rocRAND

The rocRAND project provides functions that generate pseudorandom and quasirandom numbers. The rocRAND library is implemented in the HIP programming language and optimized for AMD's latest discrete GPUs. It is designed to run on top of AMD's ROCm runtime, but it also works on CUDA-enabled GPUs.

Prior to ROCm version 5.0, this project included the hipRAND wrapper. As of version 5.0, it was split into a separate library. As of version 6.0, hipRAND can no longer be built from rocRAND.

Supported random number generators

• XORWOW
• MRG31k3p
• MRG32k3a
• Mersenne Twister (MT19937)
• Mersenne Twister for Graphic Processors (MTGP32)
• Philox (4x32, 10 rounds)
• LFSR113
• Sobol32
• Scrambled Sobol32
• Sobol64
• Scrambled Sobol64
• ThreeFry

Documentation

Documentation for rocRAND is available at https://rocm.docs.amd.com/projects/rocRAND/en/latest/

To build documentation locally, use the following code:

# Go to the docs directory
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

Requirements

• CMake (3.16 or later)
• C++ compiler with C++17 support to build the library.
  • Recommended to use at least gcc 9
  • clang uses the development headers and libraries from gcc, so a recent version of it must still be installed when compiling with clang
• C++ compiler with C++11 support to consume the library.
• For AMD platforms:
  • ROCm (1.7 or later)
  • HIP-clang compiler, which must be set as C++ compiler on ROCm platform.
• For CUDA platforms:
  • HIP
  • Latest CUDA SDK
• Python 3.6 or higher (HIP on Windows only, only required for install script)
• Visual Studio 2019 with clang support (HIP on Windows only)
• Strawberry Perl (HIP on Windows only)

Optional:

• GoogleTest (required only for tests; building tests is enabled by default)
  • Use GTEST_ROOT to specify the GoogleTest location (see also FindGTest)
  • Note: If GoogleTest is not already installed, it will be automatically downloaded and built
• Fortran compiler (required only for Fortran wrapper)
  • gfortran is recommended
• Python 3.5+ (required only for Python wrapper)
• doxygen to build the documentation

If some dependencies are missing, the CMake script automatically downloads, builds, and installs them. Setting the DEPENDENCIES_FORCE_DOWNLOAD option to ON forces the script to download all dependencies, rather than using the system-installed libraries.

Build and install

git clone https://github.com/ROCm/rocRAND.git

# Go to rocRAND directory, create and go to build directory
cd rocRAND; mkdir build; cd build

# Configure rocRAND, setup options for your system
# Build options: BUILD_TEST (off by default), BUILD_BENCHMARK (off by default), BUILD_SHARED_LIBS (on by default)
# Additionally, the ROCm installation prefix should be passed using CMAKE_PREFIX_PATH or by setting the ROCM_PATH environment variable.
#
# ! IMPORTANT !
# Set C++ compiler to HIP-clang. You can do it by adding 'CXX=<path-to-compiler>'
# before 'cmake' or setting cmake option 'CMAKE_CXX_COMPILER' to path to the compiler.
#
# The python interface do not work with static library.
#
[CXX=hipcc] cmake -DBUILD_BENCHMARK=ON ../. -DCMAKE_PREFIX_PATH=/opt/rocm # or cmake-gui ../.

# To configure rocRAND for NVIDIA platforms, the CXX compiler must be set to a host compiler. The CUDA compiler can
# be set explicitly using `-DCMAKE_CUDA_COMPILER=<path-to-nvcc>`.
# Additionally, the path to FindHIP.cmake should be passed via CMAKE_MODULE_PATH. By default, this is module is
# installed in /opt/rocm/hip/cmake.
cmake -DBUILD_BENCHMARK=ON ../. -DCMAKE_PREFIX_PATH=/opt/rocm -DCMAKE_MODULE_PATH=/opt/rocm/hip/cmake # or cmake-gui ../.
# or
[CXX=g++] cmake -DBUILD_BENCHMARK=ON -DCMAKE_CUDA_COMPILER=/usr/local/cuda/bin/nvcc -DCMAKE_PREFIX_PATH=/opt/rocm -DCMAKE_MODULE_PATH=/opt/rocm/hip/cmake ../. # or cmake-gui ../.

# Build
make -j4

# Optionally, run tests if they're enabled
ctest --output-on-failure

# Install
[sudo] make install

HIP on Windows

We've added initial support for HIP on Windows, which you can install using the rmake.py python script:

git clone https://github.com/ROCm/rocRAND.git
cd rocRAND

# the -i option will install rocPRIM to C:\hipSDK by default
python rmake.py -i

# the -c option will build all clients including unit tests
python rmake.py -c

The existing GoogleTest library in the system (especially static GoogleTest libraries built with other compilers) may cause a build failure; if you encounter errors with the existing GoogleTest library or other dependencies, you can pass the DEPENDENCIES_FORCE_DOWNLOAD flag to CMake, which can help to solve the problem.

To disable inline assembly optimizations in rocRAND (for both the host library and the device functions provided in rocrand_kernel.h), set the CMake option ENABLE_INLINE_ASM to OFF.

Running unit tests

# Go to rocRAND build directory
cd rocRAND; cd build

# To run all tests
ctest

# To run unit tests
./test/<unit-test-name>

Running benchmarks

# Go to rocRAND build directory
cd rocRAND; cd build

# To run benchmark for the host generate functions:
# The benchmarks are registered with Google Benchmark as `device_generate<engine,distribution>`, where
# engine -> xorwow, mrg31k3p, mrg32k3a, mtgp32, philox, lfsr113, mt19937,
#           threefry2x32, threefry2x64, threefry4x32, threefry4x64,
#           sobol32, scrambled_sobol32, sobol64, scrambled_sobol64
# distribution -> uniform-uint, uniform-uchar, uniform-ushort,
#                 uniform-half, uniform-float, uniform-double,
#                 normal-half, normal-float, normal-double,
#                 log-normal-half, log-normal-float, log-normal-double, poisson
# Further option can be found using --help
./benchmark/benchmark_rocrand_host_api
# To run specific benchmarks:
./benchmark/benchmark_rocrand_host_api --benchmark_filter=<regex>
# For example to run benchmarks with engine sobol64:
./benchmark/benchmark_rocrand_host_api --benchmark_filter="device_generate<sobol64*"
# To view all registered benchmarks:
./benchmark/benchmark_rocrand_host_api --benchmark_list_tests=true
# The benchmark also supports user input:
./benchmark/benchmark_rocrand_host_api --size <number> --trials <number> --offset <number> --dimensions <number> --lambda <float float float ...>
# And can print output in different formats:
./benchmark/benchmark_rocrand_host_api --benchmark_format=<console|json|csv>

# To run benchmark for device kernel functions:
# The benchmarks are registered with Google Benchmark as `device_kernel<engine,distribution>`, where
# engine -> xorwow, mrg31k3p, mrg32k3a, mtgp32, philox, lfsr113,
#           threefry2x32, threefry2x64, threefry4x32, threefry4x64,
#           sobol32, scrambled_sobol32, sobol64, scrambled_sobol64
# distribution -> uniform-uint or uniform-ullong, uniform-float, uniform-double, normal-float, normal-double,
#                 log-normal-float, log-normal-double, poisson, discrete-poisson, discrete-custom
# Further option can be found using --help
./benchmark/benchmark_rocrand_device_api
# To run specific benchmarks:
./benchmark/benchmark_rocrand_device_api --benchmark_filter=<regex>
# For example to run benchmarks with engine sobol64:
./benchmark/benchmark_rocrand_device_api --benchmark_filter="device_kernel<sobol64*"
# To view all registered benchmarks:
./benchmark/benchmark_rocrand_device_api --benchmark_list_tests=true
# The benchmark also supports user input:
./benchmark/benchmark_rocrand_device_api --size <number> --trials <number> --dimensions <number> --lambda <float float float ...>
# And can print output in different formats:
./benchmark/benchmark_rocrand_device_api --benchmark_format=<console|json|csv>

# To compare against cuRAND (cuRAND must be supported):
./benchmark/benchmark_curand_host_api [google benchmark options]
./benchmark/benchmark_curand_device_api [google benchmark options]

Legacy benchmarks

You can disable legacy benchmarks (those used prior to Google Benchmark) by setting the CMake option BUILD_LEGACY_BENCHMARK to OFF. For compatibility, the default setting is ON when BUILD_BENCHMARK is set.

Legacy benchmarks are deprecated and will be removed in a future version once all benchmarks have been migrated to the new framework.

Wrappers

• C++ wrappers for the rocRAND host API are in rocrand.hpp.
• Fortran wrappers.
• Python wrappers: rocRAND.

Support

Bugs and feature requests can be reported through the issue tracker.

Contributions and license

Contributions of any kind are most welcome! You can find more information at CONTRIBUTING.

Licensing information is located at LICENSE.