NAS Parallel Benchmarks for GPUs

This is a repository aimed at providing GPU parallel codes with different parallel APIs for the NAS Parallel Benchmarks (NPB) from a C/C++ version (NPB-CPP). The parallel versions support GPUs from NVIDIA, AMD, and Intel. You can also contribute with this project, writing issues and pull requests. 😄

🔉News: CUDA versions for pseudo-applications added and IS improved. 📅11/Feb/2021

🔉News: Parametrization support for configuring number of threads per block and CUDA parallelism optimizations. 📅25/Jul/2021

🔉News: Paper published in the journal Software: Practice and Experience (SPE). 📅29/Nov/2021

🔉News: A new GPU parallel implementation is now available using the GSParLib API. 📅15/Aug/2024

🔉News: A new GPU parallel implementation is now available using HIP (supporting GPUs from NVIDIA, AMD, and Intel). 📅30/Jan/2025

How to cite our work 👍

DOI - Araujo, G.; Griebler, D.; Rockenbach, D. A.; Danelutto, M.; Fernandes, L. G.; NAS Parallel Benchmarks with CUDA and beyond, Software: Practice and Experience (SPE), 2021.

DOI - Araujo, G.; Griebler, D.; Danelutto, M.; Fernandes, L. G.; Efficient NAS Benchmark Kernels with CUDA. 28th Euromicro International Conference on Parallel, Distributed and Network-based Processing (PDP), Västerås, 2020.

The NPB with GPUs

The parallel GPU versions were implemented from the serial version of NPB-CPP.

==================================================================

NAS Parallel Benchmarks for GPUs code contributors are:

Dalvan Griebler: [email protected]

Gabriell Araujo: [email protected]

==================================================================

Each directory is independent and contains its own implemented version:

Five kernels

• IS - Integer Sort
• EP - Embarrassingly Parallel
• CG - Conjugate Gradient
• MG - Multi-Grid
• FT - discrete 3D fast Fourier Transform

Three pseudo-application

• SP - Scalar Penta-diagonal solver
• BT - Block Tri-diagonal solver
• LU - Lower-Upper Gauss-Seidel solver

Software Requiriments

Warning: our tests were made with GCC, NVCC, HIPCC, and ChipStar

How to Compile

1. Go inside the directory of the target API (e.g., CUDA, HIP, or GSPar)

2. Update the file config/make.def informing the path to the GPU compiler

3. If you choose HIP, also update the file config/make.def exporting the vendor of the GPU you want to use (e.g., nvidia, amd, or intel):

   export HIP_PLATFORM=amd
   

4. Execute the following commands to compile and execute the NPB programs:

   make _BENCHMARK CLASS=_VERSION
   

   _BENCHMARKs are:

    CG, EP, FT, IS, MG, BT, LU, and SP 
   

   _VERSIONs are:

    + Class S: small for quick test purposes
   
    + Class W: workstation size (a 90's workstation; now likely too small)
   
    + Classes A, B, C: standard test problems; ~4X size increase going from one class to the next
   
    + Classes D, E, F: large test problems; ~16X size increase from each of the previous Classes
   

   Command example to compile:

   make ep CLASS=B
   

   Command example to run:

   bin/ep.B
   

Activating the additional timers

NPB-GPU has additional timers for profiling purpose. To activate these timers, create a dummy file 'timer.flag' in the main directory of the NPB version (e.g. CUDA/timer.flag).

Configuring the number of threads per block

NPB-GPU allows configuring the number of threads per block of each GPU kernel in the benchmarks. The user can specify the number of threads per block by editing the file gpu.config in the directory /config/. If no file is specified, all GPU kernels are executed using the warp size of the GPU as the number of threads per block.

Syntax of the gpu.config file:

<benchmark-name>_THREADS_PER_BLOCK_<gpu-kernel-name> = <interger-value>

Configuring CG benchmark as example:

CG_THREADS_PER_BLOCK_ON_KERNEL_ONE = 32
CG_THREADS_PER_BLOCK_ON_KERNEL_TWO = 128
CG_THREADS_PER_BLOCK_ON_KERNEL_THREE = 64
CG_THREADS_PER_BLOCK_ON_KERNEL_FOUR = 256
CG_THREADS_PER_BLOCK_ON_KERNEL_FIVE = 32
CG_THREADS_PER_BLOCK_ON_KERNEL_SIX = 64
CG_THREADS_PER_BLOCK_ON_KERNEL_SEVEN = 128
CG_THREADS_PER_BLOCK_ON_KERNEL_EIGHT = 64
CG_THREADS_PER_BLOCK_ON_KERNEL_NINE = 512
CG_THREADS_PER_BLOCK_ON_KERNEL_TEN = 512
CG_THREADS_PER_BLOCK_ON_KERNEL_ELEVEN = 1024

The NPB-GPU also allows changing the GPU device by providing the following syntax in the gpu.config file:

GPU_DEVICE = <interger-value>