[root]/src/dr/evomodel/newtreelikelihood
Author | Changes | Lines of Code | Lines per Change |
---|---|---|---|
Totals | 163 (100.0%) | 9413 (100.0%) | 57.7 |
rambaut | 111 (68.1%) | 6753 (71.7%) | 60.8 |
msuchard | 47 (28.8%) | 2393 (25.4%) | 50.9 |
alexei.drummond | 4 (2.5%) | 263 (2.8%) | 65.7 |
jheled | 1 (0.6%) | 4 (0.0%) | 4.0 |
uncommitting an erroneous revision
0 lines of code changed in 6 files:
Refactoring geo-specific functionality into dr.geo (step 1). Also, adding in shell classes for many different spatial distributions
0 lines of code changed in 3 files:
Moving BEAGLE/BEAST out of the main package into dr.app. This will allow us to deploy the jar file separately from BEAST.
5 lines of code changed in 8 files:
EBSP parsing fix
2 lines of code changed in 1 file:
Converted Autocorrelated Relaxed clock models to use new TreeParameterModel so that they have a similar pattern to Uncorrelated Relaxed clock models in the XML. Neither model now uses the <nodeRates> facility in TreeModel. Hopefully I haven't slowed anything down -- the test XML file suggests it gets the right answers at least!!
44 lines of code changed in 1 file:
code police
4 lines of code changed in 1 file:
I have moved all the new stuff for the 'BEAGLE' library into its own package. I am producing a set of parsers that override the default ones to use the new classes. I envisage that at some point we will want to move to the new SubstModel hierarchy as it is much cleaner. But to avoid breaking anything whilst this develops, I am compartmentalizing them.
8 lines of code changed in 9 files:
More BEAGLE tweaks (BEAGLE is the name we (Marc & Andrew) have given to the cross-platform, cross-application, CTMC likelihood evaluation library. The idea is that it produces a fast, abstract API that can be used by many programs and which can have GPU implementations (CUDA, OpenCL) as well as just fast native ones.
236 lines of code changed in 3 files:
Integrated BEAGLE into BEAST (compiles and runs but crashes in native code).
598 lines of code changed in 3 files:
Made a few changes about how parsers are dealt with. The BeastParser class can now be given a list of additional parser lists to attempt to load. For example the command line option '-vector' now adds the parser list vector_parsers.properties to the mix. In addition, parsers in the additional lists can override previously loaded ones. In vector_parsers.properties is a parser called dr.evomodel.newtreelikelihood.TreeLikelihood that overrides the default TreeLikelihood with the new one that can use the GPU. This means we can switch on GPU using the -vector command line.
32 lines of code changed in 2 files:
Created a 'BranchSiteModel' whose job is to provide an EigenDecomposition for any branch/category. HomogenousBranchSiteModel simply wraps a single SubstitutionModel and a FrequencyModel for all branches/sites.
0 lines of code changed in 2 files:
More partials scaling issues
5 lines of code changed in 1 file:
Messing around with dynamic rescaling of partials in newtreelikelihood. Trying for a rescaling mechanism that only runs after an under/over-flow
75 lines of code changed in 9 files:
Had to add a System.exit to BeastMain and changed numbering of GPU devices to 1,2,3 ...
5 lines of code changed in 4 files:
Parser option to specify GPU device number
16 lines of code changed in 2 files:
Moved ParameterChangeType -> Parameter.ChangeType
5 lines of code changed in 1 file:
added variable size functionality to Parameter
212 lines of code changed in 1 file:
Added single/multi locus option for extended BSP. GUI only at the moment - generation needs to be added for multi-loci.
4 lines of code changed in 1 file:
Hidden codons run under GPU, substitution model is still a mess
19 lines of code changed in 3 files:
Started preparing for multiple partitions on a GPU and multiple GPUs driven in different CPU threads
184 lines of code changed in 2 files:
GPU initialization no longer spits out NativeLikelihoodCore message
7 lines of code changed in 2 files:
Added flag to LikelihoodCoreFactory to set precision
6 lines of code changed in 1 file:
Fixed a couple of build errors
26 lines of code changed in 1 file:
Restored FGLC which I accidentally deleted previously; may now return results similar to GPU
410 lines of code changed in 2 files:
Reversion
1 lines of code changed in 1 file:
Packaged the whole pattern compression algorithm into a LikelihoodCore - this is still a Java version. Some initial testing before deciding whether it is worth doing a native version.
135 lines of code changed in 5 files:
Created an experimental system for compressing site patterns for individual subtrees. At the moment this is entirely implemented in Java. Some profiling will be needed to see if the overhead of the compression is less than the gain in speed. No attempt to optimize the code has been attempted yet.
1254 lines of code changed in 3 files:
Cleaning up and testing
48 lines of code changed in 7 files:
Reads GPU info correctly
60 lines of code changed in 2 files:
Making stateTipCount compatible with GPU code
44 lines of code changed in 2 files:
Added the optimization of using fixed states at tips.
283 lines of code changed in 10 files:
Added back in partialsScaling for debug scaling on GPU
86 lines of code changed in 1 file:
Modified GPULikelihoodCore wrapped to pass padded state information to native code
20 lines of code changed in 1 file:
Need to manually convert double[] freqs -> float[] freqs; array-copy does not work.
35 lines of code changed in 1 file:
14 lines of code changed in 3 files:
Created a likelihood core that uses floats to test precision management techniques.
405 lines of code changed in 1 file:
Wrote Java portion of GPULikelihoodCore.
54 lines of code changed in 3 files:
A few adjustments about what is passed from the Eigen decomposition to the likelihood core.
16 lines of code changed in 2 files:
NewNativeLikelihood now works.
1 lines of code changed in 1 file:
NewNativeLikelihood now runs but gives a NaN. Onwards and upwards...
15 lines of code changed in 3 files:
(18 more)