target/8087: sparc-sun-solaris2.7 C testsuite failures in execute/20020720-1.c w/-m64 or on sparcv9/sparc64

Roger Sayle roger@eyesopen.com
Thu Oct 3 16:41:00 GMT 2002


On Thu, 3 Oct 2002, Richard Henderson wrote:
> I don't think this is combine's job.  It's job is to smush together
> machine instructions to make cisc-y things.  It does that job
> reasonably well.
>
> What you're looking for is another kind of CSE.

In that case, the reasonable thing to do is to mark 20020720-1.c as
XFAIL on sparc64-*-* and "sparc -m64".  Those processors that benefit
from combine's ability to smush instructions into an unconditional
jump are just lucky, and failing to do so isn't a bug or deficiency
of the target backend.

Perhaps, in the middle term, I'll come back to modifying the CSE pass
to attempt multiple independent substitutions of an insn.  That should
have the benefit of bringing "all the pieces together" and checking
the REG_EQUAL notes.

Could someone suggest a patch to gcc.c-torture/execute/20020720-1.x
that disables the test on sparc-*-* only when -m64 is specified?

Roger
--



More information about the Gcc-bugs mailing list