AugurV2

AugurV2 generates compositional MCMC inference algorithms for a simple probabilistic modeling language.

Dependencies

AugurV2 additional depends on:

• GSL - GNU scientific library (tested with version 2.3)
• clang - C compiler
• Cuda - Cuda Toolkit (tested with version 7.5)

Installation

Compiler

Option 1: Build the AugurV2 compiler from source (using cabal).

$cd compiler/augur
$cabal sandbox init
$cabal install --dependencies-only
$cabal build

Option 2: Build the AugurV2 compiler from source (using stack).

$cd compiler/augur
$stack init --solver
$stack build

Runtime

Build the AugurV2 runtime.

$cd cbits
$make libcpu

If you have Cuda support with dynamic parallelism (architecture >= sm_35), you can also build the GPU library.

$make libgpu

Interface

Build the AugurV2 Python interface.

$cd pyaugur
$python setup.py install

Todos

• release optimizations

License

Apache-2.0