Alaric Snell-Pym scripsit: > If the supposed performance improvement can't be benchmarked, then > it's pointless, as nobody will actually benefit from it. Any case > where somebody can benefit from a performance improvement can be > turned into a benchmark that consists of running the code that is sped > up, and timing it. > > Benchmarks are like unit tests; they are snippets of code that perform > some operation but, rather than testing correct responses, their > emphasis is on testing resource usage.
Your clarification down-thread that a benchmark can be of any size makes this comparison rather otiose. Nobody is going to have a benchmark suite that includes tests like these: With patch #1234, application 'foo' runs in an acceptable 18 hours rather than an intolerable 25 hours. (Obviously the improvement has to be nonlinear.) With patch #2345, vectors larger than 2^40 elements show O(1) reference behavior rather than not. With patch #3456, systems running more than 25 million green threads are able to make forward progress rather than thrashing. And yet, tested in the necessary environment, these patches may be sound and even necessary. -- [W]hen I wrote it I was more than a little John Cowan febrile with foodpoisoning from an antique carrot co...@ccil.org that I foolishly ate out of an illjudged faith http://ccil.org/~cowan in the benignancy of vegetables. --And Rosta _______________________________________________ Chicken-users mailing list Chicken-users@nongnu.org https://lists.nongnu.org/mailman/listinfo/chicken-users