public inbox for gcc-bugs@sourceware.org
help / color / mirror / Atom feed
* [Bug rtl-optimization/55549] New: zero_extend and vectors
@ 2012-11-30 14:29 glisse at gcc dot gnu.org
2021-08-22 8:59 ` [Bug rtl-optimization/55549] " pinskia at gcc dot gnu.org
2021-08-23 15:32 ` segher at gcc dot gnu.org
0 siblings, 2 replies; 3+ messages in thread
From: glisse at gcc dot gnu.org @ 2012-11-30 14:29 UTC (permalink / raw)
To: gcc-bugs
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=55549
Bug #: 55549
Summary: zero_extend and vectors
Classification: Unclassified
Product: gcc
Version: 4.8.0
Status: UNCONFIRMED
Severity: enhancement
Priority: P3
Component: rtl-optimization
AssignedTo: unassigned@gcc.gnu.org
ReportedBy: glisse@gcc.gnu.org
Hello,
the documentation of zero_extend says:
@findex zero_extend
@item (zero_extend:@var{m} @var{x})
Represents the result of zero-extending the value @var{x}
to machine mode @var{m}. @var{m} must be a fixed-point mode
and @var{x} a fixed-point value of a mode narrower than @var{m}.
However, try_combine will happily call:
SUBST (*split, gen_rtx_ZERO_EXTEND (split_mode,
SUBREG_REG (*split)));
without many checks. In particular, in PR54855, it tries:
(zero_extend:V2DF (mem/u/c:DF ...))
It isn't a bug because nothing will recognize this pattern and it will be
rejected, but it seems it would be good to add a few conditions.
^ permalink raw reply [flat|nested] 3+ messages in thread
* [Bug rtl-optimization/55549] zero_extend and vectors
2012-11-30 14:29 [Bug rtl-optimization/55549] New: zero_extend and vectors glisse at gcc dot gnu.org
@ 2021-08-22 8:59 ` pinskia at gcc dot gnu.org
2021-08-23 15:32 ` segher at gcc dot gnu.org
1 sibling, 0 replies; 3+ messages in thread
From: pinskia at gcc dot gnu.org @ 2021-08-22 8:59 UTC (permalink / raw)
To: gcc-bugs
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=55549
Andrew Pinski <pinskia at gcc dot gnu.org> changed:
What |Removed |Added
----------------------------------------------------------------------------
Ever confirmed|0 |1
Keywords| |internal-improvement
Last reconfirmed| |2021-08-22
Status|UNCONFIRMED |NEW
--- Comment #1 from Andrew Pinski <pinskia at gcc dot gnu.org> ---
Still there:
#ifdef INSN_SCHEDULING
/* If *SPLIT is a paradoxical SUBREG, when we split it, it should
be written as a ZERO_EXTEND. */
if (split_code == SUBREG && MEM_P (SUBREG_REG (*split)))
{
/* Or as a SIGN_EXTEND if LOAD_EXTEND_OP says that that's
what it really is. */
if (load_extend_op (GET_MODE (SUBREG_REG (*split)))
== SIGN_EXTEND)
SUBST (*split, gen_rtx_SIGN_EXTEND (split_mode,
SUBREG_REG (*split)));
else
SUBST (*split, gen_rtx_ZERO_EXTEND (split_mode,
SUBREG_REG (*split)));
}
#endif
^ permalink raw reply [flat|nested] 3+ messages in thread
* [Bug rtl-optimization/55549] zero_extend and vectors
2012-11-30 14:29 [Bug rtl-optimization/55549] New: zero_extend and vectors glisse at gcc dot gnu.org
2021-08-22 8:59 ` [Bug rtl-optimization/55549] " pinskia at gcc dot gnu.org
@ 2021-08-23 15:32 ` segher at gcc dot gnu.org
1 sibling, 0 replies; 3+ messages in thread
From: segher at gcc dot gnu.org @ 2021-08-23 15:32 UTC (permalink / raw)
To: gcc-bugs
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=55549
--- Comment #2 from Segher Boessenkool <segher at gcc dot gnu.org> ---
Three things:
1) It was invalid *before* this, already (vector mode paradoxical subregs
make no sense);
2) The subreg needs to be fixed up *somehow*, too;
3) subreg of mem will go away any day now (or, I keep telling myself that).
^ permalink raw reply [flat|nested] 3+ messages in thread
end of thread, other threads:[~2021-08-23 15:32 UTC | newest]
Thread overview: 3+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2012-11-30 14:29 [Bug rtl-optimization/55549] New: zero_extend and vectors glisse at gcc dot gnu.org
2021-08-22 8:59 ` [Bug rtl-optimization/55549] " pinskia at gcc dot gnu.org
2021-08-23 15:32 ` segher at gcc dot gnu.org
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for read-only IMAP folder(s) and NNTP newsgroup(s).