[Preprocessing] Fix bug in TD dag matching op #19936
Closed
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
This is based on the materialize-encoding-into-nop pass, with additional patterns to handle load and store ops. For now, we materialize load of a padded input with and extract slice and store as insert slice into a larger tensor. These get folded and become partial loads/stores at the end, but we can change this later. Enable this by default in the LLVMGPU pass pipeline as it's a superset of the existing nop encoding materialization pass. --------- Signed-off-by: Jakub Kuderski <jakub@nod-labs.com>
These types are not available in NumPy so no interoperability is provided for them. --------- Signed-off-by: Boian Petkantchin <boian.petkantchin@amd.com>
…ree-org#19926) It drops the revert of llvm/llvm-project@327d627 --------- Signed-off-by: hanhanW <hanhan0912@gmail.com>
…e-org#19920) Progress on iree-org#18174, updating some stale documentation. > [!NOTE] > Demo here: https://scotttodd.github.io/iree/guides/deployment-configurations/cpu/ Changes included: * Switch examples to use ONNX instead of TensorFlow given that users are trying to use TensorFlow and failing: iree-org#19852 * Add more documentation for CPU targets and features for iree-org#18561 * Standardize some formatting across CPU/CUDA/ROCm/Vulkan pages * Adjust some parts of the ONNX guide now that support is more mature
This drops the local LLVM revert: llvm/llvm-project@de7438e Signed-off-by: hanhanW <hanhan0912@gmail.com>
…ree-org#19726) The revision adds the support for the rest of AffinityOp that have TensorPhase trait, i.e., TensorCloneOp, TensorSliceOp, TensorFillOp, and TensorUpdateOp ops. It is tricky to handle encodings for transfer ops, so only the encoding in the fill op is updated. If other operations have tensor encodings, it returns a failure for now. There are two stream tensor ops do not implement the AffinityOpInterface, so they are not supported within the revision. They are stream.tensor.load op and stream.tensor.store op. We should be able to track the resource affinity for these two ops, and it requires additional analysis. Thus, they are not scoped within the revision. The revision also adds the missing documentation to the `addLayoutsToTensorPhaseOps` method. --------- Signed-off-by: hanhanW <hanhan0912@gmail.com>
Revert commits: - llvm/llvm-project@8c1dbac The author is working on a fix, and it is not ready yet. --------- Signed-off-by: hanhanW <hanhan0912@gmail.com>
Signed-off-by: hanhanW <hanhan0912@gmail.com>
MaheshRavishankar
approved these changes
Feb 7, 2025
This skips tiling large fills for the same reasoning as in iree-org#19887
Cherry pick llvm/llvm-project@73f11ac Per @ScottTodd : "fixes Windows builds + our nightly releases". https://discord.com/channels/689900678990135345/1080178290188374049/1337527415492182137 Signed-off-by: Benoit Jacob <jacob.benoit.1@gmail.com>
We had previously cherry-picked llvm/llvm-project@73f11ac in iree-org#19939. Now we're integrating up to that commit, so it's no longer a cherry-pick. Reverting llvm/llvm-project#125789 because it breaks TorchToTosa, in torch-mlir. We will need to wait for this to be resolved in torch-mlir, then simultaneously bump torch-mlir and drop the revert. Chery-pick a Bazel fix: llvm/llvm-project@4df287a --------- Signed-off-by: Benoit Jacob <jacob.benoit.1@gmail.com>
Signed-off-by: Max Dawkins <max.dawkins@gmail.com>
Signed-off-by: Max Dawkins <max.dawkins@gmail.com>
Max191
force-pushed
the
fix-preprocessing-td-matcher
branch
from
7f16c97 to
b32f587
Compare
Max191
requested review from
stellaraccident,
hanhanW,
IanWood1,
antiagainst,
kuhar,
Groverkss,
pashu123,
benvanik and
ScottTodd
as code owners
|
Closing this to land on main instead |
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.
Fixes a bug in the
transform.iree.match.cast_compatible_dag_from_rootop when there are repeated operands.