[esp-r] Re: Errors when installing the training models on Solaris

[Joe Clarke]

Hi Daniel

Costola, D. wrote:

>I download the version from pre-release_patches on Friday.
>It compiles well, and there are no errors.
>
>The problem is that the training models calculation is strangely slow.
>I kill the process before the end, after 4.5 hours!
>Just to have a comparison, we checkout the current version from trunk, and it took only 40 minutes to fully install it.
>  
>
The cause, I suspect, is the new calculation of diffuse shading. My testing
on relatively simple models did not give me cause for concern. However,
Jon Hand has pointed out to me the long run-times with exemplar models
where these have complex shading devices. To help mitigate this problem
I changed the default in ish to an isotropic sky (i.e. one pass of the 
sky vault
per target surface and month-day) rather than anisotropic (each month-day
hour reprocessed). While this will reduce the computation load by a factor
of 10, it will still lead to long compute times in some cases.

Jon and I have discussed the problem and agreed that the solution is to 
store
all zone shading & insolation databases associated with exemplar models in
their ascii form (as produced by ish) so that an install will only 
require an
ascii-to-binary conversion. Look out for a posting to this effect in the 
near future.

Regards,

Joe

-- 
Professor J A Clarke                email: joe at esru.strath.ac.uk
ESRU, Dept. of Mechanical Eng.           phone: +44 141 548 3986
University of Strathclyde                  fax: +44 141 552 5105
Glasgow G1 1XJ, UK                 http://www.esru.strath.ac.uk/

-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://lists.strath.ac.uk/archives/esp-r/attachments/20080218/eac2a78e/attachment.html