diff options
author | Jakub Jelinek <jakub@redhat.com> | 2022-06-21 11:40:16 +0200 |
---|---|---|
committer | Jakub Jelinek <jakub@redhat.com> | 2022-06-21 11:40:16 +0200 |
commit | a0c30fe3b888f20215f3e040d21b62b603804ca9 (patch) | |
tree | b398354964de91afb6a9b2cd2078b247699c18bd | |
parent | 2df1df945fac85d7b3d084001414a66a2709d8fe (diff) | |
download | gcc-a0c30fe3b888f20215f3e040d21b62b603804ca9.zip gcc-a0c30fe3b888f20215f3e040d21b62b603804ca9.tar.gz gcc-a0c30fe3b888f20215f3e040d21b62b603804ca9.tar.bz2 |
ifcvt: Don't introduce trapping or faulting reads in noce_try_sign_mask [PR106032]
noce_try_sign_mask as documented will optimize
if (c < 0)
x = t;
else
x = 0;
into x = (c >> bitsm1) & t;
The optimization is done if either t is unconditional
(e.g. for
x = t;
if (c >= 0)
x = 0;
) or if it is cheap. We already check that t doesn't have side-effects,
but if t is conditional, we need to punt also if it may trap or fault,
as we make it unconditional.
I've briefly skimmed other noce_try* optimizations and didn't find one that
would suffer from the same problem.
2022-06-21 Jakub Jelinek <jakub@redhat.com>
PR rtl-optimization/106032
* ifcvt.cc (noce_try_sign_mask): Punt if !t_unconditional, and
t may_trap_or_fault_p, even if it is cheap.
* gcc.c-torture/execute/pr106032.c: New test.
-rw-r--r-- | gcc/ifcvt.cc | 15 | ||||
-rw-r--r-- | gcc/testsuite/gcc.c-torture/execute/pr106032.c | 21 |
2 files changed, 29 insertions, 7 deletions
diff --git a/gcc/ifcvt.cc b/gcc/ifcvt.cc index b983e87..e007b17 100644 --- a/gcc/ifcvt.cc +++ b/gcc/ifcvt.cc @@ -2833,18 +2833,19 @@ noce_try_sign_mask (struct noce_if_info *if_info) return FALSE; /* This is only profitable if T is unconditionally executed/evaluated in the - original insn sequence or T is cheap. The former happens if B is the - non-zero (T) value and if INSN_B was taken from TEST_BB, or there was no - INSN_B which can happen for e.g. conditional stores to memory. For the - cost computation use the block TEST_BB where the evaluation will end up - after the transformation. */ + original insn sequence or T is cheap and can't trap or fault. The former + happens if B is the non-zero (T) value and if INSN_B was taken from + TEST_BB, or there was no INSN_B which can happen for e.g. conditional + stores to memory. For the cost computation use the block TEST_BB where + the evaluation will end up after the transformation. */ t_unconditional = (t == if_info->b && (if_info->insn_b == NULL_RTX || BLOCK_FOR_INSN (if_info->insn_b) == if_info->test_bb)); if (!(t_unconditional - || (set_src_cost (t, mode, if_info->speed_p) - < COSTS_N_INSNS (2)))) + || ((set_src_cost (t, mode, if_info->speed_p) + < COSTS_N_INSNS (2)) + && !may_trap_or_fault_p (t)))) return FALSE; if (!noce_can_force_operand (t)) diff --git a/gcc/testsuite/gcc.c-torture/execute/pr106032.c b/gcc/testsuite/gcc.c-torture/execute/pr106032.c new file mode 100644 index 0000000..d2ccf6a5 --- /dev/null +++ b/gcc/testsuite/gcc.c-torture/execute/pr106032.c @@ -0,0 +1,21 @@ +/* PR rtl-optimization/106032 */ + +__attribute__((noipa)) int +foo (int x, int *y) +{ + int a = 0; + if (x < 0) + a = *y; + return a; +} + +int +main () +{ + int a = 42; + if (foo (0, 0) != 0 || foo (1, 0) != 0) + __builtin_abort (); + if (foo (-1, &a) != 42 || foo (-42, &a) != 42) + __builtin_abort (); + return 0; +} |