A question about hardware implementation of quick-sort in future Intel CPUs
antoinecarme opened this issue · comments
Hi,
If I understand well, this is a software C++ implementation of a quick sort using hardware AVX vector instructions.
A lot of CPUs now have instructions for cryptography, hashing etc. And Intel does.
Is there a roadmap at intel , even for the next 10 years, to have a basic quick sort performed in hardware the same way a hash is computed ? I know the hardware implementation is complex.
We need the hardware to move closer to what standard software libraries provide as a service level and a hardware sort is big game changer (I think of Data processing, Machine learning, binning, quantization, etc).
Hi @antoinecarme. Thank you for your question but unfortunately I have no idea. An instruction to sort zmm register would be useful though :)
Thanks for your answer.
I was just thinking how our code would be easier if all D. Knuth sorting algorithms were implemented inside CPUs as a sort extension besides AVX (ASX ?). Let alone the speedup improvements for numpy etc.