[PATCH] Output vextract{i, f}{32x4, 64x2} for (vec_select:(reg:Vmode) idx) when byte_offset of idx % 16 == 0.

[liuhongt via Gcc-patches]

Hi:
  As describled in PR, use vextract instead on valign when
byte_offset % 16 == 0.

  Bootstrapped and regtest on x86_64-linux-gnu{-m32,}.
  Pushed to trunk.

2020-09-13  Hongtao Liu  <hongtao....@intel.com>
            Peter Cordes  <pe...@cordes.ca>
gcc/ChangeLog:

        PR target/91103
        * config/i386/sse.md (extract_suf): Add V8SF/V8SI/V4DF/V4DI.
        (*vec_extract<mode><ssescalarmodelower>_valign): Output
        vextract{i,f}{32x4,64x2} instruction when byte_offset % 16 ==
        0.

gcc/testsuite/ChangeLog:

        PR target/91103
        * gcc.target/i386/pr91103-1.c: Add extract tests.
        * gcc.target/i386/pr91103-2.c: Ditto.
---
 gcc/config/i386/sse.md                    | 21 +++++++++++++++++----
 gcc/testsuite/gcc.target/i386/pr91103-1.c |  7 ++++++-
 gcc/testsuite/gcc.target/i386/pr91103-2.c |  4 ++++
 3 files changed, 27 insertions(+), 5 deletions(-)

diff --git a/gcc/config/i386/sse.md b/gcc/config/i386/sse.md
index 516eb4544bc..5f96016c947 100644
--- a/gcc/config/i386/sse.md
+++ b/gcc/config/i386/sse.md
@@ -9031,7 +9031,8 @@ (define_mode_attr extract_type
   [(V16SF "avx512f") (V16SI "avx512f") (V8DF "avx512dq") (V8DI "avx512dq")])
 
 (define_mode_attr extract_suf
-  [(V16SF "32x4") (V16SI "32x4") (V8DF "64x2") (V8DI "64x2")])
+  [(V16SF "32x4") (V16SI "32x4") (V8DF "64x2") (V8DI "64x2")
+   (V8SF "32x4") (V8SI "32x4") (V4DF "64x2") (V4DI "64x2")])
 
 (define_mode_iterator AVX512_VEC
   [(V8DF "TARGET_AVX512DQ") (V8DI "TARGET_AVX512DQ") V16SF V16SI])
@@ -10603,9 +10604,21 @@ (define_insn 
"*vec_extract<mode><ssescalarmodelower>_valign"
          (match_operand:V48_256_512_AVX512VL 1 "register_operand" "v")
          (parallel [(match_operand 2 "<vec_extract_imm_predicate>")])))]
   "TARGET_AVX512F
-   && INTVAL(operands[2]) >= 16 / GET_MODE_SIZE (<ssescalarmode>mode)"
-  "valign<ternlogsuffix>\t{%2, %1, %1, %<xtg_mode>0|%<xtg_mode>0, %1, %1, %2}";
-  [(set_attr "prefix" "evex")
+   && INTVAL(operands[2]) * GET_MODE_SIZE (<ssescalarmode>mode) >= 16"
+{
+  int byte_offset = INTVAL (operands[2]) * GET_MODE_SIZE (<ssescalarmode>mode);
+  if (byte_offset % 16 == 0)
+    {
+      operands[2] = GEN_INT (byte_offset / 16);
+      if (byte_offset / 16 == 1)
+       return "vextract<shuffletype><extract_suf>\t{%2, %t1, %x0|%x0, %t1, 
%2}";
+      else
+       return "vextract<shuffletype><extract_suf>\t{%2, %1, %x0|%x0, %1, %2}";
+    }
+  else
+    return "valign<ternlogsuffix>\t{%2, %1, %1, %<xtg_mode>0|%<xtg_mode>0, %1, 
%1, %2}";
+}
+  [(set_attr "prefix" "maybe_evex")
    (set_attr "mode" "<sseintvecinsnmode>")])
 
 (define_expand "avx512f_shufps512_mask"
diff --git a/gcc/testsuite/gcc.target/i386/pr91103-1.c 
b/gcc/testsuite/gcc.target/i386/pr91103-1.c
index 11caaa8bd1b..2d78a6da0f3 100644
--- a/gcc/testsuite/gcc.target/i386/pr91103-1.c
+++ b/gcc/testsuite/gcc.target/i386/pr91103-1.c
@@ -1,6 +1,7 @@
 /* { dg-do compile } */
 /* { dg-options "-mavx512vl -O2" } */
-/* { dg-final { scan-assembler-times "valign\[dq\]" 16 } } */
+/* { dg-final { scan-assembler-times "valign\[dq\]" 8 } } */
+/* { dg-final { scan-assembler-times "vextract" 12 } } */
 
 typedef float v8sf __attribute__((vector_size(32)));
 typedef float v16sf __attribute__((vector_size(64)));
@@ -23,9 +24,13 @@ EXTRACT (v8sf, float, 4);
 EXTRACT (v8sf, float, 7);
 EXTRACT (v8si, int, 4);
 EXTRACT (v8si, int, 7);
+EXTRACT (v16sf, float, 4);
 EXTRACT (v16sf, float, 8);
+EXTRACT (v16sf, float, 12);
 EXTRACT (v16sf, float, 15);
+EXTRACT (v16si, int, 4);
 EXTRACT (v16si, int, 8);
+EXTRACT (v16si, int, 12);
 EXTRACT (v16si, int, 15);
 EXTRACT (v4df, double, 2);
 EXTRACT (v4df, double, 3);
diff --git a/gcc/testsuite/gcc.target/i386/pr91103-2.c 
b/gcc/testsuite/gcc.target/i386/pr91103-2.c
index 010e4775723..a928d87f99a 100644
--- a/gcc/testsuite/gcc.target/i386/pr91103-2.c
+++ b/gcc/testsuite/gcc.target/i386/pr91103-2.c
@@ -61,9 +61,13 @@ RUNCHECK (f2, v8sf, float, 4);
 RUNCHECK (f2, v8sf, float, 7);
 RUNCHECK (di2, v8si, int, 4);
 RUNCHECK (di2, v8si, int, 7);
+RUNCHECK (f1, v16sf, float, 4);
 RUNCHECK (f1, v16sf, float, 8);
+RUNCHECK (f1, v16sf, float, 12);
 RUNCHECK (f1, v16sf, float, 15);
+RUNCHECK (di1, v16si, int, 4);
 RUNCHECK (di1, v16si, int, 8);
+RUNCHECK (di1, v16si, int, 12);
 RUNCHECK (di1, v16si, int, 15);
 RUNCHECK (d2, v4df, double, 2);
 RUNCHECK (d2, v4df, double, 3);
-- 
2.27.0