public inbox for gcc-patches@gcc.gnu.org
 help / color / mirror / Atom feed
From: Sam Feifer <sfeifer@redhat.com>
To: Richard Biener <richard.guenther@gmail.com>
Cc: Andrew Pinski <pinskia@gmail.com>, GCC Patches <gcc-patches@gcc.gnu.org>
Subject: Re: [PATCH] match.pd: Add new division pattern [PR104992]
Date: Mon, 1 Aug 2022 09:27:12 -0400	[thread overview]
Message-ID: <CAJKdq8R_gYTqZahXHqvJ_EXf5Z_=k5BEcTo2fLuee0G46mW9Bw@mail.gmail.com> (raw)
In-Reply-To: <CAFiYyc34EV_Yn7X7xXNrrfGEE4g+2uzxXmHL69OqjCHYYBak-g@mail.gmail.com>

[-- Attachment #1: Type: text/plain, Size: 5781 bytes --]

Here is the patch that I pushed. Thanks for the feedback.

Thanks
-Sam

On Thu, Jul 28, 2022 at 3:03 AM Richard Biener <richard.guenther@gmail.com>
wrote:

> On Wed, Jul 27, 2022 at 9:57 PM Sam Feifer <sfeifer@redhat.com> wrote:
> >
> >
> >> _Complex int are strange beasts, I'd simply avoid the transform for
> them.
> >>
> >
> > I added to the match.pd rule to not simplify if the operands are
> complex. There is now a test case for complex types to make sure they do
> not simplify. I had to move the "dg-do run" test to g++.dg to accommodate
> the complex type function that is included (even though there isn't a
> runtime test for complex types).
> >
> >> Can you please move the pattern next to the existing div/mod patterns,
> >> like after the related
> >>
> >
> > done :)
> >
> >> /* Simplify (A / B) * B + (A % B) -> A.  */
> >> (for div (trunc_div ceil_div floor_div round_div)
> >>      mod (trunc_mod ceil_mod floor_mod round_mod)
> >>   (simplify
> >>    (plus:c (mult:c (div @0 @1) @1) (mod @0 @1))
> >>    @0))
> >>
> >> pattern?
> >>
> >> +/* x / y * y == x -> x % y == 0.  */
> >> +(simplify
> >> +  (eq (mult (trunc_div @0 @1) @1) @0)
> >> +  (eq (trunc_mod @0 @1) { build_zero_cst TREE_TYPE(@0); }))
> >>
> >> there are parens missing around the TREE_TYPE (@0), how did you test
> >> the patch?  You probably want :s on the trunc_div and as Andrew said
> >> :c on the eq and the mult.
> >
> >
> > I made those changes to the rule. The rule worked without the
> parentheses, which is probably why I didn't notice they were missing.
> Attached is an updated patch file.
>
> +/* x / y * y == x -> x % y == 0.  */
> +(simplify
> +  (eq:c (mult:c (trunc_div:s @0 @1) @1) @0)
> +  (if (TREE_CODE (TREE_TYPE (@0)) != COMPLEX_TYPE
> +       && TREE_CODE (TREE_TYPE (@1)) != COMPLEX_TYPE)
>
> Testing this only for @0 is enough, we don't allow mixed types for
> trunc_div or mult.
>
> +    (eq (trunc_mod @0 @1) { build_zero_cst (TREE_TYPE(@0)); })))
>
> space before (@0) in 'TREE_TYPE(@0)'
>
> OK with those changes.
>
> Thanks,
> Richard.
>
>
>
> > Thanks
> > -Sam
> >
> >>
> >> Richard.
> >>
> >> > Thanks
> >> > -Sam
> >> >
> >> >
> >> > > For vector try (which works for both the C and C++ front-end):
> >> > > #define vector __attribute__((vector_size(4*sizeof(int)) ))
> >> > > vector int f(vector int x, vector int y)
> >> > > {
> >> > >   return x == x / y * y;
> >> > > }
> >> > >
> >> > > That is for the vector case, == still returns a vector type.
> >> > >
> >> > > Thanks,
> >> > > Andrew Pinski
> >> > >
> >> > > >
> >> > > > Thanks
> >> > > > -Sam
> >> > > >
> >> > > >> Thanks,
> >> > > >> Andrew Pinski
> >> > > >>
> >> > > >> > diff --git a/gcc/testsuite/gcc.dg/pr104992-1.c
> >> > > b/gcc/testsuite/gcc.dg/pr104992-1.c
> >> > > >> > new file mode 100644
> >> > > >> > index 00000000000..a80e5e180ce
> >> > > >> > --- /dev/null
> >> > > >> > +++ b/gcc/testsuite/gcc.dg/pr104992-1.c
> >> > > >> > @@ -0,0 +1,30 @@
> >> > > >> > +/* PR tree-optimization/104992 */
> >> > > >> > +/* { dg-do run } */
> >> > > >> > +/* { dg-options "-O2"} */
> >> > > >> > +
> >> > > >> > +#include "pr104992.c"
> >> > > >> > +
> >> > > >> > +int main () {
> >> > > >> > +
> >> > > >> > +    /* Should be true.  */
> >> > > >> > +    if (!foo(6, 3)
> >> > > >> > +        || !bar(12, 2)
> >> > > >> > +        || !baz(34, 17)
> >> > > >> > +        || !qux(50, 10)
> >> > > >> > +        || !fred(16, 8)
> >> > > >> > +        || !baz(-9, 3)
> >> > > >> > +        || !baz(9, -3)
> >> > > >> > +        || !baz(-9, -3)
> >> > > >> > +        ) {
> >> > > >> > +            __builtin_abort();
> >> > > >> > +         }
> >> > > >> > +
> >> > > >> > +    /* Should be false.  */
> >> > > >> > +    if (foo(5, 30)
> >> > > >> > +        || bar(72, 27)
> >> > > >> > +        || baz(42, 15)) {
> >> > > >> > +            __builtin_abort();
> >> > > >> > +        }
> >> > > >> > +
> >> > > >> > +    return 0;
> >> > > >> > +}
> >> > > >> > diff --git a/gcc/testsuite/gcc.dg/pr104992.c
> >> > > b/gcc/testsuite/gcc.dg/pr104992.c
> >> > > >> > new file mode 100644
> >> > > >> > index 00000000000..b4b0ca53118
> >> > > >> > --- /dev/null
> >> > > >> > +++ b/gcc/testsuite/gcc.dg/pr104992.c
> >> > > >> > @@ -0,0 +1,35 @@
> >> > > >> > +/* PR tree-optimization/104992 */
> >> > > >> > +/* { dg-do compile } */
> >> > > >> > +/* { dg-options "-O2 -fdump-tree-optimized" } */
> >> > > >> > +
> >> > > >> > +/* Form from PR.  */
> >> > > >> > +__attribute__((noipa)) unsigned foo(unsigned x, unsigned y)
> >> > > >> > +{
> >> > > >> > +    return x / y * y == x;
> >> > > >> > +}
> >> > > >> > +
> >> > > >> > +__attribute__((noipa)) unsigned bar(unsigned x, unsigned y) {
> >> > > >> > +    return x == x / y * y;
> >> > > >> > +}
> >> > > >> > +
> >> > > >> > +/* Signed test case.  */
> >> > > >> > +__attribute__((noipa)) unsigned baz (int x, int y) {
> >> > > >> > +    return x / y * y == x;
> >> > > >> > +}
> >> > > >> > +
> >> > > >> > +/* Changed order.  */
> >> > > >> > +__attribute__((noipa)) unsigned qux (unsigned x, unsigned y) {
> >> > > >> > +    return y * (x / y) == x;
> >> > > >> > +}
> >> > > >> > +
> >> > > >> > +/* Wrong order.  */
> >> > > >> > +__attribute__((noipa)) unsigned fred (unsigned x, unsigned y)
> {
> >> > > >> > +    return y * x / y == x;
> >> > > >> > +}
> >> > > >> > +
> >> > > >> > +/* Wrong pattern.  */
> >> > > >> > +__attribute__((noipa)) unsigned waldo (unsigned x, unsigned y,
> >> > > unsigned z) {
> >> > > >> > +    return x / y * z == x;
> >> > > >> > +}
> >> > > >> > +
> >> > > >> > +/* { dg-final {scan-tree-dump-times " % " 4 "optimized" } } */
> >> > > >> >
> >> > > >> > base-commit: 633e9920589ddfaf2d6da1c24ce99b18a2638db4
> >> > > >> > --
> >> > > >> > 2.31.1
> >> > > >> >
> >> > > >>
> >> > >
> >> > >
> >>
>
>

[-- Attachment #2: 0001-match.pd-Add-new-division-pattern-PR104992.patch --]
[-- Type: text/x-patch, Size: 4087 bytes --]

From fdabed7b0b37c02c8fb711e0d30938fae2a1ccd8 Mon Sep 17 00:00:00 2001
From: Sam Feifer <sfeifer@redhat.com>
Date: Fri, 29 Jul 2022 09:44:48 -0400
Subject: [PATCH] match.pd: Add new division pattern [PR104992]

This patch fixes a missed optimization in match.pd. It takes the pattern,
x / y * y == x, and optimizes it to x % y == 0. This produces fewer
instructions. This simplification does not happen for complex types.

This patch also adds tests for the optimization rule.

Bootstrapped/regtested on x86_64-pc-linux-gnu.

	PR tree-optimization/104992

gcc/ChangeLog:

	* match.pd (x / y * y == x): New simplification.

gcc/testsuite/ChangeLog:

	* g++.dg/pr104992-1.C: New test.
	* gcc.dg/pr104992.c: New test.
---
 gcc/match.pd                      |  6 ++++
 gcc/testsuite/g++.dg/pr104992-1.C | 30 ++++++++++++++++
 gcc/testsuite/gcc.dg/pr104992.c   | 57 +++++++++++++++++++++++++++++++
 3 files changed, 93 insertions(+)
 create mode 100644 gcc/testsuite/g++.dg/pr104992-1.C
 create mode 100644 gcc/testsuite/gcc.dg/pr104992.c

diff --git a/gcc/match.pd b/gcc/match.pd
index 330c1db0c8e..562138a8034 100644
--- a/gcc/match.pd
+++ b/gcc/match.pd
@@ -3982,6 +3982,12 @@ DEFINE_INT_AND_FLOAT_ROUND_FN (RINT)
    (plus:c (mult:c (div @0 @1) @1) (mod @0 @1))
    @0))
 
+/* x / y * y == x -> x % y == 0.  */
+(simplify
+  (eq:c (mult:c (trunc_div:s @0 @1) @1) @0)
+  (if (TREE_CODE (TREE_TYPE (@0)) != COMPLEX_TYPE)
+    (eq (trunc_mod @0 @1) { build_zero_cst (TREE_TYPE (@0)); })))
+
 /* ((X /[ex] A) +- B) * A  -->  X +- A * B.  */
 (for op (plus minus)
  (simplify
diff --git a/gcc/testsuite/g++.dg/pr104992-1.C b/gcc/testsuite/g++.dg/pr104992-1.C
new file mode 100644
index 00000000000..f5696b245ac
--- /dev/null
+++ b/gcc/testsuite/g++.dg/pr104992-1.C
@@ -0,0 +1,30 @@
+/* PR tree-optimization/104992 */
+/* { dg-do run } */
+/* { dg-options "-O2"} */
+
+#include "../gcc.dg/pr104992.c"
+
+int main () {
+
+    /* Should be true.  */
+    if (!foo(6, 3)
+        || !bar(12, 2)
+        || !baz(34, 17)
+        || !qux(50, 10)
+        || !fred(16, 8)
+        || !baz(-9, 3)
+        || !baz(9, -3)
+        || !baz(-9, -3)
+        ) {
+            __builtin_abort();
+         }
+    
+    /* Should be false.  */
+    if (foo(5, 30)
+        || bar(72, 27)
+        || baz(42, 15)) {
+            __builtin_abort();
+        }
+    
+    return 0;
+}
diff --git a/gcc/testsuite/gcc.dg/pr104992.c b/gcc/testsuite/gcc.dg/pr104992.c
new file mode 100644
index 00000000000..b9d91a13ad8
--- /dev/null
+++ b/gcc/testsuite/gcc.dg/pr104992.c
@@ -0,0 +1,57 @@
+/* PR tree-optimization/104992 */
+/* { dg-do compile } */
+/* { dg-options "-O2 -fdump-tree-optimized" } */
+
+#define vector __attribute__((vector_size(4*sizeof(int))))
+
+/* Form from PR.  */
+__attribute__((noipa)) unsigned foo(unsigned x, unsigned y)
+{
+    return x / y * y == x;
+}
+
+__attribute__((noipa)) unsigned bar(unsigned x, unsigned y) {
+    return x == x / y * y;
+}
+
+/* Signed test case.  */
+__attribute__((noipa)) unsigned baz (int x, int y) {
+    return x / y * y == x;
+}
+
+/* Changed order.  */
+__attribute__((noipa)) unsigned qux (unsigned x, unsigned y) {
+    return y * (x / y) == x;
+}
+
+/* Test for forward propogation.  */
+__attribute__((noipa)) unsigned corge(unsigned x, unsigned y) {
+    int z = x / y;
+    int q = z * y;
+    return q == x; 
+}
+
+/* Test vector case.  */
+__attribute__((noipa)) vector int thud(vector int x, vector int y) {
+    return x / y * y == x;
+}
+
+/* Complex type should not simplify because mod is different.  */
+__attribute__((noipa)) int goo(_Complex int x, _Complex int y)
+{
+    _Complex int z = x / y;
+    _Complex int q = z * y;
+    return q == x; 
+}
+
+/* Wrong order.  */
+__attribute__((noipa)) unsigned fred (unsigned x, unsigned y) {
+    return y * x / y == x;
+}
+
+/* Wrong pattern.  */
+__attribute__((noipa)) unsigned waldo (unsigned x, unsigned y, unsigned z) {
+    return x / y * z == x;
+}
+
+/* { dg-final {scan-tree-dump-times " % " 9 "optimized" } } */

base-commit: 6e0ca3fe88d8f98ba6b4009c9483e87afbcf4ee8
-- 
2.31.1


      reply	other threads:[~2022-08-01 13:27 UTC|newest]

Thread overview: 9+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2022-07-25 19:34 Sam Feifer
2022-07-25 19:49 ` Andrew Pinski
2022-07-25 20:59   ` Sam Feifer
2022-07-25 21:14     ` Andrew Pinski
2022-07-26 14:31       ` Sam Feifer
2022-07-27  8:42         ` Richard Biener
2022-07-27 19:57           ` Sam Feifer
2022-07-28  7:03             ` Richard Biener
2022-08-01 13:27               ` Sam Feifer [this message]

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to='CAJKdq8R_gYTqZahXHqvJ_EXf5Z_=k5BEcTo2fLuee0G46mW9Bw@mail.gmail.com' \
    --to=sfeifer@redhat.com \
    --cc=gcc-patches@gcc.gnu.org \
    --cc=pinskia@gmail.com \
    --cc=richard.guenther@gmail.com \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
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).