generic-library/vpx
1
0
Fork 0
Code Issues Pull Requests Releases Wiki Activity
982deebb5e
vpx/vp8/encoder/x86/variance_impl_ssse3.asm

373 lines
13 KiB
NASM
Raw Normal View History

Write SSSE3 sub-pixel filter function 1. Process 16 pixels at one time instead of 8. 2. Add check for both xoffset =0 and yoffset=0, which happens during motion search. This change gave encoder 1%~3% performance gain. Change-Id: Idaa39506b48f4f8b2fbbeb45aae8226fa32afb3e
2011-03-04 01:02:45 +01:00
;
; Copyright (c) 2010 The WebM project authors. All Rights Reserved.
;
; 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 xmm_filter_shift 7
Change encoder vp8_ and vp8cx_ public symbol prefixes to vp9_. Change-Id: Ie2e3652591b010ded10c216501ce24fd95d0aec5
2012-10-30 20:58:42 +01:00
;void vp9_filter_block2d_bil_var_ssse3
Write SSSE3 sub-pixel filter function 1. Process 16 pixels at one time instead of 8. 2. Add check for both xoffset =0 and yoffset=0, which happens during motion search. This change gave encoder 1%~3% performance gain. Change-Id: Idaa39506b48f4f8b2fbbeb45aae8226fa32afb3e
2011-03-04 01:02:45 +01:00
;(
; unsigned char *ref_ptr,
; int ref_pixels_per_line,
; unsigned char *src_ptr,
; int src_pixels_per_line,
; unsigned int Height,
; int xoffset,
; int yoffset,
; int *sum,
; unsigned int *sumsquared;;
;
;)
;Note: The filter coefficient at offset=0 is 128. Since the second register
;for Pmaddubsw is signed bytes, we must calculate zero offset seperately.
Change encoder vp8_ and vp8cx_ public symbol prefixes to vp9_. Change-Id: Ie2e3652591b010ded10c216501ce24fd95d0aec5
2012-10-30 20:58:42 +01:00
global sym(vp9_filter_block2d_bil_var_ssse3)
sym(vp9_filter_block2d_bil_var_ssse3):
Write SSSE3 sub-pixel filter function 1. Process 16 pixels at one time instead of 8. 2. Add check for both xoffset =0 and yoffset=0, which happens during motion search. This change gave encoder 1%~3% performance gain. Change-Id: Idaa39506b48f4f8b2fbbeb45aae8226fa32afb3e
2011-03-04 01:02:45 +01:00
push rbp
mov rbp, rsp
SHADOW_ARGS_TO_STACK 9
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
Write SSSE3 sub-pixel filter function 1. Process 16 pixels at one time instead of 8. 2. Add check for both xoffset =0 and yoffset=0, which happens during motion search. This change gave encoder 1%~3% performance gain. Change-Id: Idaa39506b48f4f8b2fbbeb45aae8226fa32afb3e
2011-03-04 01:02:45 +01:00
GET_GOT rbx
push rsi
push rdi
; end prolog
pxor xmm6, xmm6
pxor xmm7, xmm7
lea rcx, [GLOBAL(vp8_bilinear_filters_ssse3)]
movsxd rax, dword ptr arg(5) ; xoffset
cmp rax, 0 ; skip first_pass filter if xoffset=0
Use local labels for jumps/loops in x86 assembly. Prepend . to local labels in assembly code. This allows non unique labels within a file. Also makes profiling information more informative by keeping the function name with the loop name. Change-Id: I7a983cb3a5ba2413d5dafd0a37936b268fb9e37f
2011-08-23 00:29:41 +02:00
je .filter_block2d_bil_var_ssse3_sp_only
Write SSSE3 sub-pixel filter function 1. Process 16 pixels at one time instead of 8. 2. Add check for both xoffset =0 and yoffset=0, which happens during motion search. This change gave encoder 1%~3% performance gain. Change-Id: Idaa39506b48f4f8b2fbbeb45aae8226fa32afb3e
2011-03-04 01:02:45 +01:00
shl rax, 4 ; point to filter coeff with xoffset
lea rax, [rax + rcx] ; HFilter
movsxd rdx, dword ptr arg(6) ; yoffset
cmp rdx, 0 ; skip second_pass filter if yoffset=0
Use local labels for jumps/loops in x86 assembly. Prepend . to local labels in assembly code. This allows non unique labels within a file. Also makes profiling information more informative by keeping the function name with the loop name. Change-Id: I7a983cb3a5ba2413d5dafd0a37936b268fb9e37f
2011-08-23 00:29:41 +02:00
je .filter_block2d_bil_var_ssse3_fp_only
Write SSSE3 sub-pixel filter function 1. Process 16 pixels at one time instead of 8. 2. Add check for both xoffset =0 and yoffset=0, which happens during motion search. This change gave encoder 1%~3% performance gain. Change-Id: Idaa39506b48f4f8b2fbbeb45aae8226fa32afb3e
2011-03-04 01:02:45 +01:00
shl rdx, 4
lea rdx, [rdx + rcx] ; VFilter
mov rsi, arg(0) ;ref_ptr
mov rdi, arg(2) ;src_ptr
movsxd rcx, dword ptr arg(4) ;Height
movdqu xmm0, XMMWORD PTR [rsi]
movdqu xmm1, XMMWORD PTR [rsi+1]
movdqa xmm2, xmm0
punpcklbw xmm0, xmm1
punpckhbw xmm2, xmm1
pmaddubsw xmm0, [rax]
pmaddubsw xmm2, [rax]
paddw xmm0, [GLOBAL(xmm_bi_rd)]
paddw xmm2, [GLOBAL(xmm_bi_rd)]
psraw xmm0, xmm_filter_shift
psraw xmm2, xmm_filter_shift
packuswb xmm0, xmm2
Fix a crash while enabling shared (--enable-shared) Fixed a bug in SSSE3 sub-pixel filter functions. Change-Id: I2e2126652970eb78307ffcefcace1efd5966fb0a
2011-03-29 17:31:06 +02:00
%if ABI_IS_32BIT
add rsi, dword ptr arg(1) ;ref_pixels_per_line
%else
movsxd r8, dword ptr arg(1) ;ref_pixels_per_line
Write SSSE3 sub-pixel filter function 1. Process 16 pixels at one time instead of 8. 2. Add check for both xoffset =0 and yoffset=0, which happens during motion search. This change gave encoder 1%~3% performance gain. Change-Id: Idaa39506b48f4f8b2fbbeb45aae8226fa32afb3e
2011-03-04 01:02:45 +01:00
movsxd r9, dword ptr arg(3) ;src_pixels_per_line
Fix a crash while enabling shared (--enable-shared) Fixed a bug in SSSE3 sub-pixel filter functions. Change-Id: I2e2126652970eb78307ffcefcace1efd5966fb0a
2011-03-29 17:31:06 +02:00
lea rsi, [rsi + r8]
Write SSSE3 sub-pixel filter function 1. Process 16 pixels at one time instead of 8. 2. Add check for both xoffset =0 and yoffset=0, which happens during motion search. This change gave encoder 1%~3% performance gain. Change-Id: Idaa39506b48f4f8b2fbbeb45aae8226fa32afb3e
2011-03-04 01:02:45 +01:00
%endif
Use local labels for jumps/loops in x86 assembly. Prepend . to local labels in assembly code. This allows non unique labels within a file. Also makes profiling information more informative by keeping the function name with the loop name. Change-Id: I7a983cb3a5ba2413d5dafd0a37936b268fb9e37f
2011-08-23 00:29:41 +02:00
.filter_block2d_bil_var_ssse3_loop:
Write SSSE3 sub-pixel filter function 1. Process 16 pixels at one time instead of 8. 2. Add check for both xoffset =0 and yoffset=0, which happens during motion search. This change gave encoder 1%~3% performance gain. Change-Id: Idaa39506b48f4f8b2fbbeb45aae8226fa32afb3e
2011-03-04 01:02:45 +01:00
movdqu xmm1, XMMWORD PTR [rsi]
movdqu xmm2, XMMWORD PTR [rsi+1]
movdqa xmm3, xmm1
punpcklbw xmm1, xmm2
punpckhbw xmm3, xmm2
pmaddubsw xmm1, [rax]
pmaddubsw xmm3, [rax]
paddw xmm1, [GLOBAL(xmm_bi_rd)]
paddw xmm3, [GLOBAL(xmm_bi_rd)]
psraw xmm1, xmm_filter_shift
psraw xmm3, xmm_filter_shift
packuswb xmm1, xmm3
movdqa xmm2, xmm0
movdqa xmm0, xmm1
movdqa xmm3, xmm2
punpcklbw xmm2, xmm1
punpckhbw xmm3, xmm1
pmaddubsw xmm2, [rdx]
pmaddubsw xmm3, [rdx]
paddw xmm2, [GLOBAL(xmm_bi_rd)]
paddw xmm3, [GLOBAL(xmm_bi_rd)]
psraw xmm2, xmm_filter_shift
psraw xmm3, xmm_filter_shift
movq xmm1, QWORD PTR [rdi]
pxor xmm4, xmm4
punpcklbw xmm1, xmm4
movq xmm5, QWORD PTR [rdi+8]
punpcklbw xmm5, xmm4
psubw xmm2, xmm1
psubw xmm3, xmm5
paddw xmm6, xmm2
paddw xmm6, xmm3
pmaddwd xmm2, xmm2
pmaddwd xmm3, xmm3
paddd xmm7, xmm2
paddd xmm7, xmm3
%if ABI_IS_32BIT
Fix a crash while enabling shared (--enable-shared) Fixed a bug in SSSE3 sub-pixel filter functions. Change-Id: I2e2126652970eb78307ffcefcace1efd5966fb0a
2011-03-29 17:31:06 +02:00
add rsi, dword ptr arg(1) ;ref_pixels_per_line
Write SSSE3 sub-pixel filter function 1. Process 16 pixels at one time instead of 8. 2. Add check for both xoffset =0 and yoffset=0, which happens during motion search. This change gave encoder 1%~3% performance gain. Change-Id: Idaa39506b48f4f8b2fbbeb45aae8226fa32afb3e
2011-03-04 01:02:45 +01:00
add rdi, dword ptr arg(3) ;src_pixels_per_line
%else
Fix a crash while enabling shared (--enable-shared) Fixed a bug in SSSE3 sub-pixel filter functions. Change-Id: I2e2126652970eb78307ffcefcace1efd5966fb0a
2011-03-29 17:31:06 +02:00
lea rsi, [rsi + r8]
Write SSSE3 sub-pixel filter function 1. Process 16 pixels at one time instead of 8. 2. Add check for both xoffset =0 and yoffset=0, which happens during motion search. This change gave encoder 1%~3% performance gain. Change-Id: Idaa39506b48f4f8b2fbbeb45aae8226fa32afb3e
2011-03-04 01:02:45 +01:00
lea rdi, [rdi + r9]
%endif
sub rcx, 1
Use local labels for jumps/loops in x86 assembly. Prepend . to local labels in assembly code. This allows non unique labels within a file. Also makes profiling information more informative by keeping the function name with the loop name. Change-Id: I7a983cb3a5ba2413d5dafd0a37936b268fb9e37f
2011-08-23 00:29:41 +02:00
jnz .filter_block2d_bil_var_ssse3_loop
Write SSSE3 sub-pixel filter function 1. Process 16 pixels at one time instead of 8. 2. Add check for both xoffset =0 and yoffset=0, which happens during motion search. This change gave encoder 1%~3% performance gain. Change-Id: Idaa39506b48f4f8b2fbbeb45aae8226fa32afb3e
2011-03-04 01:02:45 +01:00
Use local labels for jumps/loops in x86 assembly. Prepend . to local labels in assembly code. This allows non unique labels within a file. Also makes profiling information more informative by keeping the function name with the loop name. Change-Id: I7a983cb3a5ba2413d5dafd0a37936b268fb9e37f
2011-08-23 00:29:41 +02:00
jmp .filter_block2d_bil_variance
Write SSSE3 sub-pixel filter function 1. Process 16 pixels at one time instead of 8. 2. Add check for both xoffset =0 and yoffset=0, which happens during motion search. This change gave encoder 1%~3% performance gain. Change-Id: Idaa39506b48f4f8b2fbbeb45aae8226fa32afb3e
2011-03-04 01:02:45 +01:00
Use local labels for jumps/loops in x86 assembly. Prepend . to local labels in assembly code. This allows non unique labels within a file. Also makes profiling information more informative by keeping the function name with the loop name. Change-Id: I7a983cb3a5ba2413d5dafd0a37936b268fb9e37f
2011-08-23 00:29:41 +02:00
.filter_block2d_bil_var_ssse3_sp_only:
Write SSSE3 sub-pixel filter function 1. Process 16 pixels at one time instead of 8. 2. Add check for both xoffset =0 and yoffset=0, which happens during motion search. This change gave encoder 1%~3% performance gain. Change-Id: Idaa39506b48f4f8b2fbbeb45aae8226fa32afb3e
2011-03-04 01:02:45 +01:00
movsxd rdx, dword ptr arg(6) ; yoffset
cmp rdx, 0 ; Both xoffset =0 and yoffset=0
Use local labels for jumps/loops in x86 assembly. Prepend . to local labels in assembly code. This allows non unique labels within a file. Also makes profiling information more informative by keeping the function name with the loop name. Change-Id: I7a983cb3a5ba2413d5dafd0a37936b268fb9e37f
2011-08-23 00:29:41 +02:00
je .filter_block2d_bil_var_ssse3_full_pixel
Write SSSE3 sub-pixel filter function 1. Process 16 pixels at one time instead of 8. 2. Add check for both xoffset =0 and yoffset=0, which happens during motion search. This change gave encoder 1%~3% performance gain. Change-Id: Idaa39506b48f4f8b2fbbeb45aae8226fa32afb3e
2011-03-04 01:02:45 +01:00
shl rdx, 4
lea rdx, [rdx + rcx] ; VFilter
mov rsi, arg(0) ;ref_ptr
mov rdi, arg(2) ;src_ptr
movsxd rcx, dword ptr arg(4) ;Height
movsxd rax, dword ptr arg(1) ;ref_pixels_per_line
movdqu xmm1, XMMWORD PTR [rsi]
movdqa xmm0, xmm1
Fix a crash while enabling shared (--enable-shared) Fixed a bug in SSSE3 sub-pixel filter functions. Change-Id: I2e2126652970eb78307ffcefcace1efd5966fb0a
2011-03-29 17:31:06 +02:00
%if ABI_IS_32BIT=0
movsxd r9, dword ptr arg(3) ;src_pixels_per_line
%endif
Write SSSE3 sub-pixel filter function 1. Process 16 pixels at one time instead of 8. 2. Add check for both xoffset =0 and yoffset=0, which happens during motion search. This change gave encoder 1%~3% performance gain. Change-Id: Idaa39506b48f4f8b2fbbeb45aae8226fa32afb3e
2011-03-04 01:02:45 +01:00
lea rsi, [rsi + rax]
Use local labels for jumps/loops in x86 assembly. Prepend . to local labels in assembly code. This allows non unique labels within a file. Also makes profiling information more informative by keeping the function name with the loop name. Change-Id: I7a983cb3a5ba2413d5dafd0a37936b268fb9e37f
2011-08-23 00:29:41 +02:00
.filter_block2d_bil_sp_only_loop:
Write SSSE3 sub-pixel filter function 1. Process 16 pixels at one time instead of 8. 2. Add check for both xoffset =0 and yoffset=0, which happens during motion search. This change gave encoder 1%~3% performance gain. Change-Id: Idaa39506b48f4f8b2fbbeb45aae8226fa32afb3e
2011-03-04 01:02:45 +01:00
movdqu xmm3, XMMWORD PTR [rsi]
movdqa xmm2, xmm1
movdqa xmm0, xmm3
punpcklbw xmm1, xmm3
punpckhbw xmm2, xmm3
pmaddubsw xmm1, [rdx]
pmaddubsw xmm2, [rdx]
paddw xmm1, [GLOBAL(xmm_bi_rd)]
paddw xmm2, [GLOBAL(xmm_bi_rd)]
psraw xmm1, xmm_filter_shift
psraw xmm2, xmm_filter_shift
movq xmm3, QWORD PTR [rdi]
pxor xmm4, xmm4
punpcklbw xmm3, xmm4
movq xmm5, QWORD PTR [rdi+8]
punpcklbw xmm5, xmm4
psubw xmm1, xmm3
psubw xmm2, xmm5
paddw xmm6, xmm1
paddw xmm6, xmm2
pmaddwd xmm1, xmm1
pmaddwd xmm2, xmm2
paddd xmm7, xmm1
paddd xmm7, xmm2
movdqa xmm1, xmm0
lea rsi, [rsi + rax] ;ref_pixels_per_line
Fix a crash while enabling shared (--enable-shared) Fixed a bug in SSSE3 sub-pixel filter functions. Change-Id: I2e2126652970eb78307ffcefcace1efd5966fb0a
2011-03-29 17:31:06 +02:00
%if ABI_IS_32BIT
add rdi, dword ptr arg(3) ;src_pixels_per_line
%else
lea rdi, [rdi + r9]
%endif
Write SSSE3 sub-pixel filter function 1. Process 16 pixels at one time instead of 8. 2. Add check for both xoffset =0 and yoffset=0, which happens during motion search. This change gave encoder 1%~3% performance gain. Change-Id: Idaa39506b48f4f8b2fbbeb45aae8226fa32afb3e
2011-03-04 01:02:45 +01:00
sub rcx, 1
Use local labels for jumps/loops in x86 assembly. Prepend . to local labels in assembly code. This allows non unique labels within a file. Also makes profiling information more informative by keeping the function name with the loop name. Change-Id: I7a983cb3a5ba2413d5dafd0a37936b268fb9e37f
2011-08-23 00:29:41 +02:00
jnz .filter_block2d_bil_sp_only_loop
Write SSSE3 sub-pixel filter function 1. Process 16 pixels at one time instead of 8. 2. Add check for both xoffset =0 and yoffset=0, which happens during motion search. This change gave encoder 1%~3% performance gain. Change-Id: Idaa39506b48f4f8b2fbbeb45aae8226fa32afb3e
2011-03-04 01:02:45 +01:00
Use local labels for jumps/loops in x86 assembly. Prepend . to local labels in assembly code. This allows non unique labels within a file. Also makes profiling information more informative by keeping the function name with the loop name. Change-Id: I7a983cb3a5ba2413d5dafd0a37936b268fb9e37f
2011-08-23 00:29:41 +02:00
jmp .filter_block2d_bil_variance
Write SSSE3 sub-pixel filter function 1. Process 16 pixels at one time instead of 8. 2. Add check for both xoffset =0 and yoffset=0, which happens during motion search. This change gave encoder 1%~3% performance gain. Change-Id: Idaa39506b48f4f8b2fbbeb45aae8226fa32afb3e
2011-03-04 01:02:45 +01:00
Use local labels for jumps/loops in x86 assembly. Prepend . to local labels in assembly code. This allows non unique labels within a file. Also makes profiling information more informative by keeping the function name with the loop name. Change-Id: I7a983cb3a5ba2413d5dafd0a37936b268fb9e37f
2011-08-23 00:29:41 +02:00
.filter_block2d_bil_var_ssse3_full_pixel:
Write SSSE3 sub-pixel filter function 1. Process 16 pixels at one time instead of 8. 2. Add check for both xoffset =0 and yoffset=0, which happens during motion search. This change gave encoder 1%~3% performance gain. Change-Id: Idaa39506b48f4f8b2fbbeb45aae8226fa32afb3e
2011-03-04 01:02:45 +01:00
mov rsi, arg(0) ;ref_ptr
mov rdi, arg(2) ;src_ptr
movsxd rcx, dword ptr arg(4) ;Height
movsxd rax, dword ptr arg(1) ;ref_pixels_per_line
Fix a crash while enabling shared (--enable-shared) Fixed a bug in SSSE3 sub-pixel filter functions. Change-Id: I2e2126652970eb78307ffcefcace1efd5966fb0a
2011-03-29 17:31:06 +02:00
movsxd rdx, dword ptr arg(3) ;src_pixels_per_line
Write SSSE3 sub-pixel filter function 1. Process 16 pixels at one time instead of 8. 2. Add check for both xoffset =0 and yoffset=0, which happens during motion search. This change gave encoder 1%~3% performance gain. Change-Id: Idaa39506b48f4f8b2fbbeb45aae8226fa32afb3e
2011-03-04 01:02:45 +01:00
pxor xmm0, xmm0
Use local labels for jumps/loops in x86 assembly. Prepend . to local labels in assembly code. This allows non unique labels within a file. Also makes profiling information more informative by keeping the function name with the loop name. Change-Id: I7a983cb3a5ba2413d5dafd0a37936b268fb9e37f
2011-08-23 00:29:41 +02:00
.filter_block2d_bil_full_pixel_loop:
Write SSSE3 sub-pixel filter function 1. Process 16 pixels at one time instead of 8. 2. Add check for both xoffset =0 and yoffset=0, which happens during motion search. This change gave encoder 1%~3% performance gain. Change-Id: Idaa39506b48f4f8b2fbbeb45aae8226fa32afb3e
2011-03-04 01:02:45 +01:00
movq xmm1, QWORD PTR [rsi]
punpcklbw xmm1, xmm0
movq xmm2, QWORD PTR [rsi+8]
punpcklbw xmm2, xmm0
movq xmm3, QWORD PTR [rdi]
punpcklbw xmm3, xmm0
movq xmm4, QWORD PTR [rdi+8]
punpcklbw xmm4, xmm0
psubw xmm1, xmm3
psubw xmm2, xmm4
paddw xmm6, xmm1
paddw xmm6, xmm2
pmaddwd xmm1, xmm1
pmaddwd xmm2, xmm2
paddd xmm7, xmm1
paddd xmm7, xmm2
lea rsi, [rsi + rax] ;ref_pixels_per_line
Fix a crash while enabling shared (--enable-shared) Fixed a bug in SSSE3 sub-pixel filter functions. Change-Id: I2e2126652970eb78307ffcefcace1efd5966fb0a
2011-03-29 17:31:06 +02:00
lea rdi, [rdi + rdx] ;src_pixels_per_line
Write SSSE3 sub-pixel filter function 1. Process 16 pixels at one time instead of 8. 2. Add check for both xoffset =0 and yoffset=0, which happens during motion search. This change gave encoder 1%~3% performance gain. Change-Id: Idaa39506b48f4f8b2fbbeb45aae8226fa32afb3e
2011-03-04 01:02:45 +01:00
sub rcx, 1
Use local labels for jumps/loops in x86 assembly. Prepend . to local labels in assembly code. This allows non unique labels within a file. Also makes profiling information more informative by keeping the function name with the loop name. Change-Id: I7a983cb3a5ba2413d5dafd0a37936b268fb9e37f
2011-08-23 00:29:41 +02:00
jnz .filter_block2d_bil_full_pixel_loop
Write SSSE3 sub-pixel filter function 1. Process 16 pixels at one time instead of 8. 2. Add check for both xoffset =0 and yoffset=0, which happens during motion search. This change gave encoder 1%~3% performance gain. Change-Id: Idaa39506b48f4f8b2fbbeb45aae8226fa32afb3e
2011-03-04 01:02:45 +01:00
Use local labels for jumps/loops in x86 assembly. Prepend . to local labels in assembly code. This allows non unique labels within a file. Also makes profiling information more informative by keeping the function name with the loop name. Change-Id: I7a983cb3a5ba2413d5dafd0a37936b268fb9e37f
2011-08-23 00:29:41 +02:00
jmp .filter_block2d_bil_variance
Write SSSE3 sub-pixel filter function 1. Process 16 pixels at one time instead of 8. 2. Add check for both xoffset =0 and yoffset=0, which happens during motion search. This change gave encoder 1%~3% performance gain. Change-Id: Idaa39506b48f4f8b2fbbeb45aae8226fa32afb3e
2011-03-04 01:02:45 +01:00
Use local labels for jumps/loops in x86 assembly. Prepend . to local labels in assembly code. This allows non unique labels within a file. Also makes profiling information more informative by keeping the function name with the loop name. Change-Id: I7a983cb3a5ba2413d5dafd0a37936b268fb9e37f
2011-08-23 00:29:41 +02:00
.filter_block2d_bil_var_ssse3_fp_only:
Write SSSE3 sub-pixel filter function 1. Process 16 pixels at one time instead of 8. 2. Add check for both xoffset =0 and yoffset=0, which happens during motion search. This change gave encoder 1%~3% performance gain. Change-Id: Idaa39506b48f4f8b2fbbeb45aae8226fa32afb3e
2011-03-04 01:02:45 +01:00
mov rsi, arg(0) ;ref_ptr
mov rdi, arg(2) ;src_ptr
movsxd rcx, dword ptr arg(4) ;Height
movsxd rdx, dword ptr arg(1) ;ref_pixels_per_line
pxor xmm0, xmm0
Fix a crash while enabling shared (--enable-shared) Fixed a bug in SSSE3 sub-pixel filter functions. Change-Id: I2e2126652970eb78307ffcefcace1efd5966fb0a
2011-03-29 17:31:06 +02:00
%if ABI_IS_32BIT=0
movsxd r9, dword ptr arg(3) ;src_pixels_per_line
%endif
Write SSSE3 sub-pixel filter function 1. Process 16 pixels at one time instead of 8. 2. Add check for both xoffset =0 and yoffset=0, which happens during motion search. This change gave encoder 1%~3% performance gain. Change-Id: Idaa39506b48f4f8b2fbbeb45aae8226fa32afb3e
2011-03-04 01:02:45 +01:00
Use local labels for jumps/loops in x86 assembly. Prepend . to local labels in assembly code. This allows non unique labels within a file. Also makes profiling information more informative by keeping the function name with the loop name. Change-Id: I7a983cb3a5ba2413d5dafd0a37936b268fb9e37f
2011-08-23 00:29:41 +02:00
.filter_block2d_bil_fp_only_loop:
Write SSSE3 sub-pixel filter function 1. Process 16 pixels at one time instead of 8. 2. Add check for both xoffset =0 and yoffset=0, which happens during motion search. This change gave encoder 1%~3% performance gain. Change-Id: Idaa39506b48f4f8b2fbbeb45aae8226fa32afb3e
2011-03-04 01:02:45 +01:00
movdqu xmm1, XMMWORD PTR [rsi]
movdqu xmm2, XMMWORD PTR [rsi+1]
movdqa xmm3, xmm1
punpcklbw xmm1, xmm2
punpckhbw xmm3, xmm2
pmaddubsw xmm1, [rax]
pmaddubsw xmm3, [rax]
paddw xmm1, [GLOBAL(xmm_bi_rd)]
paddw xmm3, [GLOBAL(xmm_bi_rd)]
psraw xmm1, xmm_filter_shift
psraw xmm3, xmm_filter_shift
movq xmm2, XMMWORD PTR [rdi]
pxor xmm4, xmm4
punpcklbw xmm2, xmm4
movq xmm5, QWORD PTR [rdi+8]
punpcklbw xmm5, xmm4
psubw xmm1, xmm2
psubw xmm3, xmm5
paddw xmm6, xmm1
paddw xmm6, xmm3
pmaddwd xmm1, xmm1
pmaddwd xmm3, xmm3
paddd xmm7, xmm1
paddd xmm7, xmm3
lea rsi, [rsi + rdx]
Fix a crash while enabling shared (--enable-shared) Fixed a bug in SSSE3 sub-pixel filter functions. Change-Id: I2e2126652970eb78307ffcefcace1efd5966fb0a
2011-03-29 17:31:06 +02:00
%if ABI_IS_32BIT
add rdi, dword ptr arg(3) ;src_pixels_per_line
%else
lea rdi, [rdi + r9]
%endif
Write SSSE3 sub-pixel filter function 1. Process 16 pixels at one time instead of 8. 2. Add check for both xoffset =0 and yoffset=0, which happens during motion search. This change gave encoder 1%~3% performance gain. Change-Id: Idaa39506b48f4f8b2fbbeb45aae8226fa32afb3e
2011-03-04 01:02:45 +01:00
sub rcx, 1
Use local labels for jumps/loops in x86 assembly. Prepend . to local labels in assembly code. This allows non unique labels within a file. Also makes profiling information more informative by keeping the function name with the loop name. Change-Id: I7a983cb3a5ba2413d5dafd0a37936b268fb9e37f
2011-08-23 00:29:41 +02:00
jnz .filter_block2d_bil_fp_only_loop
Write SSSE3 sub-pixel filter function 1. Process 16 pixels at one time instead of 8. 2. Add check for both xoffset =0 and yoffset=0, which happens during motion search. This change gave encoder 1%~3% performance gain. Change-Id: Idaa39506b48f4f8b2fbbeb45aae8226fa32afb3e
2011-03-04 01:02:45 +01:00
Use local labels for jumps/loops in x86 assembly. Prepend . to local labels in assembly code. This allows non unique labels within a file. Also makes profiling information more informative by keeping the function name with the loop name. Change-Id: I7a983cb3a5ba2413d5dafd0a37936b268fb9e37f
2011-08-23 00:29:41 +02:00
jmp .filter_block2d_bil_variance
Write SSSE3 sub-pixel filter function 1. Process 16 pixels at one time instead of 8. 2. Add check for both xoffset =0 and yoffset=0, which happens during motion search. This change gave encoder 1%~3% performance gain. Change-Id: Idaa39506b48f4f8b2fbbeb45aae8226fa32afb3e
2011-03-04 01:02:45 +01:00
Use local labels for jumps/loops in x86 assembly. Prepend . to local labels in assembly code. This allows non unique labels within a file. Also makes profiling information more informative by keeping the function name with the loop name. Change-Id: I7a983cb3a5ba2413d5dafd0a37936b268fb9e37f
2011-08-23 00:29:41 +02:00
.filter_block2d_bil_variance:
Write SSSE3 sub-pixel filter function 1. Process 16 pixels at one time instead of 8. 2. Add check for both xoffset =0 and yoffset=0, which happens during motion search. This change gave encoder 1%~3% performance gain. Change-Id: Idaa39506b48f4f8b2fbbeb45aae8226fa32afb3e
2011-03-04 01:02:45 +01:00
pxor xmm0, xmm0
pxor xmm1, xmm1
pxor xmm5, xmm5
punpcklwd xmm0, xmm6
punpckhwd xmm1, xmm6
psrad xmm0, 16
psrad xmm1, 16
paddd xmm0, xmm1
movdqa xmm1, xmm0
movdqa xmm6, xmm7
punpckldq xmm6, xmm5
punpckhdq xmm7, xmm5
paddd xmm6, xmm7
punpckldq xmm0, xmm5
punpckhdq xmm1, xmm5
paddd xmm0, xmm1
movdqa xmm7, xmm6
movdqa xmm1, xmm0
psrldq xmm7, 8
psrldq xmm1, 8
paddd xmm6, xmm7
paddd xmm0, xmm1
mov rsi, arg(7) ;[Sum]
mov rdi, arg(8) ;[SSE]
movd [rsi], xmm0
movd [rdi], xmm6
; begin epilog
pop rdi
pop rsi
RESTORE_GOT
RESTORE_XMM
UNSHADOW_ARGS
pop rbp
ret
SECTION_RODATA
align 16
xmm_bi_rd:
times 8 dw 64
align 16
WebM Experimental Codec Branch Snapshot This is a code snapshot of experimental work currently ongoing for a next-generation codec. The codebase has been cut down considerably from the libvpx baseline. For example, we are currently only supporting VBR 2-pass rate control and have removed most of the code relating to coding speed, threading, error resilience, partitions and various other features. This is in part to make the codebase easier to work on and experiment with, but also because we want to have an open discussion about how the bitstream will be structured and partitioned and not have that conversation constrained by past work. Our basic working pattern has been to initially encapsulate experiments using configure options linked to #IF CONFIG_XXX statements in the code. Once experiments have matured and we are reasonably happy that they give benefit and can be merged without breaking other experiments, we remove the conditional compile statements and merge them in. Current changes include: * Temporal coding experiment for segments (though still only 4 max, it will likely be increased). * Segment feature experiment - to allow various bits of information to be coded at the segment level. Features tested so far include mode and reference frame information, limiting end of block offset and transform size, alongside Q and loop filter parameters, but this set is very fluid. * Support for 8x8 transform - 8x8 dct with 2nd order 2x2 haar is used in MBs using 16x16 prediction modes within inter frames. * Compound prediction (combination of signals from existing predictors to create a new predictor). * 8 tap interpolation filters and 1/8th pel motion vectors. * Loop filter modifications. * Various entropy modifications and changes to how entropy contexts and updates are handled. * Extended quantizer range matched to transform precision improvements. There are also ongoing further experiments that we hope to merge in the near future: For example, coding of motion and other aspects of the prediction signal to better support larger image formats, use of larger block sizes (e.g. 32x32 and up) and lossless non-transform based coding options (especially for key frames). It is our hope that we will be able to make regular updates and we will warmly welcome community contributions. Please be warned that, at this stage, the codebase is currently slower than VP8 stable branch as most new code has not been optimized, and even the 'C' has been deliberately written to be simple and obvious, not fast. The following graphs have the initial test results, numbers in the tables measure the compression improvement in terms of percentage. The build has the following optional experiments configured: --enable-experimental --enable-enhanced_interp --enable-uvintra --enable-high_precision_mv --enable-sixteenth_subpel_uv CIF Size clips: http://getwebm.org/tmp/cif/ HD size clips: http://getwebm.org/tmp/hd/ (stable_20120309 represents encoding results of WebM master branch build as of commit#7a15907) They were encoded using the following encode parameters: --good --cpu-used=0 -t 0 --lag-in-frames=25 --min-q=0 --max-q=63 --end-usage=0 --auto-alt-ref=1 -p 2 --pass=2 --kf-max-dist=9999 --kf-min-dist=0 --drop-frame=0 --static-thresh=0 --bias-pct=50 --minsection-pct=0 --maxsection-pct=800 --sharpness=0 --arnr-maxframes=7 --arnr-strength=3(for HD,6 for CIF) --arnr-type=3 Change-Id: I5c62ed09cfff5815a2bb34e7820d6a810c23183c
2012-03-10 02:32:50 +01:00
vp8_bilinear_filters_ssse3:
times 8 db 128, 0
times 8 db 120, 8
times 8 db 112, 16
times 8 db 104, 24
times 8 db 96, 32
times 8 db 88, 40
times 8 db 80, 48
times 8 db 72, 56
times 8 db 64, 64
times 8 db 56, 72
times 8 db 48, 80
times 8 db 40, 88
times 8 db 32, 96
times 8 db 24, 104
times 8 db 16, 112
times 8 db 8, 120
Reference in New Issue Copy Permalink