____  _             
| __ )| | ___   __ _ 
|  _ \| |/ _ \ / _` |
| |_) | | (_) | (_| |
|____/|_|\___/ \__, |
               |___/ 
     _               _            
 _/_| |__   ___ _ __| |_   _  _/_ 
/ _ \ '_ \ / _ \ '__| | | | |/ _ \
  __/ |_) |  __/ |  | | |_| |  __/
\___|_.__/ \___|_|  |_|\__,_|\___|

< Previous post | Next post >

[2026-02-13][Daily]

Evaluating pytorch openmp backend

Open-mp had a bug so I fixed it. Interestingly it has similar runtime, but more avg cpu consumption, so more energy consumption.

That's pertty interesting, and it makes me want to try workload experiments on different number of threads for each dataset/framework.

Here nothing that interesting, pytorch-dp = pytorch-intraop-56, dahl-hyper = dahl

Note that here, I complety disabled set_num_threads on pytorch-intraop. This more or less matches open-mp behavior, but nothing really interesting.

< Previous post | Next post >