So, uh, what’s the catch with part 2? I just used my part 1 solution, and it worked.
xfix
bjorng
Erlang Core Team
qhwa
Here’s my solution powered by the revised homemade intcode computer. 
It took 570ms to finish part2.
It cost me ten minutes or so to debug. What surprised me was that it was a REAL diagnose program, pointing out that I had something wrong with executing opcode 3. It was true, I did not support mode 2 for reading at that moment. 
What an ingenious puzzle!
1
xfix
Interestingly, my solution is using :array and runs relatively quickly.
$ mix test test/day9_test.exs
..
Finished in 0.4 seconds
2 doctests, 0 failures
0
cblavier
It’s working but it’s really slow : https://github.com/cblavier/advent/blob/master/lib/2019/day5/day5.ex
Any idea of what I should improve in my code?
(my guess is that I’m using a lot of binaries instead of integers)
0
qhwa
I had a glance, it seems that List.replace_at/3 is used in write_memory. This is quite slow. If a list is under random access and update frequently, Map is preferred.
1
alvises
Just finished mine. As you guys said, the refactoring made to the VM in day5/7 definitively paid off!
Follow the @qhwa suggestion. Use a Map to represent the memory of the VM. Keys are addresses and values are intcodes.
After seeing the code of the people here, I’ve decided to just use a struct with memory, ip, relative_base, input, output, etc. keys. In this way each function just takes a vm (with some other arguments) and returns the updated VM. This makes easier to use pipes in the code.
1
alvises
Is there a benchmarking tool that benchmarks the code speed giving a score that’s independent from the machine’s speed?
0
sasajuric
Author of Elixir In Action
Not that I’m aware of. You’d need to run both versions on the same machine.
As an aside, I suspect that the slowness of my solution is due to the usage of streams. Moving to plain recursion might help, but I don’t even want to bother to check this because the execution time is ATM good enough. If at some later challenge it starts causing problems, I’ll see about optimizing it 
2
