shithub: libvpx

Info  •  Files  •  Log  •  Branches

Bug fix: fadst4() in vp9/encoder/vp9_dct.c

A new bug was introduced in a80bdfd "Change sinpi_{1,2,3,4}_9 from
tran_high_t to int16_t". Reverted the change in this file.

BUG=webm:1450

Failed test C/TransHT.AccuracyCheck/26.

Change-Id: Id001f57aad811803ef7d367d2b2bc008d8499991

--- a/vp9/encoder/vp9_dct.c

+++ b/vp9/encoder/vp9_dct.c

@@ -246,14 +246,13 @@

     return;

   }

-  // 32-bit result is enough for the following multiplications.

-  s0 = sinpi_1_9 * input[0];

-  s1 = sinpi_4_9 * input[0];

-  s2 = sinpi_2_9 * input[1];

-  s3 = sinpi_1_9 * input[1];

-  s4 = sinpi_3_9 * input[2];

-  s5 = sinpi_4_9 * input[3];

-  s6 = sinpi_2_9 * input[3];

+  s0 = sinpi_1_9 * x0;

+  s1 = sinpi_4_9 * x0;

+  s2 = sinpi_2_9 * x1;

+  s3 = sinpi_1_9 * x1;

+  s4 = sinpi_3_9 * x2;

+  s5 = sinpi_4_9 * x3;

+  s6 = sinpi_2_9 * x3;

   s7 = x0 + x1 - x3;

   x0 = s0 + s2 + s5;