vpx/vp8/common/x86/mask_sse3.asm
Yaowu Xu 6035da5448 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
2012-03-15 07:36:47 -07:00

485 lines
12 KiB
NASM

;
; 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"
;void int vp8_makemask_sse3(
; unsigned char *y,
; unsigned char *u,
; unsigned char *v,
; unsigned char *ym,
; unsigned char *uvm,
; int yp,
; int uvp,
; int ys,
; int us,
; int vs,
; int yt,
; int ut,
; int vt)
global sym(vp8_makemask_sse3)
sym(vp8_makemask_sse3):
push rbp
mov rbp, rsp
SHADOW_ARGS_TO_STACK 14
push rsi
push rdi
; end prolog
mov rsi, arg(0) ;y
mov rdi, arg(1) ;u
mov rcx, arg(2) ;v
mov rax, arg(3) ;ym
movsxd rbx, dword arg(4) ;yp
movsxd rdx, dword arg(5) ;uvp
pxor xmm0,xmm0
;make 16 copies of the center y value
movd xmm1, arg(6)
pshufb xmm1, xmm0
; make 16 copies of the center u value
movd xmm2, arg(7)
pshufb xmm2, xmm0
; make 16 copies of the center v value
movd xmm3, arg(8)
pshufb xmm3, xmm0
unpcklpd xmm2, xmm3
;make 16 copies of the y tolerance
movd xmm3, arg(9)
pshufb xmm3, xmm0
;make 16 copies of the u tolerance
movd xmm4, arg(10)
pshufb xmm4, xmm0
;make 16 copies of the v tolerance
movd xmm5, arg(11)
pshufb xmm5, xmm0
unpckhpd xmm4, xmm5
mov r8,8
NextPairOfRows:
;grab the y source values
movdqu xmm0, [rsi]
;compute abs difference between source and y target
movdqa xmm6, xmm1
movdqa xmm7, xmm0
psubusb xmm0, xmm1
psubusb xmm6, xmm7
por xmm0, xmm6
;compute abs difference between
movdqa xmm6, xmm3
pcmpgtb xmm6, xmm0
;grab the y source values
add rsi, rbx
movdqu xmm0, [rsi]
;compute abs difference between source and y target
movdqa xmm11, xmm1
movdqa xmm7, xmm0
psubusb xmm0, xmm1
psubusb xmm11, xmm7
por xmm0, xmm11
;compute abs difference between
movdqa xmm11, xmm3
pcmpgtb xmm11, xmm0
;grab the u and v source values
movdqu xmm7, [rdi]
movdqu xmm8, [rcx]
unpcklpd xmm7, xmm8
;compute abs difference between source and uv targets
movdqa xmm9, xmm2
movdqa xmm10, xmm7
psubusb xmm7, xmm2
psubusb xmm9, xmm10
por xmm7, xmm9
;check whether the number is < tolerance
movdqa xmm0, xmm4
pcmpgtb xmm0, xmm7
;double u and v masks
movdqa xmm8, xmm0
punpckhbw xmm0, xmm0
punpcklbw xmm8, xmm8
;mask row 0 and output
pand xmm6, xmm8
pand xmm6, xmm0
movdqa [rax],xmm6
;mask row 1 and output
pand xmm11, xmm8
pand xmm11, xmm0
movdqa [rax+16],xmm11
; to the next row or set of rows
add rsi, rbx
add rdi, rdx
add rcx, rdx
add rax,32
dec r8
jnz NextPairOfRows
; begin epilog
pop rdi
pop rsi
UNSHADOW_ARGS
pop rbp
ret
;GROW_HORIZ (register for result, source register or mem local)
; takes source and shifts left and ors with source
; then shifts right and ors with source
%macro GROW_HORIZ 2
movdqa %1, %2
movdqa xmm14, %1
movdqa xmm15, %1
pslldq xmm14, 1
psrldq xmm15, 1
por %1,xmm14
por %1,xmm15
%endmacro
;GROW_VERT (result, center row, above row, below row)
%macro GROW_VERT 4
movdqa %1,%2
por %1,%3
por %1,%4
%endmacro
;GROW_NEXTLINE (new line to grow, new source, line to write)
%macro GROW_NEXTLINE 3
GROW_HORIZ %1, %2
GROW_VERT xmm3, xmm0, xmm1, xmm2
movdqa %3,xmm3
%endmacro
;void int vp8_growmaskmb_sse3(
; unsigned char *om,
; unsigned char *nm,
global sym(vp8_growmaskmb_sse3)
sym(vp8_growmaskmb_sse3):
push rbp
mov rbp, rsp
SHADOW_ARGS_TO_STACK 2
push rsi
push rdi
; end prolog
mov rsi, arg(0) ;src
mov rdi, arg(1) ;rst
GROW_HORIZ xmm0, [rsi]
GROW_HORIZ xmm1, [rsi+16]
GROW_HORIZ xmm2, [rsi+32]
GROW_VERT xmm3, xmm0, xmm1, xmm2
por xmm0,xmm1
movdqa [rdi], xmm0
movdqa [rdi+16],xmm3
GROW_NEXTLINE xmm0,[rsi+48],[rdi+32]
GROW_NEXTLINE xmm1,[rsi+64],[rdi+48]
GROW_NEXTLINE xmm2,[rsi+80],[rdi+64]
GROW_NEXTLINE xmm0,[rsi+96],[rdi+80]
GROW_NEXTLINE xmm1,[rsi+112],[rdi+96]
GROW_NEXTLINE xmm2,[rsi+128],[rdi+112]
GROW_NEXTLINE xmm0,[rsi+144],[rdi+128]
GROW_NEXTLINE xmm1,[rsi+160],[rdi+144]
GROW_NEXTLINE xmm2,[rsi+176],[rdi+160]
GROW_NEXTLINE xmm0,[rsi+192],[rdi+176]
GROW_NEXTLINE xmm1,[rsi+208],[rdi+192]
GROW_NEXTLINE xmm2,[rsi+224],[rdi+208]
GROW_NEXTLINE xmm0,[rsi+240],[rdi+224]
por xmm0,xmm2
movdqa [rdi+240], xmm0
; begin epilog
pop rdi
pop rsi
UNSHADOW_ARGS
pop rbp
ret
;unsigned int vp8_sad16x16_masked_wmt(
; unsigned char *src_ptr,
; int src_stride,
; unsigned char *ref_ptr,
; int ref_stride,
; unsigned char *mask)
global sym(vp8_sad16x16_masked_wmt)
sym(vp8_sad16x16_masked_wmt):
push rbp
mov rbp, rsp
SHADOW_ARGS_TO_STACK 5
push rsi
push rdi
; end prolog
mov rsi, arg(0) ;src_ptr
mov rdi, arg(2) ;ref_ptr
mov rbx, arg(4) ;mask
movsxd rax, dword ptr arg(1) ;src_stride
movsxd rdx, dword ptr arg(3) ;ref_stride
mov rcx, 16
pxor xmm3, xmm3
NextSadRow:
movdqu xmm0, [rsi]
movdqu xmm1, [rdi]
movdqu xmm2, [rbx]
pand xmm0, xmm2
pand xmm1, xmm2
psadbw xmm0, xmm1
paddw xmm3, xmm0
add rsi, rax
add rdi, rdx
add rbx, 16
dec rcx
jnz NextSadRow
movdqa xmm4 , xmm3
psrldq xmm4, 8
paddw xmm3, xmm4
movq rax, xmm3
; begin epilog
pop rdi
pop rsi
UNSHADOW_ARGS
pop rbp
ret
;unsigned int vp8_sad16x16_unmasked_wmt(
; unsigned char *src_ptr,
; int src_stride,
; unsigned char *ref_ptr,
; int ref_stride,
; unsigned char *mask)
global sym(vp8_sad16x16_unmasked_wmt)
sym(vp8_sad16x16_unmasked_wmt):
push rbp
mov rbp, rsp
SHADOW_ARGS_TO_STACK 5
push rsi
push rdi
; end prolog
mov rsi, arg(0) ;src_ptr
mov rdi, arg(2) ;ref_ptr
mov rbx, arg(4) ;mask
movsxd rax, dword ptr arg(1) ;src_stride
movsxd rdx, dword ptr arg(3) ;ref_stride
mov rcx, 16
pxor xmm3, xmm3
next_vp8_sad16x16_unmasked_wmt:
movdqu xmm0, [rsi]
movdqu xmm1, [rdi]
movdqu xmm2, [rbx]
por xmm0, xmm2
por xmm1, xmm2
psadbw xmm0, xmm1
paddw xmm3, xmm0
add rsi, rax
add rdi, rdx
add rbx, 16
dec rcx
jnz next_vp8_sad16x16_unmasked_wmt
movdqa xmm4 , xmm3
psrldq xmm4, 8
paddw xmm3, xmm4
movq rax, xmm3
; begin epilog
pop rdi
pop rsi
UNSHADOW_ARGS
pop rbp
ret
;unsigned int vp8_masked_predictor_wmt(
; unsigned char *masked,
; unsigned char *unmasked,
; int src_stride,
; unsigned char *dst_ptr,
; int dst_stride,
; unsigned char *mask)
global sym(vp8_masked_predictor_wmt)
sym(vp8_masked_predictor_wmt):
push rbp
mov rbp, rsp
SHADOW_ARGS_TO_STACK 6
push rsi
push rdi
; end prolog
mov rsi, arg(0) ;src_ptr
mov rdi, arg(1) ;ref_ptr
mov rbx, arg(5) ;mask
movsxd rax, dword ptr arg(2) ;src_stride
mov r11, arg(3) ; destination
movsxd rdx, dword ptr arg(4) ;dst_stride
mov rcx, 16
pxor xmm3, xmm3
next_vp8_masked_predictor_wmt:
movdqu xmm0, [rsi]
movdqu xmm1, [rdi]
movdqu xmm2, [rbx]
pand xmm0, xmm2
pandn xmm2, xmm1
por xmm0, xmm2
movdqu [r11], xmm0
add r11, rdx
add rsi, rax
add rdi, rdx
add rbx, 16
dec rcx
jnz next_vp8_masked_predictor_wmt
; begin epilog
pop rdi
pop rsi
UNSHADOW_ARGS
pop rbp
ret
;unsigned int vp8_masked_predictor_uv_wmt(
; unsigned char *masked,
; unsigned char *unmasked,
; int src_stride,
; unsigned char *dst_ptr,
; int dst_stride,
; unsigned char *mask)
global sym(vp8_masked_predictor_uv_wmt)
sym(vp8_masked_predictor_uv_wmt):
push rbp
mov rbp, rsp
SHADOW_ARGS_TO_STACK 6
push rsi
push rdi
; end prolog
mov rsi, arg(0) ;src_ptr
mov rdi, arg(1) ;ref_ptr
mov rbx, arg(5) ;mask
movsxd rax, dword ptr arg(2) ;src_stride
mov r11, arg(3) ; destination
movsxd rdx, dword ptr arg(4) ;dst_stride
mov rcx, 8
pxor xmm3, xmm3
next_vp8_masked_predictor_uv_wmt:
movq xmm0, [rsi]
movq xmm1, [rdi]
movq xmm2, [rbx]
pand xmm0, xmm2
pandn xmm2, xmm1
por xmm0, xmm2
movq [r11], xmm0
add r11, rdx
add rsi, rax
add rdi, rax
add rbx, 8
dec rcx
jnz next_vp8_masked_predictor_uv_wmt
; begin epilog
pop rdi
pop rsi
UNSHADOW_ARGS
pop rbp
ret
;unsigned int vp8_uv_from_y_mask(
; unsigned char *ymask,
; unsigned char *uvmask)
global sym(vp8_uv_from_y_mask)
sym(vp8_uv_from_y_mask):
push rbp
mov rbp, rsp
SHADOW_ARGS_TO_STACK 6
push rsi
push rdi
; end prolog
mov rsi, arg(0) ;src_ptr
mov rdi, arg(1) ;dst_ptr
mov rcx, 8
pxor xmm3, xmm3
next_p8_uv_from_y_mask:
movdqu xmm0, [rsi]
pshufb xmm0, [shuf1b] ;[GLOBAL(shuf1b)]
movq [rdi],xmm0
add rdi, 8
add rsi,32
dec rcx
jnz next_p8_uv_from_y_mask
; begin epilog
pop rdi
pop rsi
UNSHADOW_ARGS
pop rbp
ret
SECTION_RODATA
align 16
shuf1b:
db 0, 2, 4, 6, 8, 10, 12, 14, 0, 0, 0, 0, 0, 0, 0, 0