Re: optimization/7476: constant variable tested in loop

Re: optimization/7476: constant variable tested in loop

[Daniel Berlin]

The following reply was made to PR optimization/7476; it has been noted by GNATS.

From: Daniel Berlin <dberlin@dberlin.org>
To: Kevin Ryde <user42@zip.com.au>
Cc: gcc-gnats@gcc.gnu.org, <nobody@gcc.gnu.org>,
	<gcc-prs@gcc.gnu.org>, <gcc-bugs@gcc.gnu.org>
Subject: Re: optimization/7476: constant variable tested in loop
Date: Fri, 2 Aug 2002 22:27:50 -0400 (EDT)

 Fixed by the local cprop stuff Jan added to the mainline.
         .file   "testconst.c"
         .text
         .align 2
         .p2align 4,,15
 .globl foo
         .type   foo, @function
 foo:
         pushl   %ebp
         movl    $9, %eax
         movl    %esp, %ebp
         .p2align 4,,15
 .L7:
         decl    %eax
         jns     .L7
         popl    %ebp
         ret
         .size   foo, .-foo
         .ident  "GCC: (GNU) 3.3 20020729 (experimental)
 
 On Sat, 3 Aug 2002, Kevin Ryde wrote:
 
 > 
 > >Number:         7476
 > >Category:       optimization
 > >Synopsis:       constant variable tested in loop
 > >Confidential:   no
 > >Severity:       non-critical
 > >Priority:       low
 > >Responsible:    unassigned
 > >State:          open
 > >Class:          pessimizes-code
 > >Submitter-Id:   net
 > >Arrival-Date:   Fri Aug 02 16:06:01 PDT 2002
 > >Closed-Date:
 > >Last-Modified:
 > >Originator:     
 > >Release:        3.1.1 20020606 (Debian prerelease) (Debian testing/unstable)
 > >Organization:
 > >Environment:
 > System: Linux blah 2.2.15 #1 Tue Apr 25 17:13:48 EST 2000 i586 unknown
 > Architecture: i586
 > 	<machine, os, target, libraries (multiple lines)>
 > host: i386-pc-linux-gnu
 > build: i386-pc-linux-gnu
 > target: i386-pc-linux-gnu
 > configured with: /mnt/data/gcc-3.1/gcc-3.1-3.1.1ds1/src/configure -v --enable-languages=c,c++,java,f77,proto,objc,ada --prefix=/usr --mandir=/usr/share/man --infodir=/usr/share/info --with-gxx-include-dir=/usr/include/g++-v3-3.1 --enable-shared --with-system-zlib --enable-long-long --enable-nls --without-included-gettext --enable-clocale=gnu --enable-threads=posix --enable-java-gc=boehm --enable-objc-gc i386-linux
 > 
 > >Description:
 > A constant variable tested within a loop doesn't seem to be
 > recognised as constant.
 > 
 > >How-To-Repeat:
 > The file foo.c below compiled with
 > 
 > 	gcc-3.1 -O9 -S foo.c
 > 
 > produces the foo.s below.  Notice that it still contains a call to
 > "puts", despite the "cond" variable in the source always being false.
 > 
 > Outside a loop such a construct seems to be recognised as constant and
 > the dependent code goes dead, but within a loop that doesn't seem to
 > happen.
 > 
 > I'm not sure if it's too much to expect something like this in a loop
 > to go dead.  I guess in the presence of gotos or loops it's not always
 > too easy to be certain something really is constant.  A variable
 > written only once with a constant value like this might be tractable
 > though.
 > 
 > For what it's worth this arose within gmp where the "cond" condition
 > tested was sometimes a compile-time constant, or sometimes a runtime
 > test, depending on the configuration.  It had seemed like a good idea
 > just to put the result in a variable in either case, and let the
 > optimizer drop unused code in the constant case.
 > 
 > 
 > 
 > --=-=-=
 > Content-Type: text/x-csrc
 > Content-Disposition: attachment; filename=foo.c
 > 
 > void
 > foo (void)
 > {
 >   int  cond = 0;
 >   int  i;
 > 
 >   for (i = 0; i < 10; i++)
 >     if (cond)
 >       puts ("hello");
 > }
 > 
 > --=-=-=
 > Content-Disposition: attachment; filename=foo.s
 > 
 > 	.file	"foo.c"
 > 	.section	.rodata.str1.1,"aMS",@progbits,1
 > .LC0:
 > 	.string	"hello"
 > 	.text
 > 	.align 2
 > 	.p2align 2,,3
 > .globl foo
 > 	.type	foo,@function
 > foo:
 > 	pushl	%ebp
 > 	movl	%esp, %ebp
 > 	pushl	%esi
 > 	pushl	%ebx
 > 	xorl	%esi, %esi
 > 	movl	$9, %ebx
 > 	.p2align 2,,3
 > .L7:
 > 	testl	%esi, %esi
 > 	jne	.L10
 > .L4:
 > 	decl	%ebx
 > 	jns	.L7
 > 	leal	-8(%ebp), %esp
 > 	popl	%ebx
 > 	popl	%esi
 > 	leave
 > 	ret
 > 	.p2align 2,,3
 > .L10:
 > 	subl	$12, %esp
 > 	pushl	$.LC0
 > 	call	puts
 > 	addl	$16, %esp
 > 	jmp	.L4
 > .Lfe1:
 > 	.size	foo,.Lfe1-foo
 > 	.ident	"GCC: (GNU) 3.1.1 20020703 (Debian prerelease)"
 > 
 > --=-=-=--
 > >Fix:
 > >Release-Note:
 > >Audit-Trail:
 > >Unformatted:
 >  --=-=-=
 >  
 > 
 > 
 

Re: optimization/7476: constant variable tested in loop

[bangerth]

Synopsis: constant variable tested in loop

State-Changed-From-To: open->closed
State-Changed-By: bangerth
State-Changed-When: Fri Dec  6 13:35:33 2002
State-Changed-Why:
    This works on the mainline, i.e. what will become 3.3

http://gcc.gnu.org/cgi-bin/gnatsweb.pl?cmd=view%20audit-trail&database=gcc&pr=7476

Re: optimization/7476: constant variable tested in loop

[Andrew Pinski]

The following reply was made to PR optimization/7476; it has been noted by GNATS.

From: Andrew Pinski <pinskia@physics.uc.edu>
To: Kevin Ryde <user42@zip.com.au>
Cc: gcc-gnats@gcc.gnu.org
Subject: Re: optimization/7476: constant variable tested in loop
Date: Fri, 2 Aug 2002 19:12:02 -0400

 This has been fixed in at least gcc 3.3 20020802 (experimental):
          .file   "foo.c"
          .text
          .align 2
          .p2align 4,,15
 .globl foo
          .type   foo, @function
 foo:
          pushl   %ebp
          movl    $9, %eax
          movl    %esp, %ebp
          .p2align 4,,15
 .L7:
          decl    %eax
          jns     .L7
          popl    %ebp
          ret
          .size   foo, .-foo
          .ident  "GCC: (GNU) 3.3 20020802 (experimental)"
 
 which is compared to gcc 3.1.1:
          .file   "foo.c"
          .section        .rodata.str1.1,"aMS",@progbits,1
 .LC0:
          .string "hello"
          .text
          .align 2
          .p2align 4,,15
 .globl foo
          .type   foo,@function
 foo:
          pushl   %ebp
          movl    %esp, %ebp
          pushl   %esi
          xorl    %esi, %esi
          pushl   %ebx
          movl    $9, %ebx
          subl    $16, %esp
          .p2align 4,,15
 .L7:
          testl   %esi, %esi
          jne     .L10
 .L4:
          decl    %ebx
          jns     .L7
          addl    $16, %esp
          popl    %ebx
          popl    %esi
          popl    %ebp
          ret
          .p2align 4,,7
 .L10:
          movl    $.LC0, (%esp)
          call    puts
          jmp     .L4
 .Lfe1:
          .size   foo,.Lfe1-foo
          .ident  "GCC: (GNU) 3.1.1"
 
 
 

optimization/7476: constant variable tested in loop

[Kevin Ryde]

>Number:         7476
>Category:       optimization
>Synopsis:       constant variable tested in loop
>Confidential:   no
>Severity:       non-critical
>Priority:       low
>Responsible:    unassigned
>State:          open
>Class:          pessimizes-code
>Submitter-Id:   net
>Arrival-Date:   Fri Aug 02 16:06:01 PDT 2002
>Closed-Date:
>Last-Modified:
>Originator:     
>Release:        3.1.1 20020606 (Debian prerelease) (Debian testing/unstable)
>Organization:
>Environment:
System: Linux blah 2.2.15 #1 Tue Apr 25 17:13:48 EST 2000 i586 unknown
Architecture: i586
	<machine, os, target, libraries (multiple lines)>
host: i386-pc-linux-gnu
build: i386-pc-linux-gnu
target: i386-pc-linux-gnu
configured with: /mnt/data/gcc-3.1/gcc-3.1-3.1.1ds1/src/configure -v --enable-languages=c,c++,java,f77,proto,objc,ada --prefix=/usr --mandir=/usr/share/man --infodir=/usr/share/info --with-gxx-include-dir=/usr/include/g++-v3-3.1 --enable-shared --with-system-zlib --enable-long-long --enable-nls --without-included-gettext --enable-clocale=gnu --enable-threads=posix --enable-java-gc=boehm --enable-objc-gc i386-linux

>Description:
A constant variable tested within a loop doesn't seem to be
recognised as constant.

>How-To-Repeat:
The file foo.c below compiled with

	gcc-3.1 -O9 -S foo.c

produces the foo.s below.  Notice that it still contains a call to
"puts", despite the "cond" variable in the source always being false.

Outside a loop such a construct seems to be recognised as constant and
the dependent code goes dead, but within a loop that doesn't seem to
happen.

I'm not sure if it's too much to expect something like this in a loop
to go dead.  I guess in the presence of gotos or loops it's not always
too easy to be certain something really is constant.  A variable
written only once with a constant value like this might be tractable
though.

For what it's worth this arose within gmp where the "cond" condition
tested was sometimes a compile-time constant, or sometimes a runtime
test, depending on the configuration.  It had seemed like a good idea
just to put the result in a variable in either case, and let the
optimizer drop unused code in the constant case.



--=-=-=
Content-Type: text/x-csrc
Content-Disposition: attachment; filename=foo.c

void
foo (void)
{
  int  cond = 0;
  int  i;

  for (i = 0; i < 10; i++)
    if (cond)
      puts ("hello");
}

--=-=-=
Content-Disposition: attachment; filename=foo.s

	.file	"foo.c"
	.section	.rodata.str1.1,"aMS",@progbits,1
.LC0:
	.string	"hello"
	.text
	.align 2
	.p2align 2,,3
.globl foo
	.type	foo,@function
foo:
	pushl	%ebp
	movl	%esp, %ebp
	pushl	%esi
	pushl	%ebx
	xorl	%esi, %esi
	movl	$9, %ebx
	.p2align 2,,3
.L7:
	testl	%esi, %esi
	jne	.L10
.L4:
	decl	%ebx
	jns	.L7
	leal	-8(%ebp), %esp
	popl	%ebx
	popl	%esi
	leave
	ret
	.p2align 2,,3
.L10:
	subl	$12, %esp
	pushl	$.LC0
	call	puts
	addl	$16, %esp
	jmp	.L4
.Lfe1:
	.size	foo,.Lfe1-foo
	.ident	"GCC: (GNU) 3.1.1 20020703 (Debian prerelease)"

--=-=-=--
>Fix:
>Release-Note:
>Audit-Trail:
>Unformatted:
 --=-=-=