chased@rbbb.Eng.Sun.COM (David Chase) (04/06/91)
Somewhat mystifying that anyone would speak of "a processor" achieving a certain number of cycles per instruction. The compiler also has something to do with this. When including the compiler in the measurement, be sure that you compare both CPI and total instructions executed -- all else being equal (which, of course, it isn't), I'll take 10% higher CPI if the dynamic instruction count is reduced by 20%. If *all* you care about is CPI, I'm sure that the compiler can get your code asymptotically close to the minimum (how many NOPS would you like inserted?) David Chase Sun