public inbox for gcc-patches@gcc.gnu.org
 help / color / mirror / Atom feed
From: "Kewen.Lin" <linkw@linux.ibm.com>
To: "juzhe.zhong@rivai.ai" <juzhe.zhong@rivai.ai>
Cc: Robin Dapp <rdapp.gcc@gmail.com>,
	"richard.sandiford" <richard.sandiford@arm.com>,
	rguenther <rguenther@suse.de>,
	GCC Patches <gcc-patches@gcc.gnu.org>
Subject: Re: [PATCH V4] VECT: Support loop len control on EXTRACT_LAST vectorization
Date: Mon, 14 Aug 2023 20:45:21 +0800	[thread overview]
Message-ID: <b8288bf5-e66d-0ab9-cb85-305243f86fda@linux.ibm.com> (raw)
In-Reply-To: <AAA6C829D41AF904+2023081420082138571818@rivai.ai>

Hi Juzhe,

on 2023/8/14 20:08, juzhe.zhong@rivai.ai wrote:
> Hi, Kewin.
> 
> Could you test 'can_vec_extract_var_idx_p' and send V5 patch when you pass the testing?

The below diff was bootstrapped and regress-tested on Power10 LE.  Comparing to the
previous v4, the only changes should be the proposed can_vec_extract_var_idx_p and
its required new includes as below:

+#include "memmodel.h"
+#include "optabs.h"
 
Could you have a double check?

Since I just tested it on Power10, you have the full ownership on the patch, I'd leave
the v5 posting to you.  Thanks!

BR,
Kewen
-----
diff --git a/gcc/tree-vect-loop.cc b/gcc/tree-vect-loop.cc
index bc3063c3615..5ae9f69c7eb 100644
--- a/gcc/tree-vect-loop.cc
+++ b/gcc/tree-vect-loop.cc
@@ -32,6 +32,8 @@ along with GCC; see the file COPYING3.  If not see
 #include "tree-pass.h"
 #include "ssa.h"
 #include "optabs-tree.h"
+#include "memmodel.h"
+#include "optabs.h"
 #include "diagnostic-core.h"
 #include "fold-const.h"
 #include "stor-layout.h"
@@ -10300,17 +10302,7 @@ vectorizable_live_operation (vec_info *vinfo, stmt_vec_info stmt_info,
       /* No transformation required.  */
       if (loop_vinfo && LOOP_VINFO_CAN_USE_PARTIAL_VECTORS_P (loop_vinfo))
 	{
-	  if (!direct_internal_fn_supported_p (IFN_EXTRACT_LAST, vectype,
-					       OPTIMIZE_FOR_SPEED))
-	    {
-	      if (dump_enabled_p ())
-		dump_printf_loc (MSG_MISSED_OPTIMIZATION, vect_location,
-				 "can't operate on partial vectors "
-				 "because the target doesn't support extract "
-				 "last reduction.\n");
-	      LOOP_VINFO_CAN_USE_PARTIAL_VECTORS_P (loop_vinfo) = false;
-	    }
-	  else if (slp_node)
+	  if (slp_node)
 	    {
 	      if (dump_enabled_p ())
 		dump_printf_loc (MSG_MISSED_OPTIMIZATION, vect_location,
@@ -10330,9 +10322,26 @@ vectorizable_live_operation (vec_info *vinfo, stmt_vec_info stmt_info,
 	  else
 	    {
 	      gcc_assert (ncopies == 1 && !slp_node);
-	      vect_record_loop_mask (loop_vinfo,
-				     &LOOP_VINFO_MASKS (loop_vinfo),
-				     1, vectype, NULL);
+	      if (direct_internal_fn_supported_p (IFN_EXTRACT_LAST, vectype,
+						  OPTIMIZE_FOR_SPEED))
+		vect_record_loop_mask (loop_vinfo,
+				       &LOOP_VINFO_MASKS (loop_vinfo),
+				       1, vectype, NULL);
+	      else if (can_vec_extract_var_idx_p (
+			 TYPE_MODE (vectype), TYPE_MODE (TREE_TYPE (vectype))))
+		vect_record_loop_len (loop_vinfo,
+				      &LOOP_VINFO_LENS (loop_vinfo),
+				      1, vectype, 1);
+	      else
+		{
+		  if (dump_enabled_p ())
+		    dump_printf_loc (
+		      MSG_MISSED_OPTIMIZATION, vect_location,
+		      "can't operate on partial vectors "
+		      "because the target doesn't support extract "
+		      "last reduction.\n");
+		  LOOP_VINFO_CAN_USE_PARTIAL_VECTORS_P (loop_vinfo) = false;
+		}
 	    }
 	}
       /* ???  Enable for loop costing as well.  */
@@ -10358,7 +10367,9 @@ vectorizable_live_operation (vec_info *vinfo, stmt_vec_info stmt_info,
   gimple *vec_stmt;
   if (slp_node)
     {
-      gcc_assert (!loop_vinfo || !LOOP_VINFO_FULLY_MASKED_P (loop_vinfo));
+      gcc_assert (!loop_vinfo
+		  || (!LOOP_VINFO_FULLY_MASKED_P (loop_vinfo)
+		      && !LOOP_VINFO_FULLY_WITH_LENGTH_P (loop_vinfo)));

       /* Get the correct slp vectorized stmt.  */
       vec_lhs = SLP_TREE_VEC_DEFS (slp_node)[vec_entry];
@@ -10402,7 +10413,42 @@ vectorizable_live_operation (vec_info *vinfo, stmt_vec_info stmt_info,

       gimple_seq stmts = NULL;
       tree new_tree;
-      if (LOOP_VINFO_FULLY_MASKED_P (loop_vinfo))
+      if (LOOP_VINFO_FULLY_WITH_LENGTH_P (loop_vinfo))
+	{
+	  /* Emit:
+
+	       SCALAR_RES = VEC_EXTRACT <VEC_LHS, LEN + BIAS - 1>
+
+	     where VEC_LHS is the vectorized live-out result and MASK is
+	     the loop mask for the final iteration.  */
+	  gcc_assert (ncopies == 1 && !slp_node);
+	  gimple_seq tem = NULL;
+	  gimple_stmt_iterator gsi = gsi_last (tem);
+	  tree len
+	    = vect_get_loop_len (loop_vinfo, &gsi,
+				 &LOOP_VINFO_LENS (loop_vinfo),
+				 1, vectype, 0, 0);
+
+	  /* BIAS - 1.  */
+	  signed char biasval = LOOP_VINFO_PARTIAL_LOAD_STORE_BIAS (loop_vinfo);
+	  tree bias_minus_one
+	    = int_const_binop (MINUS_EXPR,
+			       build_int_cst (TREE_TYPE (len), biasval),
+			       build_one_cst (TREE_TYPE (len)));
+
+	  /* LAST_INDEX = LEN + (BIAS - 1).  */
+	  tree last_index = gimple_build (&stmts, PLUS_EXPR, TREE_TYPE (len),
+					  len, bias_minus_one);
+
+	  /* SCALAR_RES = VEC_EXTRACT <VEC_LHS, LEN + BIAS - 1>.  */
+	  tree scalar_res
+	    = gimple_build (&stmts, CFN_VEC_EXTRACT, TREE_TYPE (vectype),
+			    vec_lhs_phi, last_index);
+
+	  /* Convert the extracted vector element to the scalar type.  */
+	  new_tree = gimple_convert (&stmts, lhs_type, scalar_res);
+	}
+      else if (LOOP_VINFO_FULLY_MASKED_P (loop_vinfo))
 	{
 	  /* Emit:

  reply	other threads:[~2023-08-14 12:45 UTC|newest]

Thread overview: 19+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2023-08-11 14:24 juzhe.zhong
2023-08-14  6:53 ` Richard Biener
2023-08-14  7:09   ` juzhe.zhong
2023-08-14  8:51     ` Kewen.Lin
2023-08-14  8:58       ` Robin Dapp
2023-08-14  9:15         ` Richard Biener
2023-08-14  9:19         ` Kewen.Lin
2023-08-14  9:26           ` juzhe.zhong
2023-08-14 12:07             ` Richard Biener
2023-08-14 12:08           ` juzhe.zhong
2023-08-14 12:45             ` Kewen.Lin [this message]
2023-08-14 13:28               ` 钟居哲
2023-08-14 13:44                 ` Richard Biener
2023-08-14 18:51               ` Stefan Schulze Frielinghaus
2023-08-15  6:25                 ` Kewen.Lin
2023-08-15  6:58                   ` Richard Biener
2023-08-15  8:52                     ` Richard Sandiford
2023-08-15  8:55                       ` Richard Biener
2023-08-15  9:05                         ` Richard Sandiford

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=b8288bf5-e66d-0ab9-cb85-305243f86fda@linux.ibm.com \
    --to=linkw@linux.ibm.com \
    --cc=gcc-patches@gcc.gnu.org \
    --cc=juzhe.zhong@rivai.ai \
    --cc=rdapp.gcc@gmail.com \
    --cc=rguenther@suse.de \
    --cc=richard.sandiford@arm.com \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for read-only IMAP folder(s) and NNTP newsgroup(s).