[PATCH] Fix x86 vinsert[if]{32x4,64x2} patterns (PR target/85177)

[Jakub Jelinek]

Hi!

The following testcase is miscompiled, because for vinserti32x4 the expander
creates an incorrect VEC_MERGE selector; it matches what the define_insn
later does, so often it works properly, but when the middle-end sees the
RTL, it can try to simplify it (in this case fwprop1, but in theory also
e.g. combiner when using simplify-rtx etc.).
The bits in the VEC_MERGE selector are ordered lowest bit determines lowest
vector element (from which of the first two VEC_MERGE operand it is taken),
but the define_expand and define_insn were instead counting them from
highest bit.

Fixed thusly, bootstrapped/regtested on x86_64-linux and i686-linux, ok for
trunk?

2018-04-03  Jakub Jelinek  <jakub@redhat.com>

	PR target/85177
	* config/i386/sse.md
	(<extract_type>_vinsert<shuffletype><extract_suf>_mask): Fix
	computation of the VEC_MERGE selector from mask.
	(<extract_type>_vinsert<shuffletype><extract_suf>_1<mask_name>):
	Fix decoding of the VEC_MERGE selector into mask.

	* gcc.target/i386/avx512f-pr85177.c: New test.

--- gcc/config/i386/sse.md.jj	2018-04-01 08:03:37.771565040 +0200
+++ gcc/config/i386/sse.md	2018-04-03 16:41:23.001090852 +0200
@@ -12627,11 +12627,11 @@ (define_expand "<extract_type>_vinsert<s
    (match_operand:<avx512fmaskmode> 5 "register_operand")]
   "TARGET_AVX512F"
 {
-  int mask,selector;
+  int mask, selector;
   mask = INTVAL (operands[3]);
-  selector = GET_MODE_UNIT_SIZE (<MODE>mode) == 4 ?
-    0xFFFF ^ (0xF000 >> mask * 4)
-    : 0xFF ^ (0xC0 >> mask * 2);
+  selector = (GET_MODE_UNIT_SIZE (<MODE>mode) == 4
+  	      ? 0xFFFF ^ (0x000F << mask * 4)
+	      : 0xFF ^ (0x03 << mask * 2));
   emit_insn (gen_<extract_type>_vinsert<shuffletype><extract_suf>_1_mask
     (operands[0], operands[1], operands[2], GEN_INT (selector),
      operands[4], operands[5]));
@@ -12650,16 +12650,16 @@ (define_insn "<mask_codefor><extract_typ
   int mask;
   int selector = INTVAL (operands[3]);
 
-  if (selector == 0xFFF || selector == 0x3F)
+  if (selector == (GET_MODE_UNIT_SIZE (<MODE>mode) == 4 ? 0xFFF0 : 0xFC))
     mask = 0;
-  else if ( selector == 0xF0FF || selector == 0xCF)
+  else if (selector == (GET_MODE_UNIT_SIZE (<MODE>mode) == 4 ? 0xFF0F : 0xF3))
     mask = 1;
-  else if ( selector == 0xFF0F || selector == 0xF3)
+  else if (selector == (GET_MODE_UNIT_SIZE (<MODE>mode) == 4 ? 0xF0FF : 0xCF))
     mask = 2;
-  else if ( selector == 0xFFF0 || selector == 0xFC)
+  else if (selector == (GET_MODE_UNIT_SIZE (<MODE>mode) == 4 ? 0x0FFF : 0x3F))
     mask = 3;
   else
-      gcc_unreachable ();
+    gcc_unreachable ();
 
   operands[3] = GEN_INT (mask);
 
--- gcc/testsuite/gcc.target/i386/avx512f-pr85177.c.jj	2018-04-03 17:05:54.166498062 +0200
+++ gcc/testsuite/gcc.target/i386/avx512f-pr85177.c	2018-04-03 17:06:16.695503852 +0200
@@ -0,0 +1,30 @@
+/* PR target/85177 */
+/* { dg-do run { target { avx512f && int128 } } } */
+/* { dg-options "-O -fno-tree-ccp -fno-tree-sra -mavx512f -mno-avx512bw" } */
+
+#include "avx512f-check.h"
+
+typedef short U __attribute__ ((vector_size (64)));
+typedef __int128 V __attribute__ ((vector_size (64)));
+
+static inline __attribute__((always_inline)) U
+foo (int i, U u)
+{
+  u[i & 1] = 1;
+  return u;
+}
+
+__attribute__((noipa)) int
+bar ()
+{
+  V x = (V) foo (0, (U) { });
+  for (unsigned i = 0; i < 4; i++)
+    if (x[i] != (i == 0)) __builtin_abort ();
+  return 0;
+}
+
+static void
+avx512f_test (void)
+{
+  bar ();
+}

	Jakub