feat: use ahash for better hash performance

Signed-off-by: xiaguan <[email protected]>

Cargo.toml

@@ -16,6 +16,7 @@ jemalloc = ["tikv-jemallocator"]
 python = ["pyo3", "pyo3-build-config"]
 
 [dependencies]
+ahash = "0.8"
 anyhow = "1"
 async-recursion = "1"
 async-stream = "0.3"
@@ -119,3 +120,6 @@ risinglight_proto = { path = "proto" }
 [lib]
 name = "risinglight"
 crate-type = ["cdylib", "rlib"]
+
+[profile.release]
+debug = true

src/executor/hash_agg.rs

@@ -1,7 +1,6 @@
 // Copyright 2024 RisingLight Project Authors. Licensed under Apache-2.0.
 
-use std::collections::HashMap;
-
+use ahash::{HashMap, HashMapExt};
 use iter_chunks::IterChunks;
 use smallvec::SmallVec;
 

src/executor/hash_join.rs

@@ -1,9 +1,9 @@
 // Copyright 2024 RisingLight Project Authors. Licensed under Apache-2.0.
 
-use std::collections::{HashMap, HashSet};
 use std::marker::ConstParamTy;
 use std::vec::Vec;
 
+use ahash::{HashMap, HashMapExt, HashSet, HashSetExt};
 use futures::TryStreamExt;
 use smallvec::SmallVec;
 

src/storage/secondary/index.rs

@@ -44,7 +44,7 @@ impl ColumnIndex {
         let checksum = footer.get_u64();
         verify_checksum(checksum_type, index_data, checksum)?;
 
-        let mut indexes = vec![];
+        let mut indexes = Vec::with_capacity(length);
         for _ in 0..length {
             let index = BlockIndex::decode_length_delimited(&mut index_data)?;
             indexes.push(index);