This commit removes all parts of code associated with the existing
MICROPY_OPT_CACHE_MAP_LOOKUP_IN_BYTECODE optimisation option, including the
-mcache-lookup-bc option to mpy-cross.
This feature originally provided a significant performance boost for Unix,
but wasn't able to be enabled for MCU targets (due to frozen bytecode), and
added significant extra complexity to generating and distributing .mpy
files.
The equivalent performance gain is now provided by the combination of
MICROPY_OPT_LOAD_ATTR_FAST_PATH and MICROPY_OPT_MAP_LOOKUP_CACHE (which has
been enabled on the unix port in the previous commit).
It's hard to provide precise performance numbers, but tests have been run
on a wide variety of architectures (x86-64, ARM Cortex, Aarch64, RISC-V,
xtensa) and they all generally agree on the qualitative improvements seen
by the combination of MICROPY_OPT_LOAD_ATTR_FAST_PATH and
MICROPY_OPT_MAP_LOOKUP_CACHE.
For example, on a "quiet" Linux x64 environment (i3-5010U @ 2.10GHz) the
change from CACHE_MAP_LOOKUP_IN_BYTECODE, to LOAD_ATTR_FAST_PATH combined
with MAP_LOOKUP_CACHE is:
diff of scores (higher is better)
N=2000 M=2000 bccache -> attrmapcache diff diff% (error%)
bm_chaos.py 13742.56 -> 13905.67 : +163.11 = +1.187% (+/-3.75%)
bm_fannkuch.py 60.13 -> 61.34 : +1.21 = +2.012% (+/-2.11%)
bm_fft.py 113083.20 -> 114793.68 : +1710.48 = +1.513% (+/-1.57%)
bm_float.py 256552.80 -> 243908.29 : -12644.51 = -4.929% (+/-1.90%)
bm_hexiom.py 521.93 -> 625.41 : +103.48 = +19.826% (+/-0.40%)
bm_nqueens.py 197544.25 -> 217713.12 : +20168.87 = +10.210% (+/-3.01%)
bm_pidigits.py 8072.98 -> 8198.75 : +125.77 = +1.558% (+/-3.22%)
misc_aes.py 17283.45 -> 16480.52 : -802.93 = -4.646% (+/-0.82%)
misc_mandel.py 99083.99 -> 128939.84 : +29855.85 = +30.132% (+/-5.88%)
misc_pystone.py 83860.10 -> 82592.56 : -1267.54 = -1.511% (+/-2.27%)
misc_raytrace.py 21490.40 -> 22227.23 : +736.83 = +3.429% (+/-1.88%)
This shows that the new optimisations are at least as good as the existing
inline-bytecode-caching, and are sometimes much better (because the new
ones apply caching to a wider variety of map lookups).
The new optimisations can also benefit code generated by the native
emitter, because they apply to the runtime rather than the generated code.
The improvement for the native emitter when LOAD_ATTR_FAST_PATH and
MAP_LOOKUP_CACHE are enabled is (same Linux environment as above):
diff of scores (higher is better)
N=2000 M=2000 native -> nat-attrmapcache diff diff% (error%)
bm_chaos.py 14130.62 -> 15464.68 : +1334.06 = +9.441% (+/-7.11%)
bm_fannkuch.py 74.96 -> 76.16 : +1.20 = +1.601% (+/-1.80%)
bm_fft.py 166682.99 -> 168221.86 : +1538.87 = +0.923% (+/-4.20%)
bm_float.py 233415.23 -> 265524.90 : +32109.67 = +13.756% (+/-2.57%)
bm_hexiom.py 628.59 -> 734.17 : +105.58 = +16.796% (+/-1.39%)
bm_nqueens.py 225418.44 -> 232926.45 : +7508.01 = +3.331% (+/-3.10%)
bm_pidigits.py 6322.00 -> 6379.52 : +57.52 = +0.910% (+/-5.62%)
misc_aes.py 20670.10 -> 27223.18 : +6553.08 = +31.703% (+/-1.56%)
misc_mandel.py 138221.11 -> 152014.01 : +13792.90 = +9.979% (+/-2.46%)
misc_pystone.py 85032.14 -> 105681.44 : +20649.30 = +24.284% (+/-2.25%)
misc_raytrace.py 19800.01 -> 23350.73 : +3550.72 = +17.933% (+/-2.79%)
In summary, compared to MICROPY_OPT_CACHE_MAP_LOOKUP_IN_BYTECODE, the new
MICROPY_OPT_LOAD_ATTR_FAST_PATH and MICROPY_OPT_MAP_LOOKUP_CACHE options:
- are simpler;
- take less code size;
- are faster (generally);
- work with code generated by the native emitter;
- can be used on embedded targets with a small and constant RAM overhead;
- allow the same .mpy bytecode to run on all targets.
See #7680 for further discussion. And see also #7653 for a discussion
about simplifying mpy-cross options.
Signed-off-by: Jim Mussared <jim.mussared@gmail.com>
- Moves definition of BOARD_FLASH_SIZE and other header files related to
flash configuration into the Makefile.
- Adds board specific clock_config.h.
- Adds board.h, pin_mux.h, and peripherals.h as they are
required by NXP MCU SDK in order to use our own clock_config.h.
- Renames board specific FlexSPI configuration files.
- Updates flash frequency of MIMXRT1020_EVK
- Creates separated flash_config files for QSPI NOR and
QSPI Hyper flash.
- Unifies VFS start address to be @ 1M for 1010 and 1020 boards.
- Unifies 1050EVK boards
- Adds support to both NOR and HyperFlash on boards with
both capabilities.
- Adds automatic FlexRAM initialization to start-up code based on
linker script and NXP HAL.
- Applies code formatting to all files in mimxrt port.
With this change the flash configuration is restructured and
organized. This simplifies the configuration process and
provides a better overview of each board's settings. With the integration
of clock_config.h, board.h, pin_mux.h, and peripherals.h we gain better
control of the settings and clock configurations. Furthermore the
implementation of an explicit FlexRAM setup improves the system
performance and allows for performance tuning.
Signed-off-by: Philipp Ebensberger
To keep things neat and tidy, we ensure that each file has 1 and only 1
newline at the end of each file.
Signed-off-by: David Lechner <david@pybricks.com>
This feature {x=} was introduced in Python 3.8 so needs a separate .exp
file to run on earlier Python versions.
See https://bugs.python.org/issue36817
Signed-off-by: Damien George <damien@micropython.org>
This makes it work like --no-follow and --no-exclusive using a mutex group
and dest. Although the current implementation with BooleanOptionAction is
neater it requires Python 3.9, so don't use this feature.
Signed-off-by: Jim Mussared <jim.mussared@gmail.com>
The --no-exclusive flag was accidentally added to the mutex group in
178198a01d.
The --soft-reset flag was accidentally added to the mutex group in
41adf17830.
These flags can be specified independently to --[no-]follow so should not
be in that mutex group.
Signed-off-by: Jim Mussared <jim.mussared@gmail.com>
And using "-B" means mpy-cross is forcefully rebuilt, sometimes with
invalid CFLAGS_EXTRA options which makes the auto-build fail.
Signed-off-by: Damien George <damien@micropython.org>
The sys.stdin.buffer and sys.stdout.buffer streams work just as well (and
are just as fast) as pyb.USB_VCP on stm32 devices, so there's no need to
have the USB_VCP specialisation code, which just adds complexity.
Also, on stm32 devices with both USB and UART (or other serial interface),
if something other than the USB_VCP port is used for the serial connection
then mpremote mount will not work because it will default to reading and
writing on USB_VCP instead of the other connected serial stream.
As part of this simplification, support for a second port as input is
removed (this feature was never exposed to the user).
Signed-off-by: Damien George <damien@micropython.org>
This implements (most of) the PEP-498 spec for f-strings and is based on
https://github.com/micropython/micropython/pull/4998 by @klardotsh.
It is implemented in the lexer as a syntax translation to `str.format`:
f"{a}" --> "{}".format(a)
It also supports:
f"{a=}" --> "a={}".format(a)
This is done by extracting the arguments into a temporary vstr buffer,
then after the string has been tokenized, the lexer input queue is saved
and the contents of the temporary vstr buffer are injected into the lexer
instead.
There are four main limitations:
- raw f-strings (`fr` or `rf` prefixes) are not supported and will raise
`SyntaxError: raw f-strings are not supported`.
- literal concatenation of f-strings with adjacent strings will fail
"{}" f"{a}" --> "{}{}".format(a) (str.format will incorrectly use
the braces from the non-f-string)
f"{a}" f"{a}" --> "{}".format(a) "{}".format(a) (cannot concatenate)
- PEP-498 requires the full parser to understand the interpolated
argument, however because this entirely runs in the lexer it cannot
resolve nested braces in expressions like
f"{'}'}"
- The !r, !s, and !a conversions are not supported.
Includes tests and cpydiffs.
Signed-off-by: Jim Mussared <jim.mussared@gmail.com>
For consistency with other board-level config macros that begin with
MICROPY_HW_USB.
Also allow boards in the mimxrt, nrf and samd ports to configure these
values.
Signed-off-by: Damien George <damien@micropython.org>
The firmware for Teensy 4.0, Teensy 4.1 and MIMXRT1020_EVK are created.
Users of other MIMXRT10xx_EVK boards should be able to build the firmware
themselves, they might need specific DEBUG settings.
The Makefile had to be changed in order to build the .bin file as well.
Coverage calculated by Codecov has the same reliability/deterministic
issues as Coveralls did, so the problem is likely to do with the output of
lcov/gcov, rather than the analysis and display of the data.
Switch from lcov to gcov for data generation to try and simplify this
process of computing coverage.
Signed-off-by: Damien George <damien@micropython.org>
Now a ctrl-C will not stop mpremote, rather this character will be passed
through to the attached device.
The mpremote version is also increased to 0.0.5.
Signed-off-by: Damien George <damien@micropython.org>
Using just the list of available ports, instead of a hard-coded list of
possible ports, means that all ports will be available for auto connection.
And the order that they will be attempted in will match what's printed by
"mpremote connect list" (and will be the same as before, trying ACMx before
USBx). Auto-connect will also now work on Mac, and will allow all COM
ports on Windows.
Signed-off-by: Damien George <damien@micropython.org>
Following on from ef16834887, this adds a
coverage build and running of the test suite on an ARM 32-bit Linux-based
architecture.
Signed-off-by: Damien George <damien@micropython.org>
This adds a coverage build and running of the test suite on a MIPS 32-bit
big endian architecture. It uses the feature of qemu to execute foreign
code as though it were native to the system (using qemu user mode). The
code compiled for MIPS will run under the qemu VM, but all syscalls made by
this code go to the host (Linux) system.
See related #7268 and #7273.
Signed-off-by: Damien George <damien@micropython.org>
As the new default behaviour, this allows PyDFU to be used with all
devices, not just the ones matching a specific set of VID/PID values. But
it's still possible to specify VID/PID if needed to narrow down the
selection of the USB device.
Signed-off-by: Tobias Thyrrestrup <tt@LEGO.com>
This can be treated by the linker the same as R_X86_64_REX_GOTPCRELX,
according to https://reviews.llvm.org/D18301.
Signed-off-by: Damien George <damien@micropython.org>
Since version 21.4b0, Black now processes one-line docstrings by stripping
leading and trailing spaces, and adding a padding space when needed to
break up """"; see https://github.com/psf/black/pull/1740
This commit makes the Python code in this repository conform to this rule.
This is now the default, but can be overridden with CLI `--no-exclusive`,
or constructing `Pyboard(..., exclusive=False)`.
Signed-off-by: Damien George <damien@micropython.org>
It's a bit of a pitfall with user C modules that including them in the
build does not automatically enable them. This commit changes the docs and
examples for user C modules to encourage writers of user C modules to
enable them unconditionally. This makes things simpler and covers most use
cases.
See discussion in issue #6960, and also #7086.
Signed-off-by: Damien George <damien@micropython.org>
Instead of raising a ZeroDivisionError, this tool now just skips any
elements in the DFU file that have zero size.
Signed-off-by: Damien George <damien@micropython.org>
Updates the zephyr docker image to the latest, v0.11.13. This updates CI
to use zephyr SDK v0.12.2 and GCC v10.2.0 for the zephyr port.
Signed-off-by: Maureen Helm <maureen.helm@nxp.com>
Refactors the zephyr build infrastructure to build MicroPython as a
cmake target, using the recently introduced core cmake rules.
This change makes it possible to build the zephyr port like most other
zephyr applications using west or cmake directly. It simplifies building
with extra cmake arguments, such as specifying an alternate conf file or
adding an Arduino shield. It also enables building the zephyr port
anywhere in the host file system, which will allow regressing across
multiple boards with the zephyr twister script.
Signed-off-by: Maureen Helm <maureen.helm@nxp.com>
This allows customising which features can be enabled in a frozen library.
e.g. `include("path.py", extra_features=True)`
in path.py:
options.defaults(standard_features=True)
if options.standard_features:
# freeze standard modules.
if options.extra_features:
# freeze extra modules.
Signed-off-by: Jim Mussared <jim.mussared@gmail.com>
So that all MicroPython ports that use tinyusb use the same version. Also
requires fewer submodule checkouts when building rp2 along with other ports
that use tinyusb.
Signed-off-by: Damien George <damien@micropython.org>
The upip module is frozen into ports supporting it, and it is included in
the source tree, so there is no need to get it from PyPi. Moreover the
PyPi package referred to is an out-of-date version of upip which is
basically unrelated to our upip.py because the source is taken from a fork
of micropython-lib instead of this repository.
The main rules enforced are:
- At most 72 characters in the subject line, with a ": " in it.
- At most 75 characters per line in the body.
- No "noreply" email addresses.
Because the version included in xtensa-lx106-elf-standalone.tar.gz needs
Python 2 (and pyserial for Python 2).
Signed-off-by: Damien George <damien@micropython.org>
The -Og optimisation level produces a more realistic build, gives a better
debugging experience, and generates smaller code than -O0, allowing debug
builds to fit in flash.
This commit also assigns variables in can.c to prevent warnings when -Og is
used, and builds a board in CI with DEBUG=1 enabled.
Signed-off-by: Damien George <damien@micropython.org>
This commit adds support to pyboard.py for the new raw REPL paste mode.
Note that this new pyboard.py is fully backwards compatible with old
devices (it detects if the device supports the new raw REPL paste mode).
Signed-off-by: Damien George <damien@micropython.org>
The msvc compiler doesn't accept zero-sized arrays so let the freezing
process generate compatible C code in case no modules are found and the
involved arrays are all empty. This doesn't affect the functionality in
any way because those arrays only get accessed when mp_frozen_mpy_names
contains names, i.e. when modules are actually found.
MP_BC_CALL_FUNCTION will leave the result on the Python stack, so that
result must be discarded by MP_BC_POP_TOP.
Signed-off-by: Damien George <damien@micropython.org>
Updating to Black v20.8b1 there are two changes that affect the code in
this repository:
- If there is a trailing comma in a list (eg [], () or function call) then
that list is now written out with one line per element. So remove such
trailing commas where the list should stay on one line.
- Spaces at the start of """ doc strings are removed.
Signed-off-by: Damien George <damien@micropython.org>
The existing implementation of mkdir() in this file is not sophisticated
enough to work correctly on all operating systems (eg Mac can raise
EISDIR). Using the standard os.makedirs() function handles all cases
correctly.
Signed-off-by: Damien George <damien@micropython.org>
Prior to this commit, pyboard.py used eval() to "parse" file data received
from the board. Using eval() on received data from a device is dangerous,
because a malicious device may inject arbitrary code execution on the PC
that is doing the operation.
Consider the following scenario:
Eve may write a malicious script to Bob's board in his absence. On return
Bob notices that something is wrong with the board, because it doesn't work
as expected anymore. He wants to read out boot.py (or any other file) to
see what is wrong. What he gets is a remote code execution on his PC.
Proof of concept:
Eve:
$ cat boot.py
_print = print
print = lambda *x, **y: _print("os.system('ls /; echo Pwned!')", end="\r\n\x04")
$ ./pyboard.py -f cp boot.py :
cp boot.py :boot.py
Bob:
$ ./pyboard.py -f cp :boot.py /tmp/foo
cp :boot.py /tmp/foo
bin chroot dev home lib32 media opt root sbin sys usr
boot config etc lib lib64 mnt proc run srv tmp var
Pwned!
There's also the possibility that the device is malfunctioning and sends
random and possibly dangerous data back to the PC, to be eval'd.
Fix this problem by using ast.literal_eval() to parse the received bytes,
instead of eval().
Signed-off-by: Michael Buesch <m@bues.ch>
If mpy-cross exits with an error be sure to print that error in a way that
is readable, instead of a long bytes object.
Signed-off-by: Damien George <damien@micropython.org>
The file `mbedtls_errors/mp_mbedtls_errors.c` can be used instead of
`mbedtls/library/error.c` to give shorter error strings, reducing the build
size of the error strings from about 12-16kB down to about 2-5kB.
This reverts commit 4d6f60d428.
This implementation used the timeout as a maximum amount of time needed for
the operation, when actually the spec and other tools suggest that it's the
minumum delay needed between subsequent USB transfers.
This adds support for freezing an entire directory while keeping the
directory as part of the import path. For example
freeze("path/to/library", "module")
will recursively freeze all scripts in "path/to/library/module" and have
them importable as "from module import ...".
Signed-off-by: Damien George <damien@micropython.org>
With only `sp_func_proto_paren = remove` set there are some cases where
uncrustify misses removing a space between the function name and the
opening '('. This sets all of the related options to `force` as well.
This is the result of running...
uncrustify -c tools/uncrustify.cfg --update-config-with-doc -o tools/uncrustify.cfg
...with some manual fixups to correct places where it changed things it
should not have.
Essentially it just adds new config parameters introduced in v0.71.0
with their default values.
Signed-off-by: David Lechner <david@pybricks.com>
Formatting for `* sizeof` was fixed in uncrustify v0.71, so we no longer
need the fixups for it. Also, there was one file where the updated
uncrustify caught a problem that the regex didn't pick up, which is updated
in this commit.
Signed-off-by: David Lechner <david@pybricks.com>
This adds a new command line option `-v` to `tools/codeformat.py` to enable
verbose printing of all files that are scanned.
Normally `uncrustify` and `black` are called with the `-q` option so
setting verbose suppresses the `-q` option and on `black` also enables the
`-v` option which makes it print out all file names matching the filter
similar to how `uncrustify` does by default.
This suppresses the Parsing: <file> as language C lines. This makes
parsing run a bit faster and on CI it makes for less scrolling through logs
(and black already uses the -q option).
Note: the uncrustify configuration is explicitly set to 'add' instead of
'force' in order not to alter the comments which use extra spaces after //
as a means of indenting text for clarity.
The decompression of error-strings is only done if the string is accessed
via printing or via er.args. Tests are added for this feature to ensure
the decompression works.
This adds the Python files in the tests/ directory to be formatted with
./tools/codeformat.py. The basics/ subdirectory is excluded for now so we
aren't changing too much at once.
In a few places `# fmt: off`/`# fmt: on` was used where the code had
special formatting for readability or where the test was actually testing
the specific formatting.
This eliminates the need for the sizeof regex fixup by rearranging things a
bit. All other bitfields already use the parentheses around expressions
with sizeof, so one case is fixed by following this convention.
VM_MAX_STATE_ON_STACK is the only remaining problem and it can be worked
around by changing the order of the operands.
When using a manifest on Windows the reference to mpy-cross compiler was
missing the .exe file extension, so add it when appropriate.
Also allow the default path to mpy-cross to be overridden by the (optional)
MICROPY_MPYCROSS environment variable, to allow full flexibility on any OS.
This commit adds a tool, codeformat.py, which will reformat C and Python
code to fit a certain style. By default the tool will reformat (almost)
all the original (ie not 3rd-party) .c, .h and .py files in this
repository. Passing filenames on the command-line to codeformat.py will
reformat only those. Reformatting is done in-place.
uncrustify is used for C reformatting, which is available for many
platforms and can be easily built from source, see
https://github.com/uncrustify/uncrustify. The configuration for uncrustify
is also added in this commit and values are chosen to best match the
existing code style. A small post-processing stage on .c and .h files is
done by codeformat.py (after running uncrustify) to fix up some minor
items:
- space inserted after * when used as multiplication with sizeof
- #if/ifdef/ifndef/elif/else/endif are dedented by one level when they are
configuring if-blocks and case-blocks.
For Python code, the formatter used is black, which can be pip-installed;
see https://github.com/psf/black. The defaults are used, except for line-
length which is set at 99 characters to match the "about 100" line-length
limit used in C code.
The formatting tools used and their configuration were chosen to strike a
balance between keeping existing style and not changing too many lines of
code, and enforcing a relatively strict style (especially for Python code).
This should help to keep the code consistent across everything, and reduce
cognitive load when writing new code to match the style.
This option makes pyboard.py exit as soon as the script/command is
successfully sent to the device, ie it does not wait for any output. This
can help to avoid hangs if the board is being rebooted with --comman (for
example).
Example usage:
$ python3 ./tools/pyboard.py --device /dev/ttyUSB0 --no-follow \
--command 'import machine; machine.reset()'
Some parts of code have been aligned to increase readability. In general
'' instead of "" were used wherever possible to keep the same convention
for entire file. Import inspect line has been moved to the top according
to hints reported by pep8 tools. A few extra spaces were removed, a few
missing spaces were added. Comments have been updated, mostly in
"read_dfu_file" function. Some other comments have been capitalized and/or
slightly updated. A few docstrings were fixed as well. No real code
changes intended.
This is an alternative to f4ed2df that adds a newline so that the output of
the test starts on a new line and the result of the test is prefixed with
"result: " to distinguish it from the test output.
Suggested-by: @dpgeorge
This patch allows executing .mpy files (including native ones) directly on
a target, eg a board over a serial connection. So there's no need to copy
the file to its filesystem to test it.
For example:
$ mpy-cross foo.py
$ pyboard.py foo.mpy
This makes the loading of viper-code-with-relocations a bit neater and
easier to understand, by treating the rodata/bss like a special object to
be loaded into the constant table (which is how it behaves).
We don't want to add a feature flag to .mpy files that indicate float
support because it will get complex and difficult to use. Instead the .mpy
is built using whatever precision it chooses (float or double) and the
native glue API will convert between this choice and what the host runtime
actually uses.
This commit adds a new tool called mpy_ld.py which is essentially a linker
that builds .mpy files directly from .o files. A new header file
(dynruntime.h) and makefile fragment (dynruntime.mk) are also included
which allow building .mpy files from C source code. Such .mpy files can
then be dynamically imported as though they were a normal Python module,
even though they are implemented in C.
Converting .o files directly (rather than pre-linked .elf files) allows the
resulting .mpy to be more efficient because it has more control over the
relocations; for example it can skip PLT indirection. Doing it this way
also allows supporting more architectures, such as Xtensa which has
specific needs for position-independent code and the GOT.
The tool supports targets of x86, x86-64, ARM Thumb and Xtensa (windowed
and non-windowed). BSS, text and rodata sections are supported, with
relocations to all internal sections and symbols, as well as relocations to
some external symbols (defined by dynruntime.h), and linking of qstrs.
Usage:
mpy-tool.py -o merged.mpy --merge mod1.mpy mod2.mpy
The constituent .mpy files are executed sequentially when the merged file
is imported, and they all use the same global namespace.
While the new manifest.py style got introduced for freezing python code
into the resulting binary, the old way - where files and modules within
ports/*/modules where baked into the resulting binary - was still
supported via `freeze('$(PORT_DIR)/modules')` within manifest.py.
However behaviour changed for symlinked directories (=modules), as those
links weren't followed anymore.
This commit restores the original behaviour by explicitly following
symlinks within a modules/ directory
When loading a manifest file, e.g. by include(), it will chdir first to the
directory of that manifest. This means that all file operations within a
manifest are relative to that manifest's location.
As a consequence of this, additional environment variables are needed to
find absolute paths, so the following are added: $(MPY_LIB_DIR),
$(PORT_DIR), $(BOARD_DIR). And rename $(MPY) to $(MPY_DIR) to be
consistent.
Existing manifests are updated to match.
This introduces a new build variable FROZEN_MANIFEST which can be set to a
manifest listing (written in Python) that describes the set of files to be
frozen in to the firmware.
Instead of encoding 4 zero bytes as placeholders for the simple_name and
source_file qstrs, and storing the qstrs after the bytecode, store the
qstrs at the location of these 4 bytes. This saves 4 bytes per bytecode
function stored in a .mpy file (for example lcd160cr.mpy drops by 232
bytes, 4x 58 functions). And resulting code size is slightly reduced on
ports that use this feature.
This patch compresses the second part of the bytecode prelude which
contains the source file name, function name, source-line-number mapping
and cell closure information. This part of the prelude now begins with a
single varible length unsigned integer which encodes 2 numbers, being the
byte-size of the following 2 sections in the header: the "source info
section" and the "closure section". After decoding this variable unsigned
integer it's possible to skip over one or both of these sections very
easily.
This scheme saves about 2 bytes for most functions compared to the original
format: one in the case that there are no closure cells, and one because
padding was eliminated.
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).
Prior to this patch mp_opcode_format would calculate the incorrect size of
the MP_BC_UNWIND_JUMP opcode, missing the additional byte. But, because
opcodes below 0x10 are unused and treated as bytes in the .mpy load/save
and freezing code, this bug did not show any symptoms, since nested unwind
jumps would rarely (if ever) reach a depth of 16 (so the extra byte of this
opcode would be between 0x01 and 0x0f and be correctly loaded/saved/frozen
simply as an undefined opcode).
This patch fixes this bug by correctly accounting for the additional byte.
.
- Split 'qemu-arm' from 'unix' for generating tests.
- Add frozen module to the qemu-arm test build.
- Add test that reproduces the requirement to half-word align native
function data.
Use "-f" to select filesystem mode, followed by the command to execute.
Optionally put ":" at the start of a filename to indicate that it's on the
remote device, if it would otherwise be ambiguous.
Examples:
$ pyboard.py -f ls
$ pyboard.py -f cat main.py
$ pyboard.py -f cp :main.py . # get from device
$ pyboard.py -f cp main.py : # put to device
$ pyboard.py -f rm main.py
Previously, when linking qstr objects in native code for ARM Thumb, the
index into the machine code was being incremented by 4, not 8. It should
be 8 to account for the size of the two machine instructions movw and movt.
This patch makes sure the index into the machine code is incremented by the
correct amount for all variations of qstr linking.
See issue #4829.
Fixes errors in the tool when 1) linking qstrs in native ARM-M code; 2)
freezing multiple files some of which use native code and some which don't.
Fixes issue #4829.
The user can now select their own package index by either passing the "-i"
command line option, or setting the upip.index_urls variable (before doing
an install).
The https://micropython.org/pi package index hosts packages from
micropython-lib and will be searched first when installing a package. If a
package is not found here then it will fallback to PyPI.
Prior to this patch, when a lot of data was output by a running script
pyboard.py would try to capture all of this output into the "data"
variable, which would gradually slow down pyboard.py to the point where it
would have large CPU and memory usage (on the host) and potentially lose
data.
This patch fixes this problem by not accumulating the data in the case that
the data is not needed, which is when "data_consumer" is used.
The qstr window size is not log-2 encoded, it's just the actual number (but
in mpy-tool.py this didn't lead to an error because the size is just used
to truncate the window so it doesn't grow arbitrarily large in memory).
Addresses issue #4635.
This system makes it a lot easier to include external libraries as static,
native modules in MicroPython. Simply pass USER_C_MODULES (like
FROZEN_MPY_DIR) as a make parameter.
When encoded in the mpy file, if qstr <= QSTR_LAST_STATIC then store two
bytes: 0, static_qstr_id. Otherwise encode the qstr as usual (either with
string data or a reference into the qstr window).
Reduces mpy file size by about 5%.
Instead of emitting two bytes in the bytecode for where the linked qstr
should be written to, it is now replaced by the actual qstr data, or a
reference into the qstr window.
Reduces mpy file size by about 10%.
This is an implementation of a sliding qstr window used to reduce the
number of qstrs stored in a .mpy file. The window size is configured to 32
entries which takes a fixed 64 bytes (16-bits each) on the C stack when
loading/saving a .mpy file. It allows to remember the most recent 32 qstrs
so they don't need to be stored again in the .mpy file. The qstr window
uses a simple least-recently-used mechanism to discard the least recently
used qstr when the window overflows (similar to dictionary compression).
This scheme only needs a single pass to save/load the .mpy file.
Reduces mpy file size by about 25% with a window size of 32.
POP_BLOCK and POP_EXCEPT are now the same, and are always followed by a
JUMP. So this optimisation reduces code size, and RAM usage of bytecode by
two bytes for each try-except handler.
Under python3 (tested with 3.6.7) bytes with a list of integers as an
argument returns a different result than under python 2.7 (tested with
2.7.15rc1) which causes pydfu.py to fail when run under 2.7. Changing
bytes to bytearray makes pydfu work properly under both Python 2.7 and
Python 3.6.
If you happen to only have a really simple frozen file that doesn't contain
any new qstrs then the generated frozen_mpy.c file contains an empty
enumeration which causes a C compile time error.
Following an equivalent fix to py/bc.c. The reason the incorrect values
for the opcode constants were not previously causing a bug is because they
were never being used: these opcodes always have qstr arguments so the part
of the code that was comparing them would never be reached.
Thanks to @malinah for finding the problem and providing the initial patch.
A DFU device must be in the idle state before it can be programmed, and
this requires either clearing the status or aborting, depending on its
current state. Code is added to do this. And the USB transfer size is now
automatically detected so devices with a size less than 2048 bytes work
correctly.
Some Python linters don't like unconditional except clauses because they
catch SystemExit and KeyboardInterrupt, which usually is not the intended
behaviour.
There appears to be an issue on Windows with CPython >= 3.6,
sys.stdout.flush() raises an exception:
OSError: [WinError 87] The parameter is incorrect
It works fine to just catch and ignore the error on the flush line. Tested
on Windows 10 x64 1803 (Build 17134.228), Python 3.6.4 amd64.
The Python documentation recommends to pass the command as a string when
using Popen(..., shell=True). This is because "sh -c <string>" is used to
execute the command and additional arguments after the command string are
passed to the shell itself (not the executing command).
https://docs.python.org/3.5/library/subprocess.html#subprocess.Popen
The first dynamic qstr pool is double the size of the 'alloc' field of
the last const qstr pool. The built in const qstr pool
(mp_qstr_const_pool) has a hardcoded alloc size of 10, meaning that the
first dynamic pool is allocated space for 20 entries. The alloc size
must be less than or equal to the actual number of qstrs in the pool
(the 'len' field) to ensure that the first dynamically created qstr
triggers the creation of a new pool.
When modules are frozen a second const pool is created (generally
mp_qstr_frozen_const_pool) and linked to the built in pool. However,
this second const pool had its 'alloc' field set to the number of qstrs
in the pool. When freezing a large quantity of modules this can result
in thousands of qstrs being in the pool. This means that the first
dynamically created qstr results in a massive allocation. This commit
sets the alloc size of the frozen qstr pool to 10 or less (if the number
of qstrs in the pool is less than 10). The result of this is that the
allocation behaviour when a dynamic qstr is created is identical with an
without frozen code.
Note that there is the potential for a slight memory inefficiency if the
frozen modules have less than 10 qstrs, as the first few dynamic
allocations will have quite a large overhead, but the geometric growth
soon deals with this.
The ST DFU bootloader supports a transfer size up to 2048 bytes, so send
that much data on each download (to device) packet. This almost halves
total download time.
Instead of passing thru more and more options from tinytest-codegen to
run-tests --list-tests, pipe output of run-tests --list-tests into
tinytest-codegen.
Gets passed to run-tests --list-tests to get actual list of tests to use.
If --target= is not given, legacy set hardcoded in tinytest-codegen itself
is used.
Also, get rid of tinytest test groups - they aren't really used for
anything, and only complicate processing. Besides, one of the next
step is to limit number of tests per a generated file to control
the binary size, which also will require "flat" list of tests.
The way tinytest was used in qemu-arm test target is that it didn't test
much. MicroPython tests are based on matching the test output against
reference output, but qemu-arm's implementation didn't do that, it
effectively tested just that there was no exception during test
execution. "upytesthelper" wrapper was introduce to fix it, and so
test generator is now switched to generate test code for it.
Also, fix PEP8 and other codestyle issues.
This patch allows the following code to run without allocating on the heap:
super().foo(...)
Before this patch such a call would allocate a super object on the heap and
then load the foo method and call it right away. The super object is only
needed to perform the lookup of the method and not needed after that. This
patch makes an optimisation to allocate the super object on the C stack and
discard it right after use.
Changes in code size due to this patch are:
bare-arm: +128
minimal: +232
unix x64: +416
unix nanbox: +364
stmhal: +184
esp8266: +340
cc3200: +128
This allows to execute a command and communicate with its stdin/stdout
via pipes ("exec") or with command-created pseudo-terminal ("execpty"),
to emulate serial access. Immediate usecase is controlling a QEMU process
which emulates board's serial via normal console, but it could be used
e.g. with helper binaries to access real board over other hadware
protocols, etc.
An example of device specification for these cases is:
--device exec:../zephyr/qemu.sh
--device execpty:../zephyr/qemu2.sh
Where qemu.sh contains long-long qemu startup line, or calls another
command. There's a special support in this patch for running the command
in a new terminal session, to support shell wrappers like that (without
new terminal session, only wrapper script would be terminated, but its
child processes would continue to run).
This patch introduces the a small framework to track differences between
uPy and CPython. The framework consists of:
- A set of "tests" which test for an individual feature that differs between
uPy and CPy. Each test is like a normal uPy test in the test suite, but
has a special comment at the start with some meta-data: a category (eg
syntax, core language), a human-readable description of the difference, a
cause, and a workaround. Following the meta-data there is a short code
snippet which demonstrates the difference. See tests/cpydiff directory
for the initial set of tests.
- A program (this patch) which runs all the tests (on uPy and CPy) and
generates nicely-formated .rst documenting the differences.
- Integration into the docs build so that everything is automatic, and the
differences appear in a way that is easy for users to read/reference (see
latter commits).
The idea with using this new framework is:
- When a new difference is found it's easy to write a short test for it,
along with a description, and add it to the existing ones. It's also easy
for contributors to submit tests for differences they find.
- When something is no longer different the tool will give an error and
difference can be removed (or promoted to a proper feature test).
Previous to this patch the qemu-arm tests were compiled with is_relp=true
meaning that the __repl_print__ function was called for all lines of code
in the outer scope. This is not the right behaviour for scripts that are
executed as though they were a file (eg tests).
With this fix the micropython/heapalloc_str.py test now works so it is
removed from the test blacklist.
With caching of map lookups in the bytecode, frozen bytecode can still
work but must be stored in RAM, not ROM. This patch allows mpy-tool.py to
generate code that works with this optimisation, but it's not recommended
to use it on embedded targets (because of lack of RAM).
Previous to this patch pyboard.py would open a new serial connection to
the target for each script that was run, and for any command that was run.
Apart from being inefficient, this meant that the board was soft-reset
between scripts/commands, which precludes scripts from accessing variables
set in a previous one.
This patch changes the behaviour of pyboard.py so that the connection to
the target is created only once, and it's not reset between scripts or any
command that is sent with the -c option.
To make its inclusion as frozen modules in multiple ports less magic.
Ports are just expected to symlink 2 files into their scripts/modules
subdirs.
Unix port updated to use this and in general follow frozen modules setup
tested and tried on baremetal ports, where there's "scripts" predefined
dir (overridable with FROZEN_DIR make var), and a user just drops Python
files there.
This helps to test floating point code on Cortex-M hardware.
As part of this patch the link-time-optimisation was disabled because it
wasn't compatible with software FP support. In particular, the linker
could not find the __aeabi_f2d, __aeabi_d2f etc functions even though they
were provided by lib/libm/math.c.
Frozen modules are now stored with extensions and with '/' as path
separator. In other words, frozen modules paths stored as they are
in normal filesystem.
When an mpy file is frozen it must know the values of certain
configuration variables. This patch provides an explicit check in the
generated C file that the configuration variables are what they are
supposed to be.
The config variable MICROPY_MODULE_FROZEN is now made of two separate
parts: MICROPY_MODULE_FROZEN_STR and MICROPY_MODULE_FROZEN_MPY. This
allows to have none, either or both of frozen strings and frozen mpy
files (aka frozen bytecode).