.TH BStri_mult 3 "4/4/1996" " " "BlockSolve95"
.SH NAME
BStri_mult \- Multiply the matrix (A - shift*B) by a block of vectors

.SH INPUT PARAMETERS
.PD 0
.TP
.B A 
- a sparse matrix
.PD 1
.PD 0
.TP
.B comm_A 
- the communication structure for A
.PD 1
.PD 0
.TP
.B B 
- a sparse matrix
.PD 1
.PD 0
.TP
.B comm_B 
- the communication structure for B
.PD 1
.PD 0
.TP
.B v1 
- the block of vectors to multiply by
.PD 1
.PD 0
.TP
.B t1 
- a block of work vectors
.PD 1
.PD 0
.TP
.B t2 
- a block of work vectors
.PD 1
.PD 0
.TP
.B shift 
- the shift value in (A-shift*B)
.PD 1
.PD 0
.TP
.B BS 
- the number of vectors in v1
.PD 1
.PD 0
.TP
.B procinfo 
- the usual processor stuff
.PD 1

.SH OUTPUT PARAMETERS
.PD 0
.TP
.B v2 
- the resulting block of vectors
.PD 1

.SH RETURNS
void

.SH NOTES
Different code is used to multiply a single vector than is used
to multiply a block of vectors (this improves efficiency).  Also
different code is used if shift=0.0.  If B is NULL, then we
assume that it is the identity matrix.

.SH SYNOPSIS
.nf
void BStri_mult(BSpar_mat *A, BScomm *comm_A, BSpar_mat *B, BScomm *comm_B,
 FLOAT *v1, FLOAT *v2, FLOAT *t1, FLOAT *t2, FLOAT shift, int BS,
 BSprocinfo *procinfo)

.fi

.SH LOCATION
 BlockSolve95/src/BStri_mult.c