* Add DSE to early passes
@ 2015-04-13 10:01 Jan Hubicka
2015-04-13 12:12 ` Richard Biener
2015-04-22 12:13 ` H.J. Lu
0 siblings, 2 replies; 5+ messages in thread
From: Jan Hubicka @ 2015-04-13 10:01 UTC (permalink / raw)
To: rguenther, gcc-patches
Hi,
this patch adds DSE to early optimizations and handles testuiste fallout.
As discussed in PR 65076 this reduces number of CLOBBER statements in tramp3d
to 50% (and those accounts 29% of all code previously). The pass also quite
often kills real stores reducing Firefox binary by over 2% due to
better inline decisions.
Bootstrapped/regtested ppc64-linux, OK?
Honza
* passes.def (early_optimizations): Add pass_dse.
* g++.dg/tree-ssa/pr61034.C: Update template.
* g++.dg/warn/Warray-bounds.C: Harden for DSE.
* gcc.dg/Warray-bounds-11.c: Likewise.
* gcc.dg/Warray-bounds.c: Likewise.
Index: passes.def
===================================================================
--- passes.def (revision 222016)
+++ passes.def (working copy)
@@ -89,6 +89,7 @@ along with GCC; see the file COPYING3.
NEXT_PASS (pass_build_ealias);
NEXT_PASS (pass_fre);
NEXT_PASS (pass_merge_phi);
+ NEXT_PASS (pass_dse);
NEXT_PASS (pass_cd_dce);
NEXT_PASS (pass_early_ipa_sra);
NEXT_PASS (pass_tail_recursion);
Index: testsuite/g++.dg/tree-ssa/pr61034.C
===================================================================
--- testsuite/g++.dg/tree-ssa/pr61034.C (revision 222016)
+++ testsuite/g++.dg/tree-ssa/pr61034.C (working copy)
@@ -42,6 +42,6 @@ bool f(I a, I b, I c, I d) {
// This works only if everything is inlined into 'f'.
// { dg-final { scan-tree-dump-times ";; Function" 1 "fre2" } }
-// { dg-final { scan-tree-dump-times "free" 19 "fre2" } }
+// { dg-final { scan-tree-dump-times "free" 18 "fre2" } }
// { dg-final { scan-tree-dump-times "unreachable" 11 "fre2" } }
// { dg-final { cleanup-tree-dump "fre2" } }
Index: testsuite/g++.dg/warn/Warray-bounds.C
===================================================================
--- testsuite/g++.dg/warn/Warray-bounds.C (revision 222016)
+++ testsuite/g++.dg/warn/Warray-bounds.C (working copy)
@@ -11,6 +11,7 @@ static inline int n(void) {
void g(int *p);
void h(int p);
+void bar (void *);
int* f(void) {
int b[10];
@@ -27,6 +28,7 @@ int* f(void) {
a[ 9] = 0;
a[10] = 0; /* { dg-warning "array subscript" } */
a[11] = 0; /* { dg-warning "array subscript" } */
+ bar (a);
a[2 * n() - 11] = 1; /* { dg-warning "array subscript" } */
a[2 * n() - 10] = 1;
a[2 * n() - 1] = 1;
@@ -38,6 +40,7 @@ int* f(void) {
b[ 9] = 0;
b[10] = 0; /* { dg-warning "array subscript" } */
b[11] = 0; /* { dg-warning "array subscript" } */
+ bar (b);
b[2 * n() - 11] = 1; /* { dg-warning "array subscript" } */
b[2 * n() - 10] = 1;
b[2 * n() - 1] = 1;
@@ -49,6 +52,7 @@ int* f(void) {
c.c[ 9] = 0;
c.c[10] = 0; /* { dg-warning "array subscript" } */
c.c[11] = 0; /* { dg-warning "array subscript" } */
+ bar (&c);
c.c[2 * n() - 11] = 1; /* { dg-warning "array subscript" } */
c.c[2 * n() - 10] = 1;
c.c[2 * n() - 1] = 1;
@@ -87,6 +91,8 @@ int* f(void) {
if (-1 >= 0)
c.c[-1] = 0;
+ bar (b);
+ bar (&c);
return a;
}
Index: testsuite/gcc.dg/Warray-bounds-11.c
===================================================================
--- testsuite/gcc.dg/Warray-bounds-11.c (revision 222016)
+++ testsuite/gcc.dg/Warray-bounds-11.c (working copy)
@@ -92,5 +92,11 @@ void foo(int (*a)[3])
bar(c);
bar(e);
bar(f.f);
+ bar(h1->j);
+ bar(h3->j);
+ bar(h3b->j);
+ bar(h1b->j);
+ bar(h->j);
+ bar(h0->j);
}
Index: testsuite/gcc.dg/Warray-bounds.c
===================================================================
--- testsuite/gcc.dg/Warray-bounds.c (revision 222016)
+++ testsuite/gcc.dg/Warray-bounds.c (working copy)
@@ -11,6 +11,8 @@ static inline int n(void) {
void g(int *p);
void h(int p);
+void bar (void *);
+
int* f(void) {
int b[10];
int i;
@@ -26,10 +28,12 @@ int* f(void) {
a[ 9] = 0;
a[10] = 0; /* { dg-warning "6:array subscript" } */
a[11] = 0; /* { dg-warning "6:array subscript" } */
+ bar (a);
a[2 * n() - 11] = 1; /* { dg-warning "6:array subscript" } */
a[2 * n() - 10] = 1;
a[2 * n() - 1] = 1;
a[2 * n() - 0] = 1; /* { dg-warning "6:array subscript" } */
+ bar (a);
b[-1] = 0; /* { dg-warning "6:array subscript" } */
b[ 0] = 0;
@@ -37,6 +41,7 @@ int* f(void) {
b[ 9] = 0;
b[10] = 0; /* { dg-warning "6:array subscript" } */
b[11] = 0; /* { dg-warning "6:array subscript" } */
+ bar (b);
b[2 * n() - 11] = 1; /* { dg-warning "6:array subscript" } */
b[2 * n() - 10] = 1;
b[2 * n() - 1] = 1;
@@ -48,6 +53,7 @@ int* f(void) {
c.c[ 9] = 0;
c.c[10] = 0; /* { dg-warning "8:array subscript" } */
c.c[11] = 0; /* { dg-warning "8:array subscript" } */
+ bar (&c);
c.c[2 * n() - 11] = 1; /* { dg-warning "8:array subscript" } */
c.c[2 * n() - 10] = 1;
c.c[2 * n() - 1] = 1;
@@ -88,6 +94,8 @@ int* f(void) {
for (i = 20; i < 30; ++i)
a[i] = 1; /* { dg-warning "15:array subscript" } */
+ bar (b);
+ bar (&c);
return a;
}
^ permalink raw reply [flat|nested] 5+ messages in thread
* Re: Add DSE to early passes
2015-04-13 10:01 Add DSE to early passes Jan Hubicka
@ 2015-04-13 12:12 ` Richard Biener
2015-04-22 12:13 ` H.J. Lu
1 sibling, 0 replies; 5+ messages in thread
From: Richard Biener @ 2015-04-13 12:12 UTC (permalink / raw)
To: Jan Hubicka; +Cc: gcc-patches
On Mon, 13 Apr 2015, Jan Hubicka wrote:
> Hi,
> this patch adds DSE to early optimizations and handles testuiste fallout.
> As discussed in PR 65076 this reduces number of CLOBBER statements in tramp3d
> to 50% (and those accounts 29% of all code previously). The pass also quite
> often kills real stores reducing Firefox binary by over 2% due to
> better inline decisions.
>
> Bootstrapped/regtested ppc64-linux, OK?
Ok, but please wait a bit with committing to maintain trunk as a
test-base for GCC 5 patches during the RC phase.
Thanks,
Richard.
> Honza
>
> * passes.def (early_optimizations): Add pass_dse.
>
> * g++.dg/tree-ssa/pr61034.C: Update template.
> * g++.dg/warn/Warray-bounds.C: Harden for DSE.
> * gcc.dg/Warray-bounds-11.c: Likewise.
> * gcc.dg/Warray-bounds.c: Likewise.
> Index: passes.def
> ===================================================================
> --- passes.def (revision 222016)
> +++ passes.def (working copy)
> @@ -89,6 +89,7 @@ along with GCC; see the file COPYING3.
> NEXT_PASS (pass_build_ealias);
> NEXT_PASS (pass_fre);
> NEXT_PASS (pass_merge_phi);
> + NEXT_PASS (pass_dse);
> NEXT_PASS (pass_cd_dce);
> NEXT_PASS (pass_early_ipa_sra);
> NEXT_PASS (pass_tail_recursion);
> Index: testsuite/g++.dg/tree-ssa/pr61034.C
> ===================================================================
> --- testsuite/g++.dg/tree-ssa/pr61034.C (revision 222016)
> +++ testsuite/g++.dg/tree-ssa/pr61034.C (working copy)
> @@ -42,6 +42,6 @@ bool f(I a, I b, I c, I d) {
> // This works only if everything is inlined into 'f'.
>
> // { dg-final { scan-tree-dump-times ";; Function" 1 "fre2" } }
> -// { dg-final { scan-tree-dump-times "free" 19 "fre2" } }
> +// { dg-final { scan-tree-dump-times "free" 18 "fre2" } }
> // { dg-final { scan-tree-dump-times "unreachable" 11 "fre2" } }
> // { dg-final { cleanup-tree-dump "fre2" } }
> Index: testsuite/g++.dg/warn/Warray-bounds.C
> ===================================================================
> --- testsuite/g++.dg/warn/Warray-bounds.C (revision 222016)
> +++ testsuite/g++.dg/warn/Warray-bounds.C (working copy)
> @@ -11,6 +11,7 @@ static inline int n(void) {
>
> void g(int *p);
> void h(int p);
> +void bar (void *);
>
> int* f(void) {
> int b[10];
> @@ -27,6 +28,7 @@ int* f(void) {
> a[ 9] = 0;
> a[10] = 0; /* { dg-warning "array subscript" } */
> a[11] = 0; /* { dg-warning "array subscript" } */
> + bar (a);
> a[2 * n() - 11] = 1; /* { dg-warning "array subscript" } */
> a[2 * n() - 10] = 1;
> a[2 * n() - 1] = 1;
> @@ -38,6 +40,7 @@ int* f(void) {
> b[ 9] = 0;
> b[10] = 0; /* { dg-warning "array subscript" } */
> b[11] = 0; /* { dg-warning "array subscript" } */
> + bar (b);
> b[2 * n() - 11] = 1; /* { dg-warning "array subscript" } */
> b[2 * n() - 10] = 1;
> b[2 * n() - 1] = 1;
> @@ -49,6 +52,7 @@ int* f(void) {
> c.c[ 9] = 0;
> c.c[10] = 0; /* { dg-warning "array subscript" } */
> c.c[11] = 0; /* { dg-warning "array subscript" } */
> + bar (&c);
> c.c[2 * n() - 11] = 1; /* { dg-warning "array subscript" } */
> c.c[2 * n() - 10] = 1;
> c.c[2 * n() - 1] = 1;
> @@ -87,6 +91,8 @@ int* f(void) {
> if (-1 >= 0)
> c.c[-1] = 0;
>
> + bar (b);
> + bar (&c);
> return a;
> }
>
> Index: testsuite/gcc.dg/Warray-bounds-11.c
> ===================================================================
> --- testsuite/gcc.dg/Warray-bounds-11.c (revision 222016)
> +++ testsuite/gcc.dg/Warray-bounds-11.c (working copy)
> @@ -92,5 +92,11 @@ void foo(int (*a)[3])
> bar(c);
> bar(e);
> bar(f.f);
> + bar(h1->j);
> + bar(h3->j);
> + bar(h3b->j);
> + bar(h1b->j);
> + bar(h->j);
> + bar(h0->j);
> }
>
> Index: testsuite/gcc.dg/Warray-bounds.c
> ===================================================================
> --- testsuite/gcc.dg/Warray-bounds.c (revision 222016)
> +++ testsuite/gcc.dg/Warray-bounds.c (working copy)
> @@ -11,6 +11,8 @@ static inline int n(void) {
> void g(int *p);
> void h(int p);
>
> +void bar (void *);
> +
> int* f(void) {
> int b[10];
> int i;
> @@ -26,10 +28,12 @@ int* f(void) {
> a[ 9] = 0;
> a[10] = 0; /* { dg-warning "6:array subscript" } */
> a[11] = 0; /* { dg-warning "6:array subscript" } */
> + bar (a);
> a[2 * n() - 11] = 1; /* { dg-warning "6:array subscript" } */
> a[2 * n() - 10] = 1;
> a[2 * n() - 1] = 1;
> a[2 * n() - 0] = 1; /* { dg-warning "6:array subscript" } */
> + bar (a);
>
> b[-1] = 0; /* { dg-warning "6:array subscript" } */
> b[ 0] = 0;
> @@ -37,6 +41,7 @@ int* f(void) {
> b[ 9] = 0;
> b[10] = 0; /* { dg-warning "6:array subscript" } */
> b[11] = 0; /* { dg-warning "6:array subscript" } */
> + bar (b);
> b[2 * n() - 11] = 1; /* { dg-warning "6:array subscript" } */
> b[2 * n() - 10] = 1;
> b[2 * n() - 1] = 1;
> @@ -48,6 +53,7 @@ int* f(void) {
> c.c[ 9] = 0;
> c.c[10] = 0; /* { dg-warning "8:array subscript" } */
> c.c[11] = 0; /* { dg-warning "8:array subscript" } */
> + bar (&c);
> c.c[2 * n() - 11] = 1; /* { dg-warning "8:array subscript" } */
> c.c[2 * n() - 10] = 1;
> c.c[2 * n() - 1] = 1;
> @@ -88,6 +94,8 @@ int* f(void) {
> for (i = 20; i < 30; ++i)
> a[i] = 1; /* { dg-warning "15:array subscript" } */
>
> + bar (b);
> + bar (&c);
> return a;
> }
>
>
>
--
Richard Biener <rguenther@suse.de>
SUSE LINUX GmbH, GF: Felix Imendoerffer, Jane Smithard, Jennifer Guild,
Dilip Upmanyu, Graham Norton HRB 21284 (AG Nuernberg)
^ permalink raw reply [flat|nested] 5+ messages in thread
* Re: Add DSE to early passes
2015-04-13 10:01 Add DSE to early passes Jan Hubicka
2015-04-13 12:12 ` Richard Biener
@ 2015-04-22 12:13 ` H.J. Lu
2015-04-22 12:17 ` Jan Hubicka
1 sibling, 1 reply; 5+ messages in thread
From: H.J. Lu @ 2015-04-22 12:13 UTC (permalink / raw)
To: Jan Hubicka; +Cc: Richard Guenther, GCC Patches
On Mon, Apr 13, 2015 at 3:01 AM, Jan Hubicka <hubicka@ucw.cz> wrote:
> Hi,
> this patch adds DSE to early optimizations and handles testuiste fallout.
> As discussed in PR 65076 this reduces number of CLOBBER statements in tramp3d
> to 50% (and those accounts 29% of all code previously). The pass also quite
> often kills real stores reducing Firefox binary by over 2% due to
> better inline decisions.
>
> Bootstrapped/regtested ppc64-linux, OK?
>
> Honza
>
> * passes.def (early_optimizations): Add pass_dse.
>
> * g++.dg/tree-ssa/pr61034.C: Update template.
> * g++.dg/warn/Warray-bounds.C: Harden for DSE.
> * gcc.dg/Warray-bounds-11.c: Likewise.
> * gcc.dg/Warray-bounds.c: Likewise.
> Index: passes.def
> ===================================================================
> --- passes.def (revision 222016)
> +++ passes.def (working copy)
> @@ -89,6 +89,7 @@ along with GCC; see the file COPYING3.
> NEXT_PASS (pass_build_ealias);
> NEXT_PASS (pass_fre);
> NEXT_PASS (pass_merge_phi);
> + NEXT_PASS (pass_dse);
> NEXT_PASS (pass_cd_dce);
> NEXT_PASS (pass_early_ipa_sra);
> NEXT_PASS (pass_tail_recursion);
> Index: testsuite/g++.dg/tree-ssa/pr61034.C
> ===================================================================
> --- testsuite/g++.dg/tree-ssa/pr61034.C (revision 222016)
> +++ testsuite/g++.dg/tree-ssa/pr61034.C (working copy)
> @@ -42,6 +42,6 @@ bool f(I a, I b, I c, I d) {
> // This works only if everything is inlined into 'f'.
>
> // { dg-final { scan-tree-dump-times ";; Function" 1 "fre2" } }
> -// { dg-final { scan-tree-dump-times "free" 19 "fre2" } }
> +// { dg-final { scan-tree-dump-times "free" 18 "fre2" } }
> // { dg-final { scan-tree-dump-times "unreachable" 11 "fre2" } }
> // { dg-final { cleanup-tree-dump "fre2" } }
I got
FAIL: g++.dg/tree-ssa/pr61034.C -std=gnu++11 scan-tree-dump-times
fre2 "free" 18
FAIL: g++.dg/tree-ssa/pr61034.C -std=gnu++14 scan-tree-dump-times
fre2 "free" 18
FAIL: g++.dg/tree-ssa/pr61034.C -std=gnu++98 scan-tree-dump-times
fre2 "free" 18
on Linux/x86.
--
H.J.
^ permalink raw reply [flat|nested] 5+ messages in thread
* Re: Add DSE to early passes
2015-04-22 12:13 ` H.J. Lu
@ 2015-04-22 12:17 ` Jan Hubicka
2015-04-22 12:27 ` Richard Biener
0 siblings, 1 reply; 5+ messages in thread
From: Jan Hubicka @ 2015-04-22 12:17 UTC (permalink / raw)
To: H.J. Lu; +Cc: Jan Hubicka, Richard Guenther, GCC Patches
> > @@ -42,6 +42,6 @@ bool f(I a, I b, I c, I d) {
> > // This works only if everything is inlined into 'f'.
> >
> > // { dg-final { scan-tree-dump-times ";; Function" 1 "fre2" } }
> > -// { dg-final { scan-tree-dump-times "free" 19 "fre2" } }
> > +// { dg-final { scan-tree-dump-times "free" 18 "fre2" } }
> > // { dg-final { scan-tree-dump-times "unreachable" 11 "fre2" } }
> > // { dg-final { cleanup-tree-dump "fre2" } }
>
> I got
>
> FAIL: g++.dg/tree-ssa/pr61034.C -std=gnu++11 scan-tree-dump-times
> fre2 "free" 18
> FAIL: g++.dg/tree-ssa/pr61034.C -std=gnu++14 scan-tree-dump-times
> fre2 "free" 18
> FAIL: g++.dg/tree-ssa/pr61034.C -std=gnu++98 scan-tree-dump-times
> fre2 "free" 18
>
> on Linux/x86.
Hmm, I tested on pcc64, but now I see the same failure on x86-64.
Perhaps this testcase became target sensitive?
Richard, does the count really matter?
Honza
^ permalink raw reply [flat|nested] 5+ messages in thread
* Re: Add DSE to early passes
2015-04-22 12:17 ` Jan Hubicka
@ 2015-04-22 12:27 ` Richard Biener
0 siblings, 0 replies; 5+ messages in thread
From: Richard Biener @ 2015-04-22 12:27 UTC (permalink / raw)
To: Jan Hubicka; +Cc: H.J. Lu, GCC Patches
On Wed, 22 Apr 2015, Jan Hubicka wrote:
> > > @@ -42,6 +42,6 @@ bool f(I a, I b, I c, I d) {
> > > // This works only if everything is inlined into 'f'.
> > >
> > > // { dg-final { scan-tree-dump-times ";; Function" 1 "fre2" } }
> > > -// { dg-final { scan-tree-dump-times "free" 19 "fre2" } }
> > > +// { dg-final { scan-tree-dump-times "free" 18 "fre2" } }
> > > // { dg-final { scan-tree-dump-times "unreachable" 11 "fre2" } }
> > > // { dg-final { cleanup-tree-dump "fre2" } }
> >
> > I got
> >
> > FAIL: g++.dg/tree-ssa/pr61034.C -std=gnu++11 scan-tree-dump-times
> > fre2 "free" 18
> > FAIL: g++.dg/tree-ssa/pr61034.C -std=gnu++14 scan-tree-dump-times
> > fre2 "free" 18
> > FAIL: g++.dg/tree-ssa/pr61034.C -std=gnu++98 scan-tree-dump-times
> > fre2 "free" 18
> >
> > on Linux/x86.
>
> Hmm, I tested on pcc64, but now I see the same failure on x86-64.
> Perhaps this testcase became target sensitive?
> Richard, does the count really matter?
Well, it verifies that (part) of the desired optimization happens.
I can't see what should be target dependent here though.
Richard.
^ permalink raw reply [flat|nested] 5+ messages in thread
end of thread, other threads:[~2015-04-22 12:27 UTC | newest]
Thread overview: 5+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2015-04-13 10:01 Add DSE to early passes Jan Hubicka
2015-04-13 12:12 ` Richard Biener
2015-04-22 12:13 ` H.J. Lu
2015-04-22 12:17 ` Jan Hubicka
2015-04-22 12:27 ` Richard Biener
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).