Re: Interpretation of Streams

From: Brad Lucier (lucier@MATH.Purdue.EDU)
Date: Mon Sep 25 1995 - 17:24:14 CDT


Thanks for your reply; that mini-table should be

-------------------------------------------
                copy scale add triad
-------------------------------------------
Sun 670,1 cpu 43.7 40.3 43.8 42.4
Sun 670,2 cpu 65.0 58.9 59.0 71.9
-------------------------------------------
speedups 1.49 1.46 1.35 1.70

with upgraded 55 Mhz Hypersparc modules.

I compiled stream_d.f with the provided f77 -O4, but things went even
more slowly.

Bottom line: Do you think we'd get significant improvement by upgrading
again to 125 Mhz modules? Do you think the L2 cache reload time is
governed by the bus cycle time or the CPU cycle time?

Brad



This archive was generated by hypermail 2b29 : Tue Apr 18 2000 - 05:23:05 CDT