TELEMAC-MASCARET Forum: Simulation is slower on new OS. (1/1)

Welcome, Guest

TOPIC: Simulation is slower on new OS.

Simulation is slower on new OS. 2 years 11 months ago #39361

Htun Pyae Sone
OFFLINE
Junior Boarder
Posts: 38
Thank you received: 2

Dear all,

Our university recently updated the HPC-OS from CentOS7 to RedHatOS8 and I compiled the Telemac from the scratch. The only changes in the configuration file is that I remove "-machinefile MPI_HOSTFILE" flag from mpirun command according to Yugi's post #38249. I used the v7p3r1 for both versions of OS. In OS7, the computation time is about 550 times faster than the simulation time with 112 cores. However, in OS8, it is only about 80 times faster which is significantly slower. I run for the same simulation case for both compilation and I am using solving scheme 14. I check for the Iteration numbers for both and they are in the similar range. What could be the possible reason? Thank you very much.

Best regards,
Htun.

The administrator has disabled public write access.

Simulation is slower on new OS. 2 years 11 months ago #39383

Htun Pyae Sone OFFLINE Junior Boarder Posts: 38 Thank you received: 2	It was solved by switching back to old version of compilers. Just learned that it can have huge impact on simulation time. Thank you.
	The administrator has disabled public write access.

Simulation is slower on new OS. 2 years 11 months ago #39384

nicogodet OFFLINE Expert Boarder Posts: 157 Thank you received: 39	I'm interested in this. Could you detail a bit ? Compiler versions ?
	The administrator has disabled public write access.

Simulation is slower on new OS. 2 years 11 months ago #39388

Htun Pyae Sone OFFLINE Junior Boarder Posts: 38 Thank you received: 2	I used gcc and openmpi. I will sum up as follow: 1. gcc 8.4.0 and OpenMPI 4.1.1 --> slow. 2. gcc 7.2.0 and OpenMPI 3.0.0 --> fast. I am not sure about the reason since I am not very familiar with compiler.
	The administrator has disabled public write access.

Simulation is slower on new OS. 2 years 11 months ago #39390

sechele

hello - i am a software developer currently tasked with deploying a 2d shallow water equation model for floodplain definition with rainfall forcing.

i am still trying to understand how domain experts use such models and consequently exactly what those words mean

i would be grateful if would you be able to share a little bit about how your university uses this tool, above the undergrad level

thanks

The administrator has disabled public write access.