# `Beaver.MLIR.Conversion` High-level, diagnostic-aware MLIR dialect conversion. For a declarative, inspectable, and reusable pipeline, see `Beaver.MLIR.Conversion.Plan`. `apply/5` runs full or partial conversion outside BEAM scheduler threads and services callback-backed targets, type converters, and conversion patterns in the calling process. Mutable pattern sets are frozen and cleaned up automatically. See the [dialect conversion guide](dialect-conversion.html) for a complete Slang-to-LLVM example. # `conversion_ir` ```elixir @type conversion_ir() :: Beaver.MLIR.Module.t() | Beaver.MLIR.Operation.t() ``` # `mode` ```elixir @type mode() :: :full | :partial ``` # `result` ```elixir @type result() :: {:ok, conversion_ir(), [term()]} | {:error, Beaver.MLIR.Conversion.Error.t()} ``` # `acc_to_llvm` Lower OpenACC operations to LLVM dialect. ### Argument name in MLIR CLI `acc-to-llvm` ### Pass name in TableGen `ConvertACCToLLVMPass` # `acc_to_llvm` # `apply` ```elixir @spec apply( mode(), conversion_ir(), Beaver.MLIR.ConversionTarget.t(), Beaver.MLIR.RewritePatternSet.t() | Beaver.MLIR.FrozenRewritePatternSet.t(), keyword() ) :: result() ``` # `apply!` ```elixir @spec apply!( mode(), conversion_ir(), Beaver.MLIR.ConversionTarget.t(), Beaver.MLIR.RewritePatternSet.t() | Beaver.MLIR.FrozenRewritePatternSet.t(), keyword() ) :: conversion_ir() ``` # `arm_neon_2d_to_intr` Convert Arm NEON structured ops to intrinsics ### Argument name in MLIR CLI `arm-neon-2d-to-intr` ### Pass name in TableGen `ConvertArmNeon2dToIntrPass` # `arm_neon_2d_to_intr` # `convert_affine_for_to_gpu` Convert top-level AffineFor Ops to GPU kernels ### Argument name in MLIR CLI `convert-affine-for-to-gpu` ### Pass name in TableGen `ConvertAffineForToGPUPass` # `convert_affine_for_to_gpu` # `convert_amdgpu_to_rocdl` Convert AMDGPU dialect to ROCDL dialect ### Argument name in MLIR CLI `convert-amdgpu-to-rocdl` ### Pass name in TableGen `ConvertAMDGPUToROCDLPass` # `convert_amdgpu_to_rocdl` # `convert_arith_to_amdgpu` Convert Arith operations to AMDGPU-specific implementations ### Argument name in MLIR CLI `convert-arith-to-amdgpu` ### Pass name in TableGen `ArithToAMDGPUConversionPass` # `convert_arith_to_amdgpu` # `convert_arith_to_apfloat` Convert Arith ops to APFloat runtime library calls ### Argument name in MLIR CLI `convert-arith-to-apfloat` ### Pass name in TableGen `ArithToAPFloatConversionPass` # `convert_arith_to_apfloat` # `convert_arith_to_arm_sme` Convert Arith dialect to ArmSME dialect ### Argument name in MLIR CLI `convert-arith-to-arm-sme` ### Pass name in TableGen `ArithToArmSMEConversionPass` # `convert_arith_to_arm_sme` # `convert_arith_to_emitc` Convert Arith dialect to EmitC dialect ### Argument name in MLIR CLI `convert-arith-to-emitc` ### Pass name in TableGen `ConvertArithToEmitC` # `convert_arith_to_emitc` # `convert_arith_to_llvm` Convert Arith dialect to LLVM dialect ### Argument name in MLIR CLI `convert-arith-to-llvm` ### Pass name in TableGen `ArithToLLVMConversionPass` # `convert_arith_to_llvm` # `convert_arith_to_spirv` Convert Arith dialect to SPIR-V dialect ### Argument name in MLIR CLI `convert-arith-to-spirv` ### Pass name in TableGen `ConvertArithToSPIRVPass` # `convert_arith_to_spirv` # `convert_arm_sme_to_llvm` Lower the operations from the ArmSME dialect into the LLVM dialect ### Argument name in MLIR CLI `convert-arm-sme-to-llvm` ### Pass name in TableGen `ConvertArmSMEToLLVM` # `convert_arm_sme_to_llvm` # `convert_arm_sme_to_scf` Lower the operations from the ArmSME dialect into the SCF dialect ### Argument name in MLIR CLI `convert-arm-sme-to-scf` ### Pass name in TableGen `ConvertArmSMEToSCFPass` # `convert_arm_sme_to_scf` # `convert_async_to_llvm` Convert the operations from the async dialect into the LLVM dialect ### Argument name in MLIR CLI `convert-async-to-llvm` ### Pass name in TableGen `ConvertAsyncToLLVMPass` # `convert_async_to_llvm` # `convert_bufferization_to_memref` Convert operations from the Bufferization dialect to the MemRef dialect ### Argument name in MLIR CLI `convert-bufferization-to-memref` ### Pass name in TableGen `ConvertBufferizationToMemRefPass` # `convert_bufferization_to_memref` # `convert_cf_to_llvm` Convert ControlFlow operations to the LLVM dialect ### Argument name in MLIR CLI `convert-cf-to-llvm` ### Pass name in TableGen `ConvertControlFlowToLLVMPass` # `convert_cf_to_llvm` # `convert_cf_to_spirv` Convert ControlFlow dialect to SPIR-V dialect ### Argument name in MLIR CLI `convert-cf-to-spirv` ### Pass name in TableGen `ConvertControlFlowToSPIRVPass` # `convert_cf_to_spirv` # `convert_complex_to_libm` Convert Complex dialect to libm calls ### Argument name in MLIR CLI `convert-complex-to-libm` ### Pass name in TableGen `ConvertComplexToLibm` # `convert_complex_to_libm` # `convert_complex_to_llvm` Convert Complex dialect to LLVM dialect ### Argument name in MLIR CLI `convert-complex-to-llvm` ### Pass name in TableGen `ConvertComplexToLLVMPass` # `convert_complex_to_llvm` # `convert_complex_to_rocdl_library_calls` Convert Complex dialect to ROCDL library calls ### Argument name in MLIR CLI `convert-complex-to-rocdl-library-calls` ### Pass name in TableGen `ConvertComplexToROCDLLibraryCalls` # `convert_complex_to_rocdl_library_calls` # `convert_complex_to_spirv` Convert Complex dialect to SPIRV dialect ### Argument name in MLIR CLI `convert-complex-to-spirv` ### Pass name in TableGen `ConvertComplexToSPIRVPass` # `convert_complex_to_spirv` # `convert_complex_to_standard` Convert Complex dialect to standard dialect ### Argument name in MLIR CLI `convert-complex-to-standard` ### Pass name in TableGen `ConvertComplexToStandardPass` # `convert_complex_to_standard` # `convert_func_to_emitc` Convert Func dialect to EmitC dialect ### Argument name in MLIR CLI `convert-func-to-emitc` ### Pass name in TableGen `ConvertFuncToEmitC` # `convert_func_to_emitc` # `convert_func_to_llvm` Convert from the Func dialect to the LLVM dialect ### Argument name in MLIR CLI `convert-func-to-llvm` ### Pass name in TableGen `ConvertFuncToLLVMPass` # `convert_func_to_llvm` # `convert_func_to_spirv` Convert Func dialect to SPIR-V dialect ### Argument name in MLIR CLI `convert-func-to-spirv` ### Pass name in TableGen `ConvertFuncToSPIRVPass` # `convert_func_to_spirv` # `convert_gpu_to_llvm_spv` Generate LLVM operations to be ingested by a SPIR-V backend for gpu operations ### Argument name in MLIR CLI `convert-gpu-to-llvm-spv` ### Pass name in TableGen `ConvertGpuOpsToLLVMSPVOps` # `convert_gpu_to_llvm_spv` # `convert_gpu_to_nvvm` Generate NVVM operations for gpu operations ### Argument name in MLIR CLI `convert-gpu-to-nvvm` ### Pass name in TableGen `ConvertGpuOpsToNVVMOps` # `convert_gpu_to_nvvm` # `convert_gpu_to_rocdl` Generate ROCDL operations for gpu operations ### Argument name in MLIR CLI `convert-gpu-to-rocdl` ### Pass name in TableGen `ConvertGpuOpsToROCDLOps` # `convert_gpu_to_rocdl` # `convert_gpu_to_spirv` Convert GPU dialect to SPIR-V dialect ### Argument name in MLIR CLI `convert-gpu-to-spirv` ### Pass name in TableGen `ConvertGPUToSPIRV` # `convert_gpu_to_spirv` # `convert_index_to_llvm` Lower the `index` dialect to the `llvm` dialect. ### Argument name in MLIR CLI `convert-index-to-llvm` ### Pass name in TableGen `ConvertIndexToLLVMPass` # `convert_index_to_llvm` # `convert_index_to_spirv` Lower the `index` dialect to the `spirv` dialect. ### Argument name in MLIR CLI `convert-index-to-spirv` ### Pass name in TableGen `ConvertIndexToSPIRVPass` # `convert_index_to_spirv` # `convert_linalg_to_std` Convert the operations from the linalg dialect into the Standard dialect ### Argument name in MLIR CLI `convert-linalg-to-std` ### Pass name in TableGen `ConvertLinalgToStandardPass` # `convert_linalg_to_std` # `convert_math_to_apfloat` Convert Math ops to APFloat runtime library calls ### Argument name in MLIR CLI `convert-math-to-apfloat` ### Pass name in TableGen `MathToAPFloatConversionPass` # `convert_math_to_apfloat` # `convert_math_to_emitc` Convert some Math operations to EmitC call_opaque operations ### Argument name in MLIR CLI `convert-math-to-emitc` ### Pass name in TableGen `ConvertMathToEmitC` # `convert_math_to_emitc` # `convert_math_to_funcs` Convert Math operations to calls of outlined implementations. ### Argument name in MLIR CLI `convert-math-to-funcs` ### Pass name in TableGen `ConvertMathToFuncs` # `convert_math_to_funcs` # `convert_math_to_libm` Convert Math dialect to libm calls ### Argument name in MLIR CLI `convert-math-to-libm` ### Pass name in TableGen `ConvertMathToLibmPass` # `convert_math_to_libm` # `convert_math_to_llvm` Convert Math dialect to LLVM dialect ### Argument name in MLIR CLI `convert-math-to-llvm` ### Pass name in TableGen `ConvertMathToLLVMPass` # `convert_math_to_llvm` # `convert_math_to_nvvm` Convert Math dialect to CUDA libdevice calls ### Argument name in MLIR CLI `convert-math-to-nvvm` ### Pass name in TableGen `ConvertMathToNVVM` # `convert_math_to_nvvm` # `convert_math_to_rocdl` Convert Math dialect to ROCDL library calls ### Argument name in MLIR CLI `convert-math-to-rocdl` ### Pass name in TableGen `ConvertMathToROCDL` # `convert_math_to_rocdl` # `convert_math_to_spirv` Convert Math dialect to SPIR-V dialect ### Argument name in MLIR CLI `convert-math-to-spirv` ### Pass name in TableGen `ConvertMathToSPIRVPass` # `convert_math_to_spirv` # `convert_math_to_xevm` Convert (fast) math operations to native XeVM/SPIRV equivalents ### Argument name in MLIR CLI `convert-math-to-xevm` ### Pass name in TableGen `ConvertMathToXeVM` # `convert_math_to_xevm` # `convert_memref_to_emitc` Convert MemRef dialect to EmitC dialect ### Argument name in MLIR CLI `convert-memref-to-emitc` ### Pass name in TableGen `ConvertMemRefToEmitC` # `convert_memref_to_emitc` # `convert_memref_to_spirv` Convert MemRef dialect to SPIR-V dialect ### Argument name in MLIR CLI `convert-memref-to-spirv` ### Pass name in TableGen `ConvertMemRefToSPIRVPass` # `convert_memref_to_spirv` # `convert_nvgpu_to_nvvm` Convert NVGPU dialect to NVVM dialect ### Argument name in MLIR CLI `convert-nvgpu-to-nvvm` ### Pass name in TableGen `ConvertNVGPUToNVVMPass` # `convert_nvgpu_to_nvvm` # `convert_nvvm_to_llvm` Convert NVVM to PTX with Inline Assembly in LLVM dialect ### Argument name in MLIR CLI `convert-nvvm-to-llvm` ### Pass name in TableGen `ConvertNVVMToLLVMPass` # `convert_nvvm_to_llvm` # `convert_openacc_to_scf` Convert the OpenACC ops to OpenACC with SCF dialect ### Argument name in MLIR CLI `convert-openacc-to-scf` ### Pass name in TableGen `ConvertOpenACCToSCFPass` # `convert_openacc_to_scf` # `convert_openmp_to_llvm` Convert the OpenMP ops to OpenMP ops with LLVM dialect ### Argument name in MLIR CLI `convert-openmp-to-llvm` ### Pass name in TableGen `ConvertOpenMPToLLVMPass` # `convert_openmp_to_llvm` # `convert_parallel_loops_to_gpu` Convert mapped scf.parallel ops to gpu launch operations ### Argument name in MLIR CLI `convert-parallel-loops-to-gpu` ### Pass name in TableGen `ConvertParallelLoopToGpuPass` # `convert_parallel_loops_to_gpu` # `convert_pdl_to_pdl_interp` Convert PDL ops to PDL interpreter ops ### Argument name in MLIR CLI `convert-pdl-to-pdl-interp` ### Pass name in TableGen `ConvertPDLToPDLInterpPass` # `convert_pdl_to_pdl_interp` # `convert_scf_to_cf` Convert SCF dialect to ControlFlow dialect, replacing structured control flow with a CFG ### Argument name in MLIR CLI `convert-scf-to-cf` ### Pass name in TableGen `SCFToControlFlowPass` # `convert_scf_to_cf` # `convert_scf_to_emitc` Convert SCF dialect to EmitC dialect, maintaining structured control flow ### Argument name in MLIR CLI `convert-scf-to-emitc` ### Pass name in TableGen `SCFToEmitC` # `convert_scf_to_emitc` # `convert_scf_to_openmp` Convert SCF parallel loop to OpenMP parallel + workshare constructs. ### Argument name in MLIR CLI `convert-scf-to-openmp` ### Pass name in TableGen `ConvertSCFToOpenMPPass` # `convert_scf_to_openmp` # `convert_scf_to_spirv` Convert SCF dialect to SPIR-V dialect. ### Argument name in MLIR CLI `convert-scf-to-spirv` ### Pass name in TableGen `SCFToSPIRV` # `convert_scf_to_spirv` # `convert_shape_constraints` Convert shape constraint operations to the standard dialect ### Argument name in MLIR CLI `convert-shape-constraints` ### Pass name in TableGen `ConvertShapeConstraintsPass` # `convert_shape_constraints` # `convert_shape_to_std` Convert operations from the shape dialect into the standard dialect ### Argument name in MLIR CLI `convert-shape-to-std` ### Pass name in TableGen `ConvertShapeToStandardPass` # `convert_shape_to_std` # `convert_shard_to_mpi` Convert Shard dialect to MPI dialect. ### Argument name in MLIR CLI `convert-shard-to-mpi` ### Pass name in TableGen `ConvertShardToMPIPass` # `convert_shard_to_mpi` # `convert_spirv_to_llvm` Convert SPIR-V dialect to LLVM dialect ### Argument name in MLIR CLI `convert-spirv-to-llvm` ### Pass name in TableGen `ConvertSPIRVToLLVMPass` # `convert_spirv_to_llvm` # `convert_tensor_to_linalg` Convert some Tensor dialect ops to Linalg dialect ### Argument name in MLIR CLI `convert-tensor-to-linalg` ### Pass name in TableGen `ConvertTensorToLinalgPass` # `convert_tensor_to_linalg` # `convert_tensor_to_spirv` Convert Tensor dialect to SPIR-V dialect ### Argument name in MLIR CLI `convert-tensor-to-spirv` ### Pass name in TableGen `ConvertTensorToSPIRVPass` # `convert_tensor_to_spirv` # `convert_to_emitc` Convert to EmitC dialect via dialect interfaces ### Argument name in MLIR CLI `convert-to-emitc` ### Pass name in TableGen `ConvertToEmitC` # `convert_to_emitc` # `convert_to_llvm` Convert to LLVM via dialect interfaces found in the input IR ### Argument name in MLIR CLI `convert-to-llvm` ### Pass name in TableGen `ConvertToLLVMPass` # `convert_to_llvm` # `convert_ub_to_llvm` Convert UB dialect to LLVM dialect ### Argument name in MLIR CLI `convert-ub-to-llvm` ### Pass name in TableGen `UBToLLVMConversionPass` # `convert_ub_to_llvm` # `convert_ub_to_spirv` Convert UB dialect to SPIR-V dialect ### Argument name in MLIR CLI `convert-ub-to-spirv` ### Pass name in TableGen `UBToSPIRVConversionPass` # `convert_ub_to_spirv` # `convert_vector_to_amx` Lower the operations from the vector dialect into the X86 dialect AMX operations ### Argument name in MLIR CLI `convert-vector-to-amx` ### Pass name in TableGen `ConvertVectorToAMX` # `convert_vector_to_amx` # `convert_vector_to_arm_sme` Lower the operations from the vector dialect into the ArmSME dialect ### Argument name in MLIR CLI `convert-vector-to-arm-sme` ### Pass name in TableGen `ConvertVectorToArmSMEPass` # `convert_vector_to_arm_sme` # `convert_vector_to_gpu` Lower the operations from the vector dialect into the GPU dialect ### Argument name in MLIR CLI `convert-vector-to-gpu` ### Pass name in TableGen `ConvertVectorToGPU` # `convert_vector_to_gpu` # `convert_vector_to_llvm` Lower the operations from the vector dialect into the LLVM dialect ### Argument name in MLIR CLI `convert-vector-to-llvm` ### Pass name in TableGen `ConvertVectorToLLVMPass` # `convert_vector_to_llvm` # `convert_vector_to_scf` Lower the operations from the vector dialect into the SCF dialect ### Argument name in MLIR CLI `convert-vector-to-scf` ### Pass name in TableGen `ConvertVectorToSCF` # `convert_vector_to_scf` # `convert_vector_to_spirv` Convert Vector dialect to SPIR-V dialect ### Argument name in MLIR CLI `convert-vector-to-spirv` ### Pass name in TableGen `ConvertVectorToSPIRVPass` # `convert_vector_to_spirv` # `convert_vector_to_xegpu` Lower the operations from the vector dialect into the XeGPU dialect ### Argument name in MLIR CLI `convert-vector-to-xegpu` ### Pass name in TableGen `ConvertVectorToXeGPU` # `convert_vector_to_xegpu` # `convert_xegpu_to_xevm` Convert XeGPU to XeVM dialect ### Argument name in MLIR CLI `convert-xegpu-to-xevm` ### Pass name in TableGen `ConvertXeGPUToXeVMPass` # `convert_xegpu_to_xevm` # `convert_xevm_to_llvm` Convert XeVM to LLVM dialect ### Argument name in MLIR CLI `convert-xevm-to-llvm` ### Pass name in TableGen `ConvertXeVMToLLVMPass` # `convert_xevm_to_llvm` # `finalize_memref_to_llvm` Finalize MemRef dialect to LLVM dialect conversion ### Argument name in MLIR CLI `finalize-memref-to-llvm` ### Pass name in TableGen `FinalizeMemRefToLLVMConversionPass` # `finalize_memref_to_llvm` # `full` ```elixir @spec full( conversion_ir(), Beaver.MLIR.ConversionTarget.t(), Beaver.MLIR.RewritePatternSet.t() | Beaver.MLIR.FrozenRewritePatternSet.t(), keyword() ) :: result() ``` # `gpu_to_llvm` Convert GPU dialect to LLVM dialect with GPU runtime calls ### Argument name in MLIR CLI `gpu-to-llvm` ### Pass name in TableGen `GpuToLLVMConversionPass` # `gpu_to_llvm` # `lift_cf_to_scf` Lift ControlFlow dialect to SCF dialect ### Argument name in MLIR CLI `lift-cf-to-scf` ### Pass name in TableGen `LiftControlFlowToSCFPass` # `lift_cf_to_scf` # `lower_affine` Lower Affine operations to a combination of Arith and SCF operations ### Argument name in MLIR CLI `lower-affine` ### Pass name in TableGen `LowerAffinePass` # `lower_affine` # `lower_host_to_llvm` Lowers the host module code and `gpu.launch_func` to LLVM ### Argument name in MLIR CLI `lower-host-to-llvm` ### Pass name in TableGen `LowerHostCodeToLLVMPass` # `lower_host_to_llvm` # `map_memref_spirv_storage_class` Map numeric MemRef memory spaces to SPIR-V storage classes ### Argument name in MLIR CLI `map-memref-spirv-storage-class` ### Pass name in TableGen `MapMemRefStorageClass` # `map_memref_spirv_storage_class` # `partial` ```elixir @spec partial( conversion_ir(), Beaver.MLIR.ConversionTarget.t(), Beaver.MLIR.RewritePatternSet.t() | Beaver.MLIR.FrozenRewritePatternSet.t(), keyword() ) :: result() ``` # `raise_scf_to_affine` Raise SCF operations to affine operations (best effort) ### Argument name in MLIR CLI `raise-scf-to-affine` ### Pass name in TableGen `RaiseSCFToAffinePass` # `raise_scf_to_affine` # `raise_wasm_mlir` Convert Wasm dialect to a group of dialect as a bridge to LLVM MLIR conversion ### Argument name in MLIR CLI `raise-wasm-mlir` ### Pass name in TableGen `RaiseWasmMLIR` # `raise_wasm_mlir` # `reconcile_unrealized_casts` Simplify and eliminate unrealized conversion casts ### Argument name in MLIR CLI `reconcile-unrealized-casts` ### Pass name in TableGen `ReconcileUnrealizedCastsPass` # `reconcile_unrealized_casts` # `set_llvm_module_datalayout` Attach a datalayout string as a module attribute ### Argument name in MLIR CLI `set-llvm-module-datalayout` ### Pass name in TableGen `SetLLVMModuleDataLayoutPass` # `set_llvm_module_datalayout` # `tosa_to_arith` Lower TOSA to the Arith dialect ### Argument name in MLIR CLI `tosa-to-arith` ### Pass name in TableGen `TosaToArithPass` # `tosa_to_arith` # `tosa_to_linalg` Lower TOSA to LinAlg on tensors ### Argument name in MLIR CLI `tosa-to-linalg` ### Pass name in TableGen `TosaToLinalg` # `tosa_to_linalg` # `tosa_to_linalg_named` Lower TOSA to LinAlg named operations ### Argument name in MLIR CLI `tosa-to-linalg-named` ### Pass name in TableGen `TosaToLinalgNamed` # `tosa_to_linalg_named` # `tosa_to_mlprogram` Lower TOSA to the MLProgram dialect ### Argument name in MLIR CLI `tosa-to-mlprogram` ### Pass name in TableGen `TosaToMLProgram` # `tosa_to_mlprogram` # `tosa_to_scf` Lower TOSA to the SCF dialect ### Argument name in MLIR CLI `tosa-to-scf` ### Pass name in TableGen `TosaToSCFPass` # `tosa_to_scf` # `tosa_to_spirv_tosa` Lower TOSA IR to SPIR-V Graph/TOSA operations ### Argument name in MLIR CLI `tosa-to-spirv-tosa` ### Pass name in TableGen `TosaToSPIRVTosa` # `tosa_to_spirv_tosa` # `tosa_to_spirv_tosa_mark_graph_constants` Mark large TOSA constants as SPIR-V Graph constants ### Argument name in MLIR CLI `tosa-to-spirv-tosa-mark-graph-constants` ### Pass name in TableGen `TosaToSPIRVTosaMarkGraphConstants` # `tosa_to_spirv_tosa_mark_graph_constants` # `tosa_to_tensor` Lower TOSA to the Tensor dialect ### Argument name in MLIR CLI `tosa-to-tensor` ### Pass name in TableGen `TosaToTensorPass` # `tosa_to_tensor`