You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
It hink this PR broke it, because it removed the lowering config #19504@bjacob I know there is this new approach to ukernels on GPU. Is the plan to fully remove the old one?
Also @benvanik, because you seem to have build the example.
Steps to reproduce your issue
Compile IREE with CUDA and samples on.
What component(s) does this issue relate to?
Compiler
Version information
No response
Additional context
No response
The text was updated successfully, but these errors were encountered:
Oh, we don't care much about that sample anymore. It was mostly just a stepping stone towards ukernels. We could just remove it, or we could not build it on CUDA.
What happened?
Hi! I have noticed that the CUDA example for custom dispatch is broken.
This is the stack trace:
The failing line is not in there, so jumping in with the debugger, this line failes:
iree/compiler/src/iree/compiler/Codegen/LLVMGPU/KernelConfig.cpp
Line 1662 in 6245db1
The op it tries to cast is the ukernel op
Seems like some special casing for that op is missing in
iree/compiler/src/iree/compiler/Codegen/LLVMGPU/KernelConfig.cpp
Line 2341 in 6245db1
It hink this PR broke it, because it removed the lowering config #19504 @bjacob I know there is this new approach to ukernels on GPU. Is the plan to fully remove the old one?
Also @benvanik, because you seem to have build the example.
Steps to reproduce your issue
Compile IREE with CUDA and samples on.
What component(s) does this issue relate to?
Compiler
Version information
No response
Additional context
No response
The text was updated successfully, but these errors were encountered: