From f04bffb04a4dae3e1a1f87f53c4591fc48bb8058 Mon Sep 17 00:00:00 2001
From: Uros Bizjak <ubizjak@gmail.com>
Date: Fri, 2 Aug 2019 17:46:02 +0200
Subject: [PATCH] re PR tree-optimization/91201 (SIMD not generated for
 horizontal sum of bytes in array)

	PR target/91201
	* config/i386/sse.md (*vec_extractv16qi_zext): New insn pattern.

testsuite/ChangeLog:

	PR target/91201
	* gcc.target/i386/sse4_1-pr91201.c: New test.

From-SVN: r274018
---
 gcc/ChangeLog                                 |  5 +++++
 gcc/config/i386/sse.md                        | 19 +++++++++++++++++++
 gcc/testsuite/ChangeLog                       |  5 +++++
 .../gcc.target/i386/sse4_1-pr91201.c          | 12 ++++++++++++
 4 files changed, 41 insertions(+)
 create mode 100644 gcc/testsuite/gcc.target/i386/sse4_1-pr91201.c

diff --git a/gcc/ChangeLog b/gcc/ChangeLog
index 6c1372efde1..e63e09d258e 100644
--- a/gcc/ChangeLog
+++ b/gcc/ChangeLog
@@ -1,3 +1,8 @@
+2019-08-02  Uroš Bizjak  <ubizjak@gmail.com>
+
+	PR target/91201
+	* config/i386/sse.md (*vec_extractv16qi_zext): New insn pattern.
+
 2019-08-02  Alexander Monakov  <amonakov@ispras.ru>
 
 	* tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Simplify casts
diff --git a/gcc/config/i386/sse.md b/gcc/config/i386/sse.md
index 56a89154e54..3391724fd8d 100644
--- a/gcc/config/i386/sse.md
+++ b/gcc/config/i386/sse.md
@@ -14970,6 +14970,25 @@
    (set_attr "prefix" "maybe_vex")
    (set_attr "mode" "TI")])
 
+(define_insn "*vec_extractv16qi_zext"
+  [(set (match_operand:HI 0 "register_operand" "=r,r")
+	(zero_extend:HI
+	  (vec_select:QI
+	    (match_operand:V16QI 1 "register_operand" "x,v")
+	    (parallel
+	      [(match_operand:SI 2 "const_0_to_15_operand")]))))]
+  "TARGET_SSE4_1"
+  "@
+   %vpextrb\t{%2, %1, %k0|%k0, %1, %2}
+   vpextrb\t{%2, %1, %k0|%k0, %1, %2}"
+  [(set_attr "isa" "*,avx512bw")
+   (set_attr "type" "sselog1")
+   (set_attr "prefix_data16" "1")
+   (set_attr "prefix_extra" "1")
+   (set_attr "length_immediate" "1")
+   (set_attr "prefix" "maybe_vex")
+   (set_attr "mode" "TI")])
+
 (define_insn "*vec_extract<mode>_mem"
   [(set (match_operand:<ssescalarmode> 0 "register_operand" "=r")
 	(vec_select:<ssescalarmode>
diff --git a/gcc/testsuite/ChangeLog b/gcc/testsuite/ChangeLog
index 00d664f961e..45d941d7e6e 100644
--- a/gcc/testsuite/ChangeLog
+++ b/gcc/testsuite/ChangeLog
@@ -1,3 +1,8 @@
+2019-08-02  Uroš Bizjak  <ubizjak@gmail.com>
+
+	PR target/91201
+	* gcc.target/i386/sse4_1-pr91201.c: New test.
+
 2019-08-02  Marek Polacek  <polacek@redhat.com>
 
 	PR c++/91230 - wrong error with __PRETTY_FUNCTION__ and generic lambda.
diff --git a/gcc/testsuite/gcc.target/i386/sse4_1-pr91201.c b/gcc/testsuite/gcc.target/i386/sse4_1-pr91201.c
new file mode 100644
index 00000000000..095d18ef895
--- /dev/null
+++ b/gcc/testsuite/gcc.target/i386/sse4_1-pr91201.c
@@ -0,0 +1,12 @@
+/* PR tree-optimization/91201 */
+/* { dg-do compile } */
+/* { dg-options "-Os -msse4.1 -masm=att" } */
+/* { dg-final { scan-assembler-not "\tmovzb(w|l)" } } */
+
+typedef unsigned char V __attribute__((vector_size (16)));
+
+unsigned short
+foo (V x)
+{
+  return x[0];
+}
-- 
2.30.2