Hi, Jason, On Jan 24, 2013, at 11:59 PM, Jason Manley wrote:
> You can export to specific versions using 2012b, but since Matlab doesn't > announce what is and isn't compatible between versions, it makes it > impossible to cater for everyone unless you export to each and every version > (which is a very time-consuming process). That's great! Why not export the casper library models to the oldest Simulink version that supports the first System Generator version that supports the ROACH2 (i.e. Virtex 6)? Assuming the new tools can open files saved in the old format, that would make the changes usable by the greatest number of people. Are there any new "must have" blocks that you would miss out on? The BitBasher and WaveScope blocks are the latest (though now relatively old) "must have" (IMHO) blocks that I'm aware of, but I don't know what's new in the latest System Generator. > While I agree that it's not fair to force users to upgrade to the latest > versions all the time, I also think it's unreasonable to expect maintainers > to backport patches to each and every old version. I don't think we need to support every old version, but it would be nice to follow the Ubuntu model and pick specific tool versions for "Long Term Support". > The precedent was set years ago when we dropped support for iBOBs and BEE2s. > When Mathworks or Xilinx stops support for certain things, so do we. > Likewise, when they change model file formats, we follow suite. Actually, I thought the library support for iBOB and BEE2s remained until relatively recently (last August). > FWIW, we had massive issues with old versions of Simulink causing files to > somehow become corrupt and it'd crash Matlab (segfault) when trying to open > them. Sometimes it was so bad that just clicking "save" would break things; > we had to constantly "save-as" with small incremental changes and hope that > they didn't break anything or else roll-back and redo all the work. The cause > of this was unknown, even though perfectly reproducable. Matlab's response > when we asked them to fix it? You guessed it!: upgrade to the new version! > And along with a host of other improvements, not least of which is a > now-useful user-interface with much better redrawing, it does seem to be > fixed in 2012b. This is an issue with the old tools, not the old library format. I understand the benefits that newer versions can provide, but I don't think that means we need to save the libraries in the latest format available. > My feeling is that if you want to use patches in your old versions of tools, > then it's up to you to get patches applied. Unfortunately, git doesn't merge > .mdl files, making this very hard. This is certainly possible, but if your fancy new tools can export a version that is compatible with the older tools, it would be a lot less error prone of a way to get those changes to those who use older versions. > If we could figure-out a clean way of scripting mask generation, then we > could regenerate the entire library on-the-fly, which'd go a long way towards > compatibility across versions. Creating a meta-library would be very interesting and, depending on how it's implemented, might be a way to get CASPER on other hardware (e.g. Altera), but I agree with Andrew that it would further limit the number of people who create/maintain meta-library blocks. Dave
