Instead of installing from source, which is complicated and time consuming, MAGMA can now be installed via conda.
Changes in this pull request:
The conda environment file and the installation documentation have been updated with info on installing MAGMA with conda.
bayeseor.gpu.GPUInterface now loads conda's libmagma.so by default.
bayeseor.posterior.PowerSpectrumPosteriorProbability now uses MAGMA functions from the GPU interface class (magma_init, magma_zpotrf, and magma_finalize) to perform the Cholesky decomposition.
Testing
I ran a test using the test data provided with the repo in a new environment using MAGMA installed with conda via conda-forge. The results agree with the results from an analysis with a separate environment where I installed MAGMA from source.
Instead of installing from source, which is complicated and time consuming, MAGMA can now be installed via conda.
Changes in this pull request:
bayeseor.gpu.GPUInterface
now loads conda'slibmagma.so
by default.bayeseor.posterior.PowerSpectrumPosteriorProbability
now uses MAGMA functions from the GPU interface class (magma_init
,magma_zpotrf
, andmagma_finalize
) to perform the Cholesky decomposition.Testing
I ran a test using the test data provided with the repo in a new environment using MAGMA installed with conda via conda-forge. The results agree with the results from an analysis with a separate environment where I installed MAGMA from source.