[PATCH]middle-end: prevent LIM from hoising vector compares from gconds if target does not support it.

[PATCH]middle-end: prevent LIM from hoising vector compares from gconds if target does not support it.

[Tamar Christina]

[-- Attachment #1: Type: text/plain, Size: 2153 bytes --]

Hi All,

LIM notices that in some cases the condition and the results are loop
invariant and tries to move them out of the loop.

While the resulting code is operationally sound, moving the compare out of the
gcond results in generating code that no longer branches, so cbranch is no
longer applicable.  As such I now add code to check during this motion to see
if the target supports flag setting vector comparison as general operation.

I have tried writing a GIMPLE testcase for this but the gimple FE seems to be
having some trouble with the vector types.  It seems to fail parsing.

The early break code testsuite however has a test for this
(vect-early-break_67.c).

Bootstrapped Regtested on aarch64-none-linux-gnu and no issues.

Ok for master?

Thanks,
Tamar

gcc/ChangeLog:

	* tree-ssa-loop-im.cc (determine_max_movement): Import insn-codes.h
	and optabs-tree.h and check for vector compare motion out of gcond.

--- inline copy of patch -- 
diff --git a/gcc/tree-ssa-loop-im.cc b/gcc/tree-ssa-loop-im.cc
index 396963b6754c7671e2e5404302a69129918555e2..2ebf6d6548c4858fd5a8b4f9ab6f332f3fe8f6cd 100644
--- a/gcc/tree-ssa-loop-im.cc
+++ b/gcc/tree-ssa-loop-im.cc
@@ -48,6 +48,8 @@ along with GCC; see the file COPYING3.  If not see
 #include "tree-dfa.h"
 #include "tree-ssa.h"
 #include "dbgcnt.h"
+#include "insn-codes.h"
+#include "optabs-tree.h"
 
 /* TODO:  Support for predicated code motion.  I.e.
 
@@ -852,6 +854,17 @@ determine_max_movement (gimple *stmt, bool must_preserve_exec)
 	  if (!extract_true_false_args_from_phi (dom, phi, NULL, NULL))
 	    return false;
 
+	/* Check if one of the depedent statement is a vector compare whether
+	   the target supports it,  otherwise it's invalid to hoist it out of
+	   the gcond it belonged to.  */
+	if (VECTOR_TYPE_P (TREE_TYPE (gimple_cond_lhs (cond))))
+	  {
+	    tree type = TREE_TYPE (gimple_cond_lhs (cond));
+	    auto code = gimple_cond_code (cond);
+	    if (!target_supports_op_p (type, code, optab_vector))
+	      return false;
+	  }
+
 	  /* Fold in dependencies and cost of the condition.  */
 	  FOR_EACH_SSA_TREE_OPERAND (val, cond, iter, SSA_OP_USE)
 	    {




-- 

[-- Attachment #2: rb18033.patch --]
[-- Type: text/plain, Size: 1195 bytes --]

diff --git a/gcc/tree-ssa-loop-im.cc b/gcc/tree-ssa-loop-im.cc
index 396963b6754c7671e2e5404302a69129918555e2..2ebf6d6548c4858fd5a8b4f9ab6f332f3fe8f6cd 100644
--- a/gcc/tree-ssa-loop-im.cc
+++ b/gcc/tree-ssa-loop-im.cc
@@ -48,6 +48,8 @@ along with GCC; see the file COPYING3.  If not see
 #include "tree-dfa.h"
 #include "tree-ssa.h"
 #include "dbgcnt.h"
+#include "insn-codes.h"
+#include "optabs-tree.h"
 
 /* TODO:  Support for predicated code motion.  I.e.
 
@@ -852,6 +854,17 @@ determine_max_movement (gimple *stmt, bool must_preserve_exec)
 	  if (!extract_true_false_args_from_phi (dom, phi, NULL, NULL))
 	    return false;
 
+	/* Check if one of the depedent statement is a vector compare whether
+	   the target supports it,  otherwise it's invalid to hoist it out of
+	   the gcond it belonged to.  */
+	if (VECTOR_TYPE_P (TREE_TYPE (gimple_cond_lhs (cond))))
+	  {
+	    tree type = TREE_TYPE (gimple_cond_lhs (cond));
+	    auto code = gimple_cond_code (cond);
+	    if (!target_supports_op_p (type, code, optab_vector))
+	      return false;
+	  }
+
 	  /* Fold in dependencies and cost of the condition.  */
 	  FOR_EACH_SSA_TREE_OPERAND (val, cond, iter, SSA_OP_USE)
 	    {

Re: [PATCH]middle-end: prevent LIM from hoising vector compares from gconds if target does not support it.

[Richard Biener]

On Mon, 27 Nov 2023, Tamar Christina wrote:

> Hi All,
> 
> LIM notices that in some cases the condition and the results are loop
> invariant and tries to move them out of the loop.
> 
> While the resulting code is operationally sound, moving the compare out of the
> gcond results in generating code that no longer branches, so cbranch is no
> longer applicable.  As such I now add code to check during this motion to see
> if the target supports flag setting vector comparison as general operation.
> 
> I have tried writing a GIMPLE testcase for this but the gimple FE seems to be
> having some trouble with the vector types.  It seems to fail parsing.
> 
> The early break code testsuite however has a test for this
> (vect-early-break_67.c).
> 
> Bootstrapped Regtested on aarch64-none-linux-gnu and no issues.
> 
> Ok for master?

OK.

Thanks,
Richard.

> Thanks,
> Tamar
> 
> gcc/ChangeLog:
> 
> 	* tree-ssa-loop-im.cc (determine_max_movement): Import insn-codes.h
> 	and optabs-tree.h and check for vector compare motion out of gcond.
> 
> --- inline copy of patch -- 
> diff --git a/gcc/tree-ssa-loop-im.cc b/gcc/tree-ssa-loop-im.cc
> index 396963b6754c7671e2e5404302a69129918555e2..2ebf6d6548c4858fd5a8b4f9ab6f332f3fe8f6cd 100644
> --- a/gcc/tree-ssa-loop-im.cc
> +++ b/gcc/tree-ssa-loop-im.cc
> @@ -48,6 +48,8 @@ along with GCC; see the file COPYING3.  If not see
>  #include "tree-dfa.h"
>  #include "tree-ssa.h"
>  #include "dbgcnt.h"
> +#include "insn-codes.h"
> +#include "optabs-tree.h"
>  
>  /* TODO:  Support for predicated code motion.  I.e.
>  
> @@ -852,6 +854,17 @@ determine_max_movement (gimple *stmt, bool must_preserve_exec)
>  	  if (!extract_true_false_args_from_phi (dom, phi, NULL, NULL))
>  	    return false;
>  
> +	/* Check if one of the depedent statement is a vector compare whether
> +	   the target supports it,  otherwise it's invalid to hoist it out of
> +	   the gcond it belonged to.  */
> +	if (VECTOR_TYPE_P (TREE_TYPE (gimple_cond_lhs (cond))))
> +	  {
> +	    tree type = TREE_TYPE (gimple_cond_lhs (cond));
> +	    auto code = gimple_cond_code (cond);
> +	    if (!target_supports_op_p (type, code, optab_vector))
> +	      return false;
> +	  }
> +
>  	  /* Fold in dependencies and cost of the condition.  */
>  	  FOR_EACH_SSA_TREE_OPERAND (val, cond, iter, SSA_OP_USE)
>  	    {
> 
> 
> 
> 
> 

-- 
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)