* [committed] arc: Fail conditional move expand patterns
@ 2022-02-25 12:50 Claudiu Zissulescu
2022-02-25 13:29 ` Robin Dapp
0 siblings, 1 reply; 3+ messages in thread
From: Claudiu Zissulescu @ 2022-02-25 12:50 UTC (permalink / raw)
To: gcc-patches, fbedard
If the movcc comparison is not valid it triggers an assert in the
current implementation. This behavior is not needed as we can FAIL
the movcc expand pattern.
gcc/
* config/arc/arc.cc (gen_compare_reg): Return NULL_RTX if the
comparison is not valid.
* config/arc/arc.md (movsicc): Fail if comparison is not valid.
(movdicc): Likewise.
(movsfcc): Likewise.
(movdfcc): Likewise.
Signed-off-by: Claudiu Zissulescu <claziss@synopsys.com>
---
gcc/config/arc/arc.cc | 3 ++-
gcc/config/arc/arc.md | 25 ++++++++++++++++++++-----
2 files changed, 22 insertions(+), 6 deletions(-)
diff --git a/gcc/config/arc/arc.cc b/gcc/config/arc/arc.cc
index 8cc173519ab..c27ba99eb60 100644
--- a/gcc/config/arc/arc.cc
+++ b/gcc/config/arc/arc.cc
@@ -2256,7 +2256,8 @@ gen_compare_reg (rtx comparison, machine_mode omode)
cmode = GET_MODE (x);
if (cmode == VOIDmode)
cmode = GET_MODE (y);
- gcc_assert (cmode == SImode || cmode == SFmode || cmode == DFmode);
+ if (cmode != SImode && cmode != SFmode && cmode != DFmode)
+ return NULL_RTX;
if (cmode == SImode)
{
if (!register_operand (x, SImode))
diff --git a/gcc/config/arc/arc.md b/gcc/config/arc/arc.md
index ace3cb70424..39b358052c1 100644
--- a/gcc/config/arc/arc.md
+++ b/gcc/config/arc/arc.md
@@ -1618,8 +1618,11 @@ (define_expand "movsicc"
(match_operand:SI 2 "nonmemory_operand" "")
(match_operand:SI 3 "register_operand" "")))]
""
- "operands[1] = gen_compare_reg (operands[1], VOIDmode);")
-
+ "
+ operands[1] = gen_compare_reg (operands[1], VOIDmode);
+ if (operands[1] == NULL_RTX)
+ FAIL;
+ ")
(define_expand "movdicc"
[(set (match_operand:DI 0 "dest_reg_operand" "")
@@ -1627,7 +1630,11 @@ (define_expand "movdicc"
(match_operand:DI 2 "nonmemory_operand" "")
(match_operand:DI 3 "register_operand" "")))]
""
- "operands[1] = gen_compare_reg (operands[1], VOIDmode);")
+ "
+ operands[1] = gen_compare_reg (operands[1], VOIDmode);
+ if (operands[1] == NULL_RTX)
+ FAIL;
+ ")
(define_expand "movsfcc"
@@ -1636,7 +1643,11 @@ (define_expand "movsfcc"
(match_operand:SF 2 "nonmemory_operand" "")
(match_operand:SF 3 "register_operand" "")))]
""
- "operands[1] = gen_compare_reg (operands[1], VOIDmode);")
+ "
+ operands[1] = gen_compare_reg (operands[1], VOIDmode);
+ if (operands[1] == NULL_RTX)
+ FAIL;
+ ")
(define_expand "movdfcc"
[(set (match_operand:DF 0 "dest_reg_operand" "")
@@ -1644,7 +1655,11 @@ (define_expand "movdfcc"
(match_operand:DF 2 "nonmemory_operand" "")
(match_operand:DF 3 "register_operand" "")))]
""
- "operands[1] = gen_compare_reg (operands[1], VOIDmode);")
+ "
+ operands[1] = gen_compare_reg (operands[1], VOIDmode);
+ if (operands[1] == NULL_RTX)
+ FAIL;
+ ")
(define_insn "*movsicc_insn"
[(set (match_operand:SI 0 "dest_reg_operand" "=w,w")
--
2.35.1
^ permalink raw reply [flat|nested] 3+ messages in thread
* Re: [committed] arc: Fail conditional move expand patterns
2022-02-25 12:50 [committed] arc: Fail conditional move expand patterns Claudiu Zissulescu
@ 2022-02-25 13:29 ` Robin Dapp
2022-02-28 9:45 ` Claudiu Zissulescu Ianculescu
0 siblings, 1 reply; 3+ messages in thread
From: Robin Dapp @ 2022-02-25 13:29 UTC (permalink / raw)
To: Claudiu Zissulescu, gcc-patches, fbedard, Jeff Law
> If the movcc comparison is not valid it triggers an assert in the
> current implementation. This behavior is not needed as we can FAIL
> the movcc expand pattern.
In case of a MODE_CC comparison you can also just return it as described
here https://gcc.gnu.org/bugzilla/show_bug.cgi?id=104154
or here: https://gcc.gnu.org/pipermail/gcc-patches/2022-February/590639.html
If there already is a "CC comparison" the backend does not need to
create one and ifcvt can make use of this, creating better sequences.
Regards
Robin
^ permalink raw reply [flat|nested] 3+ messages in thread
* Re: [committed] arc: Fail conditional move expand patterns
2022-02-25 13:29 ` Robin Dapp
@ 2022-02-28 9:45 ` Claudiu Zissulescu Ianculescu
0 siblings, 0 replies; 3+ messages in thread
From: Claudiu Zissulescu Ianculescu @ 2022-02-28 9:45 UTC (permalink / raw)
To: Robin Dapp; +Cc: gcc-patches, Francois Bedard, Jeff Law
Hi Robin,
I don't know how I missed your arc related patch, I'll bootstrap and test
your patch asap.
Thanks,
Claudiu
On Fri, Feb 25, 2022 at 3:29 PM Robin Dapp <rdapp@linux.ibm.com> wrote:
> > If the movcc comparison is not valid it triggers an assert in the
> > current implementation. This behavior is not needed as we can FAIL
> > the movcc expand pattern.
>
> In case of a MODE_CC comparison you can also just return it as described
> here https://gcc.gnu.org/bugzilla/show_bug.cgi?id=104154
>
> or here:
> https://gcc.gnu.org/pipermail/gcc-patches/2022-February/590639.html
>
> If there already is a "CC comparison" the backend does not need to
> create one and ifcvt can make use of this, creating better sequences.
>
> Regards
> Robin
>
^ permalink raw reply [flat|nested] 3+ messages in thread
end of thread, other threads:[~2022-02-28 9:45 UTC | newest]
Thread overview: 3+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2022-02-25 12:50 [committed] arc: Fail conditional move expand patterns Claudiu Zissulescu
2022-02-25 13:29 ` Robin Dapp
2022-02-28 9:45 ` Claudiu Zissulescu Ianculescu
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).