makortel / portability

Test programs for different "performance portability" frameworks and libraries

A comparison of different heterogeneous frameworks

The purpose of these test programs is to experiment with various heterogeneous frameworks and "performance portability" approaches.

Current implementations

Test	CPU (serial)	CUDA	Alpaka/Cupla	SYCL
vector_add	✔️	✔️	✔️ ✔️ ✔️ ✔️	✔️ ✔️ ✔️
shared_mem	✔️	✔️	✔️ ✔️ ✔️ ✔️	✔️ ✔️ ✔️

Support for Alpaka/Cupla is split in

• CPU serial
• CPU paraller, using OpenMP
• CPU paraller, using TBB
• GPU parallel, using CUDA

Support for SYCL is split in

• CPU (unspecified)
• genericl OpenCL
• Nvidia PTX

CUDA

These programs require a recent CUDA version (nvcc supporting C++14 and --expt-relaxed-constexpr) and a machine with GPU.

Alpaka/Cupla

Alpaka/Cupla support different backends; so far it has been tested with serial, TBB and OpenMP backends for the CPU, and the CUDA backend for Nvidia GPUs. The latter requires CUDA 9.2 through 10.1, and has been tested with gcc 7.x and gcc 8.x.

Rather than using the advertised CMake-based approach, one can use Cupla as header-only library:

# choose a base directory for Alpaka and Cupla
BASE=/opt/alpaka
mkdir -p $BASE
cd $BASE

git clone -b develop [email protected]:ComputationalRadiationPhysics/alpaka.git
git clone -b dev     [email protected]:cms-patatrack/cupla.git

The cms-patatrack for of Cupla includes

• the latest development version of Cupla;
• support for the latest development version of Alpaka ;
• merge of cupla#112 and following bug fixes;
• removal of the embedded version of Alpaka.

Note: these instructions were accurate as of July 1st, 2019.

See the vector_add/Makefile for an example. If you use a different BASE folder, updated the Makefiles accordingly.

SYCL

These programs were tested with CodePlay's ComputeCpp Community Edition, version 1.1.3 .

Other projects

Other projects with similar goals:

• BabelStream memory transfer benchmark: https://uob-hpc.github.io/BabelStream/
• Parallel Research Tools: https://github.com/ParRes/Kernels
• DoE portability study: https://performanceportability.org/