Hacker News new | past | comments | ask | show | jobs | submit login
Puget Systems:“Intel Has Every Right” to Make Intel MKL Run Slower on AMD CPUs (reddit.com)
8 points by rahuldottech 3 days ago | hide | past | web | favorite | 3 comments





The advantage of MKL is typically greatly over-rated anyhow, and I don't see why one should care. BLIS and OpenBLAS have good tuned x86 BLAS implementations, and run infinitely faster than MKL on ARM and POWER, for instance, and if you're interested in small matrices on x86_64, there's libxsmm. (I know MKL has more than BLAS, but I don't know what there is that doesn't have free rough equivalents.) BLIS performance: https://github.com/flame/blis/blob/master/docs/Performance.m...

Just as importantly, if you're working on linear algebra now you're simply less likely to use MKL because you know it's very possible your target could change and that MKL code will end up as a lock in you don't need for a relatively minor performance edge. It only really makes sense to stick with MKL if you're absolutely certain you're going to stay Intel forever.

I think the portability and ability to include in distributions is important, but most people I see don't. However, I don't know how you'd be locked in for linear algebra since you can re-link against alternatives.



Guidelines | FAQ | Support | API | Security | Lists | Bookmarklet | Legal | Apply to YC | Contact

Search: