PairReduceSum

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	√
Atlas inference product AI Core	√
Atlas inference product Vector Core	x
Atlas training product	√

Function Usage

PairReduceSum: sums up two adjacent (odd and even) elements, for example, (a1, a2, a3, a4, a5, a6, ...). The result is (a1+a2, a3+a4, a5+a6, ...). For details about reduction instructions, see How to Use Reduction Compute APIs.

Prototype

Bitwise mask mode

        
             template <typename T, bool isSetMask = true>
__aicore__ inline void PairReduceSum(const LocalTensor<T>& dst, const LocalTensor<T>& src, const int32_t repeatTime, const uint64_t mask[], const int32_t dstRepStride, const int32_t srcBlkStride, const int32_t srcRepStride)

Contiguous mask mode

        
             template <typename T, bool isSetMask = true>
__aicore__ inline void PairReduceSum(const LocalTensor<T>& dst, const LocalTensor<T>& src, const int32_t repeatTime, const int32_t mask, const int32_t dstRepStride, const int32_t srcBlkStride, const int32_t srcRepStride)

Parameters

**Table 1** Parameters in the template
Parameter	Description
T	Operand data type. For the Atlas 350 Accelerator Card, the supported data types are half and float. For the Atlas A3 training product / Atlas A3 inference product , the supported data types are half and float. For the Atlas A2 training product / Atlas A2 inference product , the supported data types are half and float. For the Atlas 200I/500 A2 inference product , the supported data types are half and float. For the Atlas inference product AI Core, the supported data types are half and float. For the Atlas training product , the supported data type is half.
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.

**Table 2** Parameters
Parameter	Input/Output	Meaning
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.
src	Input	Source operand. The type is LocalTensor, and TPosition can be VECIN, VECCALC, or VECOUT. The start address of LocalTensor must be 32-byte aligned.
repeatTime	Input	Number of iteration repeats. The value range is [0, 255]. For details about this parameter, see High-dimensional Sharding APIs.
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].
dstRepStride	Input	Address stride between adjacent iterations of the destination operand. The unit is the length after reduction of a repeat. After PairReduce is complete, the length of a repeat is halved. That is, the unit is 128 bytes. Note that this parameter cannot be set to 0 for the Atlas training product .
srcBlkStride	Input	Address stride of data blocks in a single iteration. For details, see dataBlockStride.
srcRepStride	Input	Address stride between adjacent iterations of the source operand, that is, the number of data blocks skipped from the source operand in each iteration. For details, see repeatStride.

Returns

None

Restrictions

For details about the operand address alignment requirements, see General Address Alignment Restrictions.
If the mask bits of every two elements are not configured (that is, the current two elements do not participate in the operation), the value in the corresponding destination operand is set to 0 for Atlas 200I/500 A2 inference product , and the value in the corresponding destination operand does not change for other product models. For example, if the current instruction is used for 64 elements of the float type and mask is set to 62, the last two elements are not involved in the computation. For the Atlas 200I/500 A2 inference product , the last value in the destination operand is 0. For other product models, the last value in the destination operand remains unchanged.

Examples

These examples show only part of the code used in the computation.

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

        
             // Set mask to 128, which indicates that a maximum of 128 elements are involved for computation.
int32_t mask = 256/sizeof(half);
// Each repeat contains 128 elements, and there are 128 elements in total.
int repeat = 1;
// dstLocal: destination operand tensor
// srcLocal: source operand tensor
// srcBlkStride = 1. Within a repeat, there are no gaps between blocks.
// dstRepStride = 1, srcRepStride = 8. There are no gaps between repeats.
AscendC::PairReduceSum<half>(dstLocal, srcLocal, repeat, mask, 1, 1, 8);

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

        
             // Set mask to 128, which indicates that a maximum of 128 elements are involved for computation.
uint64_t mask[2] = { UINT64_MAX, UINT64_MAX };
// Each repeat contains 128 elements, and there are 128 elements in total.
int repeat = 1;
// dstLocal: destination operand tensor
// srcLocal: source operand tensor
// srcBlkStride = 1. Within a repeat, there are no gaps between blocks.
// dstRepStride = 1, srcRepStride = 8. There are no gaps between repeats.
AscendC::PairReduceSum<half>(dstLocal, srcLocal, repeat, mask, 1, 1, 8);

Result example:

Input (src_gm):
[1, 1, 1, -1, 2, 2, -1, 2, 
 3, 3, 3, -1, 4, 4, -2, 4,
....
]
Output (dst_gm):
[2, 0, 4, 1, 6, 2, 8, 2, 
....
]

Parent topic: Reduction Computation