Commit 5647a57e authored by Martin Storsjö's avatar Martin Storsjö Committed by Jean-Baptiste Kempf
Browse files

arm64: mc: Use addp instead of addv+trn1 in warp

Before:           Cortex A53     A72     A73
warp_8x8_8bpc_neon:   1952.8  1161.3  1151.1
warp_8x8t_8bpc_neon:  1937.1  1147.5  1139.0
After:
warp_8x8_8bpc_neon:   1860.8  1068.6  1105.8
warp_8x8t_8bpc_neon:  1846.9  1056.4  1099.8
parent 3489a9c1
Pipeline #9988 passed with stages
in 5 minutes and 18 seconds
......@@ -3007,28 +3007,20 @@ function warp_filter_horz_neon
saddlp v19.4s, v19.8h
mul v22.8h, v22.8h, v5.8h
saddlp v20.4s, v20.8h
addv s23, v23.4s
saddlp v21.4s, v21.8h
addv s18, v18.4s
saddlp v22.4s, v22.8h
addv s19, v19.4s
trn1 v18.2s, v23.2s, v18.2s
addv s20, v20.4s
addp v18.4s, v23.4s, v18.4s
ext v23.16b, v16.16b, v17.16b, #2*6
trn1 v19.2s, v19.2s, v20.2s
addv s21, v21.4s
addp v19.4s, v19.4s, v20.4s
mul v23.8h, v23.8h, v6.8h
ext v20.16b, v16.16b, v17.16b, #2*7
addv s22, v22.4s
mul v20.8h, v20.8h, v7.8h
saddlp v23.4s, v23.8h
trn1 v21.2s, v21.2s, v22.2s
addp v21.4s, v21.4s, v22.4s
saddlp v20.4s, v20.8h
addv s23, v23.4s
addv s20, v20.4s
trn1 v20.2s, v23.2s, v20.2s
trn1 v18.2d, v18.2d, v19.2d
trn1 v20.2d, v21.2d, v20.2d
addp v20.4s, v23.4s, v20.4s
addp v18.4s, v18.4s, v19.4s
addp v20.4s, v21.4s, v20.4s
add w5, w5, w8
......
Markdown is supported
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