Changes between Version 1 and Version 2 of Ticket #6236, comment 14
- Timestamp:
- 2020-11-30T15:15:56Z (4 years ago)
Legend:
- Unmodified
- Added
- Removed
- Modified
-
Ticket #6236, comment 14
v1 v2 2 2 > @vitalij points out in ticket:6254#comment:8 that we have a {{{--maxSizeSolveLinearSystem}}} flag in OMC already. Is this already doing something like we want to do now? 3 3 4 For the task the defaul qrDecompositionHouseholder: 5 https://github.com/OpenModelica/OpenModelica/blob/master/OMCompiler/Compiler/BackEnd/ResolveLoops.mo#L2793 has an issues with normalization. 4 For the task the defaul [https://github.com/OpenModelica/OpenModelica/blob/master/OMCompiler/Compiler/BackEnd/ResolveLoops.mo#L2793 qrDecompositionHouseholder] has an issues with normalization. The normalization interduce a nonlinear term -> complicates the simpication, which we need. 6 5 7 I guess remove the normalization 6 I guess the normalization was 7 [https://github.com/OpenModelica/OpenModelica/blob/master/OMCompiler/Compiler/BackEnd/ResolveLoops.mo#L2869 L2869] 8 8 9 https://github.com/OpenModelica/OpenModelica/blob/master/OMCompiler/Compiler/BackEnd/ResolveLoops.mo#L2869 9 [https://github.com/OpenModelica/OpenModelica/blob/master/OMCompiler/Compiler/BackEnd/ResolveLoops.mo#L2889 L2889] 10 10 11 https://github.com/OpenModelica/OpenModelica/blob/master/OMCompiler/Compiler/BackEnd/ResolveLoops.mo#L2889 12 13 and remove `makeTmpEqnForExp`: 14 15 https://github.com/OpenModelica/OpenModelica/blob/master/OMCompiler/Compiler/BackEnd/ResolveLoops.mo#L2785 16 17 https://github.com/OpenModelica/OpenModelica/blob/master/OMCompiler/Compiler/BackEnd/ResolveLoops.mo#L2846 18 19 20 https://github.com/OpenModelica/OpenModelica/blob/master/OMCompiler/Compiler/BackEnd/ResolveLoops.mo#L2850 21 22 23 https://github.com/OpenModelica/OpenModelica/blob/master/OMCompiler/Compiler/BackEnd/ResolveLoops.mo#L2857 24 25 https://github.com/OpenModelica/OpenModelica/blob/master/OMCompiler/Compiler/BackEnd/ResolveLoops.mo#L2870 26 27 https://github.com/OpenModelica/OpenModelica/blob/master/OMCompiler/Compiler/BackEnd/ResolveLoops.mo#L2877 28 29 https://github.com/OpenModelica/OpenModelica/blob/master/OMCompiler/Compiler/BackEnd/ResolveLoops.mo#L2890 30 31 https://github.com/OpenModelica/OpenModelica/blob/master/OMCompiler/Compiler/BackEnd/ResolveLoops.mo#L2900 32 33 can work! 34 35 11 and please remove `makeTmpEqnForExp` inside qrDecompositionHouseholder. 36 12 `makeTmpEqnForExp`: Add some helper variables, which can avoid the wished simplication here! 37 13 I don't belive any more it's a good idea, to add here some variables. 38 14 39 15 Note: remove normalization make the approach numerical weak. But for small system is should work well. 16