libflame  revision_anchor
Functions
FLA_Gemm_hc_blk_var5.c File Reference

(r)

Functions

FLA_Error FLA_Gemm_hc_blk_var5 (FLA_Obj alpha, FLA_Obj A, FLA_Obj B, FLA_Obj beta, FLA_Obj C, fla_gemm_t *cntl)
 

Function Documentation

◆ FLA_Gemm_hc_blk_var5()

FLA_Error FLA_Gemm_hc_blk_var5 ( FLA_Obj  alpha,
FLA_Obj  A,
FLA_Obj  B,
FLA_Obj  beta,
FLA_Obj  C,
fla_gemm_t cntl 
)
14 {
15  FLA_Obj AT, A0,
16  AB, A1,
17  A2;
18 
19  FLA_Obj BT, B0,
20  BB, B1,
21  B2;
22 
23  dim_t b;
24 
25  FLA_Scal_internal( beta, C,
26  FLA_Cntl_sub_scal( cntl ) );
27 
28  FLA_Part_2x1( A, &AT,
29  &AB, 0, FLA_TOP );
30 
31  FLA_Part_2x1( B, &BT,
32  &BB, 0, FLA_TOP );
33 
34  while ( FLA_Obj_length( AT ) < FLA_Obj_length( A ) ){
35 
36  b = FLA_Determine_blocksize( AB, FLA_BOTTOM, FLA_Cntl_blocksize( cntl ) );
37 
38  FLA_Repart_2x1_to_3x1( AT, &A0,
39  /* ** */ /* ** */
40  &A1,
41  AB, &A2, b, FLA_BOTTOM );
42 
43  FLA_Repart_2x1_to_3x1( BT, &B0,
44  /* ** */ /* ** */
45  &B1,
46  BB, &B2, b, FLA_BOTTOM );
47 
48  /*------------------------------------------------------------*/
49 
50  /* C = alpha * A1' * B1 + C; */
51  FLA_Gemm_internal( FLA_CONJ_TRANSPOSE, FLA_CONJ_NO_TRANSPOSE,
52  alpha, A1, B1, FLA_ONE, C,
53  FLA_Cntl_sub_gemm( cntl ) );
54 
55  /*------------------------------------------------------------*/
56 
57  FLA_Cont_with_3x1_to_2x1( &AT, A0,
58  A1,
59  /* ** */ /* ** */
60  &AB, A2, FLA_TOP );
61 
62  FLA_Cont_with_3x1_to_2x1( &BT, B0,
63  B1,
64  /* ** */ /* ** */
65  &BB, B2, FLA_TOP );
66 
67  }
68 
69  return FLA_SUCCESS;
70 }
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_Scal_internal(FLA_Obj alpha, FLA_Obj A, fla_scal_t *cntl)
Definition: FLA_Scal_internal.c:16
FLA_Obj FLA_ONE
Definition: FLA_Init.c:18
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_Part_2x1(FLA_Obj A, FLA_Obj *A1, FLA_Obj *A2, dim_t mb, FLA_Side side)
Definition: FLA_View.c:76
dim_t FLA_Obj_length(FLA_Obj obj)
Definition: FLA_Query.c:116
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_3x1_to_2x1(), FLA_Determine_blocksize(), FLA_Gemm_internal(), FLA_Obj_length(), FLA_ONE, FLA_Part_2x1(), FLA_Repart_2x1_to_3x1(), and FLA_Scal_internal().

Referenced by FLA_Gemm_hc().