libflame  revision_anchor
Functions
FLA_Gemm_hn_blk_var1.c File Reference

(r)

Functions

FLA_Error FLA_Gemm_hn_blk_var1 (FLA_Obj alpha, FLA_Obj A, FLA_Obj B, FLA_Obj beta, FLA_Obj C, fla_gemm_t *cntl)
 

Function Documentation

◆ FLA_Gemm_hn_blk_var1()

FLA_Error FLA_Gemm_hn_blk_var1 ( FLA_Obj  alpha,
FLA_Obj  A,
FLA_Obj  B,
FLA_Obj  beta,
FLA_Obj  C,
fla_gemm_t cntl 
)

References FLA_Cont_with_1x3_to_1x2(), FLA_Cont_with_3x1_to_2x1(), FLA_Determine_blocksize(), FLA_Gemm_internal(), FLA_Obj_width(), FLA_Part_1x2(), FLA_Part_2x1(), FLA_Repart_1x2_to_1x3(), and FLA_Repart_2x1_to_3x1().

Referenced by FLA_Gemm_hn().

14 {
15  FLA_Obj AL, AR, A0, A1, A2;
16 
17  FLA_Obj CT, C0,
18  CB, C1,
19  C2;
20 
21  dim_t b;
22 
23  FLA_Part_1x2( A, &AL, &AR, 0, FLA_LEFT );
24 
25  FLA_Part_2x1( C, &CT,
26  &CB, 0, FLA_TOP );
27 
28  while ( FLA_Obj_width( AL ) < FLA_Obj_width( A ) ){
29 
30  b = FLA_Determine_blocksize( AR, FLA_RIGHT, FLA_Cntl_blocksize( cntl ) );
31 
32  FLA_Repart_1x2_to_1x3( AL, /**/ AR, &A0, /**/ &A1, &A2,
33  b, FLA_RIGHT );
34 
35  FLA_Repart_2x1_to_3x1( CT, &C0,
36  /* ** */ /* ** */
37  &C1,
38  CB, &C2, b, FLA_BOTTOM );
39 
40  /*------------------------------------------------------------*/
41 
42  /* C1 = alpha * A1' * B + C1; */
43  FLA_Gemm_internal( FLA_CONJ_TRANSPOSE, FLA_NO_TRANSPOSE,
44  alpha, A1, B, beta, C1,
45  FLA_Cntl_sub_gemm( cntl ) );
46 
47  /*------------------------------------------------------------*/
48 
49  FLA_Cont_with_1x3_to_1x2( &AL, /**/ &AR, A0, A1, /**/ A2,
50  FLA_LEFT );
51 
52  FLA_Cont_with_3x1_to_2x1( &CT, C0,
53  C1,
54  /* ** */ /* ** */
55  &CB, C2, FLA_TOP );
56 
57  }
58 
59  return FLA_SUCCESS;
60 }
FLA_Error FLA_Repart_2x1_to_3x1(FLA_Obj AT, FLA_Obj *A0, FLA_Obj *A1, FLA_Obj AB, FLA_Obj *A2, dim_t mb, FLA_Side side)
Definition: FLA_View.c:226
FLA_Error FLA_Repart_1x2_to_1x3(FLA_Obj AL, FLA_Obj AR, FLA_Obj *A0, FLA_Obj *A1, FLA_Obj *A2, dim_t nb, FLA_Side side)
Definition: FLA_View.c:267
FLA_Error FLA_Gemm_internal(FLA_Trans transa, FLA_Trans transb, FLA_Obj alpha, FLA_Obj A, FLA_Obj B, FLA_Obj beta, FLA_Obj C, fla_gemm_t *cntl)
Definition: FLA_Gemm_internal.c:16
unsigned long dim_t
Definition: FLA_type_defs.h:71
dim_t FLA_Determine_blocksize(FLA_Obj A_unproc, FLA_Quadrant to_dir, fla_blocksize_t *cntl_blocksizes)
Definition: FLA_Blocksize.c:234
FLA_Error FLA_Cont_with_3x1_to_2x1(FLA_Obj *AT, FLA_Obj A0, FLA_Obj A1, FLA_Obj *AB, FLA_Obj A2, FLA_Side side)
Definition: FLA_View.c:428
Definition: FLA_type_defs.h:158
dim_t FLA_Obj_width(FLA_Obj obj)
Definition: FLA_Query.c:123
FLA_Error FLA_Cont_with_1x3_to_1x2(FLA_Obj *AL, FLA_Obj *AR, FLA_Obj A0, FLA_Obj A1, FLA_Obj A2, FLA_Side side)
Definition: FLA_View.c:475
FLA_Error FLA_Part_2x1(FLA_Obj A, FLA_Obj *A1, FLA_Obj *A2, dim_t mb, FLA_Side side)
Definition: FLA_View.c:76
FLA_Error FLA_Part_1x2(FLA_Obj A, FLA_Obj *A1, FLA_Obj *A2, dim_t nb, FLA_Side side)
Definition: FLA_View.c:110