Clamp

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

The output is the input values clamped to the range [min, max], where values greater than max are set to max, values less than min are set to min, and all other values (including NaN) remain unchanged. If min is greater than max, all values (except NaN) are set to max. min and max can be a Scalar or LocalTensor value.

$\text{[math]}$

Prototype

template <const ClampConfig& config = DEFAULT_CLAMP_CONFIG, typename T, typename U, typename S>
__aicore__ inline void Clamp(const LocalTensor<T>& dst, const LocalTensor<T>& src, const U& min, const S& max, const uint32_t count)

Parameters

**Table 1** Template parameters
Parameter	Description
config	Clamp algorithm configuration. This is an optional parameter of the ClampConfig type. The code below describes the definition. isReuseSource: This parameter is reserved. Pass the default value false.
T	Operand data type. For the Atlas 350 Accelerator Card, the supported data types are int8_t, uint8_t, int16_t, uint16_t, half, bfloat16_t, int32_t, uint32_t, float, int64_t, and uint64_t.
U	LocalTensor or Scalar type. The type is automatically inferred based on the input parameter min. You do not need to set this parameter, but ensure that min meets the data type constraints. For the Atlas 350 Accelerator Card, the supported data types are int8_t, uint8_t, int16_t, uint16_t, half, bfloat16_t, int32_t, uint32_t, float, int64_t, and uint64_t.
S	LocalTensor or Scalar type. The type is automatically inferred based on the input parameter max. You do not need to set this parameter, but ensure that max meets the data type constraints. For the Atlas 350 Accelerator Card, the supported data types are int8_t, uint8_t, int16_t, uint16_t, half, bfloat16_t, int32_t, uint32_t, float, int64_t, and uint64_t.

struct ClampConfig {
    bool isReuseSource;
};

**Table 2** API parameters
Parameter	Input/Output	Description
dst	Output	Destination operand. The type is LocalTensor, and TPosition can be VECIN, VECCALC, or VECOUT.
src	Input	Source operand. The type is LocalTensor, and TPosition can be VECIN, VECCALC, or VECOUT. The source operand must have the same data type as the destination operand.
min	Input	Data lower limit. The type can be Scalar or LocalTensor. When the type is LocalTensor, the supported TPosition is VECIN, VECCALC, or VECOUT. The source and destination operands must have the same data type.
max	Input	Data upper limit. The type can be Scalar or LocalTensor. When the type is LocalTensor, the supported TPosition is VECIN, VECCALC, or VECOUT. The source and destination operands must have the same data type.
count	Input	Number of elements involved in the computation.

Returns

None

Constraints

The source operand address must not overlap the destination operand address.
For details about the operand address alignment requirements, see General Address Alignment Restrictions.

Examples

AscendC::LocalTensor<half> dst, src;
uint32_t count = 512;
half min = 30;
half max = 60;
AscendC::Clamp(dst, src, min, max, count);

Result example:

Input (src):
[13, 78, 35, 95, 83,  2,  2, 95, 51, 73, 98,  3, 55, 32, 61,  2, 40, 26, 95, ... 63]
Input (min):
[30]
Input (max):
[60]
Output (dst):
[30, 60, 35, 60, 60, 30, 30, 60, 51, 60, 60, 30, 55, 32, 60, 30, 40, 30, 60, ... 60]

Parent topic: Clamp APIs