Commit 5361e10a authored by Ronald S. Bultje's avatar Ronald S. Bultje

proresdsp: port x86 assembly to cpuflags.

parent e9da9a31
...@@ -82,8 +82,7 @@ section .text align=16 ...@@ -82,8 +82,7 @@ section .text align=16
; %1 = row or col (for rounding variable) ; %1 = row or col (for rounding variable)
; %2 = number of bits to shift at the end ; %2 = number of bits to shift at the end
; %3 = optimization %macro IDCT_1D 2
%macro IDCT_1D 3
; a0 = (W4 * row[0]) + (1 << (15 - 1)); ; a0 = (W4 * row[0]) + (1 << (15 - 1));
; a1 = a0; ; a1 = a0;
; a2 = a0; ; a2 = a0;
...@@ -330,8 +329,8 @@ section .text align=16 ...@@ -330,8 +329,8 @@ section .text align=16
; void prores_idct_put_10_<opt>(uint8_t *pixels, int stride, ; void prores_idct_put_10_<opt>(uint8_t *pixels, int stride,
; DCTELEM *block, const int16_t *qmat); ; DCTELEM *block, const int16_t *qmat);
%macro idct_put_fn 2 %macro idct_put_fn 1
cglobal prores_idct_put_10_%1, 4, 4, %2 cglobal prores_idct_put_10, 4, 4, %1
movsxd r1, r1d movsxd r1, r1d
pxor m15, m15 ; zero pxor m15, m15 ; zero
...@@ -347,7 +346,7 @@ cglobal prores_idct_put_10_%1, 4, 4, %2 ...@@ -347,7 +346,7 @@ cglobal prores_idct_put_10_%1, 4, 4, %2
pmullw m13,[r3+64] pmullw m13,[r3+64]
pmullw m12,[r3+96] pmullw m12,[r3+96]
IDCT_1D row, 17, %1 IDCT_1D row, 17
; transpose for second part of IDCT ; transpose for second part of IDCT
TRANSPOSE8x8W 8, 0, 1, 2, 4, 11, 9, 10, 3 TRANSPOSE8x8W 8, 0, 1, 2, 4, 11, 9, 10, 3
...@@ -362,7 +361,7 @@ cglobal prores_idct_put_10_%1, 4, 4, %2 ...@@ -362,7 +361,7 @@ cglobal prores_idct_put_10_%1, 4, 4, %2
; for (i = 0; i < 8; i++) ; for (i = 0; i < 8; i++)
; idctSparseColAdd(dest + i, line_size, block + i); ; idctSparseColAdd(dest + i, line_size, block + i);
IDCT_1D col, 20, %1 IDCT_1D col, 20
; clip/store ; clip/store
mova m6, [pw_512] mova m6, [pw_512]
...@@ -406,27 +405,25 @@ cglobal prores_idct_put_10_%1, 4, 4, %2 ...@@ -406,27 +405,25 @@ cglobal prores_idct_put_10_%1, 4, 4, %2
RET RET
%endmacro %endmacro
%macro signextend_sse2 3 ; dstlow, dsthigh, tmp %macro SIGNEXTEND 2-3 ; dstlow, dsthigh, tmp
%if cpuflag(sse4)
movhlps %2, %1
pmovsxwd %1, %1
pmovsxwd %2, %2
%else ; sse2
pxor %3, %3 pxor %3, %3
pcmpgtw %3, %1 pcmpgtw %3, %1
mova %2, %1 mova %2, %1
punpcklwd %1, %3 punpcklwd %1, %3
punpckhwd %2, %3 punpckhwd %2, %3
%endif
%endmacro %endmacro
%macro signextend_sse4 2-3 ; dstlow, dsthigh INIT_XMM sse2
movhlps %2, %1 idct_put_fn 16
pmovsxwd %1, %1 INIT_XMM sse4
pmovsxwd %2, %2 idct_put_fn 16
%endmacro INIT_XMM avx
idct_put_fn 16
INIT_XMM
%define SIGNEXTEND signextend_sse2
idct_put_fn sse2, 16
INIT_XMM
%define SIGNEXTEND signextend_sse4
idct_put_fn sse4, 16
INIT_AVX
idct_put_fn avx, 16
%endif %endif
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