virtualenv/FFmpeg
1
0
Fork 0
mirror of https://github.com/FFmpeg/FFmpeg.git synced 2025-08-10 06:10:52 +02:00
Code Issues Releases Activity

aarch64: vp9lpf: Interleave the start of flat8in into the calculation above

Browse Source 
This adds lots of extra .ifs, but speeds it up by a couple cycles,
by avoiding stalls.

This is cherrypicked from libav commit
b0806088d3.

Signed-off-by: Martin Storsjö <martin@martin.st>
This commit is contained in:
Martin Storsjö
2017-01-10 22:08:50 +02:00
parent 83399cf569
commit 9f3a886364
1 changed files with 11 additions and 3 deletions
Download Patch File Download Diff File Expand all files Collapse all files

14
libavcodec/aarch64/vp9lpf_neon.S
View File

@@ -338,20 +338,28 @@
uxtl_sz v0.8h, v1.8h, v22, \sz // p1 uxtl_sz v0.8h, v1.8h, v22, \sz // p1
uxtl_sz v2.8h, v3.8h, v25, \sz // q1 uxtl_sz v2.8h, v3.8h, v25, \sz // q1
.if \wd >= 8
mov x5, v6.d[0]
.ifc \sz, .16b
mov x6, v6.d[1]
.endif
.endif
saddw_sz v0.8h, v1.8h, v0.8h, v1.8h, \tmp3, \sz // p1 + f saddw_sz v0.8h, v1.8h, v0.8h, v1.8h, \tmp3, \sz // p1 + f
ssubw_sz v2.8h, v3.8h, v2.8h, v3.8h, \tmp3, \sz // q1 - f ssubw_sz v2.8h, v3.8h, v2.8h, v3.8h, \tmp3, \sz // q1 - f
sqxtun_sz v0, v0.8h, v1.8h, \sz // out p1 sqxtun_sz v0, v0.8h, v1.8h, \sz // out p1
sqxtun_sz v2, v2.8h, v3.8h, \sz // out q1 sqxtun_sz v2, v2.8h, v3.8h, \sz // out q1
.if \wd >= 8
.ifc \sz, .16b
adds x5, x5, x6
.endif
.endif
bit v22\sz, v0\sz, v5\sz // if (!hev && fm && !flat8in) bit v22\sz, v0\sz, v5\sz // if (!hev && fm && !flat8in)
bit v25\sz, v2\sz, v5\sz bit v25\sz, v2\sz, v5\sz
// If no pixels need flat8in, jump to flat8out // If no pixels need flat8in, jump to flat8out
// (or to a writeout of the inner 4 pixels, for wd=8) // (or to a writeout of the inner 4 pixels, for wd=8)
.if \wd >= 8 .if \wd >= 8
mov x5, v6.d[0]
.ifc \sz, .16b .ifc \sz, .16b
mov x6, v6.d[1]
adds x5, x5, x6
b.eq 6f b.eq 6f
.else .else
cbz x5, 6f cbz x5, 6f