as a result, concho is now able to handle hpe's amd configurations
- also added the l3 cache attribute on cpus (not used yet but I'm planning to show it in the graphs as l3cache per core affects the performance)
- added the ability to output the figure in a file instead of the screen (will be useful for continuous integration)
work related to [https://bugzilla.ipr.univ-rennes.fr/show_bug.cgi?id=4015]
This is not finished, there's still to :
- remove hardcoded base chassis id and price
- add handler for amd configurations (only intel configs are currently handled)
- also refatored Dimm class to contain more details (will be useful to estimate ram throughput)
- also added type hinting to dell.py
work related to [https://bugzilla.ipr.univ-rennes.fr/show_bug.cgi?id=4015]
- added type hinting (to help understanding the code)
- fixed dirty dynamic call to `parse()` method by adding missing abstract base class `IHtmlConfiguratorParser`
- organized data files into catalogs directory to unclutter the root
- added support for hpe web pages for which the cpu options are not present in the javascript catalog. For these pages we have to find the cpu options in the html content
- note: only the page for HP Proliant DL360 gen10 has been tested
Bug 3463 - adapter concho (outil d'aide au choix achat cpu pour hpc) au catalogue hpe
- added support for amd epyc rome processors
- handled the case where the 'additional processor' section doesn't exist in the web page (as it's the case in r6525 configurations)
- the prices are now retrived from matinfo's downloaded pages
- the derived prices (such as the price without cpu) are now automatically computed, rather than hardcoded from (hand computation)
- although it's a bit dirty, the architecture of the code is more generic:
- each configuration is now stored in a Config object, which is able to compute its price.
- dell specific code is now decoupled into a separate file.
warnings:
- the refactoring is designed to handle the old price tables but this work is not finished (the code using the old price tables is broken at the moment)
- the computed price is currently overestimated, as the price of the base memory dimms should be substracted
- disks are not currently handled
Guillaume Raffy
Guillaume Raffy