Inference Engine Memory Primitives¶

Inference Memory Primitives (C++)¶

C++

Blobs¶

InferenceEngine::Blob is the main class intended for working with memory. Using this class you can read and write memory, get information about the memory structure etc.

The right way to create Blob objects with a specific layout is to use constructors with InferenceEngine::TensorDesc.

:ref:`InferenceEngine::TensorDesc <doxid-class_inference_engine_1_1_tensor_desc>` tdesc(FP32, {1, 3, 227, 227}, InferenceEngine::Layout::NCHW);
:ref:`InferenceEngine::Blob::Ptr <doxid-class_inference_engine_1_1_blob_1abb6c4f89181e2dd6d8a29ada2dfb4060>` blob = InferenceEngine::make_shared_blob<float>(tdesc);

Layouts¶

InferenceEngine::TensorDesc is a special class that provides layout format description.

This class allows to create planar layouts using the standard formats (like InferenceEngine::Layout::NCDHW, InferenceEngine::Layout::NCHW, InferenceEngine::Layout::NC, InferenceEngine::Layout::C and etc) and also non-planar layouts using InferenceEngine::BlockingDesc.

In order to create a complex layout you should use InferenceEngine::BlockingDesc, which allows you to define the blocked memory with offsets and strides.

Examples¶

You can define a blob with dimensions {N: 1, C: 25, H: 20, W: 20} and format NHWC with using next parameters:

:ref:`InferenceEngine::BlockingDesc <doxid-class_inference_engine_1_1_blocking_desc>`({1, 20, 20, 25}, {0, 2, 3, 1}); // or
:ref:`InferenceEngine::BlockingDesc <doxid-class_inference_engine_1_1_blocking_desc>`({1, 20, 20, 25}, InferenceEngine::Layout::NHWC);

If you have a memory with real dimensions {N: 1, C: 25, H: 20, W: 20} but with channels that are blocked by 8, you can define it using next parameters:
```
:ref:`InferenceEngine::BlockingDesc <doxid-class_inference_engine_1_1_blocking_desc>`({1, 4, 20, 20, 8}, {0, 1, 2, 3, 1})
```
Also you can set strides and offsets if layout contains it.

If you have a complex blob layout and you don’t want to calculate the real offset to data you can use the InferenceEngine::TensorDesc::offset(size_t l) or InferenceEngine::TensorDesc::offset(SizeVector v) methods.

For example:

:ref:`InferenceEngine::BlockingDesc <doxid-class_inference_engine_1_1_blocking_desc>` blk({1, 4, 20, 20, 8}, {0, 1, 2, 3, 1});
:ref:`InferenceEngine::TensorDesc <doxid-class_inference_engine_1_1_tensor_desc>` tdesc(FP32, {1, 25, 20, 20}, blk);
tdesc.offset(0); // = 0
tdesc.offset(1); // = 8
tdesc.offset({0, 0, 0, 2}); // = 16
tdesc.offset({0, 1, 0, 2}); // = 17

If you would like to create a TensorDesc with a planar format and for N dimensions (N can be different 1, 2, 4 and etc), you can use the InferenceEngine::TensorDesc::getLayoutByDims method.

:ref:`InferenceEngine::TensorDesc::getLayoutByDims <doxid-class_inference_engine_1_1_tensor_desc_1a9d361162ecf211bbb86d2a380a9c8d5f>`({1}); // InferenceEngine::Layout::C
:ref:`InferenceEngine::TensorDesc::getLayoutByDims <doxid-class_inference_engine_1_1_tensor_desc_1a9d361162ecf211bbb86d2a380a9c8d5f>`({1, 2}); // InferenceEngine::Layout::NC
:ref:`InferenceEngine::TensorDesc::getLayoutByDims <doxid-class_inference_engine_1_1_tensor_desc_1a9d361162ecf211bbb86d2a380a9c8d5f>`({1, 2, 3, 4}); // InferenceEngine::Layout::NCHW
:ref:`InferenceEngine::TensorDesc::getLayoutByDims <doxid-class_inference_engine_1_1_tensor_desc_1a9d361162ecf211bbb86d2a380a9c8d5f>`({1, 2, 3}); // InferenceEngine::Layout::BLOCKED
:ref:`InferenceEngine::TensorDesc::getLayoutByDims <doxid-class_inference_engine_1_1_tensor_desc_1a9d361162ecf211bbb86d2a380a9c8d5f>`({1, 2, 3, 4, 5}); // InferenceEngine::Layout::NCDHW
:ref:`InferenceEngine::TensorDesc::getLayoutByDims <doxid-class_inference_engine_1_1_tensor_desc_1a9d361162ecf211bbb86d2a380a9c8d5f>`({1, 2, 3, 4, 5, ...}); // InferenceEngine::Layout::BLOCKED

Prev Next