Skip to content

Parallel benchmark suites with tags separating their kernels as well as their initialization, cleanup, and data transfer regions

Notifications You must be signed in to change notification settings

sed-inf-u-szeged/TaggedParallelBenchmarks

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

3 Commits
 
 
 
 
 
 
 
 
 
 

Repository files navigation

Tagged Parallel Benchmarks

This repository contains tagged versions of parts of the following 3 well-known benchmark suites: (For individual readmes, refer to the corresponding subdirectories.)

Tagging

Tagging in this context means STATIC_BEGIN / STATIC_END and DYNAMIC_BEGIN / DYNAMIC_END macro pairs at specific locations in the source code. Static macros only aim to capture the kernel functionalities while dynamic macros separate initialization/cleanup, data transfer (to and from), and kernel regions to enable individual measurements. The expansion of these macros is completely customizable, e.g., STATIC_BEGIN and STATIC_END could create a block tagged with a C++11 attribute which would help in identifying the block during static source code analysis while DYNAMIC_BEGIN and DYNAMIC_END could call a measurement library.

Platform Selection

Some OpenCL benchmarks have been modified to be able to take the target platform/device pair and also a pre-built kernel binary from the command line. The corresponding helper functionality is implemented in platform_helper.h.

License

Please refer to the licenses of the individual benchmark suites. There is no common license. When in doubt or you don't know what you're doing, please use the BSD or MIT license.

About

Parallel benchmark suites with tags separating their kernels as well as their initialization, cleanup, and data transfer regions

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published