U8String VS rust_cmd_lib

Writing performance-sensitive code in C and calling it via interop used to be the way to go during .NET Framework days but since then has become a performance trap.

Especially for small methods, calling them through interop is a deoptimization because they cannot be inlined, and involve GC frame transition (which you can suppress) as well as an indirect jump and maybe interop stub unless you are statically linking the dependency into your AOT deployment. In case the arguments are not blittable to C - marshalling too.

It is also complicates the publishing process because you have to build both .NET and C parts and then package them together, considering the matrix of [win, linux, macos] x [x64, arm64], it turns into quite an unpleasant experience.

Instead, the recommended approach is just continuing to write C# code, except with pointer and/or ref based code. This is what CoreLib itself does for the most performance-sensitive bits[0]. Naturally, it intentionally looks ugly like in Rust, but you can easily fix it with a few extension methods[1].

[0]: https://github.com/dotnet/runtime/blob/main/src/libraries/Sy...

[1]: https://github.com/U8String/U8String/blob/main/Sources/Share...

With rust_cmd_lib, you can write below code without launching shell:

zsh is also doing the variable substitution better than bash. FYI, I just released rust_cmd_lib 1.0 recently, which can do variable substitution without any quotes: https://github.com/rust-shell-script/rust_cmd_lib

With cmd_lib and rayon, it could be done in below one line code:

#!/usr/bin/env ngs # In response to https://github.com/rust-shell-script/rust_cmd_lib/blob/1f2fc2303db3f467c42589bd944b31834c8a0bca/examples/dd_test.rs # $ ./dd_test.ngs --file /dev/nvme0n1 --block_size 4096 --threads_count 4 # [LOG 2021-03-20 10:57:14 IST] Dropping caches at first # [LOG 2021-03-20 10:57:14 IST] Running command: sudo bash -c echo\ 3\ \>\ /proc/sys/vm/drop_caches # [LOG 2021-03-20 10:57:15 IST] Running with threads: 4, block size: 4096 # [LOG 2021-03-20 10:57:15 IST] Running command: sudo bash -c dd\ if=/dev/nvme0n1\ of=/dev/null\ bs=4096\ skip=1310720\ count=655360\ 2\>\&1 # [LOG 2021-03-20 10:57:15 IST] Running command: sudo bash -c dd\ if=/dev/nvme0n1\ of=/dev/null\ bs=4096\ skip=655360\ count=655360\ 2\>\&1 # [LOG 2021-03-20 10:57:15 IST] Running command: sudo bash -c dd\ if=/dev/nvme0n1\ of=/dev/null\ bs=4096\ skip=1966080\ count=655360\ 2\>\&1 # [LOG 2021-03-20 10:57:15 IST] Running command: sudo bash -c dd\ if=/dev/nvme0n1\ of=/dev/null\ bs=4096\ skip=0\ count=655360\ 2\>\&1 # [LOG 2021-03-20 10:57:22 IST] thread 0 bandwidth: 450 MB/s # [LOG 2021-03-20 10:57:22 IST] thread 1 bandwidth: 451 MB/s # [LOG 2021-03-20 10:57:22 IST] thread 2 bandwidth: 453 MB/s # [LOG 2021-03-20 10:57:22 IST] thread 3 bandwidth: 451 MB/s # [LOG 2021-03-20 10:57:22 IST] Total bandwidth: 1805 MB/s # Command line arguments automatically fed into main() # Default data size - 10G F main(file:Str, block_size:Int=4096, threads_count:Int=1, data_size:Int=10 * 1024 * 1024 * 1024) { log("Dropping caches at first") $(log: sudo bash -c "echo 3 > /proc/sys/vm/drop_caches") log("Running with threads: ${threads_count}, block size: ${block_size}") # Parallel map: each callback runs in it's own thread cnt = data_size / threads_count / block_size results = threads_count.pmap(F(i) { off = cnt * i `log: sudo bash -c "dd if=$file of=/dev/null bs=$block_size skip=$off count=$cnt 2>&1"` }) total_bandwidth = 0 results.each_idx_val(F(i, output) { # "line:" makes the first line of output as the result of `...` bandwidth = `echo $output | line: awk '/MB/ {print $10}'`.Int() log("thread ${i} bandwidth: ${bandwidth} MB/s") total_bandwidth += bandwidth }) log("Total bandwidth: $total_bandwidth MB/s") } # When running without arguments F main() { # exit() - exit code is 1 unless otherwise specified exit("Usage: ${ARGV0} --file FILE [--block_size BLOCK_SIZE] [--threads_count THREADS_COUNT] [--data_size DATA_SIZE]") }

Thanks for the inspiration from previous posts, and I tried to convert pipes.sh to rust code today. It turned out to be very straightforward with the help of rust_cmd_lib, and your can check both language versions in the project examples directory.

Yes, your concern is correct and that's why I encourage people to use macros by default: https://github.com/rust-shell-script/rust_cmd_lib#security-notes