Shiny new PM needs polishing - fails to skip extrema
|
|
Bugzilla Link |
879 |
Created on |
Oct 07, 2011 21:41 |
Resolution |
FIXED |
Resolved on |
Oct 10, 2011 17:43 |
Version |
svn |
OS |
Linux |
Architecture |
PC |
Extended Description
cd ~/sac/testsuite/optimizations/constantfolding
sac2c cfbug.sac -noewlcf -nowlf -doawlf
dies with:
arrayopt/algebraic_wlfi.c:921 Assertion "FALSE" failed!
N_prf not recognized
idx is pointing at the F_noteminval below.
iprime, the argument to PM, is _ivexi_3285_ext.
It looks like the PM "skipping" code stops after the
first skip:
/*** Partition ***/
([ 0, 0, 0 ] <= _pinl_1469_iv__SSA0_1=[_pinl_1475__eat_77, _pinl_1474__eat_76, _pinl_1473__eat_75] < [ _flat_6, _flat_7, _flat_6 ] genwidth [ _flat_6, _flat_7, _flat_6 ])
{
_ivexi_3284_ext = _noteminval_( _pinl_1475__eat_77, _flat_22);
_ivexi_3285_ext = _notemaxval_( _ivexi_3284_ext, _flat_6);
The correct result would have idx pointing at _pinl_1475_eat_77,
which is the WITHIDS_VEC.
This with:
sac2c v1.00-beta (Haggis And Apple)
developer rev 17661:MODIFIED linux-gnu_x86_64
(Fri Oct 7 15:04:36 EDT 2011 by sac)
rbe@rattler:~$