Run ETE benchmarks with MiMalloc and leave a note encouraging its usa…

…ge (#399)

fixes #395

Cargo.toml

@@ -103,6 +103,7 @@ uuid = "1.8.0"
 walkdir = "2.5.0"
 worker = "0.3.0"
 zigzag = "0.1.0"
+mimalloc = "0.1.42"
 
 [workspace.lints.rust]
 warnings = "deny"

README.md

@@ -150,6 +150,16 @@ Vortex to model more complex arrays while still exposing a logical interface. Fo
 `ChunkedArray` where the first chunk is run-length encoded and the second chunk is dictionary encoded.
 In Arrow, `RunLengthArray` and `DictionaryArray` are separate incompatible types, and so cannot be combined in this way.
 
+### Usage
+
+For best performance we recommend using [MiMalloc](https://github.com/microsoft/mimalloc) as the application's
+allocator.
+
+```rust
+#[global_allocator]
+static GLOBAL_ALLOC: MiMalloc = MiMalloc;
+```
+
 ## Contributing
 
 Please see [CONTRIBUTING.md](CONTRIBUTING.md).
@@ -159,6 +169,7 @@ Please see [CONTRIBUTING.md](CONTRIBUTING.md).
 In order to build vortex, you may also need to install the flatbuffer compiler (flatc):
 
 ### Mac
+
 ```bash
 brew install flatbuffers
 ```

bench-vortex/Cargo.toml

@@ -17,16 +17,20 @@ workspace = true
 [dependencies]
 arrow-array = { workspace = true }
 arrow-select = { workspace = true }
+bytes = { workspace = true }
 bzip2 = { workspace = true }
 csv = { workspace = true }
 enum-iterator = { workspace = true }
+flexbuffers = { workspace = true }
 futures = { workspace = true }
 humansize = { workspace = true }
 itertools = { workspace = true }
 lazy_static = { workspace = true }
 log = { workspace = true }
+mimalloc = { workspace = true }
 parquet = { workspace = true, features = [] }
 reqwest = { workspace = true }
+serde = { workspace = true }
 simplelog = { workspace = true }
 tokio = { workspace = true, features = ["full"] }
 uuid = { workspace = true, features = ["v4"] }
@@ -41,9 +45,6 @@ vortex-fastlanes = { path = "../encodings/fastlanes" }
 vortex-ipc = { path = "../vortex-ipc" }
 vortex-ree = { path = "../encodings/runend" }
 vortex-roaring = { path = "../encodings/roaring" }
-serde = { workspace = true }
-bytes = { workspace = true }
-flexbuffers = { workspace = true }
 
 [dev-dependencies]
 criterion = { workspace = true, features = ["html_reports", "async_tokio"] }

bench-vortex/benches/random_access.rs

@@ -1,11 +1,14 @@
 use bench_vortex::reader::{take_parquet, take_vortex};
 use bench_vortex::taxi_data::{taxi_data_parquet, taxi_data_vortex};
 use criterion::{black_box, criterion_group, criterion_main, Criterion};
+use mimalloc::MiMalloc;
 use tokio::runtime::Runtime;
 
+#[global_allocator]
+static GLOBAL: MiMalloc = MiMalloc;
+
 fn random_access(c: &mut Criterion) {
     let mut group = c.benchmark_group("random access");
-    group.sample_size(10);
 
     let indices = [10, 11, 12, 13, 100_000, 3_000_000];
 
@@ -16,7 +19,7 @@ fn random_access(c: &mut Criterion) {
     });
 
     let taxi_parquet = taxi_data_parquet();
-    group.bench_function("arrow", |b| {
+    group.sample_size(10).bench_function("arrow", |b| {
         b.iter(|| black_box(take_parquet(&taxi_parquet, &indices).unwrap()))
     });
 }