Commit ff41197b authored by Martin Storsjö's avatar Martin Storsjö
Browse files

arm64: mc: Schedule instructions better in the warp8x8 functions

Before:           Cortex A53     A72     A73
warp_8x8_8bpc_neon:   1997.3  1170.1  1199.9
warp_8x8t_8bpc_neon:  1982.4  1171.5  1192.6
After:
warp_8x8_8bpc_neon:   1954.6  1159.2  1153.3
warp_8x8t_8bpc_neon:  1938.5  1146.2  1136.7
parent e570088d
Pipeline #9947 passed with stages
in 5 minutes and 41 seconds
......@@ -2975,7 +2975,9 @@ function warp_filter_horz_neon
ld1 {v16.8b, v17.8b}, [x2], x3
load_filter_row d0, w12, w7
uxtl v16.8h, v16.8b
load_filter_row d1, w12, w7
uxtl v17.8h, v17.8b
load_filter_row d2, w12, w7
sxtl v0.8h, v0.8b
load_filter_row d3, w12, w7
......@@ -2988,16 +2990,12 @@ function warp_filter_horz_neon
sxtl v4.8h, v4.8b
load_filter_row d7, w12, w7
sxtl v5.8h, v5.8b
sxtl v6.8h, v6.8b
sxtl v7.8h, v7.8b
uxtl v16.8h, v16.8b
uxtl v17.8h, v17.8b
ext v18.16b, v16.16b, v17.16b, #2*1
mul v23.8h, v16.8h, v0.8h
sxtl v6.8h, v6.8b
ext v19.16b, v16.16b, v17.16b, #2*2
mul v18.8h, v18.8h, v1.8h
sxtl v7.8h, v7.8b
ext v20.16b, v16.16b, v17.16b, #2*3
mul v19.8h, v19.8h, v2.8h
ext v21.16b, v16.16b, v17.16b, #2*4
......
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