Because mpy_ld.py doesn't know the target object representation, it emits
instances of `MP_OBJ_NEW_QSTR(MP_QSTR_Foo)` as const string objects, rather
than qstrs. However this doesn't work for map keys (e.g. for a locals dict)
because the map has all_keys_are_qstrs flag is set (and also auto-complete
requires the map keys to be qstrs).
Instead, emit them as regular qstrs, and make a functioning MP_OBJ_NEW_QSTR
function available (via `native_to_obj`, also used for e.g. making
integers).
Remove the code from mpy_ld.py to emit qstrs as constant strings, but leave
behind the scaffold to emit constant objects in case we want to do use this
in the future.
Strictly this should be a .mpy sub-version bump, even though the function
table isn't changing, it does lead to a change in behavior for a new .mpy
running against old MicroPython. `mp_native_to_obj` will incorrectly return
the qstr value directly as an `mp_obj_t`, leading to unexpected results.
But given that it's broken at the moment, it seems unlikely that anyone is
relying on this, so it's not work the other downsides of a sub-version bump
(i.e. breaking pure-Python modules that use @native). The opposite case of
running an old .mpy on new MicroPython is unchanged, and remains broken in
exactly the same way.
This work was funded through GitHub Sponsors.
Signed-off-by: Jim Mussared <jim.mussared@gmail.com>
And put back our magic number, because our bytecode format differs
from upstream
drop btree & framebuf natmods, they had additional problems I didn't
want to fix right now.
To be consistent with MP_UNARY_OP_INT_FLOAT and MP_UNARY_OP_INT_COMPLEX,
and allow int() to first check if a type supports __int__ before trying
other things (as per CPython).
Signed-off-by: Damien George <damien@micropython.org>
coroutines don't have __next__; they also call themselves coroutines.
This does not change the fact that `async def` methods are generators,
but it does make them behave more like CPython.
Implements text, rodata and bss generalised relocations, as well as generic
qstr-object linking. This allows importing dynamic native modules on all
supported architectures in a unified way.
runtime0.h is part of the MicroPython ABI so it's simpler if it's
independent of config options, like MICROPY_PY_REVERSE_SPECIAL_METHODS.
What's effectively done here is to move MP_BINARY_OP_DIVMOD and
MP_BINARY_OP_CONTAINS up in the enum, then remove the #if
MICROPY_PY_REVERSE_SPECIAL_METHODS conditional.
Without this change .mpy files would need to have a feature flag for
MICROPY_PY_REVERSE_SPECIAL_METHODS (when embedding native code that uses
this enum).
This commit has no effect when MICROPY_PY_REVERSE_SPECIAL_METHODS is
disabled. With this option enabled this commit reduces code size by about
60 bytes.
The start of the bytecode prelude contains 6 numbers telling the amount of
stack needed for the Python values and exceptions, and the signature of the
function. Prior to this patch these numbers were all encoded one after the
other (2x variable unsigned integers, then 4x bytes), but using so many
bytes is unnecessary.
An entropy analysis of around 150,000 bytecode functions from the CPython
standard library showed that the optimal Shannon coding would need about
7.1 bits on average to encode these 6 numbers, compared to the existing 48
bits.
This patch attempts to get close to this optimal value by packing the 6
numbers into a single, varible-length unsigned integer via bit-wise
interleaving. The interleaving scheme is chosen to minimise the average
number of bytes needed, and at the same time keep the scheme simple enough
so it can be implemented without too much overhead in code size or speed.
The scheme requires about 10.5 bits on average to store the 6 numbers.
As a result most functions which originally took 6 bytes to encode these 6
numbers now need only 1 byte (in 80% of cases).
To make progress towards MicroPython supporting Python 3.5, adding the
matmul operator is important because it's a really "low level" part of the
language, being a new token and modifications to the grammar.
It doesn't make sense to make it configurable because 1) it would make the
grammar and lexer complicated/messy; 2) no other operators are
configurable; 3) it's not a feature that can be "dynamically plugged in"
via an import.
And matmul can be useful as a general purpose user-defined operator, it
doesn't have to be just for numpy use.
Based on work done by Jim Mussared.
So these constant objects can be loaded by dereferencing the REG_FUN_TABLE
pointer instead of loading immediate values. This reduces the size of
generated native code (when such constants are used), and means that
pointers to these constants are no longer stored in the assembly code.
This commit adds first class support for yield and yield-from in the native
emitter, including send and throw support, and yields enclosed in exception
handlers (which requires pulling down the NLR stack before yielding, then
rebuilding it when resuming).
This has been fully tested and is working on unix x86 and x86-64, and
stm32. Also basic tests have been done with the esp8266 port. Performance
of existing native code is unchanged.
This commit changes native code to handle constant objects like bytecode:
instead of storing the pointers inside the native code they are now stored
in a separate constant table (such pointers include objects like bignum,
bytes, and raw code for nested functions). This removes the need for the
GC to scan native code for root pointers, and takes a step towards making
native code independent of the runtime (eg so it can be compiled offline by
mpy-cross).
Note that the changes to the struct scope_t did not increase its size: on a
32-bit architecture it is still 48 bytes, and on a 64-bit architecture it
decreased from 80 to 72 bytes.
This commit makes viper functions have the same signature as native
functions, at the level of the emitter/assembler. This means that viper
functions can now be wrapped in the same uPy object as native functions.
Viper functions are now responsible for parsing their arguments (before it
was done by the runtime), and this makes calling them more efficient (in
most cases) because the viper entry code can be custom generated to suit
the signature of the function.
This change also opens the way forward for viper functions to take
arbitrary numbers of arguments, and for them to handle globals correctly,
among other things.
Prior to this commit a function compiled with the native decorator
@micropython.native would not work correctly when accessing global
variables, because the globals dict was not being set upon function entry.
This commit fixes this problem by, upon function entry, setting as the
current globals dict the globals dict context the function was defined
within, as per normal Python semantics, and as bytecode does. Upon
function exit the original globals dict is restored.
In order to restore the globals dict when an exception is raised the native
function must guard its internals with an nlr_push/nlr_pop pair. Because
this push/pop is relatively expensive, in both C stack usage for the
nlr_buf_t and CPU execution time, the implementation here optimises things
as much as possible. First, the compiler keeps track of whether a function
even needs to access global variables. Using this information the native
emitter then generates three different kinds of code:
1. no globals used, no exception handlers: no nlr handling code and no
setting of the globals dict.
2. globals used, no exception handlers: an nlr_buf_t is allocated on the
C stack but it is not used if the globals dict is unchanged, saving
execution time because nlr_push/nlr_pop don't need to run.
3. function has exception handlers, may use globals: an nlr_buf_t is
allocated and nlr_push/nlr_pop are always called.
In the end, native functions that don't access globals and don't have
exception handlers will run more efficiently than those that do.
Fixes issue #1573.
Before this patch MP_BINARY_OP_IN had two meanings: coming from bytecode it
meant that the args needed to be swapped, but coming from within the
runtime meant that the args were already in the correct order. This lead
to some confusion in the code and comments stating how args were reversed.
It also lead to 2 bugs: 1) containment for a subclass of a native type
didn't work; 2) the expression "{True} in True" would illegally succeed and
return True. In both of these cases it was because the args to
MP_BINARY_OP_IN ended up being reversed twice.
To fix these things this patch introduces MP_BINARY_OP_CONTAINS which
corresponds exactly to the __contains__ special method, and this is the
operator that built-in types should implement. MP_BINARY_OP_IN is now only
emitted by the compiler and is converted to MP_BINARY_OP_CONTAINS by
swapping the arguments.
Introduction of ports subdirectory where all ports are moved to
The main change in this release is the introduction of a "ports/"
subdirectory at the top-level of the repository, and all of the ports are
moved here. In the process the "stmhal" port is renamed to "stm32" to
better reflect the MCU that it targets. In addition, the STM32 CMSIS and
HAL sources are moved to a new submodule called "stm32lib".
The bytecode has changed in this release, compared to the previous release,
and as a consequence the .mpy version number has increased to version 3.
This means that scripts compiled with the previous mpy-cross must be
recompiled to work with this new version.
There have also been various enhancements and optimisations, such as:
check for valid UTF-8 when creating str objects, support for reverse
special binary operations like __radd__, full domain checking in the math
module, support for floor-division and modulo in the viper emitter,
and addition of stack overflow checking when executing a regex.
The stm32 port sees improved support for F7 MCUs, addition of a new board
B_L475E_IOT01A based on the STM32L475, and support for the Wiznet W5500
chipset along with improved socket behaviour.
A detailed list of changes follows.
py core:
- objstr: startswith, endswith: check arg to be a string
- nlrx86,x64: replace #define of defined() with portable macro usage
- objtype: handle NotImplemented return from binary special methods
- objtype: mp_obj_class_lookup: improve debug logging
- map: remove unused new/free functions
- make m_malloc_fail() have void return type, since it doesn't return
- modstruct: in struct.pack, stop converting if there are no args left
- modstruct: check and prevent buffer-read overflow in struct unpacking
- modstruct: check and prevent buffer-write overflow in struct packing
- nlrthumb: get working again on standard Thumb arch (ie not Thumb2)
- objfloat: fix binary ops with incompatible objects
- obj: fix comparison of float/complex NaN with itself
- objtype: implement fallback for instance inplace special methods
- objtuple: properly implement comparison with incompatible types
- objstr: add check for valid UTF-8 when making a str from bytes
- objlist: properly implement comparison with incompatible types
- runtime0.h: move relational ops to the beginning of mp_binary_op_t
- runtime0.h: move MP_BINARY_OP_DIVMOD to the end of mp_binary_op_t
- objtype: make sure mp_binary_op_method_name has full size again
- runtime0.h: put inplace arith ops in front of normal operations
- builtinhelp: simplify code slightly by extracting object type
- runtime: implement dispatch for "reverse op" special methods
- nlrx86: fix building for Android/x86
- builtinhelp: change signature of help text var from pointer to array
- runtime.h: change empty mp_warning macro so var-args are non empty
- modbuiltins: implement abs() by dispatching to MP_UNARY_OP_ABS
- {objfloat,objcomplex}: optimise MP_UNARY_OP_ABS by reusing variables
- mpconfig.h: add note that using computed gotos in VM is not C99
- objstr: strip: don't strip "\0" by default
- objexcept: prevent infinite recursion when allocating exceptions
- stream: remove unnecessary checks for NULL return from vstr_add_len
- vstr: raise a RuntimeError if fixed vstr buffer overflows
- vm: use lowercase letter at start of exception message
- persistentcode: define mp_raw_code_save_file() for any unix target
- add config option to print warnings/errors to stderr
- objfloat: support raising a negative number to a fractional power
- objset: simplify set and frozenset by separating their locals dicts
- objset: check that RHS of a binary op is a set/frozenset
- objset: include the failed key in a KeyError raised from set.remove
- objtype: change type of enum-to-qstr table to uint16_t to save space
- objstr: make empty bytes object have a null-terminating byte
- mpprint: only check for null string printing when NDEBUG not defined
- objtype: clean up unary- and binary-op enum-to-qstr mapping tables
- persistentcode: bump .mpy version number to version 3
- bc: update opcode_format_table to match the bytecode
- modmath: add full checks for math domain errors
- modmath: convert log2 macro into a function
- formatfloat: don't print the negative sign of a NaN value
- formatfloat: use standard isinf, isnan funcs instead of custom ones
- modbuiltins: use existing utf8_get_char helper in builtin ord func
- emitnative: implement floor-division and modulo for viper emitter
- objtype: use CPython compatible method name for sizeof
- objtype: fit qstrs for special methods in byte type
- objtype: define all special methods if requested
- objtype: introduce MICROPY_PY_ALL_INPLACE_SPECIAL_METHODS
extmod:
- modubinascii: only include uzlib/tinf.h when it's really needed
- modussl_mbedtls: allow to compile with MBEDTLS_DEBUG_C disabled
- machine_pinbase: put PinBase singleton in ROM
- re1.5: upgrade to v0.8.2, adds hook for stack overflow checking
- modure: add stack overflow checking when executing a regex
- uos_dupterm: update uos.dupterm() and helper funcs to have index
- uos_dupterm: swallow any errors from dupterm closing the stream
- vfs: replace VLA in proxy func with small, static sized array
- modussl: add finaliser support for ussl objects
- modussl_mbedtls: allow to compile with unix coverage build
lib:
- add new submodule, stm32lib containing STM32 CMSIS and HAL source
- embed/abort_: use mp_raise_msg helper function
- libm: fix tanhf so that it correctly handles +/- infinity args
- libm: remove implementation of log2f, use MP_NEED_LOG2 instead
- axtls: update, support for SSL_EAGAIN return code
- berkeley-db-1.xx: update, allow to override MINCACHE, DEFPSIZE
drivers:
- memory/spiflash: change from hard-coded soft SPI to generic SPI
- display/ssd1306.py: improve performance of graphics methods
- nrf24l01: make nRF24L01 test script more portable
- display/ssd1306: implement SSD1306_I2C poweron method
- display/ssd1306: make poweron() work the same with SSD1306_SPI
- wiznet5k: improve the performance of socket ops with threading
- wiznet5k: get low-level W5500 driver working
tools:
- upip: upgrade to 1.2.2
- pyboard: use repr() when quoting data in error messages
- pyboard: update docstring for additional device support
tests:
- object_new: better messages, check user __new__() method
- class_new: add checks for __init__ being called and other improvements
- class_new: add another testcase for __new__/__init__ interaction
- class_inplace_op: test for inplace op fallback to normal one
- run-bench-tests: update locations of executables, now in ports/
- class_reverse_op: test for reverse arith ops special methods
- run-tests: skip class_inplace_op for minimal profile
- run-tests: fix copy-paste mistake in var name
- cpydiff: add cases for locals() discrepancies
- extmod: add test for ure regexes leading to infinite recursion
- extmod: add test for '-' in character class in regex
- run-tests: close device under test using "finally"
- net_inet: update tls test to work with CPython and incl new site
unix port:
- rename modsocket.c to modusocket.c
- modusocket: remove #if MICROPY_SOCKET_EXTRA code blocks
- enable MICROPY_PY_REVERSE_SPECIAL_METHODS
stm32 port:
- modmachine: make machine.bootloader() work when MPU is enabled
- modmachine: improve support for sleep/deepsleep on F7 MCUs
- compute PLL freq table during build instead of at run time
- modmachine: for F7 MCU, save power by reducing internal volt reg
- boards/pllvalues.py: make script work with both Python 2 and 3
- Makefile: use lib/stm32lib instead of local cmsis and hal files
- remove cmsis and hal files, they are now a submodule
- Makefile: automatically fetch stm32lib submodule if needed
- update to new STM Cube HAL library
- fix clock initialisation of L4 MCUs
- rename stmhal port directory to stm32
- remove unused usbd_msc.c file
- boards: change remaining stm32f4xx_hal_conf.h to unix line ending
- boards: change linker scripts to use "K" instead of hex byte size
- boards: fix I2C1 pin mapping on NUCLEO_F401RE/F411RE boards
- i2c: when scanning for I2C devices only do 1 probe per address
- modnwwiznet5k: release the GIL on blocking network operations
- boards: add new board B_L475E_IOT01A based on STM32L475
- make-stmconst.py: make sure mpz const data lives in ROM
- timer: make pyb.Timer() instances persistent
- mpconfigport.h: add configuration for max periphs on L4 series
- usbdev: make the USBD callback struct const so it can go in ROM
- usbdev: change static function variable to non-static
- usbdev: put all CDC state in a struct
- usbdev: put all HID state in a struct
- usbdev: simplify CDC tx/rx buffer passing
- usbdev: simplify HID tx/rx buffer passing
- usbdev/core: add state parameter to all callback functions
- usbdev: put all state for the USB device driver in a struct
- usbdev: simplify pointers to MSC state and block dev operations
- usbdev: merge all global USB device state into a single struct
- usbdev: make device descriptor callbacks take a state pointer
- usbdev: move all the USB device descriptor state into its struct
- timer: enable ARPE so that timer freq can be changed smoothly
- modnwwiznet5k: get the IP address of an established socket
- boards: fix typos in stm32f767_af.csv table
- usbd_cdc_interface: don't reset CDC output buf on initialisation
- modnwwiznet5k: implement WIZNET5K.isconnected() method
- modusocket: make getaddrinfo() work when passed an IP address
- modusocket: return OSError(-2) if getaddrinfo fails
- mpconfigport.h: add MICROPY_THREAD_YIELD() macro
- modnwwiznet5k: add support for W5500 Ethernet chip
- modnwwiznet5k: increase SPI bus speed to 42MHz
- modnwwiznet5k: implement stream ioctl for the Wiznet driver
- mphalport: improve efficiency of mp_hal_stdout_tx_strn_cooked
- make uos.dupterm() conform to specs by using extmod version
cc3200 port:
- enable micropython.kbd_intr() method
- use standard implementation of keyboard interrupt
esp8266 port:
- rename axtls_helpers.c to posix_helpers.c
- posix_helpers: set ENOMEM on memory alloc failure
- set DEFPSIZE=1024, MINCACHE=3 for "btree" module
- esp_mphal: send data in chunks to mp_uos_dupterm_tx_strn
- modnetwork: add "bssid" keyword arg to WLAN.connect() method
- modules/webrepl_setup: add info about allowed password length
zephyr port:
- Makefile: revamp "test" target after ports were moved to ports/
- use CONFIG_NET_APP_SETTINGS to setup initial network addresses
- switch to interrupt-driven pull-style console
pic16bit port:
- add definition of SEEK_SET to unistd.h
docs:
- pyboard/tutorial: add "timeout=0" to UART in pass-through example
- more xrefs to "MicroPython port" in glossary
- library/network: fix ref to "socket" module (should be "usocket")
- machine.Signal: improve style/grammar and add usage example
- library: add description of "index" parameter to uos.dupterm()
- library/micropython: fix typo in RST formatting
- library/framebuf.rst: generalise constructor to all colour formats
- btree: describe page caching policy of the underlying implementation
- esp8266/tutorial: update neopixel with example of using 4 bbp
- library/network: clarify usage of "bssid" arg in connect() method
- pyboard/quickref: add info for Switch, RTC, CAN, Accel classes
- pyboard/tutorial: update now that yellow LED also supports PWM
- esp8266/quickref: add quickref info for RTC class
- library: add missing cross-ref links for classes in pyb module
- library/network: update docs to state that W5500 is supported
- uselect: document one-shot polling mode
- usocket: elaborate descriptions
- usocket: document inet_ntop(), inet_pton()
- library/network: add dhcp_hostname parameter
- reference/isr_rules: minor typo correction
- ussl: fix module name refs and use "MicroPython port" term
- esp8266/general: add section on TLS limitations
- usocket: document that settimeout() isn't supported by all ports
- ure: add "|" (alternative) to the list of supported operators
- reference/isr_rules.rst: add tutorial on use of micropython.schedule()
travis:
- use --upgrade when pip is installing cpp-coveralls
- update build command now that stm32 Wiznet config has changed
examples:
- hwconfig_console: add .on()/.off() methods
all:
- convert mp_uint_t to mp_unary_op_t/mp_binary_op_t where appropriate
- convert remaining "mp_uint_t n_args" to "size_t n_args"
- make new ports/ sub-directory and move all ports there
- update Makefiles and others to build with new ports/ dir layout
- remove inclusion of internal py header files
- use NULL instead of "" when calling mp_raise exception helpers
README:
- update "Dependencies" section
- add explicit section on contributing
- add gcc and arm-none-eabi-newlib to list of required components
.gitattributes:
- remove obsolete entries for stmhal/hal, stmhal/cmsis
- add entries for files that will move to ports/ dir
This allows user classes to implement __abs__ special method, and saves
code size (104 bytes for x86_64), even though during refactor, an issue
was fixed and few optimizations were made:
* abs() of minimum (negative) small int value is calculated properly.
* objint_longlong and objint_mpz avoid allocating new object is the
argument is already non-negative.
If, for class X, X.__add__(Y) doesn't exist (or returns NotImplemented),
try Y.__radd__(X) instead.
This patch could be simpler, but requires undoing operand swap and
operation switch to get non-confusing error message in case __radd__
doesn't exist.
This is to allow to place reverse ops immediately after normal ops, so
they can be tested as one range (which is optimization for reverse ops
introduction in the next patch).
Originally, there were grouped in blocks of 5, to make it easier e.g.
to assess and numeric code of each. But now it makes more sense to
group it by semantics/properties, and then split in chunks still,
which usually leads to chunks of ~6 ops.
It starts a dichotomy of mp_binary_op_t values which can't appear in the
bytecode. Another reason to move it is to VALUES of OP_* and OP_INPLACE_*
nicely adjacent. This also will be needed for OP_REVERSE_*, to be soon
introduced.
