Hi Nicolas, Very nice! It’s great to have a fast library for linear algebra.
Cheers, Doru > On Jan 12, 2019, at 12:33 AM, Nicolas Cellier > <[email protected]> wrote: > > Hi, > I announce the availability of Smallapack in Pharo6. > > The ConfigurationOfSmallapack is in > http://www.squeaksource.com/MetacelloRepository and there is a copy in meta > repo for Pharo 3/4/5/6. > > Currently, the ported version uses a derivative of OpalCompiler patched to > handle method with 16+ arguments. > External function calls have not been converted yet to UnifiedFFI, but the > patched compiler rather has hook to compile legacy FFI. > Though I did not install the hook to call FFI with more than 15 arguments, so > there is at least 1 unit test failing (but not crashing). > > I have auto-re-generated all the source code for using UnifiedFFI formats, so > the dependency on legacy FFI is not a necessity (apart for simplifying cross > dialect maintenance). > > But I want to review the generated code method by method rather than filing > it in blindly (the wrapper functions are also generated, and I might loose > comments or improvments if I'm careless). Unfortunately, the state of diff > tools in Pharo6, be it thru MC or worse than all, thru legacy change lists, > does not enable such a large scale review, so I think that I will edit in > Squeak and run in Pharo... > > Now that Smallapack supports Opal, there should be no major problem for > porting to Pharo7, but I did not have time to try yet. A few more MC > regressions, and the fact to forbid protocol beginning with a * was too > serious a cross compatibility hurdle for me. But I'll come back to it, tools > are generally better in ph7 than ph6. Stay tuned. > -- www.feenk.com “Programming is executable philosophy."
