AddrReg

功能说明

AddrReg即为Address Register（地址寄存器），是用于存储地址偏移量的寄存器。AddrReg应该通过CreateAddrReg API初始化，然后在循环之中使用AddrReg存储地址偏移量。AddrReg在每层循环中根据所设置的stride进行自增。

函数原型

// offset = index0 * stride0
template <typename T>
__simd_callee__ inline AddrReg CreateAddrReg(uint16_t index0, uint32_t stride0);
 
// offset = index0 * stride0 + index1 * stride1
template <typename T>
__simd_callee__ inline AddrReg CreateAddrReg(uint16_t index0, uint32_t stride0, uint16_t index1, uint32_t stride1);
 
// offset = index0 * stride0 + index1 * stride1 + index2 * stride2
template <typename T>
__simd_callee__ inline AddrReg CreateAddrReg(uint16_t index0, uint32_t stride0, uint16_t index1, uint32_t stride1, uint16_t index2, uint32_t stride2);

// offset = index0 * stride0 + index1 * stride1 + index2 * stride2 + index3 * stride3
template <typename T>
__simd_callee__ inline AddrReg CreateAddrReg(uint16_t index0, uint32_t stride0, uint16_t index1, uint32_t stride1, uint16_t index2, uint32_t stride2, uint16_t index3, uint32_t stride3);

参数说明

支持的型号

Atlas 350 加速卡

约束说明

无

调用示例

__simd_vf__ inline void CreateAddrRegVF(__ubuf__ T* dstAddr, __ubuf__ T* srcAddr, uint32_t oneRepeatSize, uint16_t repeatTimes)
{
    AscendC::Reg::MaskReg mask = AscendC::Reg::CreateMask<T>();
    AscendC::Reg::AddrReg aReg;
    for (uint16_t i = 0; i < repeatTimes; ++i) {
        aReg = AscendC::Reg::CreateAddrReg<T>(i, oneRepeatSize);
        AscendC::Reg::LoadAlign(mask, srcAddr, aReg);
        AscendC::Reg::StoreAlign(dstAddr, mask, aReg);
    }
}