Issue17 GrayBat #58
Merged
Issue17 GrayBat #58
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
| @@ -0,0 +1,74 @@ | |||
| /** | |||
| * Copyright 2013 Erik Zenker, Carlchristian Eckert, Marius Melzer | |||
There was a problem hiding this comment.
copyright should be 2015 now
|
About the CI-testing: you could edit the About your questions: How can GrayBat be integrated into haseongpu best (submodule etc.)That one is pretty difficult. As a first step, I would suggest a Is it necessary to encapsulate master and slave code into functionsMaybe not necessary, but it would make the loop-structure much more readable. Do we need non-blocking in this first version (so the master could also do computations)I don't think so, we have an issue for that (#13). The real question is: do we need to throw away all the blocking code if we want to do it properly later? Because that would be a lot of unnecessary work and we might do it non-blocking from the start Can the code readability be improved somehowWhat I like about the readability:
What could be improved:
for(Vertex v : cage.hostedVertices){
if(v == master){
// loops internally over all samples and also sends abort-messages in the end
masterFunction(v, samples);
}else{
// internally loops forever until abort-message received
slaveFunction(v);
}
} |
|
|
||
| // Create sample indices | ||
| std::vector<unsigned> samples(mesh.numberOfSamples); | ||
| std::generate(samples.begin(), samples.end(), Increment(0)); |
There was a problem hiding this comment.
This is not needed anymore, we can use
std::iota(samples.begin(), samples.end(), 0);(from algorithms which does exactly the same (and does not need a functor!))
It is c++11, but that is no problem now :)
|
Completed to work on your comments, but to conclude this pull request I first need to transform GrayBat into a clean library stucture and add it to haseongpu as a subtree depedency. I will also try to fix issue #9 by grouping all interface data into some ExperimentDescription type Furthermore, GrayBat is able to solve issue #13. I need to further investigate if this issue is totally covered by the current source state. With GrayBat, issue #12 would become obsolet, since MPI code is then not needed anymore ! |
|
so many nice things!
|
|
Oh yeah I like them too 🎈 |
|
🐳 |
|
A lot of commits later:
|
|
a(we)s(om)e |
added 4 commits
June 15, 2015 20:43
# The first commit's message is: New parallel mode: Graybat (issue ComputationalRadiationPhysics#17) * Abstraction from communication method * Equal behavior than MPI mode * Can be choosen by parallelMode=graybat * Added as subtree in include/graybat The calcPhiAse interface was reduced 4 structs (issue ComputationalRadiationPhysics#9) The threaded code was updated to C++11 threads # This is the 2nd commit message: threading code now C++11 # This is the 3rd commit message: Finished graybat hook
git-subtree-dir: include/graybat git-subtree-split: 13c673e720668804c50381e0fa6811e0b1cd490f
erikzenker
force-pushed
the
issue17-graybat
branch
from
69c597e
to
185d6a0
Compare
|
Git subtree and rebase are no good friends, so I will leave it with this five commits, but this is ne main commit ...we om ? |
|
My first thoughts for now:
There will be a more detailed review of code lines tomorrow. Probably also a full simulation with MATLAB to see if results are correct :) |
|
Unfortunately, testing on Hypnos found some problems and/or smaller annoyances build + environmentThis is probably more a list of TODO stuff (might need an issue)
setenv('LD_LIBRARY_PATH', ':/opt/pkg/devel/boost/1.56.0/gnu/4.8.2/64/opt/lib:/opt/pkg/mpi/openmpi/1.7.4/gnu/4.8.2/64/opt/lib:/opt/pkg/infiniband/psm/3.1/lib64:/opt/pkg/devel/cuda/6.5/lib64:/opt/pkg/devel/cuda/6.5/lib:/opt/pkg/numlib/mpfr/3.1.2/lib:/opt/pkg/numlib/mpc/1.0.1/lib:/opt/pkg/numlib/gmp/5.1.1/lib:/opt/torque/lib');codeThis is a more serious list of problems
|
|
Excessive testing is finished.
|
|
Could you please document your test results? |
|
I will gladly do so. Is there a preferred form of documentation? |
This was referenced Jun 22, 2015
|
I think posting those numbers here would be o.k., but in addition link to the code & version that produced these results |
|
The following tests were used to verify that the introduction of GrayBat didn't introduce any new value errors. SetupThe baseline used plain MPI for communication and the previous dev tip (268b6cd). The new code used GrayBat as introduced by this pull request (9734a6d). The programs were both executing the MATLAB example (in the respective commits, see the folder Each of the examples was launched on the Hypnos cluster, using a submit-file of the following form: #!/bin/bash
#PBS -q k20
#PBS -l nodes=2:ppn=8
#PBS -l walltime=96:00:00
#PBS -N HASE_GB_integration_testing
## ENVIRONMENT ############################
. /opt/modules-3.2.6/Modules/3.2.6/init/bash
export MODULES_NO_OUTPUT=1
module load ~/own.modules.kepler
export -n MODULES_NO_OUTPUT
uname -a
echo " "
cd ~/haseongpu_integration_GrayBat
matlab -r laserPumpCladdingExampleRun with MPI:
Run with GrayBat
ResultsTo obtain gain values for a plot of gain over time in the context of the experiment that was simulated, the values in the table below need to be processed with The results of both runs are in strong agreement with the experimental measurements.
|
Merged
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
A first try to use GrayBat as framework for communication in
haseongpu. The lines of code were reduced in comparison
to the mpi case and I think it is more intuitive. Some
code review would be nice, regarding:
IMPORTANT: This will not compile, since, GrayBat needs to be clone into the include directory:
cd include
git clone -b topic-haseongpu https://github.com/erikzenker/GrayBat.git