From: Jakub Jelinek <jakub@gcc.gnu.org>
Date: Thu, 7 Dec 2017 23:56:34 +0000 (+0100)
Subject: re PR target/81906 (Calls to rint() wrongly optimized away starting in g++ 6)
X-Git-Url: https://git.libre-soc.org/?a=commitdiff_plain;h=7f5aed1e5d7b8f02ed81178f65d84cbd27ec88cd;p=gcc.git

re PR target/81906 (Calls to rint() wrongly optimized away starting in g++ 6)

	PR target/81906
	* config/i386/i386.c (ix86_expand_rint): Handle flag_rounding_math.

	* gcc.target/i386/pr81906.c: New test.

From-SVN: r255486
---

diff --git a/gcc/ChangeLog b/gcc/ChangeLog
index 7cf988956b7..58081c1e46e 100644
--- a/gcc/ChangeLog
+++ b/gcc/ChangeLog
@@ -1,3 +1,10 @@
+2017-12-08  Joseph Myers  <joseph@codesourcery.com>
+	    Alexander Monakov  <amonakov@ispras.ru>
+	    Jakub Jelinek  <jakub@redhat.com>
+
+	PR target/81906
+	* config/i386/i386.c (ix86_expand_rint): Handle flag_rounding_math.
+
 2017-12-07  Richard Sandiford  <richard.sandiford@linaro.org>
 
 	* config/aarch64/aarch64.c (aarch64_print_address_internal): Return
diff --git a/gcc/config/i386/i386.c b/gcc/config/i386/i386.c
index 5d77f28dfa3..e323102cef5 100644
--- a/gcc/config/i386/i386.c
+++ b/gcc/config/i386/i386.c
@@ -44255,8 +44255,7 @@ ix86_expand_lfloorceil (rtx op0, rtx op1, bool do_floor)
   emit_move_insn (op0, ireg);
 }
 
-/* Expand rint (IEEE round to nearest) rounding OPERAND1 and storing the
-   result in OPERAND0.  */
+/* Expand rint rounding OPERAND1 and storing the result in OPERAND0.  */
 void
 ix86_expand_rint (rtx operand0, rtx operand1)
 {
@@ -44264,11 +44263,17 @@ ix86_expand_rint (rtx operand0, rtx operand1)
 	xa = fabs (operand1);
         if (!isless (xa, 2**52))
 	  return operand1;
-        xa = xa + 2**52 - 2**52;
+        two52 = 2**52;
+        if (flag_rounding_math)
+	  {
+	    two52 = copysign (two52, operand1);
+	    xa = operand1;
+	  }
+        xa = xa + two52 - two52;
         return copysign (xa, operand1);
    */
   machine_mode mode = GET_MODE (operand0);
-  rtx res, xa, TWO52, mask;
+  rtx res, xa, TWO52, two52, mask;
   rtx_code_label *label;
 
   res = gen_reg_rtx (mode);
@@ -44281,8 +44286,16 @@ ix86_expand_rint (rtx operand0, rtx operand1)
   TWO52 = ix86_gen_TWO52 (mode);
   label = ix86_expand_sse_compare_and_jump (UNLE, TWO52, xa, false);
 
-  xa = expand_simple_binop (mode, PLUS, xa, TWO52, NULL_RTX, 0, OPTAB_DIRECT);
-  xa = expand_simple_binop (mode, MINUS, xa, TWO52, xa, 0, OPTAB_DIRECT);
+  two52 = TWO52;
+  if (flag_rounding_math)
+    {
+      two52 = gen_reg_rtx (mode);
+      ix86_sse_copysign_to_positive (two52, TWO52, res, mask);
+      xa = res;
+    }
+
+  xa = expand_simple_binop (mode, PLUS, xa, two52, NULL_RTX, 0, OPTAB_DIRECT);
+  xa = expand_simple_binop (mode, MINUS, xa, two52, xa, 0, OPTAB_DIRECT);
 
   ix86_sse_copysign_to_positive (res, xa, res, mask);
 
diff --git a/gcc/testsuite/ChangeLog b/gcc/testsuite/ChangeLog
index d4cb0d20fdd..d12ca43c7a1 100644
--- a/gcc/testsuite/ChangeLog
+++ b/gcc/testsuite/ChangeLog
@@ -1,5 +1,8 @@
 2017-12-08  Jakub Jelinek  <jakub@redhat.com>
 
+	PR target/81906
+	* gcc.target/i386/pr81906.c: New test.
+
 	PR tree-optimization/83075
 	* gcc.dg/tree-ssa/strncpy-2.c: Use size_t instead of unsigned, add
 	separate function with noipa attribute to also verify behavior when
diff --git a/gcc/testsuite/gcc.target/i386/pr81906.c b/gcc/testsuite/gcc.target/i386/pr81906.c
new file mode 100644
index 00000000000..6c71bec41c0
--- /dev/null
+++ b/gcc/testsuite/gcc.target/i386/pr81906.c
@@ -0,0 +1,37 @@
+/* PR target/81906 */
+/* { dg-do run { target *-*-linux* *-*-gnu* } }
+/* { dg-options "-O2 -frounding-math" } */
+
+#include <fenv.h>
+
+int
+main ()
+{
+  #define N 12
+  double a[N] = { 2.0, 2.25, 2.5, 2.75, 3.5, -2.0, -2.25, -2.5, -2.75, -3.5, 0x2.0p53, -0x2.0p53 };
+  double b[N], c[N], d[N], e[N];
+  double be[N] = { 2.0, 2.0, 2.0, 3.0, 4.0, -2.0, -2.0, -2.0, -3.0, -4.0, 0x2.0p53, -0x2.0p53 };
+  double ce[N] = { 2.0, 2.0, 2.0, 2.0, 3.0, -2.0, -3.0, -3.0, -3.0, -4.0, 0x2.0p53, -0x2.0p53 };
+  double de[N] = { 2.0, 3.0, 3.0, 3.0, 4.0, -2.0, -2.0, -2.0, -2.0, -3.0, 0x2.0p53, -0x2.0p53 };
+  double ee[N] = { 2.0, 2.0, 2.0, 2.0, 3.0, -2.0, -2.0, -2.0, -2.0, -3.0, 0x2.0p53, -0x2.0p53 };
+  asm volatile ("" : : "g" (a), "g" (be), "g" (ce), "g" (de), "g" (ee) : "memory");
+
+  int i;
+  fesetround (FE_TONEAREST);
+  for (i = 0; i < N; ++i)
+    b[i] = __builtin_rint (a[i]);
+  fesetround (FE_DOWNWARD);
+  for (i = 0; i < N; ++i)
+    c[i] = __builtin_rint (a[i]);
+  fesetround (FE_UPWARD);
+  for (i = 0; i < N; ++i)
+    d[i] = __builtin_rint (a[i]);
+  fesetround (FE_TOWARDZERO);
+  for (i = 0; i < N; ++i)
+    e[i] = __builtin_rint (a[i]);
+  fesetround (FE_TONEAREST);
+  for (i = 0; i < N; ++i)
+    if (b[i] != be[i] || c[i] != ce[i] || d[i] != de[i] || e[i] != ee[i])
+      __builtin_abort ();
+  return 0;
+}