PyPy Python interpreter options

The following options can be used after translate.py targetpypystandalone or as options to py.py.

--allworkingmodules: use as many working modules as possible
--nofaking: disallow faking in the object space
--objspace -o: Object Space name
--objspace-disable_call_speedhacks: make sure that all calls go through space.call_args
--objspace-honor__builtins__: Honor the __builtins__ key of a module dictionary
--objspace-lonepycfiles: Import pyc files with no matching py file
--objspace-opcodes-CALL_LIKELY_BUILTIN: emit a special bytecode for likely calls to builtin functions
--objspace-opcodes-CALL_METHOD: emit a special bytecode for expr.name()
--objspace-std-builtinshortcut: a shortcut for operations between built-in types
--objspace-std-getattributeshortcut: track types that override __getattribute__
--objspace-std-methodcachesizeexp: 2 ** methodcachesizeexp is the size of the of the method cache
--objspace-std-multimethods: the multimethod implementation to use
--objspace-std-optimized_comparison_op: special case the comparison of integers
--objspace-std-optimized_int_add: special case the addition of two integers in BINARY_ADD
--objspace-std-optimized_list_getitem: special case the 'list[integer]' expressions
--objspace-std-sharesmallstr: always reuse the prebuilt string objects (the empty string and potentially single-char strings)
--objspace-std-withblist: good asymptotic performance for very large lists
--objspace-std-withbucketdict: use dictionaries with chained hash tables (default is open addressing)
--objspace-std-withchunklist: introducing a new nesting level to slow down list operations
--objspace-std-withfastslice: make list slicing lazy
--objspace-std-withmethodcache: try to cache method lookups
--objspace-std-withmethodcachecounter: try to cache methods and provide a counter in __pypy__. for testing purposes only.
--objspace-std-withmultidict: use dictionaries optimized for flexibility
--objspace-std-withmultilist: use lists optimized for flexibility
--objspace-std-withprebuiltchar: use prebuilt single-character string objects
--objspace-std-withprebuiltint: prebuild commonly used int objects
--objspace-std-withrangelist: enable special range list implementation that does not actually create the full list until the resulting list is mutated
--objspace-std-withrope: use ropes as the string implementation
--objspace-std-withropeunicode: use ropes for the unicode implementation
--objspace-std-withshadowtracking: track whether an instance attribute shadows a type attribute
--objspace-std-withsharingdict: use dictionaries that share the keys part
--objspace-std-withsmalldicts: handle small dictionaries differently
--objspace-std-withsmallint: use tagged integers
--objspace-std-withsmartresizablelist: only overallocate O(sqrt(n)) elements for lists
--objspace-std-withstrjoin: use strings optimized for addition
--objspace-std-withstrslice: use strings optimized for slicing
--objspace-std-withtproxy: support transparent proxies
--objspace-usepycfiles: Write and read pyc files when importing
--prebuiltintfrom: lowest integer which is prebuilt
--prebuiltintto: highest integer which is prebuilt
--pyversion: which grammar to use for app-level code
--withmod-Numeric: use module Numeric
--withmod-__builtin__: use module __builtin__
--withmod-__pypy__: use module __pypy__
--withmod-_codecs: use module _codecs
--withmod-_cslib: use module _cslib
--withmod-_demo: use module _demo
--withmod-_lsprof: use module _lsprof
--withmod-_minimal_curses: use module _minimal_curses
--withmod-_random: use module _random
--withmod-_rawffi: use module _rawffi
--withmod-_socket: use module _socket
--withmod-_sre: use module _sre
--withmod-_ssl: use module _ssl
--withmod-_stackless: use module _stackless
--withmod-_weakref: use module _weakref
--withmod-bz2: use module bz2
--withmod-cStringIO: use module cStringIO
--withmod-cclp: use module cclp
--withmod-clr: use module clr
--withmod-crypt: use module crypt
--withmod-dyngram: use module dyngram
--withmod-errno: use module errno
--withmod-fcntl: use module fcntl
--withmod-gc: use module gc
--withmod-itertools: use module itertools
--withmod-marshal: use module marshal
--withmod-math: use module math
--withmod-md5: use module md5
--withmod-mmap: use module mmap
--withmod-operator: use module operator
--withmod-posix: use module posix
--withmod-pypyjit: use module pypyjit
--withmod-rbench: use module rbench
--withmod-rctime: use module rctime
--withmod-readline: use module readline
--withmod-recparser: use module recparser
--withmod-select: use module select
--withmod-sha: use module sha
--withmod-signal: use module signal
--withmod-struct: use module struct
--withmod-symbol: use module symbol
--withmod-sys: use module sys
--withmod-termios: use module termios
--withmod-thread: use module thread
--withmod-time: use module time
--withmod-unicodedata: use module unicodedata
--withmod-zipimport: use module zipimport
--withmod-zlib: use module zlib

Internal Options

--compiler: which compiler to use for app-level code
--objspace-logbytecodes: keep track of bytecode usage
--objspace-std-logspaceoptypes: a instrumentation option: before exit, print the types seen by certain simpler bytecodes
--objspace-std-withdictmeasurement: create huge files with masses of information about dictionaries
--parser: which parser to use for app-level code
--withmod-_file: use module _file
--withmod-_pickle_support: use module _pickle_support

General translation options

The following are options of translate.py. They must be given before the targetxxx on the command line.

--opt -O: set the optimization level [0, 1, size, mem, 2, 3]

-b --backend: Backend to use for code generation
--cc: Specify compiler to use for compiling generated C
--cflags: Specify flags for the C compiler
--clever-malloc-removal: Drives inlining to remove mallocs in a clever way
--clever-malloc-removal-threshold: Threshold when to inline functions in clever malloc removal
-d --debug: Record extra annotation information
--fork-before: (UNIX) Create restartable checkpoint before step
--gc: Garbage Collection Strategy
--gcrootfinder: Strategy for finding GC Roots (framework GCs only)
--if-block-merge: Merge if ... elif chains
--inline-threshold: Threshold when to inline functions
--insist: Try hard to go on RTyping
--ldflags: Specify flags for the linker (C backend only)
--listcompr: When true, look for and special-case the sequence of operations that results from a list comprehension and attempt to pre-allocate the list
--llvm-via-c: compile llvm via C
--no-profopt: Don't use profile based optimization
--no__thread: don't use __thread for implementing TLS
--output: Output file name
--prof-based-inline: Use call count profiling to drive inlining, specify arguments
--prof-based-inline-threshold: Threshold when to inline functions for profile based inlining
--profopt: Specify profile based optimization script
--sandbox: Produce a fully-sandboxed executable
--stackless: enable stackless features during compilation
--thread: enable use of threading primitives
--translation-backendopt-constfold: Constant propagation
--translation-backendopt-heap2stack: Escape analysis and stack allocation
--translation-backendopt-inline: Do basic inlining and malloc removal
--translation-backendopt-mallocs: Remove mallocs
--translation-backendopt-none: Do not run any backend optimizations
--translation-backendopt-print_statistics: Print statistics while optimizing
--translation-backendopt-remove_asserts: Remove operations that look like 'raise AssertionError', which lets the C optimizer remove the asserts
--translation-backendopt-stack_optimization: Tranform graphs in SSI form into graphs tailored for stack based virtual machines (only for backends that support it)
--translation-cli-exception_transformer: Use exception transformer
--translation-llvm-debug: Include the llops in the source as comments
--translation-llvm-isolate: Perform an isolated import
--translation-llvm-logging: Log how long the various parts of llvm generation take
--translation-llvm-opt_options: Options passed to opt (influences level of optimization in LLVM)
--translation-ootype-mangle: Mangle names of class members
--translation-rweakref: The backend supports RPython-level weakrefs
--translation-verbose: Print extra information
--translation-withsmallfuncsets: Represent groups of less funtions than this as indices into an array
--vanilla: Try to be as portable as possible, which is not much

Internal Options

--clever-malloc-removal-heuristic: Dotted name of an heuristic function for inlining in clever malloc removal
--cli-trace-calls: Trace function calls
--inline-heuristic: Dotted name of an heuristic function for inlining
--prof-based-inline-heuristic: Dotted name of an heuristic function for profile based inlining
--raisingop2direct_call: Transform operations that can implicitly raise an exception into calls to functions that explicitly raise exceptions
--translation-simplifying: Simplify flow graphs