[SYCL][SCLA] Add CodeGen capabilities for sycl_ext_oneapi_private_alloca
#12894
Conversation
This file contains 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
victor-eds
had a problem deploying
to
WindowsCILock
GitHub Actions
Failure
premanandrao
reviewed
Mar 6, 2024
|
@premanandrao I addressed your comments. I also removed some redundant tests:
|
victor-eds
had a problem deploying
to
WindowsCILock
GitHub Actions
Failure
…loca` The [`sycl_ext_oneapi_private_alloca`](https://github.com/intel/llvm/blob/56e9067ba69809fb6ea1fd4328456ca3a009f984/sycl/doc/extensions/experimental/sycl_ext_oneapi_private_alloca.asciidoc) adds new functions returning a pointer to a specialization constant length SYCL private memory allocation. This commit adds codegen support for these functions. The `sycl::private_alloca` function is implemented as an alias to a new `__builtin_intel_sycl_alloca` builtin. This is needed to guarantee the call will lower to just an `alloca` instruction defining the private memory allocation. This builtin lowers to a SYCL builtin call to `__builtin_sycl_unique_stable_id` and a call to a new `llvm.sycl.alloca` intrinsic. This intrinsic receives three arguments encoding the specialization constant used as array length, a type hint argument encoding the allocation element type and the required alignment. Note the `sycl_ext_oneapi_private_alloca` extension defines two functions: `private_alloca` and `aligned_private_alloca`. This commit adds codegen support only for the first signature, but already prepares support for the aligned flavor by adding an argument encoding the memory allocation alignment to the `llvm.sycl.alloca` intrinsic. The intrinsic is needed as generating an `alloca` instruction right away would lead to optimization passes converting the size argument, which can be of any integral type, and thus difficulting lowering to a single SPIR-V `OpVariable` later in the pipeline. Signed-off-by: Victor Perez <victor.perez@codeplay.com>
Signed-off-by: Victor Perez <victor.perez@codeplay.com>
victor-eds
force-pushed
the
scla-clang-work
branch
from
d418d45
to
ce4d3cd
Compare
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.
The
sycl_ext_oneapi_private_allocaadds new functions returning a pointer to a specialization constant length SYCL private memory allocation. This commit adds codegen support for these functions.The
sycl::private_allocafunction is implemented as an alias to a new__builtin_intel_sycl_allocabuiltin. This is needed to guarantee the call will lower to just anallocainstruction defining the private memory allocation.This builtin lowers to a SYCL builtin call to
__builtin_sycl_unique_stable_idand a call to a newllvm.sycl.allocaintrinsic. This intrinsic receives three arguments encoding the specialization constant used as array length, a type hint argument encoding the allocation element type and the required alignment.Note the
sycl_ext_oneapi_private_allocaextension defines two functions:private_allocaandaligned_private_alloca. This commit adds codegen support only for the first signature, but already prepares support for the aligned flavor by adding an argument encoding the memory allocation alignment to thellvm.sycl.allocaintrinsic.The intrinsic is needed as generating an
allocainstruction right away would lead to optimization passes converting the size argument, which can be of any integral type, and thus difficulting lowering to a single SPIR-VOpVariablelater in the pipeline.