Generate native code using L0 sdk instead of ocloc
#5342
Closed
Conversation
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Signed-off-by: Anatoly Myachev <[email protected]>
anmyachev
commented
Oct 17, 2025
| } | ||
| } | ||
|
|
||
| extern "C" EXPORT_FUNC PyObject *get_native_code(PyObject *args) { |
There was a problem hiding this comment.
Most of this function is taken from the implementation of load_binary function, as I've decided to leave it alone for now to avoid introducing regressions. However, there's potential to reduce code duplication there.
chengjunlu
reviewed
Oct 17, 2025
| zebin = f.read() | ||
| from triton.runtime.driver import driver | ||
| # at this stage the driver is already initialized | ||
| device = driver.active.utils.get_current_device() |
There was a problem hiding this comment.
The triton.compile maybe not target to the current device. The compilation flow should not depends on the runtime. It can be used as cross compiling and AOT compiling.
| self.get_native_code = mod.get_native_code | ||
| self.get_device_properties = mod.get_device_properties | ||
| self.device_count = mod.init_devices(self.get_sycl_queue()) | ||
| # breakpoint() |
There was a problem hiding this comment.
Suggested change
| # breakpoint() |
| try { | ||
| auto [l0_module_dgrf, l0_kernel_dgrf, n_spills_dgrf] = | ||
| compileLevelZeroObjects(binary_ptr, binary_size, kernel_name, | ||
| l0_device, l0_context, build_flags(), true); |
There was a problem hiding this comment.
Suggested change
| l0_device, l0_context, build_flags(), true); | |
| l0_device, l0_context, build_flags(), true /*is_spv*/); |
|
|
||
| if (!PyArg_ParseTuple(args, "sSisi", &name, &py_bytes, &shared, | ||
| &build_flags_ptr, &devId)) { | ||
| std::cerr << "loadBinary arg parse failed" << std::endl; |
There was a problem hiding this comment.
Suggested change
| std::cerr << "loadBinary arg parse failed" << std::endl; | |
| std::cerr << "get_native_code arg parse failed" << std::endl; |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
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.
b580: https://github.com/intel/intel-xpu-backend-for-triton/actions/runs/18599949719 (to check)
a770: https://github.com/intel/intel-xpu-backend-for-triton/actions/runs/18599941172 (to check)