predict performance on other machine
in [Solaris]
|
Prev: Questions after install Solaris 10
Next: c0t2d0s3?
From: Richard B. Gilbert on 5 Jul 2010 09:43 adirtymindisajoyforever wrote: > Hi, > > I have a mainly CPU bound application that performs X good_things/ > second on > a V490, I would like to predict the number when executed on a T5220? > Or in general, is there kinda formula to convert from one sparc type > machine > to another? > What if the number of CPU's is not the same? > > Thanks in advance for all answers. Even if you *know* that the new CPU is 3.72 times faster than the old, it's extremely unlikely that your application will run exactly 3.72 times faster. It may be either faster or slower than that. The only *meaningful* answer will be found by benchmarking your application on both the old and the new hardware! Adding CPUs to the mix addresses only part of the problem. Many applications are I/O bound and probably will not run noticeably faster even with more or faster CPUs. If you want, or need, better performance, you will get much better results by determining just what and where the bottlenecks are! (Keep in mind that there may be more than one cause of the poor performance.)
From: Richard B. Gilbert on 5 Jul 2010 10:04 Thommy M. wrote: > Bill Waddington <william.waddington(a)beezmo.com> writes: > >> On Mon, 5 Jul 2010 12:20:41 +0000 (UTC), hume.spamfilter(a)bofh.ca >> wrote: >> >> >>> ("CPUs", no apostrophe...) >> My friend, that ship has sailed. >> >> On the interweb, all words ending in "s" require an apostrophe. >> >> I confess, though, that I haven't seen "hi's" or "her's" yet. Soon, >> I suspect. > > > We've seen a couple of "it's" for "its" though... ;) If that's the worst you've seen, please fasten your seatbelt!!!! Literacy in English is not common even among native speakers! I recall hearing "Her and Bruce went to a movie." from a grown man, born, raised, and "educated" in the U.S. I'm sure he drove his teachers crazy. He learned that "at his mother's knee" and wasn't about to change. BTW, did you notice the missing apostrophe in "confess"? ;-) In "words"? ;-)
From: Bill Waddington on 5 Jul 2010 10:29 On Mon, 05 Jul 2010 10:04:05 -0400, "Richard B. Gilbert" <rgilbert88(a)comcast.net> wrote: >Thommy M. wrote: >> Bill Waddington <william.waddington(a)beezmo.com> writes: >> >>> On Mon, 5 Jul 2010 12:20:41 +0000 (UTC), hume.spamfilter(a)bofh.ca >>> wrote: >>> >>> >>>> ("CPUs", no apostrophe...) >>> My friend, that ship has sailed. >>> >>> On the interweb, all words ending in "s" require an apostrophe. >>> >>> I confess, though, that I haven't seen "hi's" or "her's" yet. Soon, >>> I suspect. >> >> >> We've seen a couple of "it's" for "its" though... ;) > >If that's the worst you've seen, please fasten your seatbelt!!!! >Literacy in English is not common even among native speakers! s/even/especially/ > >I recall hearing "Her and Bruce went to a movie." from a grown man, >born, raised, and "educated" in the U.S. I'm sure he drove his teachers >crazy. He learned that "at his mother's knee" and wasn't about to change. > >BTW, did you notice the missing apostrophe in "confess"? ;-) In "words"? >;-) Sorry, I confe's's Im not that good with word's. While we're way off here in the OT weeds, how about preventive -> preventative -> preventatative -> preventatatative .... And don't get me started on "begs the question", or "starting a sentence (or horrors, a paragraph) with a conjunction", or the "excessive use of quotation marks." Oh, the humanity. Bill (back under the bridge) -- William D Waddington william.waddington(a)beezmo.com "Even bugs...are unexpected signposts on the long road of creativity..." - Ken Burtch
From: Thommy M. on 5 Jul 2010 10:31 "Richard B. Gilbert" <rgilbert88(a)comcast.net> writes: > Thommy M. wrote: >> Bill Waddington <william.waddington(a)beezmo.com> writes: >> >>> On Mon, 5 Jul 2010 12:20:41 +0000 (UTC), hume.spamfilter(a)bofh.ca >>> wrote: >>> >>> >>>> ("CPUs", no apostrophe...) >>> My friend, that ship has sailed. >>> >>> On the interweb, all words ending in "s" require an apostrophe. >>> >>> I confess, though, that I haven't seen "hi's" or "her's" yet. Soon, >>> I suspect. >> >> >> We've seen a couple of "it's" for "its" though... ;) > > If that's the worst you've seen, please fasten your seatbelt!!!! > Literacy in English is not common even among native speakers! > > I recall hearing "Her and Bruce went to a movie." from a grown man, > born, raised, and "educated" in the U.S. I'm sure he drove his > teachers crazy. He learned that "at his mother's knee" and wasn't > about to change. > > BTW, did you notice the missing apostrophe in "confess"? ;-) In "words"? > ;-) Not easy for us Swedes (or Swede's) to catch them all... ;)
From: Andrew Gabriel on 5 Jul 2010 11:40
In article <i0simp$17j$1(a)kil-nws-1.ucis.dal.ca>, hume.spamfilter(a)bofh.ca writes: > adirtymindisajoyforever <getridofthespam(a)yahoo.com> wrote: >> I have a mainly CPU bound application that performs X good_things/ >> second on >> a V490, I would like to predict the number when executed on a T5220? >> Or in general, is there kinda formula to convert from one sparc type >> machine >> to another? Not enough info to answer the question. What number/type of processors/speeds does your V490 have? How fully are you utilizing them? How will your app perform if it sees very many more, but slower CPUs? M-series would be a safer bet, unless you can provide much more detail about the app, which shows it would work well on T-series. > In general, no. You can get a VERY rough estimate by comparing clock rates. > But depending on operations, other things may matter... like floating-point > ops, memory accesses, and so on. Clock rate comparison probably isn't going to work here. T-series - probably want to half the clock rate for any sort of direct comparison, so it depends on you being able to use all those extra CPUs, in which case it's a real winner. M-series - I've seen a CPU intensive app get a 3x gain at almost the same clock speed as V-series, so this is good for single-threaded (or lower numbers of threads) apps which need higher performance per thread. >> What if the number of CPU's is not the same? > > ("CPUs", no apostrophe...) That'll only make a big difference if your app is > multithreaded, which you haven't specified, so I'm assuming not. If not, it > may make some small difference by virtue of the fact of how much contention > there is for the CPUs relative to the other machine. Yep, need to know a lot more about the app, and how it's using the existing system. -- Andrew Gabriel [email address is not usable -- followup in the newsgroup] |