libflame revision_anchor
FLASH_SA_FS.c File Reference

(r)

Functions

FLA_Error FLASH_SA_FS (FLA_Obj L, FLA_Obj D, FLA_Obj p, FLA_Obj C, FLA_Obj E, dim_t nb_alg, fla_gemm_t *cntl)
 

Function Documentation

◆ FLASH_SA_FS()

FLA_Error FLASH_SA_FS ( FLA_Obj L,
FLA_Obj D,
FLA_Obj p,
FLA_Obj C,
FLA_Obj E,
dim_t nb_alg,
fla_gemm_t * cntl )
16{
17 FLA_Obj CL, CR, C0, C1, C2;
18
19 FLA_Obj EL, ER, E0, E1, E2;
20
21 FLA_Part_1x2( C, &CL, &CR, 0, FLA_LEFT );
22
23 FLA_Part_1x2( E, &EL, &ER, 0, FLA_LEFT );
24
25 while ( FLA_Obj_width( CL ) < FLA_Obj_width( C ) )
26 {
27 FLA_Repart_1x2_to_1x3( CL, /**/ CR, &C0, /**/ &C1, &C2,
28 1, FLA_RIGHT );
29
30 FLA_Repart_1x2_to_1x3( EL, /**/ ER, &E0, /**/ &E1, &E2,
31 1, FLA_RIGHT );
32
33 /*------------------------------------------------------------*/
34
36 {
37 // Enqueue
38 ENQUEUE_FLASH_SA_FS( *FLASH_OBJ_PTR_AT( L ),
39 *FLASH_OBJ_PTR_AT( D ),
40 *FLASH_OBJ_PTR_AT( p ),
41 *FLASH_OBJ_PTR_AT( C1 ),
42 *FLASH_OBJ_PTR_AT( E1 ),
43 nb_alg,
44 FLA_Cntl_sub_gemm( cntl ) );
45 }
46 else
47 {
48 // Execute leaf
49 FLA_SA_FS_task( *FLASH_OBJ_PTR_AT( L ),
50 *FLASH_OBJ_PTR_AT( D ),
51 *FLASH_OBJ_PTR_AT( p ),
52 *FLASH_OBJ_PTR_AT( C1 ),
53 *FLASH_OBJ_PTR_AT( E1 ),
54 nb_alg,
55 FLA_Cntl_sub_gemm( cntl ) );
56 }
57
58 /*------------------------------------------------------------*/
59
60 FLA_Cont_with_1x3_to_1x2( &CL, /**/ &CR, C0, C1, /**/ C2,
61 FLA_LEFT );
62
63 FLA_Cont_with_1x3_to_1x2( &EL, /**/ &ER, E0, E1, /**/ E2,
64 FLA_LEFT );
65 }
66
67 return FLA_SUCCESS;
68}
FLA_Bool FLASH_Queue_get_enabled(void)
Definition FLASH_Queue.c:171
FLA_Error FLA_SA_FS_task(FLA_Obj L, FLA_Obj D, FLA_Obj p, FLA_Obj C, FLA_Obj E, dim_t nb_alg, fla_gemm_t *cntl)
Definition FLA_SA_FS_task.c:13
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_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
Definition FLA_type_defs.h:159

References FLA_Cont_with_1x3_to_1x2(), FLA_Obj_width(), FLA_Part_1x2(), FLA_Repart_1x2_to_1x3(), FLA_SA_FS_task(), and FLASH_Queue_get_enabled().

Referenced by FLASH_SA_LU().