https://gcc.gnu.org/bugzilla/show_bug.cgi?id=93934
--- Comment #25 from Uroš Bizjak ---
(In reply to jos...@codesourcery.com from comment #24)
> This is a fundamentally different test, because it involves (in the
> abstract machine) lvalue-to-rvalue conversion of a sNaN representation.
> That
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=93934
--- Comment #24 from joseph at codesourcery dot com ---
On Thu, 14 Oct 2021, ubizjak at gmail dot com via Gcc-bugs wrote:
> The situation is hopeless from the beginning. Please consider this testcase:
>
> --cut here--
> #include
> #include
>
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=93934
--- Comment #23 from Zoltan Vajda ---
(In reply to Uroš Bizjak from comment #20)
> (In reply to jos...@codesourcery.com from comment #16)
> > I don't think this bug is anything to do with -fsignaling-nans, for the
> > same reason as applies to
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=93934
--- Comment #22 from rguenther at suse dot de ---
On Thu, 14 Oct 2021, ubizjak at gmail dot com wrote:
> https://gcc.gnu.org/bugzilla/show_bug.cgi?id=93934
>
> --- Comment #20 from Uroš Bizjak ---
> (In reply to jos...@codesourcery.com from
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=93934
--- Comment #21 from Uroš Bizjak ---
(In reply to Zoltan Vajda from comment #19)
> The problem does not only apply for conditional moves! I can turn on sse,
> for example.
>
> https://gcc.godbolt.org/z/jP3Kne8T5
>
> Then the problematic code
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=93934
--- Comment #20 from Uroš Bizjak ---
(In reply to jos...@codesourcery.com from comment #16)
> I don't think this bug is anything to do with -fsignaling-nans, for the
> same reason as applies to bug 58416 and bug 71460.
The situation is
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=93934
--- Comment #19 from Zoltan Vajda ---
(In reply to Uroš Bizjak from comment #18)
> The following patch fixes the PR, see the comment inline:
>
> --cut here--
> diff --git a/gcc/config/i386/i386-expand.c b/gcc/config/i386/i386-expand.c
> index
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=93934
--- Comment #18 from Uroš Bizjak ---
The following patch fixes the PR, see the comment inline:
--cut here--
diff --git a/gcc/config/i386/i386-expand.c b/gcc/config/i386/i386-expand.c
index 6e2b7920d2b..b87490fe544 100644
---
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=93934
--- Comment #17 from Richard Biener ---
It might help to provide a option or tunable to disable the use of fcmov which
should reduce the attack surface a bit. I don't see any way to avoid using
fld for loading x87 float or double values (well,
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=93934
--- Comment #16 from joseph at codesourcery dot com ---
I don't think this bug is anything to do with -fsignaling-nans, for the
same reason as applies to bug 58416 and bug 71460.
The option -fsignaling-nans is only about correctly handling
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=93934
--- Comment #15 from Zoltan Vajda ---
In my special case, I have an embedded realtime application with a lot of FP
atithmetic on Intel 32 bit architecture (huge and complex legacy codebase). FPU
exceptions are enabled, so loading an SNaN results
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=93934
--- Comment #14 from Alexander Monakov ---
Zoltan, excuse me, could you please clarify what specifically you are worried
about? Your bug title says "results in UB" and the opening comment said "the
behavior [..] is unpredictable", but as far as
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=93934
--- Comment #13 from Uroš Bizjak ---
(In reply to Zoltan Vajda from comment #12)
> Using -mfpmath=sse here does not help on a 32 bit platfrom.
> https://gcc.godbolt.org/z/hs1Ef6aj4
> At line 31 the assembly code performs the speculative load.
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=93934
--- Comment #12 from Zoltan Vajda ---
Using -mfpmath=sse here does not help on a 32 bit platfrom.
https://gcc.godbolt.org/z/hs1Ef6aj4
At line 31 the assembly code performs the speculative load.
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=93934
--- Comment #11 from Uroš Bizjak ---
(In reply to Zoltan Vajda from comment #9)
> As I understand it, it is acknowledged, that this is a bug. However, the
> issue is in state NEW for a quite long time. The issue is still present in
> GCC 11.2.
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=93934
Richard Biener changed:
What|Removed |Added
CC||uros at gcc dot gnu.org
--- Comment
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=93934
--- Comment #9 from Zoltan Vajda ---
As I understand it, it is acknowledged, that this is a bug. However, the issue
is in state NEW for a quite long time. The issue is still present in GCC 11.2.
Do you see any chances for some progress in this
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=93934
--- Comment #8 from Alexander Monakov ---
I think regstack is fine as x87 only supports computations in its native 80-bit
format and conversions to/from ieee float/double happen only on memory
loads/stores.
> I suppose a fldt followed by
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=93934
--- Comment #7 from Richard Biener ---
Or the issue is in regstack handling all stack vars as 80bit and thus
introducing the conversion step in the first place.
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=93934
Richard Biener changed:
What|Removed |Added
Status|UNCONFIRMED |NEW
Last reconfirmed|
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=93934
--- Comment #5 from Alexander Monakov ---
Ah, indeed. fld won't raise FE_INVALID for 80-bit long double, but here
'result' is stored on the stack in 64-bit format.
So: fcmov and 80-bit fldt don't trap, 32-bit flds and 64-bit fldl do.
Somehow
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=93934
Richard Biener changed:
What|Removed |Added
Keywords|missed-optimization |wrong-code
--- Comment #4 from Richard
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=93934
Zoltan Vajda changed:
What|Removed |Added
Status|RESOLVED|UNCONFIRMED
Resolution|INVALID
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=93934
Alexander Monakov changed:
What|Removed |Added
Status|UNCONFIRMED |RESOLVED
CC|
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=93934
Andrew Pinski changed:
What|Removed |Added
Keywords||missed-optimization
Target|
25 matches
Mail list logo
