Elementwise
Add, sub, mul, square difference, min/max, batch norm, select, and arithmetic glue. 65 functions, declared in 3 modules.
| Function | Summary | Module |
|---|---|---|
arm_abs_s16 |
s16 elementwise absolute value | arm_nnfunctions.h |
arm_abs_s8 |
s8 elementwise absolute value | arm_nnfunctions.h |
arm_add_s16 |
s16 elementwise add of two tensors with support for broadcasting. | arm_nnfunctions.h |
arm_add_s8 |
s8 elementwise add of two tensors with support for broadcasting. | arm_nnfunctions.h |
arm_add_scalar_s16 |
s16 elementwise add of scalar and vector | arm_nnfunctions.h |
arm_add_scalar_s8 |
s8 elementwise add of scalar and vector | arm_nnfunctions.h |
arm_batch_norm_f16 |
Apply batch normalization. | NNSupport |
arm_batch_norm_f32 |
Apply batch normalization. | NNSupport |
arm_elementwise_add_broadcast_f16 |
Elementwise add with TensorFlow Lite NHWC broadcasting and an output clamp. | Elementwise Functions |
arm_elementwise_add_broadcast_f32 |
Elementwise add with TensorFlow Lite NHWC broadcasting and an output clamp. | Elementwise Functions |
arm_elementwise_add_f16 |
Elementwise add with optional output clamp. | Elementwise Functions |
arm_elementwise_add_f32 |
Elementwise add with optional output clamp. | Elementwise Functions |
arm_elementwise_add_fp16 |
Legacy float16 elementwise add with fused clamp, kept only for source compatibility with callers that predate armelementwiseaddf16(). | Elementwise Functions |
arm_elementwise_add_s16 |
s16 elementwise add of two vectors | arm_nnfunctions.h |
arm_elementwise_add_s8 |
s8 elementwise add of two vectors | arm_nnfunctions.h |
arm_elementwise_mul_broadcast_f16 |
Elementwise multiply with TensorFlow Lite NHWC broadcasting and an output clamp. | Elementwise Functions |
arm_elementwise_mul_broadcast_f32 |
Elementwise multiply with TensorFlow Lite NHWC broadcasting and an output clamp. | Elementwise Functions |
arm_elementwise_mul_f16 |
Elementwise multiply with optional output clamp. | Elementwise Functions |
arm_elementwise_mul_f32 |
Elementwise multiply with optional output clamp. | Elementwise Functions |
arm_elementwise_mul_s16 |
s16 elementwise multiplication | arm_nnfunctions.h |
arm_elementwise_mul_s8 |
s8 elementwise multiplication | arm_nnfunctions.h |
arm_elementwise_prelu_s16 |
Elementwise S16 PReLU activation function. | arm_nnfunctions.h |
arm_elementwise_prelu_s8 |
Elementwise S8 PReLU activation function. | arm_nnfunctions.h |
arm_elementwise_squared_difference_f16 |
Elementwise squared difference of two float16 vectors. | Elementwise Functions |
arm_elementwise_squared_difference_s16 |
s16 elementwise squared difference of two vectors. | arm_nnfunctions.h |
arm_elementwise_squared_difference_s8 |
s8 elementwise squared difference of two vectors. | arm_nnfunctions.h |
arm_elementwise_sub_broadcast_f16 |
Elementwise subtract with TensorFlow Lite NHWC broadcasting and an output clamp. | Elementwise Functions |
arm_elementwise_sub_broadcast_f32 |
Elementwise subtract with TensorFlow Lite NHWC broadcasting and an output clamp. | Elementwise Functions |
arm_elementwise_sub_f16 |
Elementwise subtract with optional output clamp. | Elementwise Functions |
arm_elementwise_sub_f32 |
Elementwise subtract with optional output clamp. | Elementwise Functions |
arm_elementwise_sub_s16 |
s16 elementwise subtract of two vectors | arm_nnfunctions.h |
arm_elementwise_sub_s8 |
s8 elementwise subtract of two vectors | arm_nnfunctions.h |
arm_maximum_f16 |
Elementwise maximum with TensorFlow Lite NHWC broadcasting: each dimension of the two inputs must be equal or 1, and outputdims must be their broadcast shape. | Elementwise Functions |
arm_maximum_f32 |
Elementwise maximum with TensorFlow Lite NHWC broadcasting: each dimension of the two inputs must be equal or 1, and outputdims must be their broadcast shape. | Elementwise Functions |
arm_maximum_s16 |
s16 elementwise maximum w/ support for broadcasting and scalar inputs. | arm_nnfunctions.h |
arm_maximum_s8 |
s8 elementwise maximum w/ support for broadcasting and scalar inputs. | arm_nnfunctions.h |
arm_minimum_f16 |
Elementwise minimum with TensorFlow Lite NHWC broadcasting: each dimension of the two inputs must be equal or 1, and outputdims must be their broadcast shape. | Elementwise Functions |
arm_minimum_f32 |
Elementwise minimum with TensorFlow Lite NHWC broadcasting: each dimension of the two inputs must be equal or 1, and outputdims must be their broadcast shape. | Elementwise Functions |
arm_minimum_s16 |
s16 elementwise minimum w/ support for broadcasting and scalar inputs. | arm_nnfunctions.h |
arm_minimum_s8 |
s8 elementwise minimum w/ support for broadcasting and scalar inputs. | arm_nnfunctions.h |
arm_mul_s16 |
s16 elementwise multiplication of two tensors with support for broadcasting. | arm_nnfunctions.h |
arm_mul_s8 |
s8 elementwise multiplication of two tensors with support for broadcasting. | arm_nnfunctions.h |
arm_mul_scalar_s16 |
s16 elementwise multiplication of scalar and vector | arm_nnfunctions.h |
arm_mul_scalar_s8 |
s8 elementwise multiplication of scalar and vector | arm_nnfunctions.h |
arm_nn_abs_f16 |
Elementwise absolute value. | Elementwise Functions |
arm_nn_abs_f32 |
Elementwise absolute value. | Elementwise Functions |
arm_nn_sqrt_f16 |
Elementwise square root of a float16 tensor. | Elementwise Functions |
arm_nn_sqrt_f32 |
Elementwise square root. | Elementwise Functions |
arm_rsqrt_f16 |
Elementwise reciprocal square root of a float16 tensor, 1 / sqrt(x). | Elementwise Functions |
arm_rsqrt_f32 |
Elementwise reciprocal square root, 1 / sqrt(x). | Elementwise Functions |
arm_rsqrt_s16_per_op |
INT16 reciprocal square root using a per-operator LUT. | arm_nnfunctions.h |
arm_rsqrt_s16_universal |
INT16 reciprocal square root using a shared universal LUT. | arm_nnfunctions.h |
arm_select_v2_s16 |
SELECTV2 with broadcast for int16 tensors. | arm_nnfunctions.h |
arm_select_v2_s8 |
SELECTV2 with broadcast for int8 tensors. | arm_nnfunctions.h |
arm_sqrt_s16 |
s16 elementwise square root using piecewise LUT with linear interpolation | arm_nnfunctions.h |
arm_sqrt_s16_tablefree |
s16 elementwise square root without a lookup table | arm_nnfunctions.h |
arm_sqrt_s8 |
s8 elementwise square root | arm_nnfunctions.h |
arm_squared_difference_s16 |
s16 elementwise squared difference of two tensors with support for broadcasting. | arm_nnfunctions.h |
arm_squared_difference_s8 |
s8 elementwise squared difference of two tensors with support for broadcasting. | arm_nnfunctions.h |
arm_squared_difference_scalar_s16 |
s16 elementwise squared difference of scalar and vector. | arm_nnfunctions.h |
arm_squared_difference_scalar_s8 |
s8 elementwise squared difference of scalar and vector. | arm_nnfunctions.h |
arm_sub_s16 |
s16 elementwise subtraction of two tensors with support for broadcasting. | arm_nnfunctions.h |
arm_sub_s8 |
s8 elementwise subtraction of two tensors with support for broadcasting. | arm_nnfunctions.h |
arm_sub_scalar_s16 |
s16 elementwise subtract of scalar and vector (scalar - vector) | arm_nnfunctions.h |
arm_sub_scalar_s8 |
s8 elementwise subtract of scalar and vector (scalar - vector) | arm_nnfunctions.h |