The Cray T3D, an MIMD system with NUMA shared memory capabilities and
in principle very low communications latency, can support the Canopy f
ramework for grid-oriented applications. CANOPY has been ported to the
T3D, with the intent of making it available to 8 spectrum of users. T
he performance of the T3D running Canopy has been benchmarked on five
QCD applications extensively run on ACPMAPS at Fermilab, requiring a v
ariety of data access patterns. The net performance and scaling behavi
or reveals an efficiency relative to peak Gflops almost identical to t
hat achieved on ACPMAPS. Detailed studies of the major factors impacti
ng performance are presented. Generalizations applying this analysis t
o the newly emerging crop of commercial systems reveal where their lim
itations will lie. On these applications, efficiencies of above 25% ar
e not to be expected; eliminating overheads due to Canopy will improve
matters, but by less than a factor of two.