GNU parallel is a shell tool for executing jobs in parallel using one or more computers. A job can be a single command or a small script that has to be run for each of the lines in the input. The typical input is a list of files, a list of hosts, a list of users, a list of URLs, or a list of tables. A job can also be a command that reads from a pipe. GNU parallel can then split the input into blocks and pipe a block into each command in parallel.

If you use xargs and tee today you will find GNU parallel very easy to use as GNU parallel is written to have the same options as xargs. If you write loops in shell, you will find GNU parallel may be able to replace most of the loops and make them run faster by running several jobs in parallel.

GNU parallel makes sure output from the commands is the same output as you would get had you run the commands sequentially. This makes it possible to use output from GNU parallel as input for other programs.

For each line of input GNU parallel will execute command with the line as arguments. If no command is given, the line of input is executed. Several lines will be run in parallel. GNU parallel can often be used as a substitute for xargs or cat | bash.

Reader's guide

GNU parallel includes the 4 types of documentation: Tutorial, how-to, reference and explanation.

Tutorial

If you prefer reading a book buy GNU Parallel 2018 at https://www.lulu.com/shop/ole-tange/gnu-parallel-2018/paperback/product-23558902.html or download it at: https://doi.org/10.5281/zenodo.1146014 Read at least chapter 1+2. It should take you less than 20 minutes.

Otherwise start by watching the intro videos for a quick introduction: https://youtube.com/playlist?list=PL284C9FF2488BC6D1

If you want to dive deeper: spend a couple of hours walking through the tutorial (man parallel_tutorial). Your command line will love you for it.

How-to

You can find a lot of EXAMPLEs of use after the list of OPTIONS in man parallel (Use LESS=+/EXAMPLE: man parallel). That will give you an idea of what GNU parallel is capable of, and you may find a solution you can simply adapt to your situation.

Reference

If you need a one page printable cheat sheet you can find it on: https://www.gnu.org/software/parallel/parallel_cheat.pdf

The man page is the reference for all options.

Design discussion

If you want to know the design decisions behind GNU parallel, try: man parallel_design. This is also a good intro if you intend to change GNU parallel.

OPTIONS

command

Command to execute. If command or the following arguments contain replacement strings (such as {}) every instance will be substituted with the input. If command is given, GNU parallel solve the same tasks as xargs. If command is not given GNU parallel will behave similar to cat | sh. The command must be an executable, a script, a composed command, an alias, or a function. Bash functions: export -f the function first or use env_parallel. Bash, Csh, or Tcsh aliases: Use env_parallel. Zsh, Fish, Ksh, and Pdksh functions and aliases: Use env_parallel.

{}

Input line. This replacement string will be replaced by a full line read from the input source. The input source is normally stdin (standard input), but can also be given with -a, :::, or ::::. The replacement string {} can be changed with -I. If the command line contains no replacement strings then {} will be appended to the command line. Replacement strings are normally quoted, so special characters are not parsed by the shell. The exception is if the command starts with a replacement string; then the string is not quoted.

{.}

Input line without extension. This replacement string will be replaced by the input with the extension removed. If the input line contains . after the last /, the last . until the end of the string will be removed and {.} will be replaced with the remaining. E.g. foo.jpg becomes foo, subdir/foo.jpg becomes subdir/foo, sub.dir/foo.jpg becomes sub.dir/foo, sub.dir/bar remains sub.dir/bar. If the input line does not contain . it will remain unchanged. The replacement string {.} can be changed with --er. To understand replacement strings see {}.

{/}

Basename of input line. This replacement string will be replaced by the input with the directory part removed. The replacement string {/} can be changed with --basenamereplace. To understand replacement strings see {}.

{//}

Dirname of input line. This replacement string will be replaced by the dir of the input line. See dirname(1). The replacement string {//} can be changed with --dirnamereplace. To understand replacement strings see {}.

{/.}

Basename of input line without extension. This replacement string will be replaced by the input with the directory and extension part removed. It is a combination of {/} and {.}. The replacement string {/.} can be changed with --basenameextensionreplace. To understand replacement strings see {}.

{#}

Sequence number of the job to run. This replacement string will be replaced by the sequence number of the job being run. It contains the same number as $PARALLEL_SEQ. The replacement string {#} can be changed with --seqreplace. To understand replacement strings see {}.

{%}

Job slot number. This replacement string will be replaced by the job's slot number between 1 and number of jobs to run in parallel. There will never be 2 jobs running at the same time with the same job slot number. The replacement string {%} can be changed with --slotreplace. If the job needs to be retried (e.g using --retries or --retry-failed) the job slot is not automatically updated. You should then instead use $PARALLEL_JOBSLOT: $ do_test() { id="$3 {%}=$1 PARALLEL_JOBSLOT=$2" echo run "$id"; sleep 1 # fail if {%} is odd return `echo $1%2 | bc` } $ export -f do_test $ parallel -j3 --jl mylog do_test {%} \$PARALLEL_JOBSLOT {} ::: A B C D run A {%}=1 PARALLEL_JOBSLOT=1 run B {%}=2 PARALLEL_JOBSLOT=2 run C {%}=3 PARALLEL_JOBSLOT=3 run D {%}=1 PARALLEL_JOBSLOT=1 $ parallel --retry-failed -j3 --jl mylog do_test {%} \$PARALLEL_JOBSLOT {} ::: A B C D run A {%}=1 PARALLEL_JOBSLOT=1 run C {%}=3 PARALLEL_JOBSLOT=2 run D {%}=1 PARALLEL_JOBSLOT=3 Notice how {%} and $PARALLEL_JOBSLOT differ in the retry run of C and D. To understand replacement strings see {}.

{n}

Argument from input source n or the n'th argument. This positional replacement string will be replaced by the input from input source n (when used with -a or ::::) or with the n'th argument (when used with -N). If n is negative it refers to the n'th last argument. To understand replacement strings see {}.

{n.}

Argument from input source n or the n'th argument without extension. It is a combination of {n} and {.}. This positional replacement string will be replaced by the input from input source n (when used with -a or ::::) or with the n'th argument (when used with -N). The input will have the extension removed. To understand positional replacement strings see {n}.

{n/}

Basename of argument from input source n or the n'th argument. It is a combination of {n} and {/}. This positional replacement string will be replaced by the input from input source n (when used with -a or ::::) or with the n'th argument (when used with -N). The input will have the directory (if any) removed. To understand positional replacement strings see {n}.

{n//}

Dirname of argument from input source n or the n'th argument. It is a combination of {n} and {//}. This positional replacement string will be replaced by the dir of the input from input source n (when used with -a or ::::) or with the n'th argument (when used with -N). See dirname(1). To understand positional replacement strings see {n}.

{n/.}

Basename of argument from input source n or the n'th argument without extension. It is a combination of {n}, {/}, and {.}. This positional replacement string will be replaced by the input from input source n (when used with -a or ::::) or with the n'th argument (when used with -N). The input will have the directory (if any) and extension removed. To understand positional replacement strings see {n}.

{=perl expression=}

Replace with calculated perl expression. $_ will contain the same as {}. After evaluating perl expression $_ will be used as the value. It is recommended to only change $_ but you have full access to all of GNU parallel's internal functions and data structures. The expression must give the same result if evaluated twice - otherwise the behaviour is undefined. E.g. this will not work as expected: parallel echo {= $_= ++$wrong_counter =} ::: a b c A few convenience functions and data structures have been made:

Q(string): shell quote a string
pQ(string): perl quote a string
uq() (or uq): do not quote current replacement string
hash(val): compute B::hash(val)
total_jobs(): number of jobs in total
slot(): slot number of job
seq(): sequence number of job
@arg: the arguments
skip(): skip this job (see also --filter)
yyyy_mm_dd_hh_mm_ss()

yyyy_mm_dd_hh_mm()
yyyy_mm_dd()
yyyymmddhhmmss()
yyyymmddhhmm()
yyyymmdd()

time functions

Example: seq 10 | parallel echo {} + 1 is {= $_++ =} parallel csh -c {= $_="mkdir ".Q($_) =} ::: 12" dir seq 50 | parallel echo job {#} of {= $_=total_jobs() =} See also: --rpl --parens

{=n perl expression=}

Positional equivalent to {=perl expression=}. To understand positional replacement strings see {n}. See also: {=perl expression=} {n}.

::: arguments

Use arguments from the command line as input source instead of stdin (standard input). Unlike other options for GNU parallel ::: is placed after the command and before the arguments. The following are equivalent: (echo file1; echo file2) | parallel gzip parallel gzip ::: file1 file2 parallel gzip {} ::: file1 file2 parallel --arg-sep ,, gzip {} ,, file1 file2 parallel --arg-sep ,, gzip ,, file1 file2 parallel ::: "gzip file1" "gzip file2" To avoid treating ::: as special use --arg-sep to set the argument separator to something else. See also --arg-sep. If multiple ::: are given, each group will be treated as an input source, and all combinations of input sources will be generated. E.g. ::: 1 2 ::: a b c will result in the combinations (1,a) (1,b) (1,c) (2,a) (2,b) (2,c). This is useful for replacing nested for-loops. ::: and :::: can be mixed. So these are equivalent: parallel echo {1} {2} {3} ::: 6 7 ::: 4 5 ::: 1 2 3 parallel echo {1} {2} {3} :::: <(seq 6 7) <(seq 4 5) \ :::: <(seq 1 3) parallel -a <(seq 6 7) echo {1} {2} {3} :::: <(seq 4 5) \ :::: <(seq 1 3) parallel -a <(seq 6 7) -a <(seq 4 5) echo {1} {2} {3} \ ::: 1 2 3 seq 6 7 | parallel -a - -a <(seq 4 5) echo {1} {2} {3} \ ::: 1 2 3 seq 4 5 | parallel echo {1} {2} {3} :::: <(seq 6 7) - \ ::: 1 2 3

:::+ arguments

Like ::: but linked like --link to the previous input source. Contrary to --link, values do not wrap: The shortest input source determines the length. Example: parallel echo ::: a b c :::+ 1 2 3 ::: X Y :::+ 11 22

:::: argfiles

Another way to write -a argfile1 -a argfile2 ... ::: and :::: can be mixed. See -a, ::: and --link.

::::+ argfiles

Like :::: but linked like --link to the previous input source. Contrary to --link, values do not wrap: The shortest input source determines the length.

--null

-0

Use NUL as delimiter. Normally input lines will end in \n (newline). If they end in \0 (NUL), then use this option. It is useful for processing arguments that may contain \n (newline).

--arg-file input-file

-a input-file

Use input-file as input source. If you use this option, stdin (standard input) is given to the first process run. Otherwise, stdin (standard input) is redirected from /dev/null. If multiple -a are given, each input-file will be treated as an input source, and all combinations of input sources will be generated. E.g. The file foo contains 1 2, the file bar contains a b c. -a foo -a bar will result in the combinations (1,a) (1,b) (1,c) (2,a) (2,b) (2,c). This is useful for replacing nested for-loops. See also: --link and {n}.

--arg-file-sep sep-str: Use sep-str instead of :::: as separator string between command and argument files. Useful if :::: is used for something else by the command. See also: ::::.
--arg-sep sep-str: Use sep-str instead of ::: as separator string. Useful if ::: is used for something else by the command. Also useful if you command uses ::: but you still want to read arguments from stdin (standard input): Simply change --arg-sep to a string that is not in the command line. See also: :::.
--bar: Show progress as a progress bar. In the bar is shown: % of jobs completed, estimated seconds left, and number of jobs started. It is compatible with zenity: seq 1000 | parallel -j30 --bar (echo {};sleep 0.1) \ 2> >(perl -pe BEGIN{$/="\r";$|=1};s/\r/\n/g | zenity --progress --auto-kill) | wc
--basefile file

--bf file

file will be transferred to each sshlogin before a job is started. It will be removed if --cleanup is active. The file may be a script to run or some common base data needed for the job. Multiple --bf can be specified to transfer more basefiles. The file will be transferred the same way as --transferfile.

--basenamereplace replace-str

--bnr replace-str

Use the replacement string replace-str instead of {/} for basename of input line.

--basenameextensionreplace replace-str

--bner replace-str

Use the replacement string replace-str instead of {/.} for basename of input line without extension.

--bin binexpr: Use binexpr as binning key and bin input to the jobs. binexpr is [column number|column name] [perlexpression] e.g. 3, Address, 3 $_%=100, Address s/\D//g. Each input line is split using --colsep. The value of the column is put into $_, the perl expression is executed, the resulting value is is the job slot that will be given the line. If the value is bigger than the number of jobslots the value will be modulo number of jobslots. This is similar to --shard but the hashing algorithm is a simple modulo, which makes it predictible which jobslot will receive which value. The performance is in the order of 100K rows per second. Faster if the bincol is small (<10), slower if it is big (>100). --bin requires --pipe and a fixed numeric value for --jobs. See also: --shard, --group-by, --roundrobin.
--bg: Run command in background thus GNU parallel will not wait for completion of the command before exiting. This is the default if --semaphore is set. See also: --fg, man sem. Implies --semaphore.
--bibtex

--citation

Print the citation notice and BibTeX entry for GNU parallel, silence citation notice for all future runs, and exit. It will not run any commands. If it is impossible for you to run --citation you can instead use --will-cite, which will run commands, but which will only silence the citation notice for this single run. If you use --will-cite in scripts to be run by others you are making it harder for others to see the citation notice. The development of GNU parallel is indirectly financed through citations, so if your users do not know they should cite then you are making it harder to finance development. However, if you pay 10000 EUR, you have done your part to finance future development and should feel free to use --will-cite in scripts. If you do not want to help financing future development by letting other users see the citation notice or by paying, then please use another tool instead of GNU parallel. You can find some of the alternatives in man parallel_alternatives.

--block size

--block-size size

Size of block in bytes to read at a time. The size can be postfixed with K, M, G, T, P, k, m, g, t, or p (see UNIT PREFIX). GNU parallel tries to meet the block size but can be off by the length of one record. For performance reasons size should be bigger than a two records. GNU parallel will warn you and automatically increase the size if you choose a size that is too small. If you use -N, --block-size should be bigger than N+1 records. size defaults to 1M. When using --pipepart a negative block size is not interpreted as a blocksize but as the number of blocks each jobslot should have. So this will run 10*5 = 50 jobs in total: parallel --pipepart -a myfile --block -10 -j5 wc This is an efficient alternative to --roundrobin because data is never read by GNU parallel, but you can still have very few jobslots process a large amount of data. See --pipe and --pipepart for use of this.

--blocktimeout duration

--bt duration

Time out for reading block when using --pipe. If it takes longer than duration to read a full block, use the partial block read so far. duration must be in whole seconds, but can be expressed as floats postfixed with s, m, h, or d which would multiply the float by 1, 60, 3600, or 86400. Thus these are equivalent: --blocktimeout 100000 and --blocktimeout 1d3.5h16.6m4s.

--cat: Create a temporary file with content. Normally --pipe/--pipepart will give data to the program on stdin (standard input). With --cat GNU parallel will create a temporary file with the name in {}, so you can do: parallel --pipe --cat wc {}. Implies --pipe unless --pipepart is used. See also: --fifo.
--cleanup: Remove transferred files. --cleanup will remove the transferred files on the remote computer after processing is done. find log -name *gz | parallel \ --sshlogin server.example.com --transferfile {} \ --return {.}.bz2 --cleanup "zcat {} | bzip -9 >{.}.bz2" With --transferfile {} the file transferred to the remote computer will be removed on the remote computer. Directories on the remote computer containing the file will be removed if they are empty. With --return the file transferred from the remote computer will be removed on the remote computer. Directories on the remote computer containing the file will be removed if they are empty. --cleanup is ignored when not used with --transferfile or --return.
--colsep regexp

-C regexp

Column separator. The input will be treated as a table with regexp separating the columns. The n'th column can be accessed using {n} or {n.}. E.g. {3} is the 3rd column. If there are more input sources, each input source will be separated, but the columns from each input source will be linked (see --link). parallel --colsep - echo {4} {3} {2} {1} \ ::: A-B C-D ::: e-f g-h --colsep implies --trim rl, which can be overridden with --trim n. regexp is a Perl Regular Expression: https://perldoc.perl.org/perlre.html

--compress: Compress temporary files. If the output is big and very compressible this will take up less disk space in $TMPDIR and possibly be faster due to less disk I/O. GNU parallel will try pzstd, lbzip2, pbzip2, zstd, pigz, lz4, lzop, plzip, lzip, lrz, gzip, pxz, lzma, bzip2, xz, clzip, in that order, and use the first available.
--compress-program prg

--decompress-program prg

Use prg for (de)compressing temporary files. It is assumed that prg -dc will decompress stdin (standard input) to stdout (standard output) unless --decompress-program is given.

--csv: Treat input as CSV-format. --colsep sets the field delimiter. It works very much like --colsep except it deals correctly with quoting: echo "1 big, 2 small","2""x4"" plank",12.34 | parallel --csv echo {1} of {2} at {3} Even quoted newlines are parsed correctly: (echo "Start of field 1 with newline echo Line 2 in field 1";value 2) | parallel --csv --colsep ; echo Field 1: {1} Field 2: {2} When used with --pipe only pass full CSV-records.
--ctag str: Color tag. See --tag.
--ctagstring str: Color tagstring. See --tagstring.
--delay mytime: Delay starting next job by mytime. GNU parallel will pause mytime after starting each job. mytime is normally in seconds, but can be floats postfixed with s, m, h, or d which would multiply the float by 1, 60, 3600, or 86400. Thus these are equivalent: --delay 100000 and --delay 1d3.5h16.6m4s. If you append 'auto' to mytime (e.g. 13m3sauto) GNU parallel will automatically try to find the optimal value: If a job fails, mytime is doubled. If a job succeeds, mytime is decreased by 10%.
--delimiter delim

-d delim

Input items are terminated by delim. Quotes and backslash are not special; every character in the input is taken literally. Disables the end-of-file string, which is treated like any other argument. The specified delimiter may be characters, C-style character escapes such as \n, or octal or hexadecimal escape codes. Octal and hexadecimal escape codes are understood as for the printf command. Multibyte characters are not supported.

--dirnamereplace replace-str

--dnr replace-str

Use the replacement string replace-str instead of {//} for dirname of input line.

--dry-run: Print the job to run on stdout (standard output), but do not run the job. Use -v -v to include the wrapping that GNU parallel generates (for remote jobs, --tmux, --nice, --pipe, --pipepart, --fifo and --cat). Do not count on this literally, though, as the job may be scheduled on another computer or the local computer if : is in the list.
-E eof-str: Set the end of file string to eof-str. If the end of file string occurs as a line of input, the rest of the input is not read. If neither -E nor -e is used, no end of file string is used.
--eof[=eof-str]

-e[eof-str]

This option is a synonym for the -E option. Use -E instead, because it is POSIX compliant for xargs while this option is not. If eof-str is omitted, there is no end of file string. If neither -E nor -e is used, no end of file string is used.

--embed

Embed GNU parallel in a shell script. If you need to distribute your script to someone who does not want to install GNU parallel you can embed GNU parallel in your own shell script: parallel --embed > new_script After which you add your code at the end of new_script. This is tested on ash, bash, dash, ksh, sh, and zsh.

--env var

Copy environment variable var. This will copy var to the environment that the command is run in. This is especially useful for remote execution. In Bash var can also be a Bash function - just remember to export -f the function, see command. The variable '_' is special. It will copy all exported environment variables except for the ones mentioned in ~/.parallel/ignored_vars. To copy the full environment (both exported and not exported variables, arrays, and functions) use env_parallel. See also: --record-env, --session.

--eta

Show the estimated number of seconds before finishing. This forces GNU parallel to read all jobs before starting to find the number of jobs. GNU parallel normally only reads the next job to run. The estimate is based on the runtime of finished jobs, so the first estimate will only be shown when the first job has finished. Implies --progress. See also: --bar, --progress.

--fg

Run command in foreground. With --tmux and --tmuxpane GNU parallel will start tmux in the foreground. With --semaphore GNU parallel will run the command in the foreground (opposite --bg), and wait for completion of the command before exiting. See also: --bg, man sem.

--fifo

Create a temporary fifo with content. Normally --pipe and --pipepart will give data to the program on stdin (standard input). With --fifo GNU parallel will create a temporary fifo with the name in {}, so you can do: parallel --pipe --fifo wc {}. Beware: If data is not read from the fifo, the job will block forever. Implies --pipe unless --pipepart is used. See also: --cat.

--filter filter

Only run jobs where filter is true. filter can contain replacement strings and Perl code. Example: parallel --filter {1} < {2}+1 echo ::: {1..3} ::: {1..3} Outputs: 1,1 1,2 1,3 2,2 2,3 3,3

--filter-hosts

Remove down hosts. For each remote host: check that login through ssh works. If not: do not use this host. For performance reasons, this check is performed only at the start and every time --sshloginfile is changed. If an host goes down after the first check, it will go undetected until --sshloginfile is changed; --retries can be used to mitigate this. Currently you can not put --filter-hosts in a profile, $PARALLEL, /etc/parallel/config or similar. This is because GNU parallel uses GNU parallel to compute this, so you will get an infinite loop. This will likely be fixed in a later release.

--gnu

Behave like GNU parallel. This option historically took precedence over --tollef. The --tollef option is now retired, and therefore may not be used. --gnu is kept for compatibility.

--group

Group output. Output from each job is grouped together and is only printed when the command is finished. Stdout (standard output) first followed by stderr (standard error). This takes in the order of 0.5ms per job and depends on the speed of your disk for larger output. It can be disabled with -u, but this means output from different commands can get mixed. --group is the default. Can be reversed with -u. See also: --line-buffer --ungroup

--group-by val

Group input by value. Combined with --pipe/--pipepart --group-by groups lines with the same value into a record. The value can be computed from the full line or from a single column. val can be:

column number: Use the value in the column numbered.
column name: Treat the first line as a header and use the value in the column named. (Not supported with --pipepart).
perl expression: Run the perl expression and use $_ as the value.
column number perl expression: Put the value of the column put in $_, run the perl expression, and use $_ as the value.
column name perl expression: Put the value of the column put in $_, run the perl expression, and use $_ as the value. (Not supported with --pipepart).

Example: UserID, Consumption 123, 1 123, 2 12-3, 1 221, 3 221, 1 2/21, 5 If you want to group 123, 12-3, 221, and 2/21 into 4 records and pass one record at a time to wc: tail -n +2 table.csv | \ parallel --pipe --colsep , --group-by 1 -kN1 wc Make GNU parallel treat the first line as a header: cat table.csv | \ parallel --pipe --colsep , --header : --group-by 1 -kN1 wc Address column by column name: cat table.csv | \ parallel --pipe --colsep , --header : --group-by UserID -kN1 wc If 12-3 and 123 are really the same UserID, remove non-digits in UserID when grouping: cat table.csv | parallel --pipe --colsep , --header : \ --group-by UserID s/\D//g -kN1 wc See also: --shard, --roundrobin.

--help

-h

Print a summary of the options to GNU parallel and exit.

--halt-on-error val

--halt val

When should GNU parallel terminate? In some situations it makes no sense to run all jobs. GNU parallel should simply give up as soon as a condition is met. val defaults to never, which runs all jobs no matter what. val can also take on the form of when,why. when can be 'now' which means kill all running jobs and halt immediately, or it can be 'soon' which means wait for all running jobs to complete, but start no new jobs. why can be 'fail=X', 'fail=Y%', 'success=X', 'success=Y%', 'done=X', or 'done=Y%' where X is the number of jobs that has to fail, succeed, or be done before halting, and Y is the percentage of jobs that has to fail, succeed, or be done before halting. Example:

--halt now,fail=1

exit when the first job fails. Kill running jobs.

--halt soon,fail=3

exit when 3 jobs fail, but wait for running jobs to complete.

--halt soon,fail=3%

exit when 3% of the jobs have failed, but wait for running jobs to complete.

--halt now,success=1

exit when a job succeeds. Kill running jobs.

--halt soon,success=3

exit when 3 jobs succeeds, but wait for running jobs to complete.

--halt now,success=3%

exit when 3% of the jobs have succeeded. Kill running jobs.

--halt now,done=1

exit when one of the jobs finishes. Kill running jobs.

--halt soon,done=3

exit when 3 jobs finishes, but wait for running jobs to complete.

--halt now,done=3%

exit when 3% of the jobs have finished. Kill running jobs.

For backwards compatibility these also work:

never

soon,fail=1

now,fail=1

-1

soon,success=1

-2

now,success=1

1-99%

soon,fail=1-99%

--header regexp: Use regexp as header. For normal usage the matched header (typically the first line: --header '.*\n') will be split using --colsep (which will default to '\t') and column names can be used as replacement variables: {column name}, {column name/}, {column name//}, {column name/.}, {column name.}, {=column name perl expression =}, .. For --pipe the matched header will be prepended to each output. --header : is an alias for --header '.*\n'. If regexp is a number, it is a fixed number of lines.
--hostgroups

--hgrp

Enable hostgroups on arguments. If an argument contains '@' the string after '@' will be removed and treated as a list of hostgroups on which this job is allowed to run. If there is no --sshlogin with a corresponding group, the job will run on any hostgroup. Example: parallel --hostgroups \ --sshlogin @grp1/myserver1 -S @grp1+grp2/myserver2 \ --sshlogin @grp3/myserver3 \ echo ::: my_grp1_arg@grp1 arg_for_grp2@grp2 third@grp1+grp3 my_grp1_arg may be run on either myserver1 or myserver2, third may be run on either myserver1 or myserver3, but arg_for_grp2 will only be run on myserver2. See also: --sshlogin, $PARALLEL_HOSTGROUPS, $PARALLEL_ARGHOSTGROUPS.

-I replace-str: Use the replacement string replace-str instead of {}.
--replace[=replace-str]

-i[replace-str]

This option is a synonym for -Ireplace-str if replace-str is specified, and for -I {} otherwise. This option is deprecated; use -I instead.

--joblog logfile: Logfile for executed jobs. Save a list of the executed jobs to logfile in the following TAB separated format: sequence number, sshlogin, start time as seconds since epoch, run time in seconds, bytes in files transferred, bytes in files returned, exit status, signal, and command run. For --pipe bytes transferred and bytes returned are number of input and output of bytes. If logfile is prepended with '+' log lines will be appended to the logfile. To convert the times into ISO-8601 strict do: cat logfile | perl -a -F"\t" -ne \ chomp($F[2]=`date -d \@$F[2] +%FT%T`); print join("\t",@F) If the host is long, you can use column -t to pretty print it: cat joblog | column -t See also: --resume --resume-failed.
--jobs N

-j N
--max-procs N
-P N

Number of jobslots on each machine. Run up to N jobs in parallel. 0 means as many as possible (this can take a while to determine). Default is 100% which will run one job per CPU on each machine. Due to a bug -j 0 will also evaluate replacement strings twice up to the number of joblots: # This will not count from 1 but from number-of-jobslots seq 10000 | parallel -j0 echo {= $_ = $foo++; =} | head # This will count from 1 seq 10000 | parallel -j100 echo {= $_ = $foo++; =} | head If --semaphore is set, the default is 1 thus making a mutex.

--jobs +N

-j +N
--max-procs +N
-P +N

Add N to the number of CPUs. Run this many jobs in parallel. See also --use-cores-instead-of-threads and --use-sockets-instead-of-threads.

--jobs -N

-j -N
--max-procs -N
-P -N

Subtract N from the number of CPUs. Run this many jobs in parallel. If the evaluated number is less than 1 then 1 will be used. See also --use-cores-instead-of-threads and --use-sockets-instead-of-threads.

--jobs N%

-j N%
--max-procs N%
-P N%

Multiply N% with the number of CPUs. Run this many jobs in parallel. See also --use-cores-instead-of-threads and --use-sockets-instead-of-threads.

--jobs procfile

-j procfile
--max-procs procfile
-P procfile

Read parameter from file. Use the content of procfile as parameter for -j. E.g. procfile could contain the string 100% or +2 or 10. If procfile is changed when a job completes, procfile is read again and the new number of jobs is computed. If the number is lower than before, running jobs will be allowed to finish but new jobs will not be started until the wanted number of jobs has been reached. This makes it possible to change the number of simultaneous running jobs while GNU parallel is running.

--keep-order

-k

Keep sequence of output same as the order of input. Normally the output of a job will be printed as soon as the job completes. Try this to see the difference: parallel -j4 sleep {}\; echo {} ::: 2 1 4 3 parallel -j4 -k sleep {}\; echo {} ::: 2 1 4 3 If used with --onall or --nonall the output will grouped by sshlogin in sorted order. If used with --pipe --roundrobin and the same input, the jobslots will get the same blocks in the same order in every run. -k only affects the order in which the output is printed - not the order in which jobs are run.

-L recsize: When used with --pipe: Read records of recsize. When used otherwise: Use at most recsize nonblank input lines per command line. Trailing blanks cause an input line to be logically continued on the next input line. -L 0 means read one line, but insert 0 arguments on the command line. recsize can be postfixed with K, M, G, T, P, k, m, g, t, or p (see UNIT PREFIX). Implies -X unless -m, --xargs, or --pipe is set.
--max-lines[=recsize]

-l[recsize]

When used with --pipe: Read records of recsize lines. When used otherwise: Synonym for the -L option. Unlike -L, the recsize argument is optional. If recsize is not specified, it defaults to one. The -l option is deprecated since the POSIX standard specifies -L instead. -l 0 is an alias for -l 1. Implies -X unless -m, --xargs, or --pipe is set.

--limit "command args"

Dynamic job limit. Before starting a new job run command with args. The exit value of command determines what GNU parallel will do:

Below limit. Start another job.
Over limit. Start no jobs.
Way over limit. Kill the youngest job.

You can use any shell command. There are 3 predefined commands:

"io n": Limit for I/O. The amount of disk I/O will be computed as a value 0-100, where 0 is no I/O and 100 is at least one disk is 100% saturated.
"load n": Similar to --load.
"mem n": Similar to --memfree.

--line-buffer

--lb

Buffer output on line basis. --group will keep the output together for a whole job. --ungroup allows output to mixup with half a line coming from one job and half a line coming from another job. --line-buffer fits between these two: GNU parallel will print a full line, but will allow for mixing lines of different jobs. --line-buffer takes more CPU power than both --group and --ungroup, but can be much faster than --group if the CPU is not the limiting factor. Normally --line-buffer does not buffer on disk, and can thus process an infinite amount of data, but it will buffer on disk when combined with: --keep-order, --results, --compress, and --files. This will make it as slow as --group and will limit output to the available disk space. With --keep-order --line-buffer will output lines from the first job continuously while it is running, then lines from the second job while that is running. It will buffer full lines, but jobs will not mix. Compare: parallel -j0 echo {};sleep {};echo {} ::: 1 3 2 4 parallel -j0 --lb echo {};sleep {};echo {} ::: 1 3 2 4 parallel -j0 -k --lb echo {};sleep {};echo {} ::: 1 3 2 4 See also: --group --ungroup

--xapply

--link

Link input sources. Read multiple input sources like xapply. If multiple input sources are given, one argument will be read from each of the input sources. The arguments can be accessed in the command as {1} .. {n}, so {1} will be a line from the first input source, and {6} will refer to the line with the same line number from the 6th input source. Compare these two: parallel echo {1} {2} ::: 1 2 3 ::: a b c parallel --link echo {1} {2} ::: 1 2 3 ::: a b c Arguments will be recycled if one input source has more arguments than the others: parallel --link echo {1} {2} {3} \ ::: 1 2 ::: I II III ::: a b c d e f g See also --header, :::+, ::::+.

--load max-load: Do not start new jobs on a given computer unless the number of running processes on the computer is less than max-load. max-load uses the same syntax as --jobs, so 100% for one per CPU is a valid setting. Only difference is 0 which is interpreted as 0.01.
--controlmaster

-M

Use ssh's ControlMaster to make ssh connections faster. Useful if jobs run remote and are very fast to run. This is disabled for sshlogins that specify their own ssh command.

-m: Multiple arguments. Insert as many arguments as the command line length permits. If multiple jobs are being run in parallel: distribute the arguments evenly among the jobs. Use -j1 or --xargs to avoid this. If {} is not used the arguments will be appended to the line. If {} is used multiple times each {} will be replaced with all the arguments. Support for -m with --sshlogin is limited and may fail. See also -X for context replace. If in doubt use -X as that will most likely do what is needed.
--memfree size: Minimum memory free when starting another job. The size can be postfixed with K, M, G, T, P, k, m, g, t, or p (see UNIT PREFIX). If the jobs take up very different amount of RAM, GNU parallel will only start as many as there is memory for. If less than size bytes are free, no more jobs will be started. If less than 50% size bytes are free, the youngest job will be killed, and put back on the queue to be run later. --retries must be set to determine how many times GNU parallel should retry a given job. See also: --memsuspend
--memsuspend size: Suspend jobs when there is less than 2 * size memory free. The size can be postfixed with K, M, G, T, P, k, m, g, t, or p (see UNIT PREFIX). If the available memory falls below 2 * size, GNU parallel will suspend some of the running jobs. If the available memory falls below size, only one job will be running. If a single job takes up at most size RAM, all jobs will complete without running out of memory. If you have swap available, you can usually lower size to around half the size of a single job - with the slight risk of swapping a little. Jobs will be resumed when more RAM is available - typically when the oldest job completes. --memsuspend only works on local jobs because there is no obvious way to suspend remote jobs. See also: --memfree
--minversion version: Print the version GNU parallel and exit. If the current version of GNU parallel is less than version the exit code is 255. Otherwise it is 0. This is useful for scripts that depend on features only available from a certain version of GNU parallel.
--max-args=max-args

-n max-args

Use at most max-args arguments per command line. Fewer than max-args arguments will be used if the size (see the -s option) is exceeded, unless the -x option is given, in which case GNU parallel will exit. -n 0 means read one argument, but insert 0 arguments on the command line. max-args can be postfixed with K, M, G, T, P, k, m, g, t, or p (see UNIT PREFIX). Implies -X unless -m is set.

--max-replace-args=max-args

-N max-args

Use at most max-args arguments per command line. Like -n but also makes replacement strings {1} .. {max-args} that represents argument 1 .. max-args. If too few args the {n} will be empty. -N 0 means read one argument, but insert 0 arguments on the command line. This will set the owner of the homedir to the user: tr : \n < /etc/passwd | parallel -N7 chown {1} {6} Implies -X unless -m or --pipe is set. max-args can be postfixed with K, M, G, T, P, k, m, g, t, or p (see UNIT PREFIX). When used with --pipe -N is the number of records to read. This is somewhat slower than --block.

--nonall: --onall with no arguments. Run the command on all computers given with --sshlogin but take no arguments. GNU parallel will log into --jobs number of computers in parallel and run the job on the computer. -j adjusts how many computers to log into in parallel. This is useful for running the same command (e.g. uptime) on a list of servers.
--onall: Run all the jobs on all computers given with --sshlogin. GNU parallel will log into --jobs number of computers in parallel and run one job at a time on the computer. The order of the jobs will not be changed, but some computers may finish before others. When using --group the output will be grouped by each server, so all the output from one server will be grouped together. --joblog will contain an entry for each job on each server, so there will be several job sequence 1.
--output-as-files

--outputasfiles
--files

Instead of printing the output to stdout (standard output) the output of each job is saved in a file and the filename is then printed. See also: --results

--pipe

--spreadstdin

Spread input to jobs on stdin (standard input). Read a block of data from stdin (standard input) and give one block of data as input to one job. The block size is determined by --block. The strings --recstart and --recend tell GNU parallel how a record starts and/or ends. The block read will have the final partial record removed before the block is passed on to the job. The partial record will be prepended to next block. If --recstart is given this will be used to split at record start. If --recend is given this will be used to split at record end. If both --recstart and --recend are given both will have to match to find a split position. If neither --recstart nor --recend are given --recend defaults to '\n'. To have no record separator use --recend "". --files is often used with --pipe. --pipe maxes out at around 1 GB/s input, and 100 MB/s output. If performance is important use --pipepart. See also: --recstart, --recend, --fifo, --cat, --pipepart, --files.

--pipepart

Pipe parts of a physical file. --pipepart works similar to --pipe, but is much faster. --pipepart has a few limitations:

The file must be a normal file or a block device (technically it must be seekable) and must be given using -a or ::::. The file cannot be a pipe, a fifo, or a stream as they are not seekable. If using a block device with lot of NUL bytes, remember to set --recend ''.
Record counting (-N) and line counting (-L/-l) do not work. Instead use --recstart and --recend to determine where records end.

--plain

Ignore any --profile, $PARALLEL, and ~/.parallel/config to get full control on the command line (used by GNU parallel internally when called with --sshlogin).

--plus

Activate additional replacement strings: {+/} {+.} {+..} {+...} {..} {...} {/..} {/...} {##}. The idea being that '{+foo}' matches the opposite of '{foo}' and {} = {+/}/{/} = {.}.{+.} = {+/}/{/.}.{+.} = {..}.{+..} = {+/}/{/..}.{+..} = {...}.{+...} = {+/}/{/...}.{+...} {##} is the total number of jobs to be run. It is incompatible with -X/-m/--xargs. {0%} zero-padded jobslot. {0#} zero-padded sequence number. {choose_k} is inspired by n choose k: Given a list of n elements, choose k. k is the number of input sources and n is the number of arguments in an input source. The content of the input sources must be the same and the arguments must be unique. Shorthands for variables: {slot} $PARALLEL_JOBSLOT (see {%}) {sshlogin} $PARALLEL_SSHLOGIN {host} $PARALLEL_SSHHOST {agrp} $PARALLEL_ARGHOSTGROUPS {hgrp} $PARALLEL_HOSTGROUPS The following dynamic replacement strings are also activated. They are inspired by bash's parameter expansion: {:-str} str if the value is empty {:num} remove the first num characters {:num1:num2} characters from num1 to num2 {#regexp} remove prefix regexp (non-greedy) {##regexp} remove prefix regexp (greedy) {%regexp} remove postfix regexp (non-greedy) {%%regexp} remove postfix regexp (greedy) {/regexp/str} replace regexp with str {^str} uppercase str if found at the start {^^str} uppercase str {,str} lowercase str if found at the start {,,str} lowercase str

--progress

Show progress of computations. List the computers involved in the task with number of CPUs detected and the max number of jobs to run. After that show progress for each computer: number of running jobs, number of completed jobs, and percentage of all jobs done by this computer. The percentage will only be available after all jobs have been scheduled as GNU parallel only read the next job when ready to schedule it - this is to avoid wasting time and memory by reading everything at startup. By sending GNU parallel SIGUSR2 you can toggle turning on/off --progress on a running GNU parallel process. See also: --eta and --bar.

--max-line-length-allowed

Print the maximal number of characters allowed on the command line and exit (used by GNU parallel itself to determine the line length on remote computers).

--number-of-cpus (obsolete)

Print the number of physical CPU cores and exit.

--number-of-cores

Print the number of physical CPU cores and exit (used by GNU parallel itself to determine the number of physical CPU cores on remote computers).

--number-of-sockets

Print the number of filled CPU sockets and exit (used by GNU parallel itself to determine the number of filled CPU sockets on remote computers).

--number-of-threads

Print the number of hyperthreaded CPU cores and exit (used by GNU parallel itself to determine the number of hyperthreaded CPU cores on remote computers).

--no-keep-order

Overrides an earlier --keep-order (e.g. if set in ~/.parallel/config).

--nice niceness

Run the command at this niceness. By default GNU parallel will run jobs at the same nice level as GNU parallel is started - both on the local machine and remote servers, so you are unlikely to ever use this option. Setting --nice will override this nice level. If the nice level is smaller than the current nice level, it will only affect remote jobs (e.g. if current level is 10 then --nice 5 will cause local jobs to be run at level 10, but remote jobs run at nice level 5).

--interactive

-p

Prompt the user about whether to run each command line and read a line from the terminal. Only run the command line if the response starts with 'y' or 'Y'. Implies -t.

--parens parensstring: Define start and end parenthesis for {= perl expression =}. The left and the right parenthesis can be multiple characters and are assumed to be the same length. The default is {==} giving {= as the start parenthesis and =} as the end parenthesis. Another useful setting is ,,,, which would make both parenthesis ,,: parallel --parens ,,,, echo foo is ,,s/I/O/g,, ::: FII See also: --rpl {= perl expression =}
--profile profilename

-J profilename

Use profile profilename for options. This is useful if you want to have multiple profiles. You could have one profile for running jobs in parallel on the local computer and a different profile for running jobs on remote computers. See the section PROFILE FILES for examples. profilename corresponds to the file ~/.parallel/profilename. You can give multiple profiles by repeating --profile. If parts of the profiles conflict, the later ones will be used. Default: config

--quote

-q

Quote command. If your command contains special characters that should not be interpreted by the shell (e.g. ; \ | *), use --quote to escape these. The command must be a simple command (see man bash) without redirections and without variable assignments. See the section QUOTING. Most people will not need this. Quoting is disabled by default.

--no-run-if-empty

-r

If the stdin (standard input) only contains whitespace, do not run the command. If used with --pipe this is slow.

--noswap: Do not start new jobs on a given computer if there is both swap-in and swap-out activity. The swap activity is only sampled every 10 seconds as the sampling takes 1 second to do. Swap activity is computed as (swap-in)*(swap-out) which in practice is a good value: swapping out is not a problem, swapping in is not a problem, but both swapping in and out usually indicates a problem. --memfree and --memsuspend may give better results, so try using those first.
--record-env: Record current environment variables in ~/.parallel/ignored_vars. This is useful before using --env _. See also: --env, --session.
--recstart startstring

--recend endstring

If --recstart is given startstring will be used to split at record start. If --recend is given endstring will be used to split at record end. If both --recstart and --recend are given the combined string endstringstartstring will have to match to find a split position. This is useful if either startstring or endstring match in the middle of a record. If neither --recstart nor --recend are given then --recend defaults to '\n'. To have no record separator use --recend "". --recstart and --recend are used with --pipe. Use --regexp to interpret --recstart and --recend as regular expressions. This is slow, however.

--regexp: Use --regexp to interpret --recstart and --recend as regular expressions. This is slow, however.
--remove-rec-sep

--removerecsep
--rrs

Remove the text matched by --recstart and --recend before piping it to the command. Only used with --pipe.

--results name

--res name

Save the output into files. Simple string output dir If name does not contain replacement strings and does not end in .csv/.tsv, the output will be stored in a directory tree rooted at name. Within this directory tree, each command will result in three files: name/<ARGS>/stdout and name/<ARGS>/stderr, name/<ARGS>/seq, where <ARGS> is a sequence of directories representing the header of the input source (if using --header :) or the number of the input source and corresponding values. E.g: parallel --header : --results foo echo {a} {b} \ ::: a I II ::: b III IIII will generate the files: foo/a/II/b/III/seq foo/a/II/b/III/stderr foo/a/II/b/III/stdout foo/a/II/b/IIII/seq foo/a/II/b/IIII/stderr foo/a/II/b/IIII/stdout foo/a/I/b/III/seq foo/a/I/b/III/stderr foo/a/I/b/III/stdout foo/a/I/b/IIII/seq foo/a/I/b/IIII/stderr foo/a/I/b/IIII/stdout and parallel --results foo echo {1} {2} ::: I II ::: III IIII will generate the files: foo/1/II/2/III/seq foo/1/II/2/III/stderr foo/1/II/2/III/stdout foo/1/II/2/IIII/seq foo/1/II/2/IIII/stderr foo/1/II/2/IIII/stdout foo/1/I/2/III/seq foo/1/I/2/III/stderr foo/1/I/2/III/stdout foo/1/I/2/IIII/seq foo/1/I/2/IIII/stderr foo/1/I/2/IIII/stdout CSV file output If name ends in .csv/.tsv the output will be a CSV-file named name. .csv gives a comma separated value file. .tsv gives a TAB separated value file. -.csv/-.tsv are special: It will give the file on stdout (standard output). JSON file output If name ends in .json the output will be a JSON-file named name. -.json is special: It will give the file on stdout (standard output). Replacement string output file If name contains a replacement string and the replaced result does not end in /, then the standard output will be stored in a file named by this result. Standard error will be stored in the same file name with '.err' added, and the sequence number will be stored in the same file name with '.seq' added. E.g. parallel --results my_{} echo ::: foo bar baz will generate the files: my_bar my_bar.err my_bar.seq my_baz my_baz.err my_baz.seq my_foo my_foo.err my_foo.seq Replacement string output dir If name contains a replacement string and the replaced result ends in /, then output files will be stored in the resulting dir. E.g. parallel --results my_{}/ echo ::: foo bar baz will generate the files: my_bar/seq my_bar/stderr my_bar/stdout my_baz/seq my_baz/stderr my_baz/stdout my_foo/seq my_foo/stderr my_foo/stdout See also: --files, --tag, --header, --joblog.

--resume: Resumes from the last unfinished job. By reading --joblog or the --results dir GNU parallel will figure out the last unfinished job and continue from there. As GNU parallel only looks at the sequence numbers in --joblog then the input, the command, and --joblog all have to remain unchanged; otherwise GNU parallel may run wrong commands. See also: --joblog, --results, --resume-failed, --retries.
--resume-failed: Retry all failed and resume from the last unfinished job. By reading --joblog GNU parallel will figure out the failed jobs and run those again. After that it will resume last unfinished job and continue from there. As GNU parallel only looks at the sequence numbers in --joblog then the input, the command, and --joblog all have to remain unchanged; otherwise GNU parallel may run wrong commands. See also: --joblog, --resume, --retry-failed, --retries.
--retry-failed: Retry all failed jobs in joblog. By reading --joblog GNU parallel will figure out the failed jobs and run those again. --retry-failed ignores the command and arguments on the command line: It only looks at the joblog. Differences between --resume, --resume-failed, --retry-failed In this example exit {= $_%=2 =} will cause every other job to fail. timeout -k 1 4 parallel --joblog log -j10 \ sleep {}; exit {= $_%=2 =} ::: {10..1} 4 jobs completed. 2 failed: Seq [...] Exitval Signal Command 10 [...] 1 0 sleep 1; exit 1 9 [...] 0 0 sleep 2; exit 0 8 [...] 1 0 sleep 3; exit 1 7 [...] 0 0 sleep 4; exit 0 --resume does not care about the Exitval, but only looks at Seq. If the Seq is run, it will not be run again. So if needed, you can change the command for the seqs not run yet: parallel --resume --joblog log -j10 \ sleep .{}; exit {= $_%=2 =} ::: {10..1} Seq [...] Exitval Signal Command [... as above ...] 1 [...] 0 0 sleep .10; exit 0 6 [...] 1 0 sleep .5; exit 1 5 [...] 0 0 sleep .6; exit 0 4 [...] 1 0 sleep .7; exit 1 3 [...] 0 0 sleep .8; exit 0 2 [...] 1 0 sleep .9; exit 1 --resume-failed cares about the Exitval, but also only looks at Seq to figure out which commands to run. Again this means you can change the command, but not the arguments. It will run the failed seqs and the seqs not yet run: parallel --resume-failed --joblog log -j10 \ echo {};sleep .{}; exit {= $_%=3 =} ::: {10..1} Seq [...] Exitval Signal Command [... as above ...] 10 [...] 1 0 echo 1;sleep .1; exit 1 8 [...] 0 0 echo 3;sleep .3; exit 0 6 [...] 2 0 echo 5;sleep .5; exit 2 4 [...] 1 0 echo 7;sleep .7; exit 1 2 [...] 0 0 echo 9;sleep .9; exit 0 --retry-failed cares about the Exitval, but takes the command from the joblog. It ignores any arguments or commands given on the command line: parallel --retry-failed --joblog log -j10 this part is ignored Seq [...] Exitval Signal Command [... as above ...] 10 [...] 1 0 echo 1;sleep .1; exit 1 6 [...] 2 0 echo 5;sleep .5; exit 2 4 [...] 1 0 echo 7;sleep .7; exit 1 See also: --joblog, --resume, --resume-failed, --retries.
--retries n: If a job fails, retry it on another computer on which it has not failed. Do this n times. If there are fewer than n computers in --sshlogin GNU parallel will re-use all the computers. This is useful if some jobs fail for no apparent reason (such as network failure).
--return filename: Transfer files from remote computers. --return is used with --sshlogin when the arguments are files on the remote computers. When processing is done the file filename will be transferred from the remote computer using rsync and will be put relative to the default login dir. E.g. echo foo/bar.txt | parallel --return {.}.out \ --sshlogin server.example.com touch {.}.out This will transfer the file $HOME/foo/bar.out from the computer server.example.com to the file foo/bar.out after running touch foo/bar.out on server.example.com. parallel -S server --trc out/./{}.out touch {}.out ::: in/file This will transfer the file in/file.out from the computer server.example.com to the files out/in/file.out after running touch in/file.out on server. echo /tmp/foo/bar.txt | parallel --return {.}.out \ --sshlogin server.example.com touch {.}.out This will transfer the file /tmp/foo/bar.out from the computer server.example.com to the file /tmp/foo/bar.out after running touch /tmp/foo/bar.out on server.example.com. Multiple files can be transferred by repeating the option multiple times: echo /tmp/foo/bar.txt | parallel \ --sshlogin server.example.com \ --return {.}.out --return {.}.out2 touch {.}.out {.}.out2 --return is often used with --transferfile and --cleanup. --return is ignored when used with --sshlogin : or when not used with --sshlogin. For details on transferring see --transferfile.
--round-robin

--round

Normally --pipe will give a single block to each instance of the command. With --roundrobin all blocks will at random be written to commands already running. This is useful if the command takes a long time to initialize. --keep-order will not work with --roundrobin as it is impossible to track which input block corresponds to which output. --roundrobin implies --pipe, except if --pipepart is given. See also: --group-by, --shard.

--rpl 'tag perl expression': Use tag as a replacement string for perl expression. This makes it possible to define your own replacement strings. GNU parallel's 7 replacement strings are implemented as: --rpl {} --rpl {#} 1 $_=$job->seq() --rpl {%} 1 $_=$job->slot() --rpl {/} s:.*/:: --rpl {//} $Global::use{"File::Basename"} ||= eval "use File::Basename; 1;"; $_ = dirname($_); --rpl {/.} s:.*/::; s:\.[^/.]+$::; --rpl {.} s:\.[^/.]+$:: The --plus replacement strings are implemented as: --rpl {+/} s:/[^/]*$:: --rpl {+.} s:.*\.:: --rpl {+..} s:.*\.([^.]*\.):$1: --rpl {+...} s:.*\.([^.]*\.[^.]*\.):$1: --rpl {..} s:\.[^/.]+$::; s:\.[^/.]+$:: --rpl {...} s:\.[^/.]+$::; s:\.[^/.]+$::; s:\.[^/.]+$:: --rpl {/..} s:.*/::; s:\.[^/.]+$::; s:\.[^/.]+$:: --rpl {/...} s:.*/::;s:\.[^/.]+$::;s:\.[^/.]+$::;s:\.[^/.]+$:: --rpl {##} $_=total_jobs() --rpl {:-(.+?)} $_ ||= $$1 --rpl {:(\d+?)} substr($_,0,$$1) = "" --rpl {:(\d+?):(\d+?)} $_ = substr($_,$$1,$$2); --rpl {#([^#].*?)} s/^$$1//; --rpl {%(.+?)} s/$$1$//; --rpl {/(.+?)/(.*?)} s/$$1/$$2/; --rpl {^(.+?)} s/^($$1)/uc($1)/e; --rpl {^^(.+?)} s/($$1)/uc($1)/eg; --rpl {,(.+?)} s/^($$1)/lc($1)/e; --rpl {,,(.+?)} s/($$1)/lc($1)/eg; If the user defined replacement string starts with '{' it can also be used as a positional replacement string (like {2.}). It is recommended to only change $_ but you have full access to all of GNU parallel's internal functions and data structures. Here are a few examples: Is the job sequence even or odd? --rpl {odd} $_ = seq() % 2 ? "odd" : "even" Pad job sequence with leading zeros to get equal width --rpl {0#} $f=1+int("".(log(total_jobs())/log(10))); $_=sprintf("%0${f}d",seq()) Job sequence counting from 0 --rpl {#0} $_ = seq() - 1 Job slot counting from 2 --rpl {%1} $_ = slot() + 1 Remove all extensions --rpl {:} s:(\.[^/]+)*$:: You can have dynamic replacement strings by including parenthesis in the replacement string and adding a regular expression between the parenthesis. The matching string will be inserted as $$1: parallel --rpl {%(.*?)} s/$$1// echo {%.tar.gz} ::: my.tar.gz parallel --rpl {:%(.+?)} s:$$1(\.[^/]+)*$:: \ echo {:%_file} ::: my_file.tar.gz parallel -n3 --rpl {/:%(.*?)} s:.*/(.*)$$1(\.[^/]+)*$:$1: \ echo job {#}: {2} {2.} {3/:%_1} ::: a/b.c c/d.e f/g_1.h.i You can even use multiple matches: parallel --rpl {/(.+?)/(.*?)} s/$$1/$$2/; echo {/replacethis/withthis} {/b/C} ::: a_replacethis_b parallel --rpl {(.*?)/(.*?)} $_="$$2$_$$1" \ echo {swap/these} ::: -middle- See also: {= perl expression =} --parens
--rsync-opts options: Options to pass on to rsync. Setting --rsync-opts takes precedence over setting the environment variable $PARALLEL_RSYNC_OPTS.
--max-chars=max-chars

-s max-chars

Use at most max-chars characters per command line, including the command and initial-arguments and the terminating nulls at the ends of the argument strings. The largest allowed value is system-dependent, and is calculated as the argument length limit for exec, less the size of your environment. The default value is the maximum. max-chars can be postfixed with K, M, G, T, P, k, m, g, t, or p (see UNIT PREFIX). Implies -X unless -m is set.

--show-limits: Display the limits on the command-line length which are imposed by the operating system and the -s option. Pipe the input from /dev/null (and perhaps specify --no-run-if-empty) if you don't want GNU parallel to do anything.
--semaphore: Work as a counting semaphore. --semaphore will cause GNU parallel to start command in the background. When the number of jobs given by --jobs is reached, GNU parallel will wait for one of these to complete before starting another command. --semaphore implies --bg unless --fg is specified. --semaphore implies --semaphorename `tty` unless --semaphorename is specified. Used with --fg, --wait, and --semaphorename. The command sem is an alias for parallel --semaphore. See also: man sem.
--semaphorename name

--id name

Use name as the name of the semaphore. Default is the name of the controlling tty (output from tty). The default normally works as expected when used interactively, but when used in a script name should be set. $$ or my_task_name are often a good value. The semaphore is stored in ~/.parallel/semaphores/ Implies --semaphore. See also: man sem.

--semaphoretimeout secs

--st secs

If secs > 0: If the semaphore is not released within secs seconds, take it anyway. If secs < 0: If the semaphore is not released within secs seconds, exit. Implies --semaphore. See also: man sem.

--seqreplace replace-str: Use the replacement string replace-str instead of {#} for job sequence number.
--session: Record names in current environment in $PARALLEL_IGNORED_NAMES and exit. Only used with env_parallel. Aliases, functions, and variables with names in $PARALLEL_IGNORED_NAMES will not be copied. Only supported in Ash, Bash, Dash, Ksh, Sh, and Zsh. See also: --env, --record-env.
--shard shardexpr: Use shardexpr as shard key and shard input to the jobs. shardexpr is [column number|column name] [perlexpression] e.g. 3, Address, 3 $_%=100, Address s/\d//g. Each input line is split using --colsep. The value of the column is put into $_, the perl expression is executed, the resulting value is hashed so that all lines of a given value is given to the same job slot. This is similar to sharding in databases. The performance is in the order of 100K rows per second. Faster if the shardcol is small (<10), slower if it is big (>100). --shard requires --pipe and a fixed numeric value for --jobs. See also: --bin, --group-by, --roundrobin.
--shebang

--hashbang

GNU parallel can be called as a shebang (#!) command as the first line of a script. The content of the file will be treated as inputsource. Like this: #!/usr/bin/parallel --shebang -r wget https://ftpmirror.gnu.org/parallel/parallel-20120822.tar.bz2 https://ftpmirror.gnu.org/parallel/parallel-20130822.tar.bz2 https://ftpmirror.gnu.org/parallel/parallel-20140822.tar.bz2 --shebang must be set as the first option. On FreeBSD env is needed: #!/usr/bin/env -S parallel --shebang -r wget https://ftpmirror.gnu.org/parallel/parallel-20120822.tar.bz2 https://ftpmirror.gnu.org/parallel/parallel-20130822.tar.bz2 https://ftpmirror.gnu.org/parallel/parallel-20140822.tar.bz2 There are many limitations of shebang (#!) depending on your operating system. See details on https://www.in-ulm.de/~mascheck/various/shebang/

--shebang-wrap: GNU parallel can parallelize scripts by wrapping the shebang line. If the program can be run like this: cat arguments | parallel the_program then the script can be changed to: #!/usr/bin/parallel --shebang-wrap /original/parser --options E.g. #!/usr/bin/parallel --shebang-wrap /usr/bin/python If the program can be run like this: cat data | parallel --pipe the_program then the script can be changed to: #!/usr/bin/parallel --shebang-wrap --pipe /orig/parser --opts E.g. #!/usr/bin/parallel --shebang-wrap --pipe /usr/bin/perl -w --shebang-wrap must be set as the first option.
--shellquote: Does not run the command but quotes it. Useful for making quoted composed commands for GNU parallel. Multiple --shellquote with quote the string multiple times, so parallel --shellquote | parallel --shellquote can be written as parallel --shellquote --shellquote.
--shuf: Shuffle jobs. When having multiple input sources it is hard to randomize jobs. --shuf will generate all jobs, and shuffle them before running them. This is useful to get a quick preview of the results before running the full batch.
--skip-first-line: Do not use the first line of input (used by GNU parallel itself when called with --shebang).
--sql DBURL (obsolete): Use --sqlmaster instead.
--sqlmaster DBURL: Submit jobs via SQL server. DBURL must point to a table, which will contain the same information as --joblog, the values from the input sources (stored in columns V1 .. Vn), and the output (stored in columns Stdout and Stderr). If DBURL is prepended with '+' GNU parallel assumes the table is already made with the correct columns and appends the jobs to it. If DBURL is not prepended with '+' the table will be dropped and created with the correct amount of V-columns unless --sqlmaster does not run any jobs, but it creates the values for the jobs to be run. One or more --sqlworker must be run to actually execute the jobs. If --wait is set, GNU parallel will wait for the jobs to complete. The format of a DBURL is: [sql:]vendor://[[user][:pwd]@][host][:port]/[db]/table E.g. sql:mysql://hr:hr@localhost:3306/hrdb/jobs mysql://scott:tiger@my.example.com/pardb/paralleljobs sql:oracle://scott:tiger@ora.example.com/xe/parjob postgresql://scott:tiger@pg.example.com/pgdb/parjob pg:///parjob sqlite3:///%2Ftmp%2Fpardb.sqlite/parjob csv:///%2Ftmp%2Fpardb/parjob Notice how / in the path of sqlite and CVS must be encoded as %2F. Except the last / in CSV which must be a /. It can also be an alias from ~/.sql/aliases: :myalias mysql:///mydb/paralleljobs
--sqlandworker DBURL: Shorthand for: --sqlmaster DBURL --sqlworker DBURL.
--sqlworker DBURL: Execute jobs via SQL server. Read the input sources variables from the table pointed to by DBURL. The command on the command line should be the same as given by --sqlmaster. If you have more than one --sqlworker jobs may be run more than once. If --sqlworker runs on the local machine, the hostname in the SQL table will not be ':' but instead the hostname of the machine.
--ssh sshcommand: GNU parallel defaults to using ssh for remote access. This can be overridden with --ssh. It can also be set on a per server basis (see --sshlogin).
--sshdelay mytime: Delay starting next ssh by mytime. GNU parallel will not start another ssh for the next mytime. For details on mytime see --delay.
-S [@hostgroups/][ncpus/]sshlogin[,[@hostgroups/][ncpus/]sshlogin[,...]]

-S @hostgroup
--sshlogin [@hostgroups/][ncpus/]sshlogin[,[@hostgroups/][ncpus/]sshlogin[,...]]
--sshlogin @hostgroup

Distribute jobs to remote computers. The jobs will be run on a list of remote computers. If hostgroups is given, the sshlogin will be added to that hostgroup. Multiple hostgroups are separated by '+'. The sshlogin will always be added to a hostgroup named the same as sshlogin. If only the @hostgroup is given, only the sshlogins in that hostgroup will be used. Multiple @hostgroup can be given. GNU parallel will determine the number of CPUs on the remote computers and run the number of jobs as specified by -j. If the number ncpus is given GNU parallel will use this number for number of CPUs on the host. Normally ncpus will not be needed. An sshlogin is of the form: [sshcommand [options]] [username@]hostname The sshlogin must not require a password (ssh-agent, ssh-copy-id, and sshpass may help with that). The sshlogin ':' is special, it means 'no ssh' and will therefore run on the local computer. The sshlogin '..' is special, it read sshlogins from ~/.parallel/sshloginfile or $XDG_CONFIG_HOME/parallel/sshloginfile The sshlogin '-' is special, too, it read sshlogins from stdin (standard input). To specify more sshlogins separate the sshlogins by comma, newline (in the same string), or repeat the options multiple times. For examples: see --sshloginfile. The remote host must have GNU parallel installed. --sshlogin is known to cause problems with -m and -X. --sshlogin is often used with --transferfile, --return, --cleanup, and --trc.

--sshloginfile filename

--slf filename

File with sshlogins. The file consists of sshlogins on separate lines. Empty lines and lines starting with '#' are ignored. Example: server.example.com username@server2.example.com 8/my-8-cpu-server.example.com 2/my_other_username@my-dualcore.example.net # This server has SSH running on port 2222 ssh -p 2222 server.example.net 4/ssh -p 2222 quadserver.example.net # Use a different ssh program myssh -p 2222 -l myusername hexacpu.example.net # Use a different ssh program with default number of CPUs //usr/local/bin/myssh -p 2222 -l myusername hexacpu # Use a different ssh program with 6 CPUs 6//usr/local/bin/myssh -p 2222 -l myusername hexacpu # Assume 16 CPUs on the local computer 16/: # Put server1 in hostgroup1 @hostgroup1/server1 # Put myusername@server2 in hostgroup1+hostgroup2 @hostgroup1+hostgroup2/myusername@server2 # Force 4 CPUs and put ssh -p 2222 server3 in hostgroup1 @hostgroup1/4/ssh -p 2222 server3 When using a different ssh program the last argument must be the hostname. Multiple --sshloginfile are allowed. GNU parallel will first look for the file in current dir; if that fails it look for the file in ~/.parallel. The sshloginfile '..' is special, it read sshlogins from ~/.parallel/sshloginfile The sshloginfile '.' is special, it read sshlogins from /etc/parallel/sshloginfile The sshloginfile '-' is special, too, it read sshlogins from stdin (standard input). If the sshloginfile is changed it will be re-read when a job finishes though at most once per second. This makes it possible to add and remove hosts while running. This can be used to have a daemon that updates the sshloginfile to only contain servers that are up: cp original.slf tmp2.slf while [ 1 ] ; do nice parallel --nonall -j0 -k --slf original.slf \ --tag echo | perl s/\t$// > tmp.slf if diff tmp.slf tmp2.slf; then mv tmp.slf tmp2.slf fi sleep 10 done & parallel --slf tmp2.slf ...

--slotreplace replace-str: Use the replacement string replace-str instead of {%} for job slot number.
--silent: Silent. The job to be run will not be printed. This is the default. Can be reversed with -v.
--template file=repl

--tmpl file=repl

Copy file to repl. All replacement strings in the contents of file will be replaced. All replacement strings in the name repl will be replaced. With --cleanup the new file will be removed when the job is done. If my.tmpl contains this: Xval: {x} Yval: {y} FixedValue: 9 # x with 2 decimals DecimalX: {=x $_=sprintf("%.2f",$_) =} TenX: {=x $_=$_*10 =} RandomVal: {=1 $_=rand() =} it can be used like this: myprog() { echo Using "$@"; cat "$@"; } export -f myprog parallel --cleanup --header : --tmpl my.tmpl={#}.t myprog {#}.t \ ::: x 1.234 2.345 3.45678 ::: y 1 2 3

--tty: Open terminal tty. If GNU parallel is used for starting a program that accesses the tty (such as an interactive program) then this option may be needed. It will default to starting only one job at a time (i.e. -j1), not buffer the output (i.e. -u), and it will open a tty for the job. You can of course override -j1 and -u. Using --tty unfortunately means that GNU parallel cannot kill the jobs (with --timeout, --memfree, or --halt). This is due to GNU parallel giving each child its own process group, which is then killed. Process groups are dependant on the tty.
--tag: Tag lines with arguments. Each output line will be prepended with the arguments and TAB (\t). When combined with --onall or --nonall the lines will be prepended with the sshlogin instead. --tag is ignored when using -u. --ctag gives the tag a color.
--tagstring str: Tag lines with a string. Each output line will be prepended with str and TAB (\t). str can contain replacement strings such as {}. --tagstring is ignored when using -u, --onall, and --nonall. --ctagstring gives the tag a color.
--tee: Pipe all data to all jobs. Used with --pipe/--pipepart and :::. seq 1000 | parallel --pipe --tee -v wc {} ::: -w -l -c How many numbers in 1..1000 contain 0..9, and how many bytes do they fill: seq 1000 | parallel --pipe --tee --tag \ grep {1} | wc {2} ::: {0..9} ::: -l -c How many words contain a..z and how many bytes do they fill? parallel -a /usr/share/dict/words --pipepart --tee --tag \ grep {1} | wc {2} ::: {a..z} ::: -l -c
--termseq sequence: Termination sequence. When a job is killed due to --timeout, --memfree, --halt, or abnormal termination of GNU parallel, sequence determines how the job is killed. The default is: TERM,200,TERM,100,TERM,50,KILL,25 which sends a TERM signal, waits 200 ms, sends another TERM signal, waits 100 ms, sends another TERM signal, waits 50 ms, sends a KILL signal, waits 25 ms, and exits. GNU parallel detects if a process dies before the waiting time is up.
--tmpdir dirname: Directory for temporary files. GNU parallel normally buffers output into temporary files in /tmp. By setting --tmpdir you can use a different dir for the files. Setting --tmpdir is equivalent to setting $TMPDIR.
--tmux (Long beta testing): Use tmux for output. Start a tmux session and run each job in a window in that session. No other output will be produced.
--tmuxpane (Long beta testing): Use tmux for output but put output into panes in the first window. Useful if you want to monitor the progress of less than 100 concurrent jobs.
--timeout duration: Time out for command. If the command runs for longer than duration seconds it will get killed as per --termseq. If duration is followed by a % then the timeout will dynamically be computed as a percentage of the median average runtime of successful jobs. Only values > 100% will make sense. duration is normally in seconds, but can be floats postfixed with s, m, h, or d which would multiply the float by 1, 60, 3600, or 86400. Thus these are equivalent: --timeout 100000 and --timeout 1d3.5h16.6m4s.
--verbose

-t

Print the job to be run on stderr (standard error). See also: -v, -p.

--transfer: Transfer files to remote computers. Shorthand for: --transferfile {}.
--transferfile filename

--tf filename

--transferfile is used with --sshlogin to transfer files to the remote computers. The files will be transferred using rsync and will be put relative to the work dir (see --workdir). The filename will normally contain a replacement string. If the path contains /./ the remaining path will be relative to the work dir (for details: see rsync). If the work dir is /home/user, the transferring will be as follows: /tmp/foo/bar => /tmp/foo/bar tmp/foo/bar => /home/user/tmp/foo/bar /tmp/./foo/bar => /home/user/foo/bar tmp/./foo/bar => /home/user/foo/bar Examples This will transfer the file foo/bar.txt to the computer server.example.com to the file $HOME/foo/bar.txt before running wc foo/bar.txt on server.example.com: echo foo/bar.txt | parallel --transferfile {} \ --sshlogin server.example.com wc This will transfer the file /tmp/foo/bar.txt to the computer server.example.com to the file /tmp/foo/bar.txt before running wc /tmp/foo/bar.txt on server.example.com: echo /tmp/foo/bar.txt | parallel --transferfile {} \ --sshlogin server.example.com wc This will transfer the file /tmp/foo/bar.txt to the computer server.example.com to the file foo/bar.txt before running wc ./foo/bar.txt on server.example.com: echo /tmp/./foo/bar.txt | parallel --transferfile {} \ --sshlogin server.example.com wc {= s:.*/\./:./: =} --transferfile is often used with --return and --cleanup. A shorthand for --transferfile {} is --transfer. --transferfile is ignored when used with --sshlogin : or when not used with --sshlogin.

--trc filename

Transfer, Return, Cleanup. Shorthand for: --transferfile {} --return filename --cleanup

--trim <n|l|r|lr|rl>

Trim white space in input.

No trim. Input is not modified. This is the default.
Left trim. Remove white space from start of input. E.g. a bc -> a bc .
Right trim. Remove white space from end of input. E.g. a bc -> a bc.

lr

rl

Both trim. Remove white space from both start and end of input. E.g. a bc -> a bc. This is the default if --colsep is used.

--ungroup

-u

Ungroup output. Output is printed as soon as possible and bypasses GNU parallel internal processing. This may cause output from different commands to be mixed thus should only be used if you do not care about the output. Compare these: seq 4 | parallel -j0 \ sleep {};echo -n start{};sleep {};echo {}end seq 4 | parallel -u -j0 \ sleep {};echo -n start{};sleep {};echo {}end It also disables --tag. GNU parallel outputs faster with -u. Compare the speeds of these: parallel seq ::: 300000000 >/dev/null parallel -u seq ::: 300000000 >/dev/null parallel --line-buffer seq ::: 300000000 >/dev/null Can be reversed with --group. See also: --line-buffer --group

--extensionreplace replace-str

--er replace-str

Use the replacement string replace-str instead of {.} for input line without extension.

--use-sockets-instead-of-threads

--use-cores-instead-of-threads
--use-cpus-instead-of-cores (obsolete)

Determine how GNU parallel counts the number of CPUs. GNU parallel uses this number when the number of jobslots is computed relative to the number of CPUs (e.g. 100% or +1). CPUs can be counted in three different ways:

sockets

The number of filled CPU sockets (i.e. the number of physical chips).

cores

The number of physical cores (i.e. the number of physical compute cores).

threads

The number of hyperthreaded cores (i.e. the number of virtual cores - with some of them possibly being hyperthreaded)

Normally the number of CPUs is computed as the number of CPU threads. With --use-sockets-instead-of-threads or --use-cores-instead-of-threads you can force it to be computed as the number of filled sockets or number of cores instead. Most users will not need these options. --use-cpus-instead-of-cores is a (misleading) alias for --use-sockets-instead-of-threads and is kept for backwards compatibility.

-v: Verbose. Print the job to be run on stdout (standard output). Can be reversed with --silent. See also -t. Use -v -v to print the wrapping ssh command when running remotely.
--version

-V

Print the version GNU parallel and exit.

--workdir mydir

--wd mydir

Jobs will be run in the dir mydir. The default is the current dir for the local machine, and the login dir for remote computers. Files transferred using --transferfile and --return will be relative to mydir on remote computers. The special mydir value ... will create working dirs under ~/.parallel/tmp/. If --cleanup is given these dirs will be removed. The special mydir value . uses the current working dir. If the current working dir is beneath your home dir, the value . is treated as the relative path to your home dir. This means that if your home dir is different on remote computers (e.g. if your login is different) the relative path will still be relative to your home dir. To see the difference try: parallel -S server pwd ::: "" parallel --wd . -S server pwd ::: "" parallel --wd ... -S server pwd ::: "" mydir can contain GNU parallel's replacement strings.

--wait: Wait for all commands to complete. Used with --semaphore or --sqlmaster. See also: man sem.
-X: Multiple arguments with context replace. Insert as many arguments as the command line length permits. If multiple jobs are being run in parallel: distribute the arguments evenly among the jobs. Use -j1 to avoid this. If {} is not used the arguments will be appended to the line. If {} is used as part of a word (like pic{}.jpg) then the whole word will be repeated. If {} is used multiple times each {} will be replaced with the arguments. Normally -X will do the right thing, whereas -m can give unexpected results if {} is used as part of a word. Support for -X with --sshlogin is limited and may fail. See also: -m.
--exit

-x

Exit if the size (see the -s option) is exceeded.

--xargs: Multiple arguments. Insert as many arguments as the command line length permits. If {} is not used the arguments will be appended to the line. If {} is used multiple times each {} will be replaced with all the arguments. Support for --xargs with --sshlogin is limited and may fail. See also -X for context replace. If in doubt use -X as that will most likely do what is needed.

EXAMPLES

EXAMPLE: Working as xargs -n1. Argument appending

GNU parallel can work similar to xargs -n1.

To compress all html files using gzip run:

find . -name *.html | parallel gzip --best

If the file names may contain a newline use -0. Substitute FOO BAR with FUBAR in all files in this dir and subdirs:

find . -type f -print0 | \ parallel -q0 perl -i -pe s/FOO BAR/FUBAR/g

Note -q is needed because of the space in 'FOO BAR'.

EXAMPLE: Simple network scanner

prips can generate IP-addresses from CIDR notation. With GNU parallel you can build a simple network scanner to see which addresses respond to ping:

prips 130.229.16.0/20 | \ parallel --timeout 2 -j0 \ ping -c 1 {} >/dev/null && echo {} 2>/dev/null

EXAMPLE: Reading arguments from command line

GNU parallel can take the arguments from command line instead of stdin (standard input). To compress all html files in the current dir using gzip run:

parallel gzip --best ::: *.html

To convert *.wav to *.mp3 using LAME running one process per CPU run:

parallel lame {} -o {.}.mp3 ::: *.wav

EXAMPLE: Inserting multiple arguments

When moving a lot of files like this: mv *.log destdir you will sometimes get the error:

bash: /bin/mv: Argument list too long

because there are too many files. You can instead do:

ls | grep -E \.log$ | parallel mv {} destdir

This will run mv for each file. It can be done faster if mv gets as many arguments that will fit on the line:

ls | grep -E \.log$ | parallel -m mv {} destdir

In many shells you can also use printf:

printf %s\0 *.log | parallel -0 -m mv {} destdir

EXAMPLE: Context replace

To remove the files pict0000.jpg .. pict9999.jpg you could do:

seq -w 0 9999 | parallel rm pict{}.jpg

You could also do:

seq -w 0 9999 | perl -pe s/(.*)/pict$1.jpg/ | parallel -m rm

The first will run rm 10000 times, while the last will only run rm as many times needed to keep the command line length short enough to avoid Argument list too long (it typically runs 1-2 times).

You could also run:

seq -w 0 9999 | parallel -X rm pict{}.jpg

This will also only run rm as many times needed to keep the command line length short enough.

EXAMPLE: Compute intensive jobs and substitution

If ImageMagick is installed this will generate a thumbnail of a jpg file:

convert -geometry 120 foo.jpg thumb_foo.jpg

This will run with number-of-cpus jobs in parallel for all jpg files in a directory:

ls *.jpg | parallel convert -geometry 120 {} thumb_{}

To do it recursively use find:

find . -name *.jpg | \ parallel convert -geometry 120 {} {}_thumb.jpg

Notice how the argument has to start with {} as {} will include path (e.g. running convert -geometry 120 ./foo/bar.jpg thumb_./foo/bar.jpg would clearly be wrong). The command will generate files like ./foo/bar.jpg_thumb.jpg.

Use {.} to avoid the extra .jpg in the file name. This command will make files like ./foo/bar_thumb.jpg:

find . -name *.jpg | \ parallel convert -geometry 120 {} {.}_thumb.jpg

EXAMPLE: Substitution and redirection

This will generate an uncompressed version of .gz-files next to the .gz-file:

parallel zcat {} ">"{.} ::: *.gz

Quoting of > is necessary to postpone the redirection. Another solution is to quote the whole command:

parallel "zcat {} >{.}" ::: *.gz

Other special shell characters (such as * ; $ > < | >> <<) also need to be put in quotes, as they may otherwise be interpreted by the shell and not given to GNU parallel.

EXAMPLE: Composed commands

A job can consist of several commands. This will print the number of files in each directory:

ls | parallel echo -n {}" "; ls {}|wc -l

To put the output in a file called <name>.dir:

ls | parallel (echo -n {}" "; ls {}|wc -l) >{}.dir

Even small shell scripts can be run by GNU parallel:

find . | parallel a={}; name=${a##*/}; \ upper=$(echo "$name" | tr "[:lower:]" "[:upper:]");\ echo "$name - $upper" ls | parallel mv {} "$(echo {} | tr "[:upper:]" "[:lower:]")"

Given a list of URLs, list all URLs that fail to download. Print the line number and the URL.

cat urlfile | parallel "wget {} 2>/dev/null || grep -n {} urlfile"

Create a mirror directory with the same filenames except all files and symlinks are empty files.

cp -rs /the/source/dir mirror_dir find mirror_dir -type l | parallel -m rm {} && touch {}

Find the files in a list that do not exist

cat file_list | parallel if [ ! -e {} ] ; then echo {}; fi

EXAMPLE: Composed command with perl replacement string

You have a bunch of file. You want them sorted into dirs. The dir of each file should be named the first letter of the file name.

parallel mkdir -p {=s/(.).*/$1/=}; mv {} {=s/(.).*/$1/=} ::: *

EXAMPLE: Composed command with multiple input sources

You have a dir with files named as 24 hours in 5 minute intervals: 00:00, 00:05, 00:10 .. 23:55. You want to find the files missing:

parallel [ -f {1}:{2} ] "||" echo {1}:{2} does not exist \ ::: {00..23} ::: {00..55..5}

EXAMPLE: Calling Bash functions

If the composed command is longer than a line, it becomes hard to read. In Bash you can use functions. Just remember to export -f the function.

doit() { echo Doing it for $1 sleep 2 echo Done with $1 } export -f doit parallel doit ::: 1 2 3 doubleit() { echo Doing it for $1 $2 sleep 2 echo Done with $1 $2 } export -f doubleit parallel doubleit ::: 1 2 3 ::: a b

To do this on remote servers you need to transfer the function using --env:

parallel --env doit -S server doit ::: 1 2 3 parallel --env doubleit -S server doubleit ::: 1 2 3 ::: a b

If your environment (aliases, variables, and functions) is small you can copy the full environment without having to export -f anything. See env_parallel.

EXAMPLE: Function tester

To test a program with different parameters:

tester() { if (eval "$@") >&/dev/null; then perl -e printf "\033[30;102m[ OK ]\033[0m @ARGV\n" "$@" else perl -e printf "\033[30;101m[FAIL]\033[0m @ARGV\n" "$@" fi } export -f tester parallel tester my_program ::: arg1 arg2 parallel tester exit ::: 1 0 2 0

If my_program fails a red FAIL will be printed followed by the failing command; otherwise a green OK will be printed followed by the command.

EXAMPLE: Continously show the latest line of output

It can be useful to monitor the output of running jobs.

This shows the most recent output line until a job finishes. After which the output of the job is printed in full:

parallel {} | tee >(cat >&3) ::: command 1 command 2 \ 3> >(perl -ne $|=1;chomp;printf"%.$COLUMNSs\r",$_." "x100)

EXAMPLE: Log rotate

Log rotation renames a logfile to an extension with a higher number: log.1 becomes log.2, log.2 becomes log.3, and so on. The oldest log is removed. To avoid overwriting files the process starts backwards from the high number to the low number. This will keep 10 old versions of the log:

seq 9 -1 1 | parallel -j1 mv log.{} log.{= $_++ =} mv log log.1

EXAMPLE: Removing file extension when processing files

When processing files removing the file extension using {.} is often useful.

Create a directory for each zip-file and unzip it in that dir:

parallel mkdir {.}; cd {.}; unzip ../{} ::: *.zip

Recompress all .gz files in current directory using bzip2 running 1 job per CPU in parallel:

parallel "zcat {} | bzip2 >{.}.bz2 && rm {}" ::: *.gz

Convert all WAV files to MP3 using LAME:

find sounddir -type f -name *.wav | parallel lame {} -o {.}.mp3

Put all converted in the same directory:

find sounddir -type f -name *.wav | \ parallel lame {} -o mydir/{/.}.mp3

EXAMPLE: Removing strings from the argument

If you have directory with tar.gz files and want these extracted in the corresponding dir (e.g foo.tar.gz will be extracted in the dir foo) you can do:

parallel --plus mkdir {..}; tar -C {..} -xf {} ::: *.tar.gz

If you want to remove a different ending, you can use {%string}:

parallel --plus echo {%_demo} ::: mycode_demo keep_demo_here

You can also remove a starting string with {#string}

parallel --plus echo {#demo_} ::: demo_mycode keep_demo_here

To remove a string anywhere you can use regular expressions with {/regexp/replacement} and leave the replacement empty:

parallel --plus echo {/demo_/} ::: demo_mycode remove_demo_here

EXAMPLE: Download 24 images for each of the past 30 days

Let us assume a website stores images like:

https://www.example.com/path/to/YYYYMMDD_##.jpg

where YYYYMMDD is the date and ## is the number 01-24. This will download images for the past 30 days:

getit() { date=$(date -d "today -$1 days" +%Y%m%d) num=$2 echo wget https://www.example.com/path/to/${date}_${num}.jpg } export -f getit parallel getit ::: $(seq 30) ::: $(seq -w 24)

$(date -d today -$1 days +%Y%m%d) will give the dates in YYYYMMDD with $1 days subtracted.

EXAMPLE: Download world map from NASA

NASA provides tiles to download on earthdata.nasa.gov. Download tiles for Blue Marble world map and create a 10240x20480 map.

base=https://map1a.vis.earthdata.nasa.gov/wmts-geo/wmts.cgi service="SERVICE=WMTS&REQUEST=GetTile&VERSION=1.0.0" layer="LAYER=BlueMarble_ShadedRelief_Bathymetry" set="STYLE=&TILEMATRIXSET=EPSG4326_500m&TILEMATRIX=5" tile="TILEROW={1}&TILECOL={2}" format="FORMAT=image%2Fjpeg" url="$base?$service&$layer&$set&$tile&$format" parallel -j0 -q wget "$url" -O {1}_{2}.jpg ::: {0..19} ::: {0..39} parallel eval convert +append {}_{0..39}.jpg line{}.jpg ::: {0..19} convert -append line{0..19}.jpg world.jpg

EXAMPLE: Download Apollo-11 images from NASA using jq

Search NASA using their API to get JSON for images related to 'apollo 11' and has 'moon landing' in the description.

The search query returns JSON containing URLs to JSON containing collections of pictures. One of the pictures in each of these collection is large.

wget is used to get the JSON for the search query. jq is then used to extract the URLs of the collections. parallel then calls wget to get each collection, which is passed to jq to extract the URLs of all images. grep filters out the large images, and parallel finally uses wget to fetch the images.

base="https://images-api.nasa.gov/search" q="q=apollo 11" description="description=moon landing" media_type="media_type=image" wget -O - "$base?$q&$description&$media_type" | jq -r .collection.items[].href | parallel wget -O - | jq -r .[] | grep large | parallel wget

EXAMPLE: Download video playlist in parallel

youtube-dl is an excellent tool to download videos. It can, however, not download videos in parallel. This takes a playlist and downloads 10 videos in parallel.

url=youtu.be/watch?v=0wOf2Fgi3DE&list=UU_cznB5YZZmvAmeq7Y3EriQ export url youtube-dl --flat-playlist "https://$url" | parallel --tagstring {#} --lb -j10 \ youtube-dl --playlist-start {#} --playlist-end {#} "https://$url"

EXAMPLE: Prepend last modified date (ISO8601) to file name

parallel mv {} {= $a=pQ($_); $b=$_; \ $_=qx{date -r "$a" +%FT%T}; chomp; $_="$_ $b" =} ::: *

{= and =} mark a perl expression. pQ perl-quotes the string. date +%FT%T is the date in ISO8601 with time.

EXAMPLE: Save output in ISO8601 dirs

Save output from ps aux every second into dirs named yyyy-mm-ddThh:mm:ss+zz:zz.

seq 1000 | parallel -N0 -j1 --delay 1 \ --results {= $_=`date -Isec`; chomp=}/ ps aux

EXAMPLE: Digital clock with "blinking" :

The : in a digital clock blinks. To make every other line have a ':' and the rest a ' ' a perl expression is used to look at the 3rd input source. If the value modulo 2 is 1: Use : otherwise use :

parallel -k echo {1}{=3 $_=$_%2?":":" "=}{2}{3} \ ::: {0..12} ::: {0..5} ::: {0..9}

EXAMPLE: Aggregating content of files

This:

parallel --header : echo x{X}y{Y}z{Z} \> x{X}y{Y}z{Z} \ ::: X {1..5} ::: Y {01..10} ::: Z {1..5}

will generate the files x1y01z1 .. x5y10z5. If you want to aggregate the output grouping on x and z you can do this:

parallel eval cat {=s/y01/y*/=} > {=s/y01//=} ::: *y01*

For all values of x and z it runs commands like:

cat x1y*z1 > x1z1

So you end up with x1z1 .. x5z5 each containing the content of all values of y.

EXAMPLE: Breadth first parallel web crawler/mirrorer

This script below will crawl and mirror a URL in parallel. It downloads first pages that are 1 click down, then 2 clicks down, then 3; instead of the normal depth first, where the first link link on each page is fetched first.

Run like this:

PARALLEL=-j100 ./parallel-crawl http://gatt.org.yeslab.org/

Remove the wget part if you only want a web crawler.

It works by fetching a page from a list of URLs and looking for links in that page that are within the same starting URL and that have not already been seen. These links are added to a new queue. When all the pages from the list is done, the new queue is moved to the list of URLs and the process is started over until no unseen links are found.

#!/bin/bash # E.g. http://gatt.org.yeslab.org/ URL=$1 # Stay inside the start dir BASEURL=$(echo $URL | perl -pe s:#.*::; s:(//.*/)[^/]*:$1:) URLLIST=$(mktemp urllist.XXXX) URLLIST2=$(mktemp urllist.XXXX) SEEN=$(mktemp seen.XXXX) # Spider to get the URLs echo $URL >$URLLIST cp $URLLIST $SEEN while [ -s $URLLIST ] ; do cat $URLLIST | parallel lynx -listonly -image_links -dump {} \; \ wget -qm -l1 -Q1 {} \; echo Spidered: {} \>\&2 | perl -ne s/#.*//; s/\s+\d+.\s(\S+)$/$1/ and do { $seen{$1}++ or print } | grep -F $BASEURL | grep -v -x -F -f $SEEN | tee -a $SEEN > $URLLIST2 mv $URLLIST2 $URLLIST done rm -f $URLLIST $URLLIST2 $SEEN

EXAMPLE: Process files from a tar file while unpacking

If the files to be processed are in a tar file then unpacking one file and processing it immediately may be faster than first unpacking all files.

tar xvf foo.tgz | perl -ne print $l;$l=$_;END{print $l} | \ parallel echo

The Perl one-liner is needed to make sure the file is complete before handing it to GNU parallel.

EXAMPLE: Rewriting a for-loop and a while-read-loop

for-loops like this:

(for x in `cat list` ; do do_something $x done) | process_output

and while-read-loops like this:

cat list | (while read x ; do do_something $x done) | process_output

can be written like this:

cat list | parallel do_something | process_output

For example: Find which host name in a list has IP address 1.2.3 4:

cat hosts.txt | parallel -P 100 host | grep 1.2.3.4

If the processing requires more steps the for-loop like this:

(for x in `cat list` ; do no_extension=${x%.*}; do_step1 $x scale $no_extension.jpg do_step2 <$x $no_extension done) | process_output

and while-loops like this:

cat list | (while read x ; do no_extension=${x%.*}; do_step1 $x scale $no_extension.jpg do_step2 <$x $no_extension done) | process_output

can be written like this:

cat list | parallel "do_step1 {} scale {.}.jpg ; do_step2 <{} {.}" |\ process_output

If the body of the loop is bigger, it improves readability to use a function:

(for x in `cat list` ; do do_something $x [... 100 lines that do something with $x ...] done) | process_output cat list | (while read x ; do do_something $x [... 100 lines that do something with $x ...] done) | process_output

can both be rewritten as:

doit() { x=$1 do_something $x [... 100 lines that do something with $x ...] } export -f doit cat list | parallel doit

EXAMPLE: Rewriting nested for-loops

Nested for-loops like this:

(for x in `cat xlist` ; do for y in `cat ylist` ; do do_something $x $y done done) | process_output

can be written like this:

parallel do_something {1} {2} :::: xlist ylist | process_output

Nested for-loops like this:

(for colour in red green blue ; do for size in S M L XL XXL ; do echo $colour $size done done) | sort

can be written like this:

parallel echo {1} {2} ::: red green blue ::: S M L XL XXL | sort

EXAMPLE: Finding the lowest difference between files

diff is good for finding differences in text files. diff | wc -l gives an indication of the size of the difference. To find the differences between all files in the current dir do:

parallel --tag diff {1} {2} | wc -l ::: * ::: * | sort -nk3

This way it is possible to see if some files are closer to other files.

EXAMPLE: for-loops with column names

When doing multiple nested for-loops it can be easier to keep track of the loop variable if is is named instead of just having a number. Use --header : to let the first argument be an named alias for the positional replacement string:

parallel --header : echo {colour} {size} \ ::: colour red green blue ::: size S M L XL XXL

This also works if the input file is a file with columns:

cat addressbook.tsv | \ parallel --colsep \t --header : echo {Name} {E-mail address}

EXAMPLE: All combinations in a list

GNU parallel makes all combinations when given two lists.

To make all combinations in a single list with unique values, you repeat the list and use replacement string {choose_k}:

parallel --plus echo {choose_k} ::: A B C D ::: A B C D parallel --plus echo 2{2choose_k} 1{1choose_k} ::: A B C D ::: A B C D

{choose_k} works for any number of input sources:

parallel --plus echo {choose_k} ::: A B C D ::: A B C D ::: A B C D

EXAMPLE: From a to b and b to c

Assume you have input like:

aardvark babble cab dab each

and want to run combinations like:

aardvark babble babble cab cab dab dab each

If the input is in the file in.txt:

parallel echo {1} - {2} ::::+ <(head -n -1 in.txt) <(tail -n +2 in.txt)

If the input is in the array $a here are two solutions:

seq $((${#a[@]}-1)) | \ env_parallel --env a echo ${a[{=$_--=}]} - ${a[{}]} parallel echo {1} - {2} ::: "${a[@]::${#a[@]}-1}" :::+ "${a[@]:1}"

EXAMPLE: Count the differences between all files in a dir

Using --results the results are saved in /tmp/diffcount*.

parallel --results /tmp/diffcount "diff -U 0 {1} {2} | \ tail -n +3 |grep -v ^@|wc -l" ::: * ::: *

To see the difference between file A and file B look at the file '/tmp/diffcount/1/A/2/B'.

EXAMPLE: Speeding up fast jobs

Starting a job on the local machine takes around 10 ms. This can be a big overhead if the job takes very few ms to run. Often you can group small jobs together using -X which will make the overhead less significant. Compare the speed of these:

seq -w 0 9999 | parallel touch pict{}.jpg seq -w 0 9999 | parallel -X touch pict{}.jpg

If your program cannot take multiple arguments, then you can use GNU parallel to spawn multiple GNU parallels:

seq -w 0 9999999 | \ parallel -j10 -q -I,, --pipe parallel -j0 touch pict{}.jpg

If -j0 normally spawns 252 jobs, then the above will try to spawn 2520 jobs. On a normal GNU/Linux system you can spawn 32000 jobs using this technique with no problems. To raise the 32000 jobs limit raise /proc/sys/kernel/pid_max to 4194303.

If you do not need GNU parallel to have control over each job (so no need for --retries or --joblog or similar), then it can be even faster if you can generate the command lines and pipe those to a shell. So if you can do this:

mygenerator | sh

Then that can be parallelized like this:

mygenerator | parallel --pipe --block 10M sh

E.g.

mygenerator() { seq 10000000 | perl -pe print "echo This is fast job number "; } mygenerator | parallel --pipe --block 10M sh

The overhead is 100000 times smaller namely around 100 nanoseconds per job.

EXAMPLE: Using shell variables

When using shell variables you need to quote them correctly as they may otherwise be interpreted by the shell.

Notice the difference between:

ARR=("My brothers 12\" records are worth <\$\$\$>"! Foo Bar) parallel echo ::: ${ARR[@]} # This is probably not what you want

and:

ARR=("My brothers 12\" records are worth <\$\$\$>"! Foo Bar) parallel echo ::: "${ARR[@]}"

When using variables in the actual command that contains special characters (e.g. space) you can quote them using '$VAR' or using "'s and -q:

VAR="My brothers 12\" records are worth <\$\$\$>" parallel -q echo "$VAR" ::: ! export VAR parallel echo "$VAR" ::: !

If $VAR does not contain ' then '$VAR' will also work (and does not need export):

VAR="My 12\" records are worth <\$\$\$>" parallel echo "$VAR" ::: !

If you use them in a function you just quote as you normally would do:

VAR="My brothers 12\" records are worth <\$\$\$>" export VAR myfunc() { echo "$VAR" "$1"; } export -f myfunc parallel myfunc ::: !

EXAMPLE: Group output lines

When running jobs that output data, you often do not want the output of multiple jobs to run together. GNU parallel defaults to grouping the output of each job, so the output is printed when the job finishes. If you want full lines to be printed while the job is running you can use --line-buffer. If you want output to be printed as soon as possible you can use -u.

Compare the output of:

parallel wget --limit-rate=100k \ https://ftpmirror.gnu.org/parallel/parallel-20{}0822.tar.bz2 \ ::: {12..16} parallel --line-buffer wget --limit-rate=100k \ https://ftpmirror.gnu.org/parallel/parallel-20{}0822.tar.bz2 \ ::: {12..16} parallel -u wget --limit-rate=100k \ https://ftpmirror.gnu.org/parallel/parallel-20{}0822.tar.bz2 \ ::: {12..16}

EXAMPLE: Tag output lines

GNU parallel groups the output lines, but it can be hard to see where the different jobs begin. --tag prepends the argument to make that more visible:

parallel --tag wget --limit-rate=100k \ https://ftpmirror.gnu.org/parallel/parallel-20{}0822.tar.bz2 \ ::: {12..16}

--tag works with --line-buffer but not with -u:

parallel --tag --line-buffer wget --limit-rate=100k \ https://ftpmirror.gnu.org/parallel/parallel-20{}0822.tar.bz2 \ ::: {12..16}

Check the uptime of the servers in ~/.parallel/sshloginfile:

parallel --tag -S .. --nonall uptime

EXAMPLE: Colorize output

Give each job a new color. Most terminals support ANSI colors with the escape code \033[30;3Xm where 0 <= X <= 7:

seq 10 | \ parallel --tagstring \033[30;3{=$_=++$::color%8=}m seq {} parallel --rpl {color} $_="\033[30;3".(++$::color%8)."m" \ --tagstring {color} seq {} ::: {1..10}

To get rid of the initial \t (which comes from --tagstring):

... | perl -pe s/\t//

EXAMPLE: Keep order of output same as order of input

Normally the output of a job will be printed as soon as it completes. Sometimes you want the order of the output to remain the same as the order of the input. This is often important, if the output is used as input for another system. -k will make sure the order of output will be in the same order as input even if later jobs end before earlier jobs.

Append a string to every line in a text file:

cat textfile | parallel -k echo {} append_string

If you remove -k some of the lines may come out in the wrong order.

Another example is traceroute:

parallel traceroute ::: qubes-os.org debian.org freenetproject.org

will give traceroute of qubes-os.org, debian.org and freenetproject.org, but it will be sorted according to which job completed first.

To keep the order the same as input run:

parallel -k traceroute ::: qubes-os.org debian.org freenetproject.org

This will make sure the traceroute to qubes-os.org will be printed first.

A bit more complex example is downloading a huge file in chunks in parallel: Some internet connections will deliver more data if you download files in parallel. For downloading files in parallel see: EXAMPLE: Download 10 images for each of the past 30 days. But if you are downloading a big file you can download the file in chunks in parallel.

To download byte 10000000-19999999 you can use curl:

curl -r 10000000-19999999 https://example.com/the/big/file >file.part

To download a 1 GB file we need 100 10MB chunks downloaded and combined in the correct order.

seq 0 99 | parallel -k curl -r \ {}0000000-{}9999999 https://example.com/the/big/file > file

EXAMPLE: Parallel grep

grep -r greps recursively through directories. On multicore CPUs GNU parallel can often speed this up.

find . -type f | parallel -k -j150% -n 1000 -m grep -H -n STRING {}

This will run 1.5 job per CPU, and give 1000 arguments to grep.

EXAMPLE: Grepping n lines for m regular expressions.

The simplest solution to grep a big file for a lot of regexps is:

grep -f regexps.txt bigfile

Or if the regexps are fixed strings:

grep -F -f regexps.txt bigfile

There are 3 limiting factors: CPU, RAM, and disk I/O.

RAM is easy to measure: If the grep process takes up most of your free memory (e.g. when running top), then RAM is a limiting factor.

CPU is also easy to measure: If the grep takes >90% CPU in top, then the CPU is a limiting factor, and parallelization will speed this up.

It is harder to see if disk I/O is the limiting factor, and depending on the disk system it may be faster or slower to parallelize. The only way to know for certain is to test and measure.

Limiting factor: RAM

The normal grep -f regexps.txt bigfile works no matter the size of bigfile, but if regexps.txt is so big it cannot fit into memory, then you need to split this.

grep -F takes around 100 bytes of RAM and grep takes about 500 bytes of RAM per 1 byte of regexp. So if regexps.txt is 1% of your RAM, then it may be too big.

If you can convert your regexps into fixed strings do that. E.g. if the lines you are looking for in bigfile all looks like:

ID1 foo bar baz Identifier1 quux fubar ID2 foo bar baz Identifier2

then your regexps.txt can be converted from:

ID1.*Identifier1 ID2.*Identifier2

into:

ID1 foo bar baz Identifier1 ID2 foo bar baz Identifier2

This way you can use grep -F which takes around 80% less memory and is much faster.

If it still does not fit in memory you can do this:

parallel --pipepart -a regexps.txt --block 1M grep -F -f - -n bigfile | \ sort -un | perl -pe s/^\d+://

The 1M should be your free memory divided by the number of CPU threads and divided by 200 for grep -F and by 1000 for normal grep. On GNU/Linux you can do:

free=$(awk /^((Swap)?Cached|MemFree|Buffers):/ { sum += $2 } END { print sum } /proc/meminfo) percpu=$((free / 200 / $(parallel --number-of-threads)))k parallel --pipepart -a regexps.txt --block $percpu --compress \ grep -F -f - -n bigfile | \ sort -un | perl -pe s/^\d+://

If you can live with duplicated lines and wrong order, it is faster to do:

parallel --pipepart -a regexps.txt --block $percpu --compress \ grep -F -f - bigfile

Limiting factor: CPU

If the CPU is the limiting factor parallelization should be done on the regexps:

cat regexps.txt | parallel --pipe -L1000 --roundrobin --compress \ grep -f - -n bigfile | \ sort -un | perl -pe s/^\d+://

The command will start one grep per CPU and read bigfile one time per CPU, but as that is done in parallel, all reads except the first will be cached in RAM. Depending on the size of regexps.txt it may be faster to use --block 10m instead of -L1000.

Some storage systems perform better when reading multiple chunks in parallel. This is true for some RAID systems and for some network file systems. To parallelize the reading of bigfile:

parallel --pipepart --block 100M -a bigfile -k --compress \ grep -f regexps.txt

This will split bigfile into 100MB chunks and run grep on each of these chunks. To parallelize both reading of bigfile and regexps.txt combine the two using --cat:

parallel --pipepart --block 100M -a bigfile --cat cat regexps.txt \ \| parallel --pipe -L1000 --roundrobin grep -f - {}

If a line matches multiple regexps, the line may be duplicated.

Bigger problem

If the problem is too big to be solved by this, you are probably ready for Lucene.

EXAMPLE: Using remote computers

To run commands on a remote computer SSH needs to be set up and you must be able to login without entering a password (The commands ssh-copy-id, ssh-agent, and sshpass may help you do that).

If you need to login to a whole cluster, you typically do not want to accept the host key for every host. You want to accept them the first time and be warned if they are ever changed. To do that:

# Add the servers to the sshloginfile (echo servera; echo serverb) > .parallel/my_cluster # Make sure .ssh/config exist touch .ssh/config cp .ssh/config .ssh/config.backup # Disable StrictHostKeyChecking temporarily (echo Host *; echo StrictHostKeyChecking no) >> .ssh/config parallel --slf my_cluster --nonall true # Remove the disabling of StrictHostKeyChecking mv .ssh/config.backup .ssh/config

The servers in .parallel/my_cluster are now added in .ssh/known_hosts.

To run echo on server.example.com:

seq 10 | parallel --sshlogin server.example.com echo

To run commands on more than one remote computer run:

seq 10 | parallel --sshlogin s1.example.com,s2.example.net echo

Or:

seq 10 | parallel --sshlogin server.example.com \ --sshlogin server2.example.net echo

If the login username is foo on server2.example.net use:

seq 10 | parallel --sshlogin server.example.com \ --sshlogin foo@server2.example.net echo

If your list of hosts is server1-88.example.net with login foo:

seq 10 | parallel -Sfoo@server{1..88}.example.net echo

To distribute the commands to a list of computers, make a file mycomputers with all the computers:

server.example.com foo@server2.example.com server3.example.com

Then run:

seq 10 | parallel --sshloginfile mycomputers echo

To include the local computer add the special sshlogin ':' to the list:

server.example.com foo@server2.example.com server3.example.com :

GNU parallel will try to determine the number of CPUs on each of the remote computers, and run one job per CPU - even if the remote computers do not have the same number of CPUs.

If the number of CPUs on the remote computers is not identified correctly the number of CPUs can be added in front. Here the computer has 8 CPUs.

seq 10 | parallel --sshlogin 8/server.example.com echo

EXAMPLE: Transferring of files

To recompress gzipped files with bzip2 using a remote computer run:

find logs/ -name *.gz | \ parallel --sshlogin server.example.com \ --transfer "zcat {} | bzip2 -9 >{.}.bz2"

This will list the .gz-files in the logs directory and all directories below. Then it will transfer the files to server.example.com to the corresponding directory in $HOME/logs. On server.example.com the file will be recompressed using zcat and bzip2 resulting in the corresponding file with .gz replaced with .bz2.

If you want the resulting bz2-file to be transferred back to the local computer add --return {.}.bz2:

find logs/ -name *.gz | \ parallel --sshlogin server.example.com \ --transfer --return {.}.bz2 "zcat {} | bzip2 -9 >{.}.bz2"

After the recompressing is done the .bz2-file is transferred back to the local computer and put next to the original .gz-file.

If you want to delete the transferred files on the remote computer add --cleanup. This will remove both the file transferred to the remote computer and the files transferred from the remote computer:

find logs/ -name *.gz | \ parallel --sshlogin server.example.com \ --transfer --return {.}.bz2 --cleanup "zcat {} | bzip2 -9 >{.}.bz2"

If you want run on several computers add the computers to --sshlogin either using ',' or multiple --sshlogin:

find logs/ -name *.gz | \ parallel --sshlogin server.example.com,server2.example.com \ --sshlogin server3.example.com \ --transfer --return {.}.bz2 --cleanup "zcat {} | bzip2 -9 >{.}.bz2"

You can add the local computer using --sshlogin :. This will disable the removing and transferring for the local computer only:

find logs/ -name *.gz | \ parallel --sshlogin server.example.com,server2.example.com \ --sshlogin server3.example.com \ --sshlogin : \ --transfer --return {.}.bz2 --cleanup "zcat {} | bzip2 -9 >{.}.bz2"

Often --transfer, --return and --cleanup are used together. They can be shortened to --trc:

find logs/ -name *.gz | \ parallel --sshlogin server.example.com,server2.example.com \ --sshlogin server3.example.com \ --sshlogin : \ --trc {.}.bz2 "zcat {} | bzip2 -9 >{.}.bz2"

With the file mycomputers containing the list of computers it becomes:

find logs/ -name *.gz | parallel --sshloginfile mycomputers \ --trc {.}.bz2 "zcat {} | bzip2 -9 >{.}.bz2"

If the file ~/.parallel/sshloginfile contains the list of computers the special short hand -S .. can be used:

find logs/ -name *.gz | parallel -S .. \ --trc {.}.bz2 "zcat {} | bzip2 -9 >{.}.bz2"

EXAMPLE: Advanced file transfer

Assume you have files in in/*, want them processed on server, and transferred back into /other/dir:

parallel -S server --trc /other/dir/./{/}.out \ cp {/} {/}.out ::: in/./*

EXAMPLE: Distributing work to local and remote computers

Convert *.mp3 to *.ogg running one process per CPU on local computer and server2:

parallel --trc {.}.ogg -S server2,: \ mpg321 -w - {} | oggenc -q0 - -o {.}.ogg ::: *.mp3

EXAMPLE: Running the same command on remote computers

To run the command uptime on remote computers you can do:

parallel --tag --nonall -S server1,server2 uptime

--nonall reads no arguments. If you have a list of jobs you want to run on each computer you can do:

parallel --tag --onall -S server1,server2 echo ::: 1 2 3

Remove --tag if you do not want the sshlogin added before the output.

If you have a lot of hosts use '-j0' to access more hosts in parallel.

EXAMPLE: Running 'sudo' on remote computers

Put the password into passwordfile then run:

parallel --ssh cat passwordfile | ssh --nonall \ -S user@server1,user@server2 sudo -S ls -l /root

EXAMPLE: Using remote computers behind NAT wall

If the workers are behind a NAT wall, you need some trickery to get to them.

If you can ssh to a jumphost, and reach the workers from there, then the obvious solution would be this, but it does not work:

parallel --ssh ssh jumphost ssh -S host1 echo ::: DOES NOT WORK

It does not work because the command is dequoted by ssh twice where as GNU parallel only expects it to be dequoted once.

You can use a bash function and have GNU parallel quote the command:

jumpssh() { ssh -A jumphost ssh $(parallel --shellquote ::: "$@"); } export -f jumpssh parallel --ssh jumpssh -S host1 echo ::: this works

Or you can instead put this in ~/.ssh/config:

Host host1 host2 host3 ProxyCommand ssh jumphost.domain nc -w 1 %h 22

It requires nc(netcat) to be installed on jumphost. With this you can simply:

parallel -S host1,host2,host3 echo ::: This does work

No jumphost, but port forwards

If there is no jumphost but each server has port 22 forwarded from the firewall (e.g. the firewall's port 22001 = port 22 on host1, 22002 = host2, 22003 = host3) then you can use ~/.ssh/config:

Host host1.v Port 22001 Host host2.v Port 22002 Host host3.v Port 22003 Host *.v Hostname firewall

And then use host{1..3}.v as normal hosts:

parallel -S host1.v,host2.v,host3.v echo ::: a b c

No jumphost, no port forwards

If ports cannot be forwarded, you need some sort of VPN to traverse the NAT-wall. TOR is one options for that, as it is very easy to get working.

You need to install TOR and setup a hidden service. In torrc put:

HiddenServiceDir /var/lib/tor/hidden_service/ HiddenServicePort 22 127.0.0.1:22

Then start TOR: /etc/init.d/tor restart

The TOR hostname is now in /var/lib/tor/hidden_service/hostname and is something similar to izjafdceobowklhz.onion. Now you simply prepend torsocks to ssh:

parallel --ssh torsocks ssh -S izjafdceobowklhz.onion \ -S zfcdaeiojoklbwhz.onion,auclucjzobowklhi.onion echo ::: a b c

If not all hosts are accessible through TOR:

parallel -S torsocks ssh izjafdceobowklhz.onion,host2,host3 \ echo ::: a b c

See more ssh tricks on https://en.wikibooks.org/wiki/OpenSSH/Cookbook/Proxies_and_Jump_Hosts

EXAMPLE: Use outrun instead of ssh

outrun lets you run a command on a remote server. outrun sets up a connection to access files at the source server, and automatically transfers files. outrun must be installed on the remote system.

You can use outrun in an sshlogin this way:

parallel -S outrun user@server eval command

EXAMPLE: Slurm cluster

The Slurm Workload Manager is used in many clusters.

Here is a simple example of using GNU parallel to call srun:

#!/bin/bash #SBATCH --time 00:02:00 #SBATCH --ntasks=4 #SBATCH --job-name GnuParallelDemo #SBATCH --output gnuparallel.out module purge module load gnu_parallel my_parallel="parallel --delay .2 -j $SLURM_NTASKS" my_srun="srun --export=all --exclusive -n1 --cpus-per-task=1 --cpu-bind=cores" $my_parallel "$my_srun" echo This is job {} ::: {1..20}

EXAMPLE: Parallelizing rsync

rsync is a great tool, but sometimes it will not fill up the available bandwidth. Running multiple rsync in parallel can fix this.

cd src-dir find . -type f | parallel -j10 -X rsync -zR -Ha ./{} fooserver:/dest-dir/

Adjust -j10 until you find the optimal number.

rsync -R will create the needed subdirectories, so all files are not put into a single dir. The ./ is needed so the resulting command looks similar to:

rsync -zR ././sub/dir/file fooserver:/dest-dir/

The /./ is what rsync -R works on.

If you are unable to push data, but need to pull them and the files are called digits.png (e.g. 000000.png) you might be able to do:

seq -w 0 99 | parallel rsync -Havessh fooserver:src/*{}.png destdir/

EXAMPLE: Use multiple inputs in one command

Copy files like foo.es.ext to foo.ext:

ls *.es.* | perl -pe print; s/\.es// | parallel -N2 cp {1} {2}

The perl command spits out 2 lines for each input. GNU parallel takes 2 inputs (using -N2) and replaces {1} and {2} with the inputs.

Count in binary:

parallel -k echo ::: 0 1 ::: 0 1 ::: 0 1 ::: 0 1 ::: 0 1 ::: 0 1

Print the number on the opposing sides of a six sided die:

parallel --link -a <(seq 6) -a <(seq 6 -1 1) echo parallel --link echo :::: <(seq 6) <(seq 6 -1 1)

Convert files from all subdirs to PNG-files with consecutive numbers (useful for making input PNG's for ffmpeg):

parallel --link -a <(find . -type f | sort) \ -a <(seq $(find . -type f|wc -l)) convert {1} {2}.png

Alternative version:

find . -type f | sort | parallel convert {} {#}.png

EXAMPLE: Use a table as input

Content of table_file.tsv:

foo<TAB>bar baz <TAB> quux

To run:

cmd -o bar -i foo cmd -o quux -i baz

you can run:

parallel -a table_file.tsv --colsep \t cmd -o {2} -i {1}

Note: The default for GNU parallel is to remove the spaces around the columns. To keep the spaces:

parallel -a table_file.tsv --trim n --colsep \t cmd -o {2} -i {1}

EXAMPLE: Output to database

GNU parallel can output to a database table and a CSV-file:

dburl=csv:///%2Ftmp%2Fmydir dbtableurl=$dburl/mytable.csv parallel --sqlandworker $dbtableurl seq ::: {1..10}

It is rather slow and takes up a lot of CPU time because GNU parallel parses the whole CSV file for each update.

A better approach is to use an SQLite-base and then convert that to CSV:

dburl=sqlite3:///%2Ftmp%2Fmy.sqlite dbtableurl=$dburl/mytable parallel --sqlandworker $dbtableurl seq ::: {1..10} sql $dburl .headers on .mode csv SELECT * FROM mytable;

This takes around a second per job.

If you have access to a real database system, such as PostgreSQL, it is even faster:

dburl=pg://user:pass@host/mydb dbtableurl=$dburl/mytable parallel --sqlandworker $dbtableurl seq ::: {1..10} sql $dburl \ "COPY (SELECT * FROM mytable) TO stdout DELIMITER , CSV HEADER;"

Or MySQL:

dburl=mysql://user:pass@host/mydb dbtableurl=$dburl/mytable parallel --sqlandworker $dbtableurl seq ::: {1..10} sql -p -B $dburl "SELECT * FROM mytable;" > mytable.tsv perl -pe s/"/""/g; s/\t/","/g; s/^/"/; s/$/"/; %s=("\\" => "\\", "t" => "\t", "n" => "\n"); s/\\([\\tn])/$s{$1}/g; mytable.tsv

EXAMPLE: Output to CSV-file for R

If you have no need for the advanced job distribution control that a database provides, but you simply want output into a CSV file that you can read into R or LibreCalc, then you can use --results:

parallel --results my.csv seq ::: 10 20 30 R > mydf <- read.csv("my.csv"); > print(mydf[2,]) > write(as.character(mydf[2,c("Stdout")]),)

EXAMPLE: Use XML as input

The show Aflyttet on Radio 24syv publishes an RSS feed with their audio podcasts on: http://arkiv.radio24syv.dk/audiopodcast/channel/4466232

Using xpath you can extract the URLs for 2019 and download them using GNU parallel:

wget -O - http://arkiv.radio24syv.dk/audiopodcast/channel/4466232 | \ xpath -e "//pubDate[contains(text(),2019)]/../enclosure/@url" | \ parallel -u wget {= s/ url="//; s/"//; =}

EXAMPLE: Run the same command 10 times

If you want to run the same command with the same arguments 10 times in parallel you can do:

seq 10 | parallel -n0 my_command my_args

EXAMPLE: Working as cat | sh. Resource inexpensive jobs and evaluation

GNU parallel can work similar to cat | sh.

A resource inexpensive job is a job that takes very little CPU, disk I/O and network I/O. Ping is an example of a resource inexpensive job. wget is too - if the webpages are small.

The content of the file jobs_to_run:

ping -c 1 10.0.0.1 wget http://example.com/status.cgi?ip=10.0.0.1 ping -c 1 10.0.0.2 wget http://example.com/status.cgi?ip=10.0.0.2 ... ping -c 1 10.0.0.255 wget http://example.com/status.cgi?ip=10.0.0.255

To run 100 processes simultaneously do:

parallel -j 100 < jobs_to_run

As there is not a command the jobs will be evaluated by the shell.

EXAMPLE: Call program with FASTA sequence

FASTA files have the format:

>Sequence name1 sequence sequence continued >Sequence name2 sequence sequence continued more sequence

To call myprog with the sequence as argument run:

cat file.fasta | parallel --pipe -N1 --recstart > --rrs \ read a; echo Name: "$a"; myprog $(tr -d "\n")

EXAMPLE: Call program with interleaved FASTQ records

FASTQ files have the format:

@M10991:61:000000000-A7EML:1:1101:14011:1001 1:N:0:28 CTCCTAGGTCGGCATGATGGGGGAAGGAGAGCATGGGAAGAAATGAGAGAGTAGCAAGG + #8BCCGGGGGFEFECFGGGGGGGGG@;FFGGGEG@FF<EE<@FFC,CEGCCGGFF<FGF

Interleaved FASTQ starts with a line like these:

@HWUSI-EAS100R:6:73:941:1973#0/1 @EAS139:136:FC706VJ:2:2104:15343:197393 1:Y:18:ATCACG @EAS139:136:FC706VJ:2:2104:15343:197393 1:N:18:1

where '/1' and ' 1:' determines this is read 1.

This will cut big.fq into one chunk per CPU core and pass it on stdin (standard input) to the program fastq-reader:

parallel --pipepart -a big.fq --block -1 --regexp \ --recend \n --recstart @.*(/1| 1:.*)\n[A-Za-z\n\.~] \ fastq-reader

EXAMPLE: Processing a big file using more CPUs

To process a big file or some output you can use --pipe to split up the data into blocks and pipe the blocks into the processing program.

If the program is gzip -9 you can do:

cat bigfile | parallel --pipe --recend -k gzip -9 > bigfile.gz

This will split bigfile into blocks of 1 MB and pass that to gzip -9 in parallel. One gzip will be run per CPU. The output of gzip -9 will be kept in order and saved to bigfile.gz

gzip works fine if the output is appended, but some processing does not work like that - for example sorting. For this GNU parallel can put the output of each command into a file. This will sort a big file in parallel:

cat bigfile | parallel --pipe --files sort |\ parallel -Xj1 sort -m {} ; rm {} >bigfile.sort

Here bigfile is split into blocks of around 1MB, each block ending in '\n' (which is the default for --recend). Each block is passed to sort and the output from sort is saved into files. These files are passed to the second parallel that runs sort -m on the files before it removes the files. The output is saved to bigfile.sort.

GNU parallel's --pipe maxes out at around 100 MB/s because every byte has to be copied through GNU parallel. But if bigfile is a real (seekable) file GNU parallel can by-pass the copying and send the parts directly to the program:

parallel --pipepart --block 100m -a bigfile --files sort |\ parallel -Xj1 sort -m {} ; rm {} >bigfile.sort

EXAMPLE: Grouping input lines

When processing with --pipe you may have lines grouped by a value. Here is my.csv:

Transaction Customer Item 1 a 53 2 b 65 3 b 82 4 c 96 5 c 67 6 c 13 7 d 90 8 d 43 9 d 91 10 d 84 11 e 72 12 e 102 13 e 63 14 e 56 15 e 74

Let us assume you want GNU parallel to process each customer. In other words: You want all the transactions for a single customer to be treated as a single record.

To do this we preprocess the data with a program that inserts a record separator before each customer (column 2 = $F[1]). Here we first make a 50 character random string, which we then use as the separator:

sep=`perl -e print map { ("a".."z","A".."Z")[rand(52)] } (1..50);` cat my.csv | \ perl -ape $F[1] ne $l and print "$sep"; $l = $F[1] | \ parallel --recend $sep --rrs --pipe -N1 wc

If your program can process multiple customers replace -N1 with a reasonable --blocksize.

EXAMPLE: Running more than 250 jobs workaround

If you need to run a massive amount of jobs in parallel, then you will likely hit the filehandle limit which is often around 250 jobs. If you are super user you can raise the limit in /etc/security/limits.conf but you can also use this workaround. The filehandle limit is per process. That means that if you just spawn more GNU parallels then each of them can run 250 jobs. This will spawn up to 2500 jobs:

cat myinput |\ parallel --pipe -N 50 --roundrobin -j50 parallel -j50 your_prg

This will spawn up to 62500 jobs (use with caution - you need 64 GB RAM to do this, and you may need to increase /proc/sys/kernel/pid_max):

cat myinput |\ parallel --pipe -N 250 --roundrobin -j250 parallel -j250 your_prg

EXAMPLE: Working as mutex and counting semaphore

The command sem is an alias for parallel --semaphore.

A counting semaphore will allow a given number of jobs to be started in the background. When the number of jobs are running in the background, GNU sem will wait for one of these to complete before starting another command. sem --wait will wait for all jobs to complete.

Run 10 jobs concurrently in the background:

for i in *.log ; do echo $i sem -j10 gzip $i ";" echo done done sem --wait

A mutex is a counting semaphore allowing only one job to run. This will edit the file myfile and prepends the file with lines with the numbers 1 to 3.

seq 3 | parallel sem sed -i -e 1i{} myfile

As myfile can be very big it is important only one process edits the file at the same time.

Name the semaphore to have multiple different semaphores active at the same time:

seq 3 | parallel sem --id mymutex sed -i -e 1i{} myfile

EXAMPLE: Mutex for a script

Assume a script is called from cron or from a web service, but only one instance can be run at a time. With sem and --shebang-wrap the script can be made to wait for other instances to finish. Here in bash:

#!/usr/bin/sem --shebang-wrap -u --id $0 --fg /bin/bash echo This will run sleep 5 echo exclusively

Here perl:

#!/usr/bin/sem --shebang-wrap -u --id $0 --fg /usr/bin/perl print "This will run "; sleep 5; print "exclusively\n";

Here python:

#!/usr/local/bin/sem --shebang-wrap -u --id $0 --fg /usr/bin/python import time print "This will run "; time.sleep(5) print "exclusively";

EXAMPLE: Start editor with filenames from stdin (standard input)

You can use GNU parallel to start interactive programs like emacs or vi:

cat filelist | parallel --tty -X emacs cat filelist | parallel --tty -X vi

If there are more files than will fit on a single command line, the editor will be started again with the remaining files.

EXAMPLE: Running sudo

sudo requires a password to run a command as root. It caches the access, so you only need to enter the password again if you have not used sudo for a while.

The command:

parallel sudo echo ::: This is a bad idea

is no good, as you would be prompted for the sudo password for each of the jobs. You can either do:

sudo echo This parallel sudo echo ::: is a good idea

or:

sudo parallel echo ::: This is a good idea

This way you only have to enter the sudo password once.

EXAMPLE: GNU Parallel as queue system/batch manager

GNU parallel can work as a simple job queue system or batch manager. The idea is to put the jobs into a file and have GNU parallel read from that continuously. As GNU parallel will stop at end of file we use tail to continue reading:

true >jobqueue; tail -n+0 -f jobqueue | parallel

To submit your jobs to the queue:

echo my_command my_arg >> jobqueue

You can of course use -S to distribute the jobs to remote computers:

true >jobqueue; tail -n+0 -f jobqueue | parallel -S ..

Output only will be printed when reading the next input after a job has finished: So you need to submit a job after the first has finished to see the output from the first job.

If you keep this running for a long time, jobqueue will grow. A way of removing the jobs already run is by making GNU parallel stop when it hits a special value and then restart. To use --eof to make GNU parallel exit, tail also needs to be forced to exit:

true >jobqueue; while true; do tail -n+0 -f jobqueue | (parallel -E StOpHeRe -S ..; echo GNU Parallel is now done; perl -e while(<>){/StOpHeRe/ and last};print <> jobqueue > j2; (seq 1000 >> jobqueue &); echo Done appending dummy data forcing tail to exit) echo tail exited; mv j2 jobqueue done

In some cases you can run on more CPUs and computers during the night:

# Day time echo 50% > jobfile cp day_server_list ~/.parallel/sshloginfile # Night time echo 100% > jobfile cp night_server_list ~/.parallel/sshloginfile tail -n+0 -f jobqueue | parallel --jobs jobfile -S ..

GNU parallel discovers if jobfile or ~/.parallel/sshloginfile changes.

EXAMPLE: GNU Parallel as dir processor

If you have a dir in which users drop files that needs to be processed you can do this on GNU/Linux (If you know what inotifywait is called on other platforms file a bug report):

inotifywait -qmre MOVED_TO -e CLOSE_WRITE --format %w%f my_dir |\ parallel -u echo

This will run the command echo on each file put into my_dir or subdirs of my_dir.

You can of course use -S to distribute the jobs to remote computers:

inotifywait -qmre MOVED_TO -e CLOSE_WRITE --format %w%f my_dir |\ parallel -S .. -u echo

If the files to be processed are in a tar file then unpacking one file and processing it immediately may be faster than first unpacking all files. Set up the dir processor as above and unpack into the dir.

Using GNU parallel as dir processor has the same limitations as using GNU parallel as queue system/batch manager.

EXAMPLE: Locate the missing package

If you have downloaded source and tried compiling it, you may have seen:

$ ./configure [...] checking for something.h... no configure: error: "libsomething not found"

Often it is not obvious which package you should install to get that file. Debian has `apt-file` to search for a file. `tracefile` from https://gitlab.com/ole.tange/tangetools can tell which files a program tried to access. In this case we are interested in one of the last files:

$ tracefile -un ./configure | tail | parallel -j0 apt-file search

SPREADING BLOCKS OF DATA

--round-robin, --pipe-part, --shard, --bin and --group-by are all specialized versions of --pipe.

In the following n is the number of jobslots given by --jobs. A record starts with --recstart and ends with --recend. It is typically a full line. A chunk is a number of full records that is approximately the size of a block. A block can contain half records, a chunk cannot.

--pipe starts one job per chunk. It reads blocks from stdin (standard input). It finds a record end near a block border and passes a chunk to the program.

--pipe-part starts one job per chunk - just like normal --pipe. It first finds record endings near all block borders in the file and then starts the jobs. By using --block -1 it will set the block size to 1/n * size-of-file. Used this way it will start n jobs in total.

--round-robin starts n jobs in total. It reads a block and passes a chunk to whichever job is ready to read. It does not parse the content except for identifying where a record ends to make sure it only passes full records.

--shard starts n jobs in total. It parses each line to read the value in the given column. Based on this value the line is passed to one of the n jobs. All lines having this value will be given to the same jobslot.

--bin works like --shard but the value of the column is the jobslot number it will be passed to. If the value is bigger than n, then n will be subtracted from the value until the values is smaller than or equal to n.

--group-by starts one job per chunk. Record borders are not given by --recend/--recstart. Instead a record is defined by a number of lines having the same value in a given column. So the value of a given column changes at a chunk border. With --pipe every line is parsed, with --pipe-part only a few lines are parsed to find the chunk border.

--group-by can be combined with --round-robin or --pipe-part.

UNIT PREFIX

Many numerical arguments in GNU parallel can be postfixed with K, M, G, T, P, k, m, g, t, or p which would multiply the number with 1024, 1048576, 1073741824, 1099511627776, 1125899906842624, 1000, 1000000, 1000000000, 1000000000000, or 1000000000000000, respectively.

You can even give it as a math expression. E.g. 1000000 can be written as 1M-12*2.024*2k.

QUOTING

GNU parallel is very liberal in quoting. You only need to quote characters that have special meaning in shell:

( ) $ ` " < > ; | \

and depending on context these needs to be quoted, too:

~ & # ! ? space * {

Therefore most people will never need more quoting than putting '\' in front of the special characters.

Often you can simply put \' around every ':

perl -ne /^\S+\s+\S+$/ and print $ARGV,"\n" file

can be quoted:

parallel perl -ne \/^\S+\s+\S+$/ and print $ARGV,"\n"\ ::: file

However, when you want to use a shell variable you need to quote the $-sign. Here is an example using $PARALLEL_SEQ. This variable is set by GNU parallel itself, so the evaluation of the $ must be done by the sub shell started by GNU parallel:

seq 10 | parallel -N2 echo seq:\$PARALLEL_SEQ arg1:{1} arg2:{2}

If the variable is set before GNU parallel starts you can do this:

VAR=this_is_set_before_starting echo test | parallel echo {} $VAR

Prints: test this_is_set_before_starting

It is a little more tricky if the variable contains more than one space in a row:

VAR="two spaces between each word" echo test | parallel echo {} \"$VAR"\

Prints: test two spaces between each word

If the variable should not be evaluated by the shell starting GNU parallel but be evaluated by the sub shell started by GNU parallel, then you need to quote it:

echo test | parallel VAR=this_is_set_after_starting \; echo {} \$VAR

Prints: test this_is_set_after_starting

It is a little more tricky if the variable contains space:

echo test |\ parallel VAR="two spaces between each word" echo {} \"$VAR"\

Prints: test two spaces between each word

$$ is the shell variable containing the process id of the shell. This will print the process id of the shell running GNU parallel:

seq 10 | parallel echo $$

And this will print the process ids of the sub shells started by GNU parallel.

seq 10 | parallel echo \$\$

If the special characters should not be evaluated by the sub shell then you need to protect it against evaluation from both the shell starting GNU parallel and the sub shell:

echo test | parallel echo {} \\\$VAR

Prints: test $VAR

GNU parallel can protect against evaluation by the sub shell by using -q:

echo test | parallel -q echo {} \$VAR

Prints: test $VAR

This is particularly useful if you have lots of quoting. If you want to run a perl script like this:

perl -ne /^\S+\s+\S+$/ and print $ARGV,"\n" file

It needs to be quoted like one of these:

ls | parallel perl -ne /^\\S+\\s+\\S+\$/\ and\ print\ \$ARGV,\"\\n\" ls | parallel perl -ne \/^\S+\s+\S+$/ and print $ARGV,"\n"\

Notice how spaces, \'s, "'s, and $'s need to be quoted. GNU parallel can do the quoting by using option -q:

ls | parallel -q perl -ne /^\S+\s+\S+$/ and print $ARGV,"\n"

However, this means you cannot make the sub shell interpret special characters. For example because of -q this WILL NOT WORK:

ls *.gz | parallel -q "zcat {} >{.}" ls *.gz | parallel -q "zcat {} | bzip2 >{.}.bz2"

because > and | need to be interpreted by the sub shell.

If you get errors like:

sh: -c: line 0: syntax error near unexpected token sh: Syntax error: Unterminated quoted string sh: -c: line 0: unexpected EOF while looking for matching ` sh: -c: line 1: syntax error: unexpected end of file zsh:1: no matches found:

then you might try using -q.

If you are using bash process substitution like <(cat foo) then you may try -q and prepending command with bash -c:

ls | parallel -q bash -c wc -c <(echo {})

Or for substituting output:

ls | parallel -q bash -c \ tar c {} | tee >(gzip >{}.tar.gz) | bzip2 >{}.tar.bz2

Conclusion: To avoid dealing with the quoting problems it may be easier just to write a small script or a function (remember to export -f the function) and have GNU parallel call that.

LIST RUNNING JOBS

If you want a list of the jobs currently running you can run:

killall -USR1 parallel

GNU parallel will then print the currently running jobs on stderr (standard error).

COMPLETE RUNNING JOBS BUT DO NOT START NEW JOBS

If you regret starting a lot of jobs you can simply break GNU parallel, but if you want to make sure you do not have half-completed jobs you should send the signal SIGHUP to GNU parallel:

killall -HUP parallel

This will tell GNU parallel to not start any new jobs, but wait until the currently running jobs are finished before exiting.

ENVIRONMENT VARIABLES

$PARALLEL_HOME

Dir where GNU parallel stores config files, semaphores, and caches information between invocations. Default: $HOME/.parallel.

$PARALLEL_ARGHOSTGROUPS

When using --hostgroups GNU parallel sets this to the hostgroups of the job. Remember to quote the $, so it gets evaluated by the correct shell. Or use --plus and {agrp}.

$PARALLEL_HOSTGROUPS

When using --hostgroups GNU parallel sets this to the hostgroups of the sshlogin that the job is run on. Remember to quote the $, so it gets evaluated by the correct shell. Or use --plus and {hgrp}.

$PARALLEL_JOBSLOT

Set by GNU parallel and can be used in jobs run by GNU parallel. Remember to quote the $, so it gets evaluated by the correct shell. Or use --plus and {slot}. $PARALLEL_JOBSLOT is the jobslot of the job. It is equal to {%} unless the job is being retried. See {%} for details.

$PARALLEL_PID

Set by GNU parallel and can be used in jobs run by GNU parallel. Remember to quote the $, so it gets evaluated by the correct shell. This makes it possible for the jobs to communicate directly to GNU parallel. Example: If each of the jobs tests a solution and one of jobs finds the solution the job can tell GNU parallel not to start more jobs by: kill -HUP $PARALLEL_PID. This only works on the local computer.

$PARALLEL_RSYNC_OPTS

Options to pass on to rsync. Defaults to: -rlDzR.

$PARALLEL_SHELL

Use this shell for the commands run by GNU parallel:

$PARALLEL_SHELL. If undefined use:
The shell that started GNU parallel. If that cannot be determined:
$SHELL. If undefined use:
/bin/sh

$PARALLEL_SSH

GNU parallel defaults to using the ssh command for remote access. This can be overridden with $PARALLEL_SSH, which again can be overridden with --ssh. It can also be set on a per server basis (see --sshlogin).

$PARALLEL_SSHHOST

Set by GNU parallel and can be used in jobs run by GNU parallel. Remember to quote the $, so it gets evaluated by the correct shell. Or use --plus and {host}. $PARALLEL_SSHHOST is the host part of an sshlogin line. E.g. 4//usr/bin/specialssh user@host becomes: host

$PARALLEL_SSHLOGIN

Set by GNU parallel and can be used in jobs run by GNU parallel. Remember to quote the $, so it gets evaluated by the correct shell. Or use --plus and {sshlogin}. The value is the sshlogin line with number of cores removed. E.g. 4//usr/bin/specialssh user@host becomes: /usr/bin/specialssh user@host

$PARALLEL_SEQ

Set by GNU parallel and can be used in jobs run by GNU parallel. Remember to quote the $, so it gets evaluated by the correct shell. $PARALLEL_SEQ is the sequence number of the job running. Example: seq 10 | parallel -N2 \ echo seq:$PARALLEL_SEQ arg1:{1} arg2:{2} {#} is a shorthand for $PARALLEL_SEQ.

$PARALLEL_TMUX

Path to tmux. If unset the tmux in $PATH is used.

$TMPDIR

Directory for temporary files. See: --tmpdir.

$PARALLEL

The environment variable $PARALLEL will be used as default options for GNU parallel. If the variable contains special shell characters (e.g. $, *, or space) then these need to be to be escaped with \. Example: cat list | parallel -j1 -k -v ls cat list | parallel -j1 -k -v -S"myssh user@server" ls can be written as: cat list | PARALLEL="-kvj1" parallel ls cat list | PARALLEL=-kvj1 -S myssh\ user@server \ parallel echo Notice the \ after 'myssh' is needed because 'myssh' and 'user@server' must be one argument.

DEFAULT PROFILE (CONFIG FILE)

The global configuration file /etc/parallel/config, followed by user configuration file ~/.parallel/config (formerly known as .parallelrc) will be read in turn if they exist. Lines starting with '#' will be ignored. The format can follow that of the environment variable $PARALLEL, but it is often easier to simply put each option on its own line.

Options on the command line take precedence, followed by the environment variable $PARALLEL, user configuration file ~/.parallel/config, and finally the global configuration file /etc/parallel/config.

Note that no file that is read for options, nor the environment variable $PARALLEL, may contain retired options such as --tollef.

PROFILE FILES

If --profile set, GNU parallel will read the profile from that file rather than the global or user configuration files. You can have multiple --profiles.

Profiles are searched for in ~/.parallel. If the name starts with / it is seen as an absolute path. If the name starts with ./ it is seen as a relative path from current dir.

Example: Profile for running a command on every sshlogin in ~/.ssh/sshlogins and prepend the output with the sshlogin:

echo --tag -S .. --nonall > ~/.parallel/nonall_profile parallel -J nonall_profile uptime

Example: Profile for running every command with -j-1 and nice

echo -j-1 nice > ~/.parallel/nice_profile parallel -J nice_profile bzip2 -9 ::: *

Example: Profile for running a perl script before every command:

echo "perl -e \$a=\$\$; print \$a,\" \",\$PARALLEL_SEQ,\" \";;" \ > ~/.parallel/pre_perl parallel -J pre_perl echo ::: *

Note how the $ and " need to be quoted using \.

Example: Profile for running distributed jobs with nice on the remote computers:

echo -S .. nice > ~/.parallel/dist parallel -J dist --trc {.}.bz2 bzip2 -9 ::: *

EXIT STATUS

Exit status depends on --halt-on-error if one of these is used: success=X, success=Y%, fail=Y%.

All jobs ran without error. If success=X is used: X jobs ran without error. If success=Y% is used: Y% of the jobs ran without error.

1-100: Some of the jobs failed. The exit status gives the number of failed jobs. If Y% is used the exit status is the percentage of jobs that failed.

More than 100 jobs failed.
Other error.

-1 (In joblog and SQL table): Killed by Ctrl-C, timeout, not enough memory or similar.
-2 (In joblog and SQL table): skip() was called in {= =}.
-1000 (In SQL table): Job is ready to run (set by --sqlmaster).
-1220 (In SQL table): Job is taken by worker (set by --sqlworker).

If fail=1 is used, the exit status will be the exit status of the failing job.

DIFFERENCES BETWEEN GNU Parallel AND ALTERNATIVES

See: man parallel_alternatives

BUGS

Quoting of newline

Because of the way newline is quoted this will not work:

echo 1,2,3 | parallel -vkd, "echo a{}b"

However, these will all work:

echo 1,2,3 | parallel -vkd, echo a{}b echo 1,2,3 | parallel -vkd, "echo a{}b" echo 1,2,3 | parallel -vkd, "echo a"{}"b"

Speed

Startup

GNU parallel is slow at starting up - around 250 ms the first time and 150 ms after that.

Job startup

SSH

When using multiple computers GNU parallel opens ssh connections to them to figure out how many connections can be used reliably simultaneously (Namely SSHD's MaxStartups). This test is done for each host in serial, so if your --sshloginfile contains many hosts it may be slow.

If your jobs are short you may see that there are fewer jobs running on the remote systems than expected. This is due to time spent logging in and out. -M may help here.

Disk access

A single disk can normally read data faster if it reads one file at a time instead of reading a lot of files in parallel, as this will avoid disk seeks. However, newer disk systems with multiple drives can read faster if reading from multiple files in parallel.

If the jobs are of the form read-all-compute-all-write-all, so everything is read before anything is written, it may be faster to force only one disk access at the time:

sem --id diskio cat file | compute | sem --id diskio cat > file

If the jobs are of the form read-compute-write, so writing starts before all reading is done, it may be faster to force only one reader and writer at the time:

sem --id read cat file | compute | sem --id write cat > file

If the jobs are of the form read-compute-read-compute, it may be faster to run more jobs in parallel than the system has CPUs, as some of the jobs will be stuck waiting for disk access.

--nice limits command length

The current implementation of --nice is too pessimistic in the max allowed command length. It only uses a little more than half of what it could. This affects -X and -m. If this becomes a real problem for you, file a bug-report.

Aliases and functions do not work

If you get:

Cant exec "command": No such file or directory

or:

open3: exec of by command failed

or:

/bin/bash: command: command not found

it may be because command is not known, but it could also be because command is an alias or a function. If it is a function you need to export -f the function first or use env_parallel. An alias will only work if you use env_parallel.

Database with MySQL fails randomly

The --sql* options may fail randomly with MySQL. This problem does not exist with PostgreSQL.

REPORTING BUGS

Report bugs to <bug-parallel@gnu.org> or https://savannah.gnu.org/bugs/?func=additem&group=parallel

When you write your report, please keep in mind, that you must give the reader enough information to be able to run exactly what you run. So you need to include all data and programs that you use to show the problem.

See a perfect bug report on https://lists.gnu.org/archive/html/bug-parallel/2015-01/msg00000.html

Your bug report should always include:

The error message you get (if any). If the error message is not from GNU parallel you need to show why you think GNU parallel caused this.
The complete output of parallel --version. If you are not running the latest released version (see https://ftp.gnu.org/gnu/parallel/) you should specify why you believe the problem is not fixed in that version.
A minimal, complete, and verifiable example (See description on https://stackoverflow.com/help/mcve). It should be a complete example that others can run which shows the problem including all files needed to run the example. This should preferably be small and simple, so try to remove as many options as possible. A combination of yes, seq, cat, echo, wc, and sleep can reproduce most errors. If your example requires large files, see if you can make them with something like seq 100000000 > bigfile or yes | head -n 1000000000 > file. If you need multiple columns: paste <(seq 1000) <(seq 1000 1999) If your example requires remote execution, see if you can use localhost - maybe using another login. If you have access to a different system (maybe a VirtualBox on your own machine), test if your MCVE shows the problem on that system. If it does not, read below.
The output of your example. If your problem is not easily reproduced by others, the output might help them figure out the problem.
Whether you have watched the intro videos (https://www.youtube.com/playlist?list=PL284C9FF2488BC6D1), walked through the tutorial (man parallel_tutorial), and read the EXAMPLE section in the man page (man parallel - search for EXAMPLE:).

Bug dependent on environment

If you suspect the error is dependent on your environment or distribution, please see if you can reproduce the error on one of these VirtualBox images: https://sourceforge.net/projects/virtualboximage/files/ https://www.osboxes.org/virtualbox-images/

Specifying the name of your distribution is not enough as you may have installed software that is not in the VirtualBox images.

If you cannot reproduce the error on any of the VirtualBox images above, see if you can build a VirtualBox image on which you can reproduce the error. If not you should assume the debugging will be done through you. That will put more burden on you and it is extra important you give any information that help. In general the problem will be fixed faster and with less work for you if you can reproduce the error on a VirtualBox.

In summary

Your report must include:

parallel --version
output + error message
full example including all files
VirtualBox image, if you cannot reproduce it on other systems

LICENSE

This program is free software; you can redistribute it and/or modify it under the terms of the GNU General Public License as published by the Free Software Foundation; either version 3 of the License, or at your option any later version.

This program is distributed in the hope that it will be useful, but WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License for more details.

You should have received a copy of the GNU General Public License along with this program. If not, see <https://www.gnu.org/licenses/>.

Documentation license I

Permission is granted to copy, distribute and/or modify this documentation under the terms of the GNU Free Documentation License, Version 1.3 or any later version published by the Free Software Foundation; with no Invariant Sections, with no Front-Cover Texts, and with no Back-Cover Texts. A copy of the license is included in the file LICENSES/GFDL-1.3-or-later.txt.

Documentation license II

You are free:

to Share: to copy, distribute and transmit the work
to Remix: to adapt the work

Under the following conditions:

Attribution: You must attribute the work in the manner specified by the author or licensor (but not in any way that suggests that they endorse you or your use of the work).
Share Alike: If you alter, transform, or build upon this work, you may distribute the resulting work only under the same, similar or a compatible license.

With the understanding that:

Waiver

Any of the above conditions can be waived if you get permission from the copyright holder.

Public Domain

Where the work or any of its elements is in the public domain under applicable law, that status is in no way affected by the license.

Other Rights

In no way are any of the following rights affected by the license:

Your fair dealing or fair use rights, or other applicable copyright exceptions and limitations;
The author's moral rights;
Rights other persons may have either in the work itself or in how the work is used, such as publicity or privacy rights.

Notice

For any reuse or distribution, you must make clear to others the license terms of this work.

A copy of the full license is included in the file as LICENCES/CC-BY-SA-4.0.txt

DEPENDENCIES

GNU parallel uses Perl, and the Perl modules Getopt::Long, IPC::Open3, Symbol, IO::File, POSIX, and File::Temp.

For --csv it uses the Perl module Text::CSV.

For remote usage it uses rsync with ssh.

AUTHOR

When using GNU parallel for a publication please cite:

O. Tange (2011): GNU Parallel - The Command-Line Power Tool, ;login: The USENIX Magazine, February 2011:42-47.

This helps funding further development; and it won't cost you a cent. If you pay 10000 EUR you should feel free to use GNU Parallel without citing.

Parts of the manual concerning xargs compatibility is inspired by the manual of xargs from GNU findutils 4.4.2.