When simplifing BFR of an insert, require a mode precision integral type (PR108688)
Checks
Commit Message
The same problem as PR 88739 has crept in but
this time in match.pd when simplifying bit_field_ref of
an bit_insert. That is we are generating a BIT_FIELD_REF
of a non-mode-precision integral type.
OK? Bootstrapped and tested on x86_64-linux-gnu with no regressions.
PR tree-optimization/108688
gcc/ChangeLog:
* match.pd (bit_field_ref [bit_insert]):
Avoid generating BIT_FIELD_REFs of non-mode-precision
integral operands
gcc/testsuite/ChangeLog:
* gcc.c-torture/compile/pr108688-1.c: New test.
---
gcc/match.pd | 4 +++-
gcc/testsuite/gcc.c-torture/compile/pr108688-1.c | 15 +++++++++++++++
2 files changed, 18 insertions(+), 1 deletion(-)
create mode 100644 gcc/testsuite/gcc.c-torture/compile/pr108688-1.c
Comments
On Wed, Feb 8, 2023 at 10:15 PM Andrew Pinski via Gcc-patches
<gcc-patches@gcc.gnu.org> wrote:
>
> The same problem as PR 88739 has crept in but
> this time in match.pd when simplifying bit_field_ref of
> an bit_insert. That is we are generating a BIT_FIELD_REF
> of a non-mode-precision integral type.
>
> OK? Bootstrapped and tested on x86_64-linux-gnu with no regressions.
Hmm. Can we handle the case of the extraction exactly covering the
insertion separately then and simplify to plain @1?
Thanks,
Richard.
> PR tree-optimization/108688
>
> gcc/ChangeLog:
>
> * match.pd (bit_field_ref [bit_insert]):
> Avoid generating BIT_FIELD_REFs of non-mode-precision
> integral operands
>
> gcc/testsuite/ChangeLog:
>
> * gcc.c-torture/compile/pr108688-1.c: New test.
> ---
> gcc/match.pd | 4 +++-
> gcc/testsuite/gcc.c-torture/compile/pr108688-1.c | 15 +++++++++++++++
> 2 files changed, 18 insertions(+), 1 deletion(-)
> create mode 100644 gcc/testsuite/gcc.c-torture/compile/pr108688-1.c
>
> diff --git a/gcc/match.pd b/gcc/match.pd
> index c9e8bebede2..d8353072822 100644
> --- a/gcc/match.pd
> +++ b/gcc/match.pd
> @@ -7487,7 +7487,9 @@ DEFINE_INT_AND_FLOAT_ROUND_FN (RINT)
> isize = tree_to_uhwi (TYPE_SIZE (TREE_TYPE (@1)));
> }
> (switch
> - (if (wi::leu_p (wi::to_wide (@ipos), wi::to_wide (@rpos))
> + (if ((!INTEGRAL_TYPE_P (TREE_TYPE (@1))
> + || type_has_mode_precision_p (TREE_TYPE (@1)))
> + && wi::leu_p (wi::to_wide (@ipos), wi::to_wide (@rpos))
> && wi::leu_p (wi::to_wide (@rpos) + wi::to_wide (@rsize),
> wi::to_wide (@ipos) + isize))
> (BIT_FIELD_REF @1 @rsize { wide_int_to_tree (bitsizetype,
> diff --git a/gcc/testsuite/gcc.c-torture/compile/pr108688-1.c b/gcc/testsuite/gcc.c-torture/compile/pr108688-1.c
> new file mode 100644
> index 00000000000..43d782d62bd
> --- /dev/null
> +++ b/gcc/testsuite/gcc.c-torture/compile/pr108688-1.c
> @@ -0,0 +1,15 @@
> +
> +
> +union U { signed int d : 7; signed int e : 2; } u;
> +int a, b;
> +
> +void
> +foo (void)
> +{
> + for (int i = 0; i < 64; i++)
> + {
> + u.d = a;
> + u.e ^= b;
> + }
> +}
> +
> --
> 2.17.1
>
On Thu, Feb 09, 2023 at 09:16:17AM +0100, Richard Biener via Gcc-patches wrote:
> On Wed, Feb 8, 2023 at 10:15 PM Andrew Pinski via Gcc-patches
> <gcc-patches@gcc.gnu.org> wrote:
> >
> > The same problem as PR 88739 has crept in but
> > this time in match.pd when simplifying bit_field_ref of
> > an bit_insert. That is we are generating a BIT_FIELD_REF
> > of a non-mode-precision integral type.
> >
> > OK? Bootstrapped and tested on x86_64-linux-gnu with no regressions.
>
> Hmm. Can we handle the case of the extraction exactly covering the
> insertion separately then and simplify to plain @1?
I was suggesting that in the PR. Here it is as an incremental patch
on top of Andrew's patch. Ok for trunk if it passes bootstrap/regtest?
On the newly added testcase the ifcvt-folding difference without/with the
incremental patch is:
--- pr108688.c.171t.ifcvt_ 2023-02-09 10:47:30.169916845 +0100
+++ pr108688.c.171t.ifcvt 2023-02-09 10:48:44.942793453 +0100
@@ -25,6 +25,8 @@ Number of blocks in CFG: 11
Number of blocks to update: 5 ( 45%)
+Applying pattern match.pd:7487, gimple-match.cc:243200
+Applying pattern match.pd:3987, gimple-match.cc:75423
Matching expression match.pd:1677, gimple-match.cc:209
Applying pattern match.pd:1733, gimple-match.cc:109481
Matching expression match.pd:2393, gimple-match.cc:852
@@ -70,7 +72,6 @@ void foo ()
signed char _29;
<unnamed-signed:7> _30;
unsigned int ivtmp_33;
- <unnamed-signed:7> _ifc__35;
unsigned char _ifc__37;
unsigned char _ifc__38;
unsigned char _ifc__39;
@@ -91,8 +92,7 @@ void foo ()
_2 = (<unnamed-signed:7>) a.0_1;
_ifc__38 = u.D.2741;
_ifc__39 = BIT_INSERT_EXPR <_ifc__38, _2, 0 (7 bits)>;
- _ifc__35 = BIT_FIELD_REF <_ifc__39, 7, 0>;
- _4 = (signed char) _ifc__35;
+ _4 = (signed char) _2;
b.1_5 = b;
_6 = (signed char) b.1_5;
_7 = _4 ^ _6;
2023-02-09 Jakub Jelinek <jakub@redhat.com>
PR tree-optimization/108688
* match.pd (bit_field_ref [bit_insert]): Simplify BIT_FIELD_REF
of BIT_INSERT_EXPR extracting exactly all inserted bits even
when without mode precision. Formatting fixes.
* gcc.c-torture/compile/pr108688-1.c: Add PR number as comment.
* gcc.dg/pr108688.c: New test.
--- gcc/match.pd.jj 2023-02-09 09:42:53.782418805 +0100
+++ gcc/match.pd 2023-02-09 10:48:06.859365613 +0100
@@ -7475,13 +7475,16 @@ (define_operator_list SYNC_FETCH_AND_AND
}
(switch
(if ((!INTEGRAL_TYPE_P (TREE_TYPE (@1))
- || type_has_mode_precision_p (TREE_TYPE (@1)))
- && wi::leu_p (wi::to_wide (@ipos), wi::to_wide (@rpos))
+ || type_has_mode_precision_p (TREE_TYPE (@1)))
+ && wi::leu_p (wi::to_wide (@ipos), wi::to_wide (@rpos))
&& wi::leu_p (wi::to_wide (@rpos) + wi::to_wide (@rsize),
wi::to_wide (@ipos) + isize))
(BIT_FIELD_REF @1 @rsize { wide_int_to_tree (bitsizetype,
- wi::to_wide (@rpos)
+ wi::to_wide (@rpos)
- wi::to_wide (@ipos)); }))
+ (if (wi::eq_p (wi::to_wide (@ipos), wi::to_wide (@rpos))
+ && compare_tree_int (@rsize, isize) == 0)
+ (convert @1))
(if (wi::geu_p (wi::to_wide (@ipos),
wi::to_wide (@rpos) + wi::to_wide (@rsize))
|| wi::geu_p (wi::to_wide (@rpos),
--- gcc/testsuite/gcc.c-torture/compile/pr108688-1.c.jj 2023-02-09 09:42:53.782418805 +0100
+++ gcc/testsuite/gcc.c-torture/compile/pr108688-1.c 2023-02-09 10:34:18.339817796 +0100
@@ -1,4 +1,4 @@
-
+/* PR tree-optimization/108688 */
union U { signed int d : 7; signed int e : 2; } u;
int a, b;
@@ -12,4 +12,3 @@ foo (void)
u.e ^= b;
}
}
-
--- gcc/testsuite/gcc.dg/pr108688.c.jj 2023-02-09 10:50:27.468253110 +0100
+++ gcc/testsuite/gcc.dg/pr108688.c 2023-02-09 10:50:16.383419646 +0100
@@ -0,0 +1,16 @@
+/* PR tree-optimization/108688 */
+/* { dg-do compile } */
+/* { dg-options "-O2 -fno-tree-pre -fno-tree-fre -fno-tree-dominator-opts -fno-tree-loop-im -fno-code-hoisting" } */
+
+union U { signed int d : 7; signed int e : 2; } u;
+int a, b;
+
+void
+foo (void)
+{
+ for (int i = 0; i < 64; i++)
+ {
+ u.d = a;
+ u.d ^= b;
+ }
+}
Jakub
On Thu, Feb 9, 2023 at 10:57 AM Jakub Jelinek <jakub@redhat.com> wrote:
>
> On Thu, Feb 09, 2023 at 09:16:17AM +0100, Richard Biener via Gcc-patches wrote:
> > On Wed, Feb 8, 2023 at 10:15 PM Andrew Pinski via Gcc-patches
> > <gcc-patches@gcc.gnu.org> wrote:
> > >
> > > The same problem as PR 88739 has crept in but
> > > this time in match.pd when simplifying bit_field_ref of
> > > an bit_insert. That is we are generating a BIT_FIELD_REF
> > > of a non-mode-precision integral type.
> > >
> > > OK? Bootstrapped and tested on x86_64-linux-gnu with no regressions.
> >
> > Hmm. Can we handle the case of the extraction exactly covering the
> > insertion separately then and simplify to plain @1?
>
> I was suggesting that in the PR. Here it is as an incremental patch
> on top of Andrew's patch. Ok for trunk if it passes bootstrap/regtest?
Yes.
Thanks,
Richard.
> On the newly added testcase the ifcvt-folding difference without/with the
> incremental patch is:
> --- pr108688.c.171t.ifcvt_ 2023-02-09 10:47:30.169916845 +0100
> +++ pr108688.c.171t.ifcvt 2023-02-09 10:48:44.942793453 +0100
> @@ -25,6 +25,8 @@ Number of blocks in CFG: 11
> Number of blocks to update: 5 ( 45%)
>
>
> +Applying pattern match.pd:7487, gimple-match.cc:243200
> +Applying pattern match.pd:3987, gimple-match.cc:75423
> Matching expression match.pd:1677, gimple-match.cc:209
> Applying pattern match.pd:1733, gimple-match.cc:109481
> Matching expression match.pd:2393, gimple-match.cc:852
> @@ -70,7 +72,6 @@ void foo ()
> signed char _29;
> <unnamed-signed:7> _30;
> unsigned int ivtmp_33;
> - <unnamed-signed:7> _ifc__35;
> unsigned char _ifc__37;
> unsigned char _ifc__38;
> unsigned char _ifc__39;
> @@ -91,8 +92,7 @@ void foo ()
> _2 = (<unnamed-signed:7>) a.0_1;
> _ifc__38 = u.D.2741;
> _ifc__39 = BIT_INSERT_EXPR <_ifc__38, _2, 0 (7 bits)>;
> - _ifc__35 = BIT_FIELD_REF <_ifc__39, 7, 0>;
> - _4 = (signed char) _ifc__35;
> + _4 = (signed char) _2;
> b.1_5 = b;
> _6 = (signed char) b.1_5;
> _7 = _4 ^ _6;
>
> 2023-02-09 Jakub Jelinek <jakub@redhat.com>
>
> PR tree-optimization/108688
> * match.pd (bit_field_ref [bit_insert]): Simplify BIT_FIELD_REF
> of BIT_INSERT_EXPR extracting exactly all inserted bits even
> when without mode precision. Formatting fixes.
>
> * gcc.c-torture/compile/pr108688-1.c: Add PR number as comment.
> * gcc.dg/pr108688.c: New test.
>
> --- gcc/match.pd.jj 2023-02-09 09:42:53.782418805 +0100
> +++ gcc/match.pd 2023-02-09 10:48:06.859365613 +0100
> @@ -7475,13 +7475,16 @@ (define_operator_list SYNC_FETCH_AND_AND
> }
> (switch
> (if ((!INTEGRAL_TYPE_P (TREE_TYPE (@1))
> - || type_has_mode_precision_p (TREE_TYPE (@1)))
> - && wi::leu_p (wi::to_wide (@ipos), wi::to_wide (@rpos))
> + || type_has_mode_precision_p (TREE_TYPE (@1)))
> + && wi::leu_p (wi::to_wide (@ipos), wi::to_wide (@rpos))
> && wi::leu_p (wi::to_wide (@rpos) + wi::to_wide (@rsize),
> wi::to_wide (@ipos) + isize))
> (BIT_FIELD_REF @1 @rsize { wide_int_to_tree (bitsizetype,
> - wi::to_wide (@rpos)
> + wi::to_wide (@rpos)
> - wi::to_wide (@ipos)); }))
> + (if (wi::eq_p (wi::to_wide (@ipos), wi::to_wide (@rpos))
> + && compare_tree_int (@rsize, isize) == 0)
> + (convert @1))
> (if (wi::geu_p (wi::to_wide (@ipos),
> wi::to_wide (@rpos) + wi::to_wide (@rsize))
> || wi::geu_p (wi::to_wide (@rpos),
> --- gcc/testsuite/gcc.c-torture/compile/pr108688-1.c.jj 2023-02-09 09:42:53.782418805 +0100
> +++ gcc/testsuite/gcc.c-torture/compile/pr108688-1.c 2023-02-09 10:34:18.339817796 +0100
> @@ -1,4 +1,4 @@
> -
> +/* PR tree-optimization/108688 */
>
> union U { signed int d : 7; signed int e : 2; } u;
> int a, b;
> @@ -12,4 +12,3 @@ foo (void)
> u.e ^= b;
> }
> }
> -
> --- gcc/testsuite/gcc.dg/pr108688.c.jj 2023-02-09 10:50:27.468253110 +0100
> +++ gcc/testsuite/gcc.dg/pr108688.c 2023-02-09 10:50:16.383419646 +0100
> @@ -0,0 +1,16 @@
> +/* PR tree-optimization/108688 */
> +/* { dg-do compile } */
> +/* { dg-options "-O2 -fno-tree-pre -fno-tree-fre -fno-tree-dominator-opts -fno-tree-loop-im -fno-code-hoisting" } */
> +
> +union U { signed int d : 7; signed int e : 2; } u;
> +int a, b;
> +
> +void
> +foo (void)
> +{
> + for (int i = 0; i < 64; i++)
> + {
> + u.d = a;
> + u.d ^= b;
> + }
> +}
>
>
> Jakub
>
@@ -7487,7 +7487,9 @@ DEFINE_INT_AND_FLOAT_ROUND_FN (RINT)
isize = tree_to_uhwi (TYPE_SIZE (TREE_TYPE (@1)));
}
(switch
- (if (wi::leu_p (wi::to_wide (@ipos), wi::to_wide (@rpos))
+ (if ((!INTEGRAL_TYPE_P (TREE_TYPE (@1))
+ || type_has_mode_precision_p (TREE_TYPE (@1)))
+ && wi::leu_p (wi::to_wide (@ipos), wi::to_wide (@rpos))
&& wi::leu_p (wi::to_wide (@rpos) + wi::to_wide (@rsize),
wi::to_wide (@ipos) + isize))
(BIT_FIELD_REF @1 @rsize { wide_int_to_tree (bitsizetype,
new file mode 100644
@@ -0,0 +1,15 @@
+
+
+union U { signed int d : 7; signed int e : 2; } u;
+int a, b;
+
+void
+foo (void)
+{
+ for (int i = 0; i < 64; i++)
+ {
+ u.d = a;
+ u.e ^= b;
+ }
+}
+