There is an increasing noise about GPGPU computing and how much faster than CPU (even parallel) it is. If you didn’t hear about all that, GPGPU is about using the computer’s graphics card(s) to do general purpose computations. The key to the performance lies in the parallel architecture of these devices. From what I read, an average graphics card has 64 parallel units, but they are not as versatile as the CPU of which a typical PC these days has 4 cores. That means, if the task is well suited, it can boost performance significantly, but if not, it’s nothing more than a lot of wasted work.
So I wanted to see for myself. To get started I read the book “OpenCL Programming Guide“. It gave a good overview. But now it was time to give it a try.
Continue reading “OpenCL First Steps”
Some linux distributions defaulted to use the –as-needed linker flag for a while. Ubuntu tried it in natty, but then reverted. Now with oneiric, it really is enabled by default.
I ran into this when one of my packages wouldn’t compile on oneiric. I always got linker errors with boost::filesystem and boost::system. Between natty and oneiric, the default version of the packaged libboost changed from 1.42 to 1.46, thus switching from filesystem v2 to v3. Obviously my first thought was that it must have to do with that. Also libwt which I use in the project had the same error in a previous version. So, I reduced my app until I was sure that couldn’t be the cause here. Also all my other packages didn’t have any problems with the transition to filesystem v3 apart from the regular changes for adapting to the new interfaces. But for these changes the compiler helps.
The project in question is organised more or less like that:
Continue reading “Mixing boost versions –as-needed”
Before I discovered what my Bifferboard really is, I almost disposed it, but now It found a new purpose. It’s a networked rfid Terminal for time tracking on our BORM ERP. I use a simple python script on the device because it’s easier to experiment on a device where I would rather not compile too much every time trying something. In fact, this is my first python project appart from looking through some scripts and changing a few lines here and there.
Google employees can spend 20% of their working time for their own projects. We at cubx (The CAD development department recently split from BORM) have now something similar. We get to spend every second friday afternoon for some projects of our own. The only restriction is that it has to do with computer graphics.
I chose a project that I had in mind for almost as long as I have been working for BORM. I wanted to run PointLine as a web service on a linux box. Wit the old core, that was so tightly tied to MFC, this would have been impossible. But the new unfinished core is designed to be platform independent. So, I took it as a base.
Some three years ago, I looked at vrml and x3d for the 3d in browser part. These standards didn’t have as broad support as I thought. You couldn’t do much without buying proprietary browser plugins. So I never made much progress on that. But then came WebGL. All mayor Browsers support it natively, and on top of that, my favourite web application framework recently got a WebGl widget.
So far, my research prototype doesn’t do much, but I already learned a lot that I can also use in my everyday work. I gained a better understanding of rendering pipelines, learned about shaders and ray picking …
As a side project, I set up a jenkins continous integration server that compiles the project on ubuntu and windows after every commit. Additionally, I learned about CPack, so that the result of the jenkins build is a binary deb package ready to install.
You can see the web application as it progresses here:
Usually, I would also provide a link to the sources, but this is closed source, sorry…
Last year, I spent some time rewriting internals of how groups are handled in our CAD software. Before that, groups were just to help with the selection of elements that belong together. The grouping was only one level deep, and the elements were organized flat on the document. All the calling functions had to uptate the relationships in the pointers and containers which were publicly available. This was a constant source of bugs and completely unsuitable for nested symbols that we wanted to import.
Continue reading “Propagate deletion of C++ Objects”