Fork me on GitHub
#off-topic
<
2018-04-23
>
ackerleytng11:04:51

tried the above in wolframalpha. at some value of k, say 50, there will always be some value of n where the inequality is true. when n gets bigger, the inequality is no longer true

ackerleytng11:04:30

I don't understand how log (n) = O(n^c) in this case. anyone understands this?

dpsutton12:04:28

did you read the answer to the question above? there's some confusion of terminology. log(n) is-element-of O(n^c)

dpsutton12:04:26

just as n^2 is an element of O(n^4)

ackerleytng12:04:03

yup, but still, for a value of k, say 50, once n is 10^16, the inequality fails, so for k = 50, I can't find a n_0 such that for n>=n_0, log n <= k n^c

dpsutton12:04:55

so lets ignore scaling and maybe drop the exponent a little bit

(let [n 10e15]
  (prn (Math/log n))
  (prn (Math/pow n 0.1)))
36.841361487904734
39.810717055349734

dpsutton12:04:33

with just an exponent of 0.1 it takes to about 10^15. so for your example here with the escaling and the exponent, maybe try 10^1000000000

dpsutton12:04:56

if you change the exponent to just 0.01 you can see how much you chop down that number and consequentially push out that changeover spot n_0

dpsutton12:04:10

or take a more convenient k, say 400

ackerleytng12:04:45

so i just got tricked by the plots

dpsutton12:04:59

although in clojure the 10^16 works for me:

(let [n 10e16]
  (prn (Math/log n))
  (prn (* 50 (Math/pow n 0.0000001))))
39.14394658089878
50.00019572011597

dpsutton12:04:09

(i'm using base 10 for convenience)

ackerleytng12:04:07

if i twiddle the constants a bit, the lines actually intersect twice

ackerleytng13:04:19

was seeing this and never thought it was possible that it would intersect again

Functional-Tom13:04:24

Trying to find the best place to post a blog

Functional-Tom13:04:50

it's clojure related, and wondered whether this channel is the correct place

Functional-Tom13:04:55

any help would be much appreciated

yogidevbear14:04:03

Hi @functionaltom, you could post into #news-and-articles

yogidevbear14:04:32

Ah, I see someone beat me to it

Functional-Tom14:04:57

thanks for the heads up though!

danielstockton13:04:17

It's easier to ask forgiveness rather than get permission 😛

danielstockton13:04:32

If it's Clojure related, it doesn't seem offtopic. #clojure ?

vemv14:04:44

@functionaltom #news-and-articles

thumbsup_all 4
qqq17:04:56

In WebGL, is there a way to query the runtime for "maximum # of vertex buffer objects" ? I'm not asking for the largest # of floats a single VBO can have; I want to know how many VBOs total I can have.

tbaldridge18:04:01

is there a limit?

qqq19:04:04

According to some email list, the limit on buffer objects is the lower of: 1. 2^32-1 2. when you run out of GPU memory

tbaldridge19:04:25

which really means #2, since if you assume floats or ints for verts, and you assume one triangle per VBO (don't do that, it's a horrible idea), you'd need 206GB of GPU memory to handle it all

tbaldridge19:04:46

eh, 103, since it's ^32-1

qqq19:04:55

is there a way to sorta do GPGPU in WebGL? I need a way to setup a shader where input vbo, output vbo OR input texture, output texture and just run the shader (everything stays in GPU memory)

tbaldridge20:04:27

You want to read up on render to texture options in webgl

qqq20:04:41

@tbaldridge: so the high level idea is to setup the computation in 'stages', where each 'stage' is rendering a scene to texture ?

tbaldridge21:04:48

@qqq pretty much textures look a lot like arrays inside the shader, and you simply read from one to output the pixel of the other

tbaldridge21:04:31

turning off stuff like 3d projections and the like so you can get a 1:1 mapping from pixels in the input/output to the array locations in the textures

benzap22:04:38

There's a WebCL spec, but none of the major browsers support it, which is a shame

benzap22:04:44

I wouldn't be surprised if they start implementing it once later wasm specs expose the DOM

benzap22:04:43

There's going to come a point in a decade or two, where everyone just uses thin clients to talk to the "OS"

benzap22:04:06

no need for hands anymore! Plug in your brain interface! beep boop!