lowres2 support.

The new lowres support is limited to decoders where lowres decoding
is possible in high quality.
I was not able to measure any speed difference, but if one is found
the 2-3 lines that might affect speed can be made compile time conditional

Signed-off-by: Michael Niedermayer <michaelni@gmx.at>

1 files changed, 213 insertions, 0 deletions

diff --git a/libavcodec/jrevdct.c b/libavcodec/jrevdct.c
index e33558f825..395eb8c638 100644
--- a/libavcodec/jrevdct.c
+++ b/libavcodec/jrevdct.c
@@ -940,3 +940,216 @@ void ff_j_rev_dct(DCTBLOCK data)
     dataptr++;                  /* advance pointer to next column */
   }
 }
+
+#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 DCTELEM *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. */
+          DCTELEM dcval = (DCTELEM) (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] = (DCTELEM) DESCALE(tmp10, CONST_BITS-PASS1_BITS);
+    dataptr[1] = (DCTELEM) DESCALE(tmp11, CONST_BITS-PASS1_BITS);
+    dataptr[2] = (DCTELEM) DESCALE(tmp12, CONST_BITS-PASS1_BITS);
+    dataptr[3] = (DCTELEM) 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