Saturday, July 12, 2008

well sexy toyes...

http://www.adorama.com/SGDP1.html

i really considering to get this to have a small portable camera for trips and just daily picture taking. I see so often beautiful landscapes or nice street scenes and never have a camera with me. Now you could argue you could just buy a cheap compact camera for like 150 - 200$ (cheap is relative), but there sensors are just so small and noisy and this camera is supposed to have a very nice large sensor.

well I think about it and wont buy it before adobe releases a lightroom plugin to handle the raw data of this camera.

...yes I know it got raving reviews for the picture and awfully reviews for the menu handling. And you need to go for everything into the menu. Well I guess I wait for a nikon or canon model with the same nice chip and just buy a small 35mm/f2 lens for the trips and maybe take the 90mm with me...

but the 35mm F/2 cost 300$ and the camera cost 700 - 300 == effective 400...

what do I want?

  • aps sized sensor - sigma got it
  • compact format - sigma got it
  • manual control with dedicated wheels, buttons - sigma lacks this
  • F/2.8 if possible since F/4 is kinda slow - sigma lacks in this
  • 28 or 35mm, both is fine, interchangeable would be perfect, sigma got parts of it
  • around 500$ - sigma cost 300$ more
  • fast write speed - sigma really lacks in this
basically a digital rangefinder like a leica M8, just affordable and not 8k$.

drink of the day,

http://www.drinksmixer.com/drink10428.html

bermuda mai tai recipe. Just weird to drink mai tai without amaretto...

pycuda and me,

well the last two weeks I'm sitting here and learning python and C again to utilize the calculation power of cuda and current nvidia gpus for calculations. I don't have any real use for this yet, but it's an interesting idea and maybe we can work it into something.

So far I rewrote the complete python math module to allow the execution of the standard math functions on a gpu instead of the gpu and the results are pretty impressive.

For example to calculate the acos function on 524k floats is 1000x faster than to calculate it on the gpu. Now I have no idea were I could use this, but hey it's nice to have.

I also created my own fork of pycuda under:

http://github.com/berlinguyinca/pycuda/tree/master

to speed up the development, since the original developer won't have time to merge my changes with the main repository, at least not all the time.
Benefit is I learn another source code management system. And git is defnitly the nices system I used sofar, but svn with still much simpler and has the benefit of more available tools. Specially I miss a git fisheye integration.

Now I just have to figure out why my ohloh ranking keeps dropping and does not increase over the time...

I also was not able to get an iphone yet, but atleast it will get cheaper once the demand settles and hopefully they improve it even more. I also don't know If I really need it...