Groups | Search | Server Info | Keyboard shortcuts | Login | Register [http] [https] [nntp] [nntps]
Groups > comp.lang.java.programmer > #11558
| From | Wanja Gayk <brixomatic@yahoo.com> |
|---|---|
| Newsgroups | comp.lang.java.programmer |
| Subject | Re: building performance |
| Date | 2012-01-21 11:41 +0100 |
| Organization | Netfront http://www.netfront.net/ |
| Message-ID | <MPG.2984b4ac6e8ec9969896e1@202.177.16.121> (permalink) |
| References | <c8cog7piqu9t7gvgdemmo34d3n7t5ok5g2@4ax.com> |
In article <c8cog7piqu9t7gvgdemmo34d3n7t5ok5g2@4ax.com>, see_website@mindprod.com.invalid says... > > I am in a slow process of upgrading my machine to a 6-core CPU. > > I was wondering how this might affect build speed. Yes it does make a huge difference here. I have two computers with 4-core hyperthreading CPU (so 8 virtual cores each), both are Intel core i7s, one running at 2.8(desktop) and one at 2.5 Ghz (notebook). Sitting at my desktop computer, I have just tried to bind all Java processes and eclipse.exe to one virtual core and do a clean build of all projects in my private workspace, which took about 25 seconds. Allowing it to run on all cores, the clean build time drops to about 4 to 5 Seconds. I've tried that back-to-back to prevent caching effects from skewing the outcome. Indeed the taskmanager showed that during the single-core-tests only one core was busy and during the multi-core-tests all 8 virtual cores have been loaded with work. However, I can assure you, that probably the best thing that you can do for your build-performance is to buy a Solid State Disc. My last customer has a huge project that takes some minutes to build from scratch even on a solid state disc and when I had been new to the project, and just had a common mechanical drive in my notebook, it took some ten minutes or more. I would recommend an Intel X25m "Postville", which has proven to be solid in the real world, while I've seen 4 OCZ-Vertex and one Crucial drive fail at my last customers' office (drive not showing up/not mounting properly, or taking half an hour to open a directory, data loss, etc), while the Intels have had no problem whatsoever (all running the latest firmware of course), most of them runing under Windows XP, in the course of about one year. For my notebook I had bought a OCZ Vertex 2 before I have heard of these problems and luckily had no problem so far. For my desktop computer I own an Intel x25m Postville with Windows' File compression switched on, which feels a little snappier than the OCZ on my notebook, but then again - the notebook as a slightly slower CPU and is lower on memory, so handkle this statement with care. The Intel SSD will give you the best bang for the buck and has not caused any problem at my customers', so it's a clear recommendation. Use one of these as your system drive and you'll have problems getting rid of the smile on your face in the first few days until you've grown accustomed to it. > 2. Is there anything reasonable simple to take advantage of the extra > cores? What do you mean? A list of applications? Video editing software and video transcoding software loves multicore systems in general. I don't need to point out Application Servers and stuff, I think. Some games make use of it, but I don't think that's what you had in mind. Kind regards, Wanja -- ..Alesi's problem was that the back of the car was jumping up and down dangerously - and I can assure you from having been teammate to Jean Alesi and knowing what kind of cars that he can pull up with, when Jean Alesi says that a car is dangerous - it is. [Jonathan Palmer] --- Posted via news://freenews.netfront.net/ - Complaints to news@netfront.net ---
Back to comp.lang.java.programmer | Previous | Next — Previous in thread | Next in thread | Find similar | Unroll thread
building performance Roedy Green <see_website@mindprod.com.invalid> - 2012-01-10 04:41 -0800
Re: building performance Wanja Gayk <brixomatic@yahoo.com> - 2012-01-21 11:41 +0100
Re: building performance Roedy Green <see_website@mindprod.com.invalid> - 2012-01-23 17:58 -0800
Re: building performance Arne Vajhøj <arne@vajhoej.dk> - 2012-01-21 16:02 -0500
csiph-web