| Message ID | 20211228102151.GP2646553@tucnak |
|---|---|
| State | New |
| Headers | show |
| Series | loop-invariant: Fix -fcompare-debug failure [PR103837] | expand |
On 12/28/2021 3:21 AM, Jakub Jelinek wrote: > Hi! > > In the following testcase we have a -fcompare-debug failure, because > can_move_invariant_reg doesn't ignore DEBUG_INSNs in its decisions. > In the testcase we have due to uninitialized variable: > loop_header > debug_insn using pseudo84 > pseudo84 = invariant > insn using pseudo84 > end loop > and with -g decide not to move the pseudo84 = invariant before the > loop header; in this case not resetting the debug insns might be fine. > But, we could have also: > pseudo84 = whatever > loop_header > debug_insn using pseudo84 > pseudo84 = invariant > insn using pseudo84 > end loop > and in that case not resetting the debug insns would result in wrong-debug. > And, we don't really have generally a good substitution on what pseudo84 > contains, it could inherit various values from different paths. > So, the following patch ignores DEBUG_INSNs in the decisions, and if there > are any that previously prevented the optimization, resets them before > return true. > > Bootstrapped/regtested on x86_64-linux and i686-linux, ok for trunk? > > 2021-12-28 Jakub Jelinek <jakub@redhat.com> > > PR rtl-optimization/103837 > * loop-invariant.c (can_move_invariant_reg): Ignore DEBUG_INSNs in > the decisions whether to return false or continue and right before > returning true reset those debug insns that previously caused > returning false. > > * gcc.dg/pr103837.c: New test. OK jeff
--- gcc/loop-invariant.c.jj 2021-11-25 08:33:08.619232065 +0100 +++ gcc/loop-invariant.c 2021-12-27 13:44:44.094132316 +0100 @@ -1691,6 +1691,7 @@ can_move_invariant_reg (class loop *loop unsigned int dest_regno, defs_in_loop_count = 0; rtx_insn *insn = inv->insn; basic_block bb = BLOCK_FOR_INSN (inv->insn); + auto_vec <rtx_insn *, 16> debug_insns_to_reset; /* We ignore hard register and memory access for cost and complexity reasons. Hard register are few at this stage and expensive to consider as they @@ -1725,10 +1726,13 @@ can_move_invariant_reg (class loop *loop continue; /* Don't move if a use is not dominated by def in insn. */ - if (use_bb == bb && DF_INSN_LUID (insn) >= DF_INSN_LUID (use_insn)) - return false; - if (!dominated_by_p (CDI_DOMINATORS, use_bb, bb)) - return false; + if ((use_bb == bb && DF_INSN_LUID (insn) >= DF_INSN_LUID (use_insn)) + || !dominated_by_p (CDI_DOMINATORS, use_bb, bb)) + { + if (!DEBUG_INSN_P (use_insn)) + return false; + debug_insns_to_reset.safe_push (use_insn); + } } /* Check for other defs. Any other def in the loop might reach a use @@ -1751,6 +1755,13 @@ can_move_invariant_reg (class loop *loop return false; } + /* Reset debug uses if a use is not dominated by def in insn. */ + for (auto use_insn : debug_insns_to_reset) + { + INSN_VAR_LOCATION_LOC (use_insn) = gen_rtx_UNKNOWN_VAR_LOC (); + df_insn_rescan (use_insn); + } + return true; } --- gcc/testsuite/gcc.dg/pr103837.c.jj 2021-12-27 13:42:23.262111770 +0100 +++ gcc/testsuite/gcc.dg/pr103837.c 2021-12-27 13:41:36.635767125 +0100 @@ -0,0 +1,19 @@ +/* PR rtl-optimization/103837 */ +/* { dg-do compile } */ +/* { dg-options "-Og -fcompare-debug -fmove-loop-invariants -fnon-call-exceptions -fexceptions -fdelete-dead-exceptions -fno-tree-dce -w" } */ + +unsigned long int +foo (int x) +{ + double a; + int b; + unsigned long int ret = a; + + for (;;) + { + b = !!((int) a); + a = x; + } + + return ret; +}
Hi! In the following testcase we have a -fcompare-debug failure, because can_move_invariant_reg doesn't ignore DEBUG_INSNs in its decisions. In the testcase we have due to uninitialized variable: loop_header debug_insn using pseudo84 pseudo84 = invariant insn using pseudo84 end loop and with -g decide not to move the pseudo84 = invariant before the loop header; in this case not resetting the debug insns might be fine. But, we could have also: pseudo84 = whatever loop_header debug_insn using pseudo84 pseudo84 = invariant insn using pseudo84 end loop and in that case not resetting the debug insns would result in wrong-debug. And, we don't really have generally a good substitution on what pseudo84 contains, it could inherit various values from different paths. So, the following patch ignores DEBUG_INSNs in the decisions, and if there are any that previously prevented the optimization, resets them before return true. Bootstrapped/regtested on x86_64-linux and i686-linux, ok for trunk? 2021-12-28 Jakub Jelinek <jakub@redhat.com> PR rtl-optimization/103837 * loop-invariant.c (can_move_invariant_reg): Ignore DEBUG_INSNs in the decisions whether to return false or continue and right before returning true reset those debug insns that previously caused returning false. * gcc.dg/pr103837.c: New test. Jakub