libvpx: Pull from upstream

source/libvpx/vpx_dsp/inv_txfm.c

 /*
  *  Copyright (c) 2015 The WebM project authors. All Rights Reserved.
  *
  *  Use of this source code is governed by a BSD-style license
  *  that can be found in the LICENSE file in the root of the source
  *  tree. An additional intellectual property rights grant can be found
  *  in the file PATENTS.  All contributing project authors may
  *  be found in the AUTHORS file in the root of the source tree.
  */
 
@@ skipping 152 matching lines @@
   step1[3] = input[6];
   temp1 = input[1] * cospi_28_64 - input[7] * cospi_4_64;
   temp2 = input[1] * cospi_4_64 + input[7] * cospi_28_64;
   step1[4] = WRAPLOW(dct_const_round_shift(temp1), 8);
   step1[7] = WRAPLOW(dct_const_round_shift(temp2), 8);
   temp1 = input[5] * cospi_12_64 - input[3] * cospi_20_64;
   temp2 = input[5] * cospi_20_64 + input[3] * cospi_12_64;
   step1[5] = WRAPLOW(dct_const_round_shift(temp1), 8);
   step1[6] = WRAPLOW(dct_const_round_shift(temp2), 8);
 
-  // stage 2 & stage 3 - even half
-  idct4_c(step1, step1);
-
-  // stage 2 - odd half
+  // stage 2
+  temp1 = (step1[0] + step1[2]) * cospi_16_64;
+  temp2 = (step1[0] - step1[2]) * cospi_16_64;
+  step2[0] = WRAPLOW(dct_const_round_shift(temp1), 8);
+  step2[1] = WRAPLOW(dct_const_round_shift(temp2), 8);
+  temp1 = step1[1] * cospi_24_64 - step1[3] * cospi_8_64;
+  temp2 = step1[1] * cospi_8_64 + step1[3] * cospi_24_64;
+  step2[2] = WRAPLOW(dct_const_round_shift(temp1), 8);
+  step2[3] = WRAPLOW(dct_const_round_shift(temp2), 8);
   step2[4] = WRAPLOW(step1[4] + step1[5], 8);
   step2[5] = WRAPLOW(step1[4] - step1[5], 8);
   step2[6] = WRAPLOW(-step1[6] + step1[7], 8);
   step2[7] = WRAPLOW(step1[6] + step1[7], 8);
 
-  // stage 3 -odd half
+  // stage 3
+  step1[0] = WRAPLOW(step2[0] + step2[3], 8);
+  step1[1] = WRAPLOW(step2[1] + step2[2], 8);
+  step1[2] = WRAPLOW(step2[1] - step2[2], 8);
+  step1[3] = WRAPLOW(step2[0] - step2[3], 8);
   step1[4] = step2[4];
   temp1 = (step2[6] - step2[5]) * cospi_16_64;
   temp2 = (step2[5] + step2[6]) * cospi_16_64;
   step1[5] = WRAPLOW(dct_const_round_shift(temp1), 8);
   step1[6] = WRAPLOW(dct_const_round_shift(temp2), 8);
   step1[7] = step2[7];
 
   // stage 4
   output[0] = WRAPLOW(step1[0] + step1[7], 8);
   output[1] = WRAPLOW(step1[1] + step1[6], 8);
@@ skipping 2275 matching lines @@
   out = WRAPLOW(highbd_dct_const_round_shift(out * cospi_16_64, bd), bd);
   a1 = ROUND_POWER_OF_TWO(out, 6);
 
   for (j = 0; j < 32; ++j) {
     for (i = 0; i < 32; ++i)
       dest[i] = highbd_clip_pixel_add(dest[i], a1, bd);
     dest += stride;
   }
 }
 #endif  // CONFIG_VP9_HIGHBITDEPTH