This is the mail archive of the
fortran@gcc.gnu.org
mailing list for the GNU Fortran project.
Re: Fortran whole file patch
- From: Jan Hubicka <hubicka at ucw dot cz>
- To: Paul Richard Thomas <paul dot richard dot thomas at gmail dot com>
- Cc: Dominique Dhumieres <dominiq at lps dot ens dot fr>, fortran at gcc dot gnu dot org, hubicka at ucw dot cz, Richard Guenther <richard dot guenther at gmail dot com>, Daniel Franke <franke dot daniel at gmail dot com>
- Date: Mon, 4 May 2009 11:32:57 +0200
- Subject: Re: Fortran whole file patch
- References: <20090502170421.E95243BAB9@mailhost.lps.ens.fr> <339c37f20905040224n829d065t23dc3ce657390382@mail.gmail.com>
> Dear All,
>
> Dominique has opened PR40011to help me track all the problems with the
> whole file scope patch. I wanted this to be a single PR, rather than
> a meta-bug, so that I could deal with all the problems as quickly as
> possible and then close it off.
>
> I have fixes for all the specific problems that Dominique lists.
>
> I know about the problem with module procedures and have a fix on it's way.
>
> Once this is done and tested, I'll submit a first patch.
>
> I'll then move to the polyhedron failures; once fixed, we should get
> an impression of how much difference in-lining will make to benchmark
> results.
Note that we should be able to do now also the IPA constant propagation
and optimize function calls to avoid passing by reference. I hope both
to cause noticeable improvements for fortran especially by making stuff
like loop bounds more explicit.
>
> I am aware of some problems with SPEC; could somebody post which parts
> ICE, please?
Thanks a lot for the work!
For SPEC I get the following from apsi, I didn't try the other
benchmarks.
/home/jh/apsi.f:574.72:
IF(NWINDX.EQ.1) PAUSE
1
Warning: Deleted feature: PAUSE statement at (1)
/home/jh/apsi.f:7178.3:
20 CONTINUE
1
/home/jh/apsi.f:7161.72:
IF(K.EQ.2) GO TO 20
2
Warning: Label at (1) is not in the same block as the GOTO statement at (2)
/home/jh/apsi.f:7178.3:
20 CONTINUE
1
/home/jh/apsi.f:7166.72:
IF(TOL.LE.0.01D0) GO TO 20
2
Warning: Label at (1) is not in the same block as the GOTO statement at (2)
/home/jh/apsi.f:7192.3:
40 CONTINUE
1
/home/jh/apsi.f:7187.72:
IF(DKS.GT.0.) GOTO 40
2
Warning: Label at (1) is not in the same block as the GOTO statement at (2)
/home/jh/apsi.f:4316.32:
CALL DINITU(ZET,TM,DKM,DZ,WORK(LWG),WORK(LWM),NZ,NP,UM,UG,VG,F)
1
Error: Type mismatch in argument 'wg' at (1); passed REAL(8) to COMPLEX(8)
/home/jh/apsi.f:4325.33:
CALL CRANKC(DKM,VM,HVAR,WORK(LWM),WORK(LWG),
1
Error: Type mismatch in argument 'wm' at (1); passed REAL(8) to COMPLEX(8)
/home/jh/apsi.f:4330.32:
CALL PADEC(DKM,VM,HVAR,WORK(LWM),WORK(LWG),
1
Error: Type mismatch in argument 'wm' at (1); passed REAL(8) to COMPLEX(8)
/home/jh/apsi.f:4338.20:
CALL DKZMN(TM,WORK(LWM),ZET,DKM,DZ,ZNOT,USTAR,DL,KLAS,ZMH,F,NZ)
1
Error: Type mismatch in argument 'wm' at (1); passed REAL(8) to COMPLEX(8)
/home/jh/apsi.f:4343.19:
CALL TEST(UM,WORK(LWM),NZ,DNORM,ITER,TOL,DT,TIMEM)
1
Error: Type mismatch in argument 'wm' at (1); passed REAL(8) to COMPLEX(8)
/home/jh/apsi.f:4445.22:
VM(K)=XDIMAG(WM(K))
1
Error: Type mismatch in argument 'qvar' at (1); passed COMPLEX(8) to REAL(8)
/home/jh/apsi.f:4346.9:
* WORK(LWM),ZET,DNORM,ITR,NY,DT,TIMEM)
1
Error: Type mismatch in argument 'wm' at (1); passed REAL(8) to COMPLEX(8)
f951: internal compiler error: Segmentation fault
Please submit a full bug report,
with preprocessed source if appropriate.
See <http://gcc.gnu.org/bugs.html> for instructions.