Add compression utility to benchlib and document runtime benchmark input file handling

Author: Kobzol

Cargo.lock

@@ -14,6 +14,10 @@ log = "0.4.17"
 env_logger = "0.10.0"
 clap = { version = "4.1", features = ["derive", "string"] }
 libc = "0.2"
+flate2 = { version = "1", optional = true }
 
 [target.'cfg(target_os = "linux")'.dependencies]
 perf-event = "0.4.7"
+
+[features]
+compression = ["dep:flate2"]

collector/benchlib/Cargo.toml

@@ -18,3 +18,7 @@ mod cli;
 pub mod comm;
 pub mod measure;
 pub mod process;
+mod utils;
+
+#[cfg(feature = "compression")]
+pub use utils::decompress_file;

collector/benchlib/src/lib.rs

@@ -0,0 +1,12 @@
+/// Decompresses the contents of a gzipped (.gz) file.
+#[cfg(feature = "compression")]
+pub fn decompress_file(contents: &[u8]) -> Vec<u8> {
+    use std::io::Read;
+
+    let mut tar = flate2::bufread::GzDecoder::new(contents);
+    let mut result = Vec::new();
+    tar.read_to_end(&mut result)
+        .expect("Cannot decompress file");
+
+    result
+}

collector/benchlib/src/utils.rs

@@ -22,3 +22,15 @@ not print anything to `stdout`!
 Note that each benchmark group binary will be thus invoked twice per benchmarking collection. Keep this
 in mind so that the `main` function of the group doesn't perform too much work, which will be thrown
 away when it is invoked with the `list` argument.
+
+## Benchmark input files
+Some benchmarks will need input files. Ideally, these should be stored directly inside this repository,
+e.g. in a `data` directory under the benchmark group workspace root. If some files are shared between
+multiple groups, you can put them in the `runtime-benchmarks/data` directory. To keep benchmarks
+self-contained, consider including the input files directly into the binary using the `include_bytes!`
+or `include_str!` macros.
+
+Try to keep the sizes of input files reasonable. If you want to increase their size for the benchmark,
+you can repeat their contents at the beginning of the benchmark run (in `main`). If the file has a
+format that is not easily repeatable it is large (e.g. larger than 1 MiB), consider compressing it
+using `gzip <file>` and them decompressing it in `main` using `benchlib::decompress_file`.