highperfocused/ffmpeg
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

x86: sbrdsp/fft: reuse ps_neg constant

Browse Source 
Signed-off-by: Michael Niedermayer <michaelni@gmx.at>
This commit is contained in:
Christophe Gisquet 2014-08-06 07:43:43 +00:00 committed by Michael Niedermayer
parent 51dd80e751
commit 75837e9add
4 changed files with 9 additions and 4 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

2
libavcodec/x86/constants.c
View File

@ -56,3 +56,5 @@ DECLARE_ALIGNED(32, const ymm_reg, ff_pb_3) = { 0x0303030303030303ULL, 0x030
0x0303030303030303ULL, 0x0303030303030303ULL }; 0x0303030303030303ULL, 0x0303030303030303ULL };
DECLARE_ALIGNED(16, const xmm_reg, ff_pb_80) = { 0x8080808080808080ULL, 0x8080808080808080ULL }; DECLARE_ALIGNED(16, const xmm_reg, ff_pb_80) = { 0x8080808080808080ULL, 0x8080808080808080ULL };
DECLARE_ALIGNED(8, const uint64_t, ff_pb_FC) = 0xFCFCFCFCFCFCFCFCULL; DECLARE_ALIGNED(8, const uint64_t, ff_pb_FC) = 0xFCFCFCFCFCFCFCFCULL;
DECLARE_ALIGNED(16, const xmm_reg, ff_ps_neg) = { 0x8000000080000000ULL, 0x8000000080000000ULL };

2
libavcodec/x86/constants.h
View File

@ -55,4 +55,6 @@ extern const xmm_reg ff_pb_80;
extern const xmm_reg ff_pb_F8; extern const xmm_reg ff_pb_F8;
extern const uint64_t ff_pb_FC; extern const uint64_t ff_pb_FC;
extern const xmm_reg ff_ps_neg;
#endif /* AVCODEC_X86_CONSTANTS_H */ #endif /* AVCODEC_X86_CONSTANTS_H */

7
libavcodec/x86/fft.asm
View File

@ -68,9 +68,10 @@ perm1: dd 0x00, 0x02, 0x03, 0x01, 0x03, 0x00, 0x02, 0x01
perm2: dd 0x00, 0x01, 0x02, 0x03, 0x01, 0x00, 0x02, 0x03 perm2: dd 0x00, 0x01, 0x02, 0x03, 0x01, 0x00, 0x02, 0x03
ps_p1p1m1p1root2: dd 1.0, 1.0, -1.0, 1.0, M_SQRT1_2, M_SQRT1_2, M_SQRT1_2, M_SQRT1_2 ps_p1p1m1p1root2: dd 1.0, 1.0, -1.0, 1.0, M_SQRT1_2, M_SQRT1_2, M_SQRT1_2, M_SQRT1_2
ps_m1m1p1m1p1m1m1m1: dd 1<<31, 1<<31, 0, 1<<31, 0, 1<<31, 1<<31, 1<<31 ps_m1m1p1m1p1m1m1m1: dd 1<<31, 1<<31, 0, 1<<31, 0, 1<<31, 1<<31, 1<<31
ps_m1m1m1m1: times 4 dd 1<<31
ps_m1p1: dd 1<<31, 0 ps_m1p1: dd 1<<31, 0
cextern ps_neg
%assign i 16 %assign i 16
%rep 13 %rep 13
cextern cos_ %+ i cextern cos_ %+ i
@ -685,7 +686,7 @@ cglobal imdct_calc, 3,5,3
mov r2, r3 mov r2, r3
sub r3, mmsize sub r3, mmsize
neg r2 neg r2
mova m2, [ps_m1m1m1m1] mova m2, [ps_neg]
.loop: .loop:
%if mmsize == 8 %if mmsize == 8
PSWAPD m0, [r1 + r3] PSWAPD m0, [r1 + r3]
@ -998,7 +999,7 @@ cglobal imdct_half, 3,12,8; FFTContext *s, FFTSample *output, const FFTSample *i
sub r4, r3 sub r4, r3
%endif %endif
%if notcpuflag(3dnowext) && mmsize == 8 %if notcpuflag(3dnowext) && mmsize == 8
movd m7, [ps_m1m1m1m1] movd m7, [ps_neg]
%endif %endif
.pre: .pre:
%if ARCH_X86_64 == 0 %if ARCH_X86_64 == 0

2
libavcodec/x86/sbrdsp.asm
View File

@ -25,13 +25,13 @@ SECTION_RODATA
; mask equivalent for multiply by -1.0 1.0 ; mask equivalent for multiply by -1.0 1.0
ps_mask times 2 dd 1<<31, 0 ps_mask times 2 dd 1<<31, 0
ps_mask2 times 2 dd 0, 1<<31 ps_mask2 times 2 dd 0, 1<<31
ps_neg times 4 dd 1<<31
ps_noise0 times 2 dd 1.0, 0.0, ps_noise0 times 2 dd 1.0, 0.0,
ps_noise2 times 2 dd -1.0, 0.0 ps_noise2 times 2 dd -1.0, 0.0
ps_noise13 dd 0.0, 1.0, 0.0, -1.0 ps_noise13 dd 0.0, 1.0, 0.0, -1.0
dd 0.0, -1.0, 0.0, 1.0 dd 0.0, -1.0, 0.0, 1.0
dd 0.0, 1.0, 0.0, -1.0 dd 0.0, 1.0, 0.0, -1.0
cextern sbr_noise_table cextern sbr_noise_table
cextern ps_neg
SECTION_TEXT SECTION_TEXT