venkyt-arista:venkyt/single-threaded-client - Branch - mozilla/sccache

fix: use single-threaded tokio runtime in sccache client

#2460

Comparing

venkyt-arista:venkyt/single-threaded-client

(

2f14909

) with

main

(

951e247

)

Untouched: 27

Skipped: 2

Benchmarks

Skipped (2)

These benchmarks were skipped, so their baseline results are used instead. If they were deleted in your codebase, archive them to remove them from the performance reports.
Learn more about archiving benchmarks

digest_with_delimiters

benches/sccache_bench.rs

Skipped

8.8 µs*

digest_incremental_updates

benches/sccache_bench.rs

Skipped

20.5 µs*

Passed

lru_hotcold_access_pattern

benches/sccache_bench.rs

143.5 µs143.4 µs

lru_sequential_scan_pattern

benches/sccache_bench.rs

359.5 µs359.5 µs

hash_multiple_files

benches/sccache_bench.rs

938.6 µs938.6 µs

cache_entry_create_small

benches/sccache_bench.rs

1.4 ms1.4 ms

decompression_high_ratio

benches/sccache_bench.rs

1.2 ms1.2 ms

hash_header_file

benches/sccache_bench.rs

204.7 µs204.7 µs

cache_key_generation

benches/sccache_bench.rs

89.5 µs89.5 µs

time_macro_finder_chunked

benches/sccache_bench.rs

971.8 µs971.8 µs

hash_large_data

benches/sccache_bench.rs

9.3 ms9.3 ms

time_macro_finder_no_macros

benches/sccache_bench.rs

3 ms3 ms

time_macro_finder_with_macros

benches/sccache_bench.rs

2.6 ms2.6 ms

compression_high_compressibility

benches/sccache_bench.rs

2.4 ms2.4 ms

build_workflow_incremental

benches/sccache_bench.rs

6.7 ms6.7 ms

compression_low_compressibility

benches/sccache_bench.rs

2.1 ms2.1 ms

cache_entry_batch_roundtrip

benches/sccache_bench.rs

22.8 ms22.8 ms

cache_entry_create_large

benches/sccache_bench.rs

4.9 ms4.9 ms

cache_entry_roundtrip_large

benches/sccache_bench.rs

5.5 ms5.5 ms

cache_entry_roundtrip_small

benches/sccache_bench.rs

1.9 ms1.9 ms

build_workflow_rebuild

benches/sccache_bench.rs

4.6 ms4.6 ms

decompression_low_ratio

benches/sccache_bench.rs

1.1 ms1.1 ms

lru_realistic_eviction_pressure

benches/sccache_bench.rs

9.1 ms9.1 ms

cache_entry_batch_create

benches/sccache_bench.rs

48 ms48 ms

build_workflow_initial

benches/sccache_bench.rs

29.2 ms29.2 ms

lru_cache_get_hit

benches/sccache_bench.rs

974.3 µs975.1 µs

lru_cache_eviction

benches/sccache_bench.rs

3.4 ms3.4 ms

lru_cache_insert

benches/sccache_bench.rs

2.7 ms2.7 ms

lru_cache_mixed_workload

benches/sccache_bench.rs

2.2 ms2.2 ms

Commits

Click on a commit to change the comparison range

Base

main

951e247

+0.01%

fix: use single-threaded tokio runtime in sccache client The sccache client was using Runtime::new() which creates a multi-threaded tokio runtime with worker threads equal to the number of CPU cores. On high-core-count servers running many parallel builds, this created an excessive number of threads. For example: - 96 vCPU server - 10 concurrent make invocations - Each make using -j16 - Result: 96 × 10 × 16 = 15,360 threads created by sccache wrappers While these threads are short-lived, with continuous build pipelines this constant thread creation/destruction causes unnecessary overhead. The sccache client only performs simple I/O operations (connecting to server, sending requests, receiving responses) and doesn't need worker threads. This change replaces all client-side Runtime::new() calls with Builder::new_current_thread().enable_all().build() to use a single-threaded runtime, reducing thread count from num_cpus to 1 per client invocation.

5a3519a

3 months ago

by venkyt-arista

-0.02%

Merge branch 'main' into venkyt/single-threaded-client

2d6cfc8

6 days ago

by sylvestre

-0.01%

Merge branch 'main' into venkyt/single-threaded-client

2f14909

6 days ago

by sylvestre

Home Terms Privacy Docs