]> granicus.if.org Git - libvpx/commit
Implemented DST 8x8 with SSE2 intrinsics.
authorYi Luo <luoyi@google.com>
Wed, 24 Feb 2016 00:59:38 +0000 (16:59 -0800)
committerYi Luo <luoyi@google.com>
Wed, 24 Feb 2016 22:58:01 +0000 (14:58 -0800)
commit0353f596e9b0763b621d9d15761365f4816f8761
tree22e986f477683d48a964558c46e4a0dcf704c2ff
parenta90f8b8c8f9775415e8ed5ada736deb64e538c76
Implemented DST 8x8 with SSE2 intrinsics.

Implemented fdst8_sse2() function against C version: fdst8().
Added seven DST related hybrid transform types in vp10_fht8x8_sse2().
Replaced vp10_fht8x8_c() with vp10_fht8x8_sse2() in fwd_txfm_8x8().
Speedup: 18.1%, 11.5%, 22.0% based on speed test from
city_cif.y4m, garden_sif.y4m, mobile_cif.y4m.

Change-Id: Ia4aa1ea44c7a33e494f64ce843037f8703f975e3
vp10/encoder/hybrid_fwd_txfm.c
vp10/encoder/x86/dct_sse2.c