Re: [Gnuplot-info] [External] fit is slow due to missing/wrong shared libraries?

[Garcia V. R. <Rob...@fl...>]

Indeed. 

The slower gnuplot has also the 4x slower sysbench. 

Okay. That seems to be the problem. Is gnuplot multi core capable?
I wonder why the entire performance of the CPU is not used. 
But yes, that would clearly explain the difference to begin with.

Best

Robin


Robin Garcia Victoria (M.Sc.)
Computer scientist
___________________________________________
Friedrich-Loeffler-Institut
Bundesforschungsinstitut für Tiergesundheit
Federal Research Institute for Animal Health
Höltystr. 10 | 31535 Neustadt
Tel: +49 5034 871-5156 (Kein IT-Support!)
IT-Hotline: 1699
https://www.fli.de/
¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯


________________________________________
Von: Werner Lippert <wer...@pe...>
Gesendet: Freitag, 19. April 2024 18:09
An: Garcia Victoria, Robin
Cc: gnu...@li...
Betreff: Re: [External] [Gnuplot-info] fit is slow due to missing/wrong shared libraries?

ACHTUNG: EXTERNE E-MAIL
Hi Robin,

Do I get it, the box with the slower gnuplot performance is also 4 times slower running sysbench?


Cheers,
Werner
_________________________________________________
Dr. Werner Lippert

Partner
peaq GmbH                                       Mobile +41 79 218 84 26
Neugutstrasse 12                                wer...@pe...
CH-8304 Wallisellen                                          www.peaq.ch
_________________________________________________
Get the most out of your Hitachi Storage Systems
With peaq IOportal, SAM4H, Crosscharging and Lifecycle Services

On 19 Apr 2024, at 18:02, Garcia Victoria, Robin <Rob...@fl...> wrote:

These are my results

root@rie-ex-web160,apt: sysbench --test=cpu run
sysbench 0.4.12:  multi-threaded system evaluation benchmark

Running the test with following options:
Number of threads: 1

Doing CPU performance benchmark

Threads started!
Done.

Maximum prime number checked in CPU test: 10000


Test execution summary:
   total time:                          2.2523s
   total number of events:              10000
   total time taken by event execution: 2.2517
   per-request statistics:
        min:                                  0.22ms
        avg:                                  0.23ms
        max:                                  0.60ms
        approx.  95 percentile:               0.23ms

Threads fairness:
   events (avg/stddev):           10000.0000/0.00
   execution time (avg/stddev):   2.2517/0.00

Apptainer> sysbench cpu run
sysbench 1.0.20 (using system LuaJIT 2.1.0-beta3)

Running the test with following options:
Number of threads: 1
Initializing random number generator from current time


Prime numbers limit: 10000

Initializing worker threads...

Threads started!

CPU speed:
   events per second:  1210.13

General statistics:
   total time:                          10.0007s
   total number of events:              12104

Latency (ms):
        min:                                    0.81
        avg:                                    0.83
        max:                                    1.42
        95th percentile:                        0.86
        sum:                                 9995.59

Threads fairness:
   events (avg/stddev):           12104.0000/0.00
   execution time (avg/stddev):   9.9956/0.00



Robin Garcia Victoria (M.Sc.)<http://scanmail.trustwave.com/?c=4477&d=sJei5iDDBf1wuA_tQXN7IhVnOYCKLGVGGOgi0xDacg&u=http%3a%2f%2fM%2eSc%2e%29>
Computer scientist
___________________________________________
Friedrich-Loeffler-Institut
Bundesforschungsinstitut für Tiergesundheit
Federal Research Institute for Animal Health
Höltystr. 10 | 31535 Neustadt
Tel: +49 5034 871-5156 (Kein IT-Support!)
IT-Hotline: 1699
https://www.fli.de/<https://scanmail.trustwave.com/?c=4477&d=sJei5iDDBf1wuA_tQXN7IhVnOYCKLGVGGOonhkeJdQ&u=https%3a%2f%2fwww%2efli%2ede%2f>
¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯


________________________________________
Von: Werner Lippert <wer...@pe...>
Gesendet: Freitag, 19. April 2024 17:17:39
An: Garcia Victoria, Robin
Cc: gnu...@li...
Betreff: [External] Re: [Gnuplot-info] fit is slow due to missing/wrong shared libraries?

ACHTUNG: EXTERNE E-MAIL
Hi Robin,

You should first run a CPU benchmark on both systems to get a baseline. AMD might be significantly slower than Intel.


Cheers,
Werner
_________________________________________________
Dr. Werner Lippert

Partner
peaq GmbH                                       Mobile +41 79 218 84 26
Neugutstrasse 12                                wer...@pe...
CH-8304 Wallisellen                                          www.peaq.ch
_________________________________________________
Get the most out of your Hitachi Storage Systems
With peaq IOportal, SAM4H, Crosscharging and Lifecycle Services

On 19 Apr 2024, at 15:23, Garcia Victoria, Robin <Rob...@fl...> wrote:

Hello everyone,

I have installed gnuplot 5.0.5 on a VM (on Debian 9) and gnuplot 5.4.4 (Debian 12) in a Singularity container.

The runtimes differ considerably:

On the VM with gnuplot version 5.0.5
real 0m31.142s
user 0m14.280s
sys 0m15.672s

In the container with gnuplot version 5.4.4
real 2m32.341s
user 1m31.325s
sys 1m0.593s

My specs are:
- VM: AMD EPYC 7543P 32-Core Processor and 32GB RAM
- Singularity: Intel(R) Xeon(R) Gold 6154 CPU @ 3.00GHz (72-Core) and 378GB RAM

root@rie-ex-web160,tmp: cat fit_run_gnuplottestlauf_year1950
Set terminal postscript
set output "Inbreeding_log_gnuplottestlauf_year1950.ps"
set ylabel "Coefficient" font "Times-Italic,14"
set xlabel "Days of birth" font "Times-Italic,14"
set style line 1 lt 1 lw 15
set button below
set autoscaling xy
f(x)=a*x+b
FIT_LIMIT =1E-20
fit f(x) "Inbreeding_log_gnuplottestlauf_year1950.csv" u 1:2 via a,b
Plot "Inbreeding_log_gnuplottestlauf_year1950.csv" with 1:2 title "Log of Inbreed Coeff", f(x)

root@rie-ex-web160,tmp: cat Inbreeding_log_gnuplottestlauf_year1950.csv
1 0.0000000000000000 1950
21 0.0000000000000000 1950
73 0.0000000000000000 1950
265 0.0000000000000000 1950

I have checked which shared libraries are used by gnuplot (ldd /usr/bin/gnuplot).

Apptainer>
ldd /usr/bin/gnuplot | wc -l
84

root@rie-ex-web160,~: ldd /usr/bin/gnuplot | wc -l
46

My container, where the fitting with gnuplot runs slower than in the VM, has 84 shared libraries for gnuplot, while 46 shared libraries are used by gnuplot in the VM.

Could this be the reason why the fitting is significantly slower?

I would be grateful for any tips and advice.

Best

Robin

Robin Garcia Victoria (M.Sc.)<http://scanmail.trustwave.com/?c=4477&d=mIui5jG0WgXIqtbgf5hLVlL1pSV3SXy9M8tXpkLvig&u=http%3a%2f%2fM%2eSc%2e%29>
Computer scientist
___________________________________________
Friedrich-Loeffler-Institut
Bundesforschungsinstitut für Tiergesundheit
Federal Research Institute for Animal Health
Höltystr. 10 | 31535 Neustadt
Tel: +49 5034 871-5156 (Kein IT-Support!)
IT-Hotline: 1699
https://www.fli.de/<https://scanmail.trustwave.com/?c=4477&d=sJei5iDDBf1wuA_tQXN7IhVnOYCKLGVGGLtwgRaMIg&u=https%3a%2f%2fwww%2efli%2ede%2f%3e>
¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯



_______________________________________________
gnuplot-info mailing list
gnu...@li...
Membership management via: https://lists.sourceforge.net/lists/listinfo/gnuplot-info<https://scanmail.trustwave.com/?c=4477&d=sJei5iDDBf1wuA_tQXN7IhVnOYCKLGVGGO8i1xTYeA&u=https%3a%2f%2flists%2esourceforge%2enet%2flists%2flistinfo%2fgnuplot-info%3e>