gpurbenchmark's Introduction

gpuRbenchmark

The gpuR benchmarking package. Herein is contained wrappers for benchmarking various operations that are executed on the GPU relative to base R.

Note, for Windows users, you may to execute the following prior to install in order for the build to complete successfully. The package needs to link against the float package in order to have support for "float" values. Unfortunately, the linking that can be done on Linux isn't supported on Windows and therefore requires your PATH to be updated.

float_libs_dir_rel = system.file(path, package="float")
float_libs_dir = tools::file_path_as_absolute(float_libs_dir_rel)
dll.path <- normalizePath(float_libs_dir)
dll.path <- utils::shortPathName(dll.path)
dll.path <- gsub("\\\\", "/", dll.path)

Sys.setenv(PATH = paste(dll.path, Sys.getenv("PATH"), sep = ";"))

Example GEMM benchmark

library(gpuRbenchmark)
benchmark_gemm(type="float")

gpurbenchmark's People

Contributors

Stargazers

Watchers

gpurbenchmark's Issues

error: object 'ldflags' not found

Seeing this error when installing on Linux Mint 18.*

** building package indices
** testing if installed package can be loaded
Number of platforms: 1
- platform: NVIDIA Corporation: OpenCL 1.2 CUDA 9.2.147
  - context device index: 0
    - Quadro M1200
checked all devices
completed initialization
called onLoadError: package or namespace load failed for ‘gpuRbenchmark’:
 .onLoad failed in loadNamespace() for 'gpuRbenchmark', details:
  call: get(name, envir = asNamespace(pkg), inherits = FALSE)
  error: object 'ldflags' not found
Error: loading failed
Execution halted

Windows build failing

Windows build is failing as reported in issue here. Confirmed build failure on Windows 7 as well.

Package installation

I see

Error in loadNamespace(i, c(lib.loc, .libPaths()), versionCheck = vI[[i]]) :
there is no package called ‘dplyr’
ERROR: lazy loading failed for package ‘gpuRbenchmark’

dplyr is not in DESCRIPTION (as well as ggplot2). Can we make it optional or remove?

Recommend Projects