Re: [PATCH] tree-optimization/79333 - fold stmts following SSA edges in VN

[Christophe Lyon <christophe.lyon@linaro.org>]

On Wed, 5 May 2021 at 09:56, Richard Biener <rguenther@suse.de> wrote:
>
> This makes sure to follow SSA edges when folding eliminated stmts.
> This reaps the same benefit as forwprop folding all stmts, not
> waiting for one to produce copysign in the new testcase.
>
> Bootstrapped on x86_64-unknown-linux-gnu, testing in progress.
>
> 2021-05-04  Richard Biener  <rguenther@suse.de>
>
>         PR tree-optimization/79333
>         * tree-ssa-sccvn.c (eliminate_dom_walker::eliminate_stmt):
>         Fold stmt following SSA edges.
>
>         * gcc.dg/tree-ssa/ssa-fre-94.c: New testcase.
>         * gcc.dg/graphite/fuse-1.c: Adjust.
>         * gcc.dg/pr43864-4.c: Likewise.
> ---
>  gcc/testsuite/gcc.dg/graphite/fuse-1.c     |  4 ++--
>  gcc/testsuite/gcc.dg/pr43864-4.c           |  6 +++---
>  gcc/testsuite/gcc.dg/tree-ssa/ssa-fre-94.c | 16 ++++++++++++++++
>  gcc/tree-ssa-sccvn.c                       |  2 +-
>  4 files changed, 22 insertions(+), 6 deletions(-)
>  create mode 100644 gcc/testsuite/gcc.dg/tree-ssa/ssa-fre-94.c
>
> diff --git a/gcc/testsuite/gcc.dg/graphite/fuse-1.c b/gcc/testsuite/gcc.dg/graphite/fuse-1.c
> index 204d3b20703..527b6e5c415 100644
> --- a/gcc/testsuite/gcc.dg/graphite/fuse-1.c
> +++ b/gcc/testsuite/gcc.dg/graphite/fuse-1.c
> @@ -1,6 +1,6 @@
>  /* Check that the two loops are fused and that we manage to fold the two xor
>     operations.  */
> -/* { dg-options "-O2 -floop-nest-optimize -fdump-tree-forwprop-all -fdump-tree-graphite-all" } */
> +/* { dg-options "-O2 -floop-nest-optimize -fdump-tree-forwprop4 -fdump-tree-graphite-all" } */
>
>  /* Make sure we fuse the loops like this:
>  AST generated by isl:
> @@ -12,7 +12,7 @@ for (int c0 = 0; c0 <= 99; c0 += 1) {
>  /* { dg-final { scan-tree-dump-times "AST generated by isl:.*for \\(int c0 = 0; c0 <= 99; c0 \\+= 1\\) \\{.*S_.*\\(c0\\);.*S_.*\\(c0\\);.*S_.*\\(c0\\);.*\\}" 1 "graphite" } } */
>
>  /* Check that after fusing the loops, the scalar computation is also fused.  */
> -/* { dg-final { scan-tree-dump-times "gimple_simplified to\[^\\n\]*\\^ 12" 1 "forwprop4" } } */
> +/* { dg-final { scan-tree-dump-times " \\^ 12;" 2 "forwprop4" } } */
>
>  #define MAX 100
>  int A[MAX];
> diff --git a/gcc/testsuite/gcc.dg/pr43864-4.c b/gcc/testsuite/gcc.dg/pr43864-4.c
> index 3c6cc50c5b8..8a25b0fd8ef 100644
> --- a/gcc/testsuite/gcc.dg/pr43864-4.c
> +++ b/gcc/testsuite/gcc.dg/pr43864-4.c
> @@ -22,7 +22,7 @@ int f(int c, int b, int d)
>    return r - r2;
>  }
>
> -/* { dg-final { scan-tree-dump-times "if " 0 "pre"} } */
> -/* { dg-final { scan-tree-dump-times "(?n)_.*\\+.*_" 1 "pre"} } */
> -/* { dg-final { scan-tree-dump-times "(?n)_.*-.*_" 2 "pre"} } */
> +/* During PRE elimination we should simplify this to return b * 2.  */
> +/* { dg-final { scan-tree-dump-times "if " 0 "pre" } } */
> +/* { dg-final { scan-tree-dump "_\[0-9\]+ = b_\[0-9\]+\\(D\\) \\* 2;\[\\r\\n\]\[^\\r\\n\]*return _\[0-9\]+;" "pre" } } */
>  /* { dg-final { scan-tree-dump-not "Invalid sum" "pre"} } */
> diff --git a/gcc/testsuite/gcc.dg/tree-ssa/ssa-fre-94.c b/gcc/testsuite/gcc.dg/tree-ssa/ssa-fre-94.c
> new file mode 100644
> index 00000000000..92eebf636c6
> --- /dev/null
> +++ b/gcc/testsuite/gcc.dg/tree-ssa/ssa-fre-94.c
> @@ -0,0 +1,16 @@
> +/* PR tree-optimization/79333 */
> +/* { dg-do compile } */
> +/* { dg-options "-O -ffinite-math-only -fdump-tree-fre1" } */
> +
> +extern __inline __attribute__ ((__always_inline__,__gnu_inline__))
> +double __attribute__ ((__nothrow__ , __leaf__))
> +fabs (double __x) { return __builtin_fabs (__x); }
> +
> +double f(float f)
> +{
> +  double t1 = fabs(f);
> +  double t2 = f / t1;
> +  return t2;
> +}
> +
> +/* { dg-final { scan-tree-dump "copysign" "fre1" } } */

This new testcase fails on aarch64-elf / arm-eabi with newlib.

Is that OK:
===========================
diff --git a/gcc/testsuite/gcc.dg/tree-ssa/ssa-fre-94.c
b/gcc/testsuite/gcc.dg/tree-ssa/ssa-fre-94.c
index 92eebf636c6..99c737562bb 100644
--- a/gcc/testsuite/gcc.dg/tree-ssa/ssa-fre-94.c
+++ b/gcc/testsuite/gcc.dg/tree-ssa/ssa-fre-94.c
@@ -1,5 +1,6 @@
 /* PR tree-optimization/79333 */
 /* { dg-do compile } */
+/* { dg-require-effective-target c99_runtime } */
 /* { dg-options "-O -ffinite-math-only -fdump-tree-fre1" } */

 extern __inline __attribute__ ((__always_inline__,__gnu_inline__))
===========================

Thanks,

Christophe




> diff --git a/gcc/tree-ssa-sccvn.c b/gcc/tree-ssa-sccvn.c
> index ca0974d72b8..e54a0c9065c 100644
> --- a/gcc/tree-ssa-sccvn.c
> +++ b/gcc/tree-ssa-sccvn.c
> @@ -6362,7 +6362,7 @@ eliminate_dom_walker::eliminate_stmt (basic_block b, gimple_stmt_iterator *gsi)
>         recompute_tree_invariant_for_addr_expr (gimple_assign_rhs1 (stmt));
>        gimple_stmt_iterator prev = *gsi;
>        gsi_prev (&prev);
> -      if (fold_stmt (gsi))
> +      if (fold_stmt (gsi, follow_all_ssa_edges))
>         {
>           /* fold_stmt may have created new stmts inbetween
>              the previous stmt and the folded stmt.  Mark
> --
> 2.26.2