[PATCH] bitint: Handle VCE from large/huge _BitInt SSA_NAME from load [PR114156]

[PATCH] bitint: Handle VCE from large/huge _BitInt SSA_NAME from load [PR114156]

[Jakub Jelinek]

Hi!

When adding checks in which case not to merge a VIEW_CONVERT_EXPR from
large/huge _BitInt to vector/complex etc., I missed the case of loads.
Those are handled differently later.
Anyway, I think the load case is something we can handle just fine,
so the following patch does that instead of preventing the merging
gimple_lower_bitint; we'd then copy from memory to memory and and do the
vce only on the second one, it is just better to vce the first one.

Bootstrapped/regtested on x86_64-linux and i686-linux, ok for trunk?

2024-03-01  Jakub Jelinek  <jakub@redhat.com>

	PR middle-end/114156
	* gimple-lower-bitint.cc (bitint_large_huge::lower_stmt): Allow
	rhs1 of a VCE to have no underlying variable if it is a load and
	handle that case.

	* gcc.dg/bitint-96.c: New test.

--- gcc/gimple-lower-bitint.cc.jj	2024-02-24 12:44:27.993108306 +0100
+++ gcc/gimple-lower-bitint.cc	2024-02-29 19:28:59.442020619 +0100
@@ -5329,6 +5329,22 @@ bitint_large_huge::lower_stmt (gimple *s
 		  gimple_assign_set_rhs1 (stmt, rhs1);
 		  gimple_assign_set_rhs_code (stmt, SSA_NAME);
 		}
+	      else if (m_names == NULL
+		       || !bitmap_bit_p (m_names, SSA_NAME_VERSION (rhs1)))
+		{
+		  gimple *g = SSA_NAME_DEF_STMT (rhs1);
+		  gcc_assert (gimple_assign_load_p (g));
+		  tree mem = gimple_assign_rhs1 (g);
+		  tree ltype = TREE_TYPE (lhs);
+		  addr_space_t as = TYPE_ADDR_SPACE (TREE_TYPE (mem));
+		  if (as != TYPE_ADDR_SPACE (ltype))
+		    ltype
+		      = build_qualified_type (ltype,
+					      TYPE_QUALS (ltype)
+					      | ENCODE_QUAL_ADDR_SPACE (as));
+		  rhs1 = build1 (VIEW_CONVERT_EXPR, ltype, mem);
+		  gimple_assign_set_rhs1 (stmt, rhs1);
+		}
 	      else
 		{
 		  int part = var_to_partition (m_map, rhs1);
--- gcc/testsuite/gcc.dg/bitint-96.c.jj	2024-02-29 19:37:27.441032088 +0100
+++ gcc/testsuite/gcc.dg/bitint-96.c	2024-02-29 19:36:34.815753879 +0100
@@ -0,0 +1,17 @@
+/* PR middle-end/114156 */
+/* { dg-do compile { target bitint } } */
+/* { dg-options "-O2" } */
+/* { dg-additional-options "-msse2" { target i?86-*-* x86_64-*-* } } */
+
+#if __BITINT_MAXWIDTH__ >= 128
+_BitInt(128) a, b;
+#else
+int a, b;
+#endif
+
+void
+foo (void)
+{
+  int u = b;
+  __builtin_memmove (&a, &b, sizeof (a));
+}

	Jakub

Re: [PATCH] bitint: Handle VCE from large/huge _BitInt SSA_NAME from load [PR114156]

[Richard Biener]

On Fri, 1 Mar 2024, Jakub Jelinek wrote:

> Hi!
> 
> When adding checks in which case not to merge a VIEW_CONVERT_EXPR from
> large/huge _BitInt to vector/complex etc., I missed the case of loads.
> Those are handled differently later.
> Anyway, I think the load case is something we can handle just fine,
> so the following patch does that instead of preventing the merging
> gimple_lower_bitint; we'd then copy from memory to memory and and do the
> vce only on the second one, it is just better to vce the first one.
> 
> Bootstrapped/regtested on x86_64-linux and i686-linux, ok for trunk?

OK.

Richard.

> 2024-03-01  Jakub Jelinek  <jakub@redhat.com>
> 
> 	PR middle-end/114156
> 	* gimple-lower-bitint.cc (bitint_large_huge::lower_stmt): Allow
> 	rhs1 of a VCE to have no underlying variable if it is a load and
> 	handle that case.
> 
> 	* gcc.dg/bitint-96.c: New test.
> 
> --- gcc/gimple-lower-bitint.cc.jj	2024-02-24 12:44:27.993108306 +0100
> +++ gcc/gimple-lower-bitint.cc	2024-02-29 19:28:59.442020619 +0100
> @@ -5329,6 +5329,22 @@ bitint_large_huge::lower_stmt (gimple *s
>  		  gimple_assign_set_rhs1 (stmt, rhs1);
>  		  gimple_assign_set_rhs_code (stmt, SSA_NAME);
>  		}
> +	      else if (m_names == NULL
> +		       || !bitmap_bit_p (m_names, SSA_NAME_VERSION (rhs1)))
> +		{
> +		  gimple *g = SSA_NAME_DEF_STMT (rhs1);
> +		  gcc_assert (gimple_assign_load_p (g));
> +		  tree mem = gimple_assign_rhs1 (g);
> +		  tree ltype = TREE_TYPE (lhs);
> +		  addr_space_t as = TYPE_ADDR_SPACE (TREE_TYPE (mem));
> +		  if (as != TYPE_ADDR_SPACE (ltype))
> +		    ltype
> +		      = build_qualified_type (ltype,
> +					      TYPE_QUALS (ltype)
> +					      | ENCODE_QUAL_ADDR_SPACE (as));
> +		  rhs1 = build1 (VIEW_CONVERT_EXPR, ltype, mem);
> +		  gimple_assign_set_rhs1 (stmt, rhs1);
> +		}
>  	      else
>  		{
>  		  int part = var_to_partition (m_map, rhs1);
> --- gcc/testsuite/gcc.dg/bitint-96.c.jj	2024-02-29 19:37:27.441032088 +0100
> +++ gcc/testsuite/gcc.dg/bitint-96.c	2024-02-29 19:36:34.815753879 +0100
> @@ -0,0 +1,17 @@
> +/* PR middle-end/114156 */
> +/* { dg-do compile { target bitint } } */
> +/* { dg-options "-O2" } */
> +/* { dg-additional-options "-msse2" { target i?86-*-* x86_64-*-* } } */
> +
> +#if __BITINT_MAXWIDTH__ >= 128
> +_BitInt(128) a, b;
> +#else
> +int a, b;
> +#endif
> +
> +void
> +foo (void)
> +{
> +  int u = b;
> +  __builtin_memmove (&a, &b, sizeof (a));
> +}
> 
> 	Jakub
> 
> 

-- 
Richard Biener <rguenther@suse.de>
SUSE Software Solutions Germany GmbH,
Frankenstrasse 146, 90461 Nuernberg, Germany;
GF: Ivo Totev, Andrew McDonald, Werner Knoblich; (HRB 36809, AG Nuernberg)