micropython/py/bc.c

Ignoring revisions in .git-blame-ignore-revs. Click here to bypass and see the normal blame view.

351 lines
14 KiB
C
Raw Normal View History

/*
* This file is part of the MicroPython project, http://micropython.org/
*
* The MIT License (MIT)
*
* Copyright (c) 2014 Damien P. George
* Copyright (c) 2014 Paul Sokolovsky
*
* Permission is hereby granted, free of charge, to any person obtaining a copy
* of this software and associated documentation files (the "Software"), to deal
* in the Software without restriction, including without limitation the rights
* to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
* copies of the Software, and to permit persons to whom the Software is
* furnished to do so, subject to the following conditions:
*
* The above copyright notice and this permission notice shall be included in
* all copies or substantial portions of the Software.
*
* THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
* IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
* FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
* AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
* LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
* OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN
* THE SOFTWARE.
*/
#include <stdbool.h>
#include <string.h>
#include <assert.h>
#include "py/bc0.h"
#include "py/bc.h"
py: Rework bytecode and .mpy file format to be mostly static data. 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>
2021-10-22 12:22:47 +01:00
#include "py/objfun.h"
#if MICROPY_DEBUG_VERBOSE // print debugging info
#define DEBUG_PRINT (1)
#else // don't print debugging info
#define DEBUG_PRINT (0)
#define DEBUG_printf(...) (void)0
#endif
py: Rework bytecode and .mpy file format to be mostly static data. 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>
2021-10-22 12:22:47 +01:00
void mp_encode_uint(void *env, mp_encode_uint_allocator_t allocator, mp_uint_t val) {
// We store each 7 bits in a separate byte, and that's how many bytes needed
byte buf[MP_ENCODE_UINT_MAX_BYTES];
byte *p = buf + sizeof(buf);
// We encode in little-ending order, but store in big-endian, to help decoding
do {
*--p = val & 0x7f;
val >>= 7;
} while (val != 0);
byte *c = allocator(env, buf + sizeof(buf) - p);
if (c != NULL) {
while (p != buf + sizeof(buf) - 1) {
*c++ = *p++ | 0x80;
}
*c = *p;
}
}
mp_uint_t mp_decode_uint(const byte **ptr) {
mp_uint_t unum = 0;
byte val;
const byte *p = *ptr;
do {
val = *p++;
unum = (unum << 7) | (val & 0x7f);
} while ((val & 0x80) != 0);
*ptr = p;
return unum;
}
// This function is used to help reduce stack usage at the caller, for the case when
// the caller doesn't need to increase the ptr argument. If ptr is a local variable
// and the caller uses mp_decode_uint(&ptr) instead of this function, then the compiler
// must allocate a slot on the stack for ptr, and this slot cannot be reused for
// anything else in the function because the pointer may have been stored in a global
// and reused later in the function.
mp_uint_t mp_decode_uint_value(const byte *ptr) {
return mp_decode_uint(&ptr);
}
// This function is used to help reduce stack usage at the caller, for the case when
// the caller doesn't need the actual value and just wants to skip over it.
const byte *mp_decode_uint_skip(const byte *ptr) {
while ((*ptr++) & 0x80) {
}
return ptr;
}
STATIC NORETURN void fun_pos_args_mismatch(mp_obj_fun_bc_t *f, size_t expected, size_t given) {
#if MICROPY_ERROR_REPORTING <= MICROPY_ERROR_REPORTING_TERSE
// generic message, used also for other argument issues
(void)f;
(void)expected;
(void)given;
mp_arg_error_terse_mismatch();
#elif MICROPY_ERROR_REPORTING == MICROPY_ERROR_REPORTING_NORMAL
(void)f;
mp_raise_msg_varg(&mp_type_TypeError,
MP_ERROR_TEXT("function takes %d positional arguments but %d were given"), expected, given);
#elif MICROPY_ERROR_REPORTING == MICROPY_ERROR_REPORTING_DETAILED
mp_raise_msg_varg(&mp_type_TypeError,
MP_ERROR_TEXT("%q() takes %d positional arguments but %d were given"),
mp_obj_fun_get_name(MP_OBJ_FROM_PTR(f)), expected, given);
#endif
}
#if DEBUG_PRINT
STATIC void dump_args(const mp_obj_t *a, size_t sz) {
DEBUG_printf("%p: ", a);
for (size_t i = 0; i < sz; i++) {
DEBUG_printf("%p ", a[i]);
}
DEBUG_printf("\n");
}
#else
#define dump_args(...) (void)0
#endif
// On entry code_state should be allocated somewhere (stack/heap) and
// contain the following valid entries:
// - code_state->fun_bc should contain a pointer to the function object
py: Rework bytecode and .mpy file format to be mostly static data. 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>
2021-10-22 12:22:47 +01:00
// - code_state->ip should contain a pointer to the beginning of the prelude
// - code_state->n_state should be the number of objects in the local state
void mp_setup_code_state(mp_code_state_t *code_state, size_t n_args, size_t n_kw, const mp_obj_t *args) {
// This function is pretty complicated. It's main aim is to be efficient in speed and RAM
// usage for the common case of positional only args.
// get the function object that we want to set up (could be bytecode or native code)
mp_obj_fun_bc_t *self = code_state->fun_bc;
#if MICROPY_STACKLESS
code_state->prev = NULL;
#endif
#if MICROPY_PY_SYS_SETTRACE
code_state->prev_state = NULL;
code_state->frame = NULL;
#endif
// Get cached n_state (rather than decode it again)
size_t n_state = code_state->n_state;
// Decode prelude
size_t n_state_unused, n_exc_stack_unused, scope_flags, n_pos_args, n_kwonly_args, n_def_pos_args;
MP_BC_PRELUDE_SIG_DECODE_INTO(code_state->ip, n_state_unused, n_exc_stack_unused, scope_flags, n_pos_args, n_kwonly_args, n_def_pos_args);
py: Rework bytecode and .mpy file format to be mostly static data. 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>
2021-10-22 12:22:47 +01:00
MP_BC_PRELUDE_SIZE_DECODE(code_state->ip);
(void)n_state_unused;
(void)n_exc_stack_unused;
code_state->sp = &code_state->state[0] - 1;
code_state->exc_sp_idx = 0;
// zero out the local stack to begin with
memset(code_state->state, 0, n_state * sizeof(*code_state->state));
const mp_obj_t *kwargs = args + n_args;
// var_pos_kw_args points to the stack where the var-args tuple, and var-kw dict, should go (if they are needed)
mp_obj_t *var_pos_kw_args = &code_state->state[n_state - 1 - n_pos_args - n_kwonly_args];
// check positional arguments
if (n_args > n_pos_args) {
// given more than enough arguments
if ((scope_flags & MP_SCOPE_FLAG_VARARGS) == 0) {
fun_pos_args_mismatch(self, n_pos_args, n_args);
}
// put extra arguments in varargs tuple
*var_pos_kw_args-- = mp_obj_new_tuple(n_args - n_pos_args, args + n_pos_args);
n_args = n_pos_args;
} else {
if ((scope_flags & MP_SCOPE_FLAG_VARARGS) != 0) {
DEBUG_printf("passing empty tuple as *args\n");
*var_pos_kw_args-- = mp_const_empty_tuple;
}
// Apply processing and check below only if we don't have kwargs,
// otherwise, kw handling code below has own extensive checks.
if (n_kw == 0 && (scope_flags & MP_SCOPE_FLAG_DEFKWARGS) == 0) {
if (n_args >= (size_t)(n_pos_args - n_def_pos_args)) {
// given enough arguments, but may need to use some default arguments
for (size_t i = n_args; i < n_pos_args; i++) {
code_state->state[n_state - 1 - i] = self->extra_args[i - (n_pos_args - n_def_pos_args)];
}
} else {
fun_pos_args_mismatch(self, n_pos_args - n_def_pos_args, n_args);
}
}
}
// copy positional args into state
for (size_t i = 0; i < n_args; i++) {
code_state->state[n_state - 1 - i] = args[i];
}
// check keyword arguments
if (n_kw != 0 || (scope_flags & MP_SCOPE_FLAG_DEFKWARGS) != 0) {
DEBUG_printf("Initial args: ");
dump_args(code_state->state + n_state - n_pos_args - n_kwonly_args, n_pos_args + n_kwonly_args);
mp_obj_t dict = MP_OBJ_NULL;
if ((scope_flags & MP_SCOPE_FLAG_VARKEYWORDS) != 0) {
dict = mp_obj_new_dict(n_kw); // TODO: better go conservative with 0?
*var_pos_kw_args = dict;
}
for (size_t i = 0; i < n_kw; i++) {
// the keys in kwargs are expected to be qstr objects
mp_obj_t wanted_arg_name = kwargs[2 * i];
py: Rework bytecode and .mpy file format to be mostly static data. 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>
2021-10-22 12:22:47 +01:00
// get pointer to arg_names array
const uint8_t *arg_names = code_state->ip;
arg_names = mp_decode_uint_skip(arg_names);
for (size_t j = 0; j < n_pos_args + n_kwonly_args; j++) {
py: Rework bytecode and .mpy file format to be mostly static data. 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>
2021-10-22 12:22:47 +01:00
qstr arg_qstr = mp_decode_uint(&arg_names);
#if MICROPY_EMIT_BYTECODE_USES_QSTR_TABLE
arg_qstr = self->context->constants.qstr_table[arg_qstr];
#endif
if (wanted_arg_name == MP_OBJ_NEW_QSTR(arg_qstr)) {
if (code_state->state[n_state - 1 - j] != MP_OBJ_NULL) {
mp_raise_msg_varg(&mp_type_TypeError,
MP_ERROR_TEXT("function got multiple values for argument '%q'"), MP_OBJ_QSTR_VALUE(wanted_arg_name));
}
code_state->state[n_state - 1 - j] = kwargs[2 * i + 1];
goto continue2;
}
}
// Didn't find name match with positional args
if ((scope_flags & MP_SCOPE_FLAG_VARKEYWORDS) == 0) {
#if MICROPY_ERROR_REPORTING <= MICROPY_ERROR_REPORTING_TERSE
mp_raise_TypeError(MP_ERROR_TEXT("unexpected keyword argument"));
#else
mp_raise_msg_varg(&mp_type_TypeError,
MP_ERROR_TEXT("unexpected keyword argument '%q'"), MP_OBJ_QSTR_VALUE(wanted_arg_name));
#endif
}
mp_obj_dict_store(dict, kwargs[2 * i], kwargs[2 * i + 1]);
continue2:;
}
DEBUG_printf("Args with kws flattened: ");
dump_args(code_state->state + n_state - n_pos_args - n_kwonly_args, n_pos_args + n_kwonly_args);
// fill in defaults for positional args
mp_obj_t *d = &code_state->state[n_state - n_pos_args];
mp_obj_t *s = &self->extra_args[n_def_pos_args - 1];
for (size_t i = n_def_pos_args; i > 0; i--, d++, s--) {
if (*d == MP_OBJ_NULL) {
*d = *s;
}
}
DEBUG_printf("Args after filling default positional: ");
dump_args(code_state->state + n_state - n_pos_args - n_kwonly_args, n_pos_args + n_kwonly_args);
// Check that all mandatory positional args are specified
while (d < &code_state->state[n_state]) {
if (*d++ == MP_OBJ_NULL) {
mp_raise_msg_varg(&mp_type_TypeError,
MP_ERROR_TEXT("function missing required positional argument #%d"), &code_state->state[n_state] - d);
}
}
// Check that all mandatory keyword args are specified
// Fill in default kw args if we have them
py: Rework bytecode and .mpy file format to be mostly static data. 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>
2021-10-22 12:22:47 +01:00
const uint8_t *arg_names = mp_decode_uint_skip(code_state->ip);
for (size_t i = 0; i < n_pos_args; i++) {
arg_names = mp_decode_uint_skip(arg_names);
}
for (size_t i = 0; i < n_kwonly_args; i++) {
py: Rework bytecode and .mpy file format to be mostly static data. 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>
2021-10-22 12:22:47 +01:00
qstr arg_qstr = mp_decode_uint(&arg_names);
#if MICROPY_EMIT_BYTECODE_USES_QSTR_TABLE
arg_qstr = self->context->constants.qstr_table[arg_qstr];
#endif
if (code_state->state[n_state - 1 - n_pos_args - i] == MP_OBJ_NULL) {
mp_map_elem_t *elem = NULL;
if ((scope_flags & MP_SCOPE_FLAG_DEFKWARGS) != 0) {
py: Rework bytecode and .mpy file format to be mostly static data. 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>
2021-10-22 12:22:47 +01:00
elem = mp_map_lookup(&((mp_obj_dict_t *)MP_OBJ_TO_PTR(self->extra_args[n_def_pos_args]))->map, MP_OBJ_NEW_QSTR(arg_qstr), MP_MAP_LOOKUP);
}
if (elem != NULL) {
code_state->state[n_state - 1 - n_pos_args - i] = elem->value;
} else {
mp_raise_msg_varg(&mp_type_TypeError,
py: Rework bytecode and .mpy file format to be mostly static data. 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>
2021-10-22 12:22:47 +01:00
MP_ERROR_TEXT("function missing required keyword argument '%q'"), arg_qstr);
}
}
}
} else {
// no keyword arguments given
if (n_kwonly_args != 0) {
mp_raise_TypeError(MP_ERROR_TEXT("function missing keyword-only argument"));
}
if ((scope_flags & MP_SCOPE_FLAG_VARKEYWORDS) != 0) {
*var_pos_kw_args = mp_obj_new_dict(0);
}
}
py: Rework bytecode and .mpy file format to be mostly static data. 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>
2021-10-22 12:22:47 +01:00
// jump over code info (source file, argument names and line-number mapping)
const uint8_t *ip = code_state->ip + n_info;
// bytecode prelude: initialise closed over variables
for (; n_cell; --n_cell) {
size_t local_num = *ip++;
code_state->state[n_state - 1 - local_num] =
mp_obj_new_cell(code_state->state[n_state - 1 - local_num]);
}
// now that we skipped over the prelude, set the ip for the VM
code_state->ip = ip;
DEBUG_printf("Calling: n_pos_args=%d, n_kwonly_args=%d\n", n_pos_args, n_kwonly_args);
dump_args(code_state->state + n_state - n_pos_args - n_kwonly_args, n_pos_args + n_kwonly_args);
dump_args(code_state->state, n_state);
}
#if MICROPY_PERSISTENT_CODE_LOAD || MICROPY_PERSISTENT_CODE_SAVE
// The following table encodes the number of bytes that a specific opcode
// takes up. Some opcodes have an extra byte, defined by MP_BC_MASK_EXTRA_BYTE.
uint mp_opcode_format(const byte *ip, size_t *opcode_size, bool count_var_uint) {
uint f = MP_BC_FORMAT(*ip);
const byte *ip_start = ip;
if (f == MP_BC_FORMAT_QSTR) {
ip += 3;
} else {
int extra_byte = (*ip & MP_BC_MASK_EXTRA_BYTE) == 0;
ip += 1;
if (f == MP_BC_FORMAT_VAR_UINT) {
if (count_var_uint) {
while ((*ip++ & 0x80) != 0) {
}
}
} else if (f == MP_BC_FORMAT_OFFSET) {
if ((*ip & 0x80) == 0) {
ip += 1;
} else {
ip += 2;
}
}
ip += extra_byte;
}
*opcode_size = ip - ip_start;
return f;
}
#endif // MICROPY_PERSISTENT_CODE_LOAD || MICROPY_PERSISTENT_CODE_SAVE