struct dnnl_reduction_desc_t


A descriptor of reduction operation. More…

#include <dnnl_types.h>

struct dnnl_reduction_desc_t
    // fields

    dnnl_primitive_kind_t primitive_kind;
    dnnl_alg_kind_t alg_kind;
    dnnl_memory_desc_t src_desc;
    dnnl_memory_desc_t dst_desc;
    float p;
    float eps;

Detailed Documentation

A descriptor of reduction operation.


dnnl_primitive_kind_t primitive_kind

The kind of primitive.

Used for self-identifying the primitive descriptor. Must be dnnl_reduction.

dnnl_alg_kind_t alg_kind

The kind of reduction algorithm.

Possible values: dnnl_reduction_max, dnnl_reduction_min, dnnl_reduction_sum, dnnl_reduction_mul, dnnl_reduction_mean, dnnl_reduction_norm_lp_max, dnnl_reduction_norm_lp_sum, dnnl_reduction_norm_lp_power_p_max, dnnl_reduction_norm_lp_power_p_sum.

dnnl_memory_desc_t src_desc

Source memory descriptor.

dnnl_memory_desc_t dst_desc

Destination memory descriptor.

float p

Algorithm specific parameters.

Accordance table: dnnl_reduction_max : p and eps are ignored dnnl_reduction_min : p and eps are ignored dnnl_reduction_norm_lp_max : p power, eps epsilon dnnl_reduction_norm_lp_sum : p power, eps epsilon dnnl_reduction_norm_lp_power_p_max : p power, eps epsilon dnnl_reduction_norm_lp_power_p_sum : p power, eps epsilon dnnl_reduction_sum : p and eps are ignored dnnl_reduction_mul : p and eps are ignored dnnl_reduction_mean : p and eps are ignored