Add doom showcase

[stephenctw]

• code cleanup
• download machine templates from git release (should we use IPFS instead?)
• support DEPLOY_TO_ANVIL option through env for lua scripts
• support CHAIN_ID, PRIV_KEY, ROOT_TOURNAMENT, URL options through env for dave-compute

[stephenctw]

@GCdePaula @guidanoli could you please try to run the doom-showcase locally? Let me know what you think. Also, any suggestions for testnet deployment?

[guidanoli]

• [x] Build test image
• [x] Run simple test
• [ ] Run stress test :question:
• [x] Run Doom showcase

I don't know if the stress test was successful or not. It just prints out Hello from Dave lua prototype! and exits with status code 137, which tells me it might not have been successful?

Also, I felt the timeout quite dreadful. Would it be possible to reduce it just for the purposes of running the tests locally and not having to wait such a long time? I might not be aware of any undesired side effects of such a change, so please tell me if there is a reason why we shouldn't mess with the timeout just for testing purposes.

[stephenctw]

@guidanoli It's definitely not successful with exit code 137. Is there any more information from the console logs? Also, which version did this happen? The pure lua-node version or the offchain + lua-node combination? For the timeouts, I think most of the time are spent on the machine computation. But the timing constants are set based on my machine performance, it may vary from yours. I'll try to update a measurement script that @GCdePaula wrote and you can try to run it to generate different timing constants for your machine.

[stephenctw]

@guidanoli now you can go to permissionless-arbitration/measure_constants directory and run the script to generate timing constants that fit better to your machine.

[guidanoli]

Is there any more information from the console logs?

No, just the line Hello from Dave lua prototype!.

which version did this happen?

I just ran the command that was on the README under permissionless-arbitration/lua_node:

docker run --rm \
    --env MACHINE_PATH="/app/lua_node/program/debootstrap-machine-sparsed" \
    --env DEPLOY_TO_ANVIL="true" \
    dave:latest

[guidanoli]

now you can go to permissionless-arbitration/measure_constants directory and run the script to generate better fitting timing constants for your machine.

Thanks! Tomorrow I will try it out.

[guidanoli]

• [x] Build test image
• [x] Run simple test
• [x] Run stress test
• [x] Run Doom showcase

@stephenctw and I hopped on a call to investigate the issue with the stress test not working, and it seems that the issue was the memory limit set by Docker. I set the memory limit to 3.8 GB, and swap to 2 GB, and it worked like a charm. Thanks, Stephen!

[stephenctw]

@guidanoli I add the graphic feature to the doom showcase. I only add this to the Lua prototype as it can reduce performance of the Dave process. Please let me know if I should also add this to the Rust Dave (production code).

I did this in a hacky way, suggestions are very welcomed.

• Dave Lua node snapshot the machine at the beginning of every level of tournament, also store extra information about the current commitment tree.
• A separate script monitors the snapshot directory constantly. When a new snapshot is taken, it implies the node has reached a new level of tournament. Therefore, the script loads the snapshot, run the machine for the given range of cycles (calculated from the extra information in the snapshot), and then read the game frames, process them and save to png files.
• The user runs another script on his host machine (./doom-showcase/view.sh) that will play the frames repeatedly as the png files are being generated. The game play starts fragmental and stops when the full game play is showed (of the current level).
• In reality, you'll see the same video throughout the four levels of tournaments. That's because the game play is short, and the logic of our dishonest player is designed to be simple. The dispute always goes on to the left most leaf, and thus the game play always starts from the beginning.
• In addition, enabling game graphics can greatly impact the performance. Because it will be doubling the amount of cartesi machine running in the process, I'm not sure if there's a way to avoid this. If Dave process gets timeout, you can try to raise the COMMITMENT_EFFORT from CanonicalConstants.sol by a few minutes depending how much slower you get.

  [#2][04/30/2024 18:30:11] join tournament reverted: ./lua_node/blockchain/sender.lua:76: Send transaction `joinTournament(bytes32,bytes32[],bytes32,bytes32)` reverted:
  Error: 
  (code: 3, message: execution reverted: tournament check-in elapsed, data: Some(String("0x08c379a00000000000000000000000000000000000000000000000000000000000000020000000000000000000000000000000000000000000000000000000000000001b746f75726e616d656e7420636865636b2d696e20656c61707365640000000000")))

[guidanoli]

In my case, I had to create the snapshot folder manually. Maybe worth putting instructions on the README to do so as well.

mkdir snapshots

[guidanoli]

Managed to run it! The color palette is indeed a bit glitchy. Some white and blue pixels seem out-of-place.

Screencast from 01-05-2024 09:25:36.webm

[guidanoli]

Maybe worth trying to extract the palette from the machine as feeding it into the playpal2rgb tool (which I updated to accept such palette as input rather than having it hard-coded).

[stephenctw]

Maybe worth trying to extract the palette from the machine as feeding it into the playpal2rgb tool (which I updated to accept such palette as input rather than having it hard-coded).

Sure, once Eduardo gave me the range of memory region, I'll update it.

[stephenctw]

Rebased the branch on top of main, updated all the examples to use machine-emulator 0.17.