| Message ID | 20100728154547.GZ18378@tyan-ft48-01.lab.bos.redhat.com |
|---|---|
| State | New |
| Headers | show |
On 07/28/10 09:45, Jakub Jelinek wrote: > Hi! > > If a bb ends with a DEBUG_INSN, bb_size between -g and -g0 > compilation might differ. > The problem is that hoist_code forgets to count BB_END (bb) > instruction (and doesn't even set to_bb_head for it, so it is left at 0). > If in -g0 the block ends with a real insn and with -g with > DEBUG_INSN, in the latter case the last real insn will be counted. > As discussed with Maxim on IRC, we should count even the last insn. > > Fixed by using FOR_BB_INSNS macro, bootstrapped/regtested on x86_64-linux > and i686-linux. Ok for trunk? > > 2010-07-28 Jakub Jelinek<jakub@redhat.com> > > PR debug/45105 > * gcse.c (hoist_code): Use FOR_BB_INSNS macro. > > * gcc.dg/pr45105.c: New test. OK. Jeff
--- gcc/gcse.c.jj 2010-07-28 10:36:00.000000000 +0200 +++ gcc/gcse.c 2010-07-28 15:27:56.000000000 +0200 @@ -4389,21 +4389,15 @@ hoist_code (void) FOR_EACH_BB (bb) { rtx insn; - rtx bb_end; int to_head; - insn = BB_HEAD (bb); - bb_end = BB_END (bb); to_head = 0; - - while (insn != bb_end) + FOR_BB_INSNS (bb, insn) { /* Don't count debug instructions to avoid them affecting decision choices. */ if (NONDEBUG_INSN_P (insn)) to_bb_head[INSN_UID (insn)] = to_head++; - - insn = NEXT_INSN (insn); } bb_size[bb->index] = to_head; --- gcc/testsuite/gcc.dg/pr45105.c.jj 2010-07-28 15:29:00.000000000 +0200 +++ gcc/testsuite/gcc.dg/pr45105.c 2010-07-28 15:28:42.000000000 +0200 @@ -0,0 +1,27 @@ +/* PR debug/45105 */ +/* { dg-do compile } */ +/* { dg-options "-Os -fcompare-debug" } */ + +extern int *baz (int *, int *); + +void +bar (int *p1, int *p2) +{ + int n = *baz (0, 0); + p1[n] = p2[n]; +} + +void +foo (int *p, int l) +{ + int a1[32]; + int a2[32]; + baz (a1, a2); + while (l) + { + if (l & 1) + p = baz (a2, p); + l--; + bar (a1, a2); + } +}