[Bug target/35839] New: Revision 133909 breaks bootstrap on powerpc-apple-darwin

[Bug target/35839] New: Revision 133909 breaks bootstrap on powerpc-apple-darwin

[dominiq at lps dot ens dot fr]

After revision 133909 building libgfortran is broken at:

...
libtool: compile:  /opt/gcc/darwin_buildw/./gcc/xgcc
-B/opt/gcc/darwin_buildw/./gcc/ -B/opt/gcc/gcc4.4w/powerpc-apple-darwin9/bin/
-B/opt/gcc/gcc4.4w/powerpc-apple-darwin9/lib/ -isystem
/opt/gcc/gcc4.4w/powerpc-apple-darwin9/include -isystem
/opt/gcc/gcc4.4w/powerpc-apple-darwin9/sys-include -DHAVE_CONFIG_H -I.
-I../../../gcc-4.4-work/libgfortran -I.
-iquote../../../gcc-4.4-work/libgfortran/io
-I../../../gcc-4.4-work/libgfortran/../gcc
-I../../../gcc-4.4-work/libgfortran/../gcc/config -I../.././gcc -D_GNU_SOURCE
-std=gnu99 -Wall -Wstrict-prototypes -Wmissing-prototypes
-Wold-style-definition -Wextra -Wwrite-strings -fcx-fortran-rules
-ftree-vectorize -funroll-loops -g -O2 -MT matmul_i4.lo -MD -MP -MF
.deps/matmul_i4.Tpo -c ../../../gcc-4.4-work/libgfortran/generated/matmul_i4.c 
-fno-common -DPIC -o .libs/matmul_i4.o
../../../gcc-4.4-work/libgfortran/generated/matmul_i4.c: In function
'matmul_i4':
../../../gcc-4.4-work/libgfortran/generated/matmul_i4.c:87: internal compiler
error: in rs6000_check_sdmode, at config/rs6000/rs6000.c:11229
Please submit a full bug report,
with preprocessed source if appropriate.
See <http://gcc.gnu.org/bugs.html> for instructions.
make[3]: *** [matmul_i4.lo] Error 1
make[2]: *** [all] Error 2
make[1]: *** [all-target-libgfortran] Error 2
make: *** [all] Error 2


-- 
           Summary: Revision 133909 breaks bootstrap on powerpc-apple-darwin
           Product: gcc
           Version: 4.4.0
            Status: UNCONFIRMED
          Severity: normal
          Priority: P3
         Component: target
        AssignedTo: unassigned at gcc dot gnu dot org
        ReportedBy: dominiq at lps dot ens dot fr
 GCC build triplet: powerpc-apple-darwin9
  GCC host triplet: powerpc-apple-darwin9
GCC target triplet: powerpc-apple-darwin9


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=35839

[Bug target/35839] Altivec with the vectorizer causes an ICE in rs6000_check_sdmode

[pinskia at gcc dot gnu dot org]

------- Comment #1 from pinskia at gcc dot gnu dot org  2008-04-06 19:19 -------
Since powerpc-apple-darwin9 enables Altivec by default, the vectorizer happens
in libgfortran.

Almost all the vectorizer tests are now failing too.


-- 

pinskia at gcc dot gnu dot org changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
           Severity|normal                      |blocker
             Status|UNCONFIRMED                 |NEW
     Ever Confirmed|0                           |1
  GCC build triplet|powerpc-apple-darwin9       |
   GCC host triplet|powerpc-apple-darwin9       |
 GCC target triplet|powerpc-apple-darwin9       |powerpc*-*-*
           Keywords|                            |build, ice-on-valid-code
   Last reconfirmed|0000-00-00 00:00:00         |2008-04-06 19:19:04
               date|                            |
            Summary|Revision 133909 breaks      |Altivec with the vectorizer
                   |bootstrap on powerpc-apple- |causes an ICE in
                   |darwin                      |rs6000_check_sdmode
   Target Milestone|---                         |4.4.0


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=35839

[Bug target/35839] Altivec with the vectorizer causes an ICE in rs6000_check_sdmode

[pinskia at gcc dot gnu dot org]

------- Comment #2 from pinskia at gcc dot gnu dot org  2008-04-06 19:32 -------
Simple testcase:
void
foo (float a[32], float b[2][32])
{
  int i;
  for (i = 0; i < 32; i++)
    a[i] = (b[0][i] > b[1][i]) ? b[0][i] : b[1][i];
}

Compile with -O2 -maltivec -ftree-vectorize and you get the failure.
We have a ALIGN_INDIRECT_REF here.


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=35839

[Bug target/35839] [4.4 Regression] Altivec with the vectorizer causes an ICE in rs6000_check_sdmode

[janis at gcc dot gnu dot org]

------- Comment #3 from janis at gcc dot gnu dot org  2008-04-07 16:48 -------
Argh, this is an excellent of example of why full testing is necessary for
small changes to a patch that has been fully tested; I thought I could get away
with just bootstrapping C and running tests that use decimal float types for
the modified patch.  My build of libgfortran on powerpc64-linux succeeds, but I
get lots of failures in vect tests.  I'll get to this right away.


-- 

janis at gcc dot gnu dot org changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
         AssignedTo|unassigned at gcc dot gnu   |janis at gcc dot gnu dot org
                   |dot org                     |
             Status|NEW                         |ASSIGNED
   Last reconfirmed|2008-04-06 19:19:04         |2008-04-07 16:48:41
               date|                            |


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=35839

[Bug target/35839] [4.4 Regression] Altivec with the vectorizer causes an ICE in rs6000_check_sdmode

[dominiq at lps dot ens dot fr]

------- Comment #4 from dominiq at lps dot ens dot fr  2008-04-08 15:06 -------
My understanding of "gcc_assert" is that it is meant to trigger ICEs for
forbidden paths. Hence introducing new "gcc_assert" is likely to trigger ICEs,
including during bootstrap, either because of a rampant bug or because the
condition is too broad, as it seems to be the case here. So I think maintainers
should be very careful when they add new "gcc_assert" and monitor the
gcc-testresults mailing list for missing regular testers.

In particular, such modification on powerpc should be correlated to "regress":

regress

2008-04-01 07:19 Results for 4.4.0 20080331 (experimental) (GCC) testsuite on
powerpc-apple-darwin8.5.0
2008-04-01 16:27 Results for 4.4.0 20080401 (experimental) (GCC) testsuite on
powerpc-apple-darwin8.5.0
2008-04-03 13:29 Results for 4.4.0 20080403 (experimental) (GCC) testsuite on
powerpc-apple-darwin8.5.0
2008-04-03 21:27 Results for 4.4.0 20080403 (experimental) (GCC) testsuite on
powerpc-apple-darwin8.5.0
2008-04-04 05:26 Results for 4.4.0 20080403 (experimental) (GCC) testsuite on
powerpc-apple-darwin8.5.0
2008-04-04 13:25 Results for 4.4.0 20080404 (experimental) (GCC) testsuite on
powerpc-apple-darwin8.5.0
2008-04-04 21:23 Results for 4.4.0 20080404 (experimental) (GCC) testsuite on
powerpc-apple-darwin8.5.0

which gives a new result every ~8 hours and has stopped to do so since April
4th (this is why I did a fresh bootstrap in order to know what was going
wrong).


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=35839

[Bug target/35839] [4.4 Regression] Altivec with the vectorizer causes an ICE in rs6000_check_sdmode

[aj at gcc dot gnu dot org]

------- Comment #5 from aj at gcc dot gnu dot org  2008-04-08 15:35 -------
Bootstrap fails on powerpc64-suse-linux-gnu with bootstrapping the 64-bit
compiler as well at the same place.


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=35839

[Bug target/35839] [4.4 Regression] Altivec with the vectorizer causes an ICE in rs6000_check_sdmode

[janis at gcc dot gnu dot org]

------- Comment #6 from janis at gcc dot gnu dot org  2008-04-08 20:51 -------
My bootstrap on powerpc64-linux worked fine after the fix for 35620 went in;
see http://gcc.gnu.org/ml/gcc-testresults/2008-04/msg00415.html for revision
133952.  Since the following day, however, my bootstraps have failed building
64-bit libgfortran with an ICE for a libgfortran configure check:

conftest.F: In function 'main':
conftest.F:1: internal compiler error: in rs6000_output_function_epilogue, at
config/rs6000/rs6000.c:16855
Please submit a full bug report,
with preprocessed source if appropriate.
See <http://gcc.gnu.org/bugs.html> for instructions.
configure:10346: $? = 1
configure: failed program was:
|       program main
| #ifndef __GNUC__
|        choke me
| #endif
| 
|       end

It also fails for just

      program main
      end

I'm doing a regression hunt to find out what caused that.

In the meantime, this has made it more difficult to test the patch to fix the
bug I introduced in rs6000_check_sdmode,
http://gcc.gnu.org/ml/gcc-patches/2008-04/msg00584.html.  It seems like an
obvious fix but it hasn't yet been reviewed.


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=35839

Re: [Bug target/35839] [4.4 Regression] Altivec with the vectorizer causes an ICE in rs6000_check_sdmode

[Andrew Pinski]

Sent from my iPhone

On Apr 8, 2008, at 13:51, "janis at gcc dot gnu dot org" <gcc-bugzilla@gcc.gnu.org 
 > wrote:

>
>
> ------- Comment #6 from janis at gcc dot gnu dot org  2008-04-08  
> 20:51 -------
> My bootstrap on powerpc64-linux worked fine after the fix for 35620  
> went in;
> see http://gcc.gnu.org/ml/gcc-testresults/2008-04/msg00415.html for  
> revision
> 133952.  Since the following day, however, my bootstraps have failed  
> building
> 64-bit libgfortran with an ICE for a libgfortran configure check:
>
> conftest.F: In function 'main':
> conftest.F:1: internal compiler error: in  
> rs6000_output_function_epilogue, at
> config/rs6000/rs6000.c:16855
> Please submit a full bug report,
> with preprocessed source if appropriate.
> See <http://gcc.gnu.org/bugs.html> for instructions.
> configure:10346: $? = 1
> configure: failed program was:
> |       program main
> | #ifndef __GNUC__
> |        choke me
> | #endif
> |
> |       end
>
> It also fails for just
>
>      program main
>      end
>
> I'm doing a regression hunt to find out what caused that.


No need this is caused by my rs6000 patch.

>
>
> In the meantime, this has made it more difficult to test the patch  
> to fix the
> bug I introduced in rs6000_check_sdmode,
> http://gcc.gnu.org/ml/gcc-patches/2008-04/msg00584.html.  It seems  
> like an
> obvious fix but it hasn't yet been reviewed.
>
>
> -- 
>
>
> http://gcc.gnu.org/bugzilla/show_bug.cgi?id=35839
>

[Bug target/35839] [4.4 Regression] Altivec with the vectorizer causes an ICE in rs6000_check_sdmode

[pinskia at gmail dot com]

------- Comment #7 from pinskia at gmail dot com  2008-04-08 21:05 -------
Subject: Re:  [4.4 Regression] Altivec with the vectorizer causes an ICE in
rs6000_check_sdmode



Sent from my iPhone

On Apr 8, 2008, at 13:51, "janis at gcc dot gnu dot org"
<gcc-bugzilla@gcc.gnu.org 
 > wrote:

>
>
> ------- Comment #6 from janis at gcc dot gnu dot org  2008-04-08  
> 20:51 -------
> My bootstrap on powerpc64-linux worked fine after the fix for 35620  
> went in;
> see http://gcc.gnu.org/ml/gcc-testresults/2008-04/msg00415.html for  
> revision
> 133952.  Since the following day, however, my bootstraps have failed  
> building
> 64-bit libgfortran with an ICE for a libgfortran configure check:
>
> conftest.F: In function 'main':
> conftest.F:1: internal compiler error: in  
> rs6000_output_function_epilogue, at
> config/rs6000/rs6000.c:16855
> Please submit a full bug report,
> with preprocessed source if appropriate.
> See <http://gcc.gnu.org/bugs.html> for instructions.
> configure:10346: $? = 1
> configure: failed program was:
> |       program main
> | #ifndef __GNUC__
> |        choke me
> | #endif
> |
> |       end
>
> It also fails for just
>
>      program main
>      end
>
> I'm doing a regression hunt to find out what caused that.


No need this is caused by my rs6000 patch.

>
>
> In the meantime, this has made it more difficult to test the patch  
> to fix the
> bug I introduced in rs6000_check_sdmode,
> http://gcc.gnu.org/ml/gcc-patches/2008-04/msg00584.html.  It seems  
> like an
> obvious fix but it hasn't yet been reviewed.
>
>
> -- 
>
>
> http://gcc.gnu.org/bugzilla/show_bug.cgi?id=35839
>


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=35839

Re: [Bug target/35839] [4.4 Regression] Altivec with the vectorizer causes an ICE in rs6000_check_sdmode

[Andrew Pinski]

Sent from my iPhone

On Apr 8, 2008, at 14:05, "pinskia at gmail dot com" <gcc-bugzilla@gcc.gnu.org 
 > wrote:

>
>
> ------- Comment #7 from pinskia at gmail dot com  2008-04-08 21:05  
> -------
> Subject: Re:  [4.4 Regression] Altivec with the vectorizer causes an  
> ICE in
> rs6000_check_sdmode
>
>
>
> Sent from my iPhone
>
> On Apr 8, 2008, at 13:51, "janis at gcc dot gnu dot org"
> <gcc-bugzilla@gcc.gnu.org
>> wrote:
>
>>
>>
>> ------- Comment #6 from janis at gcc dot gnu dot org  2008-04-08
>> 20:51 -------
>> My bootstrap on powerpc64-linux worked fine after the fix for 35620
>> went in;
>> see http://gcc.gnu.org/ml/gcc-testresults/2008-04/msg00415.html for
>> revision
>> 133952.  Since the following day, however, my bootstraps have failed
>> building
>> 64-bit libgfortran with an ICE for a libgfortran configure check:
>>
>> conftest.F: In function 'main':
>> conftest.F:1: internal compiler error: in
>> rs6000_output_function_epilogue, at
>> config/rs6000/rs6000.c:16855
>> Please submit a full bug report,
>> with preprocessed source if appropriate.
>> See <http://gcc.gnu.org/bugs.html> for instructions.
>> configure:10346: $? = 1
>> configure: failed program was:
>> |       program main
>> | #ifndef __GNUC__
>> |        choke me
>> | #endif
>> |
>> |       end
>>
>> It also fails for just
>>
>>     program main
>>     end
>>
>> I'm doing a regression hunt to find out what caused that.
>
>
> No need this is caused by my rs6000 patch.

Actually I take that back but it was just fixed by David's fix to  
rs6000.c.


>
>
>>
>>
>> In the meantime, this has made it more difficult to test the patch
>> to fix the
>> bug I introduced in rs6000_check_sdmode,
>> http://gcc.gnu.org/ml/gcc-patches/2008-04/msg00584.html.  It seems
>> like an
>> obvious fix but it hasn't yet been reviewed.
>>
>>
>> -- 
>>
>>
>> http://gcc.gnu.org/bugzilla/show_bug.cgi?id=35839
>>
>
>
> -- 
>
>
> http://gcc.gnu.org/bugzilla/show_bug.cgi?id=35839
>

[Bug target/35839] [4.4 Regression] Altivec with the vectorizer causes an ICE in rs6000_check_sdmode

[pinskia at gmail dot com]

------- Comment #8 from pinskia at gmail dot com  2008-04-08 21:10 -------
Subject: Re:  [4.4 Regression] Altivec with the vectorizer causes an ICE in
rs6000_check_sdmode



Sent from my iPhone

On Apr 8, 2008, at 14:05, "pinskia at gmail dot com" <gcc-bugzilla@gcc.gnu.org 
 > wrote:

>
>
> ------- Comment #7 from pinskia at gmail dot com  2008-04-08 21:05  
> -------
> Subject: Re:  [4.4 Regression] Altivec with the vectorizer causes an  
> ICE in
> rs6000_check_sdmode
>
>
>
> Sent from my iPhone
>
> On Apr 8, 2008, at 13:51, "janis at gcc dot gnu dot org"
> <gcc-bugzilla@gcc.gnu.org
>> wrote:
>
>>
>>
>> ------- Comment #6 from janis at gcc dot gnu dot org  2008-04-08
>> 20:51 -------
>> My bootstrap on powerpc64-linux worked fine after the fix for 35620
>> went in;
>> see http://gcc.gnu.org/ml/gcc-testresults/2008-04/msg00415.html for
>> revision
>> 133952.  Since the following day, however, my bootstraps have failed
>> building
>> 64-bit libgfortran with an ICE for a libgfortran configure check:
>>
>> conftest.F: In function 'main':
>> conftest.F:1: internal compiler error: in
>> rs6000_output_function_epilogue, at
>> config/rs6000/rs6000.c:16855
>> Please submit a full bug report,
>> with preprocessed source if appropriate.
>> See <http://gcc.gnu.org/bugs.html> for instructions.
>> configure:10346: $? = 1
>> configure: failed program was:
>> |       program main
>> | #ifndef __GNUC__
>> |        choke me
>> | #endif
>> |
>> |       end
>>
>> It also fails for just
>>
>>     program main
>>     end
>>
>> I'm doing a regression hunt to find out what caused that.
>
>
> No need this is caused by my rs6000 patch.

Actually I take that back but it was just fixed by David's fix to  
rs6000.c.


>
>
>>
>>
>> In the meantime, this has made it more difficult to test the patch
>> to fix the
>> bug I introduced in rs6000_check_sdmode,
>> http://gcc.gnu.org/ml/gcc-patches/2008-04/msg00584.html.  It seems
>> like an
>> obvious fix but it hasn't yet been reviewed.
>>
>>
>> -- 
>>
>>
>> http://gcc.gnu.org/bugzilla/show_bug.cgi?id=35839
>>
>
>
> -- 
>
>
> http://gcc.gnu.org/bugzilla/show_bug.cgi?id=35839
>


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=35839

[Bug target/35839] [4.4 Regression] Altivec with the vectorizer causes an ICE in rs6000_check_sdmode

[janis at gcc dot gnu dot org]

------- Comment #9 from janis at gcc dot gnu dot org  2008-04-08 21:18 -------
Subject: Bug 35839

Author: janis
Date: Tue Apr  8 21:17:16 2008
New Revision: 134107

URL: http://gcc.gnu.org/viewcvs?root=gcc&view=rev&rev=134107
Log:
        PR target/35839
        * config/rs6000/rs6000.c (rs6000_check_sdmode): Handle additional
        kinds of indirect references.

Modified:
    trunk/gcc/ChangeLog
    trunk/gcc/config/rs6000/rs6000.c


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=35839

[Bug target/35839] [4.4 Regression] Altivec with the vectorizer causes an ICE in rs6000_check_sdmode

[rguenth at gcc dot gnu dot org]

------- Comment #10 from rguenth at gcc dot gnu dot org  2008-04-20 20:43 -------
I supposed this is fixed?


-- 

rguenth at gcc dot gnu dot org changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
           Severity|blocker                     |normal
             Status|ASSIGNED                    |WAITING


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=35839

[Bug target/35839] [4.4 Regression] Altivec with the vectorizer causes an ICE in rs6000_check_sdmode

[dominiq at lps dot ens dot fr]

------- Comment #11 from dominiq at lps dot ens dot fr  2008-04-20 21:13 -------
> I supposed this is fixed?

I think so.


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=35839

[Bug target/35839] [4.4 Regression] Altivec with the vectorizer causes an ICE in rs6000_check_sdmode

[rguenth at gcc dot gnu dot org]

------- Comment #12 from rguenth at gcc dot gnu dot org  2008-05-16 19:33 -------
Fixed.


-- 

rguenth at gcc dot gnu dot org changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|WAITING                     |RESOLVED
         Resolution|                            |FIXED


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=35839