Ripple
Ripple is a framework designed to make parallelization of large-scale heterogeneous applications simple, with a focus on multiple gpus systems, but in the longer this will change to make heterogeneous computing easy across any system with accelerators.
It is designed to reduce the difficulty of programming on systems with many architectures, such as is the case for modern supercomputers, but also locally to easily access all the compute available on your system. The main idea for ripple is to write once, and run on any hardware, efficiently and in parallel, with minimal effort.
Currently it will use all available computational resources on a node (any number of CPUs and/or GPUs) and is currently being extended to support multi-node systems.
The target compute unit is the GPU, since it offers far superior performance compared to even large numbers of CPU cores, however, the user has a choice of which to use, and can use both, concurrently, if desired.
Ripple uses a number of abstractions to facilitate the above, which allow for simple, expressive code to run in parallel across many large systems (it has been used to run physics simulations on grids consisting of billions of cells).
The means for specifying computational flow is the graph interface, which allows computational operations and dependencies to be specified expressively and concisely, and from which ripple can determine efficient parallelization. The graph interface scales well, with up to 7.3x for 8 V100 GPUs for non-trivial real-world problems
Up to date documentation can be found here: ripple docs
Building Ripple
First, get ripple from github: ripple
Currently, Ripple requires CUDA, since some of the features use it, however, we are in the process of removing the CUDA dependency for the CPU only use case . Ripple has the following dependencies:
cmake >= 3.19
clang >= 9.0 with CUDA >= 9.0 or
gcc >= 6.0 with CUDA >= 11.0
NOTE: Ripple is written using C++ >= 17, which is why CUDA >= 11.0 is required if used as the device compiler, where as clang >= 9.0 has C++-17 support and can be used as both the host and device compiler.
Ripple is built using cmake, specifying various options. To see all available options for building ripple, from the project root, run
cmake -DPRINT_OPTIONS=ON .
which shows the required and optional options.
Cmake support for CUDA does not always work as expected, so to build ripple, the paths to the variable compilers, as well as cuda installation, need to be specified:
mkdir build && cd build
cmake \
-DCUDA_PATH=<path to cuda toolkit root> \
-DCMAKE_CUDA_COMPILER=<path to cuda compiler> \
-DCMAKE_CXX_COMPILER=<path to cxx compiler> \
-DCMAKE_BUILD_TYPE=Release \
-DCUDA_ARCHS=80;86 \
.
NOTE Of the above parameters, the first three are required, while the rest are optional. If the cuda compiler is clang, then the CXX compiler is automatically set to clang as well. If the cuda compiler is set to nvcc, the cuda host compiler will be set to the
CMAKE_CXX_COMPILER.
NOTE If
-DCMAKE_BUILD_TYPE=Debug, the cmake language feature sometimes fails to correctly verify that the cuda compiler is correct and working. The current fix for this is to first specify the build parameters as above using-DCMAKE_BUILD_TYPE=Release, and then simply execute-DCMAKE_BUILD_TYPE=Debug ..after.
NOTE Ripple will print out the complete build configuration at the end of the cmake command, so you can verify that the chosen parameters are correct.
This process is trying to be made simpler, but with current Cmake, this is the simplest process to ensure that the build is correct.
Getting Started
The shortest code to help with getting started is the SAXPY example, which is as simple as the following:
using Tensor = ripple::Tensor<float, 1>;
constexpr size_t size_x = 1000000;
const size_t partitions = topology().num_gpus();
// Create the tensors. Partitions is always a vector, with each component
// specifying the number of partitions in the {x, y, z} dimension. Each
// partition will be reside and therefore execute on a separate gpu.
// It's as simple as that for multi-GPU programming
Tensor a{{partitions}, size_x};
Tensor b{{partitions}, size_x};
Tensor c{{partitions}, size_x};
float x = 2.0f;
// Create the graph, which splits the execution across the partitions of the
// tensor. The arguments to the functors are iterators which point to the
// cell in the tensor at the global thread indices.
ripple::Graph graph;
graph.split([] ripple_all (auto a, auto b, auto c, float x) {
// Set a and b to thread indices:
*a = a.global_idx(ripple::dimx());
*b = b.global_idx(ripple::dimx());
// Set the result:
*c = *a * x + *b;
}, a, b, c, x);
ripple::execute(graph);
ripple::fence();
for (size_t i = 0; i < c.size(ripple::dimx()); ++i) {
fmt::format("{} {}\n", i, *c(i));
} Ripple has a lot more functionality, and the next best way to explore some of
it is to have a look through the benchmarks, which can be found in
benchmarks/. If you want to build the benchmarks, then add
-DRIPPLE_BUILD_BENCHMARKS=ON to the cmake configuration.
Current Support
Currently, ripple will work with Intel or AMD CPUs, and NVIDIA GPUs. Over time it is intended that ripple will support all GPUs, as well as any new types of accelerators, but we are not there yet.
If you are interested in using ripple, and have any specfic examlpes which would help with getting started, please raise an issue and it will be added.
Publications
Robert Clucas, Philip Blakely, Nikolaos Nikiforakis (2021). RRipple : Simplified Large-Scale Computation on Heterogeneous Architectures with Polymorphic Data Layout arXiv:2104.08571
Submitted to the Journal of Parallel and Distributed programming, to be updated shortly