HMLP: High-performance Machine Learning Primitives
•All Classes Namespaces Functions Variables Typedefs Enumerations Enumerator Pages
Public Member Functions | Public Attributes | Static Public Attributes | List of all members
gkmm_mrxnr< MR, NR, OPKERNEL, OP1, OP2, TA, TB, TC, TV > Struct Template Reference

This kernel takes opkernel, op1 and op2 to implement an MR-by-NR GKMM operation. More...

#include <fused_mrxnr.hpp>

Public Member Functions

void operator() (int k, TA *a, TB *b, TC *c, int ldc, TV *v, int ldv, aux_s< TA, TB, TC, TV > *aux) const
 
void operator() (int k, TA *a, TB *b, TV *v, int rs_c, int cs_c, aux_s< TA, TB, TC, TV > *aux) const
 

Public Attributes

OPKERNEL opkernel
 
OP1 op1
 
OP2 op2
 
TV initV
 

Static Public Attributes

static const size_t mr = MR
 
static const size_t nr = NR
 
static const size_t pack_mr = MR
 
static const size_t pack_nr = NR
 
static const size_t align_size = 32
 

Detailed Description

template<int MR, int NR, typename OPKERNEL, typename OP1, typename OP2, typename TA, typename TB, typename TC, typename TV>
struct gkmm_mrxnr< MR, NR, OPKERNEL, OP1, OP2, TA, TB, TC, TV >

This kernel takes opkernel, op1 and op2 to implement an MR-by-NR GKMM operation.

The documentation for this struct was generated from the following file: