{
int cpu_flags = av_get_cpu_flags();
- put_pixels_clamped = c->put_pixels_clamped;
- add_pixels_clamped = c->add_pixels_clamped;
-
- if (!high_bit_depth) {
- if (avctx->idct_algo == FF_IDCT_AUTO ||
+ if (!avctx->lowres && !high_bit_depth) {
+ if ((avctx->idct_algo == FF_IDCT_AUTO && !(avctx->flags & CODEC_FLAG_BITEXACT)) ||
avctx->idct_algo == FF_IDCT_ARM) {
c->idct_put = j_rev_dct_arm_put;
c->idct_add = j_rev_dct_arm_add;
void ff_fdct248_islow_10(int16_t *data);
void ff_j_rev_dct(int16_t *data);
+void ff_j_rev_dct4(int16_t *data);
+void ff_j_rev_dct2(int16_t *data);
+void ff_j_rev_dct1(int16_t *data);
+ void ff_jref_idct_put(uint8_t *dest, int line_size, int16_t *block);
+ void ff_jref_idct_add(uint8_t *dest, int line_size, int16_t *block);
#endif /* AVCODEC_DCT_H */
}
}
- void ff_put_pixels_clamped(const int16_t *block, uint8_t *av_restrict pixels,
- int line_size)
+ void (*ff_put_pixels_clamped)(const int16_t *block, uint8_t *pixels, int line_size);
+ void (*ff_add_pixels_clamped)(const int16_t *block, uint8_t *pixels, int line_size);
+
-static void put_pixels_clamped_c(const int16_t *block, uint8_t *restrict pixels,
++static void put_pixels_clamped_c(const int16_t *block, uint8_t *av_restrict pixels,
+ int line_size)
{
int i;
}
}
- void ff_add_pixels_clamped(const int16_t *block, uint8_t *av_restrict pixels,
- int line_size)
-static void add_pixels_clamped_c(const int16_t *block, uint8_t *restrict pixels,
++static void add_pixels_clamped_c(const int16_t *block, uint8_t *av_restrict pixels,
+ int line_size)
{
int i;
}
}
- static void jref_idct_put(uint8_t *dest, int line_size, int16_t *block)
- {
- ff_j_rev_dct(block);
- ff_put_pixels_clamped(block, dest, line_size);
- }
-
- static void jref_idct_add(uint8_t *dest, int line_size, int16_t *block)
- {
- ff_j_rev_dct(block);
- ff_add_pixels_clamped(block, dest, line_size);
- }
+static void add_pixels_clamped4_c(const int16_t *block, uint8_t *av_restrict pixels,
+ int line_size)
+{
+ int i;
+
+ /* read the pixels */
+ for(i=0;i<4;i++) {
+ pixels[0] = av_clip_uint8(pixels[0] + block[0]);
+ pixels[1] = av_clip_uint8(pixels[1] + block[1]);
+ pixels[2] = av_clip_uint8(pixels[2] + block[2]);
+ pixels[3] = av_clip_uint8(pixels[3] + block[3]);
+ pixels += line_size;
+ block += 8;
+ }
+}
+
+static void add_pixels_clamped2_c(const int16_t *block, uint8_t *av_restrict pixels,
+ int line_size)
+{
+ int i;
+
+ /* read the pixels */
+ for(i=0;i<2;i++) {
+ pixels[0] = av_clip_uint8(pixels[0] + block[0]);
+ pixels[1] = av_clip_uint8(pixels[1] + block[1]);
+ pixels += line_size;
+ block += 8;
+ }
+}
+
+static void ff_jref_idct4_put(uint8_t *dest, int line_size, int16_t *block)
+{
+ ff_j_rev_dct4 (block);
+ put_pixels_clamped4_c(block, dest, line_size);
+}
+static void ff_jref_idct4_add(uint8_t *dest, int line_size, int16_t *block)
+{
+ ff_j_rev_dct4 (block);
+ add_pixels_clamped4_c(block, dest, line_size);
+}
+
+static void ff_jref_idct2_put(uint8_t *dest, int line_size, int16_t *block)
+{
+ ff_j_rev_dct2 (block);
+ put_pixels_clamped2_c(block, dest, line_size);
+}
+static void ff_jref_idct2_add(uint8_t *dest, int line_size, int16_t *block)
+{
+ ff_j_rev_dct2 (block);
+ add_pixels_clamped2_c(block, dest, line_size);
+}
+
+static void ff_jref_idct1_put(uint8_t *dest, int line_size, int16_t *block)
+{
+ dest[0] = av_clip_uint8((block[0] + 4)>>3);
+}
+static void ff_jref_idct1_add(uint8_t *dest, int line_size, int16_t *block)
+{
+ dest[0] = av_clip_uint8(dest[0] + ((block[0] + 4)>>3));
+}
+
av_cold void ff_idctdsp_init(IDCTDSPContext *c, AVCodecContext *avctx)
{
const unsigned high_bit_depth = avctx->bits_per_raw_sample > 8;
- if (avctx->bits_per_raw_sample == 10) {
- c->idct_put = ff_simple_idct_put_10;
- c->idct_add = ff_simple_idct_add_10;
- c->idct = ff_simple_idct_10;
+ if (avctx->lowres==1) {
+ c->idct_put = ff_jref_idct4_put;
+ c->idct_add = ff_jref_idct4_add;
+ c->idct = ff_j_rev_dct4;
c->perm_type = FF_IDCT_PERM_NONE;
- } else if (avctx->idct_algo == FF_IDCT_INT) {
- c->idct_put = ff_jref_idct_put;
- c->idct_add = ff_jref_idct_add;
- c->idct = ff_j_rev_dct;
- c->perm_type = FF_IDCT_PERM_LIBMPEG2;
- } else if (avctx->idct_algo == FF_IDCT_FAAN) {
- c->idct_put = ff_faanidct_put;
- c->idct_add = ff_faanidct_add;
- c->idct = ff_faanidct;
+ } else if (avctx->lowres==2) {
+ c->idct_put = ff_jref_idct2_put;
+ c->idct_add = ff_jref_idct2_add;
+ c->idct = ff_j_rev_dct2;
c->perm_type = FF_IDCT_PERM_NONE;
- } else { // accurate/default
- c->idct_put = ff_simple_idct_put_8;
- c->idct_add = ff_simple_idct_add_8;
- c->idct = ff_simple_idct_8;
+ } else if (avctx->lowres==3) {
+ c->idct_put = ff_jref_idct1_put;
+ c->idct_add = ff_jref_idct1_add;
+ c->idct = ff_j_rev_dct1;
c->perm_type = FF_IDCT_PERM_NONE;
- c->idct_put = jref_idct_put;
- c->idct_add = jref_idct_add;
+ } else {
+ if (avctx->bits_per_raw_sample == 10) {
+ c->idct_put = ff_simple_idct_put_10;
+ c->idct_add = ff_simple_idct_add_10;
+ c->idct = ff_simple_idct_10;
+ c->perm_type = FF_IDCT_PERM_NONE;
+ } else if (avctx->bits_per_raw_sample == 12) {
+ c->idct_put = ff_simple_idct_put_12;
+ c->idct_add = ff_simple_idct_add_12;
+ c->idct = ff_simple_idct_12;
+ c->perm_type = FF_IDCT_PERM_NONE;
+ } else {
+ if (avctx->idct_algo == FF_IDCT_INT) {
++ c->idct_put = ff_jref_idct_put;
++ c->idct_add = ff_jref_idct_add;
+ c->idct = ff_j_rev_dct;
+ c->perm_type = FF_IDCT_PERM_LIBMPEG2;
+ } else if (avctx->idct_algo == FF_IDCT_FAAN) {
+ c->idct_put = ff_faanidct_put;
+ c->idct_add = ff_faanidct_add;
+ c->idct = ff_faanidct;
+ c->perm_type = FF_IDCT_PERM_NONE;
+ } else { // accurate/default
+ c->idct_put = ff_simple_idct_put_8;
+ c->idct_add = ff_simple_idct_add_8;
+ c->idct = ff_simple_idct_8;
+ c->perm_type = FF_IDCT_PERM_NONE;
+ }
+ }
}
- c->put_pixels_clamped = ff_put_pixels_clamped;
+ c->put_pixels_clamped = put_pixels_clamped_c;
c->put_signed_pixels_clamped = put_signed_pixels_clamped_c;
- c->add_pixels_clamped = ff_add_pixels_clamped;
+ c->add_pixels_clamped = add_pixels_clamped_c;
+
+ ff_put_pixels_clamped = c->put_pixels_clamped;
+ ff_add_pixels_clamped = c->add_pixels_clamped;
+ if (CONFIG_MPEG4_DECODER && avctx->idct_algo == FF_IDCT_XVID)
+ ff_xvid_idct_init(c, avctx);
+
+ if (ARCH_ALPHA)
+ ff_idctdsp_init_alpha(c, avctx, high_bit_depth);
if (ARCH_ARM)
ff_idctdsp_init_arm(c, avctx, high_bit_depth);
if (ARCH_PPC)
enum idct_permutation_type perm_type;
} IDCTDSPContext;
+ extern void (*ff_put_pixels_clamped)(const int16_t *block, uint8_t *pixels, int line_size);
+ extern void (*ff_add_pixels_clamped)(const int16_t *block, uint8_t *pixels, int line_size);
+
void ff_idctdsp_init(IDCTDSPContext *c, AVCodecContext *avctx);
+void ff_idctdsp_init_alpha(IDCTDSPContext *c, AVCodecContext *avctx,
+ unsigned high_bit_depth);
void ff_idctdsp_init_arm(IDCTDSPContext *c, AVCodecContext *avctx,
unsigned high_bit_depth);
void ff_idctdsp_init_ppc(IDCTDSPContext *c, AVCodecContext *avctx,
}
}
+#undef DCTSIZE
+#define DCTSIZE 4
+#define DCTSTRIDE 8
+
+void ff_j_rev_dct4(DCTBLOCK data)
+{
+ int32_t tmp0, tmp1, tmp2, tmp3;
+ int32_t tmp10, tmp11, tmp12, tmp13;
+ int32_t z1;
+ int32_t d0, d2, d4, d6;
+ register int16_t *dataptr;
+ int rowctr;
+
+ /* Pass 1: process rows. */
+ /* Note results are scaled up by sqrt(8) compared to a true IDCT; */
+ /* furthermore, we scale the results by 2**PASS1_BITS. */
+
+ data[0] += 4;
+
+ dataptr = data;
+
+ for (rowctr = DCTSIZE-1; rowctr >= 0; rowctr--) {
+ /* Due to quantization, we will usually find that many of the input
+ * coefficients are zero, especially the AC terms. We can exploit this
+ * by short-circuiting the IDCT calculation for any row in which all
+ * the AC terms are zero. In that case each output is equal to the
+ * DC coefficient (with scale factor as needed).
+ * With typical images and quantization tables, half or more of the
+ * row DCT calculations can be simplified this way.
+ */
+
+ register int *idataptr = (int*)dataptr;
+
+ d0 = dataptr[0];
+ d2 = dataptr[1];
+ d4 = dataptr[2];
+ d6 = dataptr[3];
+
+ if ((d2 | d4 | d6) == 0) {
+ /* AC terms all zero */
+ if (d0) {
+ /* Compute a 32 bit value to assign. */
+ int16_t dcval = (int16_t) (d0 << PASS1_BITS);
+ register int v = (dcval & 0xffff) | ((dcval << 16) & 0xffff0000);
+
+ idataptr[0] = v;
+ idataptr[1] = v;
+ }
+
+ dataptr += DCTSTRIDE; /* advance pointer to next row */
+ continue;
+ }
+
+ /* Even part: reverse the even part of the forward DCT. */
+ /* The rotator is sqrt(2)*c(-6). */
+ if (d6) {
+ if (d2) {
+ /* d0 != 0, d2 != 0, d4 != 0, d6 != 0 */
+ z1 = MULTIPLY(d2 + d6, FIX_0_541196100);
+ tmp2 = z1 + MULTIPLY(-d6, FIX_1_847759065);
+ tmp3 = z1 + MULTIPLY(d2, FIX_0_765366865);
+
+ tmp0 = (d0 + d4) << CONST_BITS;
+ tmp1 = (d0 - d4) << CONST_BITS;
+
+ tmp10 = tmp0 + tmp3;
+ tmp13 = tmp0 - tmp3;
+ tmp11 = tmp1 + tmp2;
+ tmp12 = tmp1 - tmp2;
+ } else {
+ /* d0 != 0, d2 == 0, d4 != 0, d6 != 0 */
+ tmp2 = MULTIPLY(-d6, FIX_1_306562965);
+ tmp3 = MULTIPLY(d6, FIX_0_541196100);
+
+ tmp0 = (d0 + d4) << CONST_BITS;
+ tmp1 = (d0 - d4) << CONST_BITS;
+
+ tmp10 = tmp0 + tmp3;
+ tmp13 = tmp0 - tmp3;
+ tmp11 = tmp1 + tmp2;
+ tmp12 = tmp1 - tmp2;
+ }
+ } else {
+ if (d2) {
+ /* d0 != 0, d2 != 0, d4 != 0, d6 == 0 */
+ tmp2 = MULTIPLY(d2, FIX_0_541196100);
+ tmp3 = MULTIPLY(d2, FIX_1_306562965);
+
+ tmp0 = (d0 + d4) << CONST_BITS;
+ tmp1 = (d0 - d4) << CONST_BITS;
+
+ tmp10 = tmp0 + tmp3;
+ tmp13 = tmp0 - tmp3;
+ tmp11 = tmp1 + tmp2;
+ tmp12 = tmp1 - tmp2;
+ } else {
+ /* d0 != 0, d2 == 0, d4 != 0, d6 == 0 */
+ tmp10 = tmp13 = (d0 + d4) << CONST_BITS;
+ tmp11 = tmp12 = (d0 - d4) << CONST_BITS;
+ }
+ }
+
+ /* Final output stage: inputs are tmp10..tmp13, tmp0..tmp3 */
+
+ dataptr[0] = (int16_t) DESCALE(tmp10, CONST_BITS-PASS1_BITS);
+ dataptr[1] = (int16_t) DESCALE(tmp11, CONST_BITS-PASS1_BITS);
+ dataptr[2] = (int16_t) DESCALE(tmp12, CONST_BITS-PASS1_BITS);
+ dataptr[3] = (int16_t) DESCALE(tmp13, CONST_BITS-PASS1_BITS);
+
+ dataptr += DCTSTRIDE; /* advance pointer to next row */
+ }
+
+ /* Pass 2: process columns. */
+ /* Note that we must descale the results by a factor of 8 == 2**3, */
+ /* and also undo the PASS1_BITS scaling. */
+
+ dataptr = data;
+ for (rowctr = DCTSIZE-1; rowctr >= 0; rowctr--) {
+ /* Columns of zeroes can be exploited in the same way as we did with rows.
+ * However, the row calculation has created many nonzero AC terms, so the
+ * simplification applies less often (typically 5% to 10% of the time).
+ * On machines with very fast multiplication, it's possible that the
+ * test takes more time than it's worth. In that case this section
+ * may be commented out.
+ */
+
+ d0 = dataptr[DCTSTRIDE*0];
+ d2 = dataptr[DCTSTRIDE*1];
+ d4 = dataptr[DCTSTRIDE*2];
+ d6 = dataptr[DCTSTRIDE*3];
+
+ /* Even part: reverse the even part of the forward DCT. */
+ /* The rotator is sqrt(2)*c(-6). */
+ if (d6) {
+ if (d2) {
+ /* d0 != 0, d2 != 0, d4 != 0, d6 != 0 */
+ z1 = MULTIPLY(d2 + d6, FIX_0_541196100);
+ tmp2 = z1 + MULTIPLY(-d6, FIX_1_847759065);
+ tmp3 = z1 + MULTIPLY(d2, FIX_0_765366865);
+
+ tmp0 = (d0 + d4) << CONST_BITS;
+ tmp1 = (d0 - d4) << CONST_BITS;
+
+ tmp10 = tmp0 + tmp3;
+ tmp13 = tmp0 - tmp3;
+ tmp11 = tmp1 + tmp2;
+ tmp12 = tmp1 - tmp2;
+ } else {
+ /* d0 != 0, d2 == 0, d4 != 0, d6 != 0 */
+ tmp2 = MULTIPLY(-d6, FIX_1_306562965);
+ tmp3 = MULTIPLY(d6, FIX_0_541196100);
+
+ tmp0 = (d0 + d4) << CONST_BITS;
+ tmp1 = (d0 - d4) << CONST_BITS;
+
+ tmp10 = tmp0 + tmp3;
+ tmp13 = tmp0 - tmp3;
+ tmp11 = tmp1 + tmp2;
+ tmp12 = tmp1 - tmp2;
+ }
+ } else {
+ if (d2) {
+ /* d0 != 0, d2 != 0, d4 != 0, d6 == 0 */
+ tmp2 = MULTIPLY(d2, FIX_0_541196100);
+ tmp3 = MULTIPLY(d2, FIX_1_306562965);
+
+ tmp0 = (d0 + d4) << CONST_BITS;
+ tmp1 = (d0 - d4) << CONST_BITS;
+
+ tmp10 = tmp0 + tmp3;
+ tmp13 = tmp0 - tmp3;
+ tmp11 = tmp1 + tmp2;
+ tmp12 = tmp1 - tmp2;
+ } else {
+ /* d0 != 0, d2 == 0, d4 != 0, d6 == 0 */
+ tmp10 = tmp13 = (d0 + d4) << CONST_BITS;
+ tmp11 = tmp12 = (d0 - d4) << CONST_BITS;
+ }
+ }
+
+ /* Final output stage: inputs are tmp10..tmp13, tmp0..tmp3 */
+
+ dataptr[DCTSTRIDE*0] = tmp10 >> (CONST_BITS+PASS1_BITS+3);
+ dataptr[DCTSTRIDE*1] = tmp11 >> (CONST_BITS+PASS1_BITS+3);
+ dataptr[DCTSTRIDE*2] = tmp12 >> (CONST_BITS+PASS1_BITS+3);
+ dataptr[DCTSTRIDE*3] = tmp13 >> (CONST_BITS+PASS1_BITS+3);
+
+ dataptr++; /* advance pointer to next column */
+ }
+}
+
+void ff_j_rev_dct2(DCTBLOCK data){
+ int d00, d01, d10, d11;
+
+ data[0] += 4;
+ d00 = data[0+0*DCTSTRIDE] + data[1+0*DCTSTRIDE];
+ d01 = data[0+0*DCTSTRIDE] - data[1+0*DCTSTRIDE];
+ d10 = data[0+1*DCTSTRIDE] + data[1+1*DCTSTRIDE];
+ d11 = data[0+1*DCTSTRIDE] - data[1+1*DCTSTRIDE];
+
+ data[0+0*DCTSTRIDE]= (d00 + d10)>>3;
+ data[1+0*DCTSTRIDE]= (d01 + d11)>>3;
+ data[0+1*DCTSTRIDE]= (d00 - d10)>>3;
+ data[1+1*DCTSTRIDE]= (d01 - d11)>>3;
+}
+
+void ff_j_rev_dct1(DCTBLOCK data){
+ data[0] = (data[0] + 4)>>3;
+}
+
+#undef FIX
+#undef CONST_BITS
++
+ void ff_jref_idct_put(uint8_t *dest, int line_size, int16_t *block)
+ {
+ ff_j_rev_dct(block);
+ ff_put_pixels_clamped(block, dest, line_size);
+ }
+
+ void ff_jref_idct_add(uint8_t *dest, int line_size, int16_t *block)
+ {
+ ff_j_rev_dct(block);
+ ff_add_pixels_clamped(block, dest, line_size);
+ }
* Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
*/
-#include "libavutil/internal.h"
#include "libavutil/mem.h"
#include "libavutil/x86/asm.h"
+
+ #include "libavcodec/idctdsp.h"
+
#include "idctdsp.h"
#include "simple_idct.h"