About the "on the fly" mode

Post here questions linked with issue while running the EPW code

Moderator: stiwari

Post Reply
leoi
Posts: 1
Joined: Thu Oct 13, 2022 3:10 pm
Affiliation: ISSP (CAS), USTC

About the "on the fly" mode

Post by leoi »

Dear experts,

I am using EPW to calculate superconductor's properties. Because my RAM is not very large (64 cores and 512 Gb RAM), the anisotropic equation always solves "on the fly". When I use "top" command to see the memory epw.x using in this mode, I saw epw.x only use ~400 MB per core.

So, I have two questions. The first one is what the difference between normal mode and "on the fly" mode? Is this mode storing things to the hard drive instead of the cache to reduce memory?
Another question is can I enlarge the memory "on the fly" mode using? Because in this mode, it actually can use 8Gb per core.

Best regards,
R. Lv



####################
temp( 1) = 2.00000 K

Solve anisotropic Eliashberg equations on imaginary-axis

Total number of frequency points nsiw( 1) = 923
Cutoff frequency wscut = 1.0000

Size of required memory per pool: ~= 11.2659 Gb

akeri is calculated on the fly since its size exceedes max_memlt
######################

####################
[user@login]$ top
top - 18:09:38 up 2 days, 3:07, 1 user, load average: 64.05, 64.03, 63.61
Tasks: 723 total, 65 running, 658 sleeping, 0 stopped, 0 zombie
%Cpu(s):100.0 us, 0.0 sy, 0.0 ni, 0.0 id, 0.0 wa, 0.0 hi, 0.0 si, 0.0 st
GiB Mem : 503.1 total, 455.5 free, 32.8 used, 14.8 buff/cache
GiB Swap: 128.0 total, 127.6 free, 0.4 used. 468.0 avail Mem

PID USER PR NI VIRT RES SHR S %CPU %MEM TIME+ COMMAND
52800 user 20 0 5952.6m 468.5m 35.1m R 100.0 0.1 68:06.75 epw.x
52804 user 20 0 5952.6m 462.9m 35.3m R 100.0 0.1 68:33.92 epw.x
52819 user 20 0 5954.6m 468.6m 35.1m R 100.0 0.1 68:29.30 epw.x
52824 user 20 0 5952.6m 460.9m 35.3m R 100.0 0.1 68:22.31 epw.x
52830 user 20 0 5950.6m 461.0m 35.5m R 100.0 0.1 68:14.06 epw.x
52849 user 20 0 5950.3m 456.6m 35.3m R 100.0 0.1 68:24.58 epw.x
52851 user 20 0 5948.3m 458.8m 35.5m R 100.0 0.1 68:26.69 epw.x
52792 user 20 0 5954.7m 463.0m 35.3m R 99.7 0.1 68:27.68 epw.x
52793 user 20 0 5950.6m 458.6m 35.1m R 99.7 0.1 68:19.89 epw.x
52794 user 20 0 5952.6m 463.1m 35.5m R 99.7 0.1 68:29.47 epw.x
52795 user 20 0 5950.6m 462.9m 35.2m R 99.7 0.1 68:36.27 epw.x
####################

hpaudya1
Posts: 190
Joined: Tue Mar 21, 2017 7:11 pm
Affiliation:

Re: About the "on the fly" mode

Post by hpaudya1 »

Hi R. Lv,

On the fly mode meaning we calculate anisotropic lambda (eqn. 18 of https://journals.aps.org/prb/pdf/10.110 ... .87.024505) for every k, q, bands, and frequencies (do not save in the memory) when it is required.

In your case, I recommend using "max_memlt = 8" flag in your input file, for more details https://docs.epw-code.org/doc/Inputs.html#max-memlt

In addition, It looks like 8 Gb is not even enough for your wscut = 1 at 1 K temperature, first; try lowering wscut, and second try running at higher temp. ex 2 K.

Happy EPWing,
Hari

Post Reply