RecordingMicroAllocator

class RecordingMicroAllocator : public tflite::MicroAllocator

Utility subclass of MicroAllocator that records all allocations inside the arena. A summary of allocations can be logged through the ErrorReporter by invoking LogAllocations(). This special allocator requires an instance of RecordingSingleArenaBufferAllocator to capture allocations in the head and tail. Arena allocation recording can be retrieved by type through the GetRecordedAllocation() function. This class should only be used for auditing memory usage or integration testing.

static RecordingMicroAllocator * tflite::RecordingMicroAllocator::Create(uint8_t *tensor_arena, size_t arena_size)

Parameters

Parameters of Create
Name	Type	Default	Description
`tensor_arena`	`uint8_t *`	Required
`arena_size`	`size_t`	Required

static size_t tflite::RecordingMicroAllocator::GetDefaultTailUsage()

Returns the fixed amount of memory overhead of RecordingMicroAllocator.

RecordedAllocation tflite::RecordingMicroAllocator::GetRecordedAllocation(RecordedAllocationType allocation_type) const

Returns the recorded allocations information for a given allocation type.

Parameters

Parameters of GetRecordedAllocation
Name	Type	Default	Description
`allocation_type`	`RecordedAllocationType`	Required

const RecordingSingleArenaBufferAllocator * tflite::RecordingMicroAllocator::GetSimpleMemoryAllocator() const

void tflite::RecordingMicroAllocator::PrintAllocations() const

Logs out through the ErrorReporter all allocation recordings by type defined in RecordedAllocationType.

void * tflite::RecordingMicroAllocator::AllocatePersistentBuffer(size_t bytes)

Allocates persistent buffer which has the same life time as the allocator. The memory is immediately available and is allocated from the tail of the arena.

Parameters

Parameters of AllocatePersistentBuffer
Name	Type	Default	Description
`bytes`	`size_t`	Required

TfLiteStatus tflite::RecordingMicroAllocator::AllocateNodeAndRegistrations(
    const Model *model,
    SubgraphAllocations *subgraph_allocations
)

Allocates an array in the arena to hold pointers to the node and registration pointers required to represent the inference graph of the model.

Parameters

Parameters of AllocateNodeAndRegistrations
Name	Type	Default	Description
`model`	`const Model *`	Required
`subgraph_allocations`	`SubgraphAllocations *`	Required

TfLiteStatus tflite::RecordingMicroAllocator::AllocateTfLiteEvalTensors(
    const Model *model,
    SubgraphAllocations *subgraph_allocations
)

Allocates the list of persistent TfLiteEvalTensors that are used for the “eval” phase of model inference. These structs will be the source of truth for all tensor buffers.

Parameters

Parameters of AllocateTfLiteEvalTensors
Name	Type	Default	Description
`model`	`const Model *`	Required
`subgraph_allocations`	`SubgraphAllocations *`	Required

TfLiteStatus tflite::RecordingMicroAllocator::AllocateVariables(
    const SubGraph *subgraph,
    TfLiteEvalTensor *eval_tensors,
    const int32_t *offline_planner_offsets
)

Allocates persistent tensor buffers for variable tensors in the subgraph. Online and offline variable tensors are handled differently hence the offline_planner_offsets parameter is needed.

Parameters

Parameters of AllocateVariables
Name	Type	Default
`subgraph`	`const SubGraph *`	Required
`eval_tensors`	`TfLiteEvalTensor *`	Required
`offline_planner_offsets`	`const int32_t *`	Required

TfLiteTensor * tflite::RecordingMicroAllocator::AllocatePersistentTfLiteTensorInternal()

TODO(b/162311891): Once all kernels have been updated to the new API drop this method. It is only used to record TfLiteTensor persistent allocations.

TfLiteStatus tflite::RecordingMicroAllocator::PopulateTfLiteTensorFromFlatbuffer(
    const Model *model,
    TfLiteTensor *tensor,
    int tensor_index,
    int subgraph_index,
    bool allocate_temp
)

TODO(b/162311891): Once all kernels have been updated to the new API drop this function since all allocations for quantized data will take place in the temp section.

Parameters

Parameters of PopulateTfLiteTensorFromFlatbuffer
Name	Type	Default
`model`	`const Model *`	Required
`tensor`	`TfLiteTensor *`	Required
`tensor_index`	`int`	Required
`subgraph_index`	`int`	Required
`allocate_temp`	`bool`	Required