The HLL Performance Gap
geoff at broadwell.org
Wed Jul 22 03:28:54 UTC 2009
On Tue, 2009-07-21 at 21:07 -0400, Austin Hastings wrote:
> 0. Is there a standard set of performance/timing tests? If not, can we
> make one?
> In particular,
> 0a. How long does it take to initialize parrot and execute a single
> "HALT" instruction?
> 0b. How long does it take to perform opcode O, 10**X-sub-O times, for
> some appropriate value of X-sub-O, for each O?
> 0c. How long does it take to run a standard program, with functions and
> data and objects and file i/o, Y times?
> 0d. How long does it take to run (some troublesome subsystem 'S')
> Z-sub-S times?
I've thought about doing something similar in the past, but I was not
yet convinced that the benefit would be worth the time spent. Seeing
that someone else independently came up with the same idea pushes me a
little further towards JFDI.
Anyone else care to weigh in?
More information about the parrot-dev