Area Correlation Library performance test (arecorlib - p2)

research (correct, but but not optimized yet) code for area correlation in image processing, implemented on C++(with Qt4).
- int32 and double types used
- Qt4 types for images used

AMD K10 (2.8GHz) - 810-920ms
ARM Cortex -A8(550MHz) - 8919 - 15850ms

Unbiased render & GPU

it's time:

first was LuxRender

- 0.8 support a GPU (under Linux - AMD Radeon)

- ~2.5 times for AMD Athlon X4 (2.7GHz) vs NVidia GeForce GT460

Indigo since 3.0 also support GPU

- CPU: 103025.85289 samples / second (9.70630 micro-seconds / sample)

- CPU+GPU: 225543.72788 samples / second (4.43373 micro-seconds / sample)

also a bit more than 2 times speedup for such hardware

only Maxwell render not support GPU (yet, I guess)

QT4 on Linux/amd64 & Android/arm32

Ext4 under Windows

https://sourceforge.net/projects/ext2read/
(latest version already support Ext4 for read)
PS. issues with 1.5GB+ files

Mesa 7.3 RC1 @ N800

2.6fps in gears in full-screen.
w/o DSP optimization, but at least - it works on real device.

Nokia N8x0 and h/w OpenGL ES

OpenGL EX 1.x accelerator is not connected to screen due to very high screen resolution of screen (i.e. 800x480x16bit) - 768kB only for sigle framebuffer.
upcomming Maemo 5.0/new TI hardware will have a bigger memory for GL accelerator and will have native OGL ES support from scratch.

NEdit - a classical text editor for X11 is working as well as Lesstif - a free implementation of Motif.

and for Mesa - glut, glsl is working as well