-
**Is your feature request related to a problem? Please describe.**
As described in #15366, we intend to adopt JIT compilation to some of our kernels using JITify/NVRTC.
JITify presently only supports …
-
-
-
### Required prerequisites
- [x] Consult the [security policy](https://github.com/NVIDIA/cuda-quantum/security/policy). If reporting a security vulnerability, do not report the bug using this form. U…
-
**Is your feature request related to a problem? Please describe.**
The [documentation](https://dppy.readthedocs.io/en/latest/finite_dpps/definition.html#definition) says that the correlation kernel c…
-
I realized that now we default to SSPKnoth in ClimaAtmos, one reason why fieldvector kernels take up so much time in https://github.com/CliMA/ClimaCore.jl/issues/2067#issue-2629142538, is that we are …
-
Currently, if we build an immersed grid with `active_cells_map = true`, specific kernels compute only on the map provided by the grid. This is very beneficial, but the implementation is a bit clunky, …
-
We would like to integrate the [cascade attention kernel](https://flashinfer.ai/2024/02/02/cascade-inference.html) from flashinfer.
Code pointers:
- Attention backend in sglang: https://github.com…
-
While building a package for Arch Linux, I found that enabling all CUDA architectures (`-DGINKGO_CUDA_ARCHITECTURES="All"`) leads to this error on the final link:
```
FAILED: lib/libginkgo_cuda.so…
-
Hi,
Flow and Eroder kernels set with a downstream direction (e.g. `apply_dir=fs.flow.FlowGraphTraversalDir.BREADTH_DOWNSTREAM`) raise `ValueError: Unsupported kernel application direction` , is th…