
Raw Normal View History

Revert "Removed ssse3 sixtap code" This reverts commit 6ea5bb85cd1547b846f4c794e8684de5abcf9f62.
2010-08-19 21:50:29 +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.
Revert "Removed ssse3 sixtap code" This reverts commit 6ea5bb85cd1547b846f4c794e8684de5abcf9f62.
2010-08-19 21:50:29 +02:00
;
; Use of this source code is governed by a BSD-style license
; that can be found in the LICENSE file in the root of the source
; tree. An additional intellectual property rights grant can be found
; in the file PATENTS. All contributing project authors may
; be found in the AUTHORS file in the root of the source tree.
;
%include "vpx_ports/x86_abi_support.asm"
%define BLOCK_HEIGHT_WIDTH 4
%define VP8_FILTER_WEIGHT 128
%define VP8_FILTER_SHIFT 7
;/************************************************************************************
; Notes: filter_block1d_h6 applies a 6 tap filter horizontally to the input pixels. The
; input pixel array has output_height rows. This routine assumes that output_height is an
; even number. This function handles 8 pixels in horizontal direction, calculating ONE
; rows each iteration to take advantage of the 128 bits operations.
;
; This is an implementation of some of the SSE optimizations first seen in ffvp8
;
;*************************************************************************************/
;void vp8_filter_block1d8_h6_ssse3
;(
; unsigned char *src_ptr,
; unsigned int src_pixels_per_line,
; unsigned char *output_ptr,
; unsigned int output_pitch,
; unsigned int output_height,
; unsigned int vp8_filter_index
;)
global sym(vp8_filter_block1d8_h6_ssse3)
sym(vp8_filter_block1d8_h6_ssse3):
push rbp
mov rbp, rsp
SHADOW_ARGS_TO_STACK 6
GET_GOT rbx
push rsi
push rdi
; end prolog
movsxd rdx, DWORD PTR arg(5) ;table index
xor rsi, rsi
shl rdx, 4
nasm: address labels 'rel label' vice 'wrt rip' nasm does not support `label wrt rip', it requires `rel label'. It is still fully compatible with yasm. 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: I488773a4e930a56e43b0cc72d867ee5291215f50
2010-10-04 23:18:58 +02:00
movdqa xmm7, [GLOBAL(rd)]
Revert "Removed ssse3 sixtap code" This reverts commit 6ea5bb85cd1547b846f4c794e8684de5abcf9f62.
2010-08-19 21:50:29 +02:00
nasm: address labels 'rel label' vice 'wrt rip' nasm does not support `label wrt rip', it requires `rel label'. It is still fully compatible with yasm. 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: I488773a4e930a56e43b0cc72d867ee5291215f50
2010-10-04 23:18:58 +02:00
lea rax, [GLOBAL(k0_k5)]
Revert "Removed ssse3 sixtap code" This reverts commit 6ea5bb85cd1547b846f4c794e8684de5abcf9f62.
2010-08-19 21:50:29 +02:00
add rax, rdx
mov rdi, arg(2) ;output_ptr
cmp esi, DWORD PTR [rax]
je vp8_filter_block1d8_h4_ssse3
movdqa xmm4, XMMWORD PTR [rax] ;k0_k5
movdqa xmm5, XMMWORD PTR [rax+256] ;k2_k4
movdqa xmm6, XMMWORD PTR [rax+128] ;k1_k3
mov rsi, arg(0) ;src_ptr
movsxd rax, dword ptr arg(1) ;src_pixels_per_line
movsxd rcx, dword ptr arg(4) ;output_height
movsxd rdx, dword ptr arg(3) ;output_pitch
sub rdi, rdx
;xmm3 free
filter_block1d8_h6_rowloop_ssse3:
Use movq instead of movdqu. Movdqu is more expensive (throughput, uops) than movq. Minimal impact for newer big cores, but ~2.25% gain on Atom. Change-Id: I62c80bb1cc01d8a91c350c4c7719462809a4ef7f
2010-09-20 18:30:49 +02:00
movq xmm0, MMWORD PTR [rsi - 2] ; -2 -1 0 1 2 3 4 5
Revert "Removed ssse3 sixtap code" This reverts commit 6ea5bb85cd1547b846f4c794e8684de5abcf9f62.
2010-08-19 21:50:29 +02:00
Use movq instead of movdqu. Movdqu is more expensive (throughput, uops) than movq. Minimal impact for newer big cores, but ~2.25% gain on Atom. Change-Id: I62c80bb1cc01d8a91c350c4c7719462809a4ef7f
2010-09-20 18:30:49 +02:00
movq xmm2, MMWORD PTR [rsi + 3] ; 3 4 5 6 7 8 9 10
Revert "Removed ssse3 sixtap code" This reverts commit 6ea5bb85cd1547b846f4c794e8684de5abcf9f62.
2010-08-19 21:50:29 +02:00
Use movq instead of movdqu. Movdqu is more expensive (throughput, uops) than movq. Minimal impact for newer big cores, but ~2.25% gain on Atom. Change-Id: I62c80bb1cc01d8a91c350c4c7719462809a4ef7f
2010-09-20 18:30:49 +02:00
punpcklbw xmm0, xmm2 ; -2 3 -1 4 0 5 1 6 2 7 3 8 4 9 5 10
Revert "Removed ssse3 sixtap code" This reverts commit 6ea5bb85cd1547b846f4c794e8684de5abcf9f62.
2010-08-19 21:50:29 +02:00
Use movq instead of movdqu. Movdqu is more expensive (throughput, uops) than movq. Minimal impact for newer big cores, but ~2.25% gain on Atom. Change-Id: I62c80bb1cc01d8a91c350c4c7719462809a4ef7f
2010-09-20 18:30:49 +02:00
movdqa xmm1, xmm0
pmaddubsw xmm0, xmm4
movdqa xmm2, xmm1
nasm: address labels 'rel label' vice 'wrt rip' nasm does not support `label wrt rip', it requires `rel label'. It is still fully compatible with yasm. 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: I488773a4e930a56e43b0cc72d867ee5291215f50
2010-10-04 23:18:58 +02:00
pshufb xmm1, [GLOBAL(shuf2bfrom1)]
Use movq instead of movdqu. Movdqu is more expensive (throughput, uops) than movq. Minimal impact for newer big cores, but ~2.25% gain on Atom. Change-Id: I62c80bb1cc01d8a91c350c4c7719462809a4ef7f
2010-09-20 18:30:49 +02:00
nasm: address labels 'rel label' vice 'wrt rip' nasm does not support `label wrt rip', it requires `rel label'. It is still fully compatible with yasm. 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: I488773a4e930a56e43b0cc72d867ee5291215f50
2010-10-04 23:18:58 +02:00
pshufb xmm2, [GLOBAL(shuf3bfrom1)]
Use movq instead of movdqu. Movdqu is more expensive (throughput, uops) than movq. Minimal impact for newer big cores, but ~2.25% gain on Atom. Change-Id: I62c80bb1cc01d8a91c350c4c7719462809a4ef7f
2010-09-20 18:30:49 +02:00
pmaddubsw xmm1, xmm5
lea rdi, [rdi + rdx]
pmaddubsw xmm2, xmm6
Revert "Removed ssse3 sixtap code" This reverts commit 6ea5bb85cd1547b846f4c794e8684de5abcf9f62.
2010-08-19 21:50:29 +02:00
lea rsi, [rsi + rax]
dec rcx
Use movq instead of movdqu. Movdqu is more expensive (throughput, uops) than movq. Minimal impact for newer big cores, but ~2.25% gain on Atom. Change-Id: I62c80bb1cc01d8a91c350c4c7719462809a4ef7f
2010-09-20 18:30:49 +02:00
paddsw xmm0, xmm1
paddsw xmm2, xmm7
paddsw xmm0, xmm2
psraw xmm0, 7
packuswb xmm0, xmm0
Revert "Removed ssse3 sixtap code" This reverts commit 6ea5bb85cd1547b846f4c794e8684de5abcf9f62.
2010-08-19 21:50:29 +02:00
movq MMWORD Ptr [rdi], xmm0
jnz filter_block1d8_h6_rowloop_ssse3
; begin epilog
pop rdi
pop rsi
RESTORE_GOT
UNSHADOW_ARGS
pop rbp
ret
vp8_filter_block1d8_h4_ssse3:
movdqa xmm5, XMMWORD PTR [rax+256] ;k2_k4
movdqa xmm6, XMMWORD PTR [rax+128] ;k1_k3
nasm: address labels 'rel label' vice 'wrt rip' nasm does not support `label wrt rip', it requires `rel label'. It is still fully compatible with yasm. 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: I488773a4e930a56e43b0cc72d867ee5291215f50
2010-10-04 23:18:58 +02:00
movdqa xmm3, XMMWORD PTR [GLOBAL(shuf2bfrom1)]
movdqa xmm4, XMMWORD PTR [GLOBAL(shuf3bfrom1)]
Revert "Removed ssse3 sixtap code" This reverts commit 6ea5bb85cd1547b846f4c794e8684de5abcf9f62.
2010-08-19 21:50:29 +02:00
mov rsi, arg(0) ;src_ptr
movsxd rax, dword ptr arg(1) ;src_pixels_per_line
movsxd rcx, dword ptr arg(4) ;output_height
movsxd rdx, dword ptr arg(3) ;output_pitch
sub rdi, rdx
Use movq instead of movdqu. Movdqu is more expensive (throughput, uops) than movq. Minimal impact for newer big cores, but ~2.25% gain on Atom. Change-Id: I62c80bb1cc01d8a91c350c4c7719462809a4ef7f
2010-09-20 18:30:49 +02:00
Revert "Removed ssse3 sixtap code" This reverts commit 6ea5bb85cd1547b846f4c794e8684de5abcf9f62.
2010-08-19 21:50:29 +02:00
filter_block1d8_h4_rowloop_ssse3:
Use movq instead of movdqu. Movdqu is more expensive (throughput, uops) than movq. Minimal impact for newer big cores, but ~2.25% gain on Atom. Change-Id: I62c80bb1cc01d8a91c350c4c7719462809a4ef7f
2010-09-20 18:30:49 +02:00
movq xmm0, MMWORD PTR [rsi - 2] ; -2 -1 0 1 2 3 4 5
Revert "Removed ssse3 sixtap code" This reverts commit 6ea5bb85cd1547b846f4c794e8684de5abcf9f62.
2010-08-19 21:50:29 +02:00
Use movq instead of movdqu. Movdqu is more expensive (throughput, uops) than movq. Minimal impact for newer big cores, but ~2.25% gain on Atom. Change-Id: I62c80bb1cc01d8a91c350c4c7719462809a4ef7f
2010-09-20 18:30:49 +02:00
movq xmm1, MMWORD PTR [rsi + 3] ; 3 4 5 6 7 8 9 10
Revert "Removed ssse3 sixtap code" This reverts commit 6ea5bb85cd1547b846f4c794e8684de5abcf9f62.
2010-08-19 21:50:29 +02:00
Use movq instead of movdqu. Movdqu is more expensive (throughput, uops) than movq. Minimal impact for newer big cores, but ~2.25% gain on Atom. Change-Id: I62c80bb1cc01d8a91c350c4c7719462809a4ef7f
2010-09-20 18:30:49 +02:00
punpcklbw xmm0, xmm1 ; -2 3 -1 4 0 5 1 6 2 7 3 8 4 9 5 10
movdqa xmm2, xmm0
pshufb xmm0, xmm3
pshufb xmm2, xmm4
pmaddubsw xmm0, xmm5
lea rdi, [rdi + rdx]
pmaddubsw xmm2, xmm6
Revert "Removed ssse3 sixtap code" This reverts commit 6ea5bb85cd1547b846f4c794e8684de5abcf9f62.
2010-08-19 21:50:29 +02:00
lea rsi, [rsi + rax]
dec rcx
Use movq instead of movdqu. Movdqu is more expensive (throughput, uops) than movq. Minimal impact for newer big cores, but ~2.25% gain on Atom. Change-Id: I62c80bb1cc01d8a91c350c4c7719462809a4ef7f
2010-09-20 18:30:49 +02:00
paddsw xmm0, xmm7
paddsw xmm0, xmm2
psraw xmm0, 7
packuswb xmm0, xmm0
Revert "Removed ssse3 sixtap code" This reverts commit 6ea5bb85cd1547b846f4c794e8684de5abcf9f62.
2010-08-19 21:50:29 +02:00
movq MMWORD Ptr [rdi], xmm0
jnz filter_block1d8_h4_rowloop_ssse3
; begin epilog
pop rdi
pop rsi
RESTORE_GOT
UNSHADOW_ARGS
pop rbp
ret
;void vp8_filter_block1d16_h6_ssse3
;(
; unsigned char *src_ptr,
; unsigned int src_pixels_per_line,
; unsigned char *output_ptr,
; unsigned int output_pitch,
; unsigned int output_height,
; unsigned int vp8_filter_index
;)
global sym(vp8_filter_block1d16_h6_ssse3)
sym(vp8_filter_block1d16_h6_ssse3):
push rbp
mov rbp, rsp
SHADOW_ARGS_TO_STACK 6
SAVE_XMM
GET_GOT rbx
push rsi
push rdi
; end prolog
Use movq instead of movdqu. Movdqu is more expensive (throughput, uops) than movq. Minimal impact for newer big cores, but ~2.25% gain on Atom. Change-Id: I62c80bb1cc01d8a91c350c4c7719462809a4ef7f
2010-09-20 18:30:49 +02:00
movsxd rdx, DWORD PTR arg(5) ;table index
Revert "Removed ssse3 sixtap code" This reverts commit 6ea5bb85cd1547b846f4c794e8684de5abcf9f62.
2010-08-19 21:50:29 +02:00
xor rsi, rsi
shl rdx, 4 ;
nasm: address labels 'rel label' vice 'wrt rip' nasm does not support `label wrt rip', it requires `rel label'. It is still fully compatible with yasm. 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: I488773a4e930a56e43b0cc72d867ee5291215f50
2010-10-04 23:18:58 +02:00
lea rax, [GLOBAL(k0_k5)]
Revert "Removed ssse3 sixtap code" This reverts commit 6ea5bb85cd1547b846f4c794e8684de5abcf9f62.
2010-08-19 21:50:29 +02:00
add rax, rdx
Use movq instead of movdqu. Movdqu is more expensive (throughput, uops) than movq. Minimal impact for newer big cores, but ~2.25% gain on Atom. Change-Id: I62c80bb1cc01d8a91c350c4c7719462809a4ef7f
2010-09-20 18:30:49 +02:00
mov rdi, arg(2) ;output_ptr
Revert "Removed ssse3 sixtap code" This reverts commit 6ea5bb85cd1547b846f4c794e8684de5abcf9f62.
2010-08-19 21:50:29 +02:00
;;
;; cmp esi, DWORD PTR [rax]
;; je vp8_filter_block1d16_h4_ssse3
Use movq instead of movdqu. Movdqu is more expensive (throughput, uops) than movq. Minimal impact for newer big cores, but ~2.25% gain on Atom. Change-Id: I62c80bb1cc01d8a91c350c4c7719462809a4ef7f
2010-09-20 18:30:49 +02:00
mov rsi, arg(0) ;src_ptr
Revert "Removed ssse3 sixtap code" This reverts commit 6ea5bb85cd1547b846f4c794e8684de5abcf9f62.
2010-08-19 21:50:29 +02:00
movdqa xmm4, XMMWORD PTR [rax] ;k0_k5
movdqa xmm5, XMMWORD PTR [rax+256] ;k2_k4
movdqa xmm6, XMMWORD PTR [rax+128] ;k1_k3
Use movq instead of movdqu. Movdqu is more expensive (throughput, uops) than movq. Minimal impact for newer big cores, but ~2.25% gain on Atom. Change-Id: I62c80bb1cc01d8a91c350c4c7719462809a4ef7f
2010-09-20 18:30:49 +02:00
movsxd rax, dword ptr arg(1) ;src_pixels_per_line
movsxd rcx, dword ptr arg(4) ;output_height
movsxd rdx, dword ptr arg(3) ;output_pitch
Revert "Removed ssse3 sixtap code" This reverts commit 6ea5bb85cd1547b846f4c794e8684de5abcf9f62.
2010-08-19 21:50:29 +02:00
filter_block1d16_h6_rowloop_ssse3:
Use movq instead of movdqu. Movdqu is more expensive (throughput, uops) than movq. Minimal impact for newer big cores, but ~2.25% gain on Atom. Change-Id: I62c80bb1cc01d8a91c350c4c7719462809a4ef7f
2010-09-20 18:30:49 +02:00
movq xmm0, MMWORD PTR [rsi - 2] ; -2 -1 0 1 2 3 4 5
Revert "Removed ssse3 sixtap code" This reverts commit 6ea5bb85cd1547b846f4c794e8684de5abcf9f62.
2010-08-19 21:50:29 +02:00
Use movq instead of movdqu. Movdqu is more expensive (throughput, uops) than movq. Minimal impact for newer big cores, but ~2.25% gain on Atom. Change-Id: I62c80bb1cc01d8a91c350c4c7719462809a4ef7f
2010-09-20 18:30:49 +02:00
movq xmm3, MMWORD PTR [rsi + 3] ; 3 4 5 6 7 8 9 10
Revert "Removed ssse3 sixtap code" This reverts commit 6ea5bb85cd1547b846f4c794e8684de5abcf9f62.
2010-08-19 21:50:29 +02:00
Use movq instead of movdqu. Movdqu is more expensive (throughput, uops) than movq. Minimal impact for newer big cores, but ~2.25% gain on Atom. Change-Id: I62c80bb1cc01d8a91c350c4c7719462809a4ef7f
2010-09-20 18:30:49 +02:00
punpcklbw xmm0, xmm3 ; -2 3 -1 4 0 5 1 6 2 7 3 8 4 9 5 10
Revert "Removed ssse3 sixtap code" This reverts commit 6ea5bb85cd1547b846f4c794e8684de5abcf9f62.
2010-08-19 21:50:29 +02:00
Use movq instead of movdqu. Movdqu is more expensive (throughput, uops) than movq. Minimal impact for newer big cores, but ~2.25% gain on Atom. Change-Id: I62c80bb1cc01d8a91c350c4c7719462809a4ef7f
2010-09-20 18:30:49 +02:00
movdqa xmm1, xmm0
pmaddubsw xmm0, xmm4
movdqa xmm2, xmm1
nasm: address labels 'rel label' vice 'wrt rip' nasm does not support `label wrt rip', it requires `rel label'. It is still fully compatible with yasm. 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: I488773a4e930a56e43b0cc72d867ee5291215f50
2010-10-04 23:18:58 +02:00
pshufb xmm1, [GLOBAL(shuf2bfrom1)]
Use movq instead of movdqu. Movdqu is more expensive (throughput, uops) than movq. Minimal impact for newer big cores, but ~2.25% gain on Atom. Change-Id: I62c80bb1cc01d8a91c350c4c7719462809a4ef7f
2010-09-20 18:30:49 +02:00
nasm: address labels 'rel label' vice 'wrt rip' nasm does not support `label wrt rip', it requires `rel label'. It is still fully compatible with yasm. 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: I488773a4e930a56e43b0cc72d867ee5291215f50
2010-10-04 23:18:58 +02:00
pshufb xmm2, [GLOBAL(shuf3bfrom1)]
Use movq instead of movdqu. Movdqu is more expensive (throughput, uops) than movq. Minimal impact for newer big cores, but ~2.25% gain on Atom. Change-Id: I62c80bb1cc01d8a91c350c4c7719462809a4ef7f
2010-09-20 18:30:49 +02:00
movq xmm3, MMWORD PTR [rsi + 6]
pmaddubsw xmm1, xmm5
movq xmm7, MMWORD PTR [rsi + 11]
pmaddubsw xmm2, xmm6
punpcklbw xmm3, xmm7
paddsw xmm0, xmm1
movdqa xmm1, xmm3
pmaddubsw xmm3, xmm4
paddsw xmm0, xmm2
movdqa xmm2, xmm1
nasm: address labels 'rel label' vice 'wrt rip' nasm does not support `label wrt rip', it requires `rel label'. It is still fully compatible with yasm. 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: I488773a4e930a56e43b0cc72d867ee5291215f50
2010-10-04 23:18:58 +02:00
paddsw xmm0, [GLOBAL(rd)]
Use movq instead of movdqu. Movdqu is more expensive (throughput, uops) than movq. Minimal impact for newer big cores, but ~2.25% gain on Atom. Change-Id: I62c80bb1cc01d8a91c350c4c7719462809a4ef7f
2010-09-20 18:30:49 +02:00
nasm: address labels 'rel label' vice 'wrt rip' nasm does not support `label wrt rip', it requires `rel label'. It is still fully compatible with yasm. 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: I488773a4e930a56e43b0cc72d867ee5291215f50
2010-10-04 23:18:58 +02:00
pshufb xmm1, [GLOBAL(shuf2bfrom1)]
pshufb xmm2, [GLOBAL(shuf3bfrom1)]
Use movq instead of movdqu. Movdqu is more expensive (throughput, uops) than movq. Minimal impact for newer big cores, but ~2.25% gain on Atom. Change-Id: I62c80bb1cc01d8a91c350c4c7719462809a4ef7f
2010-09-20 18:30:49 +02:00
psraw xmm0, 7
pmaddubsw xmm1, xmm5
pmaddubsw xmm2, xmm6
packuswb xmm0, xmm0
Revert "Removed ssse3 sixtap code" This reverts commit 6ea5bb85cd1547b846f4c794e8684de5abcf9f62.
2010-08-19 21:50:29 +02:00
lea rsi, [rsi + rax]
Use movq instead of movdqu. Movdqu is more expensive (throughput, uops) than movq. Minimal impact for newer big cores, but ~2.25% gain on Atom. Change-Id: I62c80bb1cc01d8a91c350c4c7719462809a4ef7f
2010-09-20 18:30:49 +02:00
paddsw xmm3, xmm1
Revert "Removed ssse3 sixtap code" This reverts commit 6ea5bb85cd1547b846f4c794e8684de5abcf9f62.
2010-08-19 21:50:29 +02:00
Use movq instead of movdqu. Movdqu is more expensive (throughput, uops) than movq. Minimal impact for newer big cores, but ~2.25% gain on Atom. Change-Id: I62c80bb1cc01d8a91c350c4c7719462809a4ef7f
2010-09-20 18:30:49 +02:00
paddsw xmm3, xmm2
nasm: address labels 'rel label' vice 'wrt rip' nasm does not support `label wrt rip', it requires `rel label'. It is still fully compatible with yasm. 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: I488773a4e930a56e43b0cc72d867ee5291215f50
2010-10-04 23:18:58 +02:00
paddsw xmm3, [GLOBAL(rd)]
Use movq instead of movdqu. Movdqu is more expensive (throughput, uops) than movq. Minimal impact for newer big cores, but ~2.25% gain on Atom. Change-Id: I62c80bb1cc01d8a91c350c4c7719462809a4ef7f
2010-09-20 18:30:49 +02:00
psraw xmm3, 7
packuswb xmm3, xmm3
punpcklqdq xmm0, xmm3
Revert "Removed ssse3 sixtap code" This reverts commit 6ea5bb85cd1547b846f4c794e8684de5abcf9f62.
2010-08-19 21:50:29 +02:00
movdqa XMMWORD Ptr [rdi], xmm0
Use movq instead of movdqu. Movdqu is more expensive (throughput, uops) than movq. Minimal impact for newer big cores, but ~2.25% gain on Atom. Change-Id: I62c80bb1cc01d8a91c350c4c7719462809a4ef7f
2010-09-20 18:30:49 +02:00
lea rdi, [rdi + rdx]
Revert "Removed ssse3 sixtap code" This reverts commit 6ea5bb85cd1547b846f4c794e8684de5abcf9f62.
2010-08-19 21:50:29 +02:00
dec rcx
jnz filter_block1d16_h6_rowloop_ssse3
; begin epilog
pop rdi
pop rsi
RESTORE_GOT
UNSHADOW_ARGS
pop rbp
ret
vp8_filter_block1d16_h4_ssse3:
movdqa xmm5, XMMWORD PTR [rax+256] ;k2_k4
movdqa xmm6, XMMWORD PTR [rax+128] ;k1_k3
mov rsi, arg(0) ;src_ptr
movsxd rax, dword ptr arg(1) ;src_pixels_per_line
movsxd rcx, dword ptr arg(4) ;output_height
movsxd rdx, dword ptr arg(3) ;output_pitch
filter_block1d16_h4_rowloop_ssse3:
movdqu xmm1, XMMWORD PTR [rsi - 2]
movdqa xmm2, xmm1
nasm: address labels 'rel label' vice 'wrt rip' nasm does not support `label wrt rip', it requires `rel label'. It is still fully compatible with yasm. 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: I488773a4e930a56e43b0cc72d867ee5291215f50
2010-10-04 23:18:58 +02:00
pshufb xmm1, [GLOBAL(shuf2b)]
pshufb xmm2, [GLOBAL(shuf3b)]
Revert "Removed ssse3 sixtap code" This reverts commit 6ea5bb85cd1547b846f4c794e8684de5abcf9f62.
2010-08-19 21:50:29 +02:00
pmaddubsw xmm1, xmm5
movdqu xmm3, XMMWORD PTR [rsi + 6]
pmaddubsw xmm2, xmm6
movdqa xmm0, xmm3
nasm: address labels 'rel label' vice 'wrt rip' nasm does not support `label wrt rip', it requires `rel label'. It is still fully compatible with yasm. 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: I488773a4e930a56e43b0cc72d867ee5291215f50
2010-10-04 23:18:58 +02:00
pshufb xmm3, [GLOBAL(shuf3b)]
pshufb xmm0, [GLOBAL(shuf2b)]
Revert "Removed ssse3 sixtap code" This reverts commit 6ea5bb85cd1547b846f4c794e8684de5abcf9f62.
2010-08-19 21:50:29 +02:00
nasm: address labels 'rel label' vice 'wrt rip' nasm does not support `label wrt rip', it requires `rel label'. It is still fully compatible with yasm. 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: I488773a4e930a56e43b0cc72d867ee5291215f50
2010-10-04 23:18:58 +02:00
paddsw xmm1, [GLOBAL(rd)]
Revert "Removed ssse3 sixtap code" This reverts commit 6ea5bb85cd1547b846f4c794e8684de5abcf9f62.
2010-08-19 21:50:29 +02:00
paddsw xmm1, xmm2
pmaddubsw xmm0, xmm5
pmaddubsw xmm3, xmm6
psraw xmm1, 7
packuswb xmm1, xmm1
lea rsi, [rsi + rax]
paddsw xmm3, xmm0
nasm: address labels 'rel label' vice 'wrt rip' nasm does not support `label wrt rip', it requires `rel label'. It is still fully compatible with yasm. 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: I488773a4e930a56e43b0cc72d867ee5291215f50
2010-10-04 23:18:58 +02:00
paddsw xmm3, [GLOBAL(rd)]
Revert "Removed ssse3 sixtap code" This reverts commit 6ea5bb85cd1547b846f4c794e8684de5abcf9f62.
2010-08-19 21:50:29 +02:00
psraw xmm3, 7
packuswb xmm3, xmm3
punpcklqdq xmm1, xmm3
movdqa XMMWORD Ptr [rdi], xmm1
add rdi, rdx
dec rcx
jnz filter_block1d16_h4_rowloop_ssse3
; begin epilog
pop rdi
pop rsi
RESTORE_GOT
UNSHADOW_ARGS
pop rbp
ret
;void vp8_filter_block1d4_h6_ssse3
;(
; unsigned char *src_ptr,
; unsigned int src_pixels_per_line,
; unsigned char *output_ptr,
; unsigned int output_pitch,
; unsigned int output_height,
; unsigned int vp8_filter_index
;)
global sym(vp8_filter_block1d4_h6_ssse3)
sym(vp8_filter_block1d4_h6_ssse3):
push rbp
mov rbp, rsp
SHADOW_ARGS_TO_STACK 6
GET_GOT rbx
push rsi
push rdi
; end prolog
movsxd rdx, DWORD PTR arg(5) ;table index
xor rsi, rsi
shl rdx, 4 ;
nasm: address labels 'rel label' vice 'wrt rip' nasm does not support `label wrt rip', it requires `rel label'. It is still fully compatible with yasm. 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: I488773a4e930a56e43b0cc72d867ee5291215f50
2010-10-04 23:18:58 +02:00
lea rax, [GLOBAL(k0_k5)]
Revert "Removed ssse3 sixtap code" This reverts commit 6ea5bb85cd1547b846f4c794e8684de5abcf9f62.
2010-08-19 21:50:29 +02:00
add rax, rdx
nasm: address labels 'rel label' vice 'wrt rip' nasm does not support `label wrt rip', it requires `rel label'. It is still fully compatible with yasm. 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: I488773a4e930a56e43b0cc72d867ee5291215f50
2010-10-04 23:18:58 +02:00
movdqa xmm7, [GLOBAL(rd)]
Revert "Removed ssse3 sixtap code" This reverts commit 6ea5bb85cd1547b846f4c794e8684de5abcf9f62.
2010-08-19 21:50:29 +02:00
cmp esi, DWORD PTR [rax]
je vp8_filter_block1d4_h4_ssse3
movdqa xmm4, XMMWORD PTR [rax] ;k0_k5
movdqa xmm5, XMMWORD PTR [rax+256] ;k2_k4
movdqa xmm6, XMMWORD PTR [rax+128] ;k1_k3
mov rsi, arg(0) ;src_ptr
mov rdi, arg(2) ;output_ptr
movsxd rax, dword ptr arg(1) ;src_pixels_per_line
movsxd rcx, dword ptr arg(4) ;output_height
movsxd rdx, dword ptr arg(3) ;output_pitch
;xmm3 free
filter_block1d4_h6_rowloop_ssse3:
movdqu xmm0, XMMWORD PTR [rsi - 2]
movdqa xmm1, xmm0
nasm: address labels 'rel label' vice 'wrt rip' nasm does not support `label wrt rip', it requires `rel label'. It is still fully compatible with yasm. 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: I488773a4e930a56e43b0cc72d867ee5291215f50
2010-10-04 23:18:58 +02:00
pshufb xmm0, [GLOBAL(shuf1b)]
Revert "Removed ssse3 sixtap code" This reverts commit 6ea5bb85cd1547b846f4c794e8684de5abcf9f62.
2010-08-19 21:50:29 +02:00
movdqa xmm2, xmm1
nasm: address labels 'rel label' vice 'wrt rip' nasm does not support `label wrt rip', it requires `rel label'. It is still fully compatible with yasm. 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: I488773a4e930a56e43b0cc72d867ee5291215f50
2010-10-04 23:18:58 +02:00
pshufb xmm1, [GLOBAL(shuf2b)]
Revert "Removed ssse3 sixtap code" This reverts commit 6ea5bb85cd1547b846f4c794e8684de5abcf9f62.
2010-08-19 21:50:29 +02:00
pmaddubsw xmm0, xmm4
nasm: address labels 'rel label' vice 'wrt rip' nasm does not support `label wrt rip', it requires `rel label'. It is still fully compatible with yasm. 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: I488773a4e930a56e43b0cc72d867ee5291215f50
2010-10-04 23:18:58 +02:00
pshufb xmm2, [GLOBAL(shuf3b)]
Revert "Removed ssse3 sixtap code" This reverts commit 6ea5bb85cd1547b846f4c794e8684de5abcf9f62.
2010-08-19 21:50:29 +02:00
pmaddubsw xmm1, xmm5
;--
pmaddubsw xmm2, xmm6
lea rsi, [rsi + rax]
;--
paddsw xmm0, xmm1
paddsw xmm0, xmm7
pxor xmm1, xmm1
paddsw xmm0, xmm2
psraw xmm0, 7
packuswb xmm0, xmm0
movd DWORD PTR [rdi], xmm0
add rdi, rdx
dec rcx
jnz filter_block1d4_h6_rowloop_ssse3
; begin epilog
pop rdi
pop rsi
RESTORE_GOT
UNSHADOW_ARGS
pop rbp
ret
vp8_filter_block1d4_h4_ssse3:
movdqa xmm5, XMMWORD PTR [rax+256] ;k2_k4
movdqa xmm6, XMMWORD PTR [rax+128] ;k1_k3
nasm: address labels 'rel label' vice 'wrt rip' nasm does not support `label wrt rip', it requires `rel label'. It is still fully compatible with yasm. 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: I488773a4e930a56e43b0cc72d867ee5291215f50
2010-10-04 23:18:58 +02:00
movdqa xmm0, XMMWORD PTR [GLOBAL(shuf2b)]
movdqa xmm3, XMMWORD PTR [GLOBAL(shuf3b)]
Revert "Removed ssse3 sixtap code" This reverts commit 6ea5bb85cd1547b846f4c794e8684de5abcf9f62.
2010-08-19 21:50:29 +02:00
mov rsi, arg(0) ;src_ptr
mov rdi, arg(2) ;output_ptr
movsxd rax, dword ptr arg(1) ;src_pixels_per_line
movsxd rcx, dword ptr arg(4) ;output_height
movsxd rdx, dword ptr arg(3) ;output_pitch
filter_block1d4_h4_rowloop_ssse3:
movdqu xmm1, XMMWORD PTR [rsi - 2]
movdqa xmm2, xmm1
nasm: address labels 'rel label' vice 'wrt rip' nasm does not support `label wrt rip', it requires `rel label'. It is still fully compatible with yasm. 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: I488773a4e930a56e43b0cc72d867ee5291215f50
2010-10-04 23:18:58 +02:00
pshufb xmm1, xmm0 ;;[GLOBAL(shuf2b)]
pshufb xmm2, xmm3 ;;[GLOBAL(shuf3b)]
Revert "Removed ssse3 sixtap code" This reverts commit 6ea5bb85cd1547b846f4c794e8684de5abcf9f62.
2010-08-19 21:50:29 +02:00
pmaddubsw xmm1, xmm5
;--
pmaddubsw xmm2, xmm6
lea rsi, [rsi + rax]
;--
paddsw xmm1, xmm7
paddsw xmm1, xmm2
psraw xmm1, 7
packuswb xmm1, xmm1
movd DWORD PTR [rdi], xmm1
add rdi, rdx
dec rcx
jnz filter_block1d4_h4_rowloop_ssse3
; begin epilog
pop rdi
pop rsi
RESTORE_GOT
UNSHADOW_ARGS
pop rbp
ret
;void vp8_filter_block1d16_v6_ssse3
;(
; unsigned char *src_ptr,
; unsigned int src_pitch,
; unsigned char *output_ptr,
; unsigned int out_pitch,
; unsigned int output_height,
; unsigned int vp8_filter_index
;)
global sym(vp8_filter_block1d16_v6_ssse3)
sym(vp8_filter_block1d16_v6_ssse3):
push rbp
mov rbp, rsp
SHADOW_ARGS_TO_STACK 6
GET_GOT rbx
push rsi
push rdi
; end prolog
movsxd rdx, DWORD PTR arg(5) ;table index
xor rsi, rsi
shl rdx, 4 ;
nasm: address labels 'rel label' vice 'wrt rip' nasm does not support `label wrt rip', it requires `rel label'. It is still fully compatible with yasm. 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: I488773a4e930a56e43b0cc72d867ee5291215f50
2010-10-04 23:18:58 +02:00
lea rax, [GLOBAL(k0_k5)]
Revert "Removed ssse3 sixtap code" This reverts commit 6ea5bb85cd1547b846f4c794e8684de5abcf9f62.
2010-08-19 21:50:29 +02:00
add rax, rdx
cmp esi, DWORD PTR [rax]
je vp8_filter_block1d16_v4_ssse3
movdqa xmm5, XMMWORD PTR [rax] ;k0_k5
movdqa xmm6, XMMWORD PTR [rax+256] ;k2_k4
movdqa xmm7, XMMWORD PTR [rax+128] ;k1_k3
mov rsi, arg(0) ;src_ptr
movsxd rdx, DWORD PTR arg(1) ;pixels_per_line
mov rdi, arg(2) ;output_ptr
%if ABI_IS_32BIT=0
movsxd r8, DWORD PTR arg(3) ;out_pitch
%endif
mov rax, rsi
movsxd rcx, DWORD PTR arg(4) ;output_height
add rax, rdx
vp8_filter_block1d16_v6_ssse3_loop:
movq xmm1, MMWORD PTR [rsi] ;A
movq xmm2, MMWORD PTR [rsi + rdx] ;B
movq xmm3, MMWORD PTR [rsi + rdx * 2] ;C
movq xmm4, MMWORD PTR [rax + rdx * 2] ;D
movq xmm0, MMWORD PTR [rsi + rdx * 4] ;E
punpcklbw xmm2, xmm4 ;B D
punpcklbw xmm3, xmm0 ;C E
movq xmm0, MMWORD PTR [rax + rdx * 4] ;F
pmaddubsw xmm3, xmm6
punpcklbw xmm1, xmm0 ;A F
pmaddubsw xmm2, xmm7
pmaddubsw xmm1, xmm5
paddsw xmm2, xmm3
paddsw xmm2, xmm1
nasm: address labels 'rel label' vice 'wrt rip' nasm does not support `label wrt rip', it requires `rel label'. It is still fully compatible with yasm. 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: I488773a4e930a56e43b0cc72d867ee5291215f50
2010-10-04 23:18:58 +02:00
paddsw xmm2, [GLOBAL(rd)]
Revert "Removed ssse3 sixtap code" This reverts commit 6ea5bb85cd1547b846f4c794e8684de5abcf9f62.
2010-08-19 21:50:29 +02:00
psraw xmm2, 7
packuswb xmm2, xmm2
movq MMWORD PTR [rdi], xmm2 ;store the results
movq xmm1, MMWORD PTR [rsi + 8] ;A
movq xmm2, MMWORD PTR [rsi + rdx + 8] ;B
movq xmm3, MMWORD PTR [rsi + rdx * 2 + 8] ;C
movq xmm4, MMWORD PTR [rax + rdx * 2 + 8] ;D
movq xmm0, MMWORD PTR [rsi + rdx * 4 + 8] ;E
punpcklbw xmm2, xmm4 ;B D
punpcklbw xmm3, xmm0 ;C E
movq xmm0, MMWORD PTR [rax + rdx * 4 + 8] ;F
pmaddubsw xmm3, xmm6
punpcklbw xmm1, xmm0 ;A F
pmaddubsw xmm2, xmm7
pmaddubsw xmm1, xmm5
add rsi, rdx
add rax, rdx
;--
;--
paddsw xmm2, xmm3
paddsw xmm2, xmm1
nasm: address labels 'rel label' vice 'wrt rip' nasm does not support `label wrt rip', it requires `rel label'. It is still fully compatible with yasm. 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: I488773a4e930a56e43b0cc72d867ee5291215f50
2010-10-04 23:18:58 +02:00
paddsw xmm2, [GLOBAL(rd)]
Revert "Removed ssse3 sixtap code" This reverts commit 6ea5bb85cd1547b846f4c794e8684de5abcf9f62.
2010-08-19 21:50:29 +02:00
psraw xmm2, 7
packuswb xmm2, xmm2
movq MMWORD PTR [rdi+8], xmm2
%if ABI_IS_32BIT
add rdi, DWORD PTR arg(3) ;out_pitch
%else
add rdi, r8
%endif
dec rcx
jnz vp8_filter_block1d16_v6_ssse3_loop
; begin epilog
pop rdi
pop rsi
RESTORE_GOT
UNSHADOW_ARGS
pop rbp
ret
vp8_filter_block1d16_v4_ssse3:
movdqa xmm6, XMMWORD PTR [rax+256] ;k2_k4
movdqa xmm7, XMMWORD PTR [rax+128] ;k1_k3
mov rsi, arg(0) ;src_ptr
movsxd rdx, DWORD PTR arg(1) ;pixels_per_line
mov rdi, arg(2) ;output_ptr
%if ABI_IS_32BIT=0
movsxd r8, DWORD PTR arg(3) ;out_pitch
%endif
mov rax, rsi
movsxd rcx, DWORD PTR arg(4) ;output_height
add rax, rdx
vp8_filter_block1d16_v4_ssse3_loop:
movq xmm2, MMWORD PTR [rsi + rdx] ;B
movq xmm3, MMWORD PTR [rsi + rdx * 2] ;C
movq xmm4, MMWORD PTR [rax + rdx * 2] ;D
movq xmm0, MMWORD PTR [rsi + rdx * 4] ;E
punpcklbw xmm2, xmm4 ;B D
punpcklbw xmm3, xmm0 ;C E
pmaddubsw xmm3, xmm6
pmaddubsw xmm2, xmm7
movq xmm5, MMWORD PTR [rsi + rdx + 8] ;B
movq xmm1, MMWORD PTR [rsi + rdx * 2 + 8] ;C
movq xmm4, MMWORD PTR [rax + rdx * 2 + 8] ;D
movq xmm0, MMWORD PTR [rsi + rdx * 4 + 8] ;E
nasm: address labels 'rel label' vice 'wrt rip' nasm does not support `label wrt rip', it requires `rel label'. It is still fully compatible with yasm. 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: I488773a4e930a56e43b0cc72d867ee5291215f50
2010-10-04 23:18:58 +02:00
paddsw xmm2, [GLOBAL(rd)]
Revert "Removed ssse3 sixtap code" This reverts commit 6ea5bb85cd1547b846f4c794e8684de5abcf9f62.
2010-08-19 21:50:29 +02:00
paddsw xmm2, xmm3
psraw xmm2, 7
packuswb xmm2, xmm2
punpcklbw xmm5, xmm4 ;B D
punpcklbw xmm1, xmm0 ;C E
pmaddubsw xmm1, xmm6
pmaddubsw xmm5, xmm7
nasm: address labels 'rel label' vice 'wrt rip' nasm does not support `label wrt rip', it requires `rel label'. It is still fully compatible with yasm. 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: I488773a4e930a56e43b0cc72d867ee5291215f50
2010-10-04 23:18:58 +02:00
movdqa xmm4, [GLOBAL(rd)]
Revert "Removed ssse3 sixtap code" This reverts commit 6ea5bb85cd1547b846f4c794e8684de5abcf9f62.
2010-08-19 21:50:29 +02:00
add rsi, rdx
add rax, rdx
;--
;--
paddsw xmm5, xmm1
paddsw xmm5, xmm4
psraw xmm5, 7
packuswb xmm5, xmm5
punpcklqdq xmm2, xmm5
movdqa XMMWORD PTR [rdi], xmm2
%if ABI_IS_32BIT
add rdi, DWORD PTR arg(3) ;out_pitch
%else
add rdi, r8
%endif
dec rcx
jnz vp8_filter_block1d16_v4_ssse3_loop
; begin epilog
pop rdi
pop rsi
RESTORE_GOT
UNSHADOW_ARGS
pop rbp
ret
;void vp8_filter_block1d8_v6_ssse3
;(
; unsigned char *src_ptr,
; unsigned int src_pitch,
; unsigned char *output_ptr,
; unsigned int out_pitch,
; unsigned int output_height,
; unsigned int vp8_filter_index
;)
global sym(vp8_filter_block1d8_v6_ssse3)
sym(vp8_filter_block1d8_v6_ssse3):
push rbp
mov rbp, rsp
SHADOW_ARGS_TO_STACK 6
GET_GOT rbx
push rsi
push rdi
; end prolog
movsxd rdx, DWORD PTR arg(5) ;table index
xor rsi, rsi
shl rdx, 4 ;
nasm: address labels 'rel label' vice 'wrt rip' nasm does not support `label wrt rip', it requires `rel label'. It is still fully compatible with yasm. 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: I488773a4e930a56e43b0cc72d867ee5291215f50
2010-10-04 23:18:58 +02:00
lea rax, [GLOBAL(k0_k5)]
Revert "Removed ssse3 sixtap code" This reverts commit 6ea5bb85cd1547b846f4c794e8684de5abcf9f62.
2010-08-19 21:50:29 +02:00
add rax, rdx
movsxd rdx, DWORD PTR arg(1) ;pixels_per_line
mov rdi, arg(2) ;output_ptr
%if ABI_IS_32BIT=0
movsxd r8, DWORD PTR arg(3) ; out_pitch
%endif
movsxd rcx, DWORD PTR arg(4) ;[output_height]
cmp esi, DWORD PTR [rax]
je vp8_filter_block1d8_v4_ssse3
movdqa xmm5, XMMWORD PTR [rax] ;k0_k5
movdqa xmm6, XMMWORD PTR [rax+256] ;k2_k4
movdqa xmm7, XMMWORD PTR [rax+128] ;k1_k3
mov rsi, arg(0) ;src_ptr
mov rax, rsi
add rax, rdx
vp8_filter_block1d8_v6_ssse3_loop:
movq xmm1, MMWORD PTR [rsi] ;A
movq xmm2, MMWORD PTR [rsi + rdx] ;B
movq xmm3, MMWORD PTR [rsi + rdx * 2] ;C
movq xmm4, MMWORD PTR [rax + rdx * 2] ;D
movq xmm0, MMWORD PTR [rsi + rdx * 4] ;E
punpcklbw xmm2, xmm4 ;B D
punpcklbw xmm3, xmm0 ;C E
movq xmm0, MMWORD PTR [rax + rdx * 4] ;F
nasm: address labels 'rel label' vice 'wrt rip' nasm does not support `label wrt rip', it requires `rel label'. It is still fully compatible with yasm. 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: I488773a4e930a56e43b0cc72d867ee5291215f50
2010-10-04 23:18:58 +02:00
movdqa xmm4, [GLOBAL(rd)]
Revert "Removed ssse3 sixtap code" This reverts commit 6ea5bb85cd1547b846f4c794e8684de5abcf9f62.
2010-08-19 21:50:29 +02:00
pmaddubsw xmm3, xmm6
punpcklbw xmm1, xmm0 ;A F
pmaddubsw xmm2, xmm7
pmaddubsw xmm1, xmm5
add rsi, rdx
add rax, rdx
;--
;--
paddsw xmm2, xmm3
paddsw xmm2, xmm1
paddsw xmm2, xmm4
psraw xmm2, 7
packuswb xmm2, xmm2
movq MMWORD PTR [rdi], xmm2
%if ABI_IS_32BIT
add rdi, DWORD PTR arg(3) ;[out_pitch]
%else
add rdi, r8
%endif
dec rcx
jnz vp8_filter_block1d8_v6_ssse3_loop
; begin epilog
pop rdi
pop rsi
RESTORE_GOT
UNSHADOW_ARGS
pop rbp
ret
vp8_filter_block1d8_v4_ssse3:
movdqa xmm6, XMMWORD PTR [rax+256] ;k2_k4
movdqa xmm7, XMMWORD PTR [rax+128] ;k1_k3
nasm: address labels 'rel label' vice 'wrt rip' nasm does not support `label wrt rip', it requires `rel label'. It is still fully compatible with yasm. 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: I488773a4e930a56e43b0cc72d867ee5291215f50
2010-10-04 23:18:58 +02:00
movdqa xmm5, [GLOBAL(rd)]
Revert "Removed ssse3 sixtap code" This reverts commit 6ea5bb85cd1547b846f4c794e8684de5abcf9f62.
2010-08-19 21:50:29 +02:00
mov rsi, arg(0) ;src_ptr
mov rax, rsi
add rax, rdx
vp8_filter_block1d8_v4_ssse3_loop:
movq xmm2, MMWORD PTR [rsi + rdx] ;B
movq xmm3, MMWORD PTR [rsi + rdx * 2] ;C
movq xmm4, MMWORD PTR [rax + rdx * 2] ;D
movq xmm0, MMWORD PTR [rsi + rdx * 4] ;E
punpcklbw xmm2, xmm4 ;B D
punpcklbw xmm3, xmm0 ;C E
pmaddubsw xmm3, xmm6
pmaddubsw xmm2, xmm7
add rsi, rdx
add rax, rdx
;--
;--
paddsw xmm2, xmm3
paddsw xmm2, xmm5
psraw xmm2, 7
packuswb xmm2, xmm2
movq MMWORD PTR [rdi], xmm2
%if ABI_IS_32BIT
add rdi, DWORD PTR arg(3) ;[out_pitch]
%else
add rdi, r8
%endif
dec rcx
jnz vp8_filter_block1d8_v4_ssse3_loop
; begin epilog
pop rdi
pop rsi
RESTORE_GOT
UNSHADOW_ARGS
pop rbp
ret
;void vp8_filter_block1d4_v6_ssse3
;(
; unsigned char *src_ptr,
; unsigned int src_pitch,
; unsigned char *output_ptr,
; unsigned int out_pitch,
; unsigned int output_height,
; unsigned int vp8_filter_index
;)
global sym(vp8_filter_block1d4_v6_ssse3)
sym(vp8_filter_block1d4_v6_ssse3):
push rbp
mov rbp, rsp
SHADOW_ARGS_TO_STACK 6
GET_GOT rbx
push rsi
push rdi
; end prolog
movsxd rdx, DWORD PTR arg(5) ;table index
xor rsi, rsi
shl rdx, 4 ;
nasm: address labels 'rel label' vice 'wrt rip' nasm does not support `label wrt rip', it requires `rel label'. It is still fully compatible with yasm. 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: I488773a4e930a56e43b0cc72d867ee5291215f50
2010-10-04 23:18:58 +02:00
lea rax, [GLOBAL(k0_k5)]
Revert "Removed ssse3 sixtap code" This reverts commit 6ea5bb85cd1547b846f4c794e8684de5abcf9f62.
2010-08-19 21:50:29 +02:00
add rax, rdx
movsxd rdx, DWORD PTR arg(1) ;pixels_per_line
mov rdi, arg(2) ;output_ptr
%if ABI_IS_32BIT=0
movsxd r8, DWORD PTR arg(3) ; out_pitch
%endif
movsxd rcx, DWORD PTR arg(4) ;[output_height]
cmp esi, DWORD PTR [rax]
je vp8_filter_block1d4_v4_ssse3
movq mm5, MMWORD PTR [rax] ;k0_k5
movq mm6, MMWORD PTR [rax+256] ;k2_k4
movq mm7, MMWORD PTR [rax+128] ;k1_k3
mov rsi, arg(0) ;src_ptr
mov rax, rsi
add rax, rdx
vp8_filter_block1d4_v6_ssse3_loop:
movd mm1, DWORD PTR [rsi] ;A
movd mm2, DWORD PTR [rsi + rdx] ;B
movd mm3, DWORD PTR [rsi + rdx * 2] ;C
movd mm4, DWORD PTR [rax + rdx * 2] ;D
movd mm0, DWORD PTR [rsi + rdx * 4] ;E
punpcklbw mm2, mm4 ;B D
punpcklbw mm3, mm0 ;C E
movd mm0, DWORD PTR [rax + rdx * 4] ;F
nasm: address labels 'rel label' vice 'wrt rip' nasm does not support `label wrt rip', it requires `rel label'. It is still fully compatible with yasm. 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: I488773a4e930a56e43b0cc72d867ee5291215f50
2010-10-04 23:18:58 +02:00
movq mm4, [GLOBAL(rd)]
Revert "Removed ssse3 sixtap code" This reverts commit 6ea5bb85cd1547b846f4c794e8684de5abcf9f62.
2010-08-19 21:50:29 +02:00
pmaddubsw mm3, mm6
punpcklbw mm1, mm0 ;A F
pmaddubsw mm2, mm7
pmaddubsw mm1, mm5
add rsi, rdx
add rax, rdx
;--
;--
paddsw mm2, mm3
paddsw mm2, mm1
paddsw mm2, mm4
psraw mm2, 7
packuswb mm2, mm2
movd DWORD PTR [rdi], mm2
%if ABI_IS_32BIT
add rdi, DWORD PTR arg(3) ;[out_pitch]
%else
add rdi, r8
%endif
dec rcx
jnz vp8_filter_block1d4_v6_ssse3_loop
; begin epilog
pop rdi
pop rsi
RESTORE_GOT
UNSHADOW_ARGS
pop rbp
ret
vp8_filter_block1d4_v4_ssse3:
movq mm6, MMWORD PTR [rax+256] ;k2_k4
movq mm7, MMWORD PTR [rax+128] ;k1_k3
nasm: address labels 'rel label' vice 'wrt rip' nasm does not support `label wrt rip', it requires `rel label'. It is still fully compatible with yasm. 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: I488773a4e930a56e43b0cc72d867ee5291215f50
2010-10-04 23:18:58 +02:00
movq mm5, MMWORD PTR [GLOBAL(rd)]
Revert "Removed ssse3 sixtap code" This reverts commit 6ea5bb85cd1547b846f4c794e8684de5abcf9f62.
2010-08-19 21:50:29 +02:00
mov rsi, arg(0) ;src_ptr
mov rax, rsi
add rax, rdx
vp8_filter_block1d4_v4_ssse3_loop:
movd mm2, DWORD PTR [rsi + rdx] ;B
movd mm3, DWORD PTR [rsi + rdx * 2] ;C
movd mm4, DWORD PTR [rax + rdx * 2] ;D
movd mm0, DWORD PTR [rsi + rdx * 4] ;E
punpcklbw mm2, mm4 ;B D
punpcklbw mm3, mm0 ;C E
pmaddubsw mm3, mm6
pmaddubsw mm2, mm7
add rsi, rdx
add rax, rdx
;--
;--
paddsw mm2, mm3
paddsw mm2, mm5
psraw mm2, 7
packuswb mm2, mm2
movd DWORD PTR [rdi], mm2
%if ABI_IS_32BIT
add rdi, DWORD PTR arg(3) ;[out_pitch]
%else
add rdi, r8
%endif
dec rcx
jnz vp8_filter_block1d4_v4_ssse3_loop
; begin epilog
pop rdi
pop rsi
RESTORE_GOT
UNSHADOW_ARGS
pop rbp
ret
Bilinear subpixel optimizations for ssse3. Used pmaddubsw for multiply and add of two filter taps at once for 16x16 and 8x8 blocks. Change-Id: Idccf2d6e094561624407b109fa7e80ba799355ea
2010-09-07 19:52:54 +02:00
;void vp8_bilinear_predict16x16_ssse3
;(
; unsigned char *src_ptr,
; int src_pixels_per_line,
; int xoffset,
; int yoffset,
; unsigned char *dst_ptr,
; int dst_pitch
;)
global sym(vp8_bilinear_predict16x16_ssse3)
sym(vp8_bilinear_predict16x16_ssse3):
push rbp
mov rbp, rsp
SHADOW_ARGS_TO_STACK 6
SAVE_XMM
GET_GOT rbx
push rsi
push rdi
; end prolog
nasm: address labels 'rel label' vice 'wrt rip' nasm does not support `label wrt rip', it requires `rel label'. It is still fully compatible with yasm. 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: I488773a4e930a56e43b0cc72d867ee5291215f50
2010-10-04 23:18:58 +02:00
lea rcx, [GLOBAL(vp8_bilinear_filters_ssse3)]
Bilinear subpixel optimizations for ssse3. Used pmaddubsw for multiply and add of two filter taps at once for 16x16 and 8x8 blocks. Change-Id: Idccf2d6e094561624407b109fa7e80ba799355ea
2010-09-07 19:52:54 +02:00
movsxd rax, dword ptr arg(2) ; xoffset
cmp rax, 0 ; skip first_pass filter if xoffset=0
je b16x16_sp_only
shl rax, 4
lea rax, [rax + rcx] ; HFilter
mov rdi, arg(4) ; dst_ptr
mov rsi, arg(0) ; src_ptr
movsxd rdx, dword ptr arg(5) ; dst_pitch
movdqa xmm1, [rax]
movsxd rax, dword ptr arg(3) ; yoffset
cmp rax, 0 ; skip second_pass filter if yoffset=0
je b16x16_fp_only
shl rax, 4
lea rax, [rax + rcx] ; VFilter
lea rcx, [rdi+rdx*8]
lea rcx, [rcx+rdx*8]
movsxd rdx, dword ptr arg(1) ; src_pixels_per_line
movdqa xmm2, [rax]
%if ABI_IS_32BIT=0
movsxd r8, dword ptr arg(5) ; dst_pitch
%endif
Use movq instead of movdqu. Movdqu is more expensive (throughput, uops) than movq. Minimal impact for newer big cores, but ~2.25% gain on Atom. Change-Id: I62c80bb1cc01d8a91c350c4c7719462809a4ef7f
2010-09-20 18:30:49 +02:00
movq xmm3, [rsi] ; 00 01 02 03 04 05 06 07
movq xmm5, [rsi+1] ; 01 02 03 04 05 06 07 08
Bilinear subpixel optimizations for ssse3. Used pmaddubsw for multiply and add of two filter taps at once for 16x16 and 8x8 blocks. Change-Id: Idccf2d6e094561624407b109fa7e80ba799355ea
2010-09-07 19:52:54 +02:00
Use movq instead of movdqu. Movdqu is more expensive (throughput, uops) than movq. Minimal impact for newer big cores, but ~2.25% gain on Atom. Change-Id: I62c80bb1cc01d8a91c350c4c7719462809a4ef7f
2010-09-20 18:30:49 +02:00
punpcklbw xmm3, xmm5 ; 00 01 01 02 02 03 03 04 04 05 05 06 06 07 07 08
movq xmm4, [rsi+8] ; 08 09 10 11 12 13 14 15
movq xmm5, [rsi+9] ; 09 10 11 12 13 14 15 16
Bilinear subpixel optimizations for ssse3. Used pmaddubsw for multiply and add of two filter taps at once for 16x16 and 8x8 blocks. Change-Id: Idccf2d6e094561624407b109fa7e80ba799355ea
2010-09-07 19:52:54 +02:00
lea rsi, [rsi + rdx] ; next line
pmaddubsw xmm3, xmm1 ; 00 02 04 06 08 10 12 14
Use movq instead of movdqu. Movdqu is more expensive (throughput, uops) than movq. Minimal impact for newer big cores, but ~2.25% gain on Atom. Change-Id: I62c80bb1cc01d8a91c350c4c7719462809a4ef7f
2010-09-20 18:30:49 +02:00
punpcklbw xmm4, xmm5 ; 08 09 09 10 10 11 11 12 12 13 13 14 14 15 15 16
Bilinear subpixel optimizations for ssse3. Used pmaddubsw for multiply and add of two filter taps at once for 16x16 and 8x8 blocks. Change-Id: Idccf2d6e094561624407b109fa7e80ba799355ea
2010-09-07 19:52:54 +02:00
pmaddubsw xmm4, xmm1 ; 01 03 05 07 09 11 13 15
nasm: address labels 'rel label' vice 'wrt rip' nasm does not support `label wrt rip', it requires `rel label'. It is still fully compatible with yasm. 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: I488773a4e930a56e43b0cc72d867ee5291215f50
2010-10-04 23:18:58 +02:00
paddw xmm3, [GLOBAL(rd)] ; xmm3 += round value
Bilinear subpixel optimizations for ssse3. Used pmaddubsw for multiply and add of two filter taps at once for 16x16 and 8x8 blocks. Change-Id: Idccf2d6e094561624407b109fa7e80ba799355ea
2010-09-07 19:52:54 +02:00
psraw xmm3, VP8_FILTER_SHIFT ; xmm3 /= 128
nasm: address labels 'rel label' vice 'wrt rip' nasm does not support `label wrt rip', it requires `rel label'. It is still fully compatible with yasm. 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: I488773a4e930a56e43b0cc72d867ee5291215f50
2010-10-04 23:18:58 +02:00
paddw xmm4, [GLOBAL(rd)] ; xmm4 += round value
Bilinear subpixel optimizations for ssse3. Used pmaddubsw for multiply and add of two filter taps at once for 16x16 and 8x8 blocks. Change-Id: Idccf2d6e094561624407b109fa7e80ba799355ea
2010-09-07 19:52:54 +02:00
psraw xmm4, VP8_FILTER_SHIFT ; xmm4 /= 128
movdqa xmm7, xmm3
packuswb xmm7, xmm4 ; 00 01 02 03 04 05 06 07 08 09 10 11 12 13 14 15
.next_row:
Use movq instead of movdqu. Movdqu is more expensive (throughput, uops) than movq. Minimal impact for newer big cores, but ~2.25% gain on Atom. Change-Id: I62c80bb1cc01d8a91c350c4c7719462809a4ef7f
2010-09-20 18:30:49 +02:00
movq xmm6, [rsi] ; 00 01 02 03 04 05 06 07
movq xmm5, [rsi+1] ; 01 02 03 04 05 06 07 08
Bilinear subpixel optimizations for ssse3. Used pmaddubsw for multiply and add of two filter taps at once for 16x16 and 8x8 blocks. Change-Id: Idccf2d6e094561624407b109fa7e80ba799355ea
2010-09-07 19:52:54 +02:00
Use movq instead of movdqu. Movdqu is more expensive (throughput, uops) than movq. Minimal impact for newer big cores, but ~2.25% gain on Atom. Change-Id: I62c80bb1cc01d8a91c350c4c7719462809a4ef7f
2010-09-20 18:30:49 +02:00
punpcklbw xmm6, xmm5
movq xmm4, [rsi+8] ; 08 09 10 11 12 13 14 15
Bilinear subpixel optimizations for ssse3. Used pmaddubsw for multiply and add of two filter taps at once for 16x16 and 8x8 blocks. Change-Id: Idccf2d6e094561624407b109fa7e80ba799355ea
2010-09-07 19:52:54 +02:00
Use movq instead of movdqu. Movdqu is more expensive (throughput, uops) than movq. Minimal impact for newer big cores, but ~2.25% gain on Atom. Change-Id: I62c80bb1cc01d8a91c350c4c7719462809a4ef7f
2010-09-20 18:30:49 +02:00
movq xmm5, [rsi+9] ; 09 10 11 12 13 14 15 16
Bilinear subpixel optimizations for ssse3. Used pmaddubsw for multiply and add of two filter taps at once for 16x16 and 8x8 blocks. Change-Id: Idccf2d6e094561624407b109fa7e80ba799355ea
2010-09-07 19:52:54 +02:00
lea rsi, [rsi + rdx] ; next line
pmaddubsw xmm6, xmm1
Use movq instead of movdqu. Movdqu is more expensive (throughput, uops) than movq. Minimal impact for newer big cores, but ~2.25% gain on Atom. Change-Id: I62c80bb1cc01d8a91c350c4c7719462809a4ef7f
2010-09-20 18:30:49 +02:00
punpcklbw xmm4, xmm5
Bilinear subpixel optimizations for ssse3. Used pmaddubsw for multiply and add of two filter taps at once for 16x16 and 8x8 blocks. Change-Id: Idccf2d6e094561624407b109fa7e80ba799355ea
2010-09-07 19:52:54 +02:00
pmaddubsw xmm4, xmm1
nasm: address labels 'rel label' vice 'wrt rip' nasm does not support `label wrt rip', it requires `rel label'. It is still fully compatible with yasm. 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: I488773a4e930a56e43b0cc72d867ee5291215f50
2010-10-04 23:18:58 +02:00
paddw xmm6, [GLOBAL(rd)] ; xmm6 += round value
Bilinear subpixel optimizations for ssse3. Used pmaddubsw for multiply and add of two filter taps at once for 16x16 and 8x8 blocks. Change-Id: Idccf2d6e094561624407b109fa7e80ba799355ea
2010-09-07 19:52:54 +02:00
psraw xmm6, VP8_FILTER_SHIFT ; xmm6 /= 128
nasm: address labels 'rel label' vice 'wrt rip' nasm does not support `label wrt rip', it requires `rel label'. It is still fully compatible with yasm. 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: I488773a4e930a56e43b0cc72d867ee5291215f50
2010-10-04 23:18:58 +02:00
paddw xmm4, [GLOBAL(rd)] ; xmm4 += round value
Bilinear subpixel optimizations for ssse3. Used pmaddubsw for multiply and add of two filter taps at once for 16x16 and 8x8 blocks. Change-Id: Idccf2d6e094561624407b109fa7e80ba799355ea
2010-09-07 19:52:54 +02:00
psraw xmm4, VP8_FILTER_SHIFT ; xmm4 /= 128
packuswb xmm6, xmm4
movdqa xmm5, xmm7
punpcklbw xmm5, xmm6
pmaddubsw xmm5, xmm2
punpckhbw xmm7, xmm6
pmaddubsw xmm7, xmm2
nasm: address labels 'rel label' vice 'wrt rip' nasm does not support `label wrt rip', it requires `rel label'. It is still fully compatible with yasm. 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: I488773a4e930a56e43b0cc72d867ee5291215f50
2010-10-04 23:18:58 +02:00
paddw xmm5, [GLOBAL(rd)] ; xmm5 += round value
Bilinear subpixel optimizations for ssse3. Used pmaddubsw for multiply and add of two filter taps at once for 16x16 and 8x8 blocks. Change-Id: Idccf2d6e094561624407b109fa7e80ba799355ea
2010-09-07 19:52:54 +02:00
psraw xmm5, VP8_FILTER_SHIFT ; xmm5 /= 128
nasm: address labels 'rel label' vice 'wrt rip' nasm does not support `label wrt rip', it requires `rel label'. It is still fully compatible with yasm. 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: I488773a4e930a56e43b0cc72d867ee5291215f50
2010-10-04 23:18:58 +02:00
paddw xmm7, [GLOBAL(rd)] ; xmm7 += round value
Bilinear subpixel optimizations for ssse3. Used pmaddubsw for multiply and add of two filter taps at once for 16x16 and 8x8 blocks. Change-Id: Idccf2d6e094561624407b109fa7e80ba799355ea
2010-09-07 19:52:54 +02:00
psraw xmm7, VP8_FILTER_SHIFT ; xmm7 /= 128
packuswb xmm5, xmm7
movdqa xmm7, xmm6
movdqa [rdi], xmm5 ; store the results in the destination
%if ABI_IS_32BIT
add rdi, DWORD PTR arg(5) ; dst_pitch
%else
add rdi, r8
%endif
cmp rdi, rcx
jne .next_row
jmp done
b16x16_sp_only:
movsxd rax, dword ptr arg(3) ; yoffset
shl rax, 4
lea rax, [rax + rcx] ; VFilter
mov rdi, arg(4) ; dst_ptr
mov rsi, arg(0) ; src_ptr
movsxd rdx, dword ptr arg(5) ; dst_pitch
movdqa xmm1, [rax] ; VFilter
lea rcx, [rdi+rdx*8]
lea rcx, [rcx+rdx*8]
movsxd rax, dword ptr arg(1) ; src_pixels_per_line
; get the first horizontal line done
Use movq instead of movdqu. Movdqu is more expensive (throughput, uops) than movq. Minimal impact for newer big cores, but ~2.25% gain on Atom. Change-Id: I62c80bb1cc01d8a91c350c4c7719462809a4ef7f
2010-09-20 18:30:49 +02:00
movq xmm4, [rsi] ; load row 0
movq xmm2, [rsi + 8] ; load row 0
Bilinear subpixel optimizations for ssse3. Used pmaddubsw for multiply and add of two filter taps at once for 16x16 and 8x8 blocks. Change-Id: Idccf2d6e094561624407b109fa7e80ba799355ea
2010-09-07 19:52:54 +02:00
lea rsi, [rsi + rax] ; next line
.next_row:
Use movq instead of movdqu. Movdqu is more expensive (throughput, uops) than movq. Minimal impact for newer big cores, but ~2.25% gain on Atom. Change-Id: I62c80bb1cc01d8a91c350c4c7719462809a4ef7f
2010-09-20 18:30:49 +02:00
movq xmm3, [rsi] ; load row + 1
movq xmm5, [rsi + 8] ; load row + 1
Bilinear subpixel optimizations for ssse3. Used pmaddubsw for multiply and add of two filter taps at once for 16x16 and 8x8 blocks. Change-Id: Idccf2d6e094561624407b109fa7e80ba799355ea
2010-09-07 19:52:54 +02:00
punpcklbw xmm4, xmm3
Use movq instead of movdqu. Movdqu is more expensive (throughput, uops) than movq. Minimal impact for newer big cores, but ~2.25% gain on Atom. Change-Id: I62c80bb1cc01d8a91c350c4c7719462809a4ef7f
2010-09-20 18:30:49 +02:00
punpcklbw xmm2, xmm5
Bilinear subpixel optimizations for ssse3. Used pmaddubsw for multiply and add of two filter taps at once for 16x16 and 8x8 blocks. Change-Id: Idccf2d6e094561624407b109fa7e80ba799355ea
2010-09-07 19:52:54 +02:00
pmaddubsw xmm4, xmm1
Use movq instead of movdqu. Movdqu is more expensive (throughput, uops) than movq. Minimal impact for newer big cores, but ~2.25% gain on Atom. Change-Id: I62c80bb1cc01d8a91c350c4c7719462809a4ef7f
2010-09-20 18:30:49 +02:00
movq xmm7, [rsi + rax] ; load row + 2
Bilinear subpixel optimizations for ssse3. Used pmaddubsw for multiply and add of two filter taps at once for 16x16 and 8x8 blocks. Change-Id: Idccf2d6e094561624407b109fa7e80ba799355ea
2010-09-07 19:52:54 +02:00
pmaddubsw xmm2, xmm1
Use movq instead of movdqu. Movdqu is more expensive (throughput, uops) than movq. Minimal impact for newer big cores, but ~2.25% gain on Atom. Change-Id: I62c80bb1cc01d8a91c350c4c7719462809a4ef7f
2010-09-20 18:30:49 +02:00
movq xmm6, [rsi + rax + 8] ; load row + 2
Bilinear subpixel optimizations for ssse3. Used pmaddubsw for multiply and add of two filter taps at once for 16x16 and 8x8 blocks. Change-Id: Idccf2d6e094561624407b109fa7e80ba799355ea
2010-09-07 19:52:54 +02:00
Use movq instead of movdqu. Movdqu is more expensive (throughput, uops) than movq. Minimal impact for newer big cores, but ~2.25% gain on Atom. Change-Id: I62c80bb1cc01d8a91c350c4c7719462809a4ef7f
2010-09-20 18:30:49 +02:00
punpcklbw xmm3, xmm7
punpcklbw xmm5, xmm6
Bilinear subpixel optimizations for ssse3. Used pmaddubsw for multiply and add of two filter taps at once for 16x16 and 8x8 blocks. Change-Id: Idccf2d6e094561624407b109fa7e80ba799355ea
2010-09-07 19:52:54 +02:00
Use movq instead of movdqu. Movdqu is more expensive (throughput, uops) than movq. Minimal impact for newer big cores, but ~2.25% gain on Atom. Change-Id: I62c80bb1cc01d8a91c350c4c7719462809a4ef7f
2010-09-20 18:30:49 +02:00
pmaddubsw xmm3, xmm1
nasm: address labels 'rel label' vice 'wrt rip' nasm does not support `label wrt rip', it requires `rel label'. It is still fully compatible with yasm. 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: I488773a4e930a56e43b0cc72d867ee5291215f50
2010-10-04 23:18:58 +02:00
paddw xmm4, [GLOBAL(rd)]
Bilinear subpixel optimizations for ssse3. Used pmaddubsw for multiply and add of two filter taps at once for 16x16 and 8x8 blocks. Change-Id: Idccf2d6e094561624407b109fa7e80ba799355ea
2010-09-07 19:52:54 +02:00
Use movq instead of movdqu. Movdqu is more expensive (throughput, uops) than movq. Minimal impact for newer big cores, but ~2.25% gain on Atom. Change-Id: I62c80bb1cc01d8a91c350c4c7719462809a4ef7f
2010-09-20 18:30:49 +02:00
pmaddubsw xmm5, xmm1
nasm: address labels 'rel label' vice 'wrt rip' nasm does not support `label wrt rip', it requires `rel label'. It is still fully compatible with yasm. 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: I488773a4e930a56e43b0cc72d867ee5291215f50
2010-10-04 23:18:58 +02:00
paddw xmm2, [GLOBAL(rd)]
Bilinear subpixel optimizations for ssse3. Used pmaddubsw for multiply and add of two filter taps at once for 16x16 and 8x8 blocks. Change-Id: Idccf2d6e094561624407b109fa7e80ba799355ea
2010-09-07 19:52:54 +02:00
Use movq instead of movdqu. Movdqu is more expensive (throughput, uops) than movq. Minimal impact for newer big cores, but ~2.25% gain on Atom. Change-Id: I62c80bb1cc01d8a91c350c4c7719462809a4ef7f
2010-09-20 18:30:49 +02:00
psraw xmm4, VP8_FILTER_SHIFT
Bilinear subpixel optimizations for ssse3. Used pmaddubsw for multiply and add of two filter taps at once for 16x16 and 8x8 blocks. Change-Id: Idccf2d6e094561624407b109fa7e80ba799355ea
2010-09-07 19:52:54 +02:00
psraw xmm2, VP8_FILTER_SHIFT
packuswb xmm4, xmm2
nasm: address labels 'rel label' vice 'wrt rip' nasm does not support `label wrt rip', it requires `rel label'. It is still fully compatible with yasm. 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: I488773a4e930a56e43b0cc72d867ee5291215f50
2010-10-04 23:18:58 +02:00
paddw xmm3, [GLOBAL(rd)]
Bilinear subpixel optimizations for ssse3. Used pmaddubsw for multiply and add of two filter taps at once for 16x16 and 8x8 blocks. Change-Id: Idccf2d6e094561624407b109fa7e80ba799355ea
2010-09-07 19:52:54 +02:00
movdqa [rdi], xmm4 ; store row 0
nasm: address labels 'rel label' vice 'wrt rip' nasm does not support `label wrt rip', it requires `rel label'. It is still fully compatible with yasm. 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: I488773a4e930a56e43b0cc72d867ee5291215f50
2010-10-04 23:18:58 +02:00
paddw xmm5, [GLOBAL(rd)]
Bilinear subpixel optimizations for ssse3. Used pmaddubsw for multiply and add of two filter taps at once for 16x16 and 8x8 blocks. Change-Id: Idccf2d6e094561624407b109fa7e80ba799355ea
2010-09-07 19:52:54 +02:00
psraw xmm3, VP8_FILTER_SHIFT
Use movq instead of movdqu. Movdqu is more expensive (throughput, uops) than movq. Minimal impact for newer big cores, but ~2.25% gain on Atom. Change-Id: I62c80bb1cc01d8a91c350c4c7719462809a4ef7f
2010-09-20 18:30:49 +02:00
psraw xmm5, VP8_FILTER_SHIFT
Bilinear subpixel optimizations for ssse3. Used pmaddubsw for multiply and add of two filter taps at once for 16x16 and 8x8 blocks. Change-Id: Idccf2d6e094561624407b109fa7e80ba799355ea
2010-09-07 19:52:54 +02:00
Use movq instead of movdqu. Movdqu is more expensive (throughput, uops) than movq. Minimal impact for newer big cores, but ~2.25% gain on Atom. Change-Id: I62c80bb1cc01d8a91c350c4c7719462809a4ef7f
2010-09-20 18:30:49 +02:00
packuswb xmm3, xmm5
movdqa xmm4, xmm7
movdqa [rdi + rdx],xmm3 ; store row 1
lea rsi, [rsi + 2*rax]
Bilinear subpixel optimizations for ssse3. Used pmaddubsw for multiply and add of two filter taps at once for 16x16 and 8x8 blocks. Change-Id: Idccf2d6e094561624407b109fa7e80ba799355ea
2010-09-07 19:52:54 +02:00
Use movq instead of movdqu. Movdqu is more expensive (throughput, uops) than movq. Minimal impact for newer big cores, but ~2.25% gain on Atom. Change-Id: I62c80bb1cc01d8a91c350c4c7719462809a4ef7f
2010-09-20 18:30:49 +02:00
movdqa xmm2, xmm6
Bilinear subpixel optimizations for ssse3. Used pmaddubsw for multiply and add of two filter taps at once for 16x16 and 8x8 blocks. Change-Id: Idccf2d6e094561624407b109fa7e80ba799355ea
2010-09-07 19:52:54 +02:00
lea rdi, [rdi + 2*rdx]
cmp rdi, rcx
jne .next_row
jmp done
b16x16_fp_only:
lea rcx, [rdi+rdx*8]
lea rcx, [rcx+rdx*8]
movsxd rax, dword ptr arg(1) ; src_pixels_per_line
.next_row:
Use movq instead of movdqu. Movdqu is more expensive (throughput, uops) than movq. Minimal impact for newer big cores, but ~2.25% gain on Atom. Change-Id: I62c80bb1cc01d8a91c350c4c7719462809a4ef7f
2010-09-20 18:30:49 +02:00
movq xmm2, [rsi] ; 00 01 02 03 04 05 06 07
movq xmm4, [rsi+1] ; 01 02 03 04 05 06 07 08
Bilinear subpixel optimizations for ssse3. Used pmaddubsw for multiply and add of two filter taps at once for 16x16 and 8x8 blocks. Change-Id: Idccf2d6e094561624407b109fa7e80ba799355ea
2010-09-07 19:52:54 +02:00
punpcklbw xmm2, xmm4
Use movq instead of movdqu. Movdqu is more expensive (throughput, uops) than movq. Minimal impact for newer big cores, but ~2.25% gain on Atom. Change-Id: I62c80bb1cc01d8a91c350c4c7719462809a4ef7f
2010-09-20 18:30:49 +02:00
movq xmm3, [rsi+8] ; 08 09 10 11 12 13 14 15
Bilinear subpixel optimizations for ssse3. Used pmaddubsw for multiply and add of two filter taps at once for 16x16 and 8x8 blocks. Change-Id: Idccf2d6e094561624407b109fa7e80ba799355ea
2010-09-07 19:52:54 +02:00
pmaddubsw xmm2, xmm1
Use movq instead of movdqu. Movdqu is more expensive (throughput, uops) than movq. Minimal impact for newer big cores, but ~2.25% gain on Atom. Change-Id: I62c80bb1cc01d8a91c350c4c7719462809a4ef7f
2010-09-20 18:30:49 +02:00
movq xmm4, [rsi+9] ; 09 10 11 12 13 14 15 16
Bilinear subpixel optimizations for ssse3. Used pmaddubsw for multiply and add of two filter taps at once for 16x16 and 8x8 blocks. Change-Id: Idccf2d6e094561624407b109fa7e80ba799355ea
2010-09-07 19:52:54 +02:00
Use movq instead of movdqu. Movdqu is more expensive (throughput, uops) than movq. Minimal impact for newer big cores, but ~2.25% gain on Atom. Change-Id: I62c80bb1cc01d8a91c350c4c7719462809a4ef7f
2010-09-20 18:30:49 +02:00
lea rsi, [rsi + rax] ; next line
punpcklbw xmm3, xmm4
Bilinear subpixel optimizations for ssse3. Used pmaddubsw for multiply and add of two filter taps at once for 16x16 and 8x8 blocks. Change-Id: Idccf2d6e094561624407b109fa7e80ba799355ea
2010-09-07 19:52:54 +02:00
pmaddubsw xmm3, xmm1
Use movq instead of movdqu. Movdqu is more expensive (throughput, uops) than movq. Minimal impact for newer big cores, but ~2.25% gain on Atom. Change-Id: I62c80bb1cc01d8a91c350c4c7719462809a4ef7f
2010-09-20 18:30:49 +02:00
movq xmm5, [rsi]
nasm: address labels 'rel label' vice 'wrt rip' nasm does not support `label wrt rip', it requires `rel label'. It is still fully compatible with yasm. 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: I488773a4e930a56e43b0cc72d867ee5291215f50
2010-10-04 23:18:58 +02:00
paddw xmm2, [GLOBAL(rd)]
Use movq instead of movdqu. Movdqu is more expensive (throughput, uops) than movq. Minimal impact for newer big cores, but ~2.25% gain on Atom. Change-Id: I62c80bb1cc01d8a91c350c4c7719462809a4ef7f
2010-09-20 18:30:49 +02:00
movq xmm7, [rsi+1]
Bilinear subpixel optimizations for ssse3. Used pmaddubsw for multiply and add of two filter taps at once for 16x16 and 8x8 blocks. Change-Id: Idccf2d6e094561624407b109fa7e80ba799355ea
2010-09-07 19:52:54 +02:00
Use movq instead of movdqu. Movdqu is more expensive (throughput, uops) than movq. Minimal impact for newer big cores, but ~2.25% gain on Atom. Change-Id: I62c80bb1cc01d8a91c350c4c7719462809a4ef7f
2010-09-20 18:30:49 +02:00
movq xmm6, [rsi+8]
Bilinear subpixel optimizations for ssse3. Used pmaddubsw for multiply and add of two filter taps at once for 16x16 and 8x8 blocks. Change-Id: Idccf2d6e094561624407b109fa7e80ba799355ea
2010-09-07 19:52:54 +02:00
psraw xmm2, VP8_FILTER_SHIFT
Use movq instead of movdqu. Movdqu is more expensive (throughput, uops) than movq. Minimal impact for newer big cores, but ~2.25% gain on Atom. Change-Id: I62c80bb1cc01d8a91c350c4c7719462809a4ef7f
2010-09-20 18:30:49 +02:00
Bilinear subpixel optimizations for ssse3. Used pmaddubsw for multiply and add of two filter taps at once for 16x16 and 8x8 blocks. Change-Id: Idccf2d6e094561624407b109fa7e80ba799355ea
2010-09-07 19:52:54 +02:00
punpcklbw xmm5, xmm7
Use movq instead of movdqu. Movdqu is more expensive (throughput, uops) than movq. Minimal impact for newer big cores, but ~2.25% gain on Atom. Change-Id: I62c80bb1cc01d8a91c350c4c7719462809a4ef7f
2010-09-20 18:30:49 +02:00
movq xmm7, [rsi+9]
Bilinear subpixel optimizations for ssse3. Used pmaddubsw for multiply and add of two filter taps at once for 16x16 and 8x8 blocks. Change-Id: Idccf2d6e094561624407b109fa7e80ba799355ea
2010-09-07 19:52:54 +02:00
nasm: address labels 'rel label' vice 'wrt rip' nasm does not support `label wrt rip', it requires `rel label'. It is still fully compatible with yasm. 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: I488773a4e930a56e43b0cc72d867ee5291215f50
2010-10-04 23:18:58 +02:00
paddw xmm3, [GLOBAL(rd)]
Bilinear subpixel optimizations for ssse3. Used pmaddubsw for multiply and add of two filter taps at once for 16x16 and 8x8 blocks. Change-Id: Idccf2d6e094561624407b109fa7e80ba799355ea
2010-09-07 19:52:54 +02:00
pmaddubsw xmm5, xmm1
psraw xmm3, VP8_FILTER_SHIFT
Use movq instead of movdqu. Movdqu is more expensive (throughput, uops) than movq. Minimal impact for newer big cores, but ~2.25% gain on Atom. Change-Id: I62c80bb1cc01d8a91c350c4c7719462809a4ef7f
2010-09-20 18:30:49 +02:00
punpcklbw xmm6, xmm7
Bilinear subpixel optimizations for ssse3. Used pmaddubsw for multiply and add of two filter taps at once for 16x16 and 8x8 blocks. Change-Id: Idccf2d6e094561624407b109fa7e80ba799355ea
2010-09-07 19:52:54 +02:00
packuswb xmm2, xmm3
pmaddubsw xmm6, xmm1
movdqa [rdi], xmm2 ; store the results in the destination
nasm: address labels 'rel label' vice 'wrt rip' nasm does not support `label wrt rip', it requires `rel label'. It is still fully compatible with yasm. 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: I488773a4e930a56e43b0cc72d867ee5291215f50
2010-10-04 23:18:58 +02:00
paddw xmm5, [GLOBAL(rd)]
Bilinear subpixel optimizations for ssse3. Used pmaddubsw for multiply and add of two filter taps at once for 16x16 and 8x8 blocks. Change-Id: Idccf2d6e094561624407b109fa7e80ba799355ea
2010-09-07 19:52:54 +02:00
lea rdi, [rdi + rdx] ; dst_pitch
psraw xmm5, VP8_FILTER_SHIFT
nasm: address labels 'rel label' vice 'wrt rip' nasm does not support `label wrt rip', it requires `rel label'. It is still fully compatible with yasm. 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: I488773a4e930a56e43b0cc72d867ee5291215f50
2010-10-04 23:18:58 +02:00
paddw xmm6, [GLOBAL(rd)]
Bilinear subpixel optimizations for ssse3. Used pmaddubsw for multiply and add of two filter taps at once for 16x16 and 8x8 blocks. Change-Id: Idccf2d6e094561624407b109fa7e80ba799355ea
2010-09-07 19:52:54 +02:00
psraw xmm6, VP8_FILTER_SHIFT
packuswb xmm5, xmm6
lea rsi, [rsi + rax] ; next line
movdqa [rdi], xmm5 ; store the results in the destination
lea rdi, [rdi + rdx] ; dst_pitch
cmp rdi, rcx
jne .next_row
done:
; begin epilog
pop rdi
pop rsi
RESTORE_GOT
RESTORE_XMM
UNSHADOW_ARGS
pop rbp
ret
;void vp8_bilinear_predict8x8_ssse3
;(
; unsigned char *src_ptr,
; int src_pixels_per_line,
; int xoffset,
; int yoffset,
; unsigned char *dst_ptr,
; int dst_pitch
;)
global sym(vp8_bilinear_predict8x8_ssse3)
sym(vp8_bilinear_predict8x8_ssse3):
push rbp
mov rbp, rsp
SHADOW_ARGS_TO_STACK 6
SAVE_XMM
GET_GOT rbx
push rsi
push rdi
; end prolog
ALIGN_STACK 16, rax
sub rsp, 144 ; reserve 144 bytes
nasm: address labels 'rel label' vice 'wrt rip' nasm does not support `label wrt rip', it requires `rel label'. It is still fully compatible with yasm. 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: I488773a4e930a56e43b0cc72d867ee5291215f50
2010-10-04 23:18:58 +02:00
lea rcx, [GLOBAL(vp8_bilinear_filters_ssse3)]
Bilinear subpixel optimizations for ssse3. Used pmaddubsw for multiply and add of two filter taps at once for 16x16 and 8x8 blocks. Change-Id: Idccf2d6e094561624407b109fa7e80ba799355ea
2010-09-07 19:52:54 +02:00
mov rsi, arg(0) ;src_ptr
movsxd rdx, dword ptr arg(1) ;src_pixels_per_line
;Read 9-line unaligned data in and put them on stack. This gives a big
;performance boost.
movdqu xmm0, [rsi]
lea rax, [rdx + rdx*2]
movdqu xmm1, [rsi+rdx]
movdqu xmm2, [rsi+rdx*2]
add rsi, rax
movdqu xmm3, [rsi]
movdqu xmm4, [rsi+rdx]
movdqu xmm5, [rsi+rdx*2]
add rsi, rax
movdqu xmm6, [rsi]
movdqu xmm7, [rsi+rdx]
movdqa XMMWORD PTR [rsp], xmm0
movdqu xmm0, [rsi+rdx*2]
movdqa XMMWORD PTR [rsp+16], xmm1
movdqa XMMWORD PTR [rsp+32], xmm2
movdqa XMMWORD PTR [rsp+48], xmm3
movdqa XMMWORD PTR [rsp+64], xmm4
movdqa XMMWORD PTR [rsp+80], xmm5
movdqa XMMWORD PTR [rsp+96], xmm6
movdqa XMMWORD PTR [rsp+112], xmm7
movdqa XMMWORD PTR [rsp+128], xmm0
movsxd rax, dword ptr arg(2) ; xoffset
cmp rax, 0 ; skip first_pass filter if xoffset=0
je b8x8_sp_only
shl rax, 4
add rax, rcx ; HFilter
mov rdi, arg(4) ; dst_ptr
movsxd rdx, dword ptr arg(5) ; dst_pitch
movdqa xmm0, [rax]
movsxd rax, dword ptr arg(3) ; yoffset
cmp rax, 0 ; skip second_pass filter if yoffset=0
je b8x8_fp_only
shl rax, 4
lea rax, [rax + rcx] ; VFilter
lea rcx, [rdi+rdx*8]
movdqa xmm1, [rax]
; get the first horizontal line done
movdqa xmm3, [rsp] ; 00 01 02 03 04 05 06 07 08 09 10 11 12 13 14 15
movdqa xmm5, xmm3 ; 01 02 03 04 05 06 07 08 09 10 11 12 13 14 15 xx
psrldq xmm5, 1
lea rsp, [rsp + 16] ; next line
punpcklbw xmm3, xmm5 ; 00 01 01 02 02 03 03 04 04 05 05 06 06 07 07 08
pmaddubsw xmm3, xmm0 ; 00 02 04 06 08 10 12 14
nasm: address labels 'rel label' vice 'wrt rip' nasm does not support `label wrt rip', it requires `rel label'. It is still fully compatible with yasm. 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: I488773a4e930a56e43b0cc72d867ee5291215f50
2010-10-04 23:18:58 +02:00
paddw xmm3, [GLOBAL(rd)] ; xmm3 += round value
Bilinear subpixel optimizations for ssse3. Used pmaddubsw for multiply and add of two filter taps at once for 16x16 and 8x8 blocks. Change-Id: Idccf2d6e094561624407b109fa7e80ba799355ea
2010-09-07 19:52:54 +02:00
psraw xmm3, VP8_FILTER_SHIFT ; xmm3 /= 128
movdqa xmm7, xmm3
packuswb xmm7, xmm7 ; 00 01 02 03 04 05 06 07 08 09 10 11 12 13 14 15
.next_row:
movdqa xmm6, [rsp] ; 00 01 02 03 04 05 06 07 08 09 10 11 12 13 14 15
lea rsp, [rsp + 16] ; next line
movdqa xmm5, xmm6
psrldq xmm5, 1
punpcklbw xmm6, xmm5
pmaddubsw xmm6, xmm0
nasm: address labels 'rel label' vice 'wrt rip' nasm does not support `label wrt rip', it requires `rel label'. It is still fully compatible with yasm. 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: I488773a4e930a56e43b0cc72d867ee5291215f50
2010-10-04 23:18:58 +02:00
paddw xmm6, [GLOBAL(rd)] ; xmm6 += round value
Bilinear subpixel optimizations for ssse3. Used pmaddubsw for multiply and add of two filter taps at once for 16x16 and 8x8 blocks. Change-Id: Idccf2d6e094561624407b109fa7e80ba799355ea
2010-09-07 19:52:54 +02:00
psraw xmm6, VP8_FILTER_SHIFT ; xmm6 /= 128
packuswb xmm6, xmm6
punpcklbw xmm7, xmm6
pmaddubsw xmm7, xmm1
nasm: address labels 'rel label' vice 'wrt rip' nasm does not support `label wrt rip', it requires `rel label'. It is still fully compatible with yasm. 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: I488773a4e930a56e43b0cc72d867ee5291215f50
2010-10-04 23:18:58 +02:00
paddw xmm7, [GLOBAL(rd)] ; xmm7 += round value
Bilinear subpixel optimizations for ssse3. Used pmaddubsw for multiply and add of two filter taps at once for 16x16 and 8x8 blocks. Change-Id: Idccf2d6e094561624407b109fa7e80ba799355ea
2010-09-07 19:52:54 +02:00
psraw xmm7, VP8_FILTER_SHIFT ; xmm7 /= 128
packuswb xmm7, xmm7
movq [rdi], xmm7 ; store the results in the destination
lea rdi, [rdi + rdx]
movdqa xmm7, xmm6
cmp rdi, rcx
jne .next_row
jmp done8x8
b8x8_sp_only:
movsxd rax, dword ptr arg(3) ; yoffset
shl rax, 4
lea rax, [rax + rcx] ; VFilter
mov rdi, arg(4) ;dst_ptr
movsxd rdx, dword ptr arg(5) ; dst_pitch
movdqa xmm0, [rax] ; VFilter
movq xmm1, XMMWORD PTR [rsp]
movq xmm2, XMMWORD PTR [rsp+16]
movq xmm3, XMMWORD PTR [rsp+32]
punpcklbw xmm1, xmm2
movq xmm4, XMMWORD PTR [rsp+48]
punpcklbw xmm2, xmm3
movq xmm5, XMMWORD PTR [rsp+64]
punpcklbw xmm3, xmm4
movq xmm6, XMMWORD PTR [rsp+80]
punpcklbw xmm4, xmm5
movq xmm7, XMMWORD PTR [rsp+96]
punpcklbw xmm5, xmm6
pmaddubsw xmm1, xmm0
pmaddubsw xmm2, xmm0
pmaddubsw xmm3, xmm0
pmaddubsw xmm4, xmm0
pmaddubsw xmm5, xmm0
punpcklbw xmm6, xmm7
pmaddubsw xmm6, xmm0
nasm: address labels 'rel label' vice 'wrt rip' nasm does not support `label wrt rip', it requires `rel label'. It is still fully compatible with yasm. 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: I488773a4e930a56e43b0cc72d867ee5291215f50
2010-10-04 23:18:58 +02:00
paddw xmm1, [GLOBAL(rd)]
Bilinear subpixel optimizations for ssse3. Used pmaddubsw for multiply and add of two filter taps at once for 16x16 and 8x8 blocks. Change-Id: Idccf2d6e094561624407b109fa7e80ba799355ea
2010-09-07 19:52:54 +02:00
nasm: address labels 'rel label' vice 'wrt rip' nasm does not support `label wrt rip', it requires `rel label'. It is still fully compatible with yasm. 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: I488773a4e930a56e43b0cc72d867ee5291215f50
2010-10-04 23:18:58 +02:00
paddw xmm2, [GLOBAL(rd)]
Bilinear subpixel optimizations for ssse3. Used pmaddubsw for multiply and add of two filter taps at once for 16x16 and 8x8 blocks. Change-Id: Idccf2d6e094561624407b109fa7e80ba799355ea
2010-09-07 19:52:54 +02:00
psraw xmm1, VP8_FILTER_SHIFT
nasm: address labels 'rel label' vice 'wrt rip' nasm does not support `label wrt rip', it requires `rel label'. It is still fully compatible with yasm. 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: I488773a4e930a56e43b0cc72d867ee5291215f50
2010-10-04 23:18:58 +02:00
paddw xmm3, [GLOBAL(rd)]
Bilinear subpixel optimizations for ssse3. Used pmaddubsw for multiply and add of two filter taps at once for 16x16 and 8x8 blocks. Change-Id: Idccf2d6e094561624407b109fa7e80ba799355ea
2010-09-07 19:52:54 +02:00
psraw xmm2, VP8_FILTER_SHIFT
nasm: address labels 'rel label' vice 'wrt rip' nasm does not support `label wrt rip', it requires `rel label'. It is still fully compatible with yasm. 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: I488773a4e930a56e43b0cc72d867ee5291215f50
2010-10-04 23:18:58 +02:00
paddw xmm4, [GLOBAL(rd)]
Bilinear subpixel optimizations for ssse3. Used pmaddubsw for multiply and add of two filter taps at once for 16x16 and 8x8 blocks. Change-Id: Idccf2d6e094561624407b109fa7e80ba799355ea
2010-09-07 19:52:54 +02:00
psraw xmm3, VP8_FILTER_SHIFT
nasm: address labels 'rel label' vice 'wrt rip' nasm does not support `label wrt rip', it requires `rel label'. It is still fully compatible with yasm. 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: I488773a4e930a56e43b0cc72d867ee5291215f50
2010-10-04 23:18:58 +02:00
paddw xmm5, [GLOBAL(rd)]
Bilinear subpixel optimizations for ssse3. Used pmaddubsw for multiply and add of two filter taps at once for 16x16 and 8x8 blocks. Change-Id: Idccf2d6e094561624407b109fa7e80ba799355ea
2010-09-07 19:52:54 +02:00
psraw xmm4, VP8_FILTER_SHIFT
nasm: address labels 'rel label' vice 'wrt rip' nasm does not support `label wrt rip', it requires `rel label'. It is still fully compatible with yasm. 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: I488773a4e930a56e43b0cc72d867ee5291215f50
2010-10-04 23:18:58 +02:00
paddw xmm6, [GLOBAL(rd)]
Bilinear subpixel optimizations for ssse3. Used pmaddubsw for multiply and add of two filter taps at once for 16x16 and 8x8 blocks. Change-Id: Idccf2d6e094561624407b109fa7e80ba799355ea
2010-09-07 19:52:54 +02:00
psraw xmm5, VP8_FILTER_SHIFT
psraw xmm6, VP8_FILTER_SHIFT
packuswb xmm1, xmm1
packuswb xmm2, xmm2
movq [rdi], xmm1
packuswb xmm3, xmm3
movq [rdi+rdx], xmm2
packuswb xmm4, xmm4
movq xmm1, XMMWORD PTR [rsp+112]
lea rdi, [rdi + 2*rdx]
movq xmm2, XMMWORD PTR [rsp+128]
packuswb xmm5, xmm5
movq [rdi], xmm3
packuswb xmm6, xmm6
movq [rdi+rdx], xmm4
lea rdi, [rdi + 2*rdx]
punpcklbw xmm7, xmm1
movq [rdi], xmm5
pmaddubsw xmm7, xmm0
movq [rdi+rdx], xmm6
punpcklbw xmm1, xmm2
pmaddubsw xmm1, xmm0
nasm: address labels 'rel label' vice 'wrt rip' nasm does not support `label wrt rip', it requires `rel label'. It is still fully compatible with yasm. 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: I488773a4e930a56e43b0cc72d867ee5291215f50
2010-10-04 23:18:58 +02:00
paddw xmm7, [GLOBAL(rd)]
Bilinear subpixel optimizations for ssse3. Used pmaddubsw for multiply and add of two filter taps at once for 16x16 and 8x8 blocks. Change-Id: Idccf2d6e094561624407b109fa7e80ba799355ea
2010-09-07 19:52:54 +02:00
psraw xmm7, VP8_FILTER_SHIFT
nasm: address labels 'rel label' vice 'wrt rip' nasm does not support `label wrt rip', it requires `rel label'. It is still fully compatible with yasm. 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: I488773a4e930a56e43b0cc72d867ee5291215f50
2010-10-04 23:18:58 +02:00
paddw xmm1, [GLOBAL(rd)]
Bilinear subpixel optimizations for ssse3. Used pmaddubsw for multiply and add of two filter taps at once for 16x16 and 8x8 blocks. Change-Id: Idccf2d6e094561624407b109fa7e80ba799355ea
2010-09-07 19:52:54 +02:00
psraw xmm1, VP8_FILTER_SHIFT
packuswb xmm7, xmm7
packuswb xmm1, xmm1
lea rdi, [rdi + 2*rdx]
movq [rdi], xmm7
movq [rdi+rdx], xmm1
lea rsp, [rsp + 144]
jmp done8x8
b8x8_fp_only:
lea rcx, [rdi+rdx*8]
.next_row:
movdqa xmm1, XMMWORD PTR [rsp]
movdqa xmm3, XMMWORD PTR [rsp+16]
movdqa xmm2, xmm1
movdqa xmm5, XMMWORD PTR [rsp+32]
psrldq xmm2, 1
movdqa xmm7, XMMWORD PTR [rsp+48]
movdqa xmm4, xmm3
psrldq xmm4, 1
movdqa xmm6, xmm5
psrldq xmm6, 1
punpcklbw xmm1, xmm2
pmaddubsw xmm1, xmm0
punpcklbw xmm3, xmm4
pmaddubsw xmm3, xmm0
punpcklbw xmm5, xmm6
pmaddubsw xmm5, xmm0
movdqa xmm2, xmm7
psrldq xmm2, 1
punpcklbw xmm7, xmm2
pmaddubsw xmm7, xmm0
nasm: address labels 'rel label' vice 'wrt rip' nasm does not support `label wrt rip', it requires `rel label'. It is still fully compatible with yasm. 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: I488773a4e930a56e43b0cc72d867ee5291215f50
2010-10-04 23:18:58 +02:00
paddw xmm1, [GLOBAL(rd)]
Bilinear subpixel optimizations for ssse3. Used pmaddubsw for multiply and add of two filter taps at once for 16x16 and 8x8 blocks. Change-Id: Idccf2d6e094561624407b109fa7e80ba799355ea
2010-09-07 19:52:54 +02:00
psraw xmm1, VP8_FILTER_SHIFT
nasm: address labels 'rel label' vice 'wrt rip' nasm does not support `label wrt rip', it requires `rel label'. It is still fully compatible with yasm. 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: I488773a4e930a56e43b0cc72d867ee5291215f50
2010-10-04 23:18:58 +02:00
paddw xmm3, [GLOBAL(rd)]
Bilinear subpixel optimizations for ssse3. Used pmaddubsw for multiply and add of two filter taps at once for 16x16 and 8x8 blocks. Change-Id: Idccf2d6e094561624407b109fa7e80ba799355ea
2010-09-07 19:52:54 +02:00
psraw xmm3, VP8_FILTER_SHIFT
nasm: address labels 'rel label' vice 'wrt rip' nasm does not support `label wrt rip', it requires `rel label'. It is still fully compatible with yasm. 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: I488773a4e930a56e43b0cc72d867ee5291215f50
2010-10-04 23:18:58 +02:00
paddw xmm5, [GLOBAL(rd)]
Bilinear subpixel optimizations for ssse3. Used pmaddubsw for multiply and add of two filter taps at once for 16x16 and 8x8 blocks. Change-Id: Idccf2d6e094561624407b109fa7e80ba799355ea
2010-09-07 19:52:54 +02:00
psraw xmm5, VP8_FILTER_SHIFT
nasm: address labels 'rel label' vice 'wrt rip' nasm does not support `label wrt rip', it requires `rel label'. It is still fully compatible with yasm. 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: I488773a4e930a56e43b0cc72d867ee5291215f50
2010-10-04 23:18:58 +02:00
paddw xmm7, [GLOBAL(rd)]
Bilinear subpixel optimizations for ssse3. Used pmaddubsw for multiply and add of two filter taps at once for 16x16 and 8x8 blocks. Change-Id: Idccf2d6e094561624407b109fa7e80ba799355ea
2010-09-07 19:52:54 +02:00
psraw xmm7, VP8_FILTER_SHIFT
packuswb xmm1, xmm1
packuswb xmm3, xmm3
packuswb xmm5, xmm5
movq [rdi], xmm1
packuswb xmm7, xmm7
movq [rdi+rdx], xmm3
lea rdi, [rdi + 2*rdx]
movq [rdi], xmm5
lea rsp, [rsp + 4*16]
movq [rdi+rdx], xmm7
lea rdi, [rdi + 2*rdx]
cmp rdi, rcx
jne .next_row
lea rsp, [rsp + 16]
done8x8:
;add rsp, 144
pop rsp
; begin epilog
pop rdi
pop rsi
RESTORE_GOT
RESTORE_XMM
UNSHADOW_ARGS
pop rbp
ret
Revert "Removed ssse3 sixtap code" This reverts commit 6ea5bb85cd1547b846f4c794e8684de5abcf9f62.
2010-08-19 21:50:29 +02:00
SECTION_RODATA
align 16
shuf1b:
db 0, 5, 1, 6, 2, 7, 3, 8, 4, 9, 5, 10, 6, 11, 7, 12
shuf2b:
db 2, 4, 3, 5, 4, 6, 5, 7, 6, 8, 7, 9, 8, 10, 9, 11
shuf3b:
db 1, 3, 2, 4, 3, 5, 4, 6, 5, 7, 6, 8, 7, 9, 8, 10
Use movq instead of movdqu. Movdqu is more expensive (throughput, uops) than movq. Minimal impact for newer big cores, but ~2.25% gain on Atom. Change-Id: I62c80bb1cc01d8a91c350c4c7719462809a4ef7f
2010-09-20 18:30:49 +02:00
align 16
shuf2bfrom1:
db 4, 8, 6, 1, 8, 3, 1, 5, 3, 7, 5, 9, 7,11, 9,13
align 16
shuf3bfrom1:
db 2, 6, 4, 8, 6, 1, 8, 3, 1, 5, 3, 7, 5, 9, 7,11
Revert "Removed ssse3 sixtap code" This reverts commit 6ea5bb85cd1547b846f4c794e8684de5abcf9f62.
2010-08-19 21:50:29 +02:00
align 16
rd:
times 8 dw 0x40
align 16
k0_k5:
times 8 db 0, 0 ;placeholder
times 8 db 0, 0
times 8 db 2, 1
times 8 db 0, 0
times 8 db 3, 3
times 8 db 0, 0
times 8 db 1, 2
times 8 db 0, 0
k1_k3:
times 8 db 0, 0 ;placeholder
times 8 db -6, 12
times 8 db -11, 36
times 8 db -9, 50
times 8 db -16, 77
times 8 db -6, 93
times 8 db -8, 108
times 8 db -1, 123
k2_k4:
times 8 db 128, 0 ;placeholder
times 8 db 123, -1
times 8 db 108, -8
times 8 db 93, -6
times 8 db 77, -16
times 8 db 50, -9
times 8 db 36, -11
times 8 db 12, -6
Bilinear subpixel optimizations for ssse3. Used pmaddubsw for multiply and add of two filter taps at once for 16x16 and 8x8 blocks. Change-Id: Idccf2d6e094561624407b109fa7e80ba799355ea
2010-09-07 19:52:54 +02:00
align 16
vp8_bilinear_filters_ssse3:
times 8 db 128, 0
times 8 db 112, 16
times 8 db 96, 32
times 8 db 80, 48
times 8 db 64, 64
times 8 db 48, 80
times 8 db 32, 96
times 8 db 16, 112
Revert "Removed ssse3 sixtap code" This reverts commit 6ea5bb85cd1547b846f4c794e8684de5abcf9f62.
2010-08-19 21:50:29 +02:00
Reference in New Issue Copy Permalink