Optimize std::max early

Optimize std::max early

[Jan Hubicka]

Hi,
we currently produce very bad code on loops using std::vector as a stack, since
we fail to inline push_back which in turn prevents SRA and we fail to optimize
out some store-to-load pairs (PR109849).

I looked into why this function is not inlined and it is inlined by clang.  We
currently estimate it to 66 instructions and inline limits are 15 at -O2 and 30
at -O3.  Clang has similar estimate, but still decides to inline at -O2.

I looked into reason why the body is so large and one problem I spotted is the
way std::max is implemented by taking and returning reference to the values.

  const T& max( const T& a, const T& b );

This makes it necessary to store the values to memory and load them later
(Max is used by code computing new size of vector on resize.)
Two stores, conditional and load accounts as 8 instructions, while
MAX_EXPR as 1 and has a lot better chance to fold with the surrounding
code.

We optimize this to MAX_EXPR, but only during late optimizations.  I think this
is a common enough coding pattern and we ought to make this transparent to
early opts and IPA.  The following is easist fix that simply adds phiprop pass
that turns the PHI of address values into PHI of values so later FRE can
propagate values across memory, phiopt discover the MAX_EXPR pattern and DSE
remove the memory stores.

Bootstrapped/regtested x86_64-linux, does this look resonable thing to do?

Looking into how expensive the pass is, I think it is very cheap, except
that it computes postdominator and updates ssa even if no patterns
are matched.  I will send patch to avoid that.

gcc/ChangeLog:

	PR tree-optimization/109811
	PR tree-optimization/109849
	* passes.def: Add phiprop to early optimization passes.
	* tree-ssa-phiprop.cc: Allow clonning.

gcc/testsuite/ChangeLog:

	PR tree-optimization/109811
	PR tree-optimization/109849
	* gcc.dg/tree-ssa/phiprop-1.c: New test.

diff --git a/gcc/passes.def b/gcc/passes.def
index c9a8f19747b..faa5208b26b 100644
--- a/gcc/passes.def
+++ b/gcc/passes.def
@@ -88,6 +88,8 @@ along with GCC; see the file COPYING3.  If not see
 	  /* pass_build_ealias is a dummy pass that ensures that we
 	     execute TODO_rebuild_alias at this point.  */
 	  NEXT_PASS (pass_build_ealias);
+	  /* Do phiprop before FRE so we optimize std::min and std::max well.  */
+	  NEXT_PASS (pass_phiprop);
 	  NEXT_PASS (pass_fre, true /* may_iterate */);
 	  NEXT_PASS (pass_early_vrp);
 	  NEXT_PASS (pass_merge_phi);
diff --git a/gcc/testsuite/gcc.dg/tree-ssa/phiprop-1.c b/gcc/testsuite/gcc.dg/tree-ssa/phiprop-1.c
new file mode 100644
index 00000000000..9f52c2a7298
--- /dev/null
+++ b/gcc/testsuite/gcc.dg/tree-ssa/phiprop-1.c
@@ -0,0 +1,14 @@
+/* { dg-do compile } */
+/* { dg-options "-O1 -fdump-tree-phiprop1-details -fdump-tree-release_ssa" } */
+int max(int a, int b)
+{
+        int *ptr;
+        if (a > b)
+          ptr = &a;
+        else
+          ptr = &b;
+        return *ptr;
+}
+
+/* { dg-final { scan-tree-dump-times "Inserting PHI for result of load" 1 "phiprop1"} } */
+/* { dg-final { scan-tree-dump-times "MAX_EXPR" 1 "release_ssa"} } */
diff --git a/gcc/tree-ssa-phiprop.cc b/gcc/tree-ssa-phiprop.cc
index 3cb4900b6be..5dc505df420 100644
--- a/gcc/tree-ssa-phiprop.cc
+++ b/gcc/tree-ssa-phiprop.cc
@@ -476,6 +476,7 @@ public:
   {}
 
   /* opt_pass methods: */
+  opt_pass * clone () final override { return new pass_phiprop (m_ctxt); }
   bool gate (function *) final override { return flag_tree_phiprop; }
   unsigned int execute (function *) final override;
 

Re: Optimize std::max early

[Richard Biener]

On Sun, Jun 18, 2023 at 5:55 PM Jan Hubicka via Gcc-patches
<gcc-patches@gcc.gnu.org> wrote:
>
> Hi,
> we currently produce very bad code on loops using std::vector as a stack, since
> we fail to inline push_back which in turn prevents SRA and we fail to optimize
> out some store-to-load pairs (PR109849).
>
> I looked into why this function is not inlined and it is inlined by clang.  We
> currently estimate it to 66 instructions and inline limits are 15 at -O2 and 30
> at -O3.  Clang has similar estimate, but still decides to inline at -O2.
>
> I looked into reason why the body is so large and one problem I spotted is the
> way std::max is implemented by taking and returning reference to the values.
>
>   const T& max( const T& a, const T& b );
>
> This makes it necessary to store the values to memory and load them later
> (Max is used by code computing new size of vector on resize.)
> Two stores, conditional and load accounts as 8 instructions, while
> MAX_EXPR as 1 and has a lot better chance to fold with the surrounding
> code.
>
> We optimize this to MAX_EXPR, but only during late optimizations.  I think this
> is a common enough coding pattern and we ought to make this transparent to
> early opts and IPA.  The following is easist fix that simply adds phiprop pass
> that turns the PHI of address values into PHI of values so later FRE can
> propagate values across memory, phiopt discover the MAX_EXPR pattern and DSE
> remove the memory stores.
>
> Bootstrapped/regtested x86_64-linux, does this look resonable thing to do?

OK.

Thanks,
Richard.

> Looking into how expensive the pass is, I think it is very cheap, except
> that it computes postdominator and updates ssa even if no patterns
> are matched.  I will send patch to avoid that.
>
> gcc/ChangeLog:
>
>         PR tree-optimization/109811
>         PR tree-optimization/109849
>         * passes.def: Add phiprop to early optimization passes.
>         * tree-ssa-phiprop.cc: Allow clonning.
>
> gcc/testsuite/ChangeLog:
>
>         PR tree-optimization/109811
>         PR tree-optimization/109849
>         * gcc.dg/tree-ssa/phiprop-1.c: New test.
>
> diff --git a/gcc/passes.def b/gcc/passes.def
> index c9a8f19747b..faa5208b26b 100644
> --- a/gcc/passes.def
> +++ b/gcc/passes.def
> @@ -88,6 +88,8 @@ along with GCC; see the file COPYING3.  If not see
>           /* pass_build_ealias is a dummy pass that ensures that we
>              execute TODO_rebuild_alias at this point.  */
>           NEXT_PASS (pass_build_ealias);
> +         /* Do phiprop before FRE so we optimize std::min and std::max well.  */
> +         NEXT_PASS (pass_phiprop);
>           NEXT_PASS (pass_fre, true /* may_iterate */);
>           NEXT_PASS (pass_early_vrp);
>           NEXT_PASS (pass_merge_phi);
> diff --git a/gcc/testsuite/gcc.dg/tree-ssa/phiprop-1.c b/gcc/testsuite/gcc.dg/tree-ssa/phiprop-1.c
> new file mode 100644
> index 00000000000..9f52c2a7298
> --- /dev/null
> +++ b/gcc/testsuite/gcc.dg/tree-ssa/phiprop-1.c
> @@ -0,0 +1,14 @@
> +/* { dg-do compile } */
> +/* { dg-options "-O1 -fdump-tree-phiprop1-details -fdump-tree-release_ssa" } */
> +int max(int a, int b)
> +{
> +        int *ptr;
> +        if (a > b)
> +          ptr = &a;
> +        else
> +          ptr = &b;
> +        return *ptr;
> +}
> +
> +/* { dg-final { scan-tree-dump-times "Inserting PHI for result of load" 1 "phiprop1"} } */
> +/* { dg-final { scan-tree-dump-times "MAX_EXPR" 1 "release_ssa"} } */
> diff --git a/gcc/tree-ssa-phiprop.cc b/gcc/tree-ssa-phiprop.cc
> index 3cb4900b6be..5dc505df420 100644
> --- a/gcc/tree-ssa-phiprop.cc
> +++ b/gcc/tree-ssa-phiprop.cc
> @@ -476,6 +476,7 @@ public:
>    {}
>
>    /* opt_pass methods: */
> +  opt_pass * clone () final override { return new pass_phiprop (m_ctxt); }
>    bool gate (function *) final override { return flag_tree_phiprop; }
>    unsigned int execute (function *) final override;
>