Since Parrot is both an assembler and a bytecode interpreter, it has options to control both behaviors. Some options may have changed by the time you read this, especially options related to debugging and optimization. The document imcc/docs/running.pod should have the latest details. Or just run parrot --help.
parrot [options] file [arguments]
The file is either an .pir or .pasm source file or a Parrot bytecode file. Parrot creates an
Array object to hold the command-line arguments and stores it in
P5 on program start.
Assume PASM input on
stdin. When Parrot runs a source file with a .pasm extension, it parses the file as pure PASM code. This switch turns on PASM parsing (instead of the default PIR parsing) when a source file is read from
Assume PBC file on
stdin. When Parrot runs a bytecode file with a .pbc extension, it immediately executes the file. This option tells Parrot to immediately execute a bytecode file piped in on
Turn on debugging output. The
-d switch takes an optional argument, which is a hex value of debug bits. The individual bits are shown in Table 11-3. When hexbits isn't specified, the default debugging level is 0001. If hexbits is separated from the
-d switch by whitespace, it has to start with a number.
To produce a huge output on
stderr, turn on all the debugging bits:
$ parrot -d 0ffff ...
Show debug option bits.
Print a short summary of options to
stdout and exit.
Act like an assembler. With this switch Parrot won't run code unless it's combined with the
-r switch. If the name of outputfile ends with a .pbc extension, Parrot writes a Parrot bytecode file. If outputfile ends with a .pasm extension, Parrot writes a PASM source file, even if the input file was also PASM. This can be handy to check various optimizations when you run Parrot with the
Immediately execute bytecode. This is the default unless
-o is present. The combination of
output.pbc writes a bytecode file and executes the generated PBC.
-v switch (
-v) shows which files are worked on and prints a summary of register usage and optimization statistics. Two
-v switches (
-v) also prints a line for each individual processing step.
yydebug for yacc/bison.
Show output of macro expansions and quit.
Print the program version to
stdout and exit.
Turn on optimizations. The flags currently implemented are shown in Table 11-4.
The interpreter options are mainly for selecting which run-time core to use for interpreting bytecode. The current default is the computed goto core if it's available. Otherwise the fast core is used.
Run with the slow core
Activate bounds checking. This also runs with the slow core as a side effect.
Run with the fast core.
Performs a full GC run before every op dispatch (good for debugging GC problems)
Activate profiling. This prints a summary of opcode usage and execution times after the program stops. It also runs within the slow core.
Trace execution. This also turns on the slow core.
Turn on warnings.
Turn off GC. This is for debugging only.
Wait for a keypress before running.
Cleanup up allocated memory when the final interpreter is destroyed.
Parrot destroys created interpreters (e.g. threads) on exit but doesn't normally free all memory for the last terminating interpreter, since the operating system will do this anyway. This can create false positives when
Parrot is run with a memory leak detector. To prevent this, use this option.