Re: [PATCH] tree-optimization/112623 - forwprop VEC_PACK_TRUNC generation

Re: [PATCH] tree-optimization/112623 - forwprop VEC_PACK_TRUNC generation

[Jakub Jelinek]

On Tue, Nov 21, 2023 at 02:35:02PM +0000, Richard Biener wrote:
> For vec_pack_trunc patterns there can be an ambiguity for the
> source mode for BFmode vs HFmode.  The vectorizer checks
> the insns operand mode for this, the following makes forwprop
> do the same.  That of course doesn't help if the target supports
> both conversions.
> 
> Bootstrapped and tested on x86_64-unknown-linux-gnu, OK?
> 
> Thanks,
> Richard.
> 
> 	PR tree-optimization/112623
> 	* tree-ssa-forwprop.cc (simplify_vector_constructor):
> 	Check the source mode of the insn for vector pack/unpacks.
> 
> 	* gcc.target/i386/pr112623.c: New testcase.

LGTM, thanks.

	Jakub

[PATCH] tree-optimization/112623 - forwprop VEC_PACK_TRUNC generation

[Richard Biener]

For vec_pack_trunc patterns there can be an ambiguity for the
source mode for BFmode vs HFmode.  The vectorizer checks
the insns operand mode for this, the following makes forwprop
do the same.  That of course doesn't help if the target supports
both conversions.

Bootstrapped and tested on x86_64-unknown-linux-gnu, OK?

Thanks,
Richard.

	PR tree-optimization/112623
	* tree-ssa-forwprop.cc (simplify_vector_constructor):
	Check the source mode of the insn for vector pack/unpacks.

	* gcc.target/i386/pr112623.c: New testcase.
---
 gcc/testsuite/gcc.target/i386/pr112623.c | 11 +++++++++++
 gcc/tree-ssa-forwprop.cc                 | 13 +++++++++----
 2 files changed, 20 insertions(+), 4 deletions(-)
 create mode 100644 gcc/testsuite/gcc.target/i386/pr112623.c

diff --git a/gcc/testsuite/gcc.target/i386/pr112623.c b/gcc/testsuite/gcc.target/i386/pr112623.c
new file mode 100644
index 00000000000..c4ebacec85c
--- /dev/null
+++ b/gcc/testsuite/gcc.target/i386/pr112623.c
@@ -0,0 +1,11 @@
+/* { dg-do compile } */
+/* { dg-options "-O -mavx512vl -mavx512fp16" } */
+
+typedef __bf16 __attribute__((__vector_size__ (16))) BF;
+typedef float __attribute__((__vector_size__ (32))) F;
+
+BF
+foo (F f)
+{
+  return __builtin_convertvector (f, BF);
+}
diff --git a/gcc/tree-ssa-forwprop.cc b/gcc/tree-ssa-forwprop.cc
index d39dfc1065f..0fb21e58138 100644
--- a/gcc/tree-ssa-forwprop.cc
+++ b/gcc/tree-ssa-forwprop.cc
@@ -47,6 +47,8 @@ along with GCC; see the file COPYING3.  If not see
 #include "tree-cfgcleanup.h"
 #include "cfganal.h"
 #include "optabs-tree.h"
+#include "insn-config.h"
+#include "recog.h"
 #include "tree-vector-builder.h"
 #include "vec-perm-indices.h"
 #include "internal-fn.h"
@@ -2978,6 +2980,7 @@ simplify_vector_constructor (gimple_stmt_iterator *gsi)
 	  /* Only few targets implement direct conversion patterns so try
 	     some simple special cases via VEC_[UN]PACK[_FLOAT]_LO_EXPR.  */
 	  optab optab;
+	  insn_code icode;
 	  tree halfvectype, dblvectype;
 	  enum tree_code unpack_op;
 
@@ -3015,8 +3018,9 @@ simplify_vector_constructor (gimple_stmt_iterator *gsi)
 	      && (optab = optab_for_tree_code (unpack_op,
 					       dblvectype,
 					       optab_default))
-	      && (optab_handler (optab, TYPE_MODE (dblvectype))
-		  != CODE_FOR_nothing))
+	      && ((icode = optab_handler (optab, TYPE_MODE (dblvectype)))
+		  != CODE_FOR_nothing)
+	      && (insn_data[icode].operand[0].mode == TYPE_MODE (type)))
 	    {
 	      gimple_seq stmts = NULL;
 	      tree dbl;
@@ -3054,8 +3058,9 @@ simplify_vector_constructor (gimple_stmt_iterator *gsi)
 		   && (optab = optab_for_tree_code (VEC_PACK_TRUNC_EXPR,
 						    halfvectype,
 						    optab_default))
-		   && (optab_handler (optab, TYPE_MODE (halfvectype))
-		       != CODE_FOR_nothing))
+		   && ((icode = optab_handler (optab, TYPE_MODE (halfvectype)))
+		       != CODE_FOR_nothing)
+		   && (insn_data[icode].operand[0].mode == TYPE_MODE (type)))
 	    {
 	      gimple_seq stmts = NULL;
 	      tree low = gimple_build (&stmts, BIT_FIELD_REF, halfvectype,
-- 
2.35.3