search

Warwick-Plasma / epoch

Particle-in-cell code for plasma physics simulations
https://epochpic.github.io
GNU General Public License v3.0
184 stars 58 forks source link

Will the verification of the author's computer affect the result of the calculation? #716

Closed bear-maker closed 1 month ago

bear-maker commented 1 month ago

Dear Professor, I ran epoch on the same computer with the same line of code, but the number of mpiruns was different and the results were different. What is the reason?

TomGoffrey commented 1 month ago

What do you mean by "the number of mpiruns was different"? Different number of MPI processes?

Also, when you say the results were different, could you be more specific? Depending on choices of random seed and differing domain decomposition small differences between runs would be expected.

bear-maker commented 1 month ago

What do you mean by "the number of mpiruns was different"? Different number of MPI processes?

Also, when you say the results were different, could you be more specific? Depending on choices of random seed and differing domain decomposition small differences between runs would be expected.

1727618250468

1727618250468

TomGoffrey commented 1 month ago

OK, number of MPI processes. What differences do you observe?

bear-maker commented 1 month ago

OK, number of MPI processes. What differences do you observe?

Thank you for your help. Let me elaborate on this matter in detail. Comparing the initial electron density of a certain grid, for example, the result I obtained with 80MPI is 2.1e27/cm3, while the result obtained with 40MPI is 2.09e27/cm3. This leads to different results in subsequent calculations. For example, electric and magnetic fields. A friend told me that I did not enable multi-threaded computation in the code.

Status-Mirror commented 1 month ago

This is a duplicate, see issue #721.