JMH performance benchmark for Java's native call APIs/libraries: JNI, FFM, JNA, JNADirect, JNR, JFFI, Multiffi/FFI.
5 functions to measure:
- no arguments, return void
- no arguments, return int
- 1 argument: int, return int
- no arguments, return double
- 1 argument: double, return double
- JDK 22+
- Gradle 8.10
- C compiler with C99 support
- CMake 3.24+ with a valid backend
- JAVA_HOME environment variable
First, build the native shared library:
cd ${PROJECT_ROOT}/src/main/c
cmake -S . -B ./build -DCMAKE_BUILD_TYPE=MinSizeRel
cd build
make
cp ${LIBRARY_NAME} ${PROJECT_ROOT}/Then, run the jmh main class:
cd ${PROJECT_ROOT}
./gradlew clean jmhCPU: Intel® Core™ i7-1165G7 @ 2.80GHz
OS: GNU/Linux
JDK: Eclipse Temurin JDK 23.0.1+11
VM: OpenJDK 64-Bit Server VM, 23.0.1+11
JMH version: 1.37
Blackhole mode: compiler (auto-detected)
Warmup: 3 iterations, 10 s each
Measurement: 5 iterations, 10 s each
Benchmark mode: Average time, time/op (smaller is better)
Benchmark Mode Cnt Score Error Units
FFIBenchmark.measureFFMCriticalReturnDouble avgt 5 3.921 ± 4.357 ns/op
FFIBenchmark.measureFFMCriticalReturnDoubleD avgt 5 3.990 ± 1.784 ns/op
FFIBenchmark.measureFFMCriticalReturnInt avgt 5 3.620 ± 2.541 ns/op
FFIBenchmark.measureFFMCriticalReturnIntI avgt 5 3.874 ± 1.402 ns/op
FFIBenchmark.measureFFMCriticalReturnVoid avgt 5 3.777 ± 2.609 ns/op
FFIBenchmark.measureFFMReturnDouble avgt 5 9.909 ± 5.341 ns/op
FFIBenchmark.measureFFMReturnDoubleD avgt 5 9.654 ± 4.726 ns/op
FFIBenchmark.measureFFMReturnInt avgt 5 10.528 ± 7.152 ns/op
FFIBenchmark.measureFFMReturnIntI avgt 5 10.143 ± 3.922 ns/op
FFIBenchmark.measureFFMReturnVoid avgt 5 9.925 ± 4.447 ns/op
FFIBenchmark.measureJFFIReturnDouble avgt 5 11.088 ± 2.427 ns/op
FFIBenchmark.measureJFFIReturnDoubleD avgt 5 11.163 ± 4.180 ns/op
FFIBenchmark.measureJFFIReturnInt avgt 5 11.051 ± 5.376 ns/op
FFIBenchmark.measureJFFIReturnIntI avgt 5 10.736 ± 3.835 ns/op
FFIBenchmark.measureJFFIReturnVoid avgt 5 10.802 ± 3.972 ns/op
FFIBenchmark.measureJNADirectReturnDouble avgt 5 84.891 ± 55.869 ns/op
FFIBenchmark.measureJNADirectReturnDoubleD avgt 5 107.322 ± 37.283 ns/op
FFIBenchmark.measureJNADirectReturnInt avgt 5 78.389 ± 37.819 ns/op
FFIBenchmark.measureJNADirectReturnIntI avgt 5 106.022 ± 49.758 ns/op
FFIBenchmark.measureJNADirectReturnVoid avgt 5 80.741 ± 44.313 ns/op
FFIBenchmark.measureJNAReturnDouble avgt 5 92.971 ± 37.531 ns/op
FFIBenchmark.measureJNAReturnDoubleD avgt 5 347.726 ± 372.161 ns/op
FFIBenchmark.measureJNAReturnInt avgt 5 96.692 ± 67.942 ns/op
FFIBenchmark.measureJNAReturnIntI avgt 5 262.270 ± 126.216 ns/op
FFIBenchmark.measureJNAReturnVoid avgt 5 96.983 ± 49.700 ns/op
FFIBenchmark.measureJNIReturnDouble avgt 5 10.437 ± 3.827 ns/op
FFIBenchmark.measureJNIReturnDoubleD avgt 5 10.734 ± 6.816 ns/op
FFIBenchmark.measureJNIReturnInt avgt 5 10.076 ± 4.791 ns/op
FFIBenchmark.measureJNIReturnIntI avgt 5 11.286 ± 7.431 ns/op
FFIBenchmark.measureJNIReturnVoid avgt 5 10.273 ± 5.157 ns/op
FFIBenchmark.measureJNRReturnDouble avgt 5 10.769 ± 3.098 ns/op
FFIBenchmark.measureJNRReturnDoubleD avgt 5 11.809 ± 5.242 ns/op
FFIBenchmark.measureJNRReturnInt avgt 5 10.505 ± 1.893 ns/op
FFIBenchmark.measureJNRReturnIntI avgt 5 10.995 ± 5.997 ns/op
FFIBenchmark.measureJNRReturnVoid avgt 5 11.254 ± 4.898 ns/op
FFIBenchmark.measureMultiffiFFMCriticalProxyReturnDouble avgt 5 3.488 ± 2.268 ns/op
FFIBenchmark.measureMultiffiFFMCriticalProxyReturnDoubleD avgt 5 4.113 ± 3.199 ns/op
FFIBenchmark.measureMultiffiFFMCriticalProxyReturnInt avgt 5 3.378 ± 2.383 ns/op
FFIBenchmark.measureMultiffiFFMCriticalProxyReturnIntI avgt 5 3.852 ± 1.547 ns/op
FFIBenchmark.measureMultiffiFFMCriticalProxyReturnVoid avgt 5 3.921 ± 2.163 ns/op
FFIBenchmark.measureMultiffiFFMCriticalReturnDouble avgt 5 10.910 ± 7.116 ns/op
FFIBenchmark.measureMultiffiFFMCriticalReturnDoubleD avgt 5 13.715 ± 14.789 ns/op
FFIBenchmark.measureMultiffiFFMCriticalReturnInt avgt 5 9.183 ± 4.582 ns/op
FFIBenchmark.measureMultiffiFFMCriticalReturnIntI avgt 5 10.954 ± 12.795 ns/op
FFIBenchmark.measureMultiffiFFMCriticalReturnVoid avgt 5 14.039 ± 7.318 ns/op
FFIBenchmark.measureMultiffiFFMProxyReturnDouble avgt 5 9.629 ± 3.406 ns/op
FFIBenchmark.measureMultiffiFFMProxyReturnDoubleD avgt 5 9.442 ± 4.271 ns/op
FFIBenchmark.measureMultiffiFFMProxyReturnInt avgt 5 10.136 ± 2.752 ns/op
FFIBenchmark.measureMultiffiFFMProxyReturnIntI avgt 5 11.800 ± 2.986 ns/op
FFIBenchmark.measureMultiffiFFMProxyReturnVoid avgt 5 9.595 ± 3.473 ns/op
FFIBenchmark.measureMultiffiFFMReturnDouble avgt 5 16.025 ± 6.788 ns/op
FFIBenchmark.measureMultiffiFFMReturnDoubleD avgt 5 18.300 ± 7.764 ns/op
FFIBenchmark.measureMultiffiFFMReturnInt avgt 5 15.601 ± 10.439 ns/op
FFIBenchmark.measureMultiffiFFMReturnIntI avgt 5 14.878 ± 10.161 ns/op
FFIBenchmark.measureMultiffiFFMReturnVoid avgt 5 14.357 ± 3.633 ns/op
FFIBenchmark.measureMultiffiJNANoASMProxyReturnDouble avgt 5 97.172 ± 55.662 ns/op
FFIBenchmark.measureMultiffiJNANoASMProxyReturnDoubleD avgt 5 358.443 ± 205.594 ns/op
FFIBenchmark.measureMultiffiJNANoASMProxyReturnInt avgt 5 96.205 ± 63.602 ns/op
FFIBenchmark.measureMultiffiJNANoASMProxyReturnIntI avgt 5 282.072 ± 204.915 ns/op
FFIBenchmark.measureMultiffiJNANoASMProxyReturnVoid avgt 5 78.910 ± 41.103 ns/op
FFIBenchmark.measureMultiffiJNAProxyReturnDouble avgt 5 79.186 ± 33.923 ns/op
FFIBenchmark.measureMultiffiJNAProxyReturnDoubleD avgt 5 105.324 ± 47.596 ns/op
FFIBenchmark.measureMultiffiJNAProxyReturnInt avgt 5 83.672 ± 58.382 ns/op
FFIBenchmark.measureMultiffiJNAProxyReturnIntI avgt 5 106.174 ± 91.204 ns/op
FFIBenchmark.measureMultiffiJNAProxyReturnVoid avgt 5 79.585 ± 31.512 ns/op
FFIBenchmark.measureMultiffiJNAReturnDouble avgt 5 88.017 ± 56.688 ns/op
FFIBenchmark.measureMultiffiJNAReturnDoubleD avgt 5 304.737 ± 138.740 ns/op
FFIBenchmark.measureMultiffiJNAReturnInt avgt 5 76.119 ± 33.180 ns/op
FFIBenchmark.measureMultiffiJNAReturnIntI avgt 5 234.871 ± 141.360 ns/op
FFIBenchmark.measureMultiffiJNAReturnVoid avgt 5 72.926 ± 38.152 ns/op
FFIBenchmark.measureMultiffiJNRNoASMProxyReturnDouble avgt 5 45.449 ± 20.318 ns/op
FFIBenchmark.measureMultiffiJNRNoASMProxyReturnDoubleD avgt 5 79.491 ± 55.846 ns/op
FFIBenchmark.measureMultiffiJNRNoASMProxyReturnInt avgt 5 27.725 ± 8.274 ns/op
FFIBenchmark.measureMultiffiJNRNoASMProxyReturnIntI avgt 5 49.187 ± 41.879 ns/op
FFIBenchmark.measureMultiffiJNRNoASMProxyReturnVoid avgt 5 24.496 ± 15.223 ns/op
FFIBenchmark.measureMultiffiJNRProxyReturnDouble avgt 5 10.520 ± 3.762 ns/op
FFIBenchmark.measureMultiffiJNRProxyReturnDoubleD avgt 5 11.245 ± 6.240 ns/op
FFIBenchmark.measureMultiffiJNRProxyReturnInt avgt 5 10.814 ± 3.053 ns/op
FFIBenchmark.measureMultiffiJNRProxyReturnIntI avgt 5 11.256 ± 5.799 ns/op
FFIBenchmark.measureMultiffiJNRProxyReturnVoid avgt 5 11.358 ± 5.250 ns/op
FFIBenchmark.measureMultiffiJNRReturnDouble avgt 5 39.349 ± 24.939 ns/op
FFIBenchmark.measureMultiffiJNRReturnDoubleD avgt 5 60.974 ± 47.156 ns/op
FFIBenchmark.measureMultiffiJNRReturnInt avgt 5 20.670 ± 11.473 ns/op
FFIBenchmark.measureMultiffiJNRReturnIntI avgt 5 38.172 ± 27.314 ns/op
FFIBenchmark.measureMultiffiJNRReturnVoid avgt 5 26.940 ± 15.717 ns/op