[SYCL][NFC] Fix kernel release test when only host device is available #1835
Conversation
PIMock should only be constructed from a non-host platform, make the test check for a non-host platform prior to constructing PIMock. Signed-off-by: Sergey Semenov <sergey.semenov@intel.com>
|
@bader, this should take care of the post-commit failures https://github.com/intel/llvm/runs/747446124 . Although, I wonder why these tests are run there with no non-host devices in the first place. |
Co-authored-by: Alexey Bader <alexey.bader@intel.com>
This is valid use case for DPC++ and it's expected to work. |
Sure, I just wouldn't expect it to be the only use case that's verified in post-commit (correct me if I'm wrong on that). |
That would be ideal, it's a shame that PiMock doesn't offer that option right now. |
What else do you think we should add to the post-commit? |
|
Ah, never mind, I've found the source of my confusion. I think our current setup is fine as is. |
|
@rbegam ping |
|
@intel/llvm-reviewers-runtime, ping. |
|
@smaslov-intel, thank you! |
It specifies how to interpret 'Component Type' when components of a joint matrix are storages for values of different types, for example float for TF32, unsigned short for bfloat16. At this point only tf32 type interpretation is added during SPIR-V generation. Adding it to bf16 is a breaking change and requires adaptation across drivers. Spec update: intel#8175 Signed-off-by: Sidorov, Dmitry dmitry.sidorov@intel.com Original commit: KhronosGroup/SPIRV-LLVM-Translator@b7c5218
PIMock should only be constructed from a non-host platform, make the
test check for a non-host platform prior to constructing PIMock.
Signed-off-by: Sergey Semenov sergey.semenov@intel.com