<bonda_000>
a space I think in mathematics is something that has a "metric" on it, that is, where we can find a distance between any two elements
<bonda_000>
bringing back to your comment about pythagora's theorem yesterday, which is pretty much the metric on two dimensional x,y space
<bonda_000>
a "distance" is also some form of similarity here
<bonda_000>
measurement of closeness or similarity
<bonda_000>
but if I remember correctly you can alter the base vectors that span the entire space
<bonda_000>
for instance, (0,0,1), (0,1,0), (1,0,0) span the entire x,y,z 3D space
<bonda_000>
and so do (1,2,3), (2,5,7) and (1,3,5) can also be used as base vectors
<bonda_000>
and then there are appropriate transformations for the coordinates if we want to go from one base to another
<bonda_000>
it's interesting that in case of interpolating red pixel value, the base that we have (0.48, 0.36, 0.16) and we cannot have a base such as (0.5, 0.25, 0.25) that would help us shape binary number that we want
<bonda_000>
or maybe there is a more desirable base
<bonda_000>
clever: my point is, if you view the typical binary situation like this, you can apply all or most of the linear algebra to what seems to be a discrete math problem
<bonda_000>
so in my first example when I was going from Rbig to Rapproximate by taking first 4 bits of the heaviest component, and first 2 bits of each lighter components
<bonda_000>
clever: so this transformation from 24-bit Rbig to 8-bit approximation can be considered a linear mapping from f: R3 -> R1 or some sort of a metric
<bonda_000>
much like three-dimensional pythagorean theorem d = sqrt(x^2 + y^2 + z^2)
<bonda_000>
my bad
<bonda_000>
d = sqrt((x1-x2)^2 + (y1-y2)^2 + (z1-z2)^2)
<bonda_000>
but this classic metric requires two points and we only seem to have a single point in our space
<bonda_000>
clever: but you see the math is entirely different here and usual math goes out the window because by the dot product rule
<bonda_000>
0.48 dot 0.36 = 0
<bonda_000>
as they are orthogonal
bonda_000 has quit [Ping timeout: 264 seconds]
bonda_000 has joined ##raspberrypi-internals
bonda_000 has quit [Quit: Leaving]
bonda_000 has joined ##raspberrypi-internals
<f_ridge>
<clever___/D> @x2x6_
<f_ridge>
<clever___/D> > Pi 0-3 can't drive UHS-I SD bus speeds (DDR50/SDR50/SDR104) which A1/A2 are qualified for - so yes, you're paying a premium for performance you can't use.
<bonda_000>
I'm gonna try and debayer a picture using this method but so far I only was able to put to use 2 out of 3 values
<bonda_000>
this is just a theory so far, I will need to make a circuit to see if the result is any good
<bonda_000>
on paper I tried two examples and the result was pretty darn close
<bonda_000>
clever: it's like if I took your algebra skill from school and said - do something with those binary numbers. Not a big loss - bitcode isn't interpretable by humans and operations at atomic level don't mean much as there are only two possible values
<clever>
bonda_000: when i got bored in HS, i figured out how to do long division, in binary, lol
<bonda_000>
be back soon
bonda_000 has quit [Quit: Leaving]
srk has quit [Ping timeout: 268 seconds]
srk has joined ##raspberrypi-internals
<f_ridge>
<x2x6_/D> I've compiled linux kernel to trace all requests to sdcard. To see how they make it high speed. But now I need to somehow rechannel sdcard initialization for emmc controller instead of sdhost
<f_ridge>
<x2x6_/D> because by default sdcard initialization is done by sdhost-bcm2835 driver. Hm...
<f_ridge>
<clever___/D> i see a `dtoverlay=mmc`
<f_ridge>
<clever___/D> it points the mmc controller to gpio 48-53
<f_ridge>
<clever___/D> and disables sdhost
<f_ridge>
<clever___/D> looks perfect
<f_ridge>
<clever___/D> @x2x6_ you can also `dtoverlay=mmc,overclock_50=60`
<f_ridge>
<clever___/D> so when the linux mmc core asks for 50mhz, the driver gives it 60mhz instead
<f_ridge>
<clever___/D> which lets you overclock things
<f_ridge>
<x2x6_/D> where do you see dtoverlay=mmc?
<f_ridge>
<clever___/D> in the linux source, `arch/arm/boot/dts/overlays`
<f_ridge>
<x2x6_/D> do you mean that dtoverlay=mmc is the way to configure it in config.txt?
<f_ridge>
<clever___/D> yes
<f_ridge>
<x2x6_/D> ok, lets go
<f_ridge>
<clever___/D> just keep in mind, you wont have any wifi
<f_ridge>
<x2x6_/D> Oh, yes, there is something about ttyS0
<f_ridge>
<x2x6_/D> maybe this is getty
<f_ridge>
<clever___/D> ttyS0 has an unreliable baud rate
<f_ridge>
<x2x6_/D> autologin service over uart?
<f_ridge>
<clever___/D> add `dtoverlay=disable-bt` to the `config.txt`
<f_ridge>
<x2x6_/D> same
<f_ridge>
<clever___/D> what is the first thing linux prints during boot?
<f_ridge>
<x2x6_/D> actually this worked, at first it was all total crap, I switched from 9600 to 115200 (at some point I reverted to 9600 before) and it worked. Don't remember changing back to 115200 in cmdline.txt though
<f_ridge>
<x2x6_/D> now w are talking. .. probably. I'll now drop in couple of printk's into mmc-bcm2835.c
<f_ridge>
<x2x6_/D> Can someone give me a comment why this is so complex? whats going on? How the input data looks like is Bayer filter applied on some raw pixels each having its own color component and going in chess order or something like that?