search

petrelharp / ftprime_ms

4 stars 2 forks source link

Progress on simulation benchmarks #7

Closed ashander closed 6 years ago

ashander commented 7 years ago

added 2017-12-06

molpopgen commented 6 years ago

I am running more 4Nr values for N=1e3 and N=1e4. This will allow us to plot speed w/ARG over speed w/neutral mutations. It'll all run in a day or so.

ashander commented 6 years ago

I think we've done above but added some new tasks. @molpopgen correct me if wrong but current status is below (and the c++ ones with pdel=0.0 are almost done):

molpopgen commented 6 years ago

Yep--getting there!!!!

On Wed, Dec 6, 2017 at 10:13 AM Jaime Ashander notifications@github.com wrote:

I think we've done above but added some new tasks. @molpopgen https://github.com/molpopgen correct me if wrong but current status is below (and the c++ ones with pdel=0.0 are almost done):

  • sims tracking netural with simupop (for lower range of 4Nr, N)
  • sims with pdel = 0 for (tracking netural, tracking ancestry) with fwdpy11
  • sims with pdel = 0 for (tracking neutral, tracking ancestry) with simuPOP

— You are receiving this because you were mentioned.

Reply to this email directly, view it on GitHub https://github.com/petrelharp/ftprime_ms/issues/7#issuecomment-349726135, or mute the thread https://github.com/notifications/unsubscribe-auth/AGHnH_JG9r5sYY_iGL3Ep8dwosM55miHks5s9thVgaJpZM4QBRnU .

--

Kevin Thornton

Associate Professor

Ecology and Evolutionary Biology

UC Irvine

http://www.molpopgen.org

http://github.com/ThorntonLab

http://github.com/molpopgen

molpopgen commented 6 years ago

I just started simuPOP runs with ARG tracking and selection for more 4Nr values and N=1e4. Then, it'll be on to the sims w/no selection.

ashander commented 6 years ago

Annoying issue (which will not affect the benchmarks) reported by @molpopgen with this attempt to write a logfile that would let us know how far a killed simulation got. It appears that line does not actually write a line to the logfile every 100 generations as intended, but only writes after .evolve() exits.

molpopgen commented 6 years ago

Annoying issue (which will not affect the benchmarks) reported by @molpopgen with this attempt to write a logfile that would let us know how far a killed simulation got. It appears that line does not actually write a line to the logfile every 100 generations as intended, but only writes after .evolve() exits.

Regarding this:

I am running N=1e4 with selection and tracking neutral mutations rather than the ARG. This will abort at some point due to being out of RAM.

After this, I will do N=1e3 with no selection. Unless there is serious objection, I think we can skip N=1e4 w/o selection. The results so far WITH arg tracking suggest that a 512Gb machine is not enough for simuPOP!

ashander commented 6 years ago

Amazing! Sadly not in a good way.

Sounds like a good plan to me. @petrelharp ? On Mon, Dec 11, 2017 at 14:50 Kevin R. Thornton notifications@github.com wrote:

Annoying issue (which will not affect the benchmarks) reported by @molpopgen https://github.com/molpopgen with this attempt to write a logfile that would let us know how far a killed simulation got. It appears that line does not actually write a line to the logfile every 100 generations as intended, but only writes after .evolve() exits.

Regarding this:

I am running N=1e4 with selection and tracking neutral mutations rather than the ARG. This will abort at some point due to being out of RAM.

After this, I will do N=1e3 with no selection. Unless there is serious objection, I think we can skip N=1e4 w/o selection. The results so far WITH arg tracking suggest that a 512Gb machine is not enough for simuPOP!

— You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub https://github.com/petrelharp/ftprime_ms/issues/7#issuecomment-350884737, or mute the thread https://github.com/notifications/unsubscribe-auth/AAfLOGjnQco6A1S35392xlXZ5Un9znTlks5s_bHBgaJpZM4QBRnU .

-- -Jaime

petrelharp commented 6 years ago

Sounds good.

On Mon, Dec 11, 2017 at 3:29 PM, Jaime Ashander notifications@github.com wrote:

Amazing! Sadly not in a good way.

Sounds like a good plan to me. @petrelharp ? On Mon, Dec 11, 2017 at 14:50 Kevin R. Thornton notifications@github.com wrote:

Annoying issue (which will not affect the benchmarks) reported by @molpopgen https://github.com/molpopgen with this attempt to write a logfile that would let us know how far a killed simulation got. It appears that line does not actually write a line to the logfile every 100 generations as intended, but only writes after .evolve() exits.

Regarding this:

I am running N=1e4 with selection and tracking neutral mutations rather than the ARG. This will abort at some point due to being out of RAM.

After this, I will do N=1e3 with no selection. Unless there is serious objection, I think we can skip N=1e4 w/o selection. The results so far WITH arg tracking suggest that a 512Gb machine is not enough for simuPOP!

— You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub https://github.com/petrelharp/ftprime_ms/issues/ 7#issuecomment-350884737, or mute the thread https://github.com/notifications/unsubscribe-auth/ AAfLOGjnQco6A1S35392xlXZ5Un9znTlks5s_bHBgaJpZM4QBRnU .

-- -Jaime

— You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub https://github.com/petrelharp/ftprime_ms/issues/7#issuecomment-350893178, or mute the thread https://github.com/notifications/unsubscribe-auth/AA_26alq7xcPZ41cqGaGKZtkoldnmwPWks5s_br2gaJpZM4QBRnU .