The capability to blend the message passing and shared memory programming e
nvironments is a common feature of many high performance computer architect
ures. To probe the performance benefits of this feature on a 64-processor H
P SPP-2000. a 32-processor IBM SP3. and a 128-processor SGI Origin 2000, we
propose a new hybrid parallel fast Givens QR factorization algorithm that
can interpolate between pure message passing and pure shared memory. (C) 20
01 The Franklin Institute. Published by Elsevier Science Ltd. All rights re
served.