更多样例

uint64_t mask = 64;  // 每个迭代内只计算前64个数
AscendC::Add(dstLocal, src0Local, src1Local, mask, 4, { 1, 1, 1, 8, 8, 8 });

uint64_t mask[2] = { UINT64_MAX, 0 };  // mask[0]满，mask[1]空，每次只计算前64个数
AscendC::Add(dstLocal, src0Local, src1Local, mask, 4, { 1, 1, 1, 8, 8, 8 });

uint64_t mask = 128;
// repeatTime设置为2，表示一共需要进行2次迭代
// src0BlkStride, src1BlkStride设置为1，表示每个迭代内src0参与计算的数据地址间隔为1个DataBlock
// src0RepStride设置为16, 表示相邻迭代之间src0起始地址间隔为16个datablock
AscendC::Add(dstLocal, src0Local, src1Local, mask, 2, { 1, 1, 1, 8, 16, 8 });

uint64_t mask = 128;
// repeatTime设置为2，表示一共需要进行2次迭代
// src0BlkStride设置为2，表示每个迭代内src0参与计算的数据地址间隔为2个datablock
// src0RepStride设置为16, 表示相邻迭代之间src0起始地址间隔为16个datablock
AscendC::Add(dstLocal, src0Local, src1Local, mask, 2, { 1, 2, 1, 8, 16, 8 });

#include "kernel_operator.h"
constexpr int32_t BUFFER_NUM = 2;
class KernelBinaryScalar {
public:
    __aicore__ inline KernelBinaryScalar() {}
    __aicore__ inline void Init(GM_ADDR x, GM_ADDR z, float scalar, uint32_t totalLength, uint32_t tileNum)
    {
        this->blockLength = totalLength / AscendC::GetBlockNum();
        this->scalar = scalar;
        this->tileNum = tileNum;
        ASSERT(tileNum != 0 && "tile num can not be zero!");
        this->tileLength = this->blockLength / tileNum / BUFFER_NUM;
        xGm.SetGlobalBuffer((__gm__ DTYPE_X*)x + this->blockLength * AscendC::GetBlockIdx(), this->blockLength);
        zGm.SetGlobalBuffer((__gm__ DTYPE_Z*)z + this->blockLength * AscendC::GetBlockIdx(), this->blockLength);
        pipe.InitBuffer(inQueueX, BUFFER_NUM, this->tileLength * sizeof(DTYPE_X));
        pipe.InitBuffer(outQueueZ, BUFFER_NUM, this->tileLength * sizeof(DTYPE_Z));
    }
    __aicore__ inline void Process()
    {
        int32_t loopCount = this->tileNum * BUFFER_NUM;
        for (int32_t i = 0; i < loopCount; i++) {
            CopyIn(i);
            Compute(i);
            CopyOut(i);
        }
    }
private:
    __aicore__ inline void CopyIn(int32_t progress)
    {
        AscendC::LocalTensor<DTYPE_X> xLocal = inQueueX.AllocTensor<DTYPE_X>();
        AscendC::DataCopy(xLocal, xGm[progress * this->tileLength], this->tileLength);
        inQueueX.EnQue(xLocal);
    }
    __aicore__ inline void Compute(int32_t progress)
    {
        AscendC::LocalTensor<DTYPE_X> xLocal = inQueueX.DeQue<DTYPE_X>();
        AscendC::LocalTensor<DTYPE_Z> zLocal = outQueueZ.AllocTensor<DTYPE_Z>();
        AscendC::Adds(zLocal, xLocal, (DTYPE_X)scalar, this->tileLength);
        outQueueZ.EnQue<DTYPE_Z>(zLocal);
        inQueueX.FreeTensor(xLocal);
    }
    __aicore__ inline void CopyOut(int32_t progress)
    {
        AscendC::LocalTensor<DTYPE_Z> zLocal = outQueueZ.DeQue<DTYPE_Z>();
        AscendC::DataCopy(zGm[progress * this->tileLength], zLocal, this->tileLength);
        outQueueZ.FreeTensor(zLocal);
    }
private:
    AscendC::TPipe pipe;
    AscendC::TQue<AscendC::TPosition::VECIN, BUFFER_NUM> inQueueX;
    AscendC::TQue<AscendC::TPosition::VECOUT, BUFFER_NUM> outQueueZ;
    AscendC::GlobalTensor<DTYPE_X> xGm;
    AscendC::GlobalTensor<DTYPE_Z> zGm;
    float scalar;
    uint32_t blockLength;
    uint32_t tileNum;
    uint32_t tileLength;
};
extern "C" __global__ __aicore__ void binary_scalar_simple_kernel(GM_ADDR x, GM_ADDR z, GM_ADDR workspace, GM_ADDR tiling)
{
    GET_TILING_DATA(tilingData, tiling);
    KernelBinaryScalar op;
    op.Init(x, z, tilingData.scalar, tilingData.totalLength, tilingData.tileNum);
    if (TILING_KEY_IS(1)) {
        op.Process();
    }
}