Re:Superoptimizer – A Look at the Smallest Program [pdf] (1987)
by monocasa on Tuesday November 24, 2020 @06:00PM
(#25204028)
It's from an era when program length was strongly correlated with program performance. And when it wasn't it was still a fixed known cost per instruction type, and could be statically computed.
[ Parent
| on HN ]
Re:Superoptimizer – A Look at the Smallest Program [pdf] (1987)
by djmips on Wednesday November 25, 2020 @09:28PM
(#25216320)
I was kind of nodding and agreeing with you but then I remembered a bunch of optimal code from back in those days that was very unwound and super large so it depended on the type of work being done. I guess one could say that purely algorithmic code tended to have size and speed correlated whereas code that was bottlenecked in memory read/write could benefit from being large. Cache wasn't having the big impact like today.