PR target/70300

	* config/i386/i386.md (cvtsd2ss splitter): Unpack in destination
	instead of source if operands[1] is xmm16 and above and
	!TARGET_AVX512VL.  Use avx512f_vec_dupv16sf_1 instead of
	vec_interleave_lowv4sf if we need to unpack xmm16 and above.

	* gcc.target/i386/pr70300.c: New test.


git-svn-id: svn+ssh://gcc.gnu.org/svn/gcc/trunk@234393 138bc75d-0d04-0410-961f-82ee72b054a4

4 files changed, 50 insertions, 5 deletions

diff --git a/gcc/ChangeLog b/gcc/ChangeLog
index 5193c97ecb7..f5687a507e5 100644
--- a/gcc/ChangeLog
+++ b/gcc/ChangeLog
@@ -1,5 +1,11 @@
 2016-03-22  Jakub Jelinek  <jakub@redhat.com>
 
+	PR target/70300
+	* config/i386/i386.md (cvtsd2ss splitter): Unpack in destination
+	instead of source if operands[1] is xmm16 and above and
+	!TARGET_AVX512VL.  Use avx512f_vec_dupv16sf_1 instead of
+	vec_interleave_lowv4sf if we need to unpack xmm16 and above.
+
 	PR c++/70295
 	* gimplify.c (gimplify_modify_expr): Call gimple_set_no_warning
 	on assign if (*from_p) is a comparison, set it to
diff --git a/gcc/config/i386/i386.md b/gcc/config/i386/i386.md
index 90fec1b2894..dcaef843900 100644
--- a/gcc/config/i386/i386.md
+++ b/gcc/config/i386/i386.md
@@ -4229,17 +4229,28 @@
     {
       /* If it is unsafe to overwrite upper half of source, we need
 	 to move to destination and unpack there.  */
-      if ((ORIGINAL_REGNO (operands[1]) < FIRST_PSEUDO_REGISTER
-	   || PSEUDO_REGNO_BYTES (ORIGINAL_REGNO (operands[1])) > 4)
-	  && true_regnum (operands[0]) != true_regnum (operands[1]))
+      if (((ORIGINAL_REGNO (operands[1]) < FIRST_PSEUDO_REGISTER
+	    || PSEUDO_REGNO_BYTES (ORIGINAL_REGNO (operands[1])) > 4)
+	   && true_regnum (operands[0]) != true_regnum (operands[1]))
+	  || (EXT_REX_SSE_REG_P (operands[1])
+	      && !TARGET_AVX512VL))
 	{
 	  rtx tmp = gen_rtx_REG (SFmode, true_regnum (operands[0]));
 	  emit_move_insn (tmp, operands[1]);
 	}
       else
 	operands[3] = simplify_gen_subreg (V4SFmode, operands[1], SFmode, 0);
-      emit_insn (gen_vec_interleave_lowv4sf (operands[3], operands[3],
-      		 			     operands[3]));
+      /* FIXME: vec_interleave_lowv4sf for AVX512VL should allow
+	 =v, v, then vbroadcastss will be only needed for AVX512F without
+	 AVX512VL.  */
+      if (!EXT_REX_SSE_REGNO_P (true_regnum (operands[3])))
+	emit_insn (gen_vec_interleave_lowv4sf (operands[3], operands[3],
+					       operands[3]));
+      else
+	{
+	  rtx tmp = simplify_gen_subreg (V16SFmode, operands[3], V4SFmode, 0);
+	  emit_insn (gen_avx512f_vec_dupv16sf_1 (tmp, tmp));
+	}
     }
   else
     emit_insn (gen_vec_setv4sf_0 (operands[3],
diff --git a/gcc/testsuite/ChangeLog b/gcc/testsuite/ChangeLog
index 25a36ddbc20..aad5d3f241e 100644
--- a/gcc/testsuite/ChangeLog
+++ b/gcc/testsuite/ChangeLog
@@ -1,5 +1,8 @@
 2016-03-22  Jakub Jelinek  <jakub@redhat.com>
 
+	PR target/70300
+	* gcc.target/i386/pr70300.c: New test.
+
 	PR c++/70295
 	* c-c++-common/nonnull-1.c (func): Remove parens around cp4 != 0.
 	(func2): New function for cond with parens, xfail warning for c++.
diff --git a/gcc/testsuite/gcc.target/i386/pr70300.c b/gcc/testsuite/gcc.target/i386/pr70300.c
new file mode 100644
index 00000000000..ddfadfb5957
--- /dev/null
+++ b/gcc/testsuite/gcc.target/i386/pr70300.c
@@ -0,0 +1,25 @@
+/* PR target/70300 */
+/* { dg-do compile } */
+/* { dg-options "-O2 -mtune=amdfam10 -mavx512f" } */
+
+typedef _Complex A __attribute__ ((mode (SC)));
+typedef _Complex B __attribute__ ((mode (DC)));
+typedef _Complex C __attribute__ ((mode (TC)));
+
+C
+foo (A a, B b, C c, A d, B e, C f)
+{
+  b -= a;
+  d += a;
+  a += f;
+  return a + b + d + e;
+}
+
+__attribute__((target ("avx512vl"))) C
+bar (A a, B b, C c, A d, B e, C f)
+{
+  b -= a;
+  d += a;
+  a += f;
+  return a + b + d + e;
+}