Background: .mpy files are precompiled .py files, built using mpy-cross,
that contain compiled bytecode functions (and can also contain machine
code). The benefit of using an .mpy file over a .py file is that they are
faster to import and take less memory when importing. They are also
smaller on disk.
But the real benefit of .mpy files comes when they are frozen into the
firmware. This is done by loading the .mpy file during compilation of the
firmware and turning it into a set of big C data structures (the job of
mpy-tool.py), which are then compiled and downloaded into the ROM of a
device. These C data structures can be executed in-place, ie directly from
ROM. This makes importing even faster because there is very little to do,
and also means such frozen modules take up much less RAM (because their
bytecode stays in ROM).
The downside of frozen code is that it requires recompiling and reflashing
the entire firmware. This can be a big barrier to entry, slows down
development time, and makes it harder to do OTA updates of frozen code
(because the whole firmware must be updated).
This commit attempts to solve this problem by providing a solution that
sits between loading .mpy files into RAM and freezing them into the
firmware. The .mpy file format has been reworked so that it consists of
data and bytecode which is mostly static and ready to run in-place. If
these new .mpy files are located in flash/ROM which is memory addressable,
the .mpy file can be executed (mostly) in-place.
With this approach there is still a small amount of unpacking and linking
of the .mpy file that needs to be done when it's imported, but it's still
much better than loading an .mpy from disk into RAM (although not as good
as freezing .mpy files into the firmware).
The main trick to make static .mpy files is to adjust the bytecode so any
qstrs that it references now go through a lookup table to convert from
local qstr number in the module to global qstr number in the firmware.
That means the bytecode does not need linking/rewriting of qstrs when it's
loaded. Instead only a small qstr table needs to be built (and put in RAM)
at import time. This means the bytecode itself is static/constant and can
be used directly if it's in addressable memory. Also the qstr string data
in the .mpy file, and some constant object data, can be used directly.
Note that the qstr table is global to the module (ie not per function).
In more detail, in the VM what used to be (schematically):
qst = DECODE_QSTR_VALUE;
is now (schematically):
idx = DECODE_QSTR_INDEX;
qst = qstr_table[idx];
That allows the bytecode to be fixed at compile time and not need
relinking/rewriting of the qstr values. Only qstr_table needs to be linked
when the .mpy is loaded.
Incidentally, this helps to reduce the size of bytecode because what used
to be 2-byte qstr values in the bytecode are now (mostly) 1-byte indices.
If the module uses the same qstr more than two times then the bytecode is
smaller than before.
The following changes are measured for this commit compared to the
previous (the baseline):
- average 7%-9% reduction in size of .mpy files
- frozen code size is reduced by about 5%-7%
- importing .py files uses about 5% less RAM in total
- importing .mpy files uses about 4% less RAM in total
- importing .py and .mpy files takes about the same time as before
The qstr indirection in the bytecode has only a small impact on VM
performance. For stm32 on PYBv1.0 the performance change of this commit
is:
diff of scores (higher is better)
N=100 M=100 baseline -> this-commit diff diff% (error%)
bm_chaos.py 371.07 -> 357.39 : -13.68 = -3.687% (+/-0.02%)
bm_fannkuch.py 78.72 -> 77.49 : -1.23 = -1.563% (+/-0.01%)
bm_fft.py 2591.73 -> 2539.28 : -52.45 = -2.024% (+/-0.00%)
bm_float.py 6034.93 -> 5908.30 : -126.63 = -2.098% (+/-0.01%)
bm_hexiom.py 48.96 -> 47.93 : -1.03 = -2.104% (+/-0.00%)
bm_nqueens.py 4510.63 -> 4459.94 : -50.69 = -1.124% (+/-0.00%)
bm_pidigits.py 650.28 -> 644.96 : -5.32 = -0.818% (+/-0.23%)
core_import_mpy_multi.py 564.77 -> 581.49 : +16.72 = +2.960% (+/-0.01%)
core_import_mpy_single.py 68.67 -> 67.16 : -1.51 = -2.199% (+/-0.01%)
core_qstr.py 64.16 -> 64.12 : -0.04 = -0.062% (+/-0.00%)
core_yield_from.py 362.58 -> 354.50 : -8.08 = -2.228% (+/-0.00%)
misc_aes.py 429.69 -> 405.59 : -24.10 = -5.609% (+/-0.01%)
misc_mandel.py 3485.13 -> 3416.51 : -68.62 = -1.969% (+/-0.00%)
misc_pystone.py 2496.53 -> 2405.56 : -90.97 = -3.644% (+/-0.01%)
misc_raytrace.py 381.47 -> 374.01 : -7.46 = -1.956% (+/-0.01%)
viper_call0.py 576.73 -> 572.49 : -4.24 = -0.735% (+/-0.04%)
viper_call1a.py 550.37 -> 546.21 : -4.16 = -0.756% (+/-0.09%)
viper_call1b.py 438.23 -> 435.68 : -2.55 = -0.582% (+/-0.06%)
viper_call1c.py 442.84 -> 440.04 : -2.80 = -0.632% (+/-0.08%)
viper_call2a.py 536.31 -> 532.35 : -3.96 = -0.738% (+/-0.06%)
viper_call2b.py 382.34 -> 377.07 : -5.27 = -1.378% (+/-0.03%)
And for unix on x64:
diff of scores (higher is better)
N=2000 M=2000 baseline -> this-commit diff diff% (error%)
bm_chaos.py 13594.20 -> 13073.84 : -520.36 = -3.828% (+/-5.44%)
bm_fannkuch.py 60.63 -> 59.58 : -1.05 = -1.732% (+/-3.01%)
bm_fft.py 112009.15 -> 111603.32 : -405.83 = -0.362% (+/-4.03%)
bm_float.py 246202.55 -> 247923.81 : +1721.26 = +0.699% (+/-2.79%)
bm_hexiom.py 615.65 -> 617.21 : +1.56 = +0.253% (+/-1.64%)
bm_nqueens.py 215807.95 -> 215600.96 : -206.99 = -0.096% (+/-3.52%)
bm_pidigits.py 8246.74 -> 8422.82 : +176.08 = +2.135% (+/-3.64%)
misc_aes.py 16133.00 -> 16452.74 : +319.74 = +1.982% (+/-1.50%)
misc_mandel.py 128146.69 -> 130796.43 : +2649.74 = +2.068% (+/-3.18%)
misc_pystone.py 83811.49 -> 83124.85 : -686.64 = -0.819% (+/-1.03%)
misc_raytrace.py 21688.02 -> 21385.10 : -302.92 = -1.397% (+/-3.20%)
The code size change is (firmware with a lot of frozen code benefits the
most):
bare-arm: +396 +0.697%
minimal x86: +1595 +0.979% [incl +32(data)]
unix x64: +2408 +0.470% [incl +800(data)]
unix nanbox: +1396 +0.309% [incl -96(data)]
stm32: -1256 -0.318% PYBV10
cc3200: +288 +0.157%
esp8266: -260 -0.037% GENERIC
esp32: -216 -0.014% GENERIC[incl -1072(data)]
nrf: +116 +0.067% pca10040
rp2: -664 -0.135% PICO
samd: +844 +0.607% ADAFRUIT_ITSYBITSY_M4_EXPRESS
As part of this change the .mpy file format version is bumped to version 6.
And mpy-tool.py has been improved to provide a good visualisation of the
contents of .mpy files.
In summary: this commit changes the bytecode to use qstr indirection, and
reworks the .mpy file format to be simpler and allow .mpy files to be
executed in-place. Performance is not impacted too much. Eventually it
will be possible to store such .mpy files in a linear, read-only, memory-
mappable filesystem so they can be executed from flash/ROM. This will
essentially be able to replace frozen code for most applications.
Signed-off-by: Damien George <damien@micropython.org>
uint types in viper mode can now be used for all binary operators except
floor-divide and modulo.
Fixes issue #1847 and issue #6177.
Signed-off-by: Damien George <damien@micropython.org>
The loop searches backwards for a target, but doesn't stop after finding
the first result, meaning that it'll always end up at the outermost
exception handler.
This previously made the native emitter incompatible with the bytecode
emitter, and mp_resume (and subsequently mp_obj_generator_resume) expects
the bytecode emitter behavior (i.e. throw==NULL).
In which case place the native function prelude in a bytes object, linked
from the const_table of that function. An architecture should define
N_PRELUDE_AS_BYTES_OBJ to 1 before including py/emitnative.c to emit
correct machine code, then enable MICROPY_EMIT_NATIVE_PRELUDE_AS_BYTES_OBJ
so the runtime can correctly handle the prelude being in a bytes object.
Such that args/return regs for the parent are different to args/return regs
for child calls. For an architecture to use this feature it should define
the REG_PARENT_xxx macros before including py/emitnative.c.
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).
mpy-cross uses MICROPY_DYNAMIC_COMPILER and MICROPY_EMIT_NATIVE but does
not actually need to execute native functions, and does not need
mp_fun_table. This commit makes it so mp_fun_table and all its entries are
not built when MICROPY_DYNAMIC_COMPILER is enabled, significantly reducing
the size of the mpy-cross executable and allowing it to be built on more
machines/OS's.
Prior to this commit, building the unix port with `DEBUG=1` and
`-finstrument-functions` the compilation would fail with an error like
"control reaches end of non-void function". This change fixes this by
removing the problematic "if (0)" branches. Not all branches affect
compilation, but they are all removed for consistency.
This commit adds support for saving and loading .mpy files that contain
native code (native, viper and inline-asm). A lot of the ground work was
already done for this in the form of removing pointers from generated
native code. The changes here are mainly to link in qstr values to the
native code, and change the format of .mpy files to contain native code
blocks (possibly mixed with bytecode).
A top-level summary:
- @micropython.native, @micropython.viper and @micropython.asm_thumb/
asm_xtensa are now allowed in .py files when compiling to .mpy, and they
work transparently to the user.
- Entire .py files can be compiled to native via mpy-cross -X emit=native
and for the most part the generated .mpy files should work the same as
their bytecode version.
- The .mpy file format is changed to 1) specify in the header if the file
contains native code and if so the architecture (eg x86, ARMV7M, Xtensa);
2) for each function block the kind of code is specified (bytecode,
native, viper, asm).
- When native code is loaded from a .mpy file the native code must be
modified (in place) to link qstr values in, just like bytecode (see
py/persistentcode.c:arch_link_qstr() function).
In addition, this now defines a public, native ABI for dynamically loadable
native code generated by other languages, like C.
The new compile-time option is MICROPY_DEBUG_MP_OBJ_SENTINELS, disabled by
default. This is to allow finer control of whether this debugging feature
is enabled or not (because, for example, this setting must be the same for
mpy-cross and the MicroPython main code when using native code generation).
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.
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.
The maximum index into mp_fun_table is currently less than 1024 and should
stay that way to keep things efficient for all architectures, so there is
no need to handle loading the pointer directly via a literal in this
function.
All architectures now have a dedicated register to hold the pointer to the
native function table mp_fun_table, and so they all need to load this
register at the start of the native function. This commit makes the
loading of this register uniform across architectures by passing the
pointer in the constant table for the native function, and then loading the
register from the constant table. Doing it this way means that the pointer
is not stored in the assembly code, helping to make the code more portable.
Instead of storing the function pointer directly in the assembly code.
This makes the generated code more independent of the runtime (so easier to
relocate the code), and reduces the generated code size.
Instead of storing the function pointer directly in the assembly code.
This makes the generated code more independent of the runtime (so easier to
relocate the code), and reduces the generated code size.
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.
The nlr_buf_t doesn't need to be part of the Python value stack (as it was
before this commit), it's simpler to have it separated as auxiliary state
that lives on the C stack. This will help adding yield support because in
that case the nlr_buf_t and Python value stack live in separate memory
areas (C stack and heap respectively).
This matches how bytecode does it, and matches the signature of
mp_emit_glue_assign_native. Since the native emitter doesn't support
nan-boxing uintptr_t and mp_uint_t are anyway the same bit-width.
