]> granicus.if.org Git - libvpx/commit
Improve idct32x32_34_add SSSE3 intrinsics performance
authorYi Luo <luoyi@google.com>
Wed, 1 Mar 2017 00:38:41 +0000 (16:38 -0800)
committerYi Luo <luoyi@google.com>
Wed, 1 Mar 2017 19:14:48 +0000 (11:14 -0800)
commit07c48ccfe090625e409be50ff82b6d8e6fc93578
treed14996a2f5166d22c6eb8aac8c4a1c4d296bbf1b
parentb71245683b727cdfe86f2f29eac830e88e980f8d
Improve idct32x32_34_add SSSE3 intrinsics performance

- Split the transform into first half and second half.
- Reschedule the instructions to avoid stack spillover.
- Function level speed improves ~16%.

Change-Id: I166889840d23aa8a273eca00f6fbdae8b4566f35
vpx_dsp/x86/inv_txfm_ssse3.c