Skip to content

Latest commit

 

History

History
62 lines (48 loc) · 2.61 KB

README.md

File metadata and controls

62 lines (48 loc) · 2.61 KB

miniBUDE Julia

This is an implementation of miniBUDE in Julia which contains the following variants:

  • Threaded.jl - Threaded implementation with Threads.@threads macros
  • CUDA.jl - Direct port of miniBUDE native CUDA implementation using CUDA.jl
  • AMDGPU.jl - Direct port of miniBUDE's native HIP(via CUDA) implementation using AMDGPU.jl
  • oneAPi.jl - Direct port of miniBUDE's native SYCL implementation using oneAPi.jl
  • KernelAbstractions.jl - Direct port of miniBUDE's native CUDA implementation using KernelAbstractions.jl

Build & Run

Prerequisites

  • Julia >= 1.6+

A set of reduced dependency projects are available for the following backend and implementations:

  • AMDGPU supports:
    • AMDGPU.jl
  • CUDA supports:
    • CUDA.jl
  • oneAPI supports:
    • oneAPIStream.jl
  • KernelAbstractions supports:
    • KernelAbstractions.jl
  • Threaded supports:
    • ThreadedStream.jl

With Julia on path, run the benchmark with:

> cd JuliaStream.jl
> julia --project=<BACKEND> -e 'import Pkg; Pkg.instantiate()' # only required on first run
> julia --project=<BACKEND> src/<IMPL>.jl

For example. to run the CUDA implementation:

> cd JuliaStream.jl
> julia --project=CUDA -e 'import Pkg; Pkg.instantiate()' 
> julia --project=CUDA src/CUDA.jl

Important:

  • Julia is 1-indexed, so N >= 1 in --device N (alternatively, use device name substring).
  • Thread count for Threaded.jl must be set via the JULIA_NUM_THREADS environment variable (e.g export JULIA_NUM_THREADS=$(nproc)) otherwise it defaults to 1.
  • Certain implementations such as CUDA and AMDGPU will do hardware detection at runtime and may download and/or compile further software packages for the platform.
  • If Julia is launched behind some sort of launcher (e.g aprun), you may need to specify the -H option pointing to Julia's bin directory, so that Julia can find the correct libraries at runtime, for example: aprun -n 1 -d 64 julia -H "$(dirname "$(which julia)")" ...

Alternatively, the top-level project Project.toml contains all dependencies needed to run all implementations in src. There may be instances where some packages are locked to an older version because of transitive dependency requirements.

To run the benchmark using the top-level project, run the benchmark with:

> cd JuliaStream.jl
> julia --project -e 'import Pkg; Pkg.instantiate()'  
> julia --project src/<IMPL>.jl