A benchmark of mutex vs spinlock throughput for an extremely short critical section under varying levels of contention on "average" desktop.
Inspiration and code for AmdSpinlock are from https://probablydance.com/2019/12/30/measuring-mutexes-spinlocks-and-how-bad-the-linux-scheduler-really-is/.
Summary of results:
- Spinlocks are almost always significantly worse than a good mutex, and never significantly better,
- Contention makes spinlocks relatively slower.
Biggest known caveat (apart from this being a single benchmark run on a single machine):
The best mutex implementation seems to be relatively more optimized than the best spinlock implementation.
extreme contention:
12:31:05|~/projects/lock-bench|master⚡*
λ cargo run --release 32 2 10000 100
Finished release [optimized] target(s) in 0.01s
Running `target/release/lock-bench 32 2 10000 100`
Options {
n_threads: 32,
n_locks: 2,
n_ops: 10000,
n_rounds: 100,
}
std::sync::Mutex avg 97.770106ms min 38.799445ms max 103.42306ms
parking_lot::Mutex avg 68.350542ms min 32.139233ms max 72.404877ms
spin::Mutex avg 142.257494ms min 69.860396ms max 217.587871ms
AmdSpinlock avg 127.612286ms min 50.407761ms max 219.429909ms
std::sync::Mutex avg 98.838394ms min 68.180052ms max 125.635571ms
parking_lot::Mutex avg 68.51149ms min 58.805279ms max 71.512899ms
spin::Mutex avg 139.751964ms min 54.499263ms max 193.70374ms
AmdSpinlock avg 127.757924ms min 50.249234ms max 210.452623ms
heavy contention:
12:34:39|~/projects/lock-bench|master⚡*
λ cargo run --release 32 64 10000 100
Finished release [optimized] target(s) in 0.01s
Running `target/release/lock-bench 32 64 10000 100`
Options {
n_threads: 32,
n_locks: 64,
n_ops: 10000,
n_rounds: 100,
}
std::sync::Mutex avg 21.538657ms min 11.704293ms max 23.688275ms
parking_lot::Mutex avg 10.016941ms min 6.787887ms max 11.7508ms
spin::Mutex avg 55.555043ms min 7.639845ms max 161.030869ms
AmdSpinlock avg 40.82985ms min 6.174719ms max 123.545934ms
std::sync::Mutex avg 21.489658ms min 20.344423ms max 24.05294ms
parking_lot::Mutex avg 9.640073ms min 6.782365ms max 12.600402ms
spin::Mutex avg 48.74331ms min 7.601425ms max 138.172171ms
AmdSpinlock avg 40.993328ms min 8.365127ms max 110.106416ms
light contention:
12:29:01|~/projects/lock-bench|master⚡*
λ cargo run --release 32 1000 10000 100
Finished release [optimized] target(s) in 0.01s
Running `target/release/lock-bench 32 1000 10000 100`
Options {
n_threads: 32,
n_locks: 1000,
n_ops: 10000,
n_rounds: 100,
}
std::sync::Mutex avg 13.897816ms min 8.176447ms max 15.680984ms
parking_lot::Mutex avg 6.553058ms min 3.284944ms max 8.230407ms
spin::Mutex avg 37.946399ms min 4.668167ms max 115.748116ms
AmdSpinlock avg 39.530919ms min 2.049988ms max 127.139724ms
std::sync::Mutex avg 13.922504ms min 12.885598ms max 15.28518ms
parking_lot::Mutex avg 6.80546ms min 5.621588ms max 8.932723ms
spin::Mutex avg 39.411306ms min 4.752935ms max 102.888667ms
AmdSpinlock avg 37.423773ms min 5.087086ms max 103.319751ms
no contention:
12:26:25|~/projects/lock-bench|master⚡*
λ cargo run --release 32 1000000 10000 100
Finished release [optimized] target(s) in 0.01s
Running `target/release/lock-bench 32 1000000 10000 100`
Options {
n_threads: 32,
n_locks: 1000000,
n_ops: 10000,
n_rounds: 100,
}
std::sync::Mutex avg 15.975801ms min 8.487617ms max 27.106072ms
parking_lot::Mutex avg 7.142194ms min 4.594504ms max 9.178952ms
spin::Mutex avg 5.947778ms min 4.600631ms max 8.262928ms
AmdSpinlock avg 6.512221ms min 5.059464ms max 10.408178ms
std::sync::Mutex avg 15.793032ms min 8.260458ms max 27.850666ms
parking_lot::Mutex avg 6.910369ms min 4.32757ms max 9.08715ms
spin::Mutex avg 5.890677ms min 4.424554ms max 7.622798ms
AmdSpinlock avg 6.416132ms min 5.852334ms max 7.349909ms
12:45:27|~/projects/lock-bench|master⚡*?
λ cat /etc/os-release
NAME=NixOS
ID=nixos
VERSION="19.09.1693.eab4ee0c27c (Loris)"
VERSION_CODENAME=loris
VERSION_ID="19.09.1693.eab4ee0c27c"
PRETTY_NAME="NixOS 19.09.1693.eab4ee0c27c (Loris)"
LOGO="nix-snowflake"
HOME_URL="https://nixos.org/"
DOCUMENTATION_URL="https://nixos.org/nixos/manual/index.html"
SUPPORT_URL="https://nixos.org/nixos/support.html"
BUG_REPORT_URL="https://github.com/NixOS/nixpkgs/issues"
12:45:32|~/projects/lock-bench|master⚡*?
λ uname -r
4.19.91
12:45:34|~/projects/lock-bench|master⚡*?
λ lscpu
Architecture: x86_64
CPU op-mode(s): 32-bit, 64-bit
Byte Order: Little Endian
Address sizes: 39 bits physical, 48 bits virtual
CPU(s): 8
On-line CPU(s) list: 0-7
Thread(s) per core: 2
Core(s) per socket: 4
Socket(s): 1
NUMA node(s): 1
Vendor ID: GenuineIntel
CPU family: 6
Model: 158
Model name: Intel(R) Core(TM) i7-7820HQ CPU @ 2.90GHz
Stepping: 9
CPU MHz: 1073.198
CPU max MHz: 3900.0000
CPU min MHz: 800.0000
BogoMIPS: 5808.00
Virtualization: VT-x
L1d cache: 32K
L1i cache: 32K
L2 cache: 256K
L3 cache: 8192K
NUMA node0 CPU(s): 0-7
Flags: fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca cmov pat pse36 clflush dts acpi mmx fxsr sse sse2 ss ht tm pbe syscall nx pdpe1gb rdtscp lm constant_tsc art arch_perfmon pebs bts rep_good nopl xtopology nonstop_tsc cpuid aperfmperf tsc_known_freq pni pclmulqdq dtes64 monitor ds_cpl vmx smx est tm2 ssse3 sdbg fma cx16 xtpr pdcm pcid sse4_1 sse4_2 x2apic movbe popcnt tsc_deadline_timer aes xsave avx f16c rdrand lahf_lm abm 3dnowprefetch cpuid_fault epb invpcid_single pti ssbd ibrs ibpb stibp tpr_shadow vnmi flexpriority ept vpid ept_ad fsgsbase tsc_adjust bmi1 hle avx2 smep bmi2 erms invpcid rtm mpx rdseed adx smap clflushopt intel_pt xsaveopt xsavec xgetbv1 xsaves dtherm ida arat pln pts hwp hwp_notify hwp_act_window hwp_epp flush_l1d