Ilhan Polat

I think I managed to find where it goes wrong; it's one of those ancient Fortran77-isms that I missed during the translations. @nickodell, @SamuelGabriel If you have any capacity, could...

Ouch sorry, my bad, mindless copy pasta from the wrong machine. You are obviously right. Should be ```cpp sgnd = dp * copysign(1.0, *dx); ``` Regarding the why; if `*dx`...

Then there are more places probably. In my ubuntu box it started to give 0.99999 as the results but let me check again if I am overlooking something else. Thank...

I have been obsessing with this in the last a few days. I still don't have an answer while in the meantime I tested on Fedora 42 and also cannot...

@Chris00 if you have any capacity to test with different compiler combinations that would be very valuable. So far it seems like something is off with the runtime (though not...

Because a few planets should align to keep things integer and still make the results correct. It is not a both valid and common enough edge case to consider at...

Matlab uses Tim Davis' SuiteSparse magic for sparse structure discovery and solvers and that's where the speed is coming from. We are now doing a structure discovery in dense `linalg.solve`...

It is always a bit tricky to make sense of any benchmarks but probably there are some patterns that the libraries are tuned for and maybe they are stated in...

I'm not up to speed with our current sparse array situation but if there is a way to discover structure, I think that would be fantastic addition. Then, the really...

Squaring is a strictly increasing function so the argument at the minimums agree. Hence it is the same problem with a modified formulation. This is typically not spelled out in...