Message: Re: Surprisingly poor performance of Geant4.9.2.p01 on a bi-quad-core Intel xeon machine Not Logged In (login)
 Next-in-Thread Next-in-Thread
 Next-in-Forum Next-in-Forum

None Re: Surprisingly poor performance of Geant4.9.2.p01 on a bi-quad-core Intel xeon machine 

Forum: Installation and Configuration
Re: Question Surprisingly poor performance of Geant4.9.2.p01 on a bi-quad-core Intel xeon machine
Date: 21 Sep, 2009
From: Paul Nicholas Colin Gloster <Paul Nicholas Colin Gloster>

On Mon, 21 Sep 2009, yong.kong@icxt.com submitted "Surprisingly poor
performance of Geant4.9.2.p01 on a bi-quad-core Intel xeon machine":

|-----------------------------------------------------------------------|
|"Hello, everyone!                                                      |
|                                                                       |
|Recently I tried to run my Geant4 simulations on a bi-quad-core Intel  |
|Xeon server. The installation of Geant4.9.2.p01 was sucessful. However,|
|I found that the new Quad-core machine is surprisingly much slower than|
|our old Xeon server. When Geant4 simulation is running, I noticed that |
|the cpu load is very low, only around 3%, and the CPU status (monitored|
|by top) often indicates "D" - "uninterruptable sleeping". I guess some |
|settings related to multi-core arch. might not be correctly configured.|
|Here are some comparisons:                                             |
|                                                                       |
|-------------------------------------------------                      |
|                                                                       |
|New: 2.6.30-gentoo-r4 smp x86_64 Bi-Quad-core Intel Xeon 2.80 Ghz      |
|     Geant4 built for x86_64-pc-linux-gnu                              |
|                                                                       |
|Old: 2.6.24-gentoo-r8 smp i686 intel xeon cpu 2.40 Ghz                 |
|     Geant4 built for i486-pc-linux-gnu                                |
|                                                                       |
|1. examples/novice/N02: 20000 events                                   |
|                                                                       |
|   Old:  User=42.97s Real=42.97s Sys=0s                                |
|   New:  User=18.6s  Real=18.6s Sys=0s                                 |
|                                                                       |
|2. examples/novice/N03: 20000 events                                   |
|                                                                       |
|   Old: User=51.42s Real=52.71s Sys=1.28s                              |
|   New: User=22.04s Real=175.09s Sys=0.98s                             |
|                                                                       |
|3. my own application: 20000 events                                    |
|                                                                       |
|   Old: User=3.1s Real=5.09s Sys=1.99s                                 |
|   New: User=1.86s Real=175.76s Sys=1.12s                              |
|                                                                       |
|----------------------------------------                               |
|                                                                       |
|Does someone have any idea?                                            |
|                                                                       |
|thanks!                                                                |
|                                                                       |
|Yong"                                                                  |
|-----------------------------------------------------------------------|

Hello,

You have made it clear that the newer system was using a 64bit
operating system and that the older system was using a 32bit operating
system. 32bit executables can be faster than 64bit
executables. However, you did not provide enough evidence to
conclusively deduce that the wordlength accounts for your observations
because:
* you have not made it clear that both sets of measurements were for
  Geant4.9.2.p01;
* the Linux kernels do not differ merely by wordlength: they are
  2.6.30-gentoo-r4 and 2.6.24-gentoo-r8;
and
* other potentially influential factors have been left unreported,
  such as the compilers used; and the Geant configuration options used
  when compiling and installing Geant.

Yours sincerely,
Nicholas Paul Colin Gloster

Inline Depth:
 1 1
 All All
Outline Depth:
 1 1
 2 2
 All All
Add message: (add)

1 None: Re: Surprisingly poor performance of Geant4.9.2.p01 on a bi-quad-core Intel xeon machine   (icx719 - 21 Sep, 2009)
(_ None: Re: Surprisingly poor performance of Geant4.9.2.p01 on a bi-quad-core Intel xeon machine   (Paul Nicholas Colin Gloster - 21 Sep, 2009)
(_ None: Re: Surprisingly poor performance of Geant4.9.2.p01 on a bi-quad-core Intel xeon machine   (icx719 - 21 Sep, 2009)
 Add Message Add Message
to: "Re: Surprisingly poor performance of Geant4.9.2.p01 on a bi-quad-core Intel xeon machine"

 Subscribe Subscribe

This site runs SLAC HyperNews version 1.11-slac-98, derived from the original HyperNews


[ Geant 4 Home | Geant 4 HyperNews | Search | Request New Forum | Feedback ]