[WIP] WebGPU EP [skip ci]

[fs-eire]

Description

For code review purpose only. Offer a place for discussion.

see also:

• README.md
• How to Write WebGPU EP Kernel

prerequisite:

• Add dependency dawn into deps.txt #21910
• Adding ONNX Runtime C-API for WebGPU EP #21838

related changes:

• Allow setting config entries for test runner #21894
• enhance string util functions #21893
• Enabling c++20 on linux #17816
• Update C++ to standard 20 for Windows #17706
• refactor: extract shared util function ComputeBroadcastOutputShape #21940

TODO items

• Validation Switch (allows to change the behavior of whether perform specific validation checks)
• pushErrorScope/popErrorScope
• Graph Capture
• Profiling supported by WebGPU Query Buffer
• WebGPU resources tracking (mainly for buffers)
• Global hanlders( unhandled exceptions and device lost )

[skottmckay]

Using the generic SessionOptionsAppendExecutionProvider is preferred over adding EP specific functions

[fs-eire]

Do you have any suggestions if I want to use the generic SessionOptionsAppendExecutionProvider with passing pointer (void*)?

[skottmckay]

A void* is just a number for the address so could you pass the address of the OrtWGPUProviderOptions instance?

Not quite as pure but the advantage of that is you don't need to expose a new function via all the ORT language bindings (which would be C#, Objective-C and Java for mobile).

But that said, why do the various handles have to be created outside of the EP and passed in vs the EP being passed the info to create them. I would think that would be easier to use from other languages,otherwise you're going to have to document how to create OrtWGPUProviderOptions in all those languages.

[fs-eire]

A void* is just a number for the address so could you pass the address of the OrtWGPUProviderOptions instance?

It's the same question. "How to pass a pointer to SessionOptionsAppendExecutionProvider" still need to answered.

Not quite as pure but the advantage of that is you don't need to expose a new function via all the ORT language bindings (which would be C#, Objective-C and Java for mobile).

But that said, why do the various handles have to be created outside of the EP and passed in vs the EP being passed the info to create them. I would think that would be easier to use from other languages,otherwise you're going to have to document how to create OrtWGPUProviderOptions in all those languages.

From my understanding, there is not going to be use case of passing any pointers through language bindings. The only use case (MSEdge) with passing pointers is to static link ORT and use C-API, which requires at least to pass the pointer of dawn_proc_table. So, I am totally fine with using SessionOptionsAppendExecutionProvider everywhere else with string-to-string kvp except a minimal change in C-API to enable the scenario above.

BTW, there is a seperate PR for the API change only -> #21838. Further discussion can be done there.