We have constructed a workload for comparing window system performance of different computers. It consists of key operations that window systems should universally provide. We have used this workload to measure the performance of a the Symbolics ZetaLisp window system on two Lisp Machines, the 3600 [3,4] and the CADR [ 1 ].
The workload consists of repeating a sequence of operations (which we call a run) many times. Each run starts by creating a test window, and proceeds to perform various operations on the window in the order below. Finally, the test window is deleted.
Complete paper in Adobe Acrobat (92,167 Bytes)