patrulek

2 cycles vs ~5 cycles is still nice improvement plus its more deterministic (but not that fast in optimistic case; better in pessimistic case) than `if` appraoch and probably easier to "invent" and reason about. Definitely not that cool as your `if` solution though :)

Yes you need, but the initial problem here were sequential loads from L1 into registers.

You can utilize MLP (memory-level parallelism) and reduce memory latency by reading whole 8 bytes of next_j at every iteration. Just do something like

`uint64_t packed_next_j = *(uint64_t*)(&next_j[i])`

and then just shift right to find proper j. This way you remove dependency on previous iteration. Did you tried that?

Its one of the options. Other is sniping new pools and copy trading. Im trying to develop zero/same-slot tx landing with shredstreams first, because its basically needed to be competitive.

I would say that in most cases you dont need fancy algorithms and data structures, you just need to stop treating RAM as a rough notebook. Allocating in a loop, not reusing buffers and pointer chasing are a performance killers and are quite easy to avoid before they are introduced to code. Other than that, just be more thoughtful about network and disk usage and nested loops. Its a sum of stupid things that make your program slow.

Cooking another crypto bot on Solana, as with constantly changing market conditions, my previous bot basically stopped earning anything.

> For example I have decades of Software experience, but wont know where to even begin what OP did.

Simple, you start by asking LLM for more info.

It isnt programmers fault though.

> mistakes as the programmers: taking 'rules' literally

Isnt it how we make stable, deterministic and predictable system? How do you want to create one with ambiguous rules?

Software engineering is a domain on its own, just a technical, not a business one. Good luck with looking for a once in a million data race or deadlock bugs. Good luck with synchronizing system distributed over a whole globe. And good luck with keeping your domain knowledge up-to-date while developing and maintaining your vibecoded system. Assuming you can even provide enough details and deterministic specification to AI agent, because understanding business and having knowledge is not equal to being able to write down specific and concise rules to make a system out of it.

And if you think LLM are (or will be) good enough to not care about software part, what makes you think that your domain will not be completely resolved by AI?

This is why im using ramdisk for a browser cache.

To be honest, you can feel code is too complex even without reading a single line of it.

So its true what they were saying about bitcoin. It is used by criminals.

You dont need to measure the code complexity, you can completely hallucinate those numbers if you feel that code is too complex and then watch how LLM will respond.

But it would be also possible that sentinel value used for comparison changed because of bitflip, not data structure used by program.