generic-library/vpx
1
0
Fork 0
Code Issues Pull Requests Releases Wiki Activity
66ccf5ddcf
vpx/vp9/encoder/x86/vp9_variance_impl_mmx.asm

511 lines
21 KiB
NASM
Raw Normal View History

Initial WebM release
2010-05-18 17:58:33 +02:00
;
Use WebM in copyright notice for consistency Changes 'The VP8 project' to 'The WebM project', for consistency with other webmproject.org repositories. Fixes issue #97. Change-Id: I37c13ed5fbdb9d334ceef71c6350e9febed9bbba
2010-09-09 14:16:39 +02:00
; Copyright (c) 2010 The WebM project authors. All Rights Reserved.
Initial WebM release
2010-05-18 17:58:33 +02:00
;
cosmetics: trim trailing whitespace When the license headers were updated, they accidentally contained trailing whitespace, so unfortunately we have to touch all the files again. Change-Id: I236c05fade06589e417179c0444cb39b09e4200d
2010-06-18 18:39:21 +02:00
; Use of this source code is governed by a BSD-style license
LICENSE: update with latest text Change-Id: Ieebea089095d9073b3a94932791099f614ce120c
2010-06-04 22:19:40 +02:00
; that can be found in the LICENSE file in the root of the source
; tree. An additional intellectual property rights grant can be found
cosmetics: trim trailing whitespace When the license headers were updated, they accidentally contained trailing whitespace, so unfortunately we have to touch all the files again. Change-Id: I236c05fade06589e417179c0444cb39b09e4200d
2010-06-18 18:39:21 +02:00
; in the file PATENTS. All contributing project authors may
LICENSE: update with latest text Change-Id: Ieebea089095d9073b3a94932791099f614ce120c
2010-06-04 22:19:40 +02:00
; be found in the AUTHORS file in the root of the source tree.
Initial WebM release
2010-05-18 17:58:33 +02:00
;
%include "vpx_ports/x86_abi_support.asm"
Change encoder vp8_ and vp8cx_ public symbol prefixes to vp9_. Change-Id: Ie2e3652591b010ded10c216501ce24fd95d0aec5
2012-10-30 20:58:42 +01:00
;unsigned int vp9_get_mb_ss_mmx( short *src_ptr )
add private to assembly files to insure proper chromebuild Change-Id: I6e43ca73f35401a974ed8ee27738d4318f09fd37
2012-12-20 18:40:18 +01:00
global sym(vp9_get_mb_ss_mmx) PRIVATE
Change encoder vp8_ and vp8cx_ public symbol prefixes to vp9_. Change-Id: Ie2e3652591b010ded10c216501ce24fd95d0aec5
2012-10-30 20:58:42 +01:00
sym(vp9_get_mb_ss_mmx):
Initial WebM release
2010-05-18 17:58:33 +02:00
push rbp
mov rbp, rsp
SHADOW_ARGS_TO_STACK 7
GET_GOT rbx
push rsi
push rdi
sub rsp, 8
; end prolog
mov rax, arg(0) ;src_ptr
mov rcx, 16
pxor mm4, mm4
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
.NEXTROW:
Initial WebM release
2010-05-18 17:58:33 +02:00
movq mm0, [rax]
movq mm1, [rax+8]
movq mm2, [rax+16]
movq mm3, [rax+24]
pmaddwd mm0, mm0
pmaddwd mm1, mm1
pmaddwd mm2, mm2
pmaddwd mm3, mm3
paddd mm4, mm0
paddd mm4, mm1
paddd mm4, mm2
paddd mm4, mm3
add rax, 32
dec rcx
Use local labels for jumps/loops in x86 assembly. Prepend . to local labels in assembly code. This allows non unique labels within a file. Also makes profiling information more informative by keeping the function name with the loop name. Change-Id: I7a983cb3a5ba2413d5dafd0a37936b268fb9e37f
2011-08-23 00:29:41 +02:00
ja .NEXTROW
Initial WebM release
2010-05-18 17:58:33 +02:00
movq QWORD PTR [rsp], mm4
;return sum[0]+sum[1];
movsxd rax, dword ptr [rsp]
movsxd rcx, dword ptr [rsp+4]
add rax, rcx
; begin epilog
add rsp, 8
pop rdi
pop rsi
RESTORE_GOT
UNSHADOW_ARGS
pop rbp
ret
Change encoder vp8_ and vp8cx_ public symbol prefixes to vp9_. Change-Id: Ie2e3652591b010ded10c216501ce24fd95d0aec5
2012-10-30 20:58:42 +01:00
;unsigned int vp9_get8x8var_mmx
Initial WebM release
2010-05-18 17:58:33 +02:00
;(
; unsigned char *src_ptr,
; int source_stride,
; unsigned char *ref_ptr,
; int recon_stride,
; unsigned int *SSE,
; int *Sum
;)
add private to assembly files to insure proper chromebuild Change-Id: I6e43ca73f35401a974ed8ee27738d4318f09fd37
2012-12-20 18:40:18 +01:00
global sym(vp9_get8x8var_mmx) PRIVATE
Change encoder vp8_ and vp8cx_ public symbol prefixes to vp9_. Change-Id: Ie2e3652591b010ded10c216501ce24fd95d0aec5
2012-10-30 20:58:42 +01:00
sym(vp9_get8x8var_mmx):
Initial WebM release
2010-05-18 17:58:33 +02:00
push rbp
mov rbp, rsp
SHADOW_ARGS_TO_STACK 6
push rsi
push rdi
push rbx
sub rsp, 16
; end prolog
pxor mm5, mm5 ; Blank mmx6
pxor mm6, mm6 ; Blank mmx7
pxor mm7, mm7 ; Blank mmx7
mov rax, arg(0) ;[src_ptr] ; Load base addresses
mov rbx, arg(2) ;[ref_ptr]
movsxd rcx, dword ptr arg(1) ;[source_stride]
movsxd rdx, dword ptr arg(3) ;[recon_stride]
; Row 1
movq mm0, [rax] ; Copy eight bytes to mm0
movq mm1, [rbx] ; Copy eight bytes to mm1
movq mm2, mm0 ; Take copies
movq mm3, mm1 ; Take copies
punpcklbw mm0, mm6 ; unpack to higher prrcision
punpcklbw mm1, mm6
punpckhbw mm2, mm6 ; unpack to higher prrcision
punpckhbw mm3, mm6
psubsw mm0, mm1 ; A-B (low order) to MM0
psubsw mm2, mm3 ; A-B (high order) to MM2
paddw mm5, mm0 ; accumulate differences in mm5
paddw mm5, mm2 ; accumulate differences in mm5
pmaddwd mm0, mm0 ; square and accumulate
pmaddwd mm2, mm2 ; square and accumulate
add rbx,rdx ; Inc pointer into ref data
add rax,rcx ; Inc pointer into the new data
movq mm1, [rbx] ; Copy eight bytes to mm1
paddd mm7, mm0 ; accumulate in mm7
paddd mm7, mm2 ; accumulate in mm7
; Row 2
movq mm0, [rax] ; Copy eight bytes to mm0
movq mm2, mm0 ; Take copies
movq mm3, mm1 ; Take copies
punpcklbw mm0, mm6 ; unpack to higher prrcision
punpcklbw mm1, mm6
punpckhbw mm2, mm6 ; unpack to higher prrcision
punpckhbw mm3, mm6
psubsw mm0, mm1 ; A-B (low order) to MM0
psubsw mm2, mm3 ; A-B (high order) to MM2
paddw mm5, mm0 ; accumulate differences in mm5
paddw mm5, mm2 ; accumulate differences in mm5
pmaddwd mm0, mm0 ; square and accumulate
pmaddwd mm2, mm2 ; square and accumulate
add rbx,rdx ; Inc pointer into ref data
add rax,rcx ; Inc pointer into the new data
movq mm1, [rbx] ; Copy eight bytes to mm1
paddd mm7, mm0 ; accumulate in mm7
paddd mm7, mm2 ; accumulate in mm7
; Row 3
movq mm0, [rax] ; Copy eight bytes to mm0
movq mm2, mm0 ; Take copies
movq mm3, mm1 ; Take copies
punpcklbw mm0, mm6 ; unpack to higher prrcision
punpcklbw mm1, mm6
punpckhbw mm2, mm6 ; unpack to higher prrcision
punpckhbw mm3, mm6
psubsw mm0, mm1 ; A-B (low order) to MM0
psubsw mm2, mm3 ; A-B (high order) to MM2
paddw mm5, mm0 ; accumulate differences in mm5
paddw mm5, mm2 ; accumulate differences in mm5
pmaddwd mm0, mm0 ; square and accumulate
pmaddwd mm2, mm2 ; square and accumulate
add rbx,rdx ; Inc pointer into ref data
add rax,rcx ; Inc pointer into the new data
movq mm1, [rbx] ; Copy eight bytes to mm1
paddd mm7, mm0 ; accumulate in mm7
paddd mm7, mm2 ; accumulate in mm7
; Row 4
movq mm0, [rax] ; Copy eight bytes to mm0
movq mm2, mm0 ; Take copies
movq mm3, mm1 ; Take copies
punpcklbw mm0, mm6 ; unpack to higher prrcision
punpcklbw mm1, mm6
punpckhbw mm2, mm6 ; unpack to higher prrcision
punpckhbw mm3, mm6
psubsw mm0, mm1 ; A-B (low order) to MM0
psubsw mm2, mm3 ; A-B (high order) to MM2
paddw mm5, mm0 ; accumulate differences in mm5
paddw mm5, mm2 ; accumulate differences in mm5
pmaddwd mm0, mm0 ; square and accumulate
pmaddwd mm2, mm2 ; square and accumulate
add rbx,rdx ; Inc pointer into ref data
add rax,rcx ; Inc pointer into the new data
movq mm1, [rbx] ; Copy eight bytes to mm1
paddd mm7, mm0 ; accumulate in mm7
paddd mm7, mm2 ; accumulate in mm7
; Row 5
movq mm0, [rax] ; Copy eight bytes to mm0
movq mm2, mm0 ; Take copies
movq mm3, mm1 ; Take copies
punpcklbw mm0, mm6 ; unpack to higher prrcision
punpcklbw mm1, mm6
punpckhbw mm2, mm6 ; unpack to higher prrcision
punpckhbw mm3, mm6
psubsw mm0, mm1 ; A-B (low order) to MM0
psubsw mm2, mm3 ; A-B (high order) to MM2
paddw mm5, mm0 ; accumulate differences in mm5
paddw mm5, mm2 ; accumulate differences in mm5
pmaddwd mm0, mm0 ; square and accumulate
pmaddwd mm2, mm2 ; square and accumulate
add rbx,rdx ; Inc pointer into ref data
add rax,rcx ; Inc pointer into the new data
movq mm1, [rbx] ; Copy eight bytes to mm1
; movq mm4, [rbx + rdx]
paddd mm7, mm0 ; accumulate in mm7
paddd mm7, mm2 ; accumulate in mm7
; Row 6
movq mm0, [rax] ; Copy eight bytes to mm0
movq mm2, mm0 ; Take copies
movq mm3, mm1 ; Take copies
punpcklbw mm0, mm6 ; unpack to higher prrcision
punpcklbw mm1, mm6
punpckhbw mm2, mm6 ; unpack to higher prrcision
punpckhbw mm3, mm6
psubsw mm0, mm1 ; A-B (low order) to MM0
psubsw mm2, mm3 ; A-B (high order) to MM2
paddw mm5, mm0 ; accumulate differences in mm5
paddw mm5, mm2 ; accumulate differences in mm5
pmaddwd mm0, mm0 ; square and accumulate
pmaddwd mm2, mm2 ; square and accumulate
add rbx,rdx ; Inc pointer into ref data
add rax,rcx ; Inc pointer into the new data
movq mm1, [rbx] ; Copy eight bytes to mm1
paddd mm7, mm0 ; accumulate in mm7
paddd mm7, mm2 ; accumulate in mm7
; Row 7
movq mm0, [rax] ; Copy eight bytes to mm0
movq mm2, mm0 ; Take copies
movq mm3, mm1 ; Take copies
punpcklbw mm0, mm6 ; unpack to higher prrcision
punpcklbw mm1, mm6
punpckhbw mm2, mm6 ; unpack to higher prrcision
punpckhbw mm3, mm6
psubsw mm0, mm1 ; A-B (low order) to MM0
psubsw mm2, mm3 ; A-B (high order) to MM2
paddw mm5, mm0 ; accumulate differences in mm5
paddw mm5, mm2 ; accumulate differences in mm5
pmaddwd mm0, mm0 ; square and accumulate
pmaddwd mm2, mm2 ; square and accumulate
add rbx,rdx ; Inc pointer into ref data
add rax,rcx ; Inc pointer into the new data
movq mm1, [rbx] ; Copy eight bytes to mm1
paddd mm7, mm0 ; accumulate in mm7
paddd mm7, mm2 ; accumulate in mm7
; Row 8
movq mm0, [rax] ; Copy eight bytes to mm0
movq mm2, mm0 ; Take copies
movq mm3, mm1 ; Take copies
punpcklbw mm0, mm6 ; unpack to higher prrcision
punpcklbw mm1, mm6
punpckhbw mm2, mm6 ; unpack to higher prrcision
punpckhbw mm3, mm6
psubsw mm0, mm1 ; A-B (low order) to MM0
psubsw mm2, mm3 ; A-B (high order) to MM2
paddw mm5, mm0 ; accumulate differences in mm5
paddw mm5, mm2 ; accumulate differences in mm5
pmaddwd mm0, mm0 ; square and accumulate
pmaddwd mm2, mm2 ; square and accumulate
add rbx,rdx ; Inc pointer into ref data
add rax,rcx ; Inc pointer into the new data
paddd mm7, mm0 ; accumulate in mm7
paddd mm7, mm2 ; accumulate in mm7
; Now accumulate the final results.
movq QWORD PTR [rsp+8], mm5 ; copy back accumulated results into normal memory
movq QWORD PTR [rsp], mm7 ; copy back accumulated results into normal memory
movsx rdx, WORD PTR [rsp+8]
movsx rcx, WORD PTR [rsp+10]
movsx rbx, WORD PTR [rsp+12]
movsx rax, WORD PTR [rsp+14]
add rdx, rcx
add rbx, rax
add rdx, rbx ;XSum
movsxd rax, DWORD PTR [rsp]
movsxd rcx, DWORD PTR [rsp+4]
add rax, rcx ;XXSum
mov rsi, arg(4) ;SSE
mov rdi, arg(5) ;Sum
mov dword ptr [rsi], eax
mov dword ptr [rdi], edx
xor rax, rax ; return 0
; begin epilog
add rsp, 16
pop rbx
pop rdi
pop rsi
UNSHADOW_ARGS
pop rbp
ret
;unsigned int
Change encoder vp8_ and vp8cx_ public symbol prefixes to vp9_. Change-Id: Ie2e3652591b010ded10c216501ce24fd95d0aec5
2012-10-30 20:58:42 +01:00
;vp9_get4x4var_mmx
Initial WebM release
2010-05-18 17:58:33 +02:00
;(
; unsigned char *src_ptr,
; int source_stride,
; unsigned char *ref_ptr,
; int recon_stride,
; unsigned int *SSE,
; int *Sum
;)
add private to assembly files to insure proper chromebuild Change-Id: I6e43ca73f35401a974ed8ee27738d4318f09fd37
2012-12-20 18:40:18 +01:00
global sym(vp9_get4x4var_mmx) PRIVATE
Change encoder vp8_ and vp8cx_ public symbol prefixes to vp9_. Change-Id: Ie2e3652591b010ded10c216501ce24fd95d0aec5
2012-10-30 20:58:42 +01:00
sym(vp9_get4x4var_mmx):
Initial WebM release
2010-05-18 17:58:33 +02:00
push rbp
mov rbp, rsp
SHADOW_ARGS_TO_STACK 6
push rsi
push rdi
push rbx
sub rsp, 16
; end prolog
pxor mm5, mm5 ; Blank mmx6
pxor mm6, mm6 ; Blank mmx7
pxor mm7, mm7 ; Blank mmx7
mov rax, arg(0) ;[src_ptr] ; Load base addresses
mov rbx, arg(2) ;[ref_ptr]
movsxd rcx, dword ptr arg(1) ;[source_stride]
movsxd rdx, dword ptr arg(3) ;[recon_stride]
; Row 1
movq mm0, [rax] ; Copy eight bytes to mm0
movq mm1, [rbx] ; Copy eight bytes to mm1
punpcklbw mm0, mm6 ; unpack to higher prrcision
punpcklbw mm1, mm6
psubsw mm0, mm1 ; A-B (low order) to MM0
paddw mm5, mm0 ; accumulate differences in mm5
pmaddwd mm0, mm0 ; square and accumulate
add rbx,rdx ; Inc pointer into ref data
add rax,rcx ; Inc pointer into the new data
movq mm1, [rbx] ; Copy eight bytes to mm1
paddd mm7, mm0 ; accumulate in mm7
; Row 2
movq mm0, [rax] ; Copy eight bytes to mm0
punpcklbw mm0, mm6 ; unpack to higher prrcision
punpcklbw mm1, mm6
psubsw mm0, mm1 ; A-B (low order) to MM0
paddw mm5, mm0 ; accumulate differences in mm5
pmaddwd mm0, mm0 ; square and accumulate
add rbx,rdx ; Inc pointer into ref data
add rax,rcx ; Inc pointer into the new data
movq mm1, [rbx] ; Copy eight bytes to mm1
paddd mm7, mm0 ; accumulate in mm7
; Row 3
movq mm0, [rax] ; Copy eight bytes to mm0
punpcklbw mm0, mm6 ; unpack to higher prrcision
punpcklbw mm1, mm6
psubsw mm0, mm1 ; A-B (low order) to MM0
paddw mm5, mm0 ; accumulate differences in mm5
pmaddwd mm0, mm0 ; square and accumulate
add rbx,rdx ; Inc pointer into ref data
add rax,rcx ; Inc pointer into the new data
movq mm1, [rbx] ; Copy eight bytes to mm1
paddd mm7, mm0 ; accumulate in mm7
; Row 4
movq mm0, [rax] ; Copy eight bytes to mm0
punpcklbw mm0, mm6 ; unpack to higher prrcision
punpcklbw mm1, mm6
psubsw mm0, mm1 ; A-B (low order) to MM0
paddw mm5, mm0 ; accumulate differences in mm5
pmaddwd mm0, mm0 ; square and accumulate
paddd mm7, mm0 ; accumulate in mm7
; Now accumulate the final results.
movq QWORD PTR [rsp+8], mm5 ; copy back accumulated results into normal memory
movq QWORD PTR [rsp], mm7 ; copy back accumulated results into normal memory
movsx rdx, WORD PTR [rsp+8]
movsx rcx, WORD PTR [rsp+10]
movsx rbx, WORD PTR [rsp+12]
movsx rax, WORD PTR [rsp+14]
add rdx, rcx
add rbx, rax
add rdx, rbx ;XSum
movsxd rax, DWORD PTR [rsp]
movsxd rcx, DWORD PTR [rsp+4]
add rax, rcx ;XXSum
mov rsi, arg(4) ;SSE
mov rdi, arg(5) ;Sum
mov dword ptr [rsi], eax
mov dword ptr [rdi], edx
xor rax, rax ; return 0
; begin epilog
add rsp, 16
pop rbx
pop rdi
pop rsi
UNSHADOW_ARGS
pop rbp
ret
;unsigned int
Change encoder vp8_ and vp8cx_ public symbol prefixes to vp9_. Change-Id: Ie2e3652591b010ded10c216501ce24fd95d0aec5
2012-10-30 20:58:42 +01:00
;vp9_get4x4sse_cs_mmx
Initial WebM release
2010-05-18 17:58:33 +02:00
;(
; unsigned char *src_ptr,
; int source_stride,
; unsigned char *ref_ptr,
; int recon_stride
;)
add private to assembly files to insure proper chromebuild Change-Id: I6e43ca73f35401a974ed8ee27738d4318f09fd37
2012-12-20 18:40:18 +01:00
global sym(vp9_get4x4sse_cs_mmx) PRIVATE
Change encoder vp8_ and vp8cx_ public symbol prefixes to vp9_. Change-Id: Ie2e3652591b010ded10c216501ce24fd95d0aec5
2012-10-30 20:58:42 +01:00
sym(vp9_get4x4sse_cs_mmx):
Initial WebM release
2010-05-18 17:58:33 +02:00
push rbp
mov rbp, rsp
SHADOW_ARGS_TO_STACK 4
push rsi
push rdi
push rbx
; end prolog
pxor mm6, mm6 ; Blank mmx7
pxor mm7, mm7 ; Blank mmx7
mov rax, arg(0) ;[src_ptr] ; Load base addresses
mov rbx, arg(2) ;[ref_ptr]
movsxd rcx, dword ptr arg(1) ;[source_stride]
movsxd rdx, dword ptr arg(3) ;[recon_stride]
; Row 1
movd mm0, [rax] ; Copy eight bytes to mm0
movd mm1, [rbx] ; Copy eight bytes to mm1
punpcklbw mm0, mm6 ; unpack to higher prrcision
punpcklbw mm1, mm6
psubsw mm0, mm1 ; A-B (low order) to MM0
pmaddwd mm0, mm0 ; square and accumulate
add rbx,rdx ; Inc pointer into ref data
add rax,rcx ; Inc pointer into the new data
movd mm1, [rbx] ; Copy eight bytes to mm1
paddd mm7, mm0 ; accumulate in mm7
; Row 2
movd mm0, [rax] ; Copy eight bytes to mm0
punpcklbw mm0, mm6 ; unpack to higher prrcision
punpcklbw mm1, mm6
psubsw mm0, mm1 ; A-B (low order) to MM0
pmaddwd mm0, mm0 ; square and accumulate
add rbx,rdx ; Inc pointer into ref data
add rax,rcx ; Inc pointer into the new data
movd mm1, [rbx] ; Copy eight bytes to mm1
paddd mm7, mm0 ; accumulate in mm7
; Row 3
movd mm0, [rax] ; Copy eight bytes to mm0
punpcklbw mm1, mm6
punpcklbw mm0, mm6 ; unpack to higher prrcision
psubsw mm0, mm1 ; A-B (low order) to MM0
pmaddwd mm0, mm0 ; square and accumulate
add rbx,rdx ; Inc pointer into ref data
add rax,rcx ; Inc pointer into the new data
movd mm1, [rbx] ; Copy eight bytes to mm1
paddd mm7, mm0 ; accumulate in mm7
; Row 4
movd mm0, [rax] ; Copy eight bytes to mm0
punpcklbw mm0, mm6 ; unpack to higher prrcision
punpcklbw mm1, mm6
psubsw mm0, mm1 ; A-B (low order) to MM0
pmaddwd mm0, mm0 ; square and accumulate
paddd mm7, mm0 ; accumulate in mm7
movq mm0, mm7 ;
psrlq mm7, 32
paddd mm0, mm7
nasm: match instruction length (movd/movq) to parameters nasm requires the instruction length (movd/movq) to match to its parameters. I find it more clear to really use 64bit instructions when we use 64bit registers in the assembly. Provide nasm compatibility. No binary change by this patch with yasm on {x86_64,i686}-fedora13-linux-gnu. Few longer opcodes with nasm on {x86_64,i686}-fedora13-linux-gnu have been checked as safe. Change-Id: Id9b1a5cdfb1bc05697e523c317a296df43d42a91
2010-10-04 23:19:33 +02:00
movq rax, mm0
Initial WebM release
2010-05-18 17:58:33 +02:00
; begin epilog
pop rbx
pop rdi
pop rsi
UNSHADOW_ARGS
pop rbp
ret
Reference in New Issue Copy Permalink