This is the mail archive of the
libc-alpha@sourceware.org
mailing list for the glibc project.
Re: PATCH: PR libc/10162: Use a simple loop on data shorter thansoftware pipeline
- From: "H.J. Lu" <hongjiu dot lu at intel dot com>
- To: "H.J. Lu" <hjl dot tools at gmail dot com>
- Cc: GNU C Library <libc-alpha at sourceware dot org>
- Date: Mon, 9 Nov 2009 12:37:41 -0800
- Subject: Re: PATCH: PR libc/10162: Use a simple loop on data shorter thansoftware pipeline
- References: <20091109195706.GA29166@lucon.org>
- Reply-to: "H.J. Lu" <hjl dot tools at gmail dot com>
On Mon, Nov 09, 2009 at 11:57:06AM -0800, H.J. Lu wrote:
> When data is shorter than software pipeline, recovery may fail. This
> patch avoids it by using a simple loop on data shorter than software
> pipeline.
>
Here is an updated patch. This one properly handles the shorter read
on speculative load.
H.J.
---
2009-11-09 H.J. Lu <hongjiu.lu@intel.com>
PR libc/10162
* sysdeps/ia64/memchr.S: Properly recover from shorter read.
diff --git a/sysdeps/ia64/memchr.S b/sysdeps/ia64/memchr.S
index cd062b2..cdd71ca 100644
--- a/sysdeps/ia64/memchr.S
+++ b/sysdeps/ia64/memchr.S
@@ -126,7 +126,16 @@ ENTRY(__memchr)
br.ret.sptk.many b0
.recovery:
- adds ret0 = -((MEMLAT + 1) * 8), ret0;;
+#if MEMLAT != 6
+# error "MEMLAT must be 6!"
+#endif
+(p[MEMLAT-6]) add ret0 = -8, ret0;;
+(p[MEMLAT-5]) add ret0 = -8, ret0;;
+(p[MEMLAT-4]) add ret0 = -8, ret0;;
+(p[MEMLAT-3]) add ret0 = -8, ret0;;
+(p[MEMLAT-2]) add ret0 = -8, ret0;;
+(p[MEMLAT-1]) add ret0 = -8, ret0;;
+(p[MEMLAT]) add ret0 = -8, ret0;;
(p[MEMLAT+1]) add ret0 = -8, ret0;;
(p[MEMLAT+2]) add ret0 = -8, ret0;;
.l4: