aboutsummaryrefslogtreecommitdiff
path: root/string
diff options
context:
space:
mode:
authorWilco Dijkstra <wilco.dijkstra@arm.com>2022-08-22 13:21:30 +0100
committerSzabolcs Nagy <szabolcs.nagy@arm.com>2022-08-23 10:09:05 +0100
commit87e99142bf8273215f58e4770fd7247e4b6e401b (patch)
tree65a03bdeaaab1f5794acb3083847ade7d9182a75 /string
parent802438542e4dff98109027594be5603f13708ab2 (diff)
downloadarm-optimized-routines-87e99142bf8273215f58e4770fd7247e4b6e401b.tar.gz
string: Improve strchrnul-mte
Unroll the main loop, which gives a small gain.
Diffstat (limited to 'string')
-rw-r--r--string/aarch64/strchrnul-mte.S12
1 files changed, 10 insertions, 2 deletions
diff --git a/string/aarch64/strchrnul-mte.S b/string/aarch64/strchrnul-mte.S
index 9be5cbc..b1ac4db 100644
--- a/string/aarch64/strchrnul-mte.S
+++ b/string/aarch64/strchrnul-mte.S
@@ -57,14 +57,22 @@ ENTRY (__strchrnul_aarch64_mte)
.p2align 4
L(loop):
- ldr qdata, [src, 16]!
+ ldr qdata, [src, 16]
+ cmeq vhas_chr.16b, vdata.16b, vrepchr.16b
+ cmhs vhas_chr.16b, vhas_chr.16b, vdata.16b
+ umaxp vend.16b, vhas_chr.16b, vhas_chr.16b
+ fmov tmp1, dend
+ cbnz tmp1, L(end)
+ ldr qdata, [src, 32]!
cmeq vhas_chr.16b, vdata.16b, vrepchr.16b
cmhs vhas_chr.16b, vhas_chr.16b, vdata.16b
umaxp vend.16b, vhas_chr.16b, vhas_chr.16b
fmov tmp1, dend
cbz tmp1, L(loop)
-
+ sub src, src, 16
+L(end):
shrn vend.8b, vhas_chr.8h, 4 /* 128->64 */
+ add src, src, 16
fmov tmp1, dend
#ifndef __AARCH64EB__
rbit tmp1, tmp1