Ilhan Polat
Ilhan Polat
This is really impressive work but before we go further, note the discussion on #21620 until the end. We don't really need the norm of the inverse but we need...
Indeed, I am also a bit confused about this. Sorry for the vagueness. Let me ask around for 2 and 4 and report back.
> I assume the original Fortran code you translated must have been making these same BLAS calls before, without this lack of performance ? No it is a very old...
Here is some results they provided |OPENBLAS_NUM_THREADS | Runtime | |----------------------------------------| ------------| | 1 | 5.275556 | | 2 | 5.761794 | | 3 | 6.064222 | | 4 |...
@martin-frbg Thank you regardless. I know it can be quite taxing reading some detective work. So please take your time and prioritize your well-being. I just wanted to know if...
@rgommers If you manage to have a local repro mechanism let me know so I can write dscal as a native C loop in a PR so we can test...
If I can remember how to fix this, I will try to squeeze this in today so that it catches the 1.17 train.
The only difference I can think that can affect is the `numpy.einsum` speed or your underlying BLAS/LAPACK provider (OpenBLAS or MKL) has changed. We did not have any other changes...
There is only the change from `gemm` to NumPy `matmul` calls. In small arrays that extra might be significant but as size increases should not matter too much as 4x....
@domna Can you please post your speed comparison for different sizes? I can't replicate this slowdown and also some folks previously had strange segfaults so I'm wondering if this is...