Notes on string split #1
Open
Conversation
This file contains hidden or 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
psahabu
pushed a commit
that referenced
this pull request
Aug 23, 2016
Fix issues with test/llvmDebug not working with paratest
psahabu
pushed a commit
that referenced
this pull request
Nov 24, 2016
This simple test shows the problem:
const targetLocales => setupLocaleGrid();
proc main() {
var copy = targetLocales;
writeln(copy);
}
proc setupLocaleGrid() {
return reshape(Locales#1, {0..#1, 0..#1}); // does not work
//return reshape(Locales, {0..#1, 0..#1}); // neither does this
}
This test does not currently work with the removal of array
reference counting. I'm not sure what the issue is, but since
arrays now return by value, there is no reason to use the '=>'
syntax when initializing targetLocales. Replacing it with '='
allows the test to function and so that is what I did in these
two HPL variants.
psahabu
pushed a commit
that referenced
this pull request
Nov 24, 2016
Fasta parallel study version This adds a parallel version of fasta based on the #1 entry to my study directory. The main changes that this version has from the release version: * It uses the same parallelization strategy as the current #1 entry (fasta gcc chapel-lang#7) which is to create a bunch of tasks and have them take turns locking the RNG and output stream before using them to avoid races. From my very rough experiments, it didn't seem like this really accelerated stuff much, at least given our other performance constraints. I added an execopts file for correctness testing because the default parallel chunk size doesn't actually do any chunking of the default correctness problem size and was masking a lot of bugs in my early implementations for a long time. * In accumulating the probabilities, it converts them to integers rather than storing them as floating point variables to reduce the amount of floating point math in the inner loop. This is the one thing I did that did seem to have a positive effect on performance. More code cleanup is necessary before submitting this to the website... Possible next steps: * switch from binary search to linear search for simplicity, and because the reference version does * see how much time is spent in the RNG, prob->char conversion, and I/O and focus on the relevant portion of the computation.
psahabu
pushed a commit
that referenced
this pull request
Nov 24, 2016
As I improve my study version of fasta, my plan is to propagate changes into the release version as they prove profitable. Here, I'm incorporating the change (used by the #1 entry) of having the RNG generate ints rather than floats, and to pre-scan the cumulated probabilities to convert them from floating point values to int values in order to avoid floating point math in the inner loop. As part of this change, I did some refactorings, some necessary, some for aesthetics: * moved the sumProbs() function into the randomMake() routine since it's related to that routine. * removed the nested addLine() procedure which seemed pretty pointless. * moved the RNG parameters to config params, in part because one of them has to be used in multiple scopes now, in part because these characterize the benchmarks. These changes correspond to the ones I made in my study version in PR parallelism adds a lot of code overhead and didn't result in a clear win as far as I could tell, so I'm going to save that one for last in the release version after making other improvements that seem to help my study version.
psahabu
pushed a commit
that referenced
this pull request
Nov 24, 2016
…ccum-probs Improve release fasta: use int probabilities/randoms [ratified by @lydia-duncan ] As I improve my study version of fasta, my plan is to propagate changes into the release version as they prove profitable. Here, I'm incorporating the change (used by the #1 entry) of having the RNG generate ints rather than floats, and to pre-scan the cumulated probabilities to convert them from floating point values to int values in order to avoid floating point math in the inner loop. As part of this change, I did some refactorings, some necessary, some for aesthetics: * moved the sumProbs() function into the randomMake() routine since it's related to that routine. * removed the nested addLine() procedure which seemed pretty pointless. * moved the RNG parameters to config params, in part because one of them has to be used in multiple scopes now, in part because these characterize the benchmarks. These changes correspond to the ones I made in my study version in PR parallelism adds a lot of code overhead and didn't result in a clear win as far as I could tell, so I'm going to save that one for last in the release version after making other improvements that seem to help my study version.
psahabu
pushed a commit
that referenced
this pull request
Mar 27, 2017
Update chplspell dictionary for doc/ -> doc/rst moves
This updates every entry in the chplspell dictionary that refers to a
file in ``doc/`` to instead reference ``doc/rst``.
Also remove entries for files that have been removed from the repository:
```
chips/tmp/source/modules/*
studies/hpcc/{RA,appendix}.tex
```
psahabu
pushed a commit
that referenced
this pull request
Jun 28, 2017
I'm not sure whether or not there is more that I'm missing, but this gets us a long way on our next #1 leaking test, schurComplement, and passes correctness testing for all tests that use DimensionalDist2D.
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
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.
No description provided.