[Scilab-users] Xcos simulation problem with Scilab 5.5.1

[Clément David]

Hello,

Well found, the GAINBLK and GAINBLK_f implementation differs. GAINBLK do
not call the dgemm (LAPACK) if the gain parameter is a scalar double
whereas GAINBLK_f always does.

Due to a bug on the Scilab installer the MKL automatic download failed
and Scilab silently fallback to the reference lapack implementation.

To fix the issue, please follow instruction there :
http://www.scilab.org/en/community/news/20141016

--
Clément


Le dimanche 12 octobre 2014 à 23:59 -0700, JWE a écrit :
> Hello Clement,
> 
> Thank you for checking my diagram.
> As you suggested, I stripped the diagram to a very simple one; sinus-source,
> amplifier and a scope (see attachment).
> When I run it (in Scilab 5.5.1, 64 bits) on my office machine I still have
> the slow scope updating. But when I run it on my home pc, I have no problem
> at all.
> The office pc: Core i3-2100 @ 3.1 GHz, 4 MB memory and no GPU. Windows 7, 64
> bits.
> My home pc: Core i5-3210 @ 2.5 GHz, 8 MB memory and a NVIDIA GeForce GT650M
> GPU. Windows 8.1, 64 bits.
> The main difference seems to be the presence of a dedicated GPU.
> 
> But even more strangly is the observation that when I change the gain block
> (GAINBLK_f) by a gainblok of type "GAINBLK", the simulation runs quickly!
> (When I use the gainblock "GAIN_f", it runs slow again)
> Unfortunately, when I change all the gain blocks in my original diagram by
> gain blocks of type GAINBLK, the simulation still runs slowly....
> Note that with Scilab 5.5.0 (64 bits) I don't have these problems.
> 
> Best regards,
> Jan.
> ADC_test_550_551_rev.zcos
> <http://mailinglists.scilab.org/file/n4031330/ADC_test_550_551_rev.zcos>  
> 
> 
> 
> --
> View this message in context: http://mailinglists.scilab.org/Xcos-simulation-problem-with-Scilab-5-5-1-tp4031312p4031330.html
> Sent from the Scilab users - Mailing Lists Archives mailing list archive at Nabble.com.
> _______________________________________________
> users mailing list
> users at lists.scilab.org
> http://lists.scilab.org/mailman/listinfo/users