If the backend doesn't have floating point EQ or NE comparison, dojump.c
splits it into ORDERED && UNEQ or UNORDERED || LTGT. If both comparison
operands are the same, we know the result of the second comparison though,
a == b is equivalent to a ord b and a != b is equivalent to a unord b,
and thus can just use ORDERED or UNORDERED.
On the testcase, this changes f1:
- ucomiss %xmm0, %xmm0
- movl $1, %eax
- jp .L3
- jne .L3
- ret
- .p2align 4,,10
- .p2align 3
-.L3:
xorl %eax, %eax
+ ucomiss %xmm0, %xmm0
+ setnp %al
and f3:
- ucomisd %xmm0, %xmm0
- movl $1, %eax
- jp .L8
- jne .L8
- ret
- .p2align 4,,10
- .p2align 3
-.L8:
xorl %eax, %eax
+ ucomisd %xmm0, %xmm0
+ setnp %al
while keeping the same code for f2 and f4.
2020-12-10 Jakub Jelinek <jakub@redhat.com>
PR tree-optimization/98169
* dojump.c (do_compare_rtx_and_jump): Don't split self-EQ/NE
comparisons, just use ORDERED or UNORDERED.
* gcc.target/i386/pr98169.c: New test.
/* ... or if there is no libcall for it. */
|| code_to_optab (code) == unknown_optab))
{
- enum rtx_code first_code;
+ enum rtx_code first_code, orig_code = code;
bool and_them = split_comparison (code, mode, &first_code, &code);
/* If there are no NaNs, the first comparison should always fall
if (!HONOR_NANS (mode))
gcc_assert (first_code == (and_them ? ORDERED : UNORDERED));
+ else if ((orig_code == EQ || orig_code == NE)
+ && rtx_equal_p (op0, op1))
+ /* Self-comparisons x == x or x != x can be optimized into
+ just x ord x or x nord x. */
+ code = orig_code == EQ ? ORDERED : UNORDERED;
+
else
{
profile_probability cprob
--- /dev/null
+/* PR tree-optimization/98169 */
+/* { dg-do compile } */
+/* { dg-options "-O2 -fno-finite-math-only" } */
+/* { dg-final { scan-assembler-times "\tsetn\?p\t" 4 } } */
+/* { dg-final { scan-assembler-not "\tjn\?\[ep]\t" } } */
+
+int
+f1 (float a)
+{
+ return a == a;
+}
+
+int
+f2 (float a)
+{
+ return !__builtin_isnanf (a);
+}
+
+int
+f3 (double a)
+{
+ return a == a;
+}
+
+int
+f4 (double a)
+{
+ return !__builtin_isnan (a);
+}