The Prospero Challenge • Hacker News Board

Re: The Prospero Challenge

#21

Post by ayastrebov2 » Mon, Mar 24, 2025, 10:55 PM UTC

kevmo314 wrote: post #20

montag wrote: post #19

Earlier quoted context omitted.

Impressive. Did you do all this in the past 2 hours?

Yeah, it took me an hour-ish to get the performance and then an hour-ish to figure out how to actually render it correctly. I had never seen the P2/P5 PGM format before so I didn't realize P5 files were binary. :)

Great idea and clean implementation! You seem to be interested in performance engineering - I did some work recently, check this out https://github.com/AlexanderYastrebov/wireguard-vanity-key

Re: The Prospero Challenge

#22

Post by tonetheman » Mon, Mar 24, 2025, 11:31 PM UTC

[dead]

Re: The Prospero Challenge

#23

Post by Jtsummers » Tue, Mar 25, 2025, 12:03 AM UTC

adrian17 wrote: post #15

> (Spoilers: it also allocates 60+ GB of RAM for intermediate results, so consider reducing the image size before running it on your own machine) The input algorithm is essentially written in SSA form, and it's easy to analyze when each "register" stops being used; then you can drop the arrays after their last use. Turns out the number of live registers never exceeds 160 at any point in time, and with this one additi…

I don't know why you are downvoted (at the time of this writing). That's a very good point and makes the simple Python solution feasible on many more computers (due to reduced memory footprint). It's also a straightforward pass (once you know what liveness analysis is). Since there are no loops it's just one pass starting from the end and working backwards to determine whether a variable is live or not, and then an extra check on each iteration to decide whether to delete particular arrays.

Re: The Prospero Challenge

#24

Post by tekknolagi » Tue, Mar 25, 2025, 12:15 AM UTC

Jtsummers wrote: post #23

adrian17 wrote: post #15

> (Spoilers: it also allocates 60+ GB of RAM for intermediate results, so consider reducing the image size before running it on your own machine) The input algorithm is essentially written in SSA form, and it's easy to analyze when each "register" stops being used; then you can drop the arrays after their last use. Turns out the number of live registers never exceeds 160 at any point in time, and with this one additi…

I don't know why you are downvoted (at the time of this writing). That's a very good point and makes the simple Python solution feasible on many more computers (due to reduced memory footprint). It's also a straightforward pass (once you know what liveness analysis is). Since there are no loops it's just one pass starting from the end and working backwards to determine whether a variable is live or not, and then an e…

I think getting downvoted (not by me) because if you scroll to the end of the post, that's the very first submission (by me).

Re: The Prospero Challenge

#25

Post by Jtsummers » Tue, Mar 25, 2025, 12:22 AM UTC

tekknolagi wrote: post #24

Jtsummers wrote: post #23

Earlier quoted context omitted.

I don't know why you are downvoted (at the time of this writing). That's a very good point and makes the simple Python solution feasible on many more computers (due to reduced memory footprint). It's also a straightforward pass (once you know what liveness analysis is). Since there are no loops it's just one pass starting from the end and working backwards to determine whether a variable is live or not, and then an e…

I think getting downvoted (not by me) because if you scroll to the end of the post, that's the very first submission (by me).

Ah, fair. I didn't catch that.

Re: The Prospero Challenge

#26

Post by crazygringo » Tue, Mar 25, 2025, 12:23 AM UTC

IshKebab wrote: post #11

Hash the input, if it matches the expected input output the expected image. Otherwise error. Technically correct.

Why waste time hashing?

Just output the expected image in all cases.

Still technically correct. ;)

Re: The Prospero Challenge

#27

Post by edflsafoiewq » Tue, Mar 25, 2025, 12:40 AM UTC

Converted it to GLSL that can be dropped into shadertoy.com/new. I get about 15FPS at 714x401.

https://pastebin.com/jBwNrfim

Re: The Prospero Challenge

#28

Post by Titan2189 » Tue, Mar 25, 2025, 12:53 AM UTC

cfbolztereick wrote: post #2

This is really fun, I'm getting badly nerd sniped by this.

It's been 4 hours already. How are you doing? Didn't forget to eat and drink? Just checking in on you :)

Re: The Prospero Challenge

#29

Post by James_K » Tue, Mar 25, 2025, 1:27 AM UTC

Some of these solutions seem to copy the .vm file and rewrite it in another programming language. Is this permitted? Seems like cheating to not count the time spent compiling and transforming that code in the runtime of the program.

Re: The Prospero Challenge

#30

Post by edflsafoiewq » Tue, Mar 25, 2025, 1:34 AM UTC

zomglings wrote: post #5

Can you explain how you produce the opcodes from a text/image?

Looks SDF-ish. My guess is convert the letters to curves, cut them into pieces without holes so each piece is the set of points between certain curves, write down SDFs for each curve (negative on the "inside", positive on the "outside"), combine all SDFs in a piece with intersection (max), then combine all pieces with union (min).