WorkTable.HomePage History
Hide minor edits - Show changes to output - Cancel
Changed line 7 from:
WorkTable is a programming model designed for performant portability, separation of concerns within large teams of programmers, and self documentation. Execution is event driven. Code is organized into activity-descriptions, and contexts flow among the descriptions, triggering computation when a context arrives at a description. The contexts contain both concrete data to be operated upon and abstract representations of data. The abstractions allow the runtime system to insert hardware-dependent choices, which specializes the code to the hardware during the run. Such abstractions require the ability to calculate the ID of other contexts that participate in communications.
to:
WorkTable is a programming model designed for performant portability, separation of concerns within large teams of programmers, and self documentation. Execution is event driven. Code is organized into activity-descriptions, while contexts flow among the descriptions, triggering computation when a context arrives at a description. The contexts contain both concrete data to be operated upon and abstract representations of data. The abstractions allow the runtime system to insert hardware-dependent choices, which specializes the code to the hardware during the run. Such abstractions require the ability to calculate the ID of other contexts that participate in communications.
Changed lines 4-5 from:
* [[Reference Manual]]
to:
* [[http://opensourceresearchinstitute.org/uploads/WorkTable/Sept_2011__WorkTable_tutorial.pdf | Tutorial]]
Deleted line 11:
Changed lines 1-10 from:
!Welcome to Vthread (VMS based posix threads)
* [[Vthreads.Papers | Papers]] that mention Vthread
* [[http://hg.opensourceresearchinstitute.org/cgi-bin/hgwebdir.cgi/VMS/VMS_Projects/VMS_Projects__MC_shared/Vthread/Vthread__exe_time_vs_task_size__MC_shared__Proj/|Project repo]] for a micro benchmark that measures the overhead of Vthread mutex constructs
* [[http://hg.opensourceresearchinstitute.org/cgi-bin/hgwebdir.cgi/VMS/VMS_Implementations/Vthread_impls/Vthread_MC_shared_impl/file/b94dc57e4455|Code of Vthread language]]
* [[http://hg.opensourceresearchinstitute.org/cgi-bin/hgwebdir.cgi/VMS/Applications/Vthread/Vthread__Exe_Time_Vs_Task_Size__Meas/file/fdc2f264f3d6|Code of the test application]] that generates an execution time vs task size curve
* [[Vthreads.Papers | Papers
* [[http://hg.opensourceresearchinstitute.org/cgi-bin/hgwebdir.cgi/VMS/VMS_Projects/VMS_Projects__MC_shared/Vthread/Vthread__exe_time_vs_task_size__MC_shared__Proj/|Project repo]] for a micro benchmark that measures the overhead of Vthread mutex constructs
* [[http://hg.opensourceresearchinstitute.org/cgi-bin/hgwebdir.cgi/VMS/VMS_Implementations/Vthread_impls/Vthread_MC_shared_impl/file/b94dc57e4455|Code of Vthread language]]
* [[http://hg.opensourceresearchinstitute.org/cgi-bin/hgwebdir.cgi/VMS/Applications/Vthread/Vthread__Exe_Time_Vs_Task_Size__Meas/file/fdc2f264f3d6|Code of the test application]] that generates an execution time vs task size curve
to:
!Welcome to the WorkTable HomePage
* [[ | Reference Manual]]
* [[ | Reference Manual]]
Changed lines 7-11 from:
Development time was 14 hours, including design, implementation, and testing
Vthread's runtime
to:
WorkTable is a programming model designed for performant portability, separation of concerns within large teams of programmers, and self documentation. Execution is event driven. Code is organized into activity-descriptions, and contexts flow among the descriptions, triggering computation when a context arrives at a description. The contexts contain both concrete data to be operated upon and abstract representations of data. The abstractions allow the runtime system to insert hardware-dependent choices, which specializes the code to the hardware during the run. Such abstractions require the ability to calculate the ID of other contexts that participate in communications.
* Performance portability is delivered by the abstractions, which allow the runtime system to adjust the granularity of work during the run.
* Separation of concerns is delivered by the fact that individual descriptions of activities are bits of sequential code. Parallelism aspects are only seen by the architect of the software system. The vast majority of the work lies within writing the activities. Hence, most programmers are insulated from parallelism concerns.
* Self documentation is provided by the organization of code into activity-descriptions, along with flows between descriptions. This lends itself to direct visualization, allowing the architect to visually define the system. The code that defines flow is separated, and acts as the direct representation of the visual system. Hence, creating the visual creates the code, and modifying the code modifies the visual.
* Performance portability is delivered by the abstractions, which allow the runtime system to adjust the granularity of work during the run.
* Separation of concerns is delivered by the fact that individual descriptions of activities are bits of sequential code. Parallelism aspects are only seen by the architect of the software system. The vast majority of the work lies within writing the activities. Hence, most programmers are insulated from parallelism concerns.
* Self documentation is provided by the organization of code into activity-descriptions, along with flows between descriptions. This lends itself to direct visualization, allowing the architect to visually define the system. The code that defines flow is separated, and acts as the direct representation of the visual system. Hence, creating the visual creates the code, and modifying the code modifies the visual.
Changed line 4 from:
* [[Vthread.Papers | Papers]] that mention Vthread
to:
* [[Vthreads.Papers | Papers]] that mention Vthread
Changed line 8 from:
* [[http://hg.opensourceresearchinstitute.org/cgi-bin/hgwebdir.cgi/VMS/Applications/Vthread/Vthread__Exe_Time_Vs_Task_Size__Meas/file/fdc2f264f3d6|Code of the test application that generates an execution time vs task size curve]]
to:
* [[http://hg.opensourceresearchinstitute.org/cgi-bin/hgwebdir.cgi/VMS/Applications/Vthread/Vthread__Exe_Time_Vs_Task_Size__Meas/file/fdc2f264f3d6|Code of the test application]] that generates an execution time vs task size curve
Changed lines 12-16 from:
Development time was
to:
Vthread is a VMS implementation of the posix threads mutex and condition variable constructs, embedded into C.
Development time was 14 hours, including design, implementation, and testing.
Vthread's runtime is an order of magnitude faster than the highly optimized version in the Linux Kernel, which relies upon localizing thread data and other performance enhancing tricks.
Development time was 14 hours, including design, implementation, and testing.
Vthread's runtime is an order of magnitude faster than the highly optimized version in the Linux Kernel, which relies upon localizing thread data and other performance enhancing tricks.
Added lines 1-16:
!Welcome to Vthread (VMS based posix threads)
* [[Vthread.Papers | Papers]] that mention Vthread
* [[http://hg.opensourceresearchinstitute.org/cgi-bin/hgwebdir.cgi/VMS/VMS_Projects/VMS_Projects__MC_shared/Vthread/Vthread__exe_time_vs_task_size__MC_shared__Proj/|Project repo]] for a micro benchmark that measures the overhead of Vthread mutex constructs
* [[http://hg.opensourceresearchinstitute.org/cgi-bin/hgwebdir.cgi/VMS/VMS_Implementations/Vthread_impls/Vthread_MC_shared_impl/file/b94dc57e4455|Code of Vthread language]]
* [[http://hg.opensourceresearchinstitute.org/cgi-bin/hgwebdir.cgi/VMS/Applications/Vthread/Vthread__Exe_Time_Vs_Task_Size__Meas/file/fdc2f264f3d6|Code of the test application that generates an execution time vs task size curve]]
!!Introduction and Overview
VCilk is a VMS implementation of the Cilk spawn and sync constructs, embedded into C.
Development time was 9 hours, including design, implementation, and testing.
VCilk's runtime compares favorably with the latest MIT release, Cilk version 5.4, despite the MIT version taking advantage of tricks enabled by the ultra simple nature of the language and the restrictions it places on dependency patterns that can be expressed in the language.
* [[Vthread.Papers | Papers]] that mention Vthread
* [[http://hg.opensourceresearchinstitute.org/cgi-bin/hgwebdir.cgi/VMS/VMS_Projects/VMS_Projects__MC_shared/Vthread/Vthread__exe_time_vs_task_size__MC_shared__Proj/|Project repo]] for a micro benchmark that measures the overhead of Vthread mutex constructs
* [[http://hg.opensourceresearchinstitute.org/cgi-bin/hgwebdir.cgi/VMS/VMS_Implementations/Vthread_impls/Vthread_MC_shared_impl/file/b94dc57e4455|Code of Vthread language]]
* [[http://hg.opensourceresearchinstitute.org/cgi-bin/hgwebdir.cgi/VMS/Applications/Vthread/Vthread__Exe_Time_Vs_Task_Size__Meas/file/fdc2f264f3d6|Code of the test application that generates an execution time vs task size curve]]
!!Introduction and Overview
VCilk is a VMS implementation of the Cilk spawn and sync constructs, embedded into C.
Development time was 9 hours, including design, implementation, and testing.
VCilk's runtime compares favorably with the latest MIT release, Cilk version 5.4, despite the MIT version taking advantage of tricks enabled by the ultra simple nature of the language and the restrictions it places on dependency patterns that can be expressed in the language.