From 173e6fe995e5edc26d592ce076bdc3f9c121f383 Mon Sep 17 00:00:00 2001
From: eldritch horrors <pennae@lix.systems>
Date: Mon, 28 Jul 2025 23:57:16 +0200
Subject: [PATCH] libstore: optimize nar copy stream adapter

we don't need to report progress for every read call. that's way too
much. batching like this greatly reduces CPU usage for copies out of
or into remote buidlers due to likewise greatly reduced log traffic.

Change-Id: I3db2b2ab113fbaadefc69cfde6f977fb0c6cd5ad
---
 lix/libstore/store-api.cc | 9 +++++++++
 1 file changed, 9 insertions(+)

diff --git a/lix/libstore/store-api.cc b/lix/libstore/store-api.cc
index 9a649901d..1f2a4a509 100644
--- a/lix/libstore/store-api.cc
+++ b/lix/libstore/store-api.cc
@@ -1031,8 +1031,17 @@ struct CopyPathStream : AsyncInputStream
     kj::Promise<Result<std::optional<size_t>>> read(void * data, size_t len) override
     try {
         auto result = TRY_AWAIT(inner->read(data, len));
+
+        // do not log progress on every call. nar copies cause a lot of small
+        // reads, letting each read report the current copy progress causes a
+        // huge amount of overhead (20x or more) in log traffic. reporting at
+        // 64 kiB intervals is probably enough, being about 1000 dir entries.
+        constexpr size_t CHUNK = 65536;
+        const auto doLog = !result || copied / CHUNK < (copied + *result) / CHUNK || *result < len;
         if (result) {
             copied += *result;
+        }
+        if (doLog) {
             act.progress(copied, expected);
         }
         co_return result;