diff options
author | Arpad Panyik <Arpad.Panyik@arm.com> | 2024-04-25 16:54:13 +0200 |
---|---|---|
committer | Arpad Panyik <Arpad.Panyik@arm.com> | 2024-04-25 17:02:09 +0200 |
commit | a40301b33f1b5543f8b6cab5587a184748f99ce5 (patch) | |
tree | bb9da1bb654ce7bd3760e3f9e1e558e24c8db8b0 | |
parent | b0685c387d68e4a215cec484f806ee4ab3a450d4 (diff) | |
download | libdav1d-a40301b33f1b5543f8b6cab5587a184748f99ce5.tar.gz |
AArch64: Simplify loads in *hv_filter* of DotProd path
Simplify the load sequences in *hv_filter* functions (ldr + add -> ld1)
to be more uniform and smaller. Performance is not affected.
-rw-r--r-- | src/arm/64/mc_dotprod.S | 3 |
1 files changed, 1 insertions, 2 deletions
diff --git a/src/arm/64/mc_dotprod.S b/src/arm/64/mc_dotprod.S index e076abf..31abe62 100644 --- a/src/arm/64/mc_dotprod.S +++ b/src/arm/64/mc_dotprod.S @@ -968,8 +968,7 @@ L(\type\()_6tap_hv_\isa): .align FUNC_ALIGN L(\type\()_hv_filter8_\isa): - ldr q4, [\lsrc] - add \lsrc, \lsrc, \s_strd + ld1 {v4.16b}, [\lsrc], \s_strd sub v4.16b, v4.16b, v24.16b mov v22.16b, v27.16b mov v23.16b, v27.16b |