Comments for Olhovsky

Comment on About by Eclectus

Eclectus — Thu, 02 Feb 2012 22:06:52 +0000

Hi Chris, I’d like to get in contact with you. I’m a fellow programmer within the games industry and I think we have have lots in common. Send me a mail?

Comment on Status update: Point light stress testing. by olhovsky

olhovsky — Mon, 23 Jan 2012 05:18:26 +0000

I am not targeting 30fps.

I am targeting either 60fps (16ms/frame) or 50fps (20ms/frame).

There is no rule written in stone that says that you have to pick between 30 and 60fps

Comment on Status update: Point light stress testing. by Matt

Matt — Thu, 15 Dec 2011 02:56:19 +0000

“And remember, we only have around 16-20ms per frame to do all updating.”

Are you targeting 30 or 60fps?

If you’re targeting 30fps, then you have 33mms for cpu and 33ms for gpu per frame. Using non-immediate presentation mode, the gpu and cpu will work in parallel, with the gpu rendering the previously submitted data as the cpu prepares the next frame.

If you’re targeting 60fps, then each will need to complete in 16ms (not 20).

This can be commonly misunderstood when going from PC with super fast gpu’s to the a console.

Comment on Shadow map filtering experimentation. by Xiaoxiong

Xiaoxiong — Sun, 04 Dec 2011 07:37:00 +0000

Could you show me the code you used for Edge Tap smoothing? I am working on smoothing edge after applying PCF

Comment on Cascading shadow mapping and ESM. by olhovsky

olhovsky — Wed, 19 Oct 2011 20:40:08 +0000

How much overlap you choose depends on your scene, the number of cascades you choose, and the split size of the cascades (different cascades may prefer different sized blend regions).

I don’t know of a simple formula for choosing the overlap, so I recommend that you experiment.

Larger blend regions result in lower performance, and less visible seams. As you can see from the following image, I chose a blend region that makes the seams relatively obvious (if you are far enough away and the scene is simple like this one is).
http://www.olhovsky.com/wp/wp-content/uploads/2011/07/july12_ESM_CSM_blending.png

As with everything in real time graphics, you have to tune the parameters to your application — trading visible artifacts for higher performance and better rendering quality in your other effects.

Comment on About by Naeb

Naeb — Sat, 24 Sep 2011 04:40:53 +0000

Hi,

I just recently visited your website, and I must say excellent blog on programming and not to mention your games.
I’m interested to know when did you first program? First 3d game? How long does it take to understand? What books/guides/forums are best to ask for guidance? Does your school stuff actually help in game design skills (your personal thoughts not a general answer) or was doing your own projects is what got you as far as you are. Whew! thats a lot of questions, but really that’s because I am quite impressed and inspired by your hard work and dedication.

Hope you reply,
Naeb.

Comment on Cascading shadow mapping and ESM. by CC Ricers

CC Ricers — Thu, 22 Sep 2011 17:36:28 +0000

How much overlap would you recommend for reducing seam artifacts? When I use CSM I use the a small amount, like a 0.1 – o.2% difference between the far plane of the closest cascade and the near plane of the one behind it. For example if the closest cascade has a far distance of 1000 units I would set the near plane of the next one at 998. I typically use 3 cascades so the shadow artifacts are already smaller.

Comment on Optimizing terrain culling. by olhovsky

olhovsky — Tue, 13 Sep 2011 21:04:17 +0000

200 tests is with a quadtree.

Comment on Exponential shadow map filtering (in log space). by Trosztel

Trosztel — Sun, 04 Sep 2011 11:29:05 +0000

Hi,
B2′s offset should be 1 instead of 0.

Comment on Optimizing terrain culling. by GeorgeR

GeorgeR — Tue, 30 Aug 2011 03:59:11 +0000

Good God – 200 tests? Use a quad tree!