Skip to content

F
ffmpeg.wasm-core
Commit 681a86ab authored by Diego Biurrun's avatar Diego Biurrun
Browse files
Options

x86: fft: Port to cpuflags

parent e9bb77fb
Show whitespace changes
Inline Side-by-side
Showing with 43 additions and 51 deletions
libavcodec/x86/fft.asm
View file @ 681a86ab
...@@ -190,6 +190,23 @@ SECTION .text ...@@ -190,6 +190,23 @@ SECTION .text
addps %2, %2, %5 ; {i0,i1,i2,i3} addps %2, %2, %5 ; {i0,i1,i2,i3}
%endmacro %endmacro
%macro INTERL 5
%if cpuflag(avx)
vunpckhps %3, %2, %1
vunpcklps %2, %2, %1
vextractf128 %4(%5), %2, 0
vextractf128 %4 %+ H(%5), %3, 0
vextractf128 %4(%5 + 1), %2, 1
vextractf128 %4 %+ H(%5 + 1), %3, 1
%elif cpuflag(sse)
mova %3, %2
unpcklps %2, %1
unpckhps %3, %1
mova %4(%5), %2
mova %4(%5+1), %3
%endif
%endmacro
; scheduled for cpu-bound sizes ; scheduled for cpu-bound sizes
%macro PASS_SMALL 3 ; (to load m4-m7), wre, wim %macro PASS_SMALL 3 ; (to load m4-m7), wre, wim
IF%1 mova m4, Z(4) IF%1 mova m4, Z(4)
...@@ -536,17 +553,6 @@ DEFINE_ARGS zc, w, n, o1, o3 ...@@ -536,17 +553,6 @@ DEFINE_ARGS zc, w, n, o1, o3
INIT_YMM avx INIT_YMM avx
%macro INTERL_AVX 5
vunpckhps %3, %2, %1
vunpcklps %2, %2, %1
vextractf128 %4(%5), %2, 0
vextractf128 %4 %+ H(%5), %3, 0
vextractf128 %4(%5 + 1), %2, 1
vextractf128 %4 %+ H(%5 + 1), %3, 1
%endmacro
%define INTERL INTERL_AVX
DECL_PASS pass_avx, PASS_BIG 1 DECL_PASS pass_avx, PASS_BIG 1
DECL_PASS pass_interleave_avx, PASS_BIG 0 DECL_PASS pass_interleave_avx, PASS_BIG 0
...@@ -560,16 +566,6 @@ cglobal fft_calc, 2,5,8 ...@@ -560,16 +566,6 @@ cglobal fft_calc, 2,5,8
INIT_XMM sse INIT_XMM sse
%macro INTERL_SSE 5
mova %3, %2
unpcklps %2, %1
unpckhps %3, %1
mova %4(%5), %2
mova %4(%5+1), %3
%endmacro
%define INTERL INTERL_SSE
DECL_PASS pass_sse, PASS_BIG 1 DECL_PASS pass_sse, PASS_BIG 1
DECL_PASS pass_interleave_sse, PASS_BIG 0 DECL_PASS pass_interleave_sse, PASS_BIG 0
...@@ -855,16 +851,30 @@ INIT_XMM sse ...@@ -855,16 +851,30 @@ INIT_XMM sse
%endmacro %endmacro
%macro CMUL 6 ;j, xmm0, xmm1, 3, 4, 5 %macro CMUL 6 ;j, xmm0, xmm1, 3, 4, 5
%if cpuflag(sse)
mulps m6, %3, [%5+%1] mulps m6, %3, [%5+%1]
mulps m7, %2, [%5+%1] mulps m7, %2, [%5+%1]
mulps %2, %2, [%6+%1] mulps %2, %2, [%6+%1]
mulps %3, %3, [%6+%1] mulps %3, %3, [%6+%1]
subps %2, %2, m6 subps %2, %2, m6
addps %3, %3, m7 addps %3, %3, m7
%elif cpuflag(3dnow)
mova m6, [%1+%2*2]
mova %3, [%1+%2*2+8]
mova %4, m6
mova m7, %3
pfmul m6, [%5+%2]
pfmul %3, [%6+%2]
pfmul %4, [%6+%2]
pfmul m7, [%5+%2]
pfsub %3, m6
pfadd %4, m7
%endif
%endmacro %endmacro
%macro POSROTATESHUF_AVX 5 ;j, k, z+n8, tcos+n8, tsin+n8 %macro POSROTATESHUF 5 ;j, k, z+n8, tcos+n8, tsin+n8
.post: .post:
%if cpuflag(avx)
vmovaps ymm1, [%3+%1*2] vmovaps ymm1, [%3+%1*2]
vmovaps ymm0, [%3+%1*2+0x20] vmovaps ymm0, [%3+%1*2+0x20]
vmovaps ymm3, [%3+%2*2] vmovaps ymm3, [%3+%2*2]
...@@ -893,10 +903,7 @@ INIT_XMM sse ...@@ -893,10 +903,7 @@ INIT_XMM sse
sub %2, 0x20 sub %2, 0x20
add %1, 0x20 add %1, 0x20
jl .post jl .post
%endmacro %elif cpuflag(sse)
%macro POSROTATESHUF 5 ;j, k, z+n8, tcos+n8, tsin+n8
.post:
movaps xmm1, [%3+%1*2] movaps xmm1, [%3+%1*2]
movaps xmm0, [%3+%1*2+0x10] movaps xmm0, [%3+%1*2+0x10]
CMUL %1, xmm0, xmm1, %3, %4, %5 CMUL %1, xmm0, xmm1, %3, %4, %5
...@@ -918,25 +925,9 @@ INIT_XMM sse ...@@ -918,25 +925,9 @@ INIT_XMM sse
sub %2, 0x10 sub %2, 0x10
add %1, 0x10 add %1, 0x10
jl .post jl .post
%endmacro %elif cpuflag(3dnow)
CMUL %3, %1, m0, m1, %4, %5
%macro CMUL_3DNOW 6 CMUL %3, %2, m2, m3, %4, %5
mova m6, [%1+%2*2]
mova %3, [%1+%2*2+8]
mova %4, m6
mova m7, %3
pfmul m6, [%5+%2]
pfmul %3, [%6+%2]
pfmul %4, [%6+%2]
pfmul m7, [%5+%2]
pfsub %3, m6
pfadd %4, m7
%endmacro
%macro POSROTATESHUF_3DNOW 5 ;j, k, z+n8, tcos+n8, tsin+n8
.post:
CMUL_3DNOW %3, %1, m0, m1, %4, %5
CMUL_3DNOW %3, %2, m2, m3, %4, %5
movd [%3+%1*2+ 0], m0 movd [%3+%1*2+ 0], m0
movd [%3+%2*2+12], m1 movd [%3+%2*2+12], m1
movd [%3+%2*2+ 0], m2 movd [%3+%2*2+ 0], m2
...@@ -952,9 +943,10 @@ INIT_XMM sse ...@@ -952,9 +943,10 @@ INIT_XMM sse
sub %2, 8 sub %2, 8
add %1, 8 add %1, 8
jl .post jl .post
%endif
%endmacro %endmacro
%macro DECL_IMDCT 1 %macro DECL_IMDCT 0
cglobal imdct_half, 3,12,8; FFTContext *s, FFTSample *output, const FFTSample *input cglobal imdct_half, 3,12,8; FFTContext *s, FFTSample *output, const FFTSample *input
%if ARCH_X86_64 %if ARCH_X86_64
%define rrevtab r7 %define rrevtab r7
...@@ -1060,7 +1052,7 @@ cglobal imdct_half, 3,12,8; FFTContext *s, FFTSample *output, const FFTSample *i ...@@ -1060,7 +1052,7 @@ cglobal imdct_half, 3,12,8; FFTContext *s, FFTSample *output, const FFTSample *i
neg r0 neg r0
mov r1, -mmsize mov r1, -mmsize
sub r1, r0 sub r1, r0
%1 r0, r1, r6, rtcos, rtsin POSROTATESHUF r0, r1, r6, rtcos, rtsin
%if ARCH_X86_64 == 0 %if ARCH_X86_64 == 0
add esp, 12 add esp, 12
%endif %endif
...@@ -1070,17 +1062,17 @@ cglobal imdct_half, 3,12,8; FFTContext *s, FFTSample *output, const FFTSample *i ...@@ -1070,17 +1062,17 @@ cglobal imdct_half, 3,12,8; FFTContext *s, FFTSample *output, const FFTSample *i
RET RET
%endmacro %endmacro
DECL_IMDCT POSROTATESHUF DECL_IMDCT
%if ARCH_X86_32 %if ARCH_X86_32
INIT_MMX 3dnow INIT_MMX 3dnow
DECL_IMDCT POSROTATESHUF_3DNOW DECL_IMDCT
INIT_MMX 3dnowext INIT_MMX 3dnowext
DECL_IMDCT POSROTATESHUF_3DNOW DECL_IMDCT
%endif %endif
INIT_YMM avx INIT_YMM avx
DECL_IMDCT POSROTATESHUF_AVX DECL_IMDCT
%endif ; CONFIG_MDCT %endif ; CONFIG_MDCT
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