Difference between revisions of "GPGPU-WG KnowledgeBase GLUE2 Profile Application"
Line 1: | Line 1: | ||
= Introduction = | |||
The Working Group shall define an EGI Community GLUE2 based standard that encapsulates and publishes data about the application software its environment required by Computational Accelerators. | The Working Group shall define an EGI Community GLUE2 based standard that encapsulates and publishes data about the application software its environment required by Computational Accelerators. | ||
Line 4: | Line 5: | ||
* The [[GPGPU-WG:GPGPU_Working_Group_KnowlegeBase:GLUE2:Profile:Application:ApplicationEnvironment |ApplicationEnvironment]], which is a significant extension to the GLUE1 GlueHostApplicationSoftwareRunTimeEnvironment. | * The [[GPGPU-WG:GPGPU_Working_Group_KnowlegeBase:GLUE2:Profile:Application:ApplicationEnvironment |ApplicationEnvironment]], which is a significant extension to the GLUE1 GlueHostApplicationSoftwareRunTimeEnvironment. | ||
* The [[GPGPU-WG:GPGPU_Working_Group_KnowlegeBase:GLUE2:Profile:Application:ApplicationHandle| ApplicationHandle]], a new class that can indicate how an to describe an ApplicationEnvironment instance can setup the software environment. | * The [[GPGPU-WG:GPGPU_Working_Group_KnowlegeBase:GLUE2:Profile:Application:ApplicationHandle| ApplicationHandle]], a new class that can indicate how an to describe an ApplicationEnvironment instance can setup the software environment. | ||
In addition, each application can have a reference to one or more application benchmarks. | |||
The GIP plugin ''glite-ce-glue2-applicationenvironment-dynamic'' produces ApplicationEnvironment data by converting the basic GLUE1 GlueHostApplicationSoftwareRunTimeEnvironment to a simple ApplicationEnvironment instance. | The GIP plugin ''glite-ce-glue2-applicationenvironment-dynamic'' produces ApplicationEnvironment data by converting the basic GLUE1 GlueHostApplicationSoftwareRunTimeEnvironment to a simple ApplicationEnvironment instance. | ||
Line 13: | Line 16: | ||
== Task 1 == | |||
The Working Group shall propose a convention to advertise application software that support GPGPUs and CAs. In particular we should consider the following ApplicationEnvironments: | |||
* CUDA | * CUDA | ||
* OpenCL | * OpenCL | ||
* Support Applications/Compilers for Intel Xeon Phi | * Support Applications/Compilers for Intel Xeon Phi | ||
* Other | * Application Benchmarks? | ||
* Other User Application Software? | |||
== Task 2 == | |||
The working group shall discuss the merits of using ApplicationHandle instances to help bootstrap each | |||
[[GPGPU-WG:GPGPU_Working_Group_KnowlegeBase:GLUE2:Profile:Application:ApplicationEnvironment |ApplicationEnvironment]] | [[GPGPU-WG:GPGPU_Working_Group_KnowlegeBase:GLUE2:Profile:Application:ApplicationEnvironment |ApplicationEnvironment]] | ||
---- | ---- |
Revision as of 19:52, 27 February 2014
Introduction
The Working Group shall define an EGI Community GLUE2 based standard that encapsulates and publishes data about the application software its environment required by Computational Accelerators.
There are two GLUE2 classes to consider here:
- The ApplicationEnvironment, which is a significant extension to the GLUE1 GlueHostApplicationSoftwareRunTimeEnvironment.
- The ApplicationHandle, a new class that can indicate how an to describe an ApplicationEnvironment instance can setup the software environment.
In addition, each application can have a reference to one or more application benchmarks.
The GIP plugin glite-ce-glue2-applicationenvironment-dynamic produces ApplicationEnvironment data by converting the basic GLUE1 GlueHostApplicationSoftwareRunTimeEnvironment to a simple ApplicationEnvironment instance.
The ApplicationEnvironment is a GLUE2 class that describes the software environment in which a job will run.
- Each Application is identified by a name.
- These names are not defined by the schema, but should be assigned in a way that allows applications to be uniquely identified.
Task 1
The Working Group shall propose a convention to advertise application software that support GPGPUs and CAs. In particular we should consider the following ApplicationEnvironments:
- CUDA
- OpenCL
- Support Applications/Compilers for Intel Xeon Phi
- Application Benchmarks?
- Other User Application Software?
Task 2
The working group shall discuss the merits of using ApplicationHandle instances to help bootstrap each ApplicationEnvironment