WLF doesn't work in sac/apex/testforAKD/testforAKD.sac with WLBSC. CSE fault?
|
|
Bugzilla Link |
657 |
Created on |
Jan 09, 2010 22:59 |
Resolution |
FIXED |
Resolved on |
Jan 11, 2010 15:34 |
Version |
svn |
OS |
Linux |
Architecture |
PC |
Extended Description
See also Bug #655.
Here's a code fragment from Loop_2, with Build #16708:MODIFIED:
sac2c testforAKD.sac -b11:ebt >crud
_pinl_1605__flat_5 = [ _pinl_1522__flat_95 ];
_pinl_1586__flat_110 = with {
(_wlbsc_2961_sc_bound <= _pinl_1611__flat_6=[_pinl_1612_i] < _pinl_1605__flat_5 genwidth [ _pinl_1522__flat_95 ])
{
/* empty */
} : _pinl_1612_i ;
} :
genarray( _pinl_1605__flat_5, _pinl_1517__flat_94);
_pinl_1604__wlbsc_1078_sc_bound = [ _pinl_1522__flat_95 ];
_pinl_1597_res = with {
(_wlbsc_2961_sc_bound <= _pinl_1594_iv=[_pinl_1598__eat_186] < _pinl_1605__flat_5 genwidth [ _pinl_1522__flat_95 ])
{
_pinl_1595__flat_1361 = _sel_VxA_( _pinl_1594_iv, _pinl_1586__flat_110);
} : _pinl_1595__flat_1361 ;
} :
genarray( _pinl_1605__flat_5, _pinl_1593__flat_1355);
Note that _pinl_1586__flat_110 should have been WL folded into
_pinl_1597_res, but was not. My new CWLE code should work
here, but look at these lines, which remain the same over SAACYC cycles:
_pinl_1605__flat_5 = [ _pinl_1522__flat_95 ];
_pinl_1604__wlbsc_1078_sc_bound = [ _pinl_1522__flat_95 ];
It looks to me like CSE is not picking these up, for some reason.
Time to get out the shovel and start digging.