On Thu, 23 Feb 2012, Jakub Jelinek wrote:
> Hi!
>
> IMHO we should treat CLOBBER stmts in various places like debug stmts,
> they shouldn't affect the decisions of the optimization phases, but
> if the pass does some transformation that somehow invalidates them or
> moves them (not in this pass), it needs to be adjusted. Otherwise we end
> up with the CLOBBERs inhibiting optimizations.
>
> This seems to work for ipa-split, the actual transformations DTRT
> apparently. Bootstrapped/regtested on x86_64-linux and i686-linux,
> ok for trunk?
Ok.
Thanks,
Richard.
> 2012-02-23 Jakub Jelinek <ja...@redhat.com>
>
> PR tree-optimization/52019
> * ipa-split.c (find_return_bb, find_retval, visit_bb): Ignore
> CLOBBER stmts.
>
> * gcc.dg/tree-ssa/ipa-split-6.c: New test.
>
> --- gcc/ipa-split.c.jj 2012-01-23 18:23:55.000000000 +0100
> +++ gcc/ipa-split.c 2012-02-23 12:30:21.877609719 +0100
> @@ -1,5 +1,5 @@
> /* Function splitting pass
> - Copyright (C) 2010, 2011
> + Copyright (C) 2010, 2011, 2012
> Free Software Foundation, Inc.
> Contributed by Jan Hubicka <j...@suse.cz>
>
> @@ -624,7 +624,9 @@ find_return_bb (void)
> for (bsi = gsi_last_bb (e->src); !gsi_end_p (bsi); gsi_prev (&bsi))
> {
> gimple stmt = gsi_stmt (bsi);
> - if (gimple_code (stmt) == GIMPLE_LABEL || is_gimple_debug (stmt))
> + if (gimple_code (stmt) == GIMPLE_LABEL
> + || is_gimple_debug (stmt)
> + || gimple_clobber_p (stmt))
> ;
> else if (gimple_code (stmt) == GIMPLE_ASSIGN
> && found_return
> @@ -657,7 +659,8 @@ find_retval (basic_block return_bb)
> for (bsi = gsi_start_bb (return_bb); !gsi_end_p (bsi); gsi_next (&bsi))
> if (gimple_code (gsi_stmt (bsi)) == GIMPLE_RETURN)
> return gimple_return_retval (gsi_stmt (bsi));
> - else if (gimple_code (gsi_stmt (bsi)) == GIMPLE_ASSIGN)
> + else if (gimple_code (gsi_stmt (bsi)) == GIMPLE_ASSIGN
> + && !gimple_clobber_p (gsi_stmt (bsi)))
> return gimple_assign_rhs1 (gsi_stmt (bsi));
> return NULL;
> }
> @@ -733,6 +736,9 @@ visit_bb (basic_block bb, basic_block re
> if (is_gimple_debug (stmt))
> continue;
>
> + if (gimple_clobber_p (stmt))
> + continue;
> +
> /* FIXME: We can split regions containing EH. We can not however
> split RESX, EH_DISPATCH and EH_POINTER referring to same region
> into different partitions. This would require tracking of
> --- gcc/testsuite/gcc.dg/tree-ssa/ipa-split-6.c.jj 2012-02-23
> 12:33:20.578790182 +0100
> +++ gcc/testsuite/gcc.dg/tree-ssa/ipa-split-6.c 2012-02-23
> 12:34:05.050612776 +0100
> @@ -0,0 +1,10 @@
> +/* PR tree-optimization/52019 */
> +/* { dg-do compile } */
> +/* { dg-options "-O3 -fno-tree-sra -fdump-tree-fnsplit
> -fdump-tree-optimized" } */
> +
> +#include "ipa-split-5.c"
> +
> +/* { dg-final { scan-tree-dump-times "Splitting function" 1 "fnsplit"} } */
> +/* { dg-final { cleanup-tree-dump "fnsplit" } } */
> +/* { dg-final { scan-tree-dump "part" "optimized"} } */
> +/* { dg-final { cleanup-tree-dump "optimized" } } */
>
> Jakub
>
>
--
Richard Guenther <rguent...@suse.de>
SUSE / SUSE Labs
SUSE LINUX Products GmbH - Nuernberg - AG Nuernberg - HRB 16746
GF: Jeff Hawn, Jennifer Guild, Felix Imendörffer
