projects / platform / upstream / libvpx.git / commit
? search:
summary | shortlog | log | commit | commitdiff | tree
(parent: 0fefc68) | patch
Add vpx_idct16x16_38_add_neon()
authorLinfeng Zhang <linfengz@google.com>
Tue, 7 Feb 2017 18:06:21 +0000 (10:06 -0800)
committerLinfeng Zhang <linfengz@google.com>
Wed, 8 Feb 2017 20:15:22 +0000 (12:15 -0800)
commitbc1c18e18c6001dc3db233c4dcb34c4e54923ce7
treed40a5af00293b3215adbdf7f0248e6fe8c31dcaftree | snapshot
parent0fefc6873a7f11032e52900aa1d8adf9ad8dbeedcommit | diff
Add vpx_idct16x16_38_add_neon()

The RunQuantCheck() test on it exposes 16-bit overflow in stage 7 of
pass 2. Change to use saturating add/sub for both
vpx_idct16x16_38_add_neon() and vpx_idct16x16_256_add_neon() for high
bitdepth.

Change-Id: Ibf4c107a887553a52852cc582e28d38a5a5a2712
test/partial_idct_test.cc diff | blob | history
vpx_dsp/arm/idct16x16_add_neon.c diff | blob | history
vpx_dsp/vpx_dsp_rtcd_defs.pl diff | blob | history
Domain: Multimedia / Codec;