[Bug tree-optimization/85529] [7/8/9 Regression] wrong code at -O2 and -O3 on x86_64-linux-gnu
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=85529 --- Comment #9 from Jakub Jelinek --- Author: jakub Date: Fri Apr 27 07:11:44 2018 New Revision: 259697 URL: https://gcc.gnu.org/viewcvs?rev=259697=gcc=rev Log: PR tree-optimization/85529 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): Add FIRST_BB argument. Don't call get_nonzero_bits if opcode is ERROR_MARK_NODE, rhs2 def stmt's bb is dominated by first_bb and it isn't an obvious zero extension or masking of the MSB bit. (optimize_range_tests): Add FIRST_BB argument, pass it through to optimize_range_tests_var_bound. (maybe_optimize_range_tests, reassociate_bb): Adjust optimize_range_tests callers. * gcc.c-torture/execute/pr85529-1.c: New test. * gcc.c-torture/execute/pr85529-2.c: New test. * gcc.dg/pr85529.c: New test. Added: branches/gcc-8-branch/gcc/testsuite/gcc.c-torture/execute/pr85529-1.c branches/gcc-8-branch/gcc/testsuite/gcc.c-torture/execute/pr85529-2.c branches/gcc-8-branch/gcc/testsuite/gcc.dg/pr85529.c Modified: branches/gcc-8-branch/gcc/ChangeLog branches/gcc-8-branch/gcc/testsuite/ChangeLog branches/gcc-8-branch/gcc/tree-ssa-reassoc.c
[Bug tree-optimization/85529] [7/8/9 Regression] wrong code at -O2 and -O3 on x86_64-linux-gnu
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=85529 --- Comment #8 from Jakub Jelinek --- Author: jakub Date: Fri Apr 27 07:09:51 2018 New Revision: 259696 URL: https://gcc.gnu.org/viewcvs?rev=259696=gcc=rev Log: PR tree-optimization/85529 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): Add FIRST_BB argument. Don't call get_nonzero_bits if opcode is ERROR_MARK_NODE, rhs2 def stmt's bb is dominated by first_bb and it isn't an obvious zero extension or masking of the MSB bit. (optimize_range_tests): Add FIRST_BB argument, pass it through to optimize_range_tests_var_bound. (maybe_optimize_range_tests, reassociate_bb): Adjust optimize_range_tests callers. * gcc.c-torture/execute/pr85529-1.c: New test. * gcc.c-torture/execute/pr85529-2.c: New test. * gcc.dg/pr85529.c: New test. Added: trunk/gcc/testsuite/gcc.c-torture/execute/pr85529-1.c trunk/gcc/testsuite/gcc.c-torture/execute/pr85529-2.c trunk/gcc/testsuite/gcc.dg/pr85529.c Modified: trunk/gcc/ChangeLog trunk/gcc/testsuite/ChangeLog trunk/gcc/tree-ssa-reassoc.c
[Bug tree-optimization/85529] [7/8/9 Regression] wrong code at -O2 and -O3 on x86_64-linux-gnu
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=85529 Jakub Jelinek changed: What|Removed |Added Status|NEW |ASSIGNED Assignee|unassigned at gcc dot gnu.org |jakub at gcc dot gnu.org --- Comment #7 from Jakub Jelinek --- Created attachment 44026 --> https://gcc.gnu.org/bugzilla/attachment.cgi?id=44026=edit gcc8-pr85529.patch Untested fix.
[Bug tree-optimization/85529] [7/8/9 Regression] wrong code at -O2 and -O3 on x86_64-linux-gnu
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=85529
--- Comment #6 from Jakub Jelinek ---
And finally a testcase that is miscompiled on 7.x, starting with r240858:
/* PR tree-optimization/85529 */
/* { dg-do run } */
/* { dg-options "-O2 -fno-ssa-phiopt" } */
__attribute__((noinline, noclone)) int
foo (int x)
{
x &= 31;
x -= 25;
x *= 2;
if (x < 0)
return 1;
int y = x >> 2;
if (x >= y)
return 1;
return 0;
}
int
main ()
{
int i;
for (i = 0; i < 63; i++)
if (foo (i) != 1)
__builtin_abort ();
return 0;
}
[Bug tree-optimization/85529] [7/8/9 Regression] wrong code at -O2 and -O3 on x86_64-linux-gnu
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=85529
--- Comment #5 from Jakub Jelinek ---
Slightly tweaked testcase:
struct S { int a; };
int b, c = 1, d, e, f;
static int g;
volatile struct S s;
signed char
foo (signed char i, int j)
{
return i < 0 ? i : i << j;
}
int
main ()
{
signed char k = -83;
if (!d)
goto L;
k = e || f;
L:
for (; b < 1; b++)
s.a != (k < foo (k, 2) && (c = k = g));
if (c != 1)
__builtin_abort ();
return 0;
}
Another testcase, that started being miscompiled with r255387 is:
__attribute__((noipa)) int
foo (int x)
{
x &= 63;
x -= 50;
x |= 1;
if (x < 0)
return 1;
int y = x >> 2;
if (x >= y)
return 1;
return 0;
}
int
main ()
{
int i;
for (i = 0; i < 63; i++)
if (foo (i) != 1)
__builtin_abort ();
return 0;
}
[Bug tree-optimization/85529] [7/8/9 Regression] wrong code at -O2 and -O3 on x86_64-linux-gnu
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=85529 Richard Biener changed: What|Removed |Added Target Milestone|8.0 |7.4 Summary|[8/9 Regression] wrong code |[7/8/9 Regression] wrong |at -O2 and -O3 on |code at -O2 and -O3 on |x86_64-linux-gnu|x86_64-linux-gnu --- Comment #4 from Richard Biener --- The issue is latent on the GCC 7 branch.
