> > One interesting result was that one MVCL for 1K takes about > as long as four MVCs of 256; below that MVCs are faster on > every processor I tested.
Probably not as surprising as you think. There is only one "move" instruction on the z Series. MVCL and other complex moves are implemented in millicode. That said, I always use MVCL when the length(s) are not known at assembly time because even with the setup cost, an MVCL beats an EX of an MVC. > Another surprise (?) was that two STs were faster than an STM > for two registers. Once again, no big surprise in terms of cache and memory design. It might even turn out that the advantage holds true for a larger number of registers. Try it. And in any case "who cares"? We should not be obsessing about which instruction is faster. They are all stupidly fast and it doesn't matter a lick anymore unless you are the exquisitely hand-crafted z/OS dispatcher or some similar thing that runs a billion times a minute. CC ---------------------------------------------------------------------- For IBM-MAIN subscribe / signoff / archive access instructions, send email to [EMAIL PROTECTED] with the message: GET IBM-MAIN INFO Search the archives at http://bama.ua.edu/archives/ibm-main.html
