[EPIC]: Add Hopper features to cuda::ptx
#1340
Assignees
Labels
feature request
New feature or request.
Comments
1 task
2 tasks
|
@ahendriksen can this be updated/closed? |
|
Is it okay if I keep it open? There are a few instructions that were low-priority before, but are seeing more use cases now. I am thinking of the multimem, async prefetch, and prefetch_tensormap instructions. |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Is this a duplicate?
Area
libcu++
Is your feature request related to a problem? Please describe.
Many Hopper features are not yet exposed through
cuda::ptx.Describe the solution you'd like
Add PTX instructions to
cuda::ptx.Describe alternatives you've considered
No response
Additional context
No response
Tasks
cuda::ptx::fence#1339cuda::ptx::mapa#1343cuda::ptx::getctarank#1344cuda::ptx::get_sregto support Hopper-specific special registers #1350cuda::ptx::cp_async_bulk_tensor#1398cuda::ptx::cp_reduce_async_bulk_tensor#1399cuda::ptx::cp_async_bulk_commit_group#1400cuda::ptx::cp_async_bulk_wait_group#1401cuda::ptx::cp_async_bulk#1402cuda::ptx::barrier_cluster#1365cuda::ptx::tensormap_replace#1439cuda::ptx::tensormap_cp_fenceproxy#1440cuda::ptx::elect_sync#1536cuda::ptx::cp_reduce_async_bulk#1444cuda::ptx::setmaxnregcuda::ptx::cp_async_bulk_prefetchcuda::ptx::cp_async_bulk_prefetch_tensorcuda::ptx::multimem_*cuda::ptx::prefetch_tensormapcuda::ptx::griddepcontrolcuda::ptx::mbarrier_complete_txcuda::ptx::mbarrier_expect_txcuda::ptx::stmatrixcuda::ptx::atom(cluster scope)cuda::ptx::red(cluster scope)cuda::ptx::mbarrier_init#1490The text was updated successfully, but these errors were encountered: