Skip to content

F
ffmpeg.wasm-core
Commit 0e80265b authored by Ronald S. Bultje's avatar Ronald S. Bultje
Browse files
Options

vp9: refactor 10/12bpp dc-only code in 4x4/8x8 and add to 16x16.

parent 1338fb79
Hide whitespace changes
Inline Side-by-side
Showing with 79 additions and 56 deletions
libavcodec/x86/vp9itxfm_16bpp.asm
View file @ 0e80265b
...@@ -186,6 +186,17 @@ IWHT4_FN 12, 4095 ...@@ -186,6 +186,17 @@ IWHT4_FN 12, 4095
VP9_STORE_2X 2, 3, 6, 7, 4, 5 VP9_STORE_2X 2, 3, 6, 7, 4, 5
%endmacro %endmacro
%macro DC_ONLY 2 ; shift, zero
mov coefd, dword [blockq]
movd [blockq], %2
imul coefd, 11585
add coefd, 8192
sar coefd, 14
imul coefd, 11585
add coefd, ((1 << (%1 - 1)) << 14) + 8192
sar coefd, 14 + %1
%endmacro
; 4x4 coefficients are 5+depth+sign bits, so for 10bpp, everything still fits ; 4x4 coefficients are 5+depth+sign bits, so for 10bpp, everything still fits
; in 15+1 words without additional effort, since the coefficients are 15bpp. ; in 15+1 words without additional effort, since the coefficients are 15bpp.
...@@ -195,26 +206,20 @@ cglobal vp9_idct_idct_4x4_add_10, 4, 4, 8, dst, stride, block, eob ...@@ -195,26 +206,20 @@ cglobal vp9_idct_idct_4x4_add_10, 4, 4, 8, dst, stride, block, eob
jg .idctfull jg .idctfull
; dc-only ; dc-only
pxor m4, m4
%if cpuflag(ssse3) %if cpuflag(ssse3)
movd m0, [blockq] movd m0, [blockq]
movd [blockq], m4
mova m5, [pw_11585x2] mova m5, [pw_11585x2]
pmulhrsw m0, m5 pmulhrsw m0, m5
pmulhrsw m0, m5 pmulhrsw m0, m5
%else %else
DEFINE_ARGS dst, stride, block, coef DEFINE_ARGS dst, stride, block, coef
mov coefd, dword [blockq] DC_ONLY 4, m4
imul coefd, 11585
add coefd, 8192
sar coefd, 14
imul coefd, 11585
add coefd, (8 << 14) + 8192
sar coefd, 14 + 4
movd m0, coefd movd m0, coefd
%endif %endif
pshufw m0, m0, 0 pshufw m0, m0, 0
pxor m4, m4
mova m5, [pw_1023] mova m5, [pw_1023]
movh [blockq], m4
%if cpuflag(ssse3) %if cpuflag(ssse3)
pmulhrsw m0, [pw_2048] ; (x*2048 + (1<<14))>>15 <=> (x+8)>>4 pmulhrsw m0, [pw_2048] ; (x*2048 + (1<<14))>>15 <=> (x+8)>>4
%endif %endif
...@@ -368,19 +373,12 @@ cglobal vp9_idct_idct_4x4_add_12, 4, 4, 6, dst, stride, block, eob ...@@ -368,19 +373,12 @@ cglobal vp9_idct_idct_4x4_add_12, 4, 4, 6, dst, stride, block, eob
; dword. After the final shift (4), the result is 13+sign bits, so we ; dword. After the final shift (4), the result is 13+sign bits, so we
; don't need any additional processing to fit it in a word ; don't need any additional processing to fit it in a word
DEFINE_ARGS dst, stride, block, coef DEFINE_ARGS dst, stride, block, coef
mov coefd, dword [blockq] pxor m4, m4
imul coefd, 11585 DC_ONLY 4, m4
add coefd, 8192
sar coefd, 14
imul coefd, 11585
add coefd, (8 << 14) + 8192
sar coefd, 14 + 4
movd m0, coefd movd m0, coefd
pshuflw m0, m0, q0000 pshuflw m0, m0, q0000
punpcklqdq m0, m0 punpcklqdq m0, m0
pxor m4, m4
mova m5, [pw_4095] mova m5, [pw_4095]
movd [blockq], m4
DEFINE_ARGS dst, stride, stride3 DEFINE_ARGS dst, stride, stride3
lea stride3q, [strideq*3] lea stride3q, [strideq*3]
STORE_4x4 1, 3, 0, 0, m4, m5 STORE_4x4 1, 3, 0, 0, m4, m5
...@@ -559,17 +557,17 @@ IADST4_12BPP_FN iadst, IADST4, iadst, IADST4 ...@@ -559,17 +557,17 @@ IADST4_12BPP_FN iadst, IADST4, iadst, IADST4
SWAP 0, 5, 4, 6, 2, 7 SWAP 0, 5, 4, 6, 2, 7
%endmacro %endmacro
%macro STORE_2x8 5 ; tmp1-2, reg, min, max %macro STORE_2x8 5-7 dstq, strideq ; tmp1-2, reg, min, max
mova m%1, [dstq+strideq*0] mova m%1, [%6+%7*0]
mova m%2, [dstq+strideq*1] mova m%2, [%6+%7*1]
paddw m%1, m%3 paddw m%1, m%3
paddw m%2, m%3 paddw m%2, m%3
pmaxsw m%1, %4 pmaxsw m%1, %4
pmaxsw m%2, %4 pmaxsw m%2, %4
pminsw m%1, %5 pminsw m%1, %5
pminsw m%2, %5 pminsw m%2, %5
mova [dstq+strideq*0], m%1 mova [%6+%7*0], m%1
mova [dstq+strideq*1], m%2 mova [%6+%7*1], m%2
%endmacro %endmacro
%macro PRELOAD 2-3 %macro PRELOAD 2-3
...@@ -598,18 +596,11 @@ cglobal vp9_idct_idct_8x8_add_10, 4, 6 + ARCH_X86_64, 10, \ ...@@ -598,18 +596,11 @@ cglobal vp9_idct_idct_8x8_add_10, 4, 6 + ARCH_X86_64, 10, \
; coef values are 16+sign bit, and the coef is 14bit, so 30+sign easily ; coef values are 16+sign bit, and the coef is 14bit, so 30+sign easily
; fits in 32bit ; fits in 32bit
DEFINE_ARGS dst, stride, block, coef DEFINE_ARGS dst, stride, block, coef
mov coefd, dword [blockq] pxor m2, m2
imul coefd, 11585 DC_ONLY 5, m2
add coefd, 8192
sar coefd, 14
imul coefd, 11585
add coefd, (16 << 14) + 8192
sar coefd, 14 + 5
movd m1, coefd movd m1, coefd
pshuflw m1, m1, q0000 pshuflw m1, m1, q0000
punpcklqdq m1, m1 punpcklqdq m1, m1
pxor m2, m2
movd [blockq], m2
DEFINE_ARGS dst, stride, cnt DEFINE_ARGS dst, stride, cnt
mov cntd, 4 mov cntd, 4
.loop_dc: .loop_dc:
...@@ -676,32 +667,19 @@ cglobal vp9_idct_idct_8x8_add_10, 4, 6 + ARCH_X86_64, 10, \ ...@@ -676,32 +667,19 @@ cglobal vp9_idct_idct_8x8_add_10, 4, 6 + ARCH_X86_64, 10, \
ZERO_BLOCK blockq-2*mmsize, 32, 8, m6 ZERO_BLOCK blockq-2*mmsize, 32, 8, m6
RET RET
INIT_XMM sse2 %macro DC_ONLY_64BIT 2 ; shift, zero
cglobal vp9_idct_idct_8x8_add_12, 4, 6 + ARCH_X86_64, 10, \
17 * mmsize + 2 * ARCH_X86_32 * mmsize, \
dst, stride, block, eob
mova m0, [pw_4095]
cmp eobd, 1
jg mangle(private_prefix %+ _ %+ vp9_idct_idct_8x8_add_10 %+ SUFFIX).idctfull
; dc-only - unfortunately, this one can overflow, since coefs are 18+sign
; bpp, and 18+14+sign does not fit in 32bit, so we do 2-stage multiplies
DEFINE_ARGS dst, stride, block, coef, coefl
%if ARCH_X86_64 %if ARCH_X86_64
DEFINE_ARGS dst, stride, block, coef
movsxd coefq, dword [blockq] movsxd coefq, dword [blockq]
pxor m2, m2 movd [blockq], %2
movd [blockq], m2
imul coefq, 11585 imul coefq, 11585
add coefq, 8192 add coefq, 8192
sar coefq, 14 sar coefq, 14
imul coefq, 11585 imul coefq, 11585
add coefq, (16 << 14) + 8192 add coefq, ((1 << (%1 - 1)) << 14) + 8192
sar coefq, 14 + 5 sar coefq, 14 + %1
%else %else
mov coefd, dword [blockq] mov coefd, dword [blockq]
pxor m2, m2 movd [blockq], %2
movd [blockq], m2
DEFINE_ARGS dst, stride, cnt, coef, coefl DEFINE_ARGS dst, stride, cnt, coef, coefl
mov cntd, 2 mov cntd, 2
.loop_dc_calc: .loop_dc_calc:
...@@ -715,9 +693,24 @@ cglobal vp9_idct_idct_8x8_add_12, 4, 6 + ARCH_X86_64, 10, \ ...@@ -715,9 +693,24 @@ cglobal vp9_idct_idct_8x8_add_12, 4, 6 + ARCH_X86_64, 10, \
add coefd, coefld add coefd, coefld
dec cntd dec cntd
jg .loop_dc_calc jg .loop_dc_calc
add coefd, 16 add coefd, 1 << (%1 - 1)
sar coefd, 5 sar coefd, %1
%endif %endif
%endmacro
INIT_XMM sse2
cglobal vp9_idct_idct_8x8_add_12, 4, 6 + ARCH_X86_64, 10, \
17 * mmsize + 2 * ARCH_X86_32 * mmsize, \
dst, stride, block, eob
mova m0, [pw_4095]
cmp eobd, 1
jg mangle(private_prefix %+ _ %+ vp9_idct_idct_8x8_add_10 %+ SUFFIX).idctfull
; dc-only - unfortunately, this one can overflow, since coefs are 18+sign
; bpp, and 18+14+sign does not fit in 32bit, so we do 2-stage multiplies
DEFINE_ARGS dst, stride, block, coef, coefl
pxor m2, m2
DC_ONLY_64BIT 5, m2
movd m1, coefd movd m1, coefd
pshuflw m1, m1, q0000 pshuflw m1, m1, q0000
punpcklqdq m1, m1 punpcklqdq m1, m1
...@@ -1000,7 +993,24 @@ cglobal vp9_idct_idct_16x16_add_10, 4, 6 + ARCH_X86_64, 16, \ ...@@ -1000,7 +993,24 @@ cglobal vp9_idct_idct_16x16_add_10, 4, 6 + ARCH_X86_64, 16, \
cmp eobd, 1 cmp eobd, 1
jg .idctfull jg .idctfull
; dc-only ; dc-only - the 10bit version can be done entirely in 32bit, since the max
; coef values are 17+sign bit, and the coef is 14bit, so 31+sign easily
; fits in 32bit
DEFINE_ARGS dst, stride, block, coef
pxor m2, m2
DC_ONLY 6, m2
movd m1, coefd
pshuflw m1, m1, q0000
punpcklqdq m1, m1
DEFINE_ARGS dst, stride, cnt
mov cntd, 8
.loop_dc:
STORE_2x8 3, 4, 1, m2, m0, dstq, mmsize
STORE_2x8 3, 4, 1, m2, m0, dstq+strideq, mmsize
lea dstq, [dstq+strideq*2]
dec cntd
jg .loop_dc
RET
.idctfull: .idctfull:
mova [rsp+64*mmsize], m0 mova [rsp+64*mmsize], m0
...@@ -1013,7 +1023,6 @@ cglobal vp9_idct_idct_16x16_add_10, 4, 6 + ARCH_X86_64, 16, \ ...@@ -1013,7 +1023,6 @@ cglobal vp9_idct_idct_16x16_add_10, 4, 6 + ARCH_X86_64, 16, \
mov ptrq, rsp mov ptrq, rsp
.loop_1: .loop_1:
IDCT16_1D blockq IDCT16_1D blockq
; order: 2,1,0,11,3,7,9,10,6,8,4,5,12,13,r65,15
TRANSPOSE4x4D 0, 1, 2, 3, 7 TRANSPOSE4x4D 0, 1, 2, 3, 7
mova [ptrq+ 1*mmsize], m0 mova [ptrq+ 1*mmsize], m0
...@@ -1106,6 +1115,20 @@ cglobal vp9_idct_idct_16x16_add_12, 4, 6 + ARCH_X86_64, 16, \ ...@@ -1106,6 +1115,20 @@ cglobal vp9_idct_idct_16x16_add_12, 4, 6 + ARCH_X86_64, 16, \
cmp eobd, 1 cmp eobd, 1
jg mangle(private_prefix %+ _ %+ vp9_idct_idct_16x16_add_10 %+ SUFFIX).idctfull jg mangle(private_prefix %+ _ %+ vp9_idct_idct_16x16_add_10 %+ SUFFIX).idctfull
; dc-only ; dc-only - unfortunately, this one can overflow, since coefs are 19+sign
jmp mangle(private_prefix %+ _ %+ vp9_idct_idct_16x16_add_10 %+ SUFFIX).idctfull ; bpp, and 19+14+sign does not fit in 32bit, so we do 2-stage multiplies
DEFINE_ARGS dst, stride, block, coef, coefl
pxor m2, m2
DC_ONLY_64BIT 6, m2
movd m1, coefd
pshuflw m1, m1, q0000
punpcklqdq m1, m1
DEFINE_ARGS dst, stride, cnt
mov cntd, 8
.loop_dc:
STORE_2x8 3, 4, 1, m2, m0, dstq, mmsize
STORE_2x8 3, 4, 1, m2, m0, dstq+strideq, mmsize
lea dstq, [dstq+strideq*2]
dec cntd
jg .loop_dc
RET RET
Markdown is supported
0% or
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment