parsec icon indicating copy to clipboard operation
parsec copied to clipboard

PaRSEC is a generic framework for architecture aware scheduling and management of micro-tasks on distributed, GPU accelerated, many-core heterogeneous architectures. PaRSEC assigns computation threads...

Results 115 parsec issues
Sort by recently updated
recently updated
newest added

Delegate GPU task completion to a co-manager using the MCA parameter _device_cuda_delegate_task_completion_. 1. The second CPU thread that submits the task to the GPU device is transitioned to a co-manager....

## Describe the bug Binding communication thread is unsuccessful. ``` ^[[1;37;43mW@00003^[[0m Request to bind the communication thread on core 23 failed. ^[[1;37;43mW@00000^[[0m Request to bind the communication thread on core...

bug

## Describe the bug If we do `parsec_advise_data_on_device(dta, did, PARSEC_DEV_DATA_ADVICE_PREFERRED_DEVICE )` on a `parsec_data_t *dta` that belongs to a data collection (as returned by `rank_of`, and that data is read...

bug

## Describe the bug This issue is found when running HiCMA on GPU and checking the correctness with `dplasma_potrf`, which is also on GPU, causing an error that is triggered...

bug

Using PaRSEC commit `7f81a1b` (2023-01-19) and DPLASMA commit `75012ef3f` (2023-01-23), the m, n, and k data in the trace of a GEMM is not present when using the dtd interface....

bug

## Describe the bug The first issue: `DPLASMA_WITH_RECURSIVE` and `PARSEC_HAVE_RECURSIVE` are independent; which means `PARSEC_HAVE_RECURSIVE` is always `OFF` by default even if `DPLASMA_WITH_RECURSIVE` is `ON`. The second issue: if `PARSEC_HAVE_RECURSIVE`...

bug

## Description Currently in iterative successor, the source rank (the rank where the task was executed) of a task is fixed (usually based on the data distribution). This creates a...

enhancement

## Describe the bug As part of registering a taskpool with the runtime for execution (`parsec_context_add_taskpool`), when profiling is enabled information about the taskpool is added to the global profiling...

bug

The goal of the BRAVO biased rwlock is to avoid readers taking the reader lock and thus contending for the atomic variable. Instead, readers raise a flag in an array...

enhancement

https://bitbucket.org/icldistcomp/parsec/pull-requests/332 Adds: Argobots support Adds: interface for threading library

enhancement
low priority