| Message ID | ZjswM7hdaeaORt/L@tucnak |
|---|---|
| State | New |
| Headers | show |
| Series | reassoc: Fix up optimize_range_tests_to_bit_test [PR114965] | expand |
On Wed, 8 May 2024, Jakub Jelinek wrote: > Hi! > > The optimize_range_tests_to_bit_test optimization normally emits a range > test first: > if (entry_test_needed) > { > tem = build_range_check (loc, optype, unshare_expr (exp), > false, lowi, high); > if (tem == NULL_TREE || is_gimple_val (tem)) > continue; > } > so during the bit test we already know that exp is in the [lowi, high] > range, but skips it if we have range info which tells us this isn't > necessary. > Also, normally it emits shifts by exp - lowi counter, but has an > optimization to use just exp counter if the mask isn't a more expensive > constant in that case and lowi is > 0 and high is smaller than prec. > > The following testcase is miscompiled because the two abnormal cases > are triggered. The range of exp is [43, 43][48, 48][95, 95], so we on > 64-bit arch decide we don't need the entry test, because 95 - 43 < 64. > And we also decide to use just exp as counter, because the range test > tests just for exp == 43 || exp == 48, so high is smaller than 64 too. > Because 95 is in the exp range, we can't do that, we'd either need to > do a range test first, i.e. > if (exp - 43U <= 48U - 43U) if ((1UL << exp) & mask1)) > or need to subtract lowi from the shift counter, i.e. > if ((1UL << (exp - 43)) & mask2) > but can't do both unless r.upper_bound () is < prec. > > The following patch ensures that. > > Bootstrapped/regtested on x86_64-linux and i686-linux, ok for trunk? OK. Thanks, Richard. > 2024-05-08 Jakub Jelinek <jakub@redhat.com> > > PR tree-optimization/114965 > * tree-ssa-reassoc.cc (optimize_range_tests_to_bit_test): Don't try to > optimize away exp - lowi subtraction from shift count unless entry > test is emitted or unless r.upper_bound () is smaller than prec. > > * gcc.c-torture/execute/pr114965.c: New test. > > --- gcc/tree-ssa-reassoc.cc.jj 2024-01-12 10:07:58.384848977 +0100 > +++ gcc/tree-ssa-reassoc.cc 2024-05-07 18:18:45.558814991 +0200 > @@ -3418,7 +3418,8 @@ optimize_range_tests_to_bit_test (enum t > We can avoid then subtraction of the minimum value, but the > mask constant could be perhaps more expensive. */ > if (compare_tree_int (lowi, 0) > 0 > - && compare_tree_int (high, prec) < 0) > + && compare_tree_int (high, prec) < 0 > + && (entry_test_needed || wi::ltu_p (r.upper_bound (), prec))) > { > int cost_diff; > HOST_WIDE_INT m = tree_to_uhwi (lowi); > --- gcc/testsuite/gcc.c-torture/execute/pr114965.c.jj 2024-05-07 18:17:16.767031821 +0200 > +++ gcc/testsuite/gcc.c-torture/execute/pr114965.c 2024-05-07 18:15:52.332188943 +0200 > @@ -0,0 +1,30 @@ > +/* PR tree-optimization/114965 */ > + > +static void > +foo (const char *x) > +{ > + > + char a = '0'; > + while (1) > + { > + switch (*x) > + { > + case '_': > + case '+': > + a = *x; > + x++; > + continue; > + default: > + break; > + } > + break; > + } > + if (a == '0' || a == '+') > + __builtin_abort (); > +} > + > +int > +main () > +{ > + foo ("_"); > +} > > Jakub > >
--- gcc/tree-ssa-reassoc.cc.jj 2024-01-12 10:07:58.384848977 +0100 +++ gcc/tree-ssa-reassoc.cc 2024-05-07 18:18:45.558814991 +0200 @@ -3418,7 +3418,8 @@ optimize_range_tests_to_bit_test (enum t We can avoid then subtraction of the minimum value, but the mask constant could be perhaps more expensive. */ if (compare_tree_int (lowi, 0) > 0 - && compare_tree_int (high, prec) < 0) + && compare_tree_int (high, prec) < 0 + && (entry_test_needed || wi::ltu_p (r.upper_bound (), prec))) { int cost_diff; HOST_WIDE_INT m = tree_to_uhwi (lowi); --- gcc/testsuite/gcc.c-torture/execute/pr114965.c.jj 2024-05-07 18:17:16.767031821 +0200 +++ gcc/testsuite/gcc.c-torture/execute/pr114965.c 2024-05-07 18:15:52.332188943 +0200 @@ -0,0 +1,30 @@ +/* PR tree-optimization/114965 */ + +static void +foo (const char *x) +{ + + char a = '0'; + while (1) + { + switch (*x) + { + case '_': + case '+': + a = *x; + x++; + continue; + default: + break; + } + break; + } + if (a == '0' || a == '+') + __builtin_abort (); +} + +int +main () +{ + foo ("_"); +}
Hi! The optimize_range_tests_to_bit_test optimization normally emits a range test first: if (entry_test_needed) { tem = build_range_check (loc, optype, unshare_expr (exp), false, lowi, high); if (tem == NULL_TREE || is_gimple_val (tem)) continue; } so during the bit test we already know that exp is in the [lowi, high] range, but skips it if we have range info which tells us this isn't necessary. Also, normally it emits shifts by exp - lowi counter, but has an optimization to use just exp counter if the mask isn't a more expensive constant in that case and lowi is > 0 and high is smaller than prec. The following testcase is miscompiled because the two abnormal cases are triggered. The range of exp is [43, 43][48, 48][95, 95], so we on 64-bit arch decide we don't need the entry test, because 95 - 43 < 64. And we also decide to use just exp as counter, because the range test tests just for exp == 43 || exp == 48, so high is smaller than 64 too. Because 95 is in the exp range, we can't do that, we'd either need to do a range test first, i.e. if (exp - 43U <= 48U - 43U) if ((1UL << exp) & mask1)) or need to subtract lowi from the shift counter, i.e. if ((1UL << (exp - 43)) & mask2) but can't do both unless r.upper_bound () is < prec. The following patch ensures that. Bootstrapped/regtested on x86_64-linux and i686-linux, ok for trunk? 2024-05-08 Jakub Jelinek <jakub@redhat.com> PR tree-optimization/114965 * tree-ssa-reassoc.cc (optimize_range_tests_to_bit_test): Don't try to optimize away exp - lowi subtraction from shift count unless entry test is emitted or unless r.upper_bound () is smaller than prec. * gcc.c-torture/execute/pr114965.c: New test. Jakub