]> granicus.if.org Git - libvpx/commit
quantize: copy ssse3 optimizations to intrinsics
authorJohann <johannkoenig@google.com>
Thu, 3 Aug 2017 17:22:07 +0000 (10:22 -0700)
committerJohann <johannkoenig@google.com>
Tue, 8 Aug 2017 19:22:14 +0000 (12:22 -0700)
commitd52cb5972930b68b0d0ce15b45d1a788dcd78322
treece9561fd1dcc68e820a3e826aa277994e3df6466
parentcbb83ba4aa99b40b0b4a2a407bfd6d0d8be87d1f
quantize: copy ssse3 optimizations to intrinsics

Fairly minor differences from sse2. pabsw and psignw are the big gains.
Also re-uses some values in eob calculation to avoid an extra pcmp.

Fixes test failures in HBD and OS X builds.

Allows using it in 32bit builds, where it is about 40% faster than sse2.

Substantially faster than the assembly for skip_block. 10-20% faster the
rest of the time.

Change-Id: If783bb3567e561e47667e10133b9c84414a334e2
test/vp9_quantize_test.cc
vpx_dsp/vpx_dsp.mk
vpx_dsp/vpx_dsp_rtcd_defs.pl
vpx_dsp/x86/quantize_ssse3.c [new file with mode: 0644]
vpx_dsp/x86/quantize_ssse3_x86_64.asm