* [PATCH] rtlanal: Fix up replace_rtx [PR105333]
@ 2022-04-22 7:30 Jakub Jelinek
2022-04-22 11:22 ` Richard Biener
2022-04-22 16:54 ` Eric Botcazou
0 siblings, 2 replies; 3+ messages in thread
From: Jakub Jelinek @ 2022-04-22 7:30 UTC (permalink / raw)
To: Richard Biener, Jeff Law, Eric Botcazou; +Cc: gcc-patches
Hi!
The following testcase FAILs, because replace_rtx replaces a REG with
CONST_WIDE_INT inside of a SUBREG, which is an invalid transformation
because a SUBREG relies on SUBREG_REG having non-VOIDmode but
CONST_WIDE_INT has VOIDmode.
replace_rtx already has code to deal with it, but it was doing
it only for CONST_INTs. The following patch does it also for
VOIDmode CONST_DOUBLE or CONST_WIDE_INT.
Bootstrapped/regtested on x86_64-linux and i686-linux, ok for trunk?
2022-04-22 Jakub Jelinek <jakub@redhat.com>
PR rtl-optimization/105333
* rtlanal.cc (replace_rtx): Use simplify_subreg or
simplify_unary_operation if CONST_SCALAR_INT_P rather than just
CONST_INT_P.
* gcc.dg/pr105333.c: New test.
--- gcc/rtlanal.cc.jj 2022-02-08 20:08:03.912540713 +0100
+++ gcc/rtlanal.cc 2022-04-21 15:45:23.219769629 +0200
@@ -3390,7 +3390,7 @@ replace_rtx (rtx x, rtx from, rtx to, bo
{
rtx new_rtx = replace_rtx (SUBREG_REG (x), from, to, all_regs);
- if (CONST_INT_P (new_rtx))
+ if (CONST_SCALAR_INT_P (new_rtx))
{
x = simplify_subreg (GET_MODE (x), new_rtx,
GET_MODE (SUBREG_REG (x)),
@@ -3406,7 +3406,7 @@ replace_rtx (rtx x, rtx from, rtx to, bo
{
rtx new_rtx = replace_rtx (XEXP (x, 0), from, to, all_regs);
- if (CONST_INT_P (new_rtx))
+ if (CONST_SCALAR_INT_P (new_rtx))
{
x = simplify_unary_operation (ZERO_EXTEND, GET_MODE (x),
new_rtx, GET_MODE (XEXP (x, 0)));
--- gcc/testsuite/gcc.dg/pr105333.c.jj 2022-04-21 15:48:08.310468926 +0200
+++ gcc/testsuite/gcc.dg/pr105333.c 2022-04-21 15:47:49.578729973 +0200
@@ -0,0 +1,21 @@
+/* PR rtl-optimization/105333 */
+/* { dg-do compile { target int128 } } */
+/* { dg-options "-Og -fno-tree-coalesce-vars -fno-tree-fre" } */
+
+int g;
+short s;
+
+static inline unsigned short
+bar (short a, __int128 b)
+{
+ b ^= (unsigned long) -a;
+ __builtin_strncpy ((void *) &s, (void *) &a, 1);
+ b *= 14;
+ return b;
+}
+
+void
+foo (void)
+{
+ g *= (__int128) bar (1, 1);
+}
Jakub
^ permalink raw reply [flat|nested] 3+ messages in thread
* Re: [PATCH] rtlanal: Fix up replace_rtx [PR105333]
2022-04-22 7:30 [PATCH] rtlanal: Fix up replace_rtx [PR105333] Jakub Jelinek
@ 2022-04-22 11:22 ` Richard Biener
2022-04-22 16:54 ` Eric Botcazou
1 sibling, 0 replies; 3+ messages in thread
From: Richard Biener @ 2022-04-22 11:22 UTC (permalink / raw)
To: Jakub Jelinek; +Cc: Jeff Law, Eric Botcazou, gcc-patches
On Fri, 22 Apr 2022, Jakub Jelinek wrote:
> Hi!
>
> The following testcase FAILs, because replace_rtx replaces a REG with
> CONST_WIDE_INT inside of a SUBREG, which is an invalid transformation
> because a SUBREG relies on SUBREG_REG having non-VOIDmode but
> CONST_WIDE_INT has VOIDmode.
>
> replace_rtx already has code to deal with it, but it was doing
> it only for CONST_INTs. The following patch does it also for
> VOIDmode CONST_DOUBLE or CONST_WIDE_INT.
>
> Bootstrapped/regtested on x86_64-linux and i686-linux, ok for trunk?
OK.
Richard.
> 2022-04-22 Jakub Jelinek <jakub@redhat.com>
>
> PR rtl-optimization/105333
> * rtlanal.cc (replace_rtx): Use simplify_subreg or
> simplify_unary_operation if CONST_SCALAR_INT_P rather than just
> CONST_INT_P.
>
> * gcc.dg/pr105333.c: New test.
>
> --- gcc/rtlanal.cc.jj 2022-02-08 20:08:03.912540713 +0100
> +++ gcc/rtlanal.cc 2022-04-21 15:45:23.219769629 +0200
> @@ -3390,7 +3390,7 @@ replace_rtx (rtx x, rtx from, rtx to, bo
> {
> rtx new_rtx = replace_rtx (SUBREG_REG (x), from, to, all_regs);
>
> - if (CONST_INT_P (new_rtx))
> + if (CONST_SCALAR_INT_P (new_rtx))
> {
> x = simplify_subreg (GET_MODE (x), new_rtx,
> GET_MODE (SUBREG_REG (x)),
> @@ -3406,7 +3406,7 @@ replace_rtx (rtx x, rtx from, rtx to, bo
> {
> rtx new_rtx = replace_rtx (XEXP (x, 0), from, to, all_regs);
>
> - if (CONST_INT_P (new_rtx))
> + if (CONST_SCALAR_INT_P (new_rtx))
> {
> x = simplify_unary_operation (ZERO_EXTEND, GET_MODE (x),
> new_rtx, GET_MODE (XEXP (x, 0)));
> --- gcc/testsuite/gcc.dg/pr105333.c.jj 2022-04-21 15:48:08.310468926 +0200
> +++ gcc/testsuite/gcc.dg/pr105333.c 2022-04-21 15:47:49.578729973 +0200
> @@ -0,0 +1,21 @@
> +/* PR rtl-optimization/105333 */
> +/* { dg-do compile { target int128 } } */
> +/* { dg-options "-Og -fno-tree-coalesce-vars -fno-tree-fre" } */
> +
> +int g;
> +short s;
> +
> +static inline unsigned short
> +bar (short a, __int128 b)
> +{
> + b ^= (unsigned long) -a;
> + __builtin_strncpy ((void *) &s, (void *) &a, 1);
> + b *= 14;
> + return b;
> +}
> +
> +void
> +foo (void)
> +{
> + g *= (__int128) bar (1, 1);
> +}
>
> Jakub
>
>
--
Richard Biener <rguenther@suse.de>
SUSE Software Solutions Germany GmbH, Maxfeldstrasse 5, 90409 Nuernberg,
Germany; GF: Ivo Totev; HRB 36809 (AG Nuernberg)
^ permalink raw reply [flat|nested] 3+ messages in thread
* Re: [PATCH] rtlanal: Fix up replace_rtx [PR105333]
2022-04-22 7:30 [PATCH] rtlanal: Fix up replace_rtx [PR105333] Jakub Jelinek
2022-04-22 11:22 ` Richard Biener
@ 2022-04-22 16:54 ` Eric Botcazou
1 sibling, 0 replies; 3+ messages in thread
From: Eric Botcazou @ 2022-04-22 16:54 UTC (permalink / raw)
To: Jakub Jelinek; +Cc: Richard Biener, Jeff Law, gcc-patches
> 2022-04-22 Jakub Jelinek <jakub@redhat.com>
>
> PR rtl-optimization/105333
> * rtlanal.cc (replace_rtx): Use simplify_subreg or
> simplify_unary_operation if CONST_SCALAR_INT_P rather than just
> CONST_INT_P.
>
> * gcc.dg/pr105333.c: New test.
OK, thanks.
--
Eric Botcazou
^ permalink raw reply [flat|nested] 3+ messages in thread
end of thread, other threads:[~2022-04-22 16:54 UTC | newest]
Thread overview: 3+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2022-04-22 7:30 [PATCH] rtlanal: Fix up replace_rtx [PR105333] Jakub Jelinek
2022-04-22 11:22 ` Richard Biener
2022-04-22 16:54 ` Eric Botcazou
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).