Official CUDA disasembler for sm_1.x real ISA

Yesturday NVIDIA released an official disasembler for sm_1.x (pre-Fermi) real hardware ISA. It's like an official version of DECUDA :-) (that Wladimir stopped to develop)
It takes either an ELF CUDA binary, a cubin or even an exe file, and provides the low level assembly code of the CUDA kernels.
It is only available for registered developer for now, but you can get a little more information the CUDA forum.

That's something a lot of developers have been asking for for a while. That allows to see the impact of optimizations on the real microcode, and it is particularly important for register usage for instance (since registers allocations is done after the PTX level).
Nice NVIDIA finally end up unveiling it's real hardware ISA instructions. AMD is still a little bit ahead on this since the ISA instructions and microcode is available even for the Evergreen architecture (RV870):

What exactly is a doctorate @Gizmodo

What is EXACTLY a PhD  ? The answer is there :

"Keep pushing" !

OpenGL 4.1 review @g-truc + Samples pack

OpenGL 4.1 Review:

Christophe also published an updated version of his sample pack, with a nice table showing how the new stuff works on both NVIDIA and ATI (Beta drivers):

UPDATE: Another good article about OpenGL 4.1:

Hacking D3D virtual calls + v-table hacking made useful @Humus

Direct3D API is a fully C++ object-oriented API and rely on runtime polymorphisms (virtual fonction calls) to be expendable and easily being able to provide different implementations. So all API calls are virtual calls instead of being plain C calls like in OpenGL.
Every slightly experimented C++ developer knows that virtual functions calls introduce overhead and that they should be avoided inside inner loops. Humus shows how these virtual calls can be replaced by standard calls by hacking the API objects v-table in order to keep a plain C pointer on these virtual methods !

I love this kind of hack ! But as Humus explains, D3D (like OpenGL since OpenGL 3.0) do not rely on immediate mode anymore, that means that API calls are usually consuming a slightly large amount of cycle compared to the overhead of a virtual call.
That means that in practice you wont get significant performance gain from this hack, but this is just really cool :-D And this method could still be useful to overcome performance problems in more badly design APIs !

UPDATE : D3D v-table hacking... made useful !
Humus just published another trick that shows how hacking the v-table of the D3D context can be used to... replace the default API calls by your own enhanced calls !
Humus shows how this can be useful to count the number of time an API function is called for instance. This can be done by overwriting the original object v-table pointer with the address of your own v-table. More details there :

Copyright © Icare3D Blog
Designed by Templates Next | Converted into Blogger Templates by Theme Craft