# Channel Benchmark Results This documents lists several benchmark results for different platforms in order to identify performance regressions and improvements. The are measured using the command ``` > channel_bench_* nsend ... number of threads that push objects into the channel nrec .... number of threads that pop objects from the channel buflen .. length of channel buffer queue - `0` means unbuffered channel nobj .... number of objects to pass thru the channel ``` ## AMD Ryzen 7 3800X, Ubuntu-20.04 x86_64 10000000 Objects transfered, results in Objects/µs | nsend | nrec | buflen | **V (gcc -O2)** | **V (clang)** | **V (tcc)** | **Go (glang)** | **Go (gccgo -O2)** | | :---: | :---:| :---: | :---: | :---: | :---: | :---: | :---: | | 1 | 1 | 0 | 0.95 | 0.72 | 0.66 | 4.65 | 0.56 | | 1 | 1 | 100 | 3.26 | 2.51 | 2.24 | 18.90 | 6.08 | | 4 | 4 | 0 | 0.25 | 0.26 | 0.24 | 1.84 | 0.84 | | 4 | 4 | 100 | 3.11 | 2.78 | 2.63 | 7.43 | 3.71 | ## AMD Ryzen 7 3800X, Windows 10 2004 x64 | nsend | nrec | buflen | **V (gcc -O2)** | **V (msvc /O2)** | **V (tcc)** | **Go (golang)** | | :---: | :---:| :---: | :---: | :---: | :---: | :---: | | 1 | 1 | 0 | 0.30 | 0.33 | 0.31 | 4.67 | | 1 | 1 | 100 | 1.76 | 2.05 | 1.76 | 23.31 | | 4 | 4 | 0 | 0.18 | 0.19 | 0.19 | 1.38 | | 4 | 4 | 100 | 2.54 | 2.32 | 2.07 | 4.63 | ## Raspberry Pi 3B+, Void Linux musl 32 bit 10000000 Objects transfered, results in Objects/µs | nsend | nrec | buflen | **V (gcc -O2)** | **Go (glang)** | | :---: | :---:| :---: | :---: | :---: | | 1 | 1 | 0 | 0.37 | 0.21 | | 1 | 1 | 100 | 1.03 | 0.74 | | 4 | 4 | 0 | 0.04 | 0.38 | | 4 | 4 | 100 | 2.78 | 2.63 | | 2 | 2 | 0 | 0.05 | 0.38 | | 2 | 2 | 100 | 1.26 | 0.75 |