Skip to content
GitLab
Commit 56a8a0b5 authored by Frank Galligan's avatar Frank Galligan
Browse files

Optimize vp9_tm_predictor_8x8_neon function 

Change-Id: Ia12aae491202098ff66366145aa0c3da38dc97e5
parent 8b0537f6
v1.14.0-linphone 1.4.X feature/update_to_v1.9.0-linphone feature/uwp_nuget frame_parallel highbitdepth indianrunnerduck javanwhistlingduck khakicampbell linphone linphone-android linphone-old longtailedduck m49-2623 m52-2743 m54-2840 m56-2924 m66-3359 m68-3440 mandarinduck nextgen nextgenv2 playground sandbox/Jingning/experimental sandbox/Jingning/transcode sandbox/Jingning/vpx sandbox/aconverse@google.com/ansbench sandbox/hkuang/frame_parallel sandbox/hkuang@google.com/decode sandbox/jimbankoski@google.com/proposed-aom sandbox/jingning@google.com/decoder_test_suite sandbox/jingning@google.com/experimental sandbox/jzern@google.com/test sandbox/wangch@google.com/vp9 sandbox/yaowu@google.com/mergeaom v1.12.0-linphone v1.6.1_linphone v1.7.0-linphone v1.9.0-linphone v1.9.0 v1.9.0-rc1 v1.8.2 v1.8.1 v1.8.0 v1.7.0 v1.6.1 v1.6.0 v1.5.0 v1.4.0
No related merge requests found
Hide whitespace changes
Inline Side-by-side
Showing
with 31 additions and 32 deletions
vp9/common/arm/neon/vp9_reconintra_neon.asm
+ 31
− 32
View file @ 56a8a0b5
...@@ -353,53 +353,52 @@ loop_h ...@@ -353,53 +353,52 @@ loop_h
vsubl.u8 q3, d2, d0 vsubl.u8 q3, d2, d0
; Load left row by row and compute left + (above - ytop_left) ; Load left row by row and compute left + (above - ytop_left)
vld1.u8 {d6}, [r3]
; 1st row and 2nd row ; 1st row and 2nd row
ldrb r12, [r3], #1 vdup.u8 d0, d6[0]
ldrb r2, [r3], #1 vdup.u8 d1, d6[1]
vdup.u16 q1, r12 vaddw.s16 q1, q3, d0
vdup.u16 q2, r2 vaddw.s16 q2, q3, d1
vadd.s16 q1, q1, q3
vadd.s16 q2, q2, q3
vqshrun.s16 d0, q1, #0
vqshrun.s16 d1, q2, #0
vst1.64 {d0}, [r0], r1
vst1.64 {d1}, [r0], r1
; 3rd row and 4th row ; 3rd row and 4th row
ldrb r12, [r3], #1 vdup.u8 d0, d6[2]
ldrb r2, [r3], #1 vdup.u8 d1, d6[3]
vdup.u16 q1, r12 vaddw.s16 q8, q3, d0
vdup.u16 q2, r2 vaddw.s16 q9, q3, d1
vadd.s16 q1, q1, q3
vadd.s16 q2, q2, q3
vqshrun.s16 d0, q1, #0 vqshrun.s16 d0, q1, #0
vqshrun.s16 d1, q2, #0 vqshrun.s16 d1, q2, #0
vqshrun.s16 d2, q8, #0
vqshrun.s16 d3, q9, #0
vst1.64 {d0}, [r0], r1 vst1.64 {d0}, [r0], r1
vst1.64 {d1}, [r0], r1 vst1.64 {d1}, [r0], r1
vst1.64 {d2}, [r0], r1
vst1.64 {d3}, [r0], r1
; 5th row and 6th row ; 5th row and 6th row
ldrb r12, [r3], #1 vdup.u8 d0, d6[4]
ldrb r2, [r3], #1 vdup.u8 d1, d6[5]
vdup.u16 q1, r12 vaddw.s16 q1, q3, d0
vdup.u16 q2, r2 vaddw.s16 q2, q3, d1
vadd.s16 q1, q1, q3
vadd.s16 q2, q2, q3
vqshrun.s16 d0, q1, #0
vqshrun.s16 d1, q2, #0
vst1.64 {d0}, [r0], r1
vst1.64 {d1}, [r0], r1
; 7rd row and 8th row ; 7rd row and 8th row
ldrb r12, [r3], #1 vdup.u8 d0, d6[6]
ldrb r2, [r3], #1 vdup.u8 d1, d6[7]
vdup.u16 q1, r12 vaddw.s16 q8, q3, d0
vdup.u16 q2, r2 vaddw.s16 q9, q3, d1
vadd.s16 q1, q1, q3
vadd.s16 q2, q2, q3
vqshrun.s16 d0, q1, #0 vqshrun.s16 d0, q1, #0
vqshrun.s16 d1, q2, #0 vqshrun.s16 d1, q2, #0
vqshrun.s16 d2, q8, #0
vqshrun.s16 d3, q9, #0
vst1.64 {d0}, [r0], r1 vst1.64 {d0}, [r0], r1
vst1.64 {d1}, [r0], r1 vst1.64 {d1}, [r0], r1
vst1.64 {d2}, [r0], r1
vst1.64 {d3}, [r0], r1
bx lr bx lr
ENDP ; |vp9_tm_predictor_8x8_neon| ENDP ; |vp9_tm_predictor_8x8_neon|
......
Supports Markdown
0% or .
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment