libflame  revision_anchor
Functions
FLA_Gemv_h.h File Reference

(r)

Go to the source code of this file.

Functions

FLA_Error FLA_Gemv_h_blk_var1 (FLA_Obj alpha, FLA_Obj A, FLA_Obj x, FLA_Obj beta, FLA_Obj y, fla_gemv_t *cntl)
 
FLA_Error FLA_Gemv_h_blk_var2 (FLA_Obj alpha, FLA_Obj A, FLA_Obj x, FLA_Obj beta, FLA_Obj y, fla_gemv_t *cntl)
 
FLA_Error FLA_Gemv_h_blk_var5 (FLA_Obj alpha, FLA_Obj A, FLA_Obj x, FLA_Obj beta, FLA_Obj y, fla_gemv_t *cntl)
 
FLA_Error FLA_Gemv_h_blk_var6 (FLA_Obj alpha, FLA_Obj A, FLA_Obj x, FLA_Obj beta, FLA_Obj y, fla_gemv_t *cntl)
 

Function Documentation

◆ FLA_Gemv_h_blk_var1()

FLA_Error FLA_Gemv_h_blk_var1 ( FLA_Obj  alpha,
FLA_Obj  A,
FLA_Obj  x,
FLA_Obj  beta,
FLA_Obj  y,
fla_gemv_t cntl 
)
14 {
15  FLA_Obj AL, AR, A0, A1, A2;
16 
17  FLA_Obj yT, y0,
18  yB, y1,
19  y2;
20 
21  dim_t b;
22 
23  FLA_Part_1x2( A, &AL, &AR, 0, FLA_LEFT );
24 
25  FLA_Part_2x1( y, &yT,
26  &yB, 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( yT, &y0,
36  /* ** */ /* ** */
37  &y1,
38  yB, &y2, b, FLA_BOTTOM );
39 
40  /*------------------------------------------------------------*/
41 
42  /* y1 = alpha * A1' * x + y1; */
43  FLA_Gemv_internal( FLA_CONJ_TRANSPOSE,
44  alpha, A1, x, beta, y1,
45  FLA_Cntl_sub_gemv( 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( &yT, y0,
53  y1,
54  /* ** */ /* ** */
55  &yB, y2, FLA_TOP );
56 
57  }
58 
59  return FLA_SUCCESS;
60 }
FLA_Error FLA_Gemv_internal(FLA_Trans transa, FLA_Obj alpha, FLA_Obj A, FLA_Obj x, FLA_Obj beta, FLA_Obj y, fla_gemv_t *cntl)
Definition: FLA_Gemv_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
double *restrict y1
Definition: bl1_dotsv2.c:145
Definition: FLA_type_defs.h:159

References FLA_Cont_with_1x3_to_1x2(), FLA_Cont_with_3x1_to_2x1(), FLA_Determine_blocksize(), FLA_Gemv_internal(), FLA_Obj_width(), FLA_Part_1x2(), FLA_Part_2x1(), FLA_Repart_1x2_to_1x3(), FLA_Repart_2x1_to_3x1(), and y1.

Referenced by FLA_Gemv_h().

◆ FLA_Gemv_h_blk_var2()

FLA_Error FLA_Gemv_h_blk_var2 ( FLA_Obj  alpha,
FLA_Obj  A,
FLA_Obj  x,
FLA_Obj  beta,
FLA_Obj  y,
fla_gemv_t cntl 
)
16 {
17  FLA_Obj AL, AR, A0, A1, A2;
18 
19  FLA_Obj yT, y0,
20  yB, y1,
21  y2;
22 
23  dim_t b;
24 
25  FLA_Part_1x2( A, &AL, &AR, 0, FLA_RIGHT );
26 
27  FLA_Part_2x1( y, &yT,
28  &yB, 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( yT, &y0,
38  &y1,
39  /* ** */ /* ** */
40  yB, &y2, b, FLA_TOP );
41 
42  /*------------------------------------------------------------*/
43 
44  /* y1 = alpha * A1' * x + y1; */
45  FLA_Gemv_internal( FLA_CONJ_TRANSPOSE,
46  alpha, A1, x, beta, y1,
47  FLA_Cntl_sub_gemv( 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( &yT, y0,
55  /* ** */ /* ** */
56  y1,
57  &yB, y2, FLA_BOTTOM );
58 
59  }
60 
61  return FLA_SUCCESS;
62 }

References FLA_Cont_with_1x3_to_1x2(), FLA_Cont_with_3x1_to_2x1(), FLA_Determine_blocksize(), FLA_Gemv_internal(), FLA_Obj_width(), FLA_Part_1x2(), FLA_Part_2x1(), FLA_Repart_1x2_to_1x3(), FLA_Repart_2x1_to_3x1(), and y1.

Referenced by FLA_Gemv_h().

◆ FLA_Gemv_h_blk_var5()

FLA_Error FLA_Gemv_h_blk_var5 ( FLA_Obj  alpha,
FLA_Obj  A,
FLA_Obj  x,
FLA_Obj  beta,
FLA_Obj  y,
fla_gemv_t cntl 
)
14 {
15  FLA_Obj AT, A0,
16  AB, A1,
17  A2;
18 
19  FLA_Obj xT, x0,
20  xB, x1,
21  x2;
22 
23  dim_t b;
24 
25  FLA_Scal_internal( beta, y,
26  FLA_Cntl_sub_scal( cntl ) );
27 
28  FLA_Part_2x1( A, &AT,
29  &AB, 0, FLA_TOP );
30 
31  FLA_Part_2x1( x, &xT,
32  &xB, 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( xT, &x0,
44  /* ** */ /* ** */
45  &x1,
46  xB, &x2, b, FLA_BOTTOM );
47 
48  /*------------------------------------------------------------*/
49 
50  /* y = alpha * A1' * x1 + y; */
51  FLA_Gemv_internal( FLA_CONJ_TRANSPOSE,
52  alpha, A1, x1, FLA_ONE, y,
53  FLA_Cntl_sub_gemv( 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( &xT, x0,
63  x1,
64  /* ** */ /* ** */
65  &xB, x2, FLA_TOP );
66 
67  }
68 
69  return FLA_SUCCESS;
70 }
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
dim_t FLA_Obj_length(FLA_Obj obj)
Definition: FLA_Query.c:116
x1
Definition: bl1_dotsv2.c:374

References FLA_Cont_with_3x1_to_2x1(), FLA_Determine_blocksize(), FLA_Gemv_internal(), FLA_Obj_length(), FLA_ONE, FLA_Part_2x1(), FLA_Repart_2x1_to_3x1(), FLA_Scal_internal(), and x1.

Referenced by FLA_Gemv_h().

◆ FLA_Gemv_h_blk_var6()

FLA_Error FLA_Gemv_h_blk_var6 ( FLA_Obj  alpha,
FLA_Obj  A,
FLA_Obj  x,
FLA_Obj  beta,
FLA_Obj  y,
fla_gemv_t cntl 
)
16 {
17  FLA_Obj AT, A0,
18  AB, A1,
19  A2;
20 
21  FLA_Obj xT, x0,
22  xB, x1,
23  x2;
24 
25  dim_t b;
26 
27  FLA_Scal_internal( beta, y,
28  FLA_Cntl_sub_scal( cntl ) );
29 
30  FLA_Part_2x1( A, &AT,
31  &AB, 0, FLA_BOTTOM );
32 
33  FLA_Part_2x1( x, &xT,
34  &xB, 0, FLA_BOTTOM );
35 
36  while ( FLA_Obj_length( AB ) < FLA_Obj_length( A ) ){
37 
38  b = FLA_Determine_blocksize( AT, FLA_TOP, FLA_Cntl_blocksize( cntl ) );
39 
40  FLA_Repart_2x1_to_3x1( AT, &A0,
41  &A1,
42  /* ** */ /* ** */
43  AB, &A2, b, FLA_TOP );
44 
45  FLA_Repart_2x1_to_3x1( xT, &x0,
46  &x1,
47  /* ** */ /* ** */
48  xB, &x2, b, FLA_TOP );
49 
50  /*------------------------------------------------------------*/
51 
52  /* y = alpha * A1' * x1 + y; */
53  FLA_Gemv_internal( FLA_CONJ_TRANSPOSE,
54  alpha, A1, x1, FLA_ONE, y,
55  FLA_Cntl_sub_gemv( cntl ) );
56 
57  /*------------------------------------------------------------*/
58 
59  FLA_Cont_with_3x1_to_2x1( &AT, A0,
60  /* ** */ /* ** */
61  A1,
62  &AB, A2, FLA_BOTTOM );
63 
64  FLA_Cont_with_3x1_to_2x1( &xT, x0,
65  /* ** */ /* ** */
66  x1,
67  &xB, x2, FLA_BOTTOM );
68 
69  }
70 
71  return FLA_SUCCESS;
72 }

References FLA_Cont_with_3x1_to_2x1(), FLA_Determine_blocksize(), FLA_Gemv_internal(), FLA_Obj_length(), FLA_ONE, FLA_Part_2x1(), FLA_Repart_2x1_to_3x1(), FLA_Scal_internal(), and x1.

Referenced by FLA_Gemv_h().