Semantic
— checks
Ready, set,
compile.
Put your Rx compiler to the test.
Four stages. Let’s go.
— checks in allLLVM IR
— checks
Code generation
— checks
Optimization
— checks
Your current progress
Your runs 0
| Submission | Language | Verdict | Tests |
|---|
Hmm, nothing's here..
Sign in to start your first run.
Let it run.
Current test suite commit
Loading…
Sign in to connect
Loading your run…
YOUR COMPILER / IN THE QUEUE
Ready to roll.
A little patience. The open road is just ahead.
Waiting for a worker
A LITTLE FRIENDLY COMPETITION
Fewer cycles.
Fewer cycles.
More bragging rights.
Beat the Rust baseline. Climb the ranks.
The fast crowd 0
Sign in to see the Rust baseline.
| Rank | Account | Score | Best submission | Language | Achieved |
|---|
A place for your personal best.
Sign in to see the leaderboard.
ADMINISTRATION
Manage runs.
Find a run. Follow its progress. Keep the scores fair.
Sign in as an administrator to manage runs.
Student progress
Only passes on the current test commit count.
| Student | Semantic | IR | Codegen | Optimization | Score |
|---|
All runs 0
Voided runs keep their history and do not count toward optimization scores.
| Run | Account | Suite / language | Status | Passed | Test commit | Actions |
|---|
No matching runs.
Try another search or clear the filters.
ABOUT THE JUDGE
Under the hood.
From your first build to your best score.
How scoring works
Your score is the geometric mean of the Rust baseline cycles divided by your cycles for each optimization testcase, with cycles summed across all inputs. The baseline uses the lowest correct total across Rust optimization levels 0, 1, 2, 3, s, and z. 1.0 matches Rust; scores above 1.0 beat the baseline. Your best fully accepted optimization run on the current test catalog counts; Voided runs are excluded.
Dependency preparation
After fetching your repository, the judge detects the selected language’s
dependency manifest and prepares packages before running BUILD.
Compiler builds and all tests run without network access.
The commands below run automatically inside the judge. Downloads only allow from crates.io, PyPI, npm, or Maven Central.
Rust
Rust 1.95 · CargoCommit Cargo.toml and Cargo.lock.
Fetch
cargo fetch --locked
Cargo uses the prepared cache with CARGO_NET_OFFLINE=true
during builds. Your BUILD command still needs to compile
your project, for example with cargo build --release --locked.
Python
Python 3.13 · pip · venv
Commit requirements.txt with pinned versions and hashes
for every direct and transitive dependency. Packages must have wheels
compatible with the judge’s Python version and platform.
Fetch
python3 -m pip download \
--require-hashes --only-binary=:all: \
--dest /workspace/.rx-oj-deps/wheels \
-r requirements.txtInstall offline, before BUILD
python3 -m venv /workspace/.venv
/workspace/.venv/bin/python -m pip install \
--no-index --find-links /workspace/.rx-oj-deps/wheels \
--require-hashes --only-binary=:all: \
-r requirements.txtBuild and testcase commands automatically use this virtual environment.
JavaScript
Node.js 22 · npmCommit package.json and package-lock.json.
Fetch
npm ci --no-audit --no-fund --ignore-scriptsInstall offline, before BUILD
npm ci --no-audit --no-fund --offlineJava
OpenJDK 21 · MavenCommit pom.xml and pin dependency and plugin versions.
Fetch
mvn -B -ntp -s /opt/rx-oj/maven-settings.xml \
org.apache.maven.plugins:maven-dependency-plugin:3.8.1:go-offline
Builds inherit
MAVEN_ARGS="-o -s /opt/rx-oj/maven-settings.xml" to stay
offline. Your BUILD command can run mvn package.
C / C++
GCC 14 · antlr4-(cpp)-runtimeThere is no automatic dependency fetch command.
Projects in a subdirectory
Manifests are detected at the repository root by default. To prepare
dependencies elsewhere, commit .rx-oj.toml at the root:
[dependencies]
directory = "crates/compiler"
Without a manifest, the judge skips preparation. For a fully vendored project, use
enabled = false in the [dependencies] section.