Re: optimization/6368: ia64 bad code for LAPACK with -O2 -funroll-all-loops

Re: optimization/6368: ia64 bad code for LAPACK with -O2 -funroll-all-loops

[janis]

Synopsis: ia64 bad code for LAPACK with -O2 -funroll-all-loops

Responsible-Changed-From-To: unassigned->janis
Responsible-Changed-By: janis
Responsible-Changed-When: Wed Oct  9 16:55:06 2002
Responsible-Changed-Why:
    ready to close
State-Changed-From-To: open->closed
State-Changed-By: janis
State-Changed-When: Wed Oct  9 16:55:06 2002
State-Changed-Why:
    The problem no longer occurs with the mainline or the 3.2
    branch as of 20021008.  I have no idea which patch fixed it.

http://gcc.gnu.org/cgi-bin/gnatsweb.pl?cmd=view%20audit-trail&database=gcc&pr=6368

Re: optimization/6368: ia64 bad code for LAPACK with -O2 -funroll-all-loops

[Billinghurst, David (CRTS)]

The following reply was made to PR optimization/6368; it has been noted by GNATS.

From: "Billinghurst, David (CRTS)" <David.Billinghurst@riotinto.com>
To: <janis187@us.ibm.com>,
	<gcc-gnats@gcc.gnu.org>
Cc:  
Subject: Re: optimization/6368: ia64 bad code for LAPACK with -O2 -funroll-all-loops
Date: Mon, 22 Apr 2002 00:33:24 +1000

 This may be related to either=20
  - PR/5651 -funroll-loops (fixed about the time you reported this)
  - PR/6304 which is irix -mabi=3D64 -O2 LAPACK failure
 

RE: optimization/6368: ia64 bad code for LAPACK with -O2 -funroll-all-loops

[Billinghurst, David (CRTS)]

The following reply was made to PR optimization/6368; it has been noted by GNATS.

From: "Billinghurst, David (CRTS)" <David.Billinghurst@riotinto.com>
To: <janis187@us.ibm.com>,
	<gcc-gnats@gcc.gnu.org>
Cc:  
Subject: RE: optimization/6368: ia64 bad code for LAPACK with -O2 -funroll-all-loops
Date: Fri, 19 Apr 2002 11:00:16 +1000

 Janis,
 
 I have attached one of tools for tracking down LAPACK errors.
 It is cruse, but it saves my time.
 
 You will need to adapt the makefile for the particular binary
 under test, but it is just cut and paste from the TESTING/*/Makefile
 
 The basic idea is to use a know good LAPACK library with each
 individual file from the bad LAPACK library.  Then by=20
 comparing the .out files, the bad routine(s) can be identified. =20
 
 make xeigtst-lapack-routines.txt
 for f in `cat xeigtst-lapack-routines.txt`
 do
   make F=3D${f} ssep-${f}.out
   rm xeigtsts-${f} core
 done
 
 
 ################### Makefile########################
 # LAPACK source tree
 SRC =3D /usr/local/imports/LAPACK
 
 # Bad LAPACK object tree
 L =3D /exd4/billingd/obj/LAPACK-gcc
 
 # Good LAPACK object tree
 L0 =3D /exd4/billingd/obj/LAPACK-gcc-O0
 
 vpath %.o $(L0)/TESTING/EIG:$(L0)/SRC
 vpath %.in .:$(SRC)/TESTING
 
 BLAS   =3D $(L0)/blas.a
 LAPACK =3D $(L0)/lapack.a
 TMGLIB =3D $(L0)/tmglib.a
 
 LIBS =3D $(TMGLIB) $(LAPACK) $(BLAS)
 
 # GCC object tree under test
 SS      =3D /exd4/billingd/obj/gcc
 
 
 # These options produce the failure
 FC      =3D $(SS)/gcc/g77 -B$(SS)/gcc/ -L$(SS)/mips-sgi-irix6.5/libf2c
 FFLAGS  =3D  -O3 -funroll-loops
 LD      =3D $(FC)
 LDFLAGS =3D -L$(SS)/mips-sgi-irix6.5/libf2c \
           -L$(SS)/mips-sgi-irix6.5/libf2c/.libs=20
 
 ENV =3D =
 LD_LIBRARYN32_PATH=3D$(SS)/mips-sgi-irix6.5/libf2c/.libs:$(SS)/gcc
 
 # Object files for xeigtst
 XOBJS =3D schkee.o sbdt01.o sbdt02.o sbdt03.o schkbb.o schkbd.o schkbk.o =
 \
         schkbl.o schkec.o schkgg.o schkgk.o schkgl.o schkhs.o schksb.o \
         schkst.o sckglm.o sckgqr.o sckgsv.o scklse.o sdrges.o sdrgev.o \
         sdrgsx.o sdrgvx.o sdrvbd.o sdrves.o sdrvev.o sdrvgg.o sdrvsg.o \
         sdrvst.o sdrvsx.o sdrvvx.o serrbd.o serrec.o serred.o serrgg.o \
         serrhs.o serrst.o sget02.o sget10.o sget22.o sget23.o sget24.o \
         sget31.o sget32.o sget33.o sget34.o sget35.o sget36.o sget37.o \
         sget38.o sget39.o sget51.o sget52.o sget53.o sget54.o sglmts.o \
         sgqrts.o sgrqts.o sgsvts.o shst01.o slarfy.o slarhs.o slatm4.o \
         slctes.o slctsx.o slsets.o sort01.o sort03.o ssbt21.o ssgt01.o \
         sslect.o sspt21.o sstt21.o sstt22.o ssyt21.o ssyt22.o slafts.o \
         slahd2.o slasum.o slatb9.o sstech.o sstect.o ssvdch.o ssvdct.o \
         ssxt1.o alahdg.o alasum.o alasvm.o alareq.o ilaenv.o xerbla.o \
         xlaenv.o chkxer.o
 
 xeigtsts: $(XOBJS)
 	$(LD) $(LDFLAGS) -o $@ $+ $(LIBS)
 
 ssep%.out: xeigtsts sep%.in
 	$(ENV) $(filter xeigtsts%, $+) < $(filter %.in, $+) > $@
 
 xeigtsts-$(F): $(F).o $(XOBJS)
 	$(LD) $(LDFLAGS) -o $@ $+ $(LIBS)
 
 ssep-$(F).out: xeigtsts-$(F) sep.in
 	$(ENV) $(filter xeigtsts%, $+) < $(filter %.in, $+) > $@
 
 ssg-$(F).out: xeigtsts-$(F) ssg.in
 	$(ENV) $(filter xeigtsts%, $+) < $(filter %.in, $+) > $@
 
 #----------------------------------------
 lapack-routines.txt:
 	(cd $(L)/SRC; ls ) | sed -e 's/\.o//' > $@
 
 xeigtst-nm.txt: $(L)/TESTING/xeigtsts
 	nm $+ > $@
 
 xeigtst-routines.txt: xeigtst-nm.txt=20
 	grep FUNC $+ | sed -e 's/|/ /g' \
 	|  gawk '{print $$8}' | sed -e 's/_$$//' | sort > $@
 
 xeigtst-lapack-routines.txt: lapack-routines.txt xeigtst-routines.txt
 	comm -12 $+ > $@
 
 %.o: %.f
 	$(FC) $(FFLAGS) -c -o $@ $+
 
 
 

optimization/6368: ia64 bad code for LAPACK with -O2 -funroll-all-loops

[janis187]

>Number:         6368
>Category:       optimization
>Synopsis:       ia64 bad code for LAPACK with -O2 -funroll-all-loops
>Confidential:   no
>Severity:       serious
>Priority:       medium
>Responsible:    unassigned
>State:          open
>Class:          wrong-code
>Submitter-Id:   net
>Arrival-Date:   Thu Apr 18 17:26:01 PDT 2002
>Closed-Date:
>Last-Modified:
>Originator:     Janis Johnson
>Release:        gcc version 3.1 20020418 (prerelease)
>Organization:
>Environment:
ia64-unknown-linux-gnu
Itanium, Red Hat Linux release 7.1.94 (Roswell)
>Description:
LAPACK test output file csb.out reports 120 unexpected
failures for ia64-linux when the tests are compiled with
-O2 -funroll-all-loops by GCC-3.1-20020418.  There are many
other failures as well, possibly all related to the one
described here.

LAPACK takes a couple of hours to build and test on an
Itanium system and I haven't yet checked to see if this is
a regression from 3.0.4.

I can continue to try to come up with a small test case but
in the meantime, here's a way to reproduce the problem.  If
someone looks into this using the full LAPACK distribution,
please let me know if my further efforts are not necessary.
>How-To-Repeat:
Build and test LAPACK as shown by the LAPACK build and test
guide (http://gcc.gnu.org/gcc-3.1/testing-lapack.html).
Verify that TESTING/csb.out shows no failures.  Modify the
OPTS line in make.inc to use -O2 -funroll-all-loops and do:

  cd TESTING/EIG
  rm chbt21.f
  make complex
  cd ..
  make csb.out

This time csb.out will say that 120 out of 540 tests failed
to pass the threshold.
>Fix:
Unknown
>Release-Note:
>Audit-Trail:
>Unformatted: