Had another 48hr game jam this weekend...
Brad Anderson
eco at gnuk.net
Mon Sep 16 23:39:54 PDT 2013
On Tuesday, 17 September 2013 at 06:24:20 UTC, Manu wrote:
> On 17 September 2013 15:48, deadalnix <deadalnix at gmail.com>
> wrote:
>
>> On Tuesday, 17 September 2013 at 05:32:28 UTC, Manu wrote:
>>
>>> In my experience, more memory == slower. If you care about
>>> performance,
>>> the
>>> only time it's acceptable to use more memory is if your data
>>> structures
>>> are
>>> as efficient as they can get, and the alternative is reading
>>> off the hard
>>> drive.
>>> Bandwidth isn't free, cache is only so big, and logic to
>>> process and make
>>> use of so much memory isn't free either. It usually just
>>> suggests
>>> inefficient (or just lazy) data structures, which often also
>>> implies
>>> inefficient processing logic.
>>> And the more memory an app uses, the higher chance of
>>> invoking the page
>>> file, which is a mega-killer.
>>>
>>>
>> I do agree as this is generally true. However, the problem
>> isn't really
>> cache size or bandwidth, but rather latency. We know how to
>> increase
>> bandwith or cache size, but the first one come at a cost with
>> no big
>> benefit, and the second come at increase of cost and increase
>> of latency.
>> What is capping the perf here is really latency.
>>
>
> Latency bottlenecks are usually a function of inefficient cache
> usage, or a
> working set that's too large and non-linear.
>
> That being said, less memory == more of your working set in
> cache => faster
>> program.
>
>
> Precisely.
>
> Dunno what to tell you. My VS instance is pretty light.
>>>
>>>
>> Yup, VS is one of these program that microsoft did better than
>> the
>> alternative :D
>
>
> Perhaps the only one, and also the single reason I still use
> Windows
> (despite their best efforts to ruin it more and more with
> almost every
> release!). There is STILL no realistic alternative for my
> money, well over
> a decade later...
> I don't get it. VS has been there a long time. It's not even
> perfect; farm
> from it in fact. But the fact that given over a decade of solid
> working
> example, nobody has yet managed to create a competitive product
> just blows
> my mind...
> Seriously, where is the competition? I probably use about 10%
> of VS's
> features, but the features that I do use and rely on work, and
> work well.
> Although even they could be significantly improved in some very
> simple ways.
>
> I closed about half my open tabs after my last email (~50 left
> open). Down
>>> to 93mb. You must all use some heavy plugins or something.
>>> My current solution has 10 projects, one is an entire game
>>> engine with
>>> over
>>> 500 source files, hundreds of thousands of LOC. Intellisense
>>> info for all
>>> of it... dunno what to tell you.
>>> Eclipse uses more than 4 times that much memory idling with
>>> no project
>>> open
>>> at all...
>>>
>>>
>> 4 times ? You must have a pretty light instance of eclipse !
>>
>
> It's a fairly fresh eclipse install, and I just boot it up. It
> showed the
> home screen, no project loaded. It was doing absolutely nothing
> and well
> into 400mb.
> When I do use it for android and appengine, it more or less
> works well
> enough, but the UI feels like it's held together with
> stickytape and glue,
> and it's pretty sluggish. Debugging (native code) is slow and
> clunky. How
> can I take that software seriously?
> I probably waste significant portion of my life hovering and
> waiting for
> eclipse to render the pop-up variable inspection windows. That
> shit needs
> to be instant, no excuse. It's just showing a value from ram.
> Then I press a key, it doesn't take ages for the letter to
> appear on the
> screen...
Better get used to it. The Gaben has spoken:
http://arstechnica.com/gaming/2013/09/gabe-newell-linux-is-the-future-of-gaming-new-hardware-coming-soon/
I actually agree, my experience with full blown IDEs other than
VS has been terrible (and I just spent all day fixing a VS 2010
PCH corruption bug). I've always got my beloved vim to fall back
on though.
More information about the Digitalmars-d
mailing list