torch-mlir-opt --help

torch-mlir-opt-help.txt

OVERVIEW: MLIR modular optimizer driver

Available Dialects: builtin, chlo, complex, func, linalg, memref, ml_program, quant, scf, sparse_tensor, stablehlo, tensor, tm_tensor, torch, torch_c, tosa, vhlo
USAGE: torch-mlir-opt [options] <input file>

OPTIONS:

Color Options:

  --color                                                  - Use colors in output (default=autodetect)

General options:

  --allow-unregistered-dialect                             - Allow operation with no registered dialects
  --disable-i2p-p2i-opt                                    - Disables inttoptr/ptrtoint roundtrip optimization
  --dump-pass-pipeline                                     - Print the pipeline that will be run
  --elide-resource-data-from-bytecode                      - Elide resources when generating bytecode
  --emit-bytecode                                          - Emit bytecode when generating output
  --emit-bytecode-version=<value>                          - Use specified bytecode when generating output
  --experimental-debuginfo-iterators                       - Enable communicating debuginfo positions through iterators, eliminating intrinsics. Has no effect if --preserve-input-debuginfo-format=true.
  --irdl-file=<filename>                                   - IRDL file to register before processing the input
  --load-dialect-plugin=<string>                           - Load dialects from plugin library
  --load-pass-plugin=<string>                              - Load passes from plugin library
  --log-actions-to=<string>                                - Log action execution to a file, or stderr if  '-' is passed
  --log-mlir-actions-filter=<string>                       - Comma separated list of locations to filter actions from logging
  --mlir-debug-counter=<string>                            - Comma separated list of debug counter skip and count arguments
  --mlir-disable-threading                                 - Disable multi-threading within MLIR, overrides any further call to MLIRContext::enableMultiThreading()
  --mlir-elide-elementsattrs-if-larger=<uint>              - Elide ElementsAttrs with "..." that have more elements than the given upper limit
  --mlir-elide-resource-strings-if-larger=<uint>           - Elide printing value of resources if string is too long in chars.
  --mlir-enable-debugger-hook                              - Enable Debugger hook for debugging MLIR Actions
  --mlir-generate-reproducer=<filename>                    - Generate an mlir reproducer at the provided filename (no crash required)
  --mlir-output-format=<value>                             - Output format for timing data
    =text                                                  -   display the results in text format
    =json                                                  -   display the results in JSON format
  --mlir-pass-pipeline-crash-reproducer=<string>           - Generate a .mlir reproducer file at the given output path if the pass manager crashes or fails
  --mlir-pass-pipeline-local-reproducer                    - When generating a crash reproducer, attempt to generated a reproducer with the smallest pipeline.
  --mlir-pass-statistics                                   - Display the statistics of each pass
  --mlir-pass-statistics-display=<value>                   - Display method for pass statistics
    =list                                                  -   display the results in a merged list sorted by pass name
    =pipeline                                              -   display the results with a nested pipeline view
  --mlir-pretty-debuginfo                                  - Print pretty debug info in MLIR output
  --mlir-print-debug-counter                               - Print out debug counter information after all counters have been accumulated
  --mlir-print-debuginfo                                   - Print debug info in MLIR output
  --mlir-print-elementsattrs-with-hex-if-larger=<long>     - Print DenseElementsAttrs with a hex string that have more elements than the given upper limit (use -1 to disable)
  --mlir-print-ir-after=<pass-arg>                         - Print IR after specified passes
  --mlir-print-ir-after-all                                - Print IR after each pass
  --mlir-print-ir-after-change                             - When printing the IR after a pass, only print if the IR changed
  --mlir-print-ir-after-failure                            - When printing the IR after a pass, only print if the pass failed
  --mlir-print-ir-before=<pass-arg>                        - Print IR before specified passes
  --mlir-print-ir-before-all                               - Print IR before each pass
  --mlir-print-ir-module-scope                             - When printing IR for print-ir-[before|after]{-all} always print the top-level operation
  --mlir-print-local-scope                                 - Print with local scope and inline information (eliding aliases for attributes, types, and locations
  --mlir-print-op-on-diagnostic                            - When a diagnostic is emitted on an operation, also print the operation as an attached note
  --mlir-print-skip-regions                                - Skip regions when printing ops.
  --mlir-print-stacktrace-on-diagnostic                    - When a diagnostic is emitted, also print the stack trace as an attached note
  --mlir-print-value-users                                 - Print users of operation results and block arguments as a comment
  --mlir-timing                                            - Display execution times
  --mlir-timing-display=<value>                            - Display method for timing data
    =list                                                  -   display the results in a list sorted by total time
    =tree                                                  -   display the results ina with a nested tree view
  --no-implicit-module                                     - Disable implicit addition of a top-level module op during parsing
  -o <filename>                                            - Output filename
  --output-split-marker=<string>                           - Split marker to use for merging the ouput
  --pass-pipeline=<string>                                 - Textual description of the pass pipeline to run
  --profile-actions-to=<string>                            - Profile action execution to a file, or stderr if  '-' is passed
  --run-reproducer                                         - Run the pipeline stored in the reproducer
  --show-dialects                                          - Print the list of registered dialects and exit
  --split-input-file[=<string>]                              - Split the input file into chunks using the given or default marker and process each chunk independently
  Compiler passes to run
    Passes:
      --canonicalize                                       -   Canonicalize operations
        --disable-patterns=<string>                        - Labels of patterns that should be filtered out during application
        --enable-patterns=<string>                         - Labels of patterns that should be used during application, all other patterns are filtered out
        --max-iterations=<long>                            - Max. iterations between applying patterns / simplifying regions
        --max-num-rewrites=<long>                          - Max. number of pattern rewrites within an iteration
        --region-simplify                                  - Perform control flow optimizations to the region tree
        --test-convergence                                 - Test only: Fail pass on non-convergence to detect cyclic pattern
        --top-down                                         - Seed the worklist in general top-down order
      --convert-torch-conversion-to-mlprogram              -   Convert recognized TorchConversion ops to MLProgram ops
      --convert-torch-onnx-to-torch                        -   Converts ONNX custom ops in the torch dialect to native torch ops
      --convert-torch-to-arith                             -   Convert recognized Torch ops to Std ops
      --convert-torch-to-linalg                            -   Convert recognized Torch ops to Linalg ops
      --convert-torch-to-scf                               -   Convert recognized Torch ops to SCF ops
      --convert-torch-to-stablehlo                         -   Convert Torch ops to Stablehlo ops
        --enable-i32-index                                 - Enable truncate index from i64 to i32(unsafely)
        --enable-static-shape                              - Enable static shape conversion
      --convert-torch-to-tensor                            -   Convert Torch ops to the Tensor dialect
      --convert-torch-to-tmtensor                          -   Convert recognized Torch ops to TMTensor/Linalg ops
      --convert-torch-to-tosa                              -   Convert Torch ops to TOSA ops
      --cse                                                -   Eliminate common sub-expressions
      --inline                                             -   Inline function calls
        --default-pipeline=<string>                        - The optimizer pipeline used for callables that do not have a dedicated optimizer pipeline in opPipelineList
        --inlining-threshold=<uint>                        - If the ratio between the number of the operations in the callee and the number of the operations in the caller exceeds this value (in percentage), then the callee is not inlined even if it is legal to inline it
        --max-iterations=<uint>                            - Maximum number of iterations when inlining within an SCC
        --op-pipelines=<pass-manager>                      - Callable operation specific optimizer pipelines (in the form of `dialect.op(pipeline)`)
      --loop-invariant-code-motion                         -   Hoist loop invariant instructions outside of the loop
      --print-op-stats                                     -   Print statistics of operations
        --json                                             - print the stats as JSON
      --refback-expand-ops-for-llvm                        -   Expand ops into more primitive ops before LLVM lowering.
      --refback-generalize-tensor-concat                   -   Convert tensor.concat to other tensor ops
      --refback-generalize-tensor-pad                      -   Convert tensor.pad to linalg ops
      --refback-mlprogram-bufferize                        -   Bufferize the MLProgram dialect ops
      --refback-munge-calling-conventions                  -   Munge calling conventions for calling via ExecutionEngine
      --refback-munge-memref-copy                          -   Munge memref.copy to linalg.copy
      --snapshot-op-locations                              -   Generate new locations from the current IR
        --filename=<string>                                - The filename to print the generated IR
        --tag=<string>                                     - A tag to use when fusing the new locations with the original. If unset, the locations are replaced.
      --stablehlo-aggressive-simplification                -   Canonicalizes StableHLO operations
      --stablehlo-legalize-to-linalg                       -   Legalize StableHLO to LinAlg
        --enable-primitive-ops                             - Lower to primitive Linalg ops (map, reduce and transpose) when possible, instead of linalg.generic
        --enable-sparse-ops                                - Lower to Sparse Tensor ops (sparse_tensor.concatenate)when possible, instead of linalg.generic
      --stablehlo-refine-shapes                            -   Refines shapes across a StableHLO program.
      --strip-debuginfo                                    -   Strip debug info from all operations
      --symbol-dce                                         -   Eliminate dead symbols
      --tm-tensor-bufferize                                -   Bufferize the TMTensor dialect
      --tm-tensor-to-loops                                 -   Convert TMTensor ops to loops and Linalg ops.
      --torch-adjust-calling-conventions                   -   Adjust the calling conventions of functions
      --torch-convert-custom-quant-op                      -   Convert torch custom quant op to linalg
      --torch-decompose-complex-ops                        -   Decompose complicated torch operations
        --legal-ops=<string>                               - List of operation names that should be considered legal
      --torch-drop-abstract-interp-calculations            -   Drop reified abstract interpretation calculations.
      --torch-erase-module-initializer                     -   Erase the `torch.global_slot.module_initializer` op.
      --torch-finalizing-backend-type-conversion           -   Finalizes a partial conversion to builtin tensors
      --torch-func-backend-type-conversion                 -   Convert functions to operate on builtin tensors
      --torch-fuse-quantized-ops                           -   QDQ: Fuse recognized QDQ op sequences.
      --torch-globalize-object-graph                       -   Converts TorchScript object graphs to a globalized form
      --torch-inline-global-slots                          -   Inlines torch.global_slot ops.
      --torch-lower-to-backend-contract                    -   Perform simplifications until the backend contract is satisfied.
        --backend-legal-ops=<string>                       - List of ops to be considered legal for the backend, such as 'aten.foo'.
        --decompose                                        - Decompose ops.
        --extra-library=<string>                           - MLIR module for splicing into the abstract interpretation library
        --max-iterations=<int>                             - Maximum number of invocations of the simplification pipeline.
      --torch-match-quantized-custom-ops                   -   Match quantized operations that occur in different namespace.
      --torch-maximize-value-semantics                     -   Use value-semantic tensors where possible.
      --torch-prepare-for-globalize-object-graph           -   Lowering in preparation for globalizing
      --torch-recompose-complex-ops                        -   Recompose torch operations that have been decomposed by TorchScript
      --torch-reduce-op-variants                           -   Reduces variants of ops to a smaller set of ops.
        --extra-library=<string>                           - MLIR module for verifying custom op value semantics
      --torch-refine-public-return                         -   Refine public return
      --torch-reify-dtype-calculations                     -   Reify dtype calculations.
        --extra-library=<string>                           - MLIR module for splicing into the dtype library
      --torch-reify-shape-calculations                     -   Reify shape calculations.
        --extra-library=<string>                           - MLIR module for splicing into the shape library
      --torch-scalarize-shapes                             -   Takes common shape computation operations and scalarizes them.
      --torch-simplify-dtype-calculations                  -   Simplify reified dtype calculations.
      --torch-simplify-shape-calculations                  -   Simplify reified shape calculations.
      --torch-unpack-quant-tensor                          -   Unpack quantized int4 tensor from int8 containter
      --torch-verify-backend-contract-no-decompositions    -   Check that program satisfies backend contract.
      --torch-verify-linalg-on-tensors-backend-contract    -   Verifies conformity to the linalg-on-tensors backend contract
      --torch-verify-stablehlo-backend-contract            -   Verifies conformity to the stablehlo backend contract
      --torch-verify-tosa-backend-contract                 -   Verifies conformity to the linalg-on-tensors backend contract
      --view-op-graph                                      -   Print Graphviz visualization of an operation
        --max-label-len=<uint>                             - Limit attribute/type length to number of chars
        --print-attrs                                      - Print attributes of operations
        --print-control-flow-edges                         - Print control flow edges
        --print-data-flow-edges                            - Print data flow edges
        --print-result-types                               - Print result types of operations
    Pass Pipelines:
      --torch-backend-to-linalg-on-tensors-backend-pipeline-   Pipeline lowering torch backend contract to linalg-on-tensors backend contract.
      --torch-backend-to-stablehlo-backend-pipeline        -   Pipeline lowering torch backend contract to StableHLO backend contract.
        --enable-i32-index                                 - Enable truncate index from i64 to i32(unsafely)
        --enable-static-shape                              - Enable static shape conversion.
      --torch-backend-to-tosa-backend-pipeline             -   Pipeline lowering torch backend contract to TOSA backend contract.
      --torch-function-to-torch-backend-pipeline           -   Pipeline lowering a Torch function to Torch backend form.
        --backend-legal-ops=<string>                       - List of ops to be considered legal for the backend, such as 'aten.foo'.
        --decompose-complex-ops                            - Decompose complex operations.
        --extra-library=<string>                           - Filename of MLIR module for splicing into the abstract interpretation library.
        --max-iterations=<int>                             - Maximum number of invocations of the simplification pipeline.
      --torch-shape-refinement-pipeline                    -   Pipeline refining shapes of tensors.
        --backend-legal-ops=<string>                       - List of ops to be considered legal for the backend, such as 'aten.foo'.
        --decompose-complex-ops                            - Decompose complex operations.
        --extra-library=<string>                           - Filename of MLIR module for splicing into the abstract interpretation library.
        --max-iterations=<int>                             - Maximum number of invocations of the simplification pipeline.
      --torch-simplification-pipeline                      -   Pipeline simplifying computations in the program.
        --backend-legal-ops=<string>                       - List of ops to be considered legal for the backend, such as 'aten.foo'.
        --decompose-complex-ops                            - Decompose complex operations.
        --extra-library=<string>                           - Filename of MLIR module for splicing into the abstract interpretation library.
        --max-iterations=<int>                             - Maximum number of invocations of the simplification pipeline.
      --torchscript-module-to-torch-backend-pipeline       -   Pipeline lowering TorchScript object graph IR to Torch backend form.
        --backend-legal-ops=<string>                       - List of ops to be considered legal for the backend, such as 'aten.foo'.
        --decompose-complex-ops                            - Decompose complex operations.
        --extra-library=<string>                           - Filename of MLIR module for splicing into the abstract interpretation library.
        --max-iterations=<int>                             - Maximum number of invocations of the simplification pipeline.
  --verify-diagnostics                                     - Check that emitted diagnostics match expected-* lines on the corresponding line
  --verify-each                                            - Run the verifier after each transformation pass
  --verify-roundtrip                                       - Round-trip the IR after parsing and ensure it succeeds

Generic Options:

  --help                                                   - Display available options (--help-hidden for more)
  --help-list                                              - Display list of available options (--help-list-hidden for more)
  --version                                                - Display the version of this program