Difference between revisions of "GEOS-Chem 1-month benchmark timing results"
(→Timing results from 1-month benchmarks) |
(→Wall time spent in model components) |
||
(6 intermediate revisions by one other user not shown) | |||
Line 14: | Line 14: | ||
== Timing results from 1-month benchmarks == | == Timing results from 1-month benchmarks == | ||
+ | |||
+ | To access the entire history of GEOS-Chem benchmarks, please see our [[GEOS-Chem versions]] page. | ||
+ | |||
+ | === Wall time spent in model components === | ||
[[Image:GEOS-Chem Timers from 1-month benchmarks.png]] | [[Image:GEOS-Chem Timers from 1-month benchmarks.png]] | ||
+ | |||
+ | The above plot displays the "wall clock" time spent in each model component as observed in the [[GEOS-Chem timers]] output from recent public and internal 1-month benchmark simulations (Credit: Melissa Sulprizio). This type of plot can be very helpful in determining if updates made to GEOS-Chem cause significant computational bottlenecks. | ||
+ | |||
+ | Timing information is obtained from the GEOS-Chem "Timers" output, which is printed to the log file. Timers output is activated when compiling GEOS-Chem with <tt>TIMERS=y</tt> (GNU Make) or <tt>-DTIMERS=y</tt>. | ||
+ | |||
+ | --[[User:Bmy|Bob Yantosca]] ([[User talk:Bmy|talk]]) 20:58, 21 February 2020 (UTC) | ||
+ | |||
+ | === Overall wall time === | ||
[[Image:GC_1mon_benchmark_timing.png]] | [[Image:GC_1mon_benchmark_timing.png]] | ||
The above plot displays the "wall clock" time from recent [[GEOS-Chem_benchmarking#1-month_benchmark|1-month GEOS-Chem benchmarks]]. All results displayed are from simulations that ran on all 24 CPUs of the holyjacob*.rc.fas.harvard.edu nodes (Intel(R) Xeon(R) CPU E5-2680 v3 @ 2.50 GHz) on Harvard's Odyssey cluster. | The above plot displays the "wall clock" time from recent [[GEOS-Chem_benchmarking#1-month_benchmark|1-month GEOS-Chem benchmarks]]. All results displayed are from simulations that ran on all 24 CPUs of the holyjacob*.rc.fas.harvard.edu nodes (Intel(R) Xeon(R) CPU E5-2680 v3 @ 2.50 GHz) on Harvard's Odyssey cluster. | ||
− | |||
− | |||
--[[User:Melissa Payer|Melissa Sulprizio]] ([[User talk:Melissa Payer|talk]]) 13:14, 20 September 2018 (UTC) | --[[User:Melissa Payer|Melissa Sulprizio]] ([[User talk:Melissa Payer|talk]]) 13:14, 20 September 2018 (UTC) |
Revision as of 16:40, 26 May 2020
Previous | Next | Guide to GEOS-Chem performance
- Parallelizing GEOS-Chem
- GEOS-Chem 7-day timing tests
- GEOS-Chem scalability
- GEOS-Chem 1-month benchmark timing results
- Profiling GEOS-Chem with the TAU performance system
- Speeding up GEOS-Chem
Contents
Overview
On this page, we provide timing results obtained from 1-month benchmark GEOS-Chem simualtions.
Timing results from 1-month benchmarks
To access the entire history of GEOS-Chem benchmarks, please see our GEOS-Chem versions page.
Wall time spent in model components
The above plot displays the "wall clock" time spent in each model component as observed in the GEOS-Chem timers output from recent public and internal 1-month benchmark simulations (Credit: Melissa Sulprizio). This type of plot can be very helpful in determining if updates made to GEOS-Chem cause significant computational bottlenecks.
Timing information is obtained from the GEOS-Chem "Timers" output, which is printed to the log file. Timers output is activated when compiling GEOS-Chem with TIMERS=y (GNU Make) or -DTIMERS=y.
--Bob Yantosca (talk) 20:58, 21 February 2020 (UTC)
Overall wall time
The above plot displays the "wall clock" time from recent 1-month GEOS-Chem benchmarks. All results displayed are from simulations that ran on all 24 CPUs of the holyjacob*.rc.fas.harvard.edu nodes (Intel(R) Xeon(R) CPU E5-2680 v3 @ 2.50 GHz) on Harvard's Odyssey cluster.
--Melissa Sulprizio (talk) 13:14, 20 September 2018 (UTC)
For more information