Gianluca Frison
Gianluca Frison
Ok so now with this commit https://github.com/giaf/blasfeo/commit/b2b1377e6ea54a4440df0c447afd6d4107c42168 alignment should be managed in a portable way using a macro. @RoyiAvital could you check that this still works for you, then I...
Yes sorry I made a bit of mess, first I went for ALIGN and then finally ALIGNED, but I forgot some apparently. Strange that it compiled for me. I'll fix...
Here it is https://github.com/giaf/blasfeo/blob/windows_support/include/blasfeo_common.h#L51 I just pushed the fixes ALIGN => ALIGNED. If this was the only issue, it should work now.
There was another issue, apparently the correct macro for gcc is `__GNUC__` instead of `_GCC_`. I fixed also that, the code is here https://github.com/giaf/blasfeo/tree/windows_support
Whether you should add the conversion cost to the timings depends on your application. The panel-major format used in the BLASFEO API is used for both input and output arguments,...
Now there are some tips on how to use BLASFEO on Android https://github.com/giaf/blasfeo/blob/master/guidelines.md
This https://github.com/giaf/blasfeo/commit/f786b0da61bcc365ed5fef5e9e72c01980a04751 should fix (I can't try it out, maybe make the changes locally to check it out) I don't remember where the acados version stands, but recently a lot...
Yes what you propose would indeed reduce the latency by 1 clock cycle: from 5 of hadd to 4=1(unpackhi)+3(add). But at the end, the reduction code is not so important,...
Sure if you want to make the changes and make a PR, I would be happy to merge it. But otherwise I would leave it as it is for now,...