Differences between revisions 9 and 10
⇤ ← Revision 9 as of 2009-04-09 10:31:31 →
Size: 2982
Editor: SébastienVillemot
Comment:
← Revision 10 as of 2009-04-09 16:08:13 → ⇥
Size: 3063
Editor: SébastienVillemot
Comment:
Deletions are marked like this. Additions are marked like this.
Line 8: Line 8:
 * If USE_DLL is not specified as an option in the model block any mex file (resulting from previous simulations) should be erased before any call to _static or _dynamic, otherwise matlab considers the mex file and not the m files.
Line 12: Line 11:
 * If USE_DLL is not specified as an option in the model block any mex file (resulting from previous simulations) should be erased before any call to _static or _dynamic, otherwise matlab considers the mex file and not the m files.
  '''Fixed in r2583''' -- SébastienVillemot <<DateTime(2009-04-09T16:08:13Z)>>

Outstanding bugs

  • The macroprocessor fails when the file ends with @#endif or @#endfor without a new line at the EOF

  • ExprNode::collectEndogenous() returns the empty set when called on a model local variable node (is this really a bug?) -- SébastienVillemot 2008-04-08 17:26:12

  • The block estimated_params_init does not initialize the standard deviations of the shocks and dsge_prior_weight.

  • temporary terms for static file are incorrectly computed in SparseDLL mode (should separate temporary terms for the static file from those for the dynamic file since the output langage is not the same in the two situations)

Resolved bugs

  • If USE_DLL is not specified as an option in the model block any mex file (resulting from previous simulations) should be erased before any call to _static or _dynamic, otherwise matlab considers the mex file and not the m files.

  • mex files produced with USE_DLL should return space matrices. Currently it conflicts with sparse_hessian_times_B_kronecker_C at order=2 -- MichelJuillard 2008-08-15 13:16:18

  • estimation statement doesn't accept order option

  • Since r1717, second order derivatives always computed, even with order=1 -- SébastienVillemot 2008-02-29 16:05:21

  • missleading error message when LN is used out of place. LN should be defined as a token different from LOG but trigger the same actions in DynareBison. -- MichelJuillard 2007-06-11 21:02:19

  • Bug with USE_DLL: the generated c codes seem correct (I found no differences between _dynamic.c and _dynamic_.m), but the jacobian obtained using the mex files is wrong. I've got an example where only the first two columns of the jacobian matrix are (wrongly) filled. This results in a lot of NaN eigenvalues (Blanchard & Kahn conditions are not satisfied)

  • Bugs related to the USE_DLL option:

    • Near the top of <MOD_FILE_NAME>.m, if exist(<MOD_FILE_NAME>_static.c)) should be replaced by if exist('<MOD_FILE_NAME>_static.c')

    • Also if exist(<MOD_FILE_NAME>_dynamic.c)) should be replaced by if exist('<MOD_FILE_NAME>_dynamic.c')

    • A space is missing between the optimization flag (-O) and the name of the c files to be compiled (mex)
    • The # statement does not work with USE_DLL option

    • Fixed in r1301 -- SébastienVillemot 2007-06-06 10:30:04

DynareWiki: PreProcessorBugs (last edited 2009-04-09 16:08:13 by SébastienVillemot)