BLART is an implementation of an adaptive radix tree, used as backing for map and set data structures. Adaptive radix trees offer great space efficiency and performance on keys that decompose into byte strings.
Here is an example of using the TreeMap type (blatantly stolen from the standard library):
use blart::TreeMap;
// type inference lets us omit an explicit type signature (which
// would be `TreeMap<&str, &str>` in this example).
let mut movie_reviews: TreeMap<_, _> = TreeMap::new();
// review some movies.
let _ = movie_reviews.try_insert("Office Space", "Deals with real issues in the workplace.").unwrap();
let _ = movie_reviews.try_insert("Pulp Fiction", "Masterpiece.").unwrap();
let _ = movie_reviews.try_insert("The Godfather", "Very enjoyable.").unwrap();
let _ = movie_reviews.try_insert("The Blues Brothers", "Eye lyked it a lot.").unwrap();
// check for a specific one.
if !movie_reviews.contains_key("Les Misérables") {
println!("We've got {} reviews, but Les Misérables ain't one.",
movie_reviews.len());
}
// oops, this review has a lot of spelling mistakes, let's delete it.
movie_reviews.remove("The Blues Brothers");
// look up the values associated with some keys.
let to_find = ["Up!", "Office Space"];
for movie in &to_find {
match movie_reviews.get(movie) {
Some(review) => println!("{movie}: {review}"),
None => println!("{movie} is unreviewed.")
}
}
// Look up the value for a key (will panic if the key is not found).
println!("Movie review: {}", movie_reviews["Office Space"]);
// iterate over everything.
for (movie, review) in &movie_reviews {
println!("{movie}: \"{review}\"");
}Currently we're using some specific crates (sptr and in the future back to core::ptr::*) to ensure that we're compatible with Strict Provenance. The following MIRIFLAGS setup should enable checking to make sure that we're compatible.
MIRIFLAGS="-Zmiri-strict-provenance -Zmiri-symbolic-alignment-check" cargo +nightly miri testI think this is useful because we're doing some pointer times with our tagged pointers implementation, mutating the contents of the pointer to store bits of data.
To run the fuzzer I use the command:
cargo +nightly fuzz run -j 8 -s address fuzz_tree_map_api -- -max_len=32768 -max_total_time=3600 && cargo +nightly fuzz cmin fuzz_tree_map_apiThis will run the fuzzer for a total of 3600 seconds (1 hour), using 8 jobs (half of the total number of cores on my dev box), and using the address sanitizer. The cmin command is used to compact the corpus after generating new entries.
To generate coverage reports from fuzzing corpus:
# replace with own triple as required
TARGET_TRIPLE="x86_64-unknown-linux-gnu"
cargo +nightly fuzz coverage fuzz_tree_map_api && cargo cov -- show fuzz/target/"$TARGET_TRIPLE"/release/fuzz_tree_map_api \
--format=html \
-instr-profile=fuzz/coverage/fuzz_tree_map_api/coverage.profdata \
> index.htmlTARGET_TRIPLE="x86_64-unknown-linux-gnu"
/home/declan/.rustup/toolchains/nightly-x86_64-unknown-linux-gnu/lib/rustlib/x86_64-unknown-linux-gnu/bin/llvm-cov show -format=html \
-instr-profile=fuzz/coverage/fuzz_tree_map_api/coverage.profdata \
-Xdemangler=rustfilt \
-ignore-filename-regex=\.cargo/registry \
fuzz/target/x86_64-unknown-linux-gnu/release/fuzz_tree_map_api
> cov.html
/home/declan/.rustup/toolchains/nightly-x86_64-unknown-linux-gnu/lib/rustlib/x86_64-unknown-linux-gnu/bin/llvm-cov show \
--instr-profile=/home/declan/repos/github/declanvk/blart/fuzz/coverage/fuzz_tree_map_api/coverage.profdata \
--show-instantiations --show-line-counts-or-regions --Xdemangler=rustfilt \
--format=html --ignore-filename-regex=/home/declan/.cargo/registry --ignore-filename-regex=/home/declan/.rustup/\
target/x86_64-unknown-linux-gnu/coverage/x86_64-unknown-linux-gnu/release/fuzz_tree_map_api > coverage.htmlTo run the benchmarks, install cargo-criterion, then run:
cargo +nightly criterion --history-id "$(git rev-parse --short HEAD)-0" --features bench-perf-eventsor
cargo bench --bench <bench_name> --features bench-perf-eventsIf you get a "Permission denied" error, update perf_event_paranoid:
sudo sh -c 'echo 1 >/proc/sys/kernel/perf_event_paranoid'For further details please take a look at the following link.
I use a somewhat realistic benchmark: counting words in a text file. To get started, download a text file like:
curl -o data/Ulysses.txt https://www.gutenberg.org/cache/epub/4300/pg4300.txtThen build the word count example using the profiling profile:
RUSTFLAGS="-C force-frame-pointers=yes" cargo build --profile profiling --examplesThen run the count words workload on the downloaded data while profiling:
samply record ./target/profiling/examples/count_words blart data/book-chapters-combined.txtLicensed under either of
- Apache License, Version 2.0 (LICENSE-APACHE or http://www.apache.org/licenses/LICENSE-2.0)
- MIT license (LICENSE-MIT or http://opensource.org/licenses/MIT)
at your option.
Unless you explicitly state otherwise, any contribution intentionally submitted for inclusion in the work by you, as defined in the Apache-2.0 license, shall be dual licensed as above, without any additional terms or conditions.