Implemented override for Cuda device instruction set. #931
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
This PR attempts to deal with the issue of newly introduced Cuda Instruction Sets that are unknown at the time ILGPU was compiled and published to Nuget (see #880).
Using the RTX 4090 as an example, Nvidia introduced SM 9.0 and ISA 7.8. The architecture SM 9.0 can be determined by querying the device at runtime, so that's fine. However, the instruction set ISA 7.8 is unknown to ILGPU, and cannot be determined at runtime.
This PR introduces a new overload to the
Cuda()
method when configuring aContext
that allows the caller to force the device to use a specific instruction set. This should hopefully make ILGPU versions usable as new Cuda devices are released, without forcing developers to update to the latest version of ILGPU.