[PATCH] Jump bypassing and improved cprop (take 3)
Sat Jun 1 11:08:00 GMT 2002
On Sat, 1 Jun 2002, Roger Sayle wrote:
> As Hans-Peter Nilsson pointed out, I hadn't tested my "take 2"
> patch on a cc0 target even though the patch modified code paths
> that were guarded by #ifdef HAVE_cc0.
> I've now also sucessfully tested the patch on v850-sim-elf, with
> "make check-gcc" with no new regressions.
Great, thanks. I did this too, just finished for cris-elf,
v850-elf, mn10300-elf and mmix-knuth-mmixware (the last one not
a cc0 machine) for your previous two patches; this one being
equivalent to the "first" one sent on "Thu, 30 May 2002 18:27:54
> A pleasant surprise
> is that in addition to the two gcc.dg/uninit-A.c XFAILs that
> this fixes on all platforms, it also appears to cure the v850
> specific unexpected failure 20020402-3.c.
It unfortunately introduces a libstdc++-v3 failure for
26_numerics/complex_inserters_extractors.cc execution test for
cris-elf. Can't put it on you to investigate that one, though.
(Sigh.) Don't let this stop that patch; it might be a latent
bug elsewhere. It's just a call to take an extra look at the
correctness of the patch.
BTW, your previous patch, "Thu, 30 May 2002 22:31:45 -0600
(MDT)" introduced no regressions (and no improvements at all
either for that matter) on neither cris-elf, v850-elf,
mn10300-elf or mmix-knuth-mmixware.
> Is there a definitive list of cc0 targets that GCC currently
> supports (in addition to v850 and mn10300)? I seem to recall
> a discussion that it isn't always immediately obvious from the
> machine description.
If cc0 appears in the machine description, *outside* any
define_expand:s, then it's a cc0 machine. The generated
insn-config.h then says "#define HAVE_cc0 1".
More information about the Gcc-patches