Re: [PATCH 5/8] vect: Use inbranch simdclones in masked loops

[Richard Biener <rguenther@suse.de>]

[-- Attachment #1: Type: text/plain, Size: 1646 bytes --]

On Wed, 18 Oct 2023, Andre Vieira (lists) wrote:

> Rebased, needs review.

+      tree parm_type = NULL_TREE;
+      if(i < args.length())
+       {

space before (

+/* Return SSA name of the result of the conversion of OPERAND into type 
TYPE.
+   The conversion statement is inserted at GSI.  */
+ 
+static tree                
+vect_convert (vec_info *vinfo, stmt_vec_info stmt_info, tree type, tree 
operand,
+             gimple_stmt_iterator *gsi)                 
+{                                  
+  operand = build1 (VIEW_CONVERT_EXPR, type, operand);
+  gassign *new_stmt = gimple_build_assign (make_ssa_name (type),
+                                          operand);

I don't like this much, it's got one use in your patch only.  Please
leave this abstraction out.

OK with the above two changes.

Thanks,
Richard.

> On 30/08/2023 10:13, Andre Vieira (lists) via Gcc-patches wrote:
> > This patch enables the compiler to use inbranch simdclones when generating
> > masked loops in autovectorization.
> > 
> > gcc/ChangeLog:
> > 
> >      * omp-simd-clone.cc (simd_clone_adjust_argument_types): Make function
> >      compatible with mask parameters in clone.
> >      * tree-vect-stmts.cc (vect_convert): New helper function.
> >      (vect_build_all_ones_mask): Allow vector boolean typed masks.
> >      (vectorizable_simd_clone_call): Enable the use of masked clones in
> >      fully masked loops.
> 

-- 
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)