libflame revision_anchor
|
Functions | |
FLA_Error | FLASH_Apply_Q_UT_inc_create_workspace (FLA_Obj TW, FLA_Obj B, FLA_Obj *W) |
References FLA_Abort(), FLA_Obj_datatype(), FLA_Obj_width(), FLA_Print_message(), FLASH_Obj_create_ext(), FLASH_Obj_depth(), FLASH_Obj_scalar_length_tl(), and FLASH_Obj_scalar_width_tl().
Referenced by FLASH_QR_UT_inc_solve().
{ FLA_Datatype datatype; dim_t depth; dim_t b_alg; dim_t b_flash; dim_t m, n; // Query the depth. depth = FLASH_Obj_depth( TW ); // *** The current Apply_Q_UT_inc algorithm implemented assumes that // the matrix has a hierarchical depth of 1. We check for that here // because we anticipate that we'll use a more general algorithm in the // future, and we don't want to forget to remove the constraint. *** if ( depth != 1 ) { FLA_Print_message( "FLASH_Apply_Q_UT_inc() currently only supports matrices of depth 1", __FILE__, __LINE__ ); FLA_Abort(); } // Query the datatype of matrix TW. datatype = FLA_Obj_datatype( TW ); // Inspect the length of a the top-left element of TW to get the // algorithmic blocksize we'll use throughout the Apply_Q_UT_inc // algorithm. b_alg = FLASH_Obj_scalar_length_tl( TW ); // The width of the top-left element gives us the storage blocksize. b_flash = FLASH_Obj_scalar_width_tl( TW ); // The element length of W is 1. m = 1; // Query the element (not scalar) width of the right-hand side // matrix B. This is done so we can create W with full blocks for the // right "edge cases" of B. n = FLA_Obj_width( B ); // Create hierarchical matrix W. FLASH_Obj_create_ext( datatype, m * b_alg, n * b_flash, depth, &b_alg, &b_flash, W ); return FLA_SUCCESS; }