hipRAND is a RAND marshalling library with multiple supported backends. It sits between your application and the backend RAND library, where it marshals inputs to the backend and results to the application. hipRAND exports an interface that doesn't require the client to change, regardless of the chosen backend.

hipRAND supports rocRAND and cuRAND.

You can find our documentation at hipRAND User Guide/.

To build our documentation, use the following commands:

# Go to hipRAND docs directory
cd hipRAND; cd docs

# Install Python dependencies
python3 -m pip install -r sphinx/requirements.txt

# Build the documentation
python3 -m sphinx -T -E -b html -d _build/doctrees -D language=en . _build/html

# E.g. serve the HTML docs locally
cd _build/html
python3 -m http.server

You must have the following installed to use hipRAND:

• CMake (3.16 or later)
• For AMD GPUs:
  • AMD ROCm Software (5.0.0 or later)
  • rocRAND library
• For NVIDIA GPUs:
  • CUDA Toolkit
  • cuRAND library

You can download pre-built packages by following the ROCm Install Guide, or by clicking the github releases tab (this option could have a newer version).

Once downloaded, use the following command to install hipRAND:

sudo apt update && sudo apt install hiprand

To build hipRAND, you can use the bash helper script (Ubuntu only) or build manually (for all supported platforms):

• Bash helper build script:

  The helper script install is located in the root repository. Note that this method doesn't take many options and hard-codes a configuration that you can specify by invoking CMake directly.

  A few commands in the script need sudo access, so it may prompt you for a password.

  • ./install -h: Shows help
  • ./install -id: Builds library, dependencies, and installs (the -d flag only needs to be passed once on a system)

• Manual build:

  If you use a distribution other than Ubuntu, or want more control over the build process, the hipRAND build wiki has helpful information on how to configure CMake and build manually.

You can find a list of exported functions on the wiki.

The hipRAND interface is compatible with rocRAND and cuRAND-v2 APIs. Porting a CUDA application that calls the cuRAND API to an application that calls the hipRAND API is relatively straightforward. For example, to create a generator:

hiprandStatus_t
hiprandCreateGenerator(
  hiprandGenerator_t* generator,
  hiprandRngType_t rng_type
)

Here is an example that generates a log-normally distributed float from a generator (these functions are templated for all generators).

__device__ double
hiprand_log_normal_double(
  hiprandStateSobol64_t* state,
  double mean,
  double stddev
)