Commit 5fecfb7d authored by Loren Merritt's avatar Loren Merritt
Browse files

clear_block mmx

Originally committed as revision 16045 to svn://svn.ffmpeg.org/ffmpeg/trunk
parent 5fac2776
......@@ -3420,6 +3420,11 @@ void ff_set_cmp(DSPContext* c, me_cmp_func *cmp, int type){
}
}
static void clear_block_c(DCTELEM *block)
{
memset(block, 0, sizeof(DCTELEM)*64);
}
/**
* memset(blocks, 0, sizeof(DCTELEM)*6*64)
*/
......@@ -4288,6 +4293,7 @@ void dsputil_init(DSPContext* c, AVCodecContext *avctx)
c->sum_abs_dctelem = sum_abs_dctelem_c;
c->gmc1 = gmc1_c;
c->gmc = ff_gmc_c;
c->clear_block = clear_block_c;
c->clear_blocks = clear_blocks_c;
c->pix_sum = pix_sum_c;
c->pix_norm1 = pix_norm1_c;
......
......@@ -203,6 +203,7 @@ typedef struct DSPContext {
*/
void (*gmc )(uint8_t *dst/*align 8*/, uint8_t *src/*align 1*/, int stride, int h, int ox, int oy,
int dxx, int dxy, int dyx, int dyy, int shift, int r, int width, int height);
void (*clear_block)(DCTELEM *block/*align 16*/);
void (*clear_blocks)(DCTELEM *blocks/*align 16*/);
int (*pix_sum)(uint8_t * pix, int line_size);
int (*pix_norm1)(uint8_t * pix, int line_size);
......
......@@ -810,7 +810,7 @@ static inline int get_p_cbp(MpegEncContext * s,
for (i = 0; i < 6; i++) {
if (s->block_last_index[i] >= 0 && ((cbp >> (5 - i))&1)==0 ){
s->block_last_index[i]= -1;
memset(s->block[i], 0, sizeof(DCTELEM)*64);
s->dsp.clear_block(s->block[i]);
}
}
}else{
......@@ -853,7 +853,7 @@ static inline int get_b_cbp(MpegEncContext * s, DCTELEM block[6][64],
for (i = 0; i < 6; i++) {
if (s->block_last_index[i] >= 0 && ((cbp >> (5 - i))&1)==0 ){
s->block_last_index[i]= -1;
memset(s->block[i], 0, sizeof(DCTELEM)*64);
s->dsp.clear_block(s->block[i]);
}
}
}else{
......@@ -4651,7 +4651,7 @@ retry:
rl = &rl_intra_aic;
i = 0;
s->gb= gb;
memset(block, 0, sizeof(DCTELEM)*64);
s->dsp.clear_block(block);
goto retry;
}
av_log(s->avctx, AV_LOG_ERROR, "run overflow at %dx%d i:%d\n", s->mb_x, s->mb_y, s->mb_intra);
......
......@@ -464,21 +464,42 @@ static void avg_pixels16_sse2(uint8_t *block, const uint8_t *pixels, int line_si
);
}
static void clear_blocks_mmx(DCTELEM *blocks)
#define CLEAR_BLOCKS(name,n) \
static void name(DCTELEM *blocks)\
{\
__asm__ volatile(\
"pxor %%mm7, %%mm7 \n\t"\
"mov %1, %%"REG_a" \n\t"\
"1: \n\t"\
"movq %%mm7, (%0, %%"REG_a") \n\t"\
"movq %%mm7, 8(%0, %%"REG_a") \n\t"\
"movq %%mm7, 16(%0, %%"REG_a") \n\t"\
"movq %%mm7, 24(%0, %%"REG_a") \n\t"\
"add $32, %%"REG_a" \n\t"\
" js 1b \n\t"\
: : "r" (((uint8_t *)blocks)+128*n),\
"i" (-128*n)\
: "%"REG_a\
);\
}
CLEAR_BLOCKS(clear_blocks_mmx, 6)
CLEAR_BLOCKS(clear_block_mmx, 1)
static void clear_block_sse(DCTELEM *block)
{
__asm__ volatile(
"pxor %%mm7, %%mm7 \n\t"
"mov $-128*6, %%"REG_a" \n\t"
"1: \n\t"
"movq %%mm7, (%0, %%"REG_a") \n\t"
"movq %%mm7, 8(%0, %%"REG_a") \n\t"
"movq %%mm7, 16(%0, %%"REG_a") \n\t"
"movq %%mm7, 24(%0, %%"REG_a") \n\t"
"add $32, %%"REG_a" \n\t"
" js 1b \n\t"
: : "r" (((uint8_t *)blocks)+128*6)
: "%"REG_a
);
"xorps %%xmm0, %%xmm0 \n"
"movaps %%xmm0, (%0) \n"
"movaps %%xmm0, 16(%0) \n"
"movaps %%xmm0, 32(%0) \n"
"movaps %%xmm0, 48(%0) \n"
"movaps %%xmm0, 64(%0) \n"
"movaps %%xmm0, 80(%0) \n"
"movaps %%xmm0, 96(%0) \n"
"movaps %%xmm0, 112(%0) \n"
:: "r"(block)
: "memory"
);
}
static void add_bytes_mmx(uint8_t *dst, uint8_t *src, int w){
......@@ -2569,7 +2590,10 @@ void dsputil_init_mmx(DSPContext* c, AVCodecContext *avctx)
c->put_pixels_clamped = put_pixels_clamped_mmx;
c->put_signed_pixels_clamped = put_signed_pixels_clamped_mmx;
c->add_pixels_clamped = add_pixels_clamped_mmx;
c->clear_block = clear_block_mmx;
c->clear_blocks = clear_blocks_mmx;
if (mm_flags & FF_MM_SSE)
c->clear_block = clear_block_sse;
#define SET_HPEL_FUNCS(PFX, IDX, SIZE, CPU) \
c->PFX ## _pixels_tab[IDX][0] = PFX ## _pixels ## SIZE ## _ ## CPU; \
......
......@@ -511,7 +511,7 @@ static int x8_decode_intra_mb(IntraX8Context* const w, const int chroma){
int sign;
assert(w->orient<12);
memset(s->block[0],0x00,64*sizeof(DCTELEM));
s->dsp.clear_block(s->block[0]);
if(chroma){
dc_mode=2;
......
......@@ -163,7 +163,7 @@ static int vlc_decode_block(MimicContext *ctx, int num_coeffs, int qscale)
DCTELEM *block = ctx->dct_block;
unsigned int pos;
memset(block, 0, 64 * sizeof(DCTELEM));
ctx->dsp.clear_block(block);
block[0] = get_bits(&ctx->gb, 8) << 3;
......
......@@ -444,7 +444,7 @@ static int decode_dc_progressive(MJpegDecodeContext *s, DCTELEM *block, int comp
int dc_index, int16_t *quant_matrix, int Al)
{
int val;
memset(block, 0, 64*sizeof(DCTELEM));
s->dsp.clear_block(block);
val = mjpeg_decode_dc(s, dc_index);
if (val == 0xffff) {
av_log(s->avctx, AV_LOG_ERROR, "error dc\n");
......@@ -800,7 +800,7 @@ static int mjpeg_decode_scan(MJpegDecodeContext *s, int nb_components, int Ah, i
if(s->interlaced && s->bottom_field)
ptr += linesize[c] >> 1;
if(!s->progressive) {
memset(s->block, 0, sizeof(s->block));
s->dsp.clear_block(s->block);
if(decode_block(s, s->block, i,
s->dc_index[i], s->ac_index[i],
s->quant_matrixes[ s->quant_index[c] ]) < 0) {
......
......@@ -1402,14 +1402,14 @@ static void render_slice(Vp3DecodeContext *s, int slice)
/* dequantize the DCT coefficients */
if(s->avctx->idct_algo==FF_IDCT_VP3){
Coeff *coeff= s->coeffs + i;
memset(block, 0, sizeof(block));
s->dsp.clear_block(block);
while(coeff->next){
block[coeff->index]= coeff->coeff * dequantizer[coeff->index];
coeff= coeff->next;
}
}else{
Coeff *coeff= s->coeffs + i;
memset(block, 0, sizeof(block));
s->dsp.clear_block(block);
while(coeff->next){
block[coeff->index]= (coeff->coeff * dequantizer[coeff->index] + 2)>>2;
coeff= coeff->next;
......
......@@ -405,7 +405,7 @@ static void vp56_decode_mb(vp56_context_t *s, int row, int col, int is_alpha)
mb_type = vp56_decode_mv(s, row, col);
ref_frame = vp56_reference_frame[mb_type];
memset(s->block_coeff, 0, sizeof(s->block_coeff));
s->dsp.clear_blocks(*s->block_coeff);
s->parse_coeff(s);
......
......@@ -43,12 +43,12 @@ static void wmv2_add_block(Wmv2Context *w, DCTELEM *block1, uint8_t *dst, int st
case 1:
ff_simple_idct84_add(dst , stride, block1);
ff_simple_idct84_add(dst + 4*stride, stride, w->abt_block2[n]);
memset(w->abt_block2[n], 0, 64*sizeof(DCTELEM));
s->dsp.clear_block(w->abt_block2[n]);
break;
case 2:
ff_simple_idct48_add(dst , stride, block1);
ff_simple_idct48_add(dst + 4 , stride, w->abt_block2[n]);
memset(w->abt_block2[n], 0, 64*sizeof(DCTELEM));
s->dsp.clear_block(w->abt_block2[n]);
break;
default:
av_log(s->avctx, AV_LOG_ERROR, "internal error in WMV2 abt\n");
......
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