AddDeqRelu

Applicability

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

Function Usage

Performs element-wise addition, then applies dequantization to the result, followed by a ReLU computation (comparing the result with 0 and taking the larger value). The formula is as follows:

$\text{[math]}$

The formula for calculating Deq is as follows:

$\text{[math]}$

In the formula, dividing by 2¹⁷ and then multiplying by 2¹⁷ prevents overflow caused by multiplying x by DeqScale. DeqScale must be set via SetDeqScale. For details, see SetDeqScale.

Prototype

Computation of the first n pieces of data of a tensor

        
             __aicore__ inline void AddDeqRelu(const LocalTensor<half>& dst, const LocalTensor<int32_t>& src0, const LocalTensor<int32_t>& src1, const int32_t& count)

High-dimensional tensor sharding computation

Bitwise mask mode

          
               template <bool isSetMask = true>
__aicore__ inline void AddDeqRelu(const LocalTensor<half>& dst, const LocalTensor<int32_t>& src0, const LocalTensor<int32_t>& src1, uint64_t mask[], const uint8_t repeatTime, const BinaryRepeatParams& repeatParams)

Contiguous mask mode

          
               template <bool isSetMask = true>
__aicore__ inline void AddDeqRelu(const LocalTensor<half>& dst, const LocalTensor<int32_t>& src0, const LocalTensor<int32_t>& src1, uint64_t mask, const uint8_t repeatTime, const BinaryRepeatParams& repeatParams)

When the operand is of the TensorTrait type, the template parameter is required for the LocalTensor. The following APIs are provided to support the input of operand data types as template parameters:

Computation of the first n pieces of data of a tensor

         
              template <typename T, typename U>
__aicore__ inline void AddDeqRelu(const LocalTensor<T>& dst, const LocalTensor<U>& src0, const LocalTensor<U>& src1, const int32_t& count)

High-dimensional tensor sharding computation

Bitwise mask mode

           
                template <typename T, typename U, bool isSetMask = true>
__aicore__ inline void AddDeqRelu(const LocalTensor<T>& dst, const LocalTensor<U>& src0, const LocalTensor<U>& src1, uint64_t mask[], const uint8_t repeatTime, const BinaryRepeatParams& repeatParams)

Contiguous mask mode

           
                template <typename T, typename U, bool isSetMask = true>
__aicore__ inline void AddDeqRelu(const LocalTensor<T>& dst, const LocalTensor<U>& src0, const LocalTensor<U>& src1, uint64_t mask, const uint8_t repeatTime, const BinaryRepeatParams& repeatParams)

Parameters

**Table 1** Template parameters
Parameter	Description
isSetMask	Indicates whether to set mask inside the API. true: sets mask inside the API. false: sets mask outside the API. Developers need to use the SetVectorMask API to set the mask value. In this mode, the mask value in the input parameter of this API must be set to the placeholder MASK_PLACEHOLDER.
T	Data type of the destination operand. For the Atlas 350 Accelerator Card, the supported data type is half. For the Atlas A3 training product / Atlas A3 inference product , the supported data type is half. For the Atlas A2 training product / Atlas A2 inference product , the supported data type is half. For the Atlas inference product AI Core, the supported data type is half.
U	Data type of the source operand. For the Atlas 350 Accelerator Card, the supported data type is int32_t. For the Atlas A3 training product / Atlas A3 inference product , the supported data type is int32_t. For the Atlas A2 training product / Atlas A2 inference product , the supported data type is int32_t. For the Atlas inference product AI Core, the supported data type is int32_t.

**Table 2** Parameters
Parameter	Input/Output	Description
dst	Output	Destination operand. The type is LocalTensor, and TPosition can be VECIN, VECCALC, or VECOUT. The start address of LocalTensor must be 32-byte aligned.
src0/src1	Input	Source operands. The type is LocalTensor, and TPosition can be VECIN, VECCALC, or VECOUT. The start address of LocalTensor must be 32-byte aligned.
count	Input	Number of elements involved in the computation.
mask[]/mask	Input	mask controls the elements that participate in computation in each iteration. Bitwise mode: controls the elements that participate in computation by bit. If a bit is set to 1, the corresponding element participates in the computation. If a bit is set to 0, the corresponding element is masked from the computation. The mask value is an array. The array length and the value range of the array elements are related to the operand data type. When the operand is 16-bit, the array length is 2, with mask[0] and mask[1] each in the range [0, 2⁶⁴ – 1], and they cannot both be 0 at the same time. When the operand is 32-bit, the array length is 1, with mask[0] in the range (0, 2⁶⁴ – 1]. When the operand is 64-bit, the array length is 1, with mask[0] in the range (0, 2³² – 1]. For example, if mask = [0, 8] and 8 = 0b1000, only the fourth element participates in computation. Contiguous mode: indicates the number of contiguous elements that participate in computation. The value range is related to the operand data type. The maximum number of elements that can be processed in each iteration varies according to the data type. When the operand is 16-bit, mask ∈ [1, 128]. When the operand is 32-bit, mask ∈ [1, 64]. When the operand is 64-bit, mask ∈ [1, 32]. When the number of bits of the source operand is different from that of the destination operand, the data type with more bytes is used for the computation.
repeatTime	Input	Number of iteration repeats. The Vector Unit reads 256 bytes of contiguous data for computation each time. To read the complete data for processing, the unit needs to read the input data in multiple repeats. repeatTime indicates the number of iterations. For details about this parameter, see High-dimensional Sharding APIs.
repeatParams	Input	Parameters that control the operand address strides. They are of the BinaryRepeatParams type, and contain such parameters as those that specify the address stride of the operand for the same data block between adjacent iterations and address stride of the operand between different data blocks in a single iteration. For details about the address stride of the operand between adjacent iterations, see repeatStride. For details about the address stride of the operand between different data blocks in a single iteration, see dataBlockStride.

Returns

None

Restrictions

For details about the operand address alignment requirements, see General Address Alignment Restrictions.
The destination operand address cannot overlap with the source operand address.

Examples

In the examples, srcLocal is of int32_t type and dstLocal is of half type. mask is calculated based on the int32_t type.

Example of high-dimensional tensor sharding computation (contiguous mask mode)

        
             uint64_t mask = 256 / sizeof(int32_t); // 64
// repeatTime = 4. 64 elements are computed in one iteration, and 256 elements are computed in total.
// dstBlkStride, src0BlkStride, src1BlkStride = 1. Data is continuously read and written in a single iteration.
// dstRepStride = 4, src0RepStride, src1RepStride = 8. Data is continuously read and written between adjacent iterations.
half scale = 0.1;
AscendC::SetDeqScale(scale);
AscendC::AddDeqRelu(dstLocal, src0Local, src1Local, mask, 4, { 1, 1, 1, 4, 8, 8 });

Example of high-dimensional tensor sharding computation (bitwise mask mode)

        
             uint64_t mask[2] = { UINT64_MAX, UINT64_MAX };
// repeatTime = 4. 64 elements are computed in one iteration, and 256 elements are computed in total.
// dstBlkStride, src0BlkStride, src1BlkStride = 1. Data is continuously read and written in a single iteration.
// dstRepStride = 4, src0RepStride, src1RepStride = 8. Data is continuously read and written between adjacent iterations.
half scale = 0.1;
AscendC::SetDeqScale(scale);
AscendC::AddDeqRelu(dstLocal, src0Local, src1Local, mask, 4, { 1, 1, 1, 4, 8, 8 });

Example of computing the first n pieces of data of a tensor

        
             half scale = 0.1;
AscendC::SetDeqScale(scale);
AscendC::AddDeqRelu(dstLocal, src0Local, src1Local, 512);

Result example:

Input (src0Local): [70 36 43 54 28 49 27 82 95 ...]
Input (src1Local): [19 33 34 50 42  2 97 93 99 ...]
Output (dstLocal): [8.9 6.9 7.7 10.4 7.0 5.1 12.4 17.5 19.4 ...]

Parent topic: Compound Computation