This is an attempt at recreating the functionality of GNU Parallel, a work-stealer for the command-line, in Rust under a MIT license. The end goal will be to support much of the functionality of GNU Parallel
and then to extend the functionality further for the next generation of command-line utilities written in Rust. While functionality is important, with the application being developed in Rust, the goal is to also be as fast and efficient as possible.
A nightly version of the Rust compiler is required to build this application.
This is required to disable jemalloc
because it incurs a performance overhead on Linux systems.
In addition to using a nightly compiler, it is recommended to compile with a musl target.
rustup install nightly
rustup target add x86_64-unknown-linux-musl
rustup run nightly cargo build --release --target x86_64-unknown-linux-musl
Additionally, transparent_hugepages
causes serious performance issues with the implementation
of parallel, so it is recommended to ensure that your Linux distribution has the parameter set
to madvise
:
sudo sh -c "echo madvise > /sys/kernel/mm/transparent_hugepage/enabled"
It's a good idea to install the dash
shell as this implementation of Parallel will try to use it by default.
Dash is basically a superior implementation of sh
that's many times faster and more secure.
If dash
cannot be found, it will default to sh
. Although on Windows it will default to cmd
.
See the to-do list for features and improvements that have yet to be done. If you want to contribute, pull requests are welcome. If you have an idea for improvement which isn't listed in the to-do list, feel free to email me and I will consider implementing that idea.
~/D/parallel (master) $ seq 1 10000 | time -v /usr/bin/parallel echo > /dev/null
User time (seconds): 194.73
System time (seconds): 66.49
Percent of CPU this job got: 230%
Elapsed (wall clock) time (h:mm:ss or m:ss): 1:53.08
Maximum resident set size (kbytes): 16140
~/D/parallel (master) $ time -v /usr/bin/parallel cat ::: /usr/bin/* > /dev/null
User time (seconds): 71.71
System time (seconds): 27.67
Percent of CPU this job got: 222%
Elapsed (wall clock) time (h:mm:ss or m:ss): 0:44.62
Maximum resident set size (kbytes): 17576
~/D/parallel (master) $ time -v /usr/bin/parallel --joblog log echo ::: $(seq 1 1000) > /dev/null
User time (seconds): 21.27
System time (seconds): 7.44
Percent of CPU this job got: 238%
Elapsed (wall clock) time (h:mm:ss or m:ss): 0:12.05
Maximum resident set size (kbytes): 16624
It's highly recommend to compile Parallel with MUSL instead of glibc, as this reduces memory consumption in half and doubles performance.
~/D/parallel (master) $ seq 1 10000 | time -v target/release/x86_64-unknown-linux-musl/parallel echo > /dev/null
User time (seconds): 0.48
System time (seconds): 2.85
Percent of CPU this job got: 104%
Elapsed (wall clock) time (h:mm:ss or m:ss): 0:03.20
Maximum resident set size (kbytes): 1768
~/D/parallel (master) $ time -v target/release/x86_64-unknown-linux-musl/release/parallel cat ::: /usr/bin/* > /dev/null
User time (seconds): 0.96
System time (seconds): 4.61
Percent of CPU this job got: 192%
Elapsed (wall clock) time (h:mm:ss or m:ss): 0:02.89
Maximum resident set size (kbytes): 1868
~/D/parallel (master) $ time -v target/x86_64-unknown-linux-musl/release/parallel --joblog log echo ::: $(seq 1 1000) > /dev/null
User time (seconds): 0.04
System time (seconds): 0.29
Percent of CPU this job got: 82%
Elapsed (wall clock) time (h:mm:ss or m:ss): 0:00.40
Maximum resident set size (kbytes): 1768
The following syntax is supported:
```sh parallel 'echo {}' ::: * // {} will be replaced with each input found. parallel echo ::: * // If no placeholders are used, it is automatically assumed. parallel echo :::: list1 list2 list3 // Read newline-delimited arguments stored in files. parallel echo ::: arg1 ::::+ list :::+ arg2 // Interchangeably read arguments from the command line and files. parallel echo ::: 1 2 3 ::: A B C ::: D E F // Permutate the inputs. parallel echo {} {1} {2} {3.} ::: 1 2 file.mkv // {N} tokens are replaced by the Nth input argument parallel ::: "echo 1" "echo 2" "echo 3" // If no command is supplied, the input arguments become commands. parallel 'cd {}; echo Directory: {}; echo - {}' // Commands may be chained in the platform\'s shell. seq 1 10 | parallel 'echo {}' // If no input arguments are supplied, stdin will be read. seq 1 10 | parallel --pipe cat // Piping arguments to the standard input of the given command.
```
Parallel parallelizes otherwise non-parallel command-line tasks. When there are a number of commands that need to be executed, which may be executed in parallel, the Parallel application will evenly distribute tasks to all available CPU cores. There are three basic methods for how commands are supplied:
A COMMAND may be defined, followed by an which denotes that all following arguments will be usde as INPUTS for the command.
If no COMMAND is provided, then the INPUTS will be interpreted as COMMANDS.
If no INPUTS are provided, then standard input will be read for INPUTS.
Parallel groups the standard output and error of each child process so that
outputs are printed in the order that they are given, as if the tasks were
executed serially in a traditional for loop. In addition, commands are
executed in the platform's preferred shell by default, which is sh -c
on
Unix systems, and cmd /C
on Windows. This comes at a performance cost, so
it can be disabled with the --no-shell option.
Input modes are used to determine whether the following inputs are files that contain inputs or inputs themselves. Files with inputs have each input stored on a separate line, and each line is considered an entire input.When there are multiple collected lists of inputs, each individual input list will be permutated together into a single list.
:::
Denotes that the input arguments that follow are input arguments. Additionally, those arguments will be collected into a new list.
:::+
Denotes that the input arguments that follow are input arguments. Additionally, those arguments will be added to the current list.
::::
Denotes that the input arguments that follow are files with inputs. Additionally, those arguments will be collected into a new list.
::::+
Denotes that the input arguments that follow are files with inputs. Additionally, those arguments will be added to the current list.
COMMANDs are typically formed the same way that you would normally in the shell, only that you will replace your input arguments with placeholder tokens like {}, {.}, {/}, {//} and {/.}. If no tokens are provided, it is inferred that the final argument in the command will be {}. These tokens will perform text manipulation on the inputs to mangle them in the way you like. Ideas for more tokens are welcome.
Options may also be supplied to the program to change how the program operates:
less
).YYYY-MM-DD hh:mm:ss
ffmpeg is a highly useful application for converting music and videos. However, audio transcoding is limited to a a single core. If you have a large FLAC archive and you wanted to compress it into the efficient Opus codec, it would take forever with the fastest processor to complete, unless you were to take advantage of all cores in your CPU.
sh
parallel 'ffmpeg -v 0 -i "{}" -c:a libopus -b:a 128k "{.}.opus"' ::: $(find -type f -name '*.flac')
VP9 has one glaring flaw in regards to encoding: it can only use about three cores at any given point in time. If you have an eight core processor and a dozen or more episodes of a TV series to transcode, you can use the parallel program to run three jobs at the same time, provided you also have enough memory for that.
sh
vp9_params="-c:v libvpx-vp9 -tile-columns 6 -frame-parallel 1 -rc_lookahead 25 -threads 4 -speed 1 -b:v 0 -crf 18"
opus_params="-c:a libopus -b:a 128k"
parallel -j 3 'ffmpeg -v 0 -i "{}" $vp9_params $opus_params -f webm "{.}.webm"' ::: $(find -type f -name '*.mkv')
There are a number of methods that you can use to install the application. I provide binary packages for AMD64 systems that are available for download:
I have a personal Gentoo layman overlay that provides this application for installation.
A PKGBUILD is available for Arch Linux users from the AUR.
sh
rustup component add x86_64-unknown-linux-musl
wget https://github.com/mmstick/parallel/archive/master.zip
unzip master.zip
cd parallel-master
cargo build --release --target x86_64-unknown-linux-musl
sudo install target/x86_64-unknown-linux-musl/release/parallel /usr/local/bin/parallel