Would you like to have GPU accelerated compiling tools? - Developers

My friend could be able to edit the source codes of vvis, vrad and vbsp to make them use the graphics card to compile. Would you be interested in this project? Vote in the poll.

If it could harness the energy of both the CPU and the GPU at the same time, then yes.

[QUOTE=magravn;23563704]If it could harness the energy of both the CPU and the GPU at the same time, then yes.[/QUOTE] Yes OpenCL is able to do that.

Yes!

Well fuck yeah, would be excellent!

[QUOTE=magravn;23563704]If it could harness the energy of both the CPU and the GPU at the same time, then yes.[/QUOTE] To be honest I'd prefer it if it ran entirely on the GPU. That way I could do other stuff whilst it compiles, which you can do at the moment but at the risk of a considerably slower compile and possible crashes.

[QUOTE=Lord Pirate;23565069]To be honest I'd prefer it if it ran entirely on the GPU. That way I could do other stuff whilst it compiles, which you can do at the moment but at the risk of a considerably slower compile and possible crashes.[/QUOTE] Of course it could be possible to decide.

VBCT supports VMPI (GPU calculation). The issue is, that it wont support OB-source (valve needs to update their tools to support it. Got a word from Quicksilver, that patch is done ages ago, but for some reason valve havent released it.), only EP1(2006)-source.

[QUOTE=killer89;23566448]VBCT supports VMPI (GPU calculation). The issue is, that it wont support OB-source (valve needs to update their tools to support it. Got a word from Quicksilver, that patch is done ages ago, but for some reason valve havent released it.), only EP1(2006)-source.[/QUOTE] VMPI isn't GPU calculation :tinfoil: it's for compiling over a network... [editline]05:50PM[/editline] [QUOTE=Lord Pirate;23565069]To be honest I'd prefer it if it ran entirely on the GPU. That way I could do other stuff whilst it compiles, which you can do at the moment but at the risk of a considerably slower compile and possible crashes.[/QUOTE] Get a multicore pc, use the -threads compile parameter, set the number of threads to # of cores - 1, voilà, you can do other stuff while you compile.

[QUOTE=metallics;23566566]VMPI isn't GPU calculation :tinfoil: it's for compiling over a network... [editline]05:50PM[/editline] Get a multicore pc, use the -threads compile parameter, set the number of threads to # of cores - 1, voilà, you can do other stuff while you compile.[/QUOTE] I have used GPU for VMPI calculation in VBCT. Watched the GPU's usage with GPU-Z while the compile was on, and GPU was clearly in medium-heavy use.

I agree that CPU and GPU would be amazing. [editline]06:21PM[/editline] [QUOTE=metallics;23566566]VMPI isn't GPU calculation :tinfoil: it's for compiling over a network... [editline]05:50PM[/editline] Get a multicore pc, use the -threads compile parameter, set the number of threads to # of cores - 1, voilà, you can do other stuff while you compile.[/QUOTE] Does the hammer editor automatically use all the cores your PC has or do you have to set it manually? If so I am bummed.

It uses all by default.

Who put no on the poll?

[QUOTE=Rombishead;23567206]Who put no on the poll?[/QUOTE] Me, because I wouldn't use it. My maps are correctly optimised and I have 4 cores so I can assign it to 3 and listen to music and browse whilst it's working for all of 10 minutes. You only asked if [B]I[/B] would like to have it though, I'm sure it would be helpful for others out there with lower spec machines.

[QUOTE=metallics;23567376]Me, because I wouldn't use it. My maps are correctly optimised and I have 4 cores so I can assign it to 3 and listen to music and browse whilst it's working for all of 10 minutes. You only asked if [B]I[/B] would like to have it though, I'm sure it would be helpful for others out there with lower spec machines.[/QUOTE] It's not a matter of optimisation. You can have a map that runs vis in less than 1 minute (like every proper mapper does), but rad can take hours too, expecially with HDR enabled (meaning it has to be exectued 2 times). If you make final compiles of a very complex and detailed map (say a TF2 map), rad can take more than an hour with Static Prop Lighting enabled. [editline]06:06PM[/editline] [QUOTE=killer89;23566448]VBCT supports VMPI (GPU calculation). The issue is, that it wont support OB-source (valve needs to update their tools to support it. Got a word from Quicksilver, that patch is done ages ago, but for some reason valve havent released it.), only EP1(2006)-source.[/QUOTE] VMPI is a tool used for distributed compiling (sharing the compile job over multiple computers) and it does not involve in any way usage of GPUs.

[QUOTE=TerabyteS;23567894]It's not a matter of optimisation. You can have a map that runs vis in less than 1 minute (like every proper mapper does), but rad can take hours too, expecially with HDR enabled (meaning it has to be exectued 2 times). If you make final compiles of a very complex and detailed map (say a TF2 map), rad can take more than an hour with Static Prop Lighting enabled. [editline]06:06PM[/editline] VMPI is a tool used for distributed compiling (sharing the compile job over multiple computers) and it does not involve in any way usage of GPUs.[/QUOTE] Be more careful with lightmap scale usage (that's optimization too!)? I've never had anything take more than 45 minutes and that was for a full final HDR compile (where I may or may not have been somewhat careless when picking lightmap scales)

But you could have a better PC could you not?

[QUOTE=metallics;23567938]Be more careful with lightmap scale usage (that's optimization too!)? I've never had anything take more than 45 minutes and that was for a full final HDR compile (where I may or may not have been somewhat careless when picking lightmap scales)[/QUOTE] My pc is pretty old and sucky (2008), and TF2 maps tend to get a lot heavier than other maps for the amount of geometry detail that goes in. Even if I didn't do much lightmaps fixing, I can still say it takes a while.

[QUOTE=TerabyteS;23567984]My pc is pretty old and sucky (2008), and TF2 maps tend to get a lot heavier than other maps for the amount of geometry detail that goes in. Even if I didn't do much lightmaps fixing, I can still say it takes a while.[/QUOTE] You pair could, you know read my posts. I've already admitted to having a decent mid-range pc, and you asked if [I]I[/I] would want it, and I've said no and explained why, because it would be of no use to me. [editline]07:43PM[/editline] And as better multicores become more common, it's less and less likely to become useful. 2 years ago, I'd have said yes please gimme gimme, but you're a bit late.

It would be nice if the feature was there for a final compile, but it is not necessary for anything else.

[QUOTE=metallics;23567376]Me, because I wouldn't use it. My maps are correctly optimised and I have 4 cores so I can assign it to 3 and listen to music and browse whilst it's working for all of 10 minutes. You only asked if [B]I[/B] would like to have it though, I'm sure it would be helpful for others out there with lower spec machines.[/QUOTE] Are you trying to boast or something. Why wouldn't you want it faster? You're just being stupid.

[QUOTE=Hostel;23569232]It would be nice if the feature was there for a final compile, but it is not necessary for anything else.[/QUOTE] Reducing 1 min to 10 seconds would help too.

Again, i have 4 cores, and my gfx card is only an 8800gt. I would have no use for something like this. But then, who doesn't have 4 cores now, they are so damn cheap.

:sigh: I have 2 [editline]03:23PM[/editline] Mainly cause my mother board doesn't support more than 2 cores.

[QUOTE=IronPhoenix;23570700]Again, i have 4 cores, and my gfx card is only an 8800gt. I would have no use for something like this. But then, who doesn't have 4 cores now, they are so damn cheap.[/QUOTE] You do understand, that cpu and gpu dont battle on the same serie, what comes to computing power? (gpu has passed cpu on that ages ago.)

[QUOTE=IronPhoenix;23570700]Again, i have 4 cores, and my gfx card is only an 8800gt. I would have no use for something like this. But then, who doesn't have 4 cores now, they are so damn cheap.[/QUOTE] A GPU can beat a CPU 20-0 on not-heavy calculations that need to be iterated a huge number of times. In fact a GPU has less processing power than a CPU, but it can execute the same tasks around 500 times while the CPU does it once.

[QUOTE=Firegod522;23570742]:sigh: I have 2 [editline]03:23PM[/editline] Mainly cause my mother board doesn't support more than 2 cores.[/QUOTE] Don't worry, if I remember correctly two core processors are usually faster than 4 core ones and a lot of games only utilise 2 so you aren't missing out.

Also, valve generally know what they are doing. This gpu compiling will probably cause more problems than it solves.

[QUOTE=IronPhoenix;23572000]Also, valve generally know what they are doing. This gpu compiling will probably cause more problems than it solves.[/QUOTE] What the hell are you talking about. Why would it?

Some parts will be faster Like VRAD probbably would run alot faster on the GPU