libflame  revision_anchor
Functions
FLA_Gemm_tn_blk_var2.c File Reference

(r)

Functions

FLA_Error FLA_Gemm_tn_blk_var2 (FLA_Obj alpha, FLA_Obj A, FLA_Obj B, FLA_Obj beta, FLA_Obj C, fla_gemm_t *cntl)
 

Function Documentation

◆ FLA_Gemm_tn_blk_var2()

FLA_Error FLA_Gemm_tn_blk_var2 ( FLA_Obj  alpha,
FLA_Obj  A,
FLA_Obj  B,
FLA_Obj  beta,
FLA_Obj  C,
fla_gemm_t cntl 
)
16 {
17  FLA_Obj AL, AR, A0, A1, A2;
18 
19  FLA_Obj CT, C0,
20  CB, C1,
21  C2;
22 
23  dim_t b;
24 
25  FLA_Part_1x2( A, &AL, &AR, 0, FLA_RIGHT );
26 
27  FLA_Part_2x1( C, &CT,
28  &CB, 0, FLA_BOTTOM );
29 
30  while ( FLA_Obj_width( AR ) < FLA_Obj_width( A ) ){
31 
32  b = FLA_Determine_blocksize( AL, FLA_LEFT, FLA_Cntl_blocksize( cntl ) );
33 
34  FLA_Repart_1x2_to_1x3( AL, /**/ AR, &A0, &A1, /**/ &A2,
35  b, FLA_LEFT );
36 
37  FLA_Repart_2x1_to_3x1( CT, &C0,
38  &C1,
39  /* ** */ /* ** */
40  CB, &C2, b, FLA_TOP );
41 
42  /*------------------------------------------------------------*/
43 
44  /* C1 = alpha * A1' * B + C1; */
45  FLA_Gemm_internal( FLA_TRANSPOSE, FLA_NO_TRANSPOSE,
46  alpha, A1, B, beta, C1,
47  FLA_Cntl_sub_gemm( cntl ) );
48 
49  /*------------------------------------------------------------*/
50 
51  FLA_Cont_with_1x3_to_1x2( &AL, /**/ &AR, A0, /**/ A1, A2,
52  FLA_RIGHT );
53 
54  FLA_Cont_with_3x1_to_2x1( &CT, C0,
55  /* ** */ /* ** */
56  C1,
57  &CB, C2, FLA_BOTTOM );
58 
59  }
60 
61  return FLA_SUCCESS;
62 }
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
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
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_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
dim_t FLA_Obj_width(FLA_Obj obj)
Definition: FLA_Query.c:123
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
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_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
dim_t FLA_Determine_blocksize(FLA_Obj A_unproc, FLA_Quadrant to_dir, fla_blocksize_t *cntl_blocksizes)
Definition: FLA_Blocksize.c:234
unsigned long dim_t
Definition: FLA_type_defs.h:71
Definition: FLA_type_defs.h:159

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_tn().