Open Moredread opened 10 years ago
8 million is a lot. Can you specify what "very slow" means? Also, since this is Python would not have expected it to be fast.
2 million is already unwieldy slow and that runs in under 50 CPU hours with my simulation parameters. 8 millions might be interesting for benchmarking for me. The script takes about 3 minutes, but I can time that later. I run 7 simulations atm, so it adds up.
On April 12, 2014 1:48:01 PM CEST, Eduard Bopp notifications@github.com wrote:
8 million is a lot. Can you specify what "very slow" means? Also, since this is Python would not have expected it to be fast.
Reply to this email directly or view it on GitHub: https://github.com/aepsil0n/jelly/issues/50#issuecomment-40278426
3 minutes for 2 million
On April 12, 2014 2:05:22 PM CEST, Andre Bubel kontakt@andre-bubel.de wrote:
2 million is already unwieldy slow and that runs in under 50 CPU hours with my simulation parameters. 8 millions might be interesting for benchmarking for me. The script takes about 3 minutes, but I can time that later. I run 7 simulations atm, so it adds up.
On April 12, 2014 1:48:01 PM CEST, Eduard Bopp notifications@github.com wrote:
8 million is a lot. Can you specify what "very slow" means? Also, since this is Python would not have expected it to be fast.
Reply to this email directly or view it on GitHub: https://github.com/aepsil0n/jelly/issues/50#issuecomment-40278426
Well, as I said, all the number crunching is done in Python. Does pypy help with this issue? I suspect that converting all the performance critical stuff to, say, Cython might make the code somewhat more difficult to maintain.
My IC generation code is very slow for only 8,000,000 particles.
We should look which code is performance critical.