ظ: Run time gadget2 vs. gadget4

From: HU, Rui <1155168718_at_link.cuhk.edu.hk>
Date: Tue, 13 Sep 2022 12:43:23 +0000

Hi Xiaolong,

Thanks a lot for your help. I checked the cpu.txt file, but actually the most expensive part is treegrav. For example, I used 4 CPUs to run G2/G4, in G2, it spend 2014s/2762s for treegrav and only 235s/2762s for PM. But in G4, it spends 5301s/6154s for treegrav and only 304s/6154s for PM calculation.

Best,
Rui

: Xiaolong Du <xdu_at_carnegiescience.edu>
ʱ: 2022910 20:44
ռ: Gadget General Discussion <gadget-list_at_MPA-Garching.MPG.DE>
: Re: [gadget-list] Run time gadget2 vs. gadget4

Hi Rui,

I have seen similar things when comparing Gadget 2 and Gadget 4. It seems that if one switches on PMGRID, Gadget 4 tends to update the long range force more frequently than Gadget 2. Maybe that explains why Gadget 4 took more time. One thing you can check is the time for PM computation in the log file cpu.txt.

Best,
Xiaolong


On Sep 10, 2022, at 12:49 AM, HU, Rui <1155168718_at_link.cuhk.edu.hk<mailto:1155168718_at_link.cuhk.edu.hk>> wrote:

Hi all,

I am try to use the same IC (generated by 2LPTic), same setting and same threads in gadget2 and gadget4. Then I found gadget4 (2300s for 16 threads) is almost twice as slow as gadget2 (1226s for 16 threads). Is that because gadget4 do not do the merge and do two half-step kicks? Or maybe I misunderstand the Figure 61 in the Gadget4-code paper?

Bests,
Rui


-----------------------------------------------------------

If you wish to unsubscribe from this mailing, send mail to
minimalist_at_MPA-Garching.MPG.de<mailto:minimalist_at_MPA-Garching.MPG.de> with a subject of: unsubscribe gadget-list
A web-archive of this mailing list is available here:
http://www.mpa-garching.mpg.de/gadget/gadget-list

Received on 2022-09-13 14:43:45

This archive was generated by hypermail 2.3.0 : 2023-01-10 10:01:33 CET