From: Jakub Jelinek <jakub@redhat.com>
To: Ulrich Drepper <drepper@redhat.com>,
Steven Munroe <munroesj@linux.vnet.ibm.com>,
David Woodhouse <dwmw2@infradead.org>
Cc: Glibc hackers <libc-hacker@sources.redhat.com>
Subject: [PATCH] Avoid clobbering call-saved cr registers in ppc hand written assembly
Date: Thu, 12 Jun 2008 19:18:00 -0000 [thread overview]
Message-ID: <20080612192933.GA3726@sunsite.mff.cuni.cz> (raw)
Hi!
gcc says:
cr7, cr6 (not saved or special)
cr1 (not saved, but used for FP operations)
cr0 (not saved, but used for arithmetic operations)
cr4, cr3, cr2 (saved)
...
cr5, r1, r2, ap, xer (fixed)
But apparently some *lround* routines set cr[34] registers, which are
call-saved. This causes e.g. http://bugzilla.redhat.com/450790
From the above I believe using cr1 and cr7 instead of cr3 and cr4 should
be ok (though the patch is untested).
2008-06-12 Jakub Jelinek <jakub@redhat.com>
* sysdeps/powerpc/powerpc64/fpu/s_llround.S (__llround): Avoid using
cr[34] registers.
* sysdeps/powerpc/powerpc64/fpu/s_llroundf.S (__llroundf): Likewise.
* sysdeps/powerpc/powerpc32/power4/fpu/s_llround.S (__llround): Likewise.
* sysdeps/powerpc/powerpc32/fpu/s_lround.S (__lround): Avoid using cr3
register.
--- libc/sysdeps/powerpc/powerpc64/fpu/s_llround.S.jj 2008-04-17 10:32:49.000000000 +0200
+++ libc/sysdeps/powerpc/powerpc64/fpu/s_llround.S 2008-06-12 20:53:40.000000000 +0200
@@ -1,5 +1,5 @@
/* llround function. PowerPC64 version.
- Copyright (C) 2004, 2006, 2007 Free Software Foundation, Inc.
+ Copyright (C) 2004, 2006, 2007, 2008 Free Software Foundation, Inc.
This file is part of the GNU C Library.
The GNU C Library is free software; you can redistribute it and/or
@@ -52,12 +52,12 @@ ENTRY (__llround)
fabs fp2,fp1 /* Get the absolute value of x. */
fsub fp12,fp10,fp10 /* Compute 0.0 into fp12. */
fcmpu cr6,fp2,fp10 /* if |x| < 0.5 */
- fcmpu cr4,fp2,fp9 /* if |x| >= 2^52 */
- fcmpu cr3,fp1,fp12 /* x is negative? x < 0.0 */
+ fcmpu cr7,fp2,fp9 /* if |x| >= 2^52 */
+ fcmpu cr1,fp1,fp12 /* x is negative? x < 0.0 */
blt- cr6,.Lretzero /* 0.5 > x < -0.5 so just return 0. */
- bge- cr4,.Lnobias /* 2^52 > x < -2^52 just convert with no bias. */
+ bge- cr7,.Lnobias /* 2^52 > x < -2^52 just convert with no bias. */
fadd fp3,fp2,fp10 /* |x|+=0.5 bias to prepare to round. */
- bge cr3,.Lconvert /* x is positive so don't negate x. */
+ bge cr1,.Lconvert /* x is positive so don't negate x. */
fnabs fp3,fp3 /* -(|x|+=0.5) */
.Lconvert:
fctidz fp4,fp3 /* Convert to Integer double word round toward 0. */
--- libc/sysdeps/powerpc/powerpc64/fpu/s_llroundf.S.jj 2008-04-17 10:32:49.000000000 +0200
+++ libc/sysdeps/powerpc/powerpc64/fpu/s_llroundf.S 2008-06-12 20:54:17.000000000 +0200
@@ -1,5 +1,5 @@
/* llroundf function. PowerPC64 version.
- Copyright (C) 2004, 2006, 2007 Free Software Foundation, Inc.
+ Copyright (C) 2004, 2006, 2007, 2008 Free Software Foundation, Inc.
This file is part of the GNU C Library.
The GNU C Library is free software; you can redistribute it and/or
@@ -51,12 +51,12 @@ ENTRY (__llroundf)
fabs fp2,fp1 /* Get the absolute value of x. */
fsub fp12,fp10,fp10 /* Compute 0.0 into fp12. */
fcmpu cr6,fp2,fp10 /* if |x| < 0.5 */
- fcmpu cr4,fp2,fp9 /* if |x| >= 2^23 */
- fcmpu cr3,fp1,fp12 /* x is negative? x < 0.0 */
+ fcmpu cr7,fp2,fp9 /* if |x| >= 2^23 */
+ fcmpu cr1,fp1,fp12 /* x is negative? x < 0.0 */
blt- cr6,.Lretzero /* 0.5 > x < -0.5 so just return 0. */
- bge- cr4,.Lnobias /* 2^23 > x < -2^23 just convert with no bias. */
+ bge- cr7,.Lnobias /* 2^23 > x < -2^23 just convert with no bias. */
fadd fp3,fp2,fp10 /* |x|+=0.5 bias to prepare to round. */
- bge cr3,.Lconvert /* x is positive so don't negate x. */
+ bge cr1,.Lconvert /* x is positive so don't negate x. */
fnabs fp3,fp3 /* -(|x|+=0.5) */
.Lconvert:
fctidz fp4,fp3 /* Convert to Integer double word round toward 0. */
--- libc/sysdeps/powerpc/powerpc32/power4/fpu/s_llround.S.jj 2008-04-17 10:32:49.000000000 +0200
+++ libc/sysdeps/powerpc/powerpc32/power4/fpu/s_llround.S 2008-06-12 20:54:50.000000000 +0200
@@ -1,5 +1,5 @@
/* llround function. PowerPC32 on PowerPC64 version.
- Copyright (C) 2004, 2006, 2007 Free Software Foundation, Inc.
+ Copyright (C) 2004, 2006, 2007, 2008 Free Software Foundation, Inc.
This file is part of the GNU C Library.
The GNU C Library is free software; you can redistribute it and/or
@@ -75,12 +75,12 @@ ENTRY (__llround)
fabs fp2,fp1 /* Get the absolute value of x. */
fsub fp12,fp10,fp10 /* Compute 0.0 into fpr12. */
fcmpu cr6,fp2,fp10 /* if |x| < 0.5 */
- fcmpu cr4,fp2,fp9 /* if |x| >= 2^52 */
- fcmpu cr3,fp1,fp12 /* x is negative? x < 0.0 */
+ fcmpu cr7,fp2,fp9 /* if |x| >= 2^52 */
+ fcmpu cr1,fp1,fp12 /* x is negative? x < 0.0 */
blt- cr6,.Lretzero /* 0.5 > x < -0.5 so just return 0. */
- bge- cr4,.Lnobias /* 2^52 > x < -2^52 just convert with no bias. */
+ bge- cr7,.Lnobias /* 2^52 > x < -2^52 just convert with no bias. */
fadd fp3,fp2,fp10 /* |x|+=0.5 bias to prepare to round. */
- bge cr3,.Lconvert /* x is positive so don't negate x. */
+ bge cr1,.Lconvert /* x is positive so don't negate x. */
fnabs fp3,fp3 /* -(|x|+=0.5) */
.Lconvert:
fctidz fp4,fp3 /* Convert to Integer double word round toward 0. */
--- libc/sysdeps/powerpc/powerpc32/fpu/s_lround.S.jj 2008-04-17 10:32:49.000000000 +0200
+++ libc/sysdeps/powerpc/powerpc32/fpu/s_lround.S 2008-06-12 20:55:16.000000000 +0200
@@ -1,5 +1,5 @@
/* lround function. PowerPC32 version.
- Copyright (C) 2004, 2006, 2007 Free Software Foundation, Inc.
+ Copyright (C) 2004, 2006, 2007, 2008 Free Software Foundation, Inc.
This file is part of the GNU C Library.
The GNU C Library is free software; you can redistribute it and/or
@@ -65,10 +65,10 @@ ENTRY (__lround)
fabs fp2, fp1 /* Get the absolute value of x. */
fsub fp12,fp10,fp10 /* Compute 0.0. */
fcmpu cr6, fp2, fp10 /* if |x| < 0.5 */
- fcmpu cr3, fp1, fp12 /* x is negative? x < 0.0 */
+ fcmpu cr7, fp1, fp12 /* x is negative? x < 0.0 */
blt- cr6,.Lretzero
fadd fp3,fp2,fp10 /* |x|+=0.5 bias to prepare to round. */
- bge cr3,.Lconvert /* x is positive so don't negate x. */
+ bge cr7,.Lconvert /* x is positive so don't negate x. */
fnabs fp3,fp3 /* -(|x|+=0.5) */
.Lconvert:
fctiwz fp4,fp3 /* Convert to Integer word lround toward 0. */
Jakub
reply other threads:[~2008-06-12 19:18 UTC|newest]
Thread overview: [no followups] expand[flat|nested] mbox.gz Atom feed
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=20080612192933.GA3726@sunsite.mff.cuni.cz \
--to=jakub@redhat.com \
--cc=drepper@redhat.com \
--cc=dwmw2@infradead.org \
--cc=libc-hacker@sources.redhat.com \
--cc=munroesj@linux.vnet.ibm.com \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for read-only IMAP folder(s) and NNTP newsgroup(s).