Hello,

I bumped into roots() problem again - SMath 0.93
Here is the file - Primer43.sm.
Or a slight modification - Primer43a.sm
It's been again solved by Newton-Raphson - linear problem, but roots() failed.

In WinXP if you try to get the results by roots() - crash will happen.


In Ubuntu - stack overflow error will be issued.


Regards,
Radovan

Wrote

Hello,

I bumped into roots() problem again - SMath 0.93
Here is the file - Primer43.sm.
It's been again solved by Newton-Raphson - linear problem, but roots() failed.

In WinXP if you try to get the results by roots() crash will happen
[...]

In Ubuntu - stack overflow error will be issued
[...]

Regards,
Radovan

Same problem with Win 7

Thank you. Will fix it.

Hello Andrey,

Here is another similar example Primer44.sm . SMath will crash with roots() again.
Fortunately, Newton-Raphson will do the job.

Regards,
Radovan

Hello Andrey,

Here is another example Primer46.sm

This time there is no roots() involved.


Crash will happen (stack overflow in Ubuntu) if the optimization of the region calculating X is Symbolical. If we set the Numerical optimization everything is fine.

Regards,
Radovan

Ok. Will check all these examples.

Regards.

Hello Andrey,

I was pleasantly surprised that roots() worked in my example Pimer44.sm


Primer44OK.sm

Primer43.sm and Primer43a.sm did not cause stack overflow but only "No real roots" message. Not quite sure why because those problems are similar.
Primer46.sm still crashes with Symbolic optimization.

I think it is an improvement of roots() in 0.94 regarding these examples. Thank you for your efforts.

Regards,
Radovan

Hello,

Wrote

Primer46.sm still crashes with Symbolic optimization.

Unfortunately, It happened in v0.95 as well. I changed this file a bit, regarding the change in stack() function in order to get a function for which Jacobian can be found. It looks quite innocent - but still crashes for symbolic optimization .
Here is the end of this file:

and now when enabling calculation of the last region

The Jacobian matrix J is problematic here. If you use Symbolic optimization of J inverse - crash will happen.


Here is the file: Primer46-crash-nostack-Ok.sm

Why this causes the crash - have no idea. Could someone read this file and check it out, please?

Regards,
Radovan

Hi omorr,

I can see the same issue, but I've no ideas about origins (could be the symbolic evaluation of submatrix()?)


regards,

w3b5urf3r

Thanks for answering w3b5urf3r,

Wrote

I can see the same issue, but I've no ideas about origins (could be the symbolic evalutation of submatrix()?)

It seems submatrix() is not the reason. I have another example where roots() will work - although the function is defined with submatrix() and the Jacobian is not with all constants (nonlinear equations):


Here is the file: Primer47-Ok-1.sm.

I do not know what is the reason of this, but SMath should not crash anyway.

Regards,
Radovan

if may be useful, in this file (0.95_Jacob_symbolic.sm.zip), I've minimized all possible crash causes (lines(), for loops, etc...)

I think that this could reduce the issue options (mathematical problem or code problem with Jacob()<->symbolic evaluation)...




regards,

w3b5urf3r

Thank you w3b5urf3r

You gave me again the idea where to "poke" about this. It seems there is nothing to do with Jacob or anything else (lines(), for loops, etc...). It is about this particular matrix and numbers in it. Just put this matrix with particular numbers in an empty sheet and again crash will happen.



Here is the file : strange matrix crash.sm

Just put some other numbers and choose Symbolic for, say, J/J and crash will happen again. Put some other numbers and crash will not happen.
Very strange

Regards,
radovan

Wrote

Just put some other numbers and choose Symbolic for, say, J/J and crash will happen again. Put some other numbers and crash will not happen.
Very strange

really very strange, thus the problem it's the inversion but not for any matrix...

in any case Jacob() seems to work fine


regards,

w3b5urf3r