summaryrefslogtreecommitdiff
path: root/lib/libc
diff options
context:
space:
mode:
authorrin <rin@NetBSD.org>2022-05-30 14:43:37 +0000
committerrin <rin@NetBSD.org>2022-05-30 14:43:37 +0000
commitde6f934b84dcd69e441882c19146ffed4f2e20e2 (patch)
tree2186aeebb99ffb58862ae6cb05b62ed941412aa9 /lib/libc
parent1af073c0238e77b2e27ed7a9ad2bdcba284eb892 (diff)
Obsolete hack for evbppc to replace memcmp(9), memcpy(9), and memmove(9)
with strictly-aligned versions. Now all 32-bit powerpc ports share the same libc binary. This change together with the preceding similar change in libkern slightly improve performance for DHT (ibm4xx/405GP) and RB800 (MPC8533E). See changes in bytebench scores: - DHT https://gist.github.com/rokuyama/301063355de9733bea515b84ef574c0a - RB800 https://gist.github.com/rokuyama/60ad665d367d6d110b79ec44707f39ff Improvements may be negligible, but this does not cause performance regressions at least. This hack was for 403, but unaligned memory access is now emulated by kernel. This should result in serious performance regression for 403. We will provide strictly-aligned versions by ld.so.conf.
Diffstat (limited to 'lib/libc')
-rw-r--r--lib/libc/arch/powerpc/string/Makefile.inc13
1 files changed, 2 insertions, 11 deletions
diff --git a/lib/libc/arch/powerpc/string/Makefile.inc b/lib/libc/arch/powerpc/string/Makefile.inc
index cb970cd7665..66e4bdc6be1 100644
--- a/lib/libc/arch/powerpc/string/Makefile.inc
+++ b/lib/libc/arch/powerpc/string/Makefile.inc
@@ -1,16 +1,7 @@
-# $NetBSD: Makefile.inc,v 1.15 2021/07/26 12:49:13 rin Exp $
+# $NetBSD: Makefile.inc,v 1.16 2022/05/30 14:43:37 rin Exp $
SRCS+= bzero.S ffs.S strlen.S
NO_SRCS+= memset.S
-# XXX
-# Disable asm versions that use unaligned memory access and thus break 403.
-.if ${MACHINE} == "evbppc"
-. for name in bcopy memcmp memcpy memmove
-. for suffix in o po pico go d
-${name}.${suffix}: ${name}.c
-. endfor
-. endfor
-.else
+# with unaligned memory access
SRCS+= memcmp.S bcopy.S memcpy.S memmove.S
-.endif