Comments for RasterGrid Blog

Comment on Pocket Soccer’s story so far… by Claudio

Claudio — Mon, 16 Apr 2012 23:29:18 +0000

Hello,

Your game became a mass addiction between me and my co-workers here in Brazil. We are even creating tornaments with Around 20 people, awesome. However, not so many of us have android phones or tablets and many of us have IOs devides, this game would be perfect running on an IPad, are you considering developing for Apple as well? I think you should! Thanks.

Comment on Efficient Gaussian blur with linear sampling by Andrew Moffat

Andrew Moffat — Sat, 07 Apr 2012 20:56:08 +0000

Hi Daniel,

This article was incredibly clear and informative. Thank you for taking the time to put it together. I’m still learning this stuff, so this was very helpful.

I wrote up a small Python script based on this post to generate the gaussian texture lookups, optimized for linear sampling. Maybe it can be useful to someone else too:

https://gist.github.com/2332010

Thanks again

Comment on Just released my first Android game by Daniel Rákos

Daniel Rákos — Fri, 06 Apr 2012 22:23:30 +0000

I have no interest in supporting java mobile phones. Usually the phones supporting only java games are simply not powerful enough, besides that, these platforms tend to disappear from the market. The platforms of interest are Android, iOS and maybe Windows Phone and Blackberry, but besides that…

Comment on Just released my first Android game by StereoRepublic

StereoRepublic — Fri, 06 Apr 2012 21:38:38 +0000

Hi, I am just woundering, will you make support for java mobile phones?
game is greatest and i love it

Comment on Pocket Soccer’s story so far… by Nicsor Morales

Nicsor Morales — Tue, 03 Apr 2012 14:50:57 +0000

I love this game !!!! It’s very interesting the Tournament mode and the ranking. A great record for this gameplay. I will be waiting the World Class mode. Congratulations.

Comment on GPU based dynamic geometry LOD by Dmitry

Dmitry — Fri, 30 Mar 2012 08:36:57 +0000

Good day. Thank you for your articles, they are very interesting and help to understand the technology.
But I have questions.
I’m developing a project, where it is necessary to render vast areas with a large number of objects (grass, trees, shrubs, etc.).
The surface terrain is sufficiently large (in principle, it is generally not limited to), so I smashed it on the tiles. Each tile is a container of resources, including vegetation. Each tile and its resources are loaded as needed.
Now I have implemented an approach where each type of object (vegetation) correspond to a Instance object, and correspondingly its TBO. Initial tests showed that the overall performance rather depends on the number of Instance objects and drops sharply at a sufficiently large number of TBO. Then I decided to take advantage of your approach ICR, but the result I was very impressed (with the number of objects can be seen when more than 500k less than 20
I got the best performance), but when I added a test case for a few more Instance objects of another type, the performance dropped significantly (60fps -> 32fps) I thought that maybe it’s because of frequent switching feedback. Then I realized the dynamic LOD. As a result, I got a winning performance.
In the test case I used the model tree of your demo mountains, with the number of visible objects 3339, I only got 11fps. When two visible objects – 30fps. The total number of objects in the scene 5000.
Now I’m at a loss, what could I lose sight of the implementation. It can not be created for each object type a TBO? Or does not inevitably decline when using FeedbackBuffer? Or is the problem of using TBO?
Data for your demos
Nature 1.2 200-840fps
Nature 2.0 200-830fps
mountains:
No-culling 16-70fps
No-culling + dynamicLOD 33-250fps
ICR 20-1054fps
ICR + dynamicLOD 67-1050fps
ICR + Hi-Z 130-550fps
ICR + Hi-Z + dynamicLOD 250-400fps

Description of the computer:
CPU Intel Core2Duo E4400 – 2Gz
memory – 3Gb
GPU NVIDIA GTX 550 Ti
Windows XP sp3

Comment on Flexible static analysis for C++ code bases by Alex

Alex — Mon, 26 Mar 2012 20:42:40 +0000

You told a lot about code metrics, but what about actual issue detection? Were you able to detect any interesting issues in your projects?

Comment on Sad facts about OpenGL extension libraries by SS

SS — Fri, 16 Mar 2012 00:01:50 +0000

P.S. SFML is made of awesome. I also recommend SFGUI, which is an additional library that uses SFML and OpenGL (if I’m not mistaken) to give you the tools to write forms applications (sorta like .Net… except portable and less bloaty).

Comment on Sad facts about OpenGL extension libraries by SS

SS — Fri, 16 Mar 2012 00:00:19 +0000

I had the same problem, but eventually discovered (maybe this is different after all this time?) that if I included the GLEW paths in my project search directories, defined “GLEW_STATIC” in the #DEFINE project build options tab (I did mention I’m running Code::Blocks?), and linked the following (with the latest MinGW):
-l glew32s
-l glu32
Then the binaries work after all, despite everyone saying that MinGW has trouble with them precisely because they’re for VC++. (I did have a lot of trouble finding the right options to make this work, but somehow I did and it does. Maybe it’s sub-optimal, though; see note below.)

However, I will have to try the latest GLEE and compare 1) ease of installation/use and 2) speed of end result. I have an example that someone compiled with GLEE and I compiled with GLEW and their binary is faster, but they might’ve set higher optimisation flags than I or something.

Comment on Efficient Gaussian blur with linear sampling by djmj

djmj — Tue, 06 Mar 2012 22:35:46 +0000

You are looking up 10 and not 9 pixels! Coordinate 0.0 intersects two pixels and so you take full advantage of linear interpolation! Else you need to offset by half a pixel!

Where is the benefit of the usual easier way? Lookup at coordinates [-4, -2, 0, 2, 4] to get 10 pixel range and then apply the weighting of each two texels and calculate final value?