Re: [tuples][patch] Dead Store Elimination

[Diego Novillo <dnovillo at google dot com>]

On 3/26/08 6:13 AM, Jakub Staszak wrote:

> Previously I wrote there was "no regression", but there were four 
> regressions:
> < FAIL: gcc.dg/tree-ssa/ssa-dse-1.c scan-tree-dump-times dse1 " = 4" 0
> < FAIL: gcc.dg/tree-ssa/ssa-dse-1.c scan-tree-dump-times dse1 " return 4" 1
> < FAIL: gcc.dg/tree-ssa/ssa-dse-6.c scan-tree-dump-times dse1 "local2 = " 1
> < FAIL: gcc.dg/tree-ssa/ssa-dse-7.c scan-tree-dump-times dse1 "return 5" 1
>
> These regressions occurred because of "{dg-final}" instructions for dse 
> only and lack
> of pass_copyprop. These regressions should disappear after pass_copyprop 
> is ready.

That's fine.


>   * gimple.h (gimple_uid, gimple_set_uid): Defined.
>   (gimple_statement_base): Field UID added.
>   * tree-ssa-dse.c (execute_simple_dse): #if 0 removed, bitmap_empty_p 
> condition
>   added. (memory_ssa_name_same, memory_address_same, get_kill_of_stmt_lhs,
>   dse_possible_dead_store_p, dse_optimize_stmt, dse_record_phis,
>   dse_finalize_block, tree_ssa_dse): Tuplified.
>   * passes.c (init_optimization_passes): pass_simple_dse and pass_dse 
> enabled.
>   * testsuite/gcc.dg/tree-ssa/ssa-dse-10.c: {dg-final} changed.
>   * testsuite/gcc.dg/tree-ssa/pr30375.c: {dg-final} changed.

OK with one nit:

> -  return stmt_ann (stmt)->uid;
> +  return gimple_uid (stmt);
>  }
>  
> +  phis = phi_nodes (bb);
> +
> +  for (gsi = gsi_start (phis); !gsi_end_p (gsi); gsi_next (&gsi))

You can call gsi_start_phis() here.

For future patches, it makes things a bit easier to review if you use 
the -p switch to diff.

Thanks for the quick fix.


Diego.