Skip to content

README.md

Latest commit

 

History

History
59 lines (43 loc) · 3.18 KB

README.md

File metadata and controls

59 lines (43 loc) · 3.18 KB
Raw

bots

简体中文

Bot sources and test maps.

path what it is
abyss/ the final bot, exactly as submitted (C++, compiled to WASM by unswbc)
fixtures/ two tiny test bots for the worst-case CPU census; not competition bots
maps/unseen50/ 50 generated maps that no bot was tuned on, for robustness checks

The official contest maps and the template starter bot are not part of this repository. They ship with the unswbc toolkit (MIT licence, by UNSW CPMSoc), and the commands below create them.

Setting up a workspace

The quickest way is the repository's own bots/ folder (see the root README's quick start; the root .gitignore covers what unswbc writes there). A separate workspace works too. Install the toolkit, then make a workspace folder outside this repository:

uv tool install unswbc==1.2.9     # or, on Python >= 3.11: pip install unswbc==1.2.9 (the version the final bot was built and verified with)
mkdir abyss-ws && cd abyss-ws
unswbc maps maps                  # copies the bundled official maps into ./maps (existing files are kept)
unswbc init cpp starter           # creates the template bot ./starter (the usual baseline opponent)
cp -R path/to/repo/bots/abyss ./abyss
cp -R path/to/repo/bots/maps/unseen50 ./maps/unseen50

Then:

unswbc run maps/arena.map abyss starter                       # one native game, replay in replays/
unswbc run maps/arena.map abyss starter --sandbox --no-replay # as the judge runs it: WASM, CPU metered in points

Replays open in the VS Code viewer (unswbc vscode) or at https://game.battlecode.au.

A separate workspace should live outside the repository. unswbc run writes replays/, and the build caches live in ${XDG_CACHE_HOME:-$HOME/.cache}/unswbc/.

The final bot was built and verified with unswbc 1.2.9. To reproduce the exact submitted WASM (sha256 0f39d15f10a503b5d5602144660b8c3d3e62cae53ed1fa94e792c88e395e3b0e), use that version. See abyss/README.md.

Fixtures

fixtures/ holds two minimal bots used to test the worst-case CPU census (tools/abyss_tools/run/cpu_census.py). The same two bots also ship with that tool, in tools/abyss_tools/run/census_bots/. They do not play the game.

bot behaviour
fixtures/flooder every dragon casts four sonar rays a turn with random 64-bit values and circles a 2x2 square. The bot under test therefore reads a full inbox every turn. It stops acting at round CENSUS_END (default 60), so games end by elimination.
fixtures/cpuhog burns a known number of CPU points per turn: 50M normally, 95.5M plus a fallback marker (LOG fb95 selftest) on rounds 5, 15 and 25 (round % 10 == 5), and 101M once (round 7, id 1), which the sandbox kills. A census must report exactly these costs. It stops acting at round CENSUS_END (default 30).

Both are plain single-file C++ (main.cpp + bot.toml), with no helper.hpp and no dependencies. Run them like any bot, for example unswbc run maps/arena.map flooder cpuhog --sandbox --no-replay after copying them into a workspace.