[Fwd: Fixing PR optimization/9258 and PR fortran/7681 on 3.3 branch.]

[Fwd: Fixing PR optimization/9258 and PR fortran/7681 on 3.3 branch.]

[Toon Moene]

[-- Attachment #1: Type: text/plain, Size: 460 bytes --]

Could someone with global write privileges please ACK this patch, which 
has gone into 3.2.2 and 3.4, so that 3.3 is now in regression with 
respect to this failure ?

Thanks in advance,

-- 
Toon Moene - mailto:toon@moene.indiv.nluug.nl - phoneto: +31 346 214290
Saturnushof 14, 3738 XG  Maartensdijk, The Netherlands
Maintainer, GNU Fortran 77: http://gcc.gnu.org/onlinedocs/g77_news.html
GNU Fortran 95: http://gcc-g95.sourceforge.net/ (under construction)

[-- Attachment #2: Fixing PR optimization/9258 and PR fortran/7681 on 3.3 branch. --]
[-- Type: message/rfc822, Size: 2987 bytes --]

[-- Attachment #2.1.1: Type: text/plain, Size: 728 bytes --]

L.S.,

As far as I can determine, the attached patch by Jan Hubicka which 
solves the indicated PR's should still be applied to the 3.3 branch.

Just for checks I performed make bootstrap, make -k check on 
powerpc-unknown-linux-gnu, but the problem exists on the ix86 architecture.

Andris, could you do a bootstrap & check with this patch to see if it 
fixes the pr9258.f failures that occur in your submission to 
gcc-testresults ?

Thanks in advance,

-- 
Toon Moene - mailto:toon@moene.indiv.nluug.nl - phoneto: +31 346 214290
Saturnushof 14, 3738 XG  Maartensdijk, The Netherlands
Maintainer, GNU Fortran 77: http://gcc.gnu.org/onlinedocs/g77_news.html
GNU Fortran 95: http://gcc-g95.sourceforge.net/ (under construction)

[-- Attachment #2.1.2: 9258.patch --]
[-- Type: text/plain, Size: 1267 bytes --]

2003-02-09  Jan Hubicka  <jh@suse.cz>

	PR optimization/9258, PR fortran/7681
	* global.c (struct allocno): Add no_stack_reg.
	(global_conflicts): Set it.
	(find_reg): Use it.

*** global.c.orig	Fri Jan  3 23:24:35 2003
--- global.c	Sun Feb  9 11:57:54 2003
*************** struct allocno
*** 131,134 ****
--- 131,139 ----
  
    HARD_REG_SET regs_someone_prefers;
+ 
+ #ifdef STACK_REGS
+   /* Set to true if allocno can't be allocated in the stack register.  */
+   bool no_stack_reg;
+ #endif
  };
  
*************** global_conflicts ()
*** 707,712 ****
  	      break;
  	  if (e != NULL)
! 	    for (ax = FIRST_STACK_REG; ax <= LAST_STACK_REG; ax++)
! 	      record_one_conflict (ax);
  	}
  #endif
--- 712,723 ----
  	      break;
  	  if (e != NULL)
! 	    {
! 	      EXECUTE_IF_SET_IN_ALLOCNO_SET (allocnos_live, ax,
! 		{
! 		  allocno[ax].no_stack_reg = 1;
! 		});
! 	      for (ax = FIRST_STACK_REG; ax <= LAST_STACK_REG; ax++)
! 	        record_one_conflict (ax);
! 	    }
  	}
  #endif
*************** find_reg (num, losers, alt_regs_p, accep
*** 1205,1208 ****
--- 1216,1223 ----
  					  mode)
  #endif
+ #ifdef STACK_REGS
+ 	      && (!allocno[num].no_stack_reg
+ 		  || regno < FIRST_STACK_REG || regno > LAST_STACK_REG)
+ #endif
  	      )
  	    {

Re: [Fwd: Fixing PR optimization/9258 and PR fortran/7681 on 3.3 branch.]

[Richard Henderson]

On Fri, Feb 14, 2003 at 10:35:42PM +0100, Toon Moene wrote:
> Could someone with global write privileges please ACK this patch, which 
> has gone into 3.2.2 and 3.4, so that 3.3 is now in regression with 
> respect to this failure ?

Sure.


r~

Re: [Fwd: Fixing PR optimization/9258 and PR fortran/7681 on 3.3 branch.]

[Mark Mitchell]

--On Friday, February 14, 2003 10:35:42 PM +0100 Toon Moene 
<toon@moene.indiv.nluug.nl> wrote:

> Could someone with global write privileges please ACK this patch, which
> has gone into 3.2.2 and 3.4, so that 3.3 is now in regression with
> respect to this failure ?

If it's in both 3.2.2 and 3.4, it can certainly go into 3.3.  Please go 
ahead.

Thanks,

-- 
Mark Mitchell                mark@codesourcery.com
CodeSourcery, LLC            http://www.codesourcery.com

Re: [Fwd: Fixing PR optimization/9258 and PR fortran/7681 on 3.3 branch.]

[Toon Moene]

Mark Mitchell wrote:

> --On Friday, February 14, 2003 10:35:42 PM +0100 Toon Moene 
> <toon@moene.indiv.nluug.nl> wrote:
> 
>> Could someone with global write privileges please ACK this patch, which
>> has gone into 3.2.2 and 3.4, so that 3.3 is now in regression with
>> respect to this failure ?

> If it's in both 3.2.2 and 3.4, it can certainly go into 3.3.  Please go 
> ahead.

Done (it was already approved by Richard Henderson, but thanks anyway :-)

-- 
Toon Moene - mailto:toon@moene.indiv.nluug.nl - phoneto: +31 346 214290
Saturnushof 14, 3738 XG  Maartensdijk, The Netherlands
Maintainer, GNU Fortran 77: http://gcc.gnu.org/onlinedocs/g77_news.html
GNU Fortran 95: http://gcc-g95.sourceforge.net/ (under construction)