http://llvm.org/bugs/show_bug.cgi?id=13040
Bug #: 13040
Summary: -fplugin-arg-dragonegg-llvm-option=-vectorize produces
bad bitcode for some SPEC benchmarks
Product: dragonegg
Version: trunk
Platform: PC
OS/Version: Linux
Status: NEW
Keywords: compile-fail
Severity: enhancement
Priority: P
Component: New Bugs
AssignedTo: baldrick at free.fr
ReportedBy: scott+llvm+bugzilla at pakin.org
CC: llvmbugs at cs.uiuc.edu
Classification: Unclassified
I think I've seen this on other SPEC benchmarks as well, but I've recently been
working with tonto:
$ gfortran -g -fplugin=/usr/local/lib/dragonegg.so
-fplugin-arg-dragonegg-emit-ir -O2 -fno-strict-aliasing
-fplugin-arg-dragonegg-llvm-option=-vectorize -S textfile.fppized.f90 -o
textfile.fppized.ll
$ llvm-as textfile.fppized.ll
llvm-as: assembly parsed, but does not verify as correct!
Range types must match load type!
%4 = load <2 x i32>* %3, align 4, !dbg !1765, !range !1766
Broken module found, compilation terminated.
Range types must match load type!
%20 = load <2 x i32>* %19, align 4, !dbg !1779, !range !1780
Broken module found, compilation terminated.
...
Take out the -vectorize and all is well:
$ gfortran -g -fplugin=/usr/local/lib/dragonegg.so
-fplugin-arg-dragonegg-emit-ir -O2 -fno-strict-aliasing -S textfile.fppized.f90
-o textfile.fppized.ll
$ llvm-as textfile.fppized.ll
This is with gfortran 4.5.2 and LLVM/DragonEgg revision 158077.
-- Scott
--
Configure bugmail: http://llvm.org/bugs/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are on the CC list for the bug.