LoadDataUnzip

Applicability

Product	Supported
Atlas 350 Accelerator Card	x
Atlas A3 training product/Atlas A3 inference product	x
Atlas A2 training product/Atlas A2 inference product	x
Atlas 200I/500 A2 inference product	x
Atlas inference product AI Core	√
Atlas inference product Vector Core	x
Atlas training product	x

Function Usage

Decompress the data on the GM and load the data to A1, B1, and B2. LoadUnzipIndex needs to be executed to load the compression index table before this API is called.

Prototype

template <typename T>
__aicore__ inline void LoadDataUnzip(const LocalTensor<T>& dst, const GlobalTensor<T>& src)

Parameters

**Table 1** Parameters
Parameter	Input/Output	Description
dst	Output	Destination operand. Type: LocalTensor. Supported TPosition: A1/B1/B2. When TPosition is A1/B1, the start address of LocalTensor must be 32-byte aligned. When TPosition is B2, the start address of LocalTensor must be 512-byte aligned. The supported data type is int8_t.
src	Input	Source operand of the GlobalTensor type. Its data type must be the same as that of dst.

Restrictions

For details about the operand address alignment requirements, see General Address Alignment Restrictions.

Returns

None

Examples

This example is supported on the Atlas inference product AI Core.

uint32_t srcLen = 896, dstLen = 1024, numOfIndexTabEntry = 1;
AscendC::LocalTensor<int8_t> weightB1 = inQueueB1.AllocTensor<int8_t>();
AscendC::LoadUnzipIndex(indexGlobal, numOfIndexTabEntry); // Load index data and transfer the compressed index table stored in GM into internal registers.
AscendC::LoadDataUnzip(weightB1, weGlobal); // Load data based on the index table in internal registers.

Parent topic: Data Movement