public inbox for gcc-patches@gcc.gnu.org
 help / color / mirror / Atom feed
From: Juzhe-Zhong <juzhe.zhong@rivai.ai>
To: gcc-patches@gcc.gnu.org
Cc: richard.sandiford@arm.com, rguenther@suse.de,
	Juzhe-Zhong <juzhe.zhong@rivai.ai>
Subject: [PATCH] Middle-end: Do not model address cost for SELECT_VL style vectorization
Date: Thu, 14 Dec 2023 21:45:59 +0800	[thread overview]
Message-ID: <20231214134559.137710-1-juzhe.zhong@rivai.ai> (raw)

Follow Richard's suggestions, we should not model address cost in the loop
vectorizer for select_vl or decrement IV since other style vectorization doesn't
do that.

To make cost model comparison apple to apple.
This patch set COST from 2 to 1 which turns out have better codegen
in various codegen for RVV.

Ok for trunk ?

	PR target/111153

gcc/ChangeLog:

	* tree-vect-loop.cc (vect_estimate_min_profitable_iters): Remove address cost for select_vl/decrement IV.

gcc/testsuite/ChangeLog:

	* gcc.dg/vect/costmodel/riscv/rvv/pr111153.c: Moved to...
	* gcc.dg/vect/costmodel/riscv/rvv/pr11153-2.c: ...here.
	* gcc.dg/vect/costmodel/riscv/rvv/pr111153-1.c: New test.

---
 .../vect/costmodel/riscv/rvv/pr111153-1.c      | 18 ++++++++++++++++++
 .../riscv/rvv/{pr111153.c => pr11153-2.c}      |  4 ++--
 gcc/tree-vect-loop.cc                          | 10 ++++------
 3 files changed, 24 insertions(+), 8 deletions(-)
 create mode 100644 gcc/testsuite/gcc.dg/vect/costmodel/riscv/rvv/pr111153-1.c
 rename gcc/testsuite/gcc.dg/vect/costmodel/riscv/rvv/{pr111153.c => pr11153-2.c} (93%)

diff --git a/gcc/testsuite/gcc.dg/vect/costmodel/riscv/rvv/pr111153-1.c b/gcc/testsuite/gcc.dg/vect/costmodel/riscv/rvv/pr111153-1.c
new file mode 100644
index 00000000000..51c91f7410c
--- /dev/null
+++ b/gcc/testsuite/gcc.dg/vect/costmodel/riscv/rvv/pr111153-1.c
@@ -0,0 +1,18 @@
+/* { dg-do compile } */
+/* { dg-options "-march=rv64gcv -mabi=lp64d -O3 -ftree-vectorize -mtune=generic-ooo -ffast-math" } */
+
+#define DEF_REDUC_PLUS(TYPE)                                                   \
+  TYPE __attribute__ ((noinline, noclone))                                     \
+  reduc_plus_##TYPE (TYPE *__restrict a, int n)                                \
+  {                                                                            \
+    TYPE r = 0;                                                                \
+    for (int i = 0; i < n; ++i)                                                \
+      r += a[i];                                                               \
+    return r;                                                                  \
+  }
+
+#define TEST_PLUS(T) T (int) T (float)
+
+TEST_PLUS (DEF_REDUC_PLUS)
+
+/* { dg-final { scan-assembler-not {vsetivli\s+zero,\s*4} } } */
diff --git a/gcc/testsuite/gcc.dg/vect/costmodel/riscv/rvv/pr111153.c b/gcc/testsuite/gcc.dg/vect/costmodel/riscv/rvv/pr11153-2.c
similarity index 93%
rename from gcc/testsuite/gcc.dg/vect/costmodel/riscv/rvv/pr111153.c
rename to gcc/testsuite/gcc.dg/vect/costmodel/riscv/rvv/pr11153-2.c
index 06e08ec5f2e..d361f1fc7fa 100644
--- a/gcc/testsuite/gcc.dg/vect/costmodel/riscv/rvv/pr111153.c
+++ b/gcc/testsuite/gcc.dg/vect/costmodel/riscv/rvv/pr11153-2.c
@@ -1,5 +1,5 @@
 /* { dg-do compile } */
-/* { dg-options "-march=rv64gcv -mabi=lp64d -O3 -ftree-vectorize -mtune=generic-ooo" } */
+/* { dg-options "-march=rv64gcv -mabi=lp64d -O3 -ftree-vectorize -ffast-math" } */
 
 #define DEF_REDUC_PLUS(TYPE)                                                   \
   TYPE __attribute__ ((noinline, noclone))                                     \
@@ -11,7 +11,7 @@
     return r;                                                                  \
   }
 
-#define TEST_PLUS(T) T (int)
+#define TEST_PLUS(T) T (int) T (float)
 
 TEST_PLUS (DEF_REDUC_PLUS)
 
diff --git a/gcc/tree-vect-loop.cc b/gcc/tree-vect-loop.cc
index 19e38b8637b..7a3db5f098b 100644
--- a/gcc/tree-vect-loop.cc
+++ b/gcc/tree-vect-loop.cc
@@ -4872,12 +4872,10 @@ vect_estimate_min_profitable_iters (loop_vec_info loop_vinfo,
 
 	    unsigned int length_update_cost = 0;
 	    if (LOOP_VINFO_USING_DECREMENTING_IV_P (loop_vinfo))
-	      /* For decrement IV style, we use a single SELECT_VL since
-		 beginning to calculate the number of elements need to be
-		 processed in current iteration, and a SHIFT operation to
-		 compute the next memory address instead of adding vectorization
-		 factor.  */
-	      length_update_cost = 2;
+	      /* For decrement IV style, Each only need a single SELECT_VL
+		 or MIN since beginning to calculate the number of elements
+		 need to be processed in current iteration.  */
+	      length_update_cost = 1;
 	    else
 	      /* For increment IV stype, Each may need two MINs and one MINUS to
 		 update lengths in body for next iteration.  */
-- 
2.36.1


             reply	other threads:[~2023-12-14 13:46 UTC|newest]

Thread overview: 3+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2023-12-14 13:45 Juzhe-Zhong [this message]
2023-12-14 15:10 ` Richard Biener
2023-12-15  0:51   ` juzhe.zhong

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=20231214134559.137710-1-juzhe.zhong@rivai.ai \
    --to=juzhe.zhong@rivai.ai \
    --cc=gcc-patches@gcc.gnu.org \
    --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).