//------------------------------------------------------------------------------ // GrB_vxm: vector-matrix multiply //------------------------------------------------------------------------------ // SuiteSparse:GraphBLAS, Timothy A. Davis, (c) 2017-2023, All Rights Reserved. // SPDX-License-Identifier: Apache-2.0 //------------------------------------------------------------------------------ // w' = accum (w',t) where t = u'*A or u'*A' // Rows w', u', and M' are simply columns w, u, and M. Thus: // w = accum (w,t) where t = A'*u or A*u, but with the multiply operator // flipped. The input descriptor for A, inp1, is also negated. #include "GB_mxm.h" #include "GB_get_mask.h" GrB_Info GrB_vxm // w' = accum (w', u'*A) ( GrB_Vector w, // input/output vector for results const GrB_Vector M_in, // optional mask for w, unused if NULL const GrB_BinaryOp accum, // optional accum for z=accum(w,t) const GrB_Semiring semiring, // defines '+' and '*' for matrix multiply const GrB_Vector u, // first input: vector u const GrB_Matrix A, // second input: matrix A const GrB_Descriptor desc // descriptor for w, M, and A ) { //-------------------------------------------------------------------------- // check inputs //-------------------------------------------------------------------------- GB_WHERE (w, "GrB_vxm (w, M, accum, semiring, u, A, desc)") ; GB_BURBLE_START ("GrB_vxm") ; GB_RETURN_IF_NULL_OR_FAULTY (w) ; GB_RETURN_IF_FAULTY (M_in) ; GB_RETURN_IF_NULL_OR_FAULTY (u) ; GB_RETURN_IF_NULL_OR_FAULTY (A) ; ASSERT (GB_VECTOR_OK (w)) ; ASSERT (M_in == NULL || GB_VECTOR_OK (M_in)) ; ASSERT (GB_VECTOR_OK (u)) ; // get the descriptor GB_GET_DESCRIPTOR (info, desc, C_replace, Mask_comp, Mask_struct, xx, A_transpose, AxB_method, do_sort) ; // get the mask GrB_Matrix M = GB_get_mask ((GrB_Matrix) M_in, &Mask_comp, &Mask_struct) ; //-------------------------------------------------------------------------- // w' = accum (w',u'*A) and variations, using the mxm kernel //-------------------------------------------------------------------------- // w, M, and u are treated as column vectors and passed as n-by-1 matrices // to GB_mxm A and u are swapped, and A_transpose is negated: // u'*A == A'*u // u'*A' == A*u // Since A and u are swapped, in all the matrix multiply kernels, // the multiplier must be flipped, so flipxy is passed in as true. info = GB_mxm ( (GrB_Matrix) w, C_replace, // w and its descriptor M, Mask_comp, Mask_struct, // mask and its descriptor accum, // for accum (w,t) semiring, // definition of matrix multiply A, !A_transpose, // allow A to be transposed (GrB_Matrix) u, false, // u is never transposed true, // fmult(y,x), flipxy = true AxB_method, do_sort, // algorithm selector Werk) ; GB_BURBLE_END ; return (info) ; }