mlir.dialects._bufferization_transform_ops_gen¶

Attributes¶

_ods_ir

Classes¶

`BufferLoopHoistingOp`	Hoist buffer allocations ("memref.alloc" and "memref.alloca") from loops
`EliminateEmptyTensorsOp`	Try to eliminate all `tensor.empty` ops within the targeted op by replacing
`EmptyTensorToAllocTensorOp`	Replace a tensor.empty with a bufferization.tensor_alloc.
`OneShotBufferizeOp`	Indicates that the given `target` op should be bufferized with One-Shot

Functions¶

`bufferization_buffer_loop_hoisting`(→ BufferLoopHoistingOp)
`bufferization_eliminate_empty_tensors`(...)
`bufferization_empty_tensor_to_alloc_tensor`(→ _ods_ir)
`bufferization_one_shot_bufferize`(→ _ods_ir)

Module Contents¶

mlir.dialects._bufferization_transform_ops_gen._ods_ir¶

class mlir.dialects._bufferization_transform_ops_gen.BufferLoopHoistingOp(target, *, loc=None, ip=None)¶

Bases: _ods_ir

Hoist buffer allocations (“memref.alloc” and “memref.alloca”) from loops within the targeted op. This transform assumes that there are no buffer deallocation ops in the IR.

This transform reads the target handle and modifies the payload.

OPERATION_NAME = 'transform.bufferization.buffer_loop_hoisting'¶

_ODS_REGIONS = (0, True)¶

target() → _ods_ir¶

mlir.dialects._bufferization_transform_ops_gen.bufferization_buffer_loop_hoisting(target, *, loc=None, ip=None) → BufferLoopHoistingOp¶

class mlir.dialects._bufferization_transform_ops_gen.EliminateEmptyTensorsOp(target, *, loc=None, ip=None)¶

Bases: _ods_ir

Try to eliminate all tensor.empty ops within the targeted op by replacing them with another destination tensor.

“tensor.empty” ops cannot be bufferized. They can either be converted to “bufferization.alloc_tensor” or replaced with another tensor (via this transform). “tensor.empty” does not specify the contents of the returned tensor so their results can be replaced with arbitrary tensor values as long as the dimensions match.

This transformation looks for subset ops that insert a tensor that originates from a “tensor.empty” (as per the reverse use-def chain). Such “tensor.empty” ops are replaced with the destination subset.

Example:

%0 = tensor.empty() : tensor<5xf32>
%1 = linalg.fill ... outs(%0)
%2 = tensor.insert_slice %1 into %t[1][5][1]

Is rewritten with:

%0 = tensor.extract_slice %t[1][5][1]
%1 = linalg.fill ... outs(%0)
%2 = tensor.insert_slice %1 into %t[1][5][1]

In the above example, the subset op is “tensor.insert_slice”. When tracing back the reverse use-def chain of a the source, we end up at a “tensor.empty” op.

The above example can bufferize without an allocation (in the absence of other conflicts) because there is no longer a tensor.empty op.

See -eliminate-empty-tensors for more details.

Return modes¶

This transform reads the target handle and modifies the payload. It does not produce any handle.

OPERATION_NAME = 'transform.bufferization.eliminate_empty_tensors'¶

_ODS_REGIONS = (0, True)¶

target() → _ods_ir¶

mlir.dialects._bufferization_transform_ops_gen.bufferization_eliminate_empty_tensors(target, *, loc=None, ip=None) → EliminateEmptyTensorsOp¶

class mlir.dialects._bufferization_transform_ops_gen.EmptyTensorToAllocTensorOp(transformed, target, *, loc=None, ip=None)¶

Bases: _ods_ir

Replace a tensor.empty with a bufferization.tensor_alloc.

Return modes¶

This operation consumes the target handle and produces the transformed handle. target is expected to be a tensor.empty operation. The transform always succeeds.

OPERATION_NAME = 'transform.bufferization.empty_tensor_to_alloc_tensor'¶

_ODS_REGIONS = (0, True)¶

target() → _ods_ir¶

transformed() → _ods_ir¶

mlir.dialects._bufferization_transform_ops_gen.bufferization_empty_tensor_to_alloc_tensor(transformed, target, *, loc=None, ip=None) → _ods_ir¶

class mlir.dialects._bufferization_transform_ops_gen.OneShotBufferizeOp(transformed, target, *, function_boundary_type_conversion=None, allow_return_allocs_from_loops=None, allow_unknown_ops=None, bufferize_function_boundaries=None, dump_alias_sets=None, test_analysis_only=None, print_conflicts=None, check_parallel_regions=None, memcpy_op=None, loc=None, ip=None)¶

Bases: _ods_ir

Indicates that the given target op should be bufferized with One-Shot Bufferize. The bufferization can be configured with various attributes that corresponding to options in BufferizationOptions and the one-shot-bufferize pass. More information can be found in the pass documentation.

The targeted ops must be modules or functions. This is because there is always a single, bufferized replacement op for such targets.

Note: Only ops that implement BufferizableOpInterface are bufferized. All other ops are ignored if allow_unknown_ops. If allow_unknown_ops is unset, this transform fails when an unknown/non-bufferizable op is found. Many ops implement BufferizableOpInterface via an external model. These external models must be registered when applying this transform op; otherwise, said ops would be considered non-bufferizable.

Return modes¶

This operation consumes the target handle and produces the transformed handle.

OPERATION_NAME = 'transform.bufferization.one_shot_bufferize'¶

_ODS_REGIONS = (0, True)¶

target() → _ods_ir¶

function_boundary_type_conversion() → _ods_ir | None¶

allow_return_allocs_from_loops() → _ods_ir¶

allow_unknown_ops() → _ods_ir¶

bufferize_function_boundaries() → _ods_ir¶

dump_alias_sets() → _ods_ir¶

test_analysis_only() → _ods_ir¶

print_conflicts() → _ods_ir¶

check_parallel_regions() → _ods_ir¶

memcpy_op() → _ods_ir¶

transformed() → _ods_ir¶

mlir.dialects._bufferization_transform_ops_gen.bufferization_one_shot_bufferize(transformed, target, *, function_boundary_type_conversion=None, allow_return_allocs_from_loops=None, allow_unknown_ops=None, bufferize_function_boundaries=None, dump_alias_sets=None, test_analysis_only=None, print_conflicts=None, check_parallel_regions=None, memcpy_op=None, loc=None, ip=None) → _ods_ir¶