On Wed, 14 Nov 2018, Jakub Jelinek wrote:
> On Tue, Nov 13, 2018 at 10:19:10AM +0100, Jakub Jelinek wrote:
> > > > Though, in this particular case the sqrt call is
> > > > optimized away, so it wouldn't make a difference.
> > > >
> > > > Bootstrapped/regtested on x86_64-linux and i686-linux, ok for trunk, or
> > > > should I do the gimple_build_assign + gsi_replace change?
> > >
> > > I think that would be cleaner.
> >
> > Ok, will do. Thanks.
>
> Here it is. Bootstrapped/regtested on x86_64-linux and i686-linux, ok for
> trunk?
OK.
Richard.
> 2018-11-13 Jakub Jelinek <ja...@redhat.com>
>
> PR tree-optimization/87977
> * tree-ssa-math-opts.c (optimize_recip_sqrt): Don't reuse division
> stmt, build a new one and replace the old one with it. Formatting fix.
> Call release_ssa_name (x) if !has_other_use and !delete_div.
> (pass_cse_reciprocals::execute): Before calling optimize_recip_sqrt
> verify lhs of stmt is still def.
>
> * gcc.dg/recip_sqrt_mult_1.c: Add -fcompare-debug to dg-options.
> * gcc.dg/recip_sqrt_mult_2.c: Likewise.
> * gcc.dg/recip_sqrt_mult_3.c: Likewise.
> * gcc.dg/recip_sqrt_mult_4.c: Likewise.
> * gcc.dg/recip_sqrt_mult_5.c: Likewise.
>
> --- gcc/tree-ssa-math-opts.c.jj 2018-11-12 20:01:19.224793981 +0100
> +++ gcc/tree-ssa-math-opts.c 2018-11-13 11:30:16.326203020 +0100
> @@ -652,10 +652,14 @@ optimize_recip_sqrt (gimple_stmt_iterato
> print_gimple_stmt (dump_file, stmt, 0, TDF_NONE);
> fprintf (dump_file, "with new division\n");
> }
> - gimple_assign_set_lhs (stmt, sqr_ssa_name);
> - gimple_assign_set_rhs2 (stmt, a);
> + stmt
> + = gimple_build_assign (sqr_ssa_name, gimple_assign_rhs_code (stmt),
> + gimple_assign_rhs1 (stmt), a);
> + gsi_insert_before (def_gsi, stmt, GSI_SAME_STMT);
> + gsi_remove (def_gsi, true);
> + *def_gsi = gsi_for_stmt (stmt);
> fold_stmt_inplace (def_gsi);
> update_stmt (stmt);
>
> if (dump_file)
> print_gimple_stmt (dump_file, stmt, 0, TDF_NONE);
> @@ -704,7 +707,7 @@ optimize_recip_sqrt (gimple_stmt_iterato
>
> gimple *new_stmt
> = gimple_build_assign (x, MULT_EXPR,
> - orig_sqrt_ssa_name, sqr_ssa_name);
> + orig_sqrt_ssa_name, sqr_ssa_name);
> gsi_insert_after (def_gsi, new_stmt, GSI_NEW_STMT);
> update_stmt (stmt);
> }
> @@ -715,6 +718,8 @@ optimize_recip_sqrt (gimple_stmt_iterato
> gsi_remove (&gsi2, true);
> release_defs (stmt);
> }
> + else
> + release_ssa_name (x);
> }
>
> /* Look for floating-point divisions among DEF's uses, and try to
> @@ -951,6 +956,7 @@ pass_cse_reciprocals::execute (function
> stmt = gsi_stmt (gsi);
> if (flag_unsafe_math_optimizations
> && is_gimple_assign (stmt)
> + && gimple_assign_lhs (stmt) == def
> && !stmt_can_throw_internal (cfun, stmt)
> && gimple_assign_rhs_code (stmt) == RDIV_EXPR)
> optimize_recip_sqrt (&gsi, def);
> --- gcc/testsuite/gcc.dg/recip_sqrt_mult_1.c.jj 2018-11-12
> 20:01:19.315792497 +0100
> +++ gcc/testsuite/gcc.dg/recip_sqrt_mult_1.c 2018-11-13 11:17:41.667541915
> +0100
> @@ -1,5 +1,5 @@
> /* { dg-do compile } */
> -/* { dg-options "-Ofast -fdump-tree-recip" } */
> +/* { dg-options "-Ofast -fdump-tree-recip -fcompare-debug" } */
>
> double res, res2, tmp;
> void
> --- gcc/testsuite/gcc.dg/recip_sqrt_mult_2.c.jj 2018-11-12
> 20:01:19.320792415 +0100
> +++ gcc/testsuite/gcc.dg/recip_sqrt_mult_2.c 2018-11-13 11:17:41.668541899
> +0100
> @@ -1,5 +1,5 @@
> /* { dg-do compile } */
> -/* { dg-options "-Ofast -fdump-tree-optimized" } */
> +/* { dg-options "-Ofast -fdump-tree-optimized -fcompare-debug" } */
>
> float
> foo (float a)
> --- gcc/testsuite/gcc.dg/recip_sqrt_mult_3.c.jj 2018-11-12
> 20:01:19.363791713 +0100
> +++ gcc/testsuite/gcc.dg/recip_sqrt_mult_3.c 2018-11-13 11:17:41.671541850
> +0100
> @@ -1,5 +1,5 @@
> /* { dg-do compile } */
> -/* { dg-options "-Ofast -fdump-tree-optimized" } */
> +/* { dg-options "-Ofast -fdump-tree-optimized -fcompare-debug" } */
>
> double
> foo (double a)
> --- gcc/testsuite/gcc.dg/recip_sqrt_mult_4.c.jj 2018-11-12
> 20:01:19.371791582 +0100
> +++ gcc/testsuite/gcc.dg/recip_sqrt_mult_4.c 2018-11-13 11:17:41.686541603
> +0100
> @@ -1,5 +1,5 @@
> /* { dg-do compile } */
> -/* { dg-options "-Ofast -fdump-tree-recip" } */
> +/* { dg-options "-Ofast -fdump-tree-recip -fcompare-debug" } */
>
> /* The main path doesn't have any multiplications.
> Avoid introducing them in the recip pass. */
> --- gcc/testsuite/gcc.dg/recip_sqrt_mult_5.c.jj 2018-11-12
> 20:01:19.386791336 +0100
> +++ gcc/testsuite/gcc.dg/recip_sqrt_mult_5.c 2018-11-13 11:17:41.705541293
> +0100
> @@ -1,5 +1,5 @@
> /* { dg-do compile } */
> -/* { dg-options "-Ofast -fdump-tree-recip" } */
> +/* { dg-options "-Ofast -fdump-tree-recip -fcompare-debug" } */
>
> /* We want to do the recip_sqrt transformations here there is already
> a multiplication on the main path. */
>
>
> Jakub
>
>
--
Richard Biener <rguent...@suse.de>
SUSE LINUX GmbH, GF: Felix Imendoerffer, Jane Smithard, Graham Norton, HRB
21284 (AG Nuernberg)
