Truncate

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

Truncates the float elements of the source operand to their integer parts, while retaining the data type of the source operand.

Prototype

template <typename T = DefaultType, RoundMode roundMode = RoundMode::CAST_NONE, MaskMergeMode mode = MaskMergeMode::ZEROING, typename S>
__simd_callee__ inline void Truncate(S& dstReg, S& srcReg, MaskReg& mask)

Parameters

**Table 1** Template parameters
Parameter	Description
T	Operand data type. For the Atlas 350 Accelerator Card, the supported data types of dstReg/srcReg are half, float, and bfloat16_t.
roundMode	Rounding mode. The options are as follows: RoundMode::CAST_NONE: default value. In scenarios where precision loss occurs during conversion, this value is equivalent to CAST_RINT. In other scenarios, this value has no effect. RoundMode::CAST_RINT: returns the nearest integer. If two integers are equally close, the even integer is returned. RoundMode::CAST_ROUND: round mode, rounding off RoundMode::CAST_FLOOR: floor mode, rounding down RoundMode::CAST_CEIL: ceil mode, rounding up RoundMode::CAST_TRUNC: truncation mode, rounding toward zero and discarding the fractional part
mode	Set it to MERGING or ZEROING. ZEROING: The elements that are not filtered by mask are set to zero in dst. MERGING: not supported currently.
S	srcReg or dstReg type, for example, RegTensor<float>. It is automatically inferred by the compiler and does not need to be specified.

**Table 2** Parameters
Parameter	Input/Output	Description
dstReg	Output	Destination operand. The type is RegTensor.
srcReg	Input	Source operand. The type is RegTensor. The source operand must have the same data type as the destination operand.
mask	Input	Valid indication of the source operand element operation. For details, see MaskReg.

Returns

None

Constraints

In non-saturation mode, if the input is NaN, the output is NaN. If the input value exceeds the value range of the input type, the output is +/-inf.
In saturation mode, if the input is NaN, the output is 0. If the input value exceeds the value range of the input type, the output is saturated to the maximum or minimum value.
The float type supports only the non-saturation mode.

Example

template<typename T>
__simd_vf__ inline void TruncVF(__ubuf__ T* dstAddr, __ubuf__ T* srcAddr, uint32_t count, uint32_t oneRepeatSize, uint16_t repeatTimes)
{
    AscendC::Reg::RegTensor<T> srcReg;
    AscendC::Reg::RegTensor<T> dstReg;
    AscendC::Reg::MaskReg mask;
    for (uint16_t i = 0; i < repeatTimes; i++) {
        AscendC::Reg::LoadAlign(srcReg, srcAddr + i * oneRepeatSize);
        mask = AscendC::Reg::UpdateMask<T>(count);
        AscendC::Reg::Truncate<T, AscendC::RoundMode::CAST_FLOOR, AscendC::Reg::MaskMergeMode::ZEROING>(dstReg, srcReg, mask);
        AscendC::Reg::StoreAlign(dstAddr + i * oneRepeatSize, dstReg, mask);
    }
}

Parent topic: Type Conversion