generic-library/vpx
1
0
Fork 0
Code Issues Pull Requests Releases Wiki Activity

Merge "tweak vp8_regular_quantize_b_sse2"

Browse Source
This commit is contained in:
Johann 2011-04-04 08:56:25 -07:00 committed by Code Review
commit 610dd90288
1 changed files with 42 additions and 52 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

94
vp8/encoder/x86/quantize_sse2.asm
View File

@ -22,35 +22,36 @@ sym(vp8_regular_quantize_b_sse2):
mov rbp, rsp mov rbp, rsp
SAVE_XMM SAVE_XMM
GET_GOT rbx GET_GOT rbx
push rsi
%if ABI_IS_32BIT %if ABI_IS_32BIT
push rdi push rdi
push rsi
%else %else
%ifidn __OUTPUT_FORMAT__,x64 %ifidn __OUTPUT_FORMAT__,x64
push rdi push rdi
push rsi
%endif %endif
%endif %endif
ALIGN_STACK 16, rax ALIGN_STACK 16, rax
%define BLOCKD_d 0 ; 8 %define zrun_zbin_boost 0 ; 8
%define zrun_zbin_boost 8 ; 8 %define abs_minus_zbin 8 ; 32
%define abs_minus_zbin 16 ; 32 %define temp_qcoeff 40 ; 32
%define temp_qcoeff 48 ; 32 %define qcoeff 72 ; 32
%define qcoeff 80 ; 32 %define stack_size 104
%define stack_size 112
sub rsp, stack_size sub rsp, stack_size
; end prolog ; end prolog
%if ABI_IS_32BIT %if ABI_IS_32BIT
mov rdi, arg(0) mov rdi, arg(0) ; BLOCK *b
mov rsi, arg(1) ; BLOCKD *d
%else %else
%ifidn __OUTPUT_FORMAT__,x64 %ifidn __OUTPUT_FORMAT__,x64
mov rdi, rcx ; BLOCK *b mov rdi, rcx ; BLOCK *b
mov [rsp + BLOCKD_d], rdx mov rsi, rdx ; BLOCKD *d
%else %else
;mov rdi, rdi ; BLOCK *b ;mov rdi, rdi ; BLOCK *b
mov [rsp + BLOCKD_d], rsi ;mov rsi, rsi ; BLOCKD *d
%endif %endif
%endif %endif
@ -125,59 +126,52 @@ sym(vp8_regular_quantize_b_sse2):
movdqa [rsp + qcoeff], xmm6 movdqa [rsp + qcoeff], xmm6
movdqa [rsp + qcoeff + 16], xmm6 movdqa [rsp + qcoeff + 16], xmm6
mov rsi, [rdi + vp8_block_zrun_zbin_boost] ; zbin_boost_ptr mov rdx, [rdi + vp8_block_zrun_zbin_boost] ; zbin_boost_ptr
mov rax, [rdi + vp8_block_quant_shift] ; quant_shift_ptr mov rax, [rdi + vp8_block_quant_shift] ; quant_shift_ptr
mov [rsp + zrun_zbin_boost], rsi mov [rsp + zrun_zbin_boost], rdx
%macro ZIGZAG_LOOP 1 %macro ZIGZAG_LOOP 1
movsx edx, WORD PTR[GLOBAL(zig_zag + (%1 * 2))] ; rc
; x ; x
movsx ecx, WORD PTR[rsp + abs_minus_zbin + rdx *2] movsx ecx, WORD PTR[rsp + abs_minus_zbin + %1 * 2]
; if (x >= zbin) ; if (x >= zbin)
sub cx, WORD PTR[rsi] ; x - zbin sub cx, WORD PTR[rdx] ; x - zbin
lea rsi, [rsi + 2] ; zbin_boost_ptr++ lea rdx, [rdx + 2] ; zbin_boost_ptr++
jl rq_zigzag_loop_%1 ; x < zbin jl rq_zigzag_loop_%1 ; x < zbin
movsx edi, WORD PTR[rsp + temp_qcoeff + rdx *2] movsx edi, WORD PTR[rsp + temp_qcoeff + %1 * 2]
; downshift by quant_shift[rdx] ; downshift by quant_shift[rc]
movsx ecx, WORD PTR[rax + rdx*2] ; quant_shift_ptr[rc] movsx ecx, WORD PTR[rax + %1 * 2] ; quant_shift_ptr[rc]
sar edi, cl ; also sets Z bit sar edi, cl ; also sets Z bit
je rq_zigzag_loop_%1 ; !y je rq_zigzag_loop_%1 ; !y
mov WORD PTR[rsp + qcoeff + rdx*2], di ;qcoeff_ptr[rc] = temp_qcoeff[rc] mov WORD PTR[rsp + qcoeff + %1 * 2], di ;qcoeff_ptr[rc] = temp_qcoeff[rc]
mov rsi, [rsp + zrun_zbin_boost] ; reset to b->zrun_zbin_boost mov rdx, [rsp + zrun_zbin_boost] ; reset to b->zrun_zbin_boost
rq_zigzag_loop_%1: rq_zigzag_loop_%1:
%endmacro %endmacro
ZIGZAG_LOOP 0 ; in vp8_default_zig_zag1d order: see vp8/common/entropy.c
ZIGZAG_LOOP 1 ZIGZAG_LOOP 0
ZIGZAG_LOOP 2 ZIGZAG_LOOP 1
ZIGZAG_LOOP 3 ZIGZAG_LOOP 4
ZIGZAG_LOOP 4 ZIGZAG_LOOP 8
ZIGZAG_LOOP 5 ZIGZAG_LOOP 5
ZIGZAG_LOOP 6 ZIGZAG_LOOP 2
ZIGZAG_LOOP 7 ZIGZAG_LOOP 3
ZIGZAG_LOOP 8 ZIGZAG_LOOP 6
ZIGZAG_LOOP 9 ZIGZAG_LOOP 9
ZIGZAG_LOOP 10
ZIGZAG_LOOP 11
ZIGZAG_LOOP 12 ZIGZAG_LOOP 12
ZIGZAG_LOOP 13 ZIGZAG_LOOP 13
ZIGZAG_LOOP 10
ZIGZAG_LOOP 7
ZIGZAG_LOOP 11
ZIGZAG_LOOP 14 ZIGZAG_LOOP 14
ZIGZAG_LOOP 15 ZIGZAG_LOOP 15
movdqa xmm2, [rsp + qcoeff] movdqa xmm2, [rsp + qcoeff]
movdqa xmm3, [rsp + qcoeff + 16] movdqa xmm3, [rsp + qcoeff + 16]
%if ABI_IS_32BIT mov rcx, [rsi + vp8_blockd_dequant] ; dequant_ptr
mov rdi, arg(1) mov rdi, [rsi + vp8_blockd_dqcoeff] ; dqcoeff_ptr
%else
mov rdi, [rsp + BLOCKD_d]
%endif
mov rcx, [rdi + vp8_blockd_dequant] ; dequant_ptr
mov rsi, [rdi + vp8_blockd_dqcoeff] ; dqcoeff_ptr
; y ^ sz ; y ^ sz
pxor xmm2, xmm0 pxor xmm2, xmm0
@ -190,15 +184,15 @@ ZIGZAG_LOOP 15
movdqa xmm0, [rcx] movdqa xmm0, [rcx]
movdqa xmm1, [rcx + 16] movdqa xmm1, [rcx + 16]
mov rcx, [rdi + vp8_blockd_qcoeff] ; qcoeff_ptr mov rcx, [rsi + vp8_blockd_qcoeff] ; qcoeff_ptr
pmullw xmm0, xmm2 pmullw xmm0, xmm2
pmullw xmm1, xmm3 pmullw xmm1, xmm3
movdqa [rcx], xmm2 ; store qcoeff movdqa [rcx], xmm2 ; store qcoeff
movdqa [rcx + 16], xmm3 movdqa [rcx + 16], xmm3
movdqa [rsi], xmm0 ; store dqcoeff movdqa [rdi], xmm0 ; store dqcoeff
movdqa [rsi + 16], xmm1 movdqa [rdi + 16], xmm1
; select the last value (in zig_zag order) for EOB ; select the last value (in zig_zag order) for EOB
pcmpeqw xmm2, xmm6 pcmpeqw xmm2, xmm6
@ -220,19 +214,20 @@ ZIGZAG_LOOP 15
pmaxsw xmm2, xmm3 pmaxsw xmm2, xmm3
movd eax, xmm2 movd eax, xmm2
and eax, 0xff and eax, 0xff
mov [rdi + vp8_blockd_eob], eax mov [rsi + vp8_blockd_eob], eax
; begin epilog ; begin epilog
add rsp, stack_size add rsp, stack_size
pop rsp pop rsp
%if ABI_IS_32BIT %if ABI_IS_32BIT
pop rsi
pop rdi pop rdi
%else %else
%ifidn __OUTPUT_FORMAT__,x64 %ifidn __OUTPUT_FORMAT__,x64
pop rsi
pop rdi pop rdi
%endif %endif
%endif %endif
pop rsi
RESTORE_GOT RESTORE_GOT
RESTORE_XMM RESTORE_XMM
pop rbp pop rbp
@ -347,11 +342,6 @@ sym(vp8_fast_quantize_b_impl_sse2):
SECTION_RODATA SECTION_RODATA
align 16 align 16
zig_zag:
dw 0x0000, 0x0001, 0x0004, 0x0008
dw 0x0005, 0x0002, 0x0003, 0x0006
dw 0x0009, 0x000c, 0x000d, 0x000a
dw 0x0007, 0x000b, 0x000e, 0x000f
inv_zig_zag: inv_zig_zag:
dw 0x0001, 0x0002, 0x0006, 0x0007 dw 0x0001, 0x0002, 0x0006, 0x0007
dw 0x0003, 0x0005, 0x0008, 0x000d dw 0x0003, 0x0005, 0x0008, 0x000d