diff options
| author | Geza Lore <gezalore@gmail.com> | 2015-10-28 14:35:04 +0000 |
|---|---|---|
| committer | Geza Lore <gezalore@gmail.com> | 2015-11-11 14:03:47 +0000 |
| commit | 5eefd3ebfdf61f76676de4f86e128e3d101311a2 (patch) | |
| tree | a763404e3e9890907b57fc522408fa2d63fd9ce1 /vp9/vp9cx.mk | |
| parent | 420e8d6d039c2224e00c13aba7f8908b68868359 (diff) | |
| download | libvpx-5eefd3ebfdf61f76676de4f86e128e3d101311a2.tar.gz | |
Add AVX vectorized vp9_diamond_search_sad
This function now has an AVX intrinsics version which is about 80%
faster compared to the C implementation. This provides a 2-4% total
speed-up for encode, depending on encoding parameters. The function
utilizes 3 properties of the cost function lookup table, constructed
in 'cal_nmvjointsadcost' and 'cal_nmvsadcosts'.
For the joint cost:
- mvjointsadcost[1] == mvjointsadcost[2] == mvjointsadcost[3]
For the component costs:
- For all i: mvsadcost[0][i] == mvsadcost[1][i]
(equal per component cost)
- For all i: mvsadcost[0][i] == mvsadcost[0][-i]
(Cost function is even)
These must hold, otherwise the AVX version of the function cannot be used.
Change-Id: I6c2791d43022822a9e6ab43cd124a773946d0bdc
Diffstat (limited to 'vp9/vp9cx.mk')
| -rw-r--r-- | vp9/vp9cx.mk | 1 |
1 files changed, 1 insertions, 0 deletions
diff --git a/vp9/vp9cx.mk b/vp9/vp9cx.mk index 3f3bdef96..5918240e2 100644 --- a/vp9/vp9cx.mk +++ b/vp9/vp9cx.mk @@ -96,6 +96,7 @@ VP9_CX_SRCS-yes += encoder/vp9_mbgraph.h VP9_CX_SRCS-$(HAVE_SSE2) += encoder/x86/vp9_avg_intrin_sse2.c VP9_CX_SRCS-$(HAVE_SSE2) += encoder/x86/vp9_temporal_filter_apply_sse2.asm VP9_CX_SRCS-$(HAVE_SSE2) += encoder/x86/vp9_quantize_sse2.c +VP9_CX_SRCS-$(HAVE_AVX) += encoder/x86/vp9_diamond_search_sad_avx.c ifeq ($(CONFIG_VP9_HIGHBITDEPTH),yes) VP9_CX_SRCS-$(HAVE_SSE2) += encoder/x86/vp9_highbd_block_error_intrin_sse2.c endif |