simplify graph update

[aharon-abramson]

The suggested graph update mechanism seems computationally complex: an implementation has to traverse both the modifiable and executable graphs to identify any changes to either inputs or outputs.
I better like the OpenCL approach, where one directly updates a cl_mutable_command_khr handle. We can add an update method to class node. We can also add graph::get_node(size_t index) for obtaining the nodes created during queue recording.

[EwanC]

Hi Aharon, thanks for your input. The current draft update mechanism is aimed at library use cases, where an application creates an graphs from recording a library call, where they don't have access to SYCL objects like buffers/USM to update.

A more fine grained update mechanism is something we're definitely looking at adding for future versions of the spec, so will keep a design like this in mind.

[aharon-abramson]

Great, thanks.

[EwanC]

@aharon-abramson Now that this feature is being developed on intel/llvm we're closing the issues on this repo and moving any ones still relevant over to that repo. So I'm closing this issue with a status update, but please open an issue on intel/llvm if you'd like to give further feedback.

We now have individual node update APIs https://github.com/intel/llvm/blob/sycl/sycl/doc/extensions/experimental/sycl_ext_oneapi_graph.asciidoc#individual-node-update & https://github.com/intel/llvm/blob/sycl/sycl/doc/extensions/experimental/sycl_ext_oneapi_graph.asciidoc#742-dynamic-parameters . There's an example here of how to use that https://github.com/intel/llvm/blob/sycl/sycl/doc/syclgraph/SYCLGraphUsageGuide.md#dynamic-parameter-update However, the dynamic_parameter object only works with kernel bundles right now rather than lambdas.

We're currently working on expanding this API further with better compiler integration so it can be used with lambda defined kernels as well (intel#16712 if you want to leave feedback on spec design), which we're in the middle of implementing