PairReduceSum

Applicability

Product	Supported/Unsupported
Atlas A3 training products / Atlas A3 inference products	√
Atlas A2 training products / Atlas A2 inference products	√
Atlas 200I/500 A2 inference products	√
Atlas inference product 's AI Core	√
Atlas inference product 's Vector Core	x
Atlas training products	√

Functions

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 A3 training products / Atlas A3 inference products , the supported data types are half and float. For the Atlas A2 training products / Atlas A2 inference products , the supported data types are half and float. For the Atlas 200I/500 A2 inference products , the supported data types are half and float. For the Atlas inference product 's AI Core, the supported data types are half and float. For the Atlas training products , 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	Description
dst	Output	Destination operand. The type is LocalTensor, and the supported TPosition is VECIN, VECCALC, or VECOUT. The start address of the LocalTensor must be 32-byte aligned.
src	Input	Source operand. The type is LocalTensor, and the supported TPosition is VECIN, VECCALC, or VECOUT. The start address of the 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	The mask parameter is used to control the elements involved 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 in the computation. The mask is in array form. The array length and the value range of the array elements are related to the data type of the operand. When the operand is 16-bit, the array length is 2. In this case, mask[0] and mask[1] must be in the range of [0, 2⁶⁴ – 1] and cannot be 0 at the same time. When the operand is 32-bit, the array length is 1. In this case, mask[0] must be in the range of (0, 2⁶⁴ – 1]. When the operand is 64-bit, the array length is 1. In this case, mask[0] must be in the range of (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 repeat 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 products .
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 of 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 products , 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 in the float scenario and mask is set to 62, the last two elements are not involved in the operation. For the Atlas 200I/500 A2 Inference Product , the last value of the destination operand is 0. For other product models, the last value of the destination operand does not change.

Examples

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

        
             int32_t mask = 256/sizeof(half);
int repeat = 1;
// repeat = 1, 128 elements one repeat, 128 elements total
// srcBlkStride = 1, no gap between blocks in one repeat
// dstRepStride = 1, srcRepStride = 8, no gap between repeats
AscendC::PairReduceSum<half>(dstLocal, srcLocal, repeat, mask, 1, 1, 8);

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

        
             uint64_t mask[2] = { UINT64_MAX, UINT64_MAX };
int repeat = 1;
// repeat = 1, 128 elements one repeat, 128 elements total
// srcBlkStride = 1, no gap between blocks in one repeat
// dstRepStride = 1, srcRepStride = 8, no gap between repeats
AscendC::PairReduceSum<half>(dstLocal, srcLocal, repeat, mask, 1, 1, 8);

Result example:

Input (src_gm):
[-3.441, 7.246, -0.02759, -6.324, 3.693, -7.984, -4.246, 6.332, -3.734, -2.699, -6.91, 7.887, -3.631, 5.219, 6.539, 8.688, 6.523, -6.789, -8.547, 4.258, 1.344, -8.469, -0.9253, -3.914, 3.293, -9.828, 7.082, 5.961, 2.133, 1.959, 3.928, -1.062, 9.18, -1.725, -3.645, 1.457, -2.328, -0.9487, -0.2849, -2.998, -9.281, 3.137, 0.4028, 5.961, -6.25, 2.406, -6.203, -2.699, 4.914, 1.653, -6.383, 6.855, 9.164, 0.6646, -2.854, 3.18, -0.5884, 0.4258, -5.773, -2.152, 4.258, 4.129, -8.719, -8.828, 6.145, 7.387, 1.386, -4.684, 6.324, -1.275, -1.816, 3.357, 6.832, -1.059, -9.852, -8.539, 2.938, -2.002, 9.625, -4.387, -1.309, 8.289, 2.906, -1.035, 7.723, 4.727, -6.477, 2.389, 6.75, -6.688, -0.04248, -6.613, -3.424, 7.145, 4.836, -5.617, -5.855, -5.234, -9.422, -9.852, -8.531, 2.115, 5.109, -8.094, -6.238, 9.898, -6.848, -6.051, 7.109, 4.227, -0.6187, -3.492, -4.352, 1.344, 1.526, 2.572, 2.16, -1.135, 9.812, 1.426, -8, 3.291, -2.039, 5.93, -5.52, -5.156, -9.422, 0.2236]  
Output (dst_gm):
[3.805, -6.352, -4.289, 2.086, -6.434, 0.9766, 1.588, 15.23, -0.2656, -4.289, -7.125, -4.84, -6.535, 13.05, 4.094, 2.865, 7.453, -2.188, -3.277, -3.283, -6.145, 6.363, -3.844, -8.906, 6.566, 0.4727, 9.828, 0.3262, -0.1626, -7.926, 8.391, -17.55, 13.53, -3.297, 5.047, 1.541, 5.773, -18.39, 0.9355, 5.238, 6.98, 1.871, 12.45, -4.086, 0.0625, -6.656, 3.721, -0.7812, -11.09, -19.28, -6.414, -2.984, 3.66, -12.9, 11.34, -4.109, -3.008, 4.098, 1.025, 11.23, -4.711, 3.891, -10.67, -9.195]

Complete code example:

        
         
           
           
             #include "kernel_operator.h"
class KernelReduce {
public:
    __aicore__ inline KernelReduce() {}
    __aicore__ inline void Init(__gm__ uint8_t* src, __gm__ uint8_t* dstGm)
    {
        srcGlobal.SetGlobalBuffer((__gm__ half*)src);
        dstGlobal.SetGlobalBuffer((__gm__ half*)dstGm);
        pipe.InitBuffer(inQueueSrc, 1, srcDataSize * sizeof(half));
        pipe.InitBuffer(outQueueDst, 1, dstDataSize * sizeof(half));
    }
    __aicore__ inline void Process()
    {
        CopyIn();
        Compute();
        CopyOut();
    }
private:
    __aicore__ inline void CopyIn()
    {
        AscendC::LocalTensor<half> srcLocal = inQueueSrc.AllocTensor<half>();
        AscendC::DataCopy(srcLocal, srcGlobal, srcDataSize);
        inQueueSrc.EnQue(srcLocal);
    }
    __aicore__ inline void Compute()
    {
        AscendC::LocalTensor<half> srcLocal = inQueueSrc.DeQue<half>();
        AscendC::LocalTensor<half> dstLocal = outQueueDst.AllocTensor<half>();
        half zero(0);
        AscendC::Duplicate(dstLocal, zero, dstDataSize);
        // Command execution part (replace with the preceding code)
        outQueueDst.EnQue<half>(dstLocal);
        inQueueSrc.FreeTensor(srcLocal);
    }
    __aicore__ inline void CopyOut()
    {
        AscendC::LocalTensor<half> dstLocal = outQueueDst.DeQue<half>();
        AscendC::DataCopy(dstGlobal, dstLocal, dstDataSize);
        outQueueDst.FreeTensor(dstLocal);
    }
private:
    AscendC::TPipe pipe;
    AscendC::TQue<AscendC::TPosition::VECIN, 1> inQueueSrc;
    AscendC::TQue<AscendC::TPosition::VECOUT, 1> outQueueDst;
    AscendC::GlobalTensor<half> srcGlobal, dstGlobal;
    int srcDataSize = 128;
    int dstDataSize = 64;
};
extern "C" __global__ __aicore__ void reduce_simple_kernel(__gm__ uint8_t* src, __gm__ uint8_t* dstGm)
{
    KernelReduce op;
    op.Init(src, dstGm);
    op.Process();
}

            

          

        
       

Parent topic: Reduction computation