static const unsigned long long int mm_wone __attribute__ ((aligned(8))) = 0x0001000100010001ULL;
-static void dct_unquantize_h263_mmx(MpegEncContext *s,
+static void dct_unquantize_h263_intra_mmx(MpegEncContext *s,
DCTELEM *block, int n, int qscale)
{
int level, qmul, qadd, nCoeffs;
qmul = qscale << 1;
- qadd = (qscale - 1) | 1;
assert(s->block_last_index[n]>=0 || s->h263_aic);
- if (s->mb_intra) {
- if (!s->h263_aic) {
- if (n < 4)
- level = block[0] * s->y_dc_scale;
- else
- level = block[0] * s->c_dc_scale;
- }else{
- qadd = 0;
- level= block[0];
- }
- nCoeffs=63;
- } else {
- nCoeffs= s->inter_scantable.raster_end[ s->block_last_index[n] ];
- level = 0;/* keep gcc quiet */
+ if (!s->h263_aic) {
+ if (n < 4)
+ level = block[0] * s->y_dc_scale;
+ else
+ level = block[0] * s->c_dc_scale;
+ qadd = (qscale - 1) | 1;
+ }else{
+ qadd = 0;
+ level= block[0];
}
+ if(s->ac_pred)
+ nCoeffs=63;
+ else
+ nCoeffs= s->inter_scantable.raster_end[ s->block_last_index[n] ];
+//printf("%d %d ", qmul, qadd);
+asm volatile(
+ "movd %1, %%mm6 \n\t" //qmul
+ "packssdw %%mm6, %%mm6 \n\t"
+ "packssdw %%mm6, %%mm6 \n\t"
+ "movd %2, %%mm5 \n\t" //qadd
+ "pxor %%mm7, %%mm7 \n\t"
+ "packssdw %%mm5, %%mm5 \n\t"
+ "packssdw %%mm5, %%mm5 \n\t"
+ "psubw %%mm5, %%mm7 \n\t"
+ "pxor %%mm4, %%mm4 \n\t"
+ ".balign 16\n\t"
+ "1: \n\t"
+ "movq (%0, %3), %%mm0 \n\t"
+ "movq 8(%0, %3), %%mm1 \n\t"
+
+ "pmullw %%mm6, %%mm0 \n\t"
+ "pmullw %%mm6, %%mm1 \n\t"
+
+ "movq (%0, %3), %%mm2 \n\t"
+ "movq 8(%0, %3), %%mm3 \n\t"
+
+ "pcmpgtw %%mm4, %%mm2 \n\t" // block[i] < 0 ? -1 : 0
+ "pcmpgtw %%mm4, %%mm3 \n\t" // block[i] < 0 ? -1 : 0
+
+ "pxor %%mm2, %%mm0 \n\t"
+ "pxor %%mm3, %%mm1 \n\t"
+
+ "paddw %%mm7, %%mm0 \n\t"
+ "paddw %%mm7, %%mm1 \n\t"
+
+ "pxor %%mm0, %%mm2 \n\t"
+ "pxor %%mm1, %%mm3 \n\t"
+
+ "pcmpeqw %%mm7, %%mm0 \n\t" // block[i] == 0 ? -1 : 0
+ "pcmpeqw %%mm7, %%mm1 \n\t" // block[i] == 0 ? -1 : 0
+
+ "pandn %%mm2, %%mm0 \n\t"
+ "pandn %%mm3, %%mm1 \n\t"
+
+ "movq %%mm0, (%0, %3) \n\t"
+ "movq %%mm1, 8(%0, %3) \n\t"
+
+ "addl $16, %3 \n\t"
+ "jng 1b \n\t"
+ ::"r" (block+nCoeffs), "g"(qmul), "g" (qadd), "r" (2*(-nCoeffs))
+ : "memory"
+ );
+ block[0]= level;
+}
+
+
+static void dct_unquantize_h263_inter_mmx(MpegEncContext *s,
+ DCTELEM *block, int n, int qscale)
+{
+ int level, qmul, qadd, nCoeffs;
+
+ qmul = qscale << 1;
+ qadd = (qscale - 1) | 1;
+
+ assert(s->block_last_index[n]>=0 || s->h263_aic);
+
+ nCoeffs= s->inter_scantable.raster_end[ s->block_last_index[n] ];
//printf("%d %d ", qmul, qadd);
asm volatile(
"movd %1, %%mm6 \n\t" //qmul
::"r" (block+nCoeffs), "g"(qmul), "g" (qadd), "r" (2*(-nCoeffs))
: "memory"
);
- if(s->mb_intra)
- block[0]= level;
}
high3:low3 = low1*low2
high3 += tlow1
*/
-static void dct_unquantize_mpeg1_mmx(MpegEncContext *s,
+static void dct_unquantize_mpeg1_intra_mmx(MpegEncContext *s,
DCTELEM *block, int n, int qscale)
{
int nCoeffs;
const uint16_t *quant_matrix;
+ int block0;
assert(s->block_last_index[n]>=0);
nCoeffs= s->intra_scantable.raster_end[ s->block_last_index[n] ]+1;
- if (s->mb_intra) {
- int block0;
- if (n < 4)
- block0 = block[0] * s->y_dc_scale;
- else
- block0 = block[0] * s->c_dc_scale;
- /* XXX: only mpeg1 */
- quant_matrix = s->intra_matrix;
+ if (n < 4)
+ block0 = block[0] * s->y_dc_scale;
+ else
+ block0 = block[0] * s->c_dc_scale;
+ /* XXX: only mpeg1 */
+ quant_matrix = s->intra_matrix;
asm volatile(
"pcmpeqw %%mm7, %%mm7 \n\t"
"psrlw $15, %%mm7 \n\t"
::"r" (block+nCoeffs), "r"(quant_matrix+nCoeffs), "g" (qscale), "g" (-2*nCoeffs)
: "%eax", "memory"
);
- block[0]= block0;
+ block[0]= block0;
+}
+
+static void dct_unquantize_mpeg1_inter_mmx(MpegEncContext *s,
+ DCTELEM *block, int n, int qscale)
+{
+ int nCoeffs;
+ const uint16_t *quant_matrix;
+
+ assert(s->block_last_index[n]>=0);
+
+ nCoeffs= s->intra_scantable.raster_end[ s->block_last_index[n] ]+1;
- } else {
quant_matrix = s->inter_matrix;
asm volatile(
"pcmpeqw %%mm7, %%mm7 \n\t"
::"r" (block+nCoeffs), "r"(quant_matrix+nCoeffs), "g" (qscale), "g" (-2*nCoeffs)
: "%eax", "memory"
);
- }
-
}
-static void dct_unquantize_mpeg2_mmx(MpegEncContext *s,
+static void dct_unquantize_mpeg2_intra_mmx(MpegEncContext *s,
DCTELEM *block, int n, int qscale)
{
int nCoeffs;
const uint16_t *quant_matrix;
+ int block0;
assert(s->block_last_index[n]>=0);
if(s->alternate_scan) nCoeffs= 63; //FIXME
else nCoeffs= s->intra_scantable.raster_end[ s->block_last_index[n] ];
- if (s->mb_intra) {
- int block0;
- if (n < 4)
- block0 = block[0] * s->y_dc_scale;
- else
- block0 = block[0] * s->c_dc_scale;
- quant_matrix = s->intra_matrix;
+ if (n < 4)
+ block0 = block[0] * s->y_dc_scale;
+ else
+ block0 = block[0] * s->c_dc_scale;
+ quant_matrix = s->intra_matrix;
asm volatile(
"pcmpeqw %%mm7, %%mm7 \n\t"
"psrlw $15, %%mm7 \n\t"
::"r" (block+nCoeffs), "r"(quant_matrix+nCoeffs), "g" (qscale), "g" (-2*nCoeffs)
: "%eax", "memory"
);
- block[0]= block0;
+ block[0]= block0;
//Note, we dont do mismatch control for intra as errors cannot accumulate
+}
+
+static void dct_unquantize_mpeg2_inter_mmx(MpegEncContext *s,
+ DCTELEM *block, int n, int qscale)
+{
+ int nCoeffs;
+ const uint16_t *quant_matrix;
+
+ assert(s->block_last_index[n]>=0);
+
+ if(s->alternate_scan) nCoeffs= 63; //FIXME
+ else nCoeffs= s->intra_scantable.raster_end[ s->block_last_index[n] ];
- } else {
quant_matrix = s->inter_matrix;
asm volatile(
"pcmpeqw %%mm7, %%mm7 \n\t"
::"r" (block+nCoeffs), "r"(quant_matrix+nCoeffs), "g" (qscale), "r" (-2*nCoeffs)
: "%eax", "memory"
);
- }
}
/* draw the edges of width 'w' of an image of size width, height
if (mm_flags & MM_MMX) {
const int dct_algo = s->avctx->dct_algo;
- s->dct_unquantize_h263 = dct_unquantize_h263_mmx;
- s->dct_unquantize_mpeg1 = dct_unquantize_mpeg1_mmx;
- s->dct_unquantize_mpeg2 = dct_unquantize_mpeg2_mmx;
+ s->dct_unquantize_h263_intra = dct_unquantize_h263_intra_mmx;
+ s->dct_unquantize_h263_inter = dct_unquantize_h263_inter_mmx;
+ s->dct_unquantize_mpeg1_intra = dct_unquantize_mpeg1_intra_mmx;
+ s->dct_unquantize_mpeg1_inter = dct_unquantize_mpeg1_inter_mmx;
+ s->dct_unquantize_mpeg2_intra = dct_unquantize_mpeg2_intra_mmx;
+ s->dct_unquantize_mpeg2_inter = dct_unquantize_mpeg2_inter_mmx;
draw_edges = draw_edges_mmx;
#ifdef CONFIG_ENCODERS
static void encode_picture(MpegEncContext *s, int picture_number);
#endif //CONFIG_ENCODERS
-static void dct_unquantize_mpeg1_c(MpegEncContext *s,
+static void dct_unquantize_mpeg1_intra_c(MpegEncContext *s,
DCTELEM *block, int n, int qscale);
-static void dct_unquantize_mpeg2_c(MpegEncContext *s,
+static void dct_unquantize_mpeg1_inter_c(MpegEncContext *s,
DCTELEM *block, int n, int qscale);
-static void dct_unquantize_h263_c(MpegEncContext *s,
+static void dct_unquantize_mpeg2_intra_c(MpegEncContext *s,
+ DCTELEM *block, int n, int qscale);
+static void dct_unquantize_mpeg2_inter_c(MpegEncContext *s,
+ DCTELEM *block, int n, int qscale);
+static void dct_unquantize_h263_intra_c(MpegEncContext *s,
+ DCTELEM *block, int n, int qscale);
+static void dct_unquantize_h263_inter_c(MpegEncContext *s,
DCTELEM *block, int n, int qscale);
static void draw_edges_c(uint8_t *buf, int wrap, int width, int height, int w);
#ifdef CONFIG_ENCODERS
/* init common dct for both encoder and decoder */
int DCT_common_init(MpegEncContext *s)
{
- s->dct_unquantize_h263 = dct_unquantize_h263_c;
- s->dct_unquantize_mpeg1 = dct_unquantize_mpeg1_c;
- s->dct_unquantize_mpeg2 = dct_unquantize_mpeg2_c;
+ s->dct_unquantize_h263_intra = dct_unquantize_h263_intra_c;
+ s->dct_unquantize_h263_inter = dct_unquantize_h263_inter_c;
+ s->dct_unquantize_mpeg1_intra = dct_unquantize_mpeg1_intra_c;
+ s->dct_unquantize_mpeg1_inter = dct_unquantize_mpeg1_inter_c;
+ s->dct_unquantize_mpeg2_intra = dct_unquantize_mpeg2_intra_c;
+ s->dct_unquantize_mpeg2_inter = dct_unquantize_mpeg2_inter_c;
#ifdef CONFIG_ENCODERS
s->dct_quantize= dct_quantize_c;
/* set dequantizer, we cant do it during init as it might change for mpeg4
and we cant do it in the header decode as init isnt called for mpeg4 there yet */
- if(s->mpeg_quant || s->codec_id == CODEC_ID_MPEG2VIDEO)
- s->dct_unquantize = s->dct_unquantize_mpeg2;
- else if(s->out_format == FMT_H263)
- s->dct_unquantize = s->dct_unquantize_h263;
- else
- s->dct_unquantize = s->dct_unquantize_mpeg1;
+ if(s->mpeg_quant || s->codec_id == CODEC_ID_MPEG2VIDEO){
+ s->dct_unquantize_intra = s->dct_unquantize_mpeg2_intra;
+ s->dct_unquantize_inter = s->dct_unquantize_mpeg2_inter;
+ }else if(s->out_format == FMT_H263){
+ s->dct_unquantize_intra = s->dct_unquantize_h263_intra;
+ s->dct_unquantize_inter = s->dct_unquantize_h263_inter;
+ }else{
+ s->dct_unquantize_intra = s->dct_unquantize_mpeg1_intra;
+ s->dct_unquantize_inter = s->dct_unquantize_mpeg1_inter;
+ }
if(s->dct_error_sum){
assert(s->avctx->noise_reduction && s->encoding);
static inline void put_dct(MpegEncContext *s,
DCTELEM *block, int i, uint8_t *dest, int line_size, int qscale)
{
- s->dct_unquantize(s, block, i, qscale);
+ s->dct_unquantize_intra(s, block, i, qscale);
s->dsp.idct_put (dest, line_size, block);
}
DCTELEM *block, int i, uint8_t *dest, int line_size, int qscale)
{
if (s->block_last_index[i] >= 0) {
- s->dct_unquantize(s, block, i, qscale);
+ s->dct_unquantize_inter(s, block, i, qscale);
s->dsp.idct_add (dest, line_size, block);
}
#endif //CONFIG_ENCODERS
-static void dct_unquantize_mpeg1_c(MpegEncContext *s,
+static void dct_unquantize_mpeg1_intra_c(MpegEncContext *s,
DCTELEM *block, int n, int qscale)
{
int i, level, nCoeffs;
nCoeffs= s->block_last_index[n];
- if (s->mb_intra) {
- if (n < 4)
- block[0] = block[0] * s->y_dc_scale;
- else
- block[0] = block[0] * s->c_dc_scale;
- /* XXX: only mpeg1 */
- quant_matrix = s->intra_matrix;
- for(i=1;i<=nCoeffs;i++) {
- int j= s->intra_scantable.permutated[i];
- level = block[j];
- if (level) {
- if (level < 0) {
- level = -level;
- level = (int)(level * qscale * quant_matrix[j]) >> 3;
- level = (level - 1) | 1;
- level = -level;
- } else {
- level = (int)(level * qscale * quant_matrix[j]) >> 3;
- level = (level - 1) | 1;
- }
-#ifdef PARANOID
- if (level < -2048 || level > 2047)
- fprintf(stderr, "unquant error %d %d\n", i, level);
-#endif
- block[j] = level;
+ if (n < 4)
+ block[0] = block[0] * s->y_dc_scale;
+ else
+ block[0] = block[0] * s->c_dc_scale;
+ /* XXX: only mpeg1 */
+ quant_matrix = s->intra_matrix;
+ for(i=1;i<=nCoeffs;i++) {
+ int j= s->intra_scantable.permutated[i];
+ level = block[j];
+ if (level) {
+ if (level < 0) {
+ level = -level;
+ level = (int)(level * qscale * quant_matrix[j]) >> 3;
+ level = (level - 1) | 1;
+ level = -level;
+ } else {
+ level = (int)(level * qscale * quant_matrix[j]) >> 3;
+ level = (level - 1) | 1;
}
+ block[j] = level;
}
- } else {
- i = 0;
- quant_matrix = s->inter_matrix;
- for(;i<=nCoeffs;i++) {
- int j= s->intra_scantable.permutated[i];
- level = block[j];
- if (level) {
- if (level < 0) {
- level = -level;
- level = (((level << 1) + 1) * qscale *
- ((int) (quant_matrix[j]))) >> 4;
- level = (level - 1) | 1;
- level = -level;
- } else {
- level = (((level << 1) + 1) * qscale *
- ((int) (quant_matrix[j]))) >> 4;
- level = (level - 1) | 1;
- }
-#ifdef PARANOID
- if (level < -2048 || level > 2047)
- fprintf(stderr, "unquant error %d %d\n", i, level);
-#endif
- block[j] = level;
+ }
+}
+
+static void dct_unquantize_mpeg1_inter_c(MpegEncContext *s,
+ DCTELEM *block, int n, int qscale)
+{
+ int i, level, nCoeffs;
+ const uint16_t *quant_matrix;
+
+ nCoeffs= s->block_last_index[n];
+
+ quant_matrix = s->inter_matrix;
+ for(i=0; i<=nCoeffs; i++) {
+ int j= s->intra_scantable.permutated[i];
+ level = block[j];
+ if (level) {
+ if (level < 0) {
+ level = -level;
+ level = (((level << 1) + 1) * qscale *
+ ((int) (quant_matrix[j]))) >> 4;
+ level = (level - 1) | 1;
+ level = -level;
+ } else {
+ level = (((level << 1) + 1) * qscale *
+ ((int) (quant_matrix[j]))) >> 4;
+ level = (level - 1) | 1;
}
+ block[j] = level;
}
}
}
-static void dct_unquantize_mpeg2_c(MpegEncContext *s,
+static void dct_unquantize_mpeg2_intra_c(MpegEncContext *s,
DCTELEM *block, int n, int qscale)
{
int i, level, nCoeffs;
if(s->alternate_scan) nCoeffs= 63;
else nCoeffs= s->block_last_index[n];
- if (s->mb_intra) {
+ if (n < 4)
+ block[0] = block[0] * s->y_dc_scale;
+ else
+ block[0] = block[0] * s->c_dc_scale;
+ quant_matrix = s->intra_matrix;
+ for(i=1;i<=nCoeffs;i++) {
+ int j= s->intra_scantable.permutated[i];
+ level = block[j];
+ if (level) {
+ if (level < 0) {
+ level = -level;
+ level = (int)(level * qscale * quant_matrix[j]) >> 3;
+ level = -level;
+ } else {
+ level = (int)(level * qscale * quant_matrix[j]) >> 3;
+ }
+ block[j] = level;
+ }
+ }
+}
+
+static void dct_unquantize_mpeg2_inter_c(MpegEncContext *s,
+ DCTELEM *block, int n, int qscale)
+{
+ int i, level, nCoeffs;
+ const uint16_t *quant_matrix;
+ int sum=-1;
+
+ if(s->alternate_scan) nCoeffs= 63;
+ else nCoeffs= s->block_last_index[n];
+
+ quant_matrix = s->inter_matrix;
+ for(i=0; i<=nCoeffs; i++) {
+ int j= s->intra_scantable.permutated[i];
+ level = block[j];
+ if (level) {
+ if (level < 0) {
+ level = -level;
+ level = (((level << 1) + 1) * qscale *
+ ((int) (quant_matrix[j]))) >> 4;
+ level = -level;
+ } else {
+ level = (((level << 1) + 1) * qscale *
+ ((int) (quant_matrix[j]))) >> 4;
+ }
+ block[j] = level;
+ sum+=level;
+ }
+ }
+ block[63]^=sum&1;
+}
+
+static void dct_unquantize_h263_intra_c(MpegEncContext *s,
+ DCTELEM *block, int n, int qscale)
+{
+ int i, level, qmul, qadd;
+ int nCoeffs;
+
+ assert(s->block_last_index[n]>=0);
+
+ qmul = qscale << 1;
+
+ if (!s->h263_aic) {
if (n < 4)
block[0] = block[0] * s->y_dc_scale;
else
block[0] = block[0] * s->c_dc_scale;
- quant_matrix = s->intra_matrix;
- for(i=1;i<=nCoeffs;i++) {
- int j= s->intra_scantable.permutated[i];
- level = block[j];
- if (level) {
- if (level < 0) {
- level = -level;
- level = (int)(level * qscale * quant_matrix[j]) >> 3;
- level = -level;
- } else {
- level = (int)(level * qscale * quant_matrix[j]) >> 3;
- }
-#ifdef PARANOID
- if (level < -2048 || level > 2047)
- fprintf(stderr, "unquant error %d %d\n", i, level);
-#endif
- block[j] = level;
- }
- }
- } else {
- int sum=-1;
- i = 0;
- quant_matrix = s->inter_matrix;
- for(;i<=nCoeffs;i++) {
- int j= s->intra_scantable.permutated[i];
- level = block[j];
- if (level) {
- if (level < 0) {
- level = -level;
- level = (((level << 1) + 1) * qscale *
- ((int) (quant_matrix[j]))) >> 4;
- level = -level;
- } else {
- level = (((level << 1) + 1) * qscale *
- ((int) (quant_matrix[j]))) >> 4;
- }
-#ifdef PARANOID
- if (level < -2048 || level > 2047)
- fprintf(stderr, "unquant error %d %d\n", i, level);
-#endif
- block[j] = level;
- sum+=level;
+ qadd = (qscale - 1) | 1;
+ }else{
+ qadd = 0;
+ }
+ if(s->ac_pred)
+ nCoeffs=63;
+ else
+ nCoeffs= s->inter_scantable.raster_end[ s->block_last_index[n] ];
+
+ for(i=1; i<=nCoeffs; i++) {
+ level = block[i];
+ if (level) {
+ if (level < 0) {
+ level = level * qmul - qadd;
+ } else {
+ level = level * qmul + qadd;
}
+ block[i] = level;
}
- block[63]^=sum&1;
}
}
-
-static void dct_unquantize_h263_c(MpegEncContext *s,
+static void dct_unquantize_h263_inter_c(MpegEncContext *s,
DCTELEM *block, int n, int qscale)
{
int i, level, qmul, qadd;
qadd = (qscale - 1) | 1;
qmul = qscale << 1;
- if (s->mb_intra) {
- if (!s->h263_aic) {
- if (n < 4)
- block[0] = block[0] * s->y_dc_scale;
- else
- block[0] = block[0] * s->c_dc_scale;
- }else
- qadd = 0;
- i = 1;
- nCoeffs= 63; //does not allways use zigzag table
- } else {
- i = 0;
- nCoeffs= s->inter_scantable.raster_end[ s->block_last_index[n] ];
- }
+ nCoeffs= s->inter_scantable.raster_end[ s->block_last_index[n] ];
- for(;i<=nCoeffs;i++) {
+ for(i=0; i<=nCoeffs; i++) {
level = block[i];
if (level) {
if (level < 0) {
} else {
level = level * qmul + qadd;
}
-#ifdef PARANOID
- if (level < -2048 || level > 2047)
- fprintf(stderr, "unquant error %d %d\n", i, level);
-#endif
block[i] = level;
}
}
}
-
static const AVOption mpeg4_options[] =
{
AVOPTION_CODEC_INT("bitrate", "desired video bitrate", bit_rate, 4, 240000000, 800000),
projects / platform / upstream / libav.git / commitdiff