CeilDivision

Applicability

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

Function Usage

Rounds up the result of dividing two integers, num1 and num2.

Prototype

__aicore__ constexpr inline int32_t CeilDivision(int32_t num1, int32_t num2)

Parameters

**Table 1** Template parameters
Parameter	Description
num1	Dividend.
num2	Divisor.

Returns

Rounded-up result of dividing two integers.

Restrictions

If num2 is 0, the result is 0.
This API can only be used when num1 and num2 are positive numbers.

Example

In this example, CeilDivision is used to calculate the number of iterations (repeatTimes). By performing ceiling division on the data size (count) and the amount of data processed at a time, it ensures that all data (including the tail block) is fully processed.

template <typename T>
__aicore__ inline void AddCustomImpl(__local_mem__ T *dst, __local_mem__ T *src0, __local_mem__ T *src1,
    uint32_t count)
{
    AscendC::Reg::RegTensor<T> srcReg0;
    AscendC::Reg::RegTensor<T> srcReg1;
    AscendC::Reg::RegTensor<T> dstReg;
    AscendC::Reg::MaskReg mask;
    constexpr uint32_t oneRepeatSize = AscendC::GetVecLen() / sizeof(T);
    uint16_t repeatTime = AscendC::CeilDivision(count, oneRepeatSize);
    for (uint16_t i = 0; i < repeatTime; ++i) {
        mask = AscendC::Reg::UpdateMask<T>(calCount);
        AscendC::Reg::LoadAlign(srcReg0, src0 + i * oneRepeatSize );
        AscendC::Reg::LoadAlign(srcReg1, src1 + i * oneRepeatSize );
        AscendC::Reg::Add(dstReg, srcReg0, srcReg1, mask);
        AscendC::Reg::StoreAlign(dst + i * repeatElm, reg2, mask);
    }
}

Parent topic: Tool Functions