diff options
author | Damien George <damien@micropython.org> | 2021-10-22 22:22:47 +1100 |
---|---|---|
committer | Damien George <damien@micropython.org> | 2022-02-24 18:08:43 +1100 |
commit | f2040bfc7ee033e48acef9f289790f3b4e6b74e5 (patch) | |
tree | 53402caf1b0e7321bf772278a94f5a87a9e7bf0d /py/emitbc.c | |
parent | 64bfaae7ab33e628f28ca3b53b10893fb047b48e (diff) |
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>
Diffstat (limited to 'py/emitbc.c')
-rw-r--r-- | py/emitbc.c | 156 |
1 files changed, 31 insertions, 125 deletions
diff --git a/py/emitbc.c b/py/emitbc.c index ca7404603..c04701ca7 100644 --- a/py/emitbc.c +++ b/py/emitbc.c @@ -36,8 +36,7 @@ #if MICROPY_ENABLE_COMPILER -#define BYTES_FOR_INT ((MP_BYTES_PER_OBJ_WORD * 8 + 6) / 7) -#define DUMMY_DATA_SIZE (BYTES_FOR_INT) +#define DUMMY_DATA_SIZE (MP_ENCODE_UINT_MAX_BYTES) struct _emit_t { // Accessed as mp_obj_t, so must be aligned as such, and we rely on the @@ -50,6 +49,7 @@ struct _emit_t { int stack_size; + mp_emit_common_t *emit_common; scope_t *scope; mp_uint_t last_source_line_offset; @@ -66,17 +66,11 @@ struct _emit_t { size_t n_info; size_t n_cell; - - #if MICROPY_PERSISTENT_CODE - uint16_t ct_cur_obj; - uint16_t ct_num_obj; - uint16_t ct_cur_raw_code; - #endif - mp_uint_t *const_table; }; -emit_t *emit_bc_new(void) { +emit_t *emit_bc_new(mp_emit_common_t *emit_common) { emit_t *emit = m_new0(emit_t, 1); + emit->emit_common = emit_common; return emit; } @@ -90,26 +84,9 @@ void emit_bc_free(emit_t *emit) { m_del_obj(emit_t, emit); } -typedef byte *(*emit_allocator_t)(emit_t *emit, int nbytes); - -STATIC void emit_write_uint(emit_t *emit, emit_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[BYTES_FOR_INT]; - 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(emit, buf + sizeof(buf) - p); - while (p != buf + sizeof(buf) - 1) { - *c++ = *p++ | 0x80; - } - *c = *p; -} - // all functions must go through this one to emit code info -STATIC byte *emit_get_cur_to_write_code_info(emit_t *emit, int num_bytes_to_write) { +STATIC uint8_t *emit_get_cur_to_write_code_info(void *emit_in, size_t num_bytes_to_write) { + emit_t *emit = emit_in; if (emit->pass < MP_PASS_EMIT) { emit->code_info_offset += num_bytes_to_write; return emit->dummy_data; @@ -126,14 +103,7 @@ STATIC void emit_write_code_info_byte(emit_t *emit, byte val) { } STATIC void emit_write_code_info_qstr(emit_t *emit, qstr qst) { - #if MICROPY_PERSISTENT_CODE - assert((qst >> 16) == 0); - byte *c = emit_get_cur_to_write_code_info(emit, 2); - c[0] = qst; - c[1] = qst >> 8; - #else - emit_write_uint(emit, emit_get_cur_to_write_code_info, qst); - #endif + mp_encode_uint(emit, emit_get_cur_to_write_code_info, mp_emit_common_use_qstr(emit->emit_common, qst)); } #if MICROPY_ENABLE_SOURCE_LINE @@ -166,7 +136,8 @@ STATIC void emit_write_code_info_bytes_lines(emit_t *emit, mp_uint_t bytes_to_sk #endif // all functions must go through this one to emit byte code -STATIC byte *emit_get_cur_to_write_bytecode(emit_t *emit, int num_bytes_to_write) { +STATIC uint8_t *emit_get_cur_to_write_bytecode(void *emit_in, size_t num_bytes_to_write) { + emit_t *emit = emit_in; if (emit->pass < MP_PASS_EMIT) { emit->bytecode_offset += num_bytes_to_write; return emit->dummy_data; @@ -189,12 +160,12 @@ STATIC void emit_write_bytecode_byte(emit_t *emit, int stack_adj, byte b1) { c[0] = b1; } -// Similar to emit_write_bytecode_uint(), just some extra handling to encode sign +// Similar to mp_encode_uint(), just some extra handling to encode sign STATIC void emit_write_bytecode_byte_int(emit_t *emit, int stack_adj, byte b1, mp_int_t num) { emit_write_bytecode_byte(emit, stack_adj, b1); // We store each 7 bits in a separate byte, and that's how many bytes needed - byte buf[BYTES_FOR_INT]; + 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 { @@ -218,61 +189,25 @@ STATIC void emit_write_bytecode_byte_int(emit_t *emit, int stack_adj, byte b1, m STATIC void emit_write_bytecode_byte_uint(emit_t *emit, int stack_adj, byte b, mp_uint_t val) { emit_write_bytecode_byte(emit, stack_adj, b); - emit_write_uint(emit, emit_get_cur_to_write_bytecode, val); + mp_encode_uint(emit, emit_get_cur_to_write_bytecode, val); } -#if MICROPY_PERSISTENT_CODE -STATIC void emit_write_bytecode_byte_const(emit_t *emit, int stack_adj, byte b, mp_uint_t n, mp_uint_t c) { - if (emit->pass == MP_PASS_EMIT) { - emit->const_table[n] = c; - } +STATIC void emit_write_bytecode_byte_const(emit_t *emit, int stack_adj, byte b, mp_uint_t n) { emit_write_bytecode_byte_uint(emit, stack_adj, b, n); } -#endif STATIC void emit_write_bytecode_byte_qstr(emit_t *emit, int stack_adj, byte b, qstr qst) { - #if MICROPY_PERSISTENT_CODE - assert((qst >> 16) == 0); - mp_emit_bc_adjust_stack_size(emit, stack_adj); - byte *c = emit_get_cur_to_write_bytecode(emit, 3); - c[0] = b; - c[1] = qst; - c[2] = qst >> 8; - #else - emit_write_bytecode_byte_uint(emit, stack_adj, b, qst); - #endif + emit_write_bytecode_byte_uint(emit, stack_adj, b, mp_emit_common_use_qstr(emit->emit_common, qst)); } STATIC void emit_write_bytecode_byte_obj(emit_t *emit, int stack_adj, byte b, mp_obj_t obj) { - #if MICROPY_PERSISTENT_CODE emit_write_bytecode_byte_const(emit, stack_adj, b, - emit->scope->num_pos_args + emit->scope->num_kwonly_args - + emit->ct_cur_obj++, (mp_uint_t)obj); - #else - // aligns the pointer so it is friendly to GC - emit_write_bytecode_byte(emit, stack_adj, b); - emit->bytecode_offset = (size_t)MP_ALIGN(emit->bytecode_offset, sizeof(mp_obj_t)); - mp_obj_t *c = (mp_obj_t *)emit_get_cur_to_write_bytecode(emit, sizeof(mp_obj_t)); - // Verify thar c is already uint-aligned - assert(c == MP_ALIGN(c, sizeof(mp_obj_t))); - *c = obj; - #endif + mp_emit_common_alloc_const_obj(emit->emit_common, obj)); } -STATIC void emit_write_bytecode_byte_raw_code(emit_t *emit, int stack_adj, byte b, mp_raw_code_t *rc) { - #if MICROPY_PERSISTENT_CODE +STATIC void emit_write_bytecode_byte_child(emit_t *emit, int stack_adj, byte b, mp_raw_code_t *rc) { emit_write_bytecode_byte_const(emit, stack_adj, b, - emit->scope->num_pos_args + emit->scope->num_kwonly_args - + emit->ct_num_obj + emit->ct_cur_raw_code++, (mp_uint_t)(uintptr_t)rc); - #else - // aligns the pointer so it is friendly to GC - emit_write_bytecode_byte(emit, stack_adj, b); - emit->bytecode_offset = (size_t)MP_ALIGN(emit->bytecode_offset, sizeof(void *)); - void **c = (void **)emit_get_cur_to_write_bytecode(emit, sizeof(void *)); - // Verify thar c is already uint-aligned - assert(c == MP_ALIGN(c, sizeof(void *))); - *c = rc; - #endif + mp_emit_common_alloc_const_child(emit->emit_common, rc)); #if MICROPY_PY_SYS_SETTRACE rc->line_of_definition = emit->last_source_line; #endif @@ -343,27 +278,19 @@ void mp_emit_bc_start_pass(emit_t *emit, pass_kind_t pass, scope_t *scope) { } // Write number of cells and size of the source code info - if (pass >= MP_PASS_CODE_SIZE) { - MP_BC_PRELUDE_SIZE_ENCODE(emit->n_info, emit->n_cell, emit_write_code_info_byte, emit); + if (emit->pass >= MP_PASS_CODE_SIZE) { + size_t n_info = emit->n_info; + size_t n_cell = emit->n_cell; + MP_BC_PRELUDE_SIZE_ENCODE(n_info, n_cell, emit_write_code_info_byte, emit); } emit->n_info = emit->code_info_offset; - // Write the name and source file of this function. + // Write the name of this function. emit_write_code_info_qstr(emit, scope->simple_name); - emit_write_code_info_qstr(emit, scope->source_file); - - #if MICROPY_PERSISTENT_CODE - emit->ct_cur_obj = 0; - emit->ct_cur_raw_code = 0; - #endif - - if (pass == MP_PASS_EMIT) { - // Write argument names (needed to resolve positional args passed as - // keywords). We store them as full word-sized objects for efficient access - // in mp_setup_code_state this is the start of the prelude and is guaranteed - // to be aligned on a word boundary. + // Write argument names, needed to resolve positional args passed as keywords. + { // For a given argument position (indexed by i) we need to find the // corresponding id_info which is a parameter, as it has the correct // qstr name to use as the argument name. Note that it's not a simple @@ -383,7 +310,7 @@ void mp_emit_bc_start_pass(emit_t *emit, pass_kind_t pass, scope_t *scope) { break; } } - emit->const_table[i] = (mp_uint_t)MP_OBJ_NEW_QSTR(qst); + emit_write_code_info_qstr(emit, qst); } } } @@ -396,8 +323,6 @@ void mp_emit_bc_end_pass(emit_t *emit) { // check stack is back to zero size assert(emit->stack_size == 0); - emit_write_code_info_byte(emit, 0); // end of line number info - // Calculate size of source code info section emit->n_info = emit->code_info_offset - emit->n_info; @@ -412,39 +337,20 @@ void mp_emit_bc_end_pass(emit_t *emit) { } } - #if MICROPY_PERSISTENT_CODE - assert(emit->pass <= MP_PASS_STACK_SIZE || (emit->ct_num_obj == emit->ct_cur_obj)); - emit->ct_num_obj = emit->ct_cur_obj; - #endif - if (emit->pass == MP_PASS_CODE_SIZE) { - #if !MICROPY_PERSISTENT_CODE - // so bytecode is aligned - emit->code_info_offset = (size_t)MP_ALIGN(emit->code_info_offset, sizeof(mp_uint_t)); - #endif - // calculate size of total code-info + bytecode, in bytes emit->code_info_size = emit->code_info_offset; emit->bytecode_size = emit->bytecode_offset; emit->code_base = m_new0(byte, emit->code_info_size + emit->bytecode_size); - #if MICROPY_PERSISTENT_CODE - emit->const_table = m_new0(mp_uint_t, - emit->scope->num_pos_args + emit->scope->num_kwonly_args - + emit->ct_cur_obj + emit->ct_cur_raw_code); - #else - emit->const_table = m_new0(mp_uint_t, - emit->scope->num_pos_args + emit->scope->num_kwonly_args); - #endif - } else if (emit->pass == MP_PASS_EMIT) { mp_emit_glue_assign_bytecode(emit->scope->raw_code, emit->code_base, #if MICROPY_PERSISTENT_CODE_SAVE || MICROPY_DEBUG_PRINTERS emit->code_info_size + emit->bytecode_size, #endif - emit->const_table, + emit->emit_common->children, #if MICROPY_PERSISTENT_CODE_SAVE - emit->ct_cur_obj, emit->ct_cur_raw_code, + emit->emit_common->ct_cur_child, #endif emit->scope->scope_flags); } @@ -783,21 +689,21 @@ void mp_emit_bc_unpack_ex(emit_t *emit, mp_uint_t n_left, mp_uint_t n_right) { void mp_emit_bc_make_function(emit_t *emit, scope_t *scope, mp_uint_t n_pos_defaults, mp_uint_t n_kw_defaults) { if (n_pos_defaults == 0 && n_kw_defaults == 0) { - emit_write_bytecode_byte_raw_code(emit, 1, MP_BC_MAKE_FUNCTION, scope->raw_code); + emit_write_bytecode_byte_child(emit, 1, MP_BC_MAKE_FUNCTION, scope->raw_code); } else { - emit_write_bytecode_byte_raw_code(emit, -1, MP_BC_MAKE_FUNCTION_DEFARGS, scope->raw_code); + emit_write_bytecode_byte_child(emit, -1, MP_BC_MAKE_FUNCTION_DEFARGS, scope->raw_code); } } void mp_emit_bc_make_closure(emit_t *emit, scope_t *scope, mp_uint_t n_closed_over, mp_uint_t n_pos_defaults, mp_uint_t n_kw_defaults) { if (n_pos_defaults == 0 && n_kw_defaults == 0) { int stack_adj = -n_closed_over + 1; - emit_write_bytecode_byte_raw_code(emit, stack_adj, MP_BC_MAKE_CLOSURE, scope->raw_code); + emit_write_bytecode_byte_child(emit, stack_adj, MP_BC_MAKE_CLOSURE, scope->raw_code); emit_write_bytecode_raw_byte(emit, n_closed_over); } else { assert(n_closed_over <= 255); int stack_adj = -2 - (mp_int_t)n_closed_over + 1; - emit_write_bytecode_byte_raw_code(emit, stack_adj, MP_BC_MAKE_CLOSURE_DEFARGS, scope->raw_code); + emit_write_bytecode_byte_child(emit, stack_adj, MP_BC_MAKE_CLOSURE_DEFARGS, scope->raw_code); emit_write_bytecode_raw_byte(emit, n_closed_over); } } |