Do not wrap single unset_encoding op into dispatch region. #20377
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
There are few pieces to plumb it through the software stack. What the revision does is: - Do not treat unset_encoding op as a root op, so it won't get fused into dispatches. - Add the unset_encoding ops to clonable list, so the CloneProducersIntoDispatchRegions pass can pull it into the consumer dispatch. - If there is no consumer dispatch, the unset_encoding op is converted to flow.tensor.encode op. - Similar to other tensor encode op, the op is folded away if an identity layout is recognized. Otherwise, the MaterializeEncodingsPass materializes uniqued executables for `stream.tensor.encode` ops and replaces them with dispatches to those executables. The [IR dump](https://gist.github.com/hanhanW/581d4ae9cd8ce26e47c723a27192f7d9) shows that there are no additional dispatches when the backend does not support encoding. To repro, run ```bash iree-compile \ --output-format=vm-bytecode \ --iree-hal-target-backends=vulkan-spirv \ --iree-global-opt-enable-early-materialization=false \ --mlir-print-ir-after=iree-stream-materialize-encodings \ --mlir-print-ir-before=iree-stream-materialize-encodings \ --mlir-disable-threading \ matmul_f32.mlir -o /tmp/matmul_f32.vmfb ``` ```mlir func.func @main(%lhs: tensor<?x?xf32>, %rhs: tensor<?x?xf32>) -> tensor<?x?xf32> { %c0 = arith.constant 0 : index %c1 = arith.constant 1 : index %M = tensor.dim %lhs, %c0 : tensor<?x?xf32> %N = tensor.dim %rhs, %c1 : tensor<?x?xf32> %cst = arith.constant 0.0 : f32 %init = tensor.empty(%M, %N) : tensor<?x?xf32> %fill = linalg.fill ins(%cst : f32) outs(%init : tensor<?x?xf32>) -> tensor<?x?xf32> %op = linalg.matmul ins(%lhs, %rhs : tensor<?x?xf32>, tensor<?x?xf32>) outs(%fill : tensor<?x?xf32>) -> tensor<?x?xf32> return %op : tensor<?x?xf32> } ``` Signed-off-by: hanhanW <hanhan0912@gmail.com>
Signed-off-by: hanhanW <hanhan0912@gmail.com>
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.
There are few pieces to plumb it through the software stack. What the revision does is:
stream.tensor.encodeops and replaces them with dispatches to those executables.The IR dump shows that there are no additional dispatches when the backend does not support encoding.
To repro, run
MLIR Source:
The revision deletes two old tests from
dispatch_linalg_on_tensors.mlirbecause the functionality is tested in each pass and the pipeline tests.Fixes #20187