https://gcc.gnu.org/bugzilla/show_bug.cgi?id=91635
Jim Wilson changed:
What|Removed |Added
Status|NEW |RESOLVED
Resolution|---
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=91635
--- Comment #26 from Kito Cheng ---
Author: kito
Date: Thu Sep 19 06:38:23 2019
New Revision: 275929
URL: https://gcc.gnu.org/viewcvs?rev=275929&root=gcc&view=rev
Log:
RISC-V: Fix bad insn splits with paradoxical subregs.
Shifting by more than
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=91635
--- Comment #25 from Jim Wilson ---
Author: wilson
Date: Thu Sep 5 20:32:55 2019
New Revision: 275444
URL: https://gcc.gnu.org/viewcvs?rev=275444&root=gcc&view=rev
Log:
RISC-V: Fix bad insn splits with paradoxical subregs.
Shifting by more tha
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=91635
--- Comment #24 from Segher Boessenkool ---
(clobber of a match_operand I mean, sigh).
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=91635
--- Comment #23 from Segher Boessenkool ---
If a splitter wants a new register during combine, it should do a match_scratch
for that. This is documented.
You do not normally get new registers during combine. combine cannot really
handle those.
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=91635
--- Comment #22 from Jim Wilson ---
> First of all, I think you need to use :DI instead of :GPR for the last
> define_split, as it shouldn't be iterated with.
Yes, sloppy copy paste. I will fix before committing.
> Second, doesn't this mean th
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=91635
Jakub Jelinek changed:
What|Removed |Added
CC||segher at gcc dot gnu.org
--- Comment #2
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=91635
--- Comment #20 from Jim Wilson ---
Created attachment 46830
--> https://gcc.gnu.org/bugzilla/attachment.cgi?id=46830&action=edit
proposed patch to fix paradoxical reg in splitter problem
I get better code size with this alternative patch. I
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=91635
--- Comment #19 from Jim Wilson ---
I did cross toolchain builds and checks for rv32i/newlib and rv64gc/glibc.
There were no regressions.
Since the splitters exist to reduce code size, I looked at the libc and
libstdc++ sizes with and without t
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=91635
--- Comment #18 from Jim Wilson ---
I was willing to write a patch, I just needed a day to catch up as I'm
hopelessly overloaded with work. But since Jakub wrote one, it looks right to
me, and I think we should use that. I'm not sure exactly wh
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=91635
--- Comment #17 from Kito Cheng ---
Hi Jakub:
Thanks your patch, I've run with gcc testsuite and no new fails, and I am
ruining more gcc testsuite regression with different arch/abi combination for
that.
I am amazing that seems like RISC-V is f
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=91635
--- Comment #16 from Jakub Jelinek ---
Created attachment 46809
--> https://gcc.gnu.org/bugzilla/attachment.cgi?id=46809&action=edit
gcc10-pr91635.patch
If you want full patch from me, here it is, but as I said I can't easily test
it (I don't
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=91635
--- Comment #15 from Jakub Jelinek ---
Then gen_rtx_REG (mode, REGNO (SUBREG_REG (x))) case I meant only in case some
splitter would be required for proper operation rather than just being an
optimization. If they are all an optimization, the !p
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=91635
--- Comment #14 from Zdenek Sojka ---
Thank you all for looking into this! I can't provide testcases for other
issues, if they exist at all, since I am hitting the first one too often. If
you fix only the first issue, there is a chance the other
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=91635
--- Comment #13 from Jim Wilson ---
I see 5 broken patterns which matches the list already given. The four
testcases are all triggering on the same splitter, which is the first
define_split, lshrsi3_zero_extend_3+1. The second define_split,
lsh
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=91635
--- Comment #12 from Kito Cheng ---
Hi Zdenek:
I can reproduce for all new 3 testcases, it seems like cause by same issue,
thanks you provide more testcase for that!
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=91635
--- Comment #11 from Zdenek Sojka ---
Created attachment 46803
--> https://gcc.gnu.org/bugzilla/attachment.cgi?id=46803&action=edit
another testcase
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=91635
--- Comment #10 from Zdenek Sojka ---
Created attachment 46802
--> https://gcc.gnu.org/bugzilla/attachment.cgi?id=46802&action=edit
another testcase
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=91635
--- Comment #9 from Zdenek Sojka ---
Created attachment 46801
--> https://gcc.gnu.org/bugzilla/attachment.cgi?id=46801&action=edit
another testcase
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=91635
--- Comment #8 from Jakub Jelinek ---
Certainly both register_operand predicate as well as the usual register
constraints don't rule out subregs even after reload.
can_create_pseudo_p () is false after reload of course.
In theory, what you could
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=91635
--- Comment #7 from Kito Cheng ---
Hi Jakub:
> that doesn't mean paradoxical subregs can't appear there, just it will be
> less likely.
Does it mean paradoxical subregs will appear during intermediate result even
after reload, so for such spli
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=91635
--- Comment #6 from Jakub Jelinek ---
Other problematic splitters are
zero_extendsidi2, zero_extendhi2, extend2,
dunno if I haven't missed something else. While they are all post-reload
splitters, that doesn't mean paradoxical subregs can't appe
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=91635
--- Comment #5 from Kito Cheng ---
Hi Zdenek:
Could you share more testcase? I've a patch is based on Jakub's one.
Thanks :)
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=91635
--- Comment #4 from Zdenek Sojka ---
(In reply to Jakub Jelinek from comment #3)
> Looks like a backend bug to me.
...
> fixes this, but I don't have a way to test riscv, plus the backend has
> several similar problems elsewhere.
I can provide s
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=91635
Jakub Jelinek changed:
What|Removed |Added
CC||andrew at sifive dot com,
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=91635
Jakub Jelinek changed:
What|Removed |Added
Status|UNCONFIRMED |NEW
Last reconfirmed|
26 matches
Mail list logo
