CTZ not quite dead yet
Bugzilla Link | 1091 |
Created on | Sep 27, 2013 18:31 |
Version | svn |
OS | Linux |
Architecture | PC |
Attachments | cubslcrash.sac |
Extended Description
Created an attachment (id=988) source code to reproduce fault I am trying to kill off -doctz, but there are still a few outstanding problems in doing so. One of them can be seen in the CF unit test cubslcrash.sac (wlcondoAKD in disguise). If I compile it under Build #18310 with these options: sac2c cubslcrash.sac -doawlf -nowlf -bopt -noctz > crud I get this: * Partn */ ([ _pinl_2639__flat_33 ] <= _pinl_4008_iv=[_pinl_4011__eat_53] < [ N__SSA0_1 ] genwidth [ _wlsimp_18791 ]) { _dup_15405__pinl_4022__ea_2463__icc_2038 = _accu_( _pinl_4008_iv); _dl_17285 = _mul_SxS_( 2, _pinl_4011__eat_53); _dl_17286 = _add_SxS_( _dl_17285, _dl_17423); _dup_15420__pinl_2650__flat_507 = _eq_SxS_( _pinl_4011__eat_53, _pinl_2639__flat_33); _dup_15421__pinl_2697__flat_507 = _eq_SxS_( 0, _pinl_4011__eat_53); _dup_15422__pinl_2746__flat_995 = _or_SxS_( _dup_15420__pinl_2650__flat_507, _dup_15421__pinl_2697__flat_507); _dup_15423__pinl_3708__flat_1002 = _not_S_( _dup_15422__pinl_2746__flat_995); _dup_15424__pinl_3754__flat_1282 = _toi_S_( _dup_15423__pinl_3708__flat_1002); _dup_15425__pinl_3803__flat_21 = _mul_SxS_( _dl_17286, _dup_15424__pinl_3754__flat_1282); _dup_15426__pinl_3855__flat_1282 = _toi_S_( _dup_15422__pinl_2746__flat_995); _dup_15427__pinl_3904__flat_21 = _mul_SxS_( _dup_15426__pinl_3855__flat_1282, _pinl_4011__eat_53); _dup_15428__pinl_3959__flat_5 = _add_SxS_( _dup_15425__pinl_3803__flat_21, _dup_15427__pinl_3904__flat_21); _dup_15429__pinl_4027__flat_32 = _add_SxS_( _dup_15405__pinl_4022__ea_2463__icc_2038, _dup_15428__pinl_3959__flat_5); } : _dup_15429__pinl_4027__flat_32 ; , If I compile it this way: sac2c cubslcrash.sac -doawlf -nowlf -bopt > crud I get this much more delightful code: /* Partn */ ([ 1 ] <= _pinl_4008_iv=[_pinl_4011__eat_53] < [ _pinl_2639__flat_33 ] genwidth [ _wlsimp_19688 ]) { _dup_18517__pinl_4022__ea_2463__icc_2038 = _accu_( _pinl_4008_iv); _dup_18521__dl_18404 = _mul_SxS_( 2, _pinl_4011__eat_53); _dup_18534__pinl_4027__flat_32 = _add_SxS_( _dup_18517__pinl_4022__ea_2463__icc_2038, _dup_18521__dl_18404); } : _dup_18534__pinl_4027__flat_32 ; , I'll go see if this is fixable. My desire is to disable CTZ (or kill it) as CF and some of its friends are supposed to have enough smarts to handle the sort of functional compositions that arise...