Comments on: Software always gets to decide what to do next http://blog.reverberate.org/2007/02/18/software-always-gets-to-decide-what-to-do-next/ parsing, performance, minimalism with C99 Mon, 06 Feb 2012 23:44:51 +0000 hourly 1 http://wordpress.org/?v=3.3.1 By: Jason G. http://blog.reverberate.org/2007/02/18/software-always-gets-to-decide-what-to-do-next/comment-page-1/#comment-205 Jason G. Wed, 21 Mar 2007 18:41:18 +0000 http://blog.reverberate.org/2007/02/18/software-always-gets-to-decide-what-to-do-next/#comment-205 Hey Josh, just tracked you down using a search engine - your blog is very interesting. Niche, but still a great read. Neat to see you have so many bold ambitions ahead of you. If you do write an OS, can you drop me in one of the shout outs via an Easter Egg? I have just concluded you are far more intelligent than I ever will be. Which is why I need to hire you some day... Hey Josh, just tracked you down using a search engine – your blog is very interesting. Niche, but still a great read. Neat to see you have so many bold ambitions ahead of you. If you do write an OS, can you drop me in one of the shout outs via an Easter Egg?

I have just concluded you are far more intelligent than I ever will be. Which is why I need to hire you some day…

]]> By: josh http://blog.reverberate.org/2007/02/18/software-always-gets-to-decide-what-to-do-next/comment-page-1/#comment-14 josh Sun, 18 Feb 2007 23:22:49 +0000 http://blog.reverberate.org/2007/02/18/software-always-gets-to-decide-what-to-do-next/#comment-14 If real-time applications don't understand their own resource requirements this finely, then they will simply glitch when they need more resources than the OS happens to give them. All it means is that those applications can't guarantee their users anything about their real-time performance. Yes, realistically games will probably fall in this category, just because they are so complex. Games certainly <i>could</i> be written to have fixed and knowable resource requirements, but it would be too complicated and too expensive to be worth it. Console games know that they will get 100% of the resources of a highly standardized hardware setup, so the game developers try it out until it runs fast enough and has infrequent-enough slowdowns, and they call it good. Computer games end up being as reliable as your computer is modern (as long as you're not running other resource-hogging software at the same time). Computer games certainly have the option of doing more intelligent things than loading models the split-second before they are needed, and I bet they do (otherwise games would be way less fluid than they are). They can either load everything into memory ahead of time, or determine what data is likely to be needed "soon" based on what is in proximity of the player but isn't yet visible. If real-time applications don’t understand their own resource requirements this finely, then they will simply glitch when they need more resources than the OS happens to give them. All it means is that those applications can’t guarantee their users anything about their real-time performance.

Yes, realistically games will probably fall in this category, just because they are so complex. Games certainly could be written to have fixed and knowable resource requirements, but it would be too complicated and too expensive to be worth it. Console games know that they will get 100% of the resources of a highly standardized hardware setup, so the game developers try it out until it runs fast enough and has infrequent-enough slowdowns, and they call it good. Computer games end up being as reliable as your computer is modern (as long as you’re not running other resource-hogging software at the same time).

Computer games certainly have the option of doing more intelligent things than loading models the split-second before they are needed, and I bet they do (otherwise games would be way less fluid than they are). They can either load everything into memory ahead of time, or determine what data is likely to be needed “soon” based on what is in proximity of the player but isn’t yet visible.

]]> By: Buffalo http://blog.reverberate.org/2007/02/18/software-always-gets-to-decide-what-to-do-next/comment-page-1/#comment-13 Buffalo Sun, 18 Feb 2007 22:23:41 +0000 http://blog.reverberate.org/2007/02/18/software-always-gets-to-decide-what-to-do-next/#comment-13 I don't understand how your applications know such explicit details about their own resource requirements. I imagine in some applications you can set up a direct correlation between time and the amount of resources you need but surely it cannot universally be the case. Just think of a video game - the amount of processor time I need is a function of the scene I'm currently viewing - even an upper bound really isn't possible because (say) I might using a fan-built map that's more complex than the maps that came with the game. And almost everything is pretty bursty right? The game needs basically no disk bandwidth right until the second you dash into the room with 4 models you've never seen before and the game needs to load them from disk - then the game need tons of disk bandwidth, all within the instant it takes to render 1 frame. I don’t understand how your applications know such explicit details about their own resource requirements. I imagine in some applications you can set up a direct correlation between time and the amount of resources you need but surely it cannot universally be the case. Just think of a video game – the amount of processor time I need is a function of the scene I’m currently viewing – even an upper bound really isn’t possible because (say) I might using a fan-built map that’s more complex than the maps that came with the game.

And almost everything is pretty bursty right? The game needs basically no disk bandwidth right until the second you dash into the room with 4 models you’ve never seen before and the game needs to load them from disk – then the game need tons of disk bandwidth, all within the instant it takes to render 1 frame.

]]>