diff options
| author | Johann <johannkoenig@google.com> | 2017-02-16 17:57:44 -0800 |
|---|---|---|
| committer | Johann Koenig <johannkoenig@google.com> | 2017-02-24 05:25:26 +0000 |
| commit | 904b957ae965bd3d67f15a75cd9db7954f810d33 (patch) | |
| tree | 3d3db25d9600a19e32d105bf9a149429478f3782 /vp9/common | |
| parent | aa911e8b41733a950dfe96866dbf8118940bc996 (diff) | |
| download | libvpx-904b957ae965bd3d67f15a75cd9db7954f810d33.tar libvpx-904b957ae965bd3d67f15a75cd9db7954f810d33.tar.gz libvpx-904b957ae965bd3d67f15a75cd9db7954f810d33.tar.bz2 libvpx-904b957ae965bd3d67f15a75cd9db7954f810d33.zip | |
consolidate block_error functions
vp9_highbd_block_error_8bit_c was a very simple wrapper around
vp9_block_error_c. The SSE2 implemention was practically identical to
the non-HBD one. It was missing some minor improvements which only
went into the original version.
In quick speed tests, the AVX implementation showed minimal
improvement over SSE2 when it does not detect overflow. However, when
overflow is detected the function is run a second time. The
OperationCheck test seems to trigger this case and reverses any
speed benefits by running ~60% slower. AVX2 on the other hand is
always 30-40% faster.
Change-Id: I9fcb9afbcb560f234c7ae1b13ddb69eca3988ba1
Diffstat (limited to 'vp9/common')
| -rw-r--r-- | vp9/common/vp9_rtcd_defs.pl | 3 |
1 files changed, 0 insertions, 3 deletions
diff --git a/vp9/common/vp9_rtcd_defs.pl b/vp9/common/vp9_rtcd_defs.pl index 57af79d5b..77bebc7b9 100644 --- a/vp9/common/vp9_rtcd_defs.pl +++ b/vp9/common/vp9_rtcd_defs.pl @@ -130,9 +130,6 @@ if (vpx_config("CONFIG_VP9_HIGHBITDEPTH") eq "yes") { add_proto qw/int64_t vp9_highbd_block_error/, "const tran_low_t *coeff, const tran_low_t *dqcoeff, intptr_t block_size, int64_t *ssz, int bd"; specialize qw/vp9_highbd_block_error sse2/; - add_proto qw/int64_t vp9_highbd_block_error_8bit/, "const tran_low_t *coeff, const tran_low_t *dqcoeff, intptr_t block_size, int64_t *ssz"; - specialize qw/vp9_highbd_block_error_8bit sse2 avx/; - add_proto qw/int64_t vp9_block_error_fp/, "const tran_low_t *coeff, const tran_low_t *dqcoeff, int block_size"; specialize qw/vp9_block_error_fp sse2/; |