From 1cf4bcc883a4ff0fd7ec4e04d4a80ae0c593a513 Mon Sep 17 00:00:00 2001
From: not-matthias <matthias@codspeed.io>
Date: Thu, 1 Oct 2026 16:59:28 +0200
Subject: [PATCH] perf(runner-shared): buffer the decompressed memtrack event
 stream

`MemtrackArtifact::decode_streamed` handed the zstd decoder straight to
rmp_serde, which reads 1 to 8 bytes per call, so every small read went
through zstd's stream API. Put a `BufReader` between the two.

On a large memtrack artifact, decoding all events drops from 21-33 s to
11-21 s, with no change in peak memory.

Refs COD-3704
---
 crates/runner-shared/src/artifacts/memtrack/mod.rs | 7 +++++--
 1 file changed, 5 insertions(+), 2 deletions(-)

diff --git a/crates/runner-shared/src/artifacts/memtrack/mod.rs b/crates/runner-shared/src/artifacts/memtrack/mod.rs
index 7560ec9c..226d5750 100644
--- a/crates/runner-shared/src/artifacts/memtrack/mod.rs
+++ b/crates/runner-shared/src/artifacts/memtrack/mod.rs
@@ -24,12 +24,15 @@ impl super::ArtifactExt for MemtrackArtifact {
 }
 
 impl MemtrackArtifact {
+    /// The msgpack decoder reads a few bytes per call, so the decompressed
+    /// stream is buffered rather than read straight from the zstd decoder.
+    #[allow(clippy::type_complexity)]
     pub fn decode_streamed<R: std::io::Read>(
         reader: R,
-    ) -> anyhow::Result<MemtrackEventStream<zstd::Decoder<'static, std::io::BufReader<R>>>> {
+    ) -> anyhow::Result<MemtrackEventStream<BufReader<zstd::Decoder<'static, BufReader<R>>>>> {
         let decoder = zstd::Decoder::new(reader)?;
         Ok(MemtrackEventStream {
-            deserializer: rmp_serde::Deserializer::new(decoder),
+            deserializer: rmp_serde::Deserializer::new(BufReader::new(decoder)),
         })
     }