On Mon, 19 Jan 2009, Mark Taylor wrote:
> build attempt continues... the next hiccup is in SPECX. Lots of warnings
> ("Argument foo is one precision at (2) but some other precision at (1)" etc)
> which probably(?) can be ignored.
Yes, we all do.
> Then the rather surprising:
>
> /bin/sh ../libtool --mode=compile g77 -I/mbt/starauto/include
> -fno-second-underscore -I../inc -g -O2 -c -o foursp.lo foursp.f
> g77 -I/mbt/starauto/include -fno-second-underscore -I../inc -g -O2 -c
> foursp.f -fPIC -o .libs/foursp.o
> foursp.f: In subroutine `foursp':
> In file included from foursp.f:0:
> foursp.f:79: error: unrecognizable insn:
> (insn 441 440 416 2 foursp.f:46 (set (reg:V4SF 23 xmm2)
> (and:V4SF (not:V4SF (reg:SF 22 xmm1 [141]))
> (reg:V4SF 21 xmm0 [81]))) -1 (nil)
> (nil))
> foursp.f:79: internal compiler error: in extract_insn, at recog.c:2083
> Please submit a full bug report,
> with preprocessed source if appropriate.
> See <URL:http://bugzilla.redhat.com/bugzilla> for instructions.
> make[2]: *** [foursp.lo] Error 1
>
> Looks like FORTRAN's moved on since I last read the standard. Has
> Norman infiltrated ISO/IEC 1539-1?
I suspect something less scary, like an optimizer bug. Try:
make CFLAGS=-g
in applications/specx. Could be time to update those compilers to the
GCC4 era.
Peter.
|