| Message ID | 20100824191926.GS702@tyan-ft48-01.lab.bos.redhat.com |
|---|---|
| State | New |
| Headers | show |
On Tue, Aug 24, 2010 at 9:19 PM, Jakub Jelinek <jakub@redhat.com> wrote: > Hi! > > In some cases the operands of the comparison won't have the desired type, > e.g. if there were any (useless) differences in types of the operands. > Having a NOP_EXPR as gimple comparison's operand is invalid though. > > Fixed thusly, bootstrapped/regtested on x86_64-linux and i686-linux. > Ok for trunk? Ok. Thanks, Richard. > 2010-08-24 Jakub Jelinek <jakub@redhat.com> > > PR tree-optimization/45059 > * tree-ssa-reassoc.c (eliminate_redundant_comparison): Strip useless > type conversions from newop{1,2}. Assert t is a comparison and > newop{1,2} after the stripping are gimple vals. > > * gcc.c-torture/compile/pr45059.c: New test. > > --- gcc/tree-ssa-reassoc.c.jj 2010-08-20 16:05:41.000000000 +0200 > +++ gcc/tree-ssa-reassoc.c 2010-08-24 19:16:57.000000000 +0200 > @@ -1314,9 +1314,14 @@ eliminate_redundant_comparison (enum tre > enum tree_code subcode; > tree newop1; > tree newop2; > + gcc_assert (COMPARISON_CLASS_P (t)); > tmpvar = create_tmp_var (TREE_TYPE (t), NULL); > add_referenced_var (tmpvar); > extract_ops_from_tree (t, &subcode, &newop1, &newop2); > + STRIP_USELESS_TYPE_CONVERSION (newop1); > + STRIP_USELESS_TYPE_CONVERSION (newop2); > + gcc_checking_assert (is_gimple_val (newop1) > + && is_gimple_val (newop2)); > sum = build_and_add_sum (tmpvar, newop1, newop2, subcode); > curr->op = gimple_get_lhs (sum); > } > --- gcc/testsuite/gcc.c-torture/compile/pr45059.c.jj 2010-08-24 19:05:44.000000000 +0200 > +++ gcc/testsuite/gcc.c-torture/compile/pr45059.c 2010-08-24 19:05:25.000000000 +0200 > @@ -0,0 +1,23 @@ > +/* PR tree-optimization/45059 */ > + > +typedef unsigned int T; > +extern void foo (signed char *, int); > + > +static signed char a; > +static T b[1] = { -1 }; > +static unsigned char c; > + > +static inline short int > +bar (short v) > +{ > + c |= a < b[0]; > + return 0; > +} > + > +int > +main () > +{ > + signed char *e = &a; > + foo (e, bar (bar (c))); > + return 0; > +} > > Jakub >
--- gcc/tree-ssa-reassoc.c.jj 2010-08-20 16:05:41.000000000 +0200 +++ gcc/tree-ssa-reassoc.c 2010-08-24 19:16:57.000000000 +0200 @@ -1314,9 +1314,14 @@ eliminate_redundant_comparison (enum tre enum tree_code subcode; tree newop1; tree newop2; + gcc_assert (COMPARISON_CLASS_P (t)); tmpvar = create_tmp_var (TREE_TYPE (t), NULL); add_referenced_var (tmpvar); extract_ops_from_tree (t, &subcode, &newop1, &newop2); + STRIP_USELESS_TYPE_CONVERSION (newop1); + STRIP_USELESS_TYPE_CONVERSION (newop2); + gcc_checking_assert (is_gimple_val (newop1) + && is_gimple_val (newop2)); sum = build_and_add_sum (tmpvar, newop1, newop2, subcode); curr->op = gimple_get_lhs (sum); } --- gcc/testsuite/gcc.c-torture/compile/pr45059.c.jj 2010-08-24 19:05:44.000000000 +0200 +++ gcc/testsuite/gcc.c-torture/compile/pr45059.c 2010-08-24 19:05:25.000000000 +0200 @@ -0,0 +1,23 @@ +/* PR tree-optimization/45059 */ + +typedef unsigned int T; +extern void foo (signed char *, int); + +static signed char a; +static T b[1] = { -1 }; +static unsigned char c; + +static inline short int +bar (short v) +{ + c |= a < b[0]; + return 0; +} + +int +main () +{ + signed char *e = &a; + foo (e, bar (bar (c))); + return 0; +}
Hi! In some cases the operands of the comparison won't have the desired type, e.g. if there were any (useless) differences in types of the operands. Having a NOP_EXPR as gimple comparison's operand is invalid though. Fixed thusly, bootstrapped/regtested on x86_64-linux and i686-linux. Ok for trunk? 2010-08-24 Jakub Jelinek <jakub@redhat.com> PR tree-optimization/45059 * tree-ssa-reassoc.c (eliminate_redundant_comparison): Strip useless type conversions from newop{1,2}. Assert t is a comparison and newop{1,2} after the stripping are gimple vals. * gcc.c-torture/compile/pr45059.c: New test. Jakub