RepoMirrors
/
ffmpeg
mirror of https://git.ffmpeg.org/ffmpeg.git
1
0
Fork 0
Code Issues Projects Releases Wiki Activity

sse2 version of fdct_col. 

k8: 72->61 cycles, core2: 51->26 cycles.

Originally committed as revision 8966 to svn://svn.ffmpeg.org/ffmpeg/trunk
This commit is contained in:
Loren Merritt 2007-05-10 03:13:41 +00:00
parent a4461664cd
commit 7c3a9fe2a3
1 changed files with 94 additions and 92 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

186
libavcodec/i386/fdct_mmx.c
View File

@ -52,18 +52,20 @@
#define RND_FRW_ROW (1 << (SHIFT_FRW_ROW-1)) #define RND_FRW_ROW (1 << (SHIFT_FRW_ROW-1))
//#define RND_FRW_COL (1 << (SHIFT_FRW_COL-1)) //#define RND_FRW_COL (1 << (SHIFT_FRW_COL-1))
#define X8(x) x,x,x,x,x,x,x,x
//concatenated table, for forward DCT transformation //concatenated table, for forward DCT transformation
static const int16_t fdct_tg_all_16[] ATTR_ALIGN(8) = { static const int16_t fdct_tg_all_16[24] ATTR_ALIGN(16) = {
13036, 13036, 13036, 13036, // tg * (2<<16) + 0.5 X8(13036), // tg * (2<<16) + 0.5
27146, 27146, 27146, 27146, // tg * (2<<16) + 0.5 X8(27146), // tg * (2<<16) + 0.5
-21746, -21746, -21746, -21746, // tg * (2<<16) + 0.5 X8(-21746) // tg * (2<<16) + 0.5
}; };
static const int16_t ocos_4_16[4] ATTR_ALIGN(8) = { static const int16_t ocos_4_16[8] ATTR_ALIGN(16) = {
23170, 23170, 23170, 23170, //cos * (2<<15) + 0.5 X8(23170) //cos * (2<<15) + 0.5
}; };
static const int64_t fdct_one_corr ATTR_ALIGN(8) = 0x0001000100010001LL; static const int16_t fdct_one_corr[8] ATTR_ALIGN(16) = { X8(1) };
static const int32_t fdct_r_row[2] ATTR_ALIGN(8) = {RND_FRW_ROW, RND_FRW_ROW }; static const int32_t fdct_r_row[2] ATTR_ALIGN(8) = {RND_FRW_ROW, RND_FRW_ROW };
@ -283,86 +285,88 @@ TABLE_SSE2
TABLE_SSE2 TABLE_SSE2
}}; }};
#define FDCT_COL(cpu, mm, mov)\
static av_always_inline void fdct_col(const int16_t *in, int16_t *out, int offset) static av_always_inline void fdct_col_##cpu(const int16_t *in, int16_t *out, int offset)\
{ {\
movq_m2r(*(in + offset + 1 * 8), mm0); mov##_m2r(*(in + offset + 1 * 8), mm##0);\
movq_m2r(*(in + offset + 6 * 8), mm1); mov##_m2r(*(in + offset + 6 * 8), mm##1);\
movq_r2r(mm0, mm2); mov##_r2r(mm##0, mm##2);\
movq_m2r(*(in + offset + 2 * 8), mm3); mov##_m2r(*(in + offset + 2 * 8), mm##3);\
paddsw_r2r(mm1, mm0); paddsw_r2r(mm##1, mm##0);\
movq_m2r(*(in + offset + 5 * 8), mm4); mov##_m2r(*(in + offset + 5 * 8), mm##4);\
psllw_i2r(SHIFT_FRW_COL, mm0); psllw_i2r(SHIFT_FRW_COL, mm##0);\
movq_m2r(*(in + offset + 0 * 8), mm5); mov##_m2r(*(in + offset + 0 * 8), mm##5);\
paddsw_r2r(mm3, mm4); paddsw_r2r(mm##3, mm##4);\
paddsw_m2r(*(in + offset + 7 * 8), mm5); paddsw_m2r(*(in + offset + 7 * 8), mm##5);\
psllw_i2r(SHIFT_FRW_COL, mm4); psllw_i2r(SHIFT_FRW_COL, mm##4);\
movq_r2r(mm0, mm6); mov##_r2r(mm##0, mm##6);\
psubsw_r2r(mm1, mm2); psubsw_r2r(mm##1, mm##2);\
movq_m2r(*(fdct_tg_all_16 + 4), mm1); mov##_m2r(*(fdct_tg_all_16 + 8), mm##1);\
psubsw_r2r(mm4, mm0); psubsw_r2r(mm##4, mm##0);\
movq_m2r(*(in + offset + 3 * 8), mm7); mov##_m2r(*(in + offset + 3 * 8), mm##7);\
pmulhw_r2r(mm0, mm1); pmulhw_r2r(mm##0, mm##1);\
paddsw_m2r(*(in + offset + 4 * 8), mm7); paddsw_m2r(*(in + offset + 4 * 8), mm##7);\
psllw_i2r(SHIFT_FRW_COL, mm5); psllw_i2r(SHIFT_FRW_COL, mm##5);\
paddsw_r2r(mm4, mm6); paddsw_r2r(mm##4, mm##6);\
psllw_i2r(SHIFT_FRW_COL, mm7); psllw_i2r(SHIFT_FRW_COL, mm##7);\
movq_r2r(mm5, mm4); mov##_r2r(mm##5, mm##4);\
psubsw_r2r(mm7, mm5); psubsw_r2r(mm##7, mm##5);\
paddsw_r2r(mm5, mm1); paddsw_r2r(mm##5, mm##1);\
paddsw_r2r(mm7, mm4); paddsw_r2r(mm##7, mm##4);\
por_m2r(fdct_one_corr, mm1); por_m2r(*fdct_one_corr, mm##1);\
psllw_i2r(SHIFT_FRW_COL + 1, mm2); psllw_i2r(SHIFT_FRW_COL + 1, mm##2);\
pmulhw_m2r(*(fdct_tg_all_16 + 4), mm5); pmulhw_m2r(*(fdct_tg_all_16 + 8), mm##5);\
movq_r2r(mm4, mm7); mov##_r2r(mm##4, mm##7);\
psubsw_m2r(*(in + offset + 5 * 8), mm3); psubsw_m2r(*(in + offset + 5 * 8), mm##3);\
psubsw_r2r(mm6, mm4); psubsw_r2r(mm##6, mm##4);\
movq_r2m(mm1, *(out + offset + 2 * 8)); mov##_r2m(mm##1, *(out + offset + 2 * 8));\
paddsw_r2r(mm6, mm7); paddsw_r2r(mm##6, mm##7);\
movq_m2r(*(in + offset + 3 * 8), mm1); mov##_m2r(*(in + offset + 3 * 8), mm##1);\
psllw_i2r(SHIFT_FRW_COL + 1, mm3); psllw_i2r(SHIFT_FRW_COL + 1, mm##3);\
psubsw_m2r(*(in + offset + 4 * 8), mm1); psubsw_m2r(*(in + offset + 4 * 8), mm##1);\
movq_r2r(mm2, mm6); mov##_r2r(mm##2, mm##6);\
movq_r2m(mm4, *(out + offset + 4 * 8)); mov##_r2m(mm##4, *(out + offset + 4 * 8));\
paddsw_r2r(mm3, mm2); paddsw_r2r(mm##3, mm##2);\
pmulhw_m2r(*ocos_4_16, mm2); pmulhw_m2r(*ocos_4_16, mm##2);\
psubsw_r2r(mm3, mm6); psubsw_r2r(mm##3, mm##6);\
pmulhw_m2r(*ocos_4_16, mm6); pmulhw_m2r(*ocos_4_16, mm##6);\
psubsw_r2r(mm0, mm5); psubsw_r2r(mm##0, mm##5);\
por_m2r(fdct_one_corr, mm5); por_m2r(*fdct_one_corr, mm##5);\
psllw_i2r(SHIFT_FRW_COL, mm1); psllw_i2r(SHIFT_FRW_COL, mm##1);\
por_m2r(fdct_one_corr, mm2); por_m2r(*fdct_one_corr, mm##2);\
movq_r2r(mm1, mm4); mov##_r2r(mm##1, mm##4);\
movq_m2r(*(in + offset + 0 * 8), mm3); mov##_m2r(*(in + offset + 0 * 8), mm##3);\
paddsw_r2r(mm6, mm1); paddsw_r2r(mm##6, mm##1);\
psubsw_m2r(*(in + offset + 7 * 8), mm3); psubsw_m2r(*(in + offset + 7 * 8), mm##3);\
psubsw_r2r(mm6, mm4); psubsw_r2r(mm##6, mm##4);\
movq_m2r(*(fdct_tg_all_16 + 0), mm0); mov##_m2r(*(fdct_tg_all_16 + 0), mm##0);\
psllw_i2r(SHIFT_FRW_COL, mm3); psllw_i2r(SHIFT_FRW_COL, mm##3);\
movq_m2r(*(fdct_tg_all_16 + 8), mm6); mov##_m2r(*(fdct_tg_all_16 + 16), mm##6);\
pmulhw_r2r(mm1, mm0); pmulhw_r2r(mm##1, mm##0);\
movq_r2m(mm7, *(out + offset + 0 * 8)); mov##_r2m(mm##7, *(out + offset + 0 * 8));\
pmulhw_r2r(mm4, mm6); pmulhw_r2r(mm##4, mm##6);\
movq_r2m(mm5, *(out + offset + 6 * 8)); mov##_r2m(mm##5, *(out + offset + 6 * 8));\
movq_r2r(mm3, mm7); mov##_r2r(mm##3, mm##7);\
movq_m2r(*(fdct_tg_all_16 + 8), mm5); mov##_m2r(*(fdct_tg_all_16 + 16), mm##5);\
psubsw_r2r(mm2, mm7); psubsw_r2r(mm##2, mm##7);\
paddsw_r2r(mm2, mm3); paddsw_r2r(mm##2, mm##3);\
pmulhw_r2r(mm7, mm5); pmulhw_r2r(mm##7, mm##5);\
paddsw_r2r(mm3, mm0); paddsw_r2r(mm##3, mm##0);\
paddsw_r2r(mm4, mm6); paddsw_r2r(mm##4, mm##6);\
pmulhw_m2r(*(fdct_tg_all_16 + 0), mm3); pmulhw_m2r(*(fdct_tg_all_16 + 0), mm##3);\
por_m2r(fdct_one_corr, mm0); por_m2r(*fdct_one_corr, mm##0);\
paddsw_r2r(mm7, mm5); paddsw_r2r(mm##7, mm##5);\
psubsw_r2r(mm6, mm7); psubsw_r2r(mm##6, mm##7);\
movq_r2m(mm0, *(out + offset + 1 * 8)); mov##_r2m(mm##0, *(out + offset + 1 * 8));\
paddsw_r2r(mm4, mm5); paddsw_r2r(mm##4, mm##5);\
movq_r2m(mm7, *(out + offset + 3 * 8)); mov##_r2m(mm##7, *(out + offset + 3 * 8));\
psubsw_r2r(mm1, mm3); psubsw_r2r(mm##1, mm##3);\
movq_r2m(mm5, *(out + offset + 5 * 8)); mov##_r2m(mm##5, *(out + offset + 5 * 8));\
movq_r2m(mm3, *(out + offset + 7 * 8)); mov##_r2m(mm##3, *(out + offset + 7 * 8));\
} }
FDCT_COL(mmx, mm, movq)
FDCT_COL(sse2, xmm, movdqa)
static av_always_inline void fdct_row_sse2(const int16_t *in, int16_t *out) static av_always_inline void fdct_row_sse2(const int16_t *in, int16_t *out)
{ {
@ -524,8 +528,8 @@ void ff_fdct_mmx(int16_t *block)
const int16_t *table= tab_frw_01234567; const int16_t *table= tab_frw_01234567;
int i; int i;
fdct_col(block, block1, 0); fdct_col_mmx(block, block1, 0);
fdct_col(block, block1, 4); fdct_col_mmx(block, block1, 4);
for(i=8;i>0;i--) { for(i=8;i>0;i--) {
fdct_row_mmx(block1, block, table); fdct_row_mmx(block1, block, table);
@ -542,8 +546,8 @@ void ff_fdct_mmx2(int16_t *block)
const int16_t *table= tab_frw_01234567; const int16_t *table= tab_frw_01234567;
int i; int i;
fdct_col(block, block1, 0); fdct_col_mmx(block, block1, 0);
fdct_col(block, block1, 4); fdct_col_mmx(block, block1, 4);
for(i=8;i>0;i--) { for(i=8;i>0;i--) {
fdct_row_mmx2(block1, block, table); fdct_row_mmx2(block1, block, table);
@ -558,9 +562,7 @@ void ff_fdct_sse2(int16_t *block)
int64_t align_tmp[16] ATTR_ALIGN(16); int64_t align_tmp[16] ATTR_ALIGN(16);
int16_t * const block1= (int16_t*)align_tmp; int16_t * const block1= (int16_t*)align_tmp;
fdct_col(block, block1, 0); fdct_col_sse2(block, block1, 0);
fdct_col(block, block1, 4);
fdct_row_sse2(block1, block); fdct_row_sse2(block1, block);
} }