djsundog/circuitpython
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity
388d419ba3
circuitpython/py/py.mk
Jim Mussared 154b4eb354 py: Implement "common word" compression scheme for error messages. 
The idea here is that there's a moderate amount of ROM used up by exception
text.  Obviously we try to keep the messages short, and the code can enable
terse errors, but it still adds up.  Listed below is the total string data
size for various ports:

    bare-arm 2860
    minimal 2876
    stm32 8926  (PYBV11)
    cc3200 3751
    esp32 5721

This commit implements compression of these strings.  It takes advantage of
the fact that these strings are all 7-bit ascii and extracts the top 128
frequently used words from the messages and stores them packed (dropping
their null-terminator), then uses (0x80 | index) inside strings to refer to
these common words.  Spaces are automatically added around words, saving
more bytes.  This happens transparently in the build process, mirroring the
steps that are used to generate the QSTR data.  The MP_COMPRESSED_ROM_TEXT
macro wraps any literal string that should compressed, and it's
automatically decompressed in mp_decompress_rom_string.

There are many schemes that could be used for the compression, and some are
included in py/makecompresseddata.py for reference (space, Huffman, ngram,
common word).  Results showed that the common-word compression gets better
results.  This is before counting the increased cost of the Huffman
decoder.  This might be slightly counter-intuitive, but this data is
extremely repetitive at a word-level, and the byte-level entropy coder
can't quite exploit that as efficiently.  Ideally one would combine both
approaches, but for now the common-word approach is the one that is used.

For additional comparison, the size of the raw data compressed with gzip
and zlib is calculated, as a sort of proxy for a lower entropy bound.  With
this scheme we come within 15% on stm32, and 30% on bare-arm (i.e. we use
x% more bytes than the data compressed with gzip -- not counting the code
overhead of a decoder, and how this would be hypothetically implemented).

The feature is disabled by default and can be enabled by setting
MICROPY_ROM_TEXT_COMPRESSION at the Makefile-level.
2020-04-05 14:20:57 +10:00

289 lines
7.4 KiB
Makefile
Raw Blame History

# where py object files go (they have a name prefix to prevent filename clashes)
PY_BUILD = $(BUILD)/py
# where autogenerated header files go
HEADER_BUILD = $(BUILD)/genhdr
# file containing qstr defs for the core Python bit
PY_QSTR_DEFS = $(PY_SRC)/qstrdefs.h
# If qstr autogeneration is not disabled we specify the output header
# for all collected qstrings.
ifneq ($(QSTR_AUTOGEN_DISABLE),1)
QSTR_DEFS_COLLECTED = $(HEADER_BUILD)/qstrdefs.collected.h
endif
# Any files listed by these variables will cause a full regeneration of qstrs
# DEPENDENCIES: included in qstr processing; REQUIREMENTS: not included
QSTR_GLOBAL_DEPENDENCIES += $(PY_SRC)/mpconfig.h mpconfigport.h
QSTR_GLOBAL_REQUIREMENTS += $(HEADER_BUILD)/mpversion.h
# some code is performance bottleneck and compiled with other optimization options
CSUPEROPT = -O3
# Enable building 32-bit code on 64-bit host.
ifeq ($(MICROPY_FORCE_32BIT),1)
CC += -m32
CXX += -m32
LD += -m32
endif
# External modules written in C.
ifneq ($(USER_C_MODULES),)
# pre-define USERMOD variables as expanded so that variables are immediate
# expanded as they're added to them
SRC_USERMOD :=
CFLAGS_USERMOD :=
LDFLAGS_USERMOD :=
$(foreach module, $(wildcard $(USER_C_MODULES)/*/micropython.mk), \
$(eval USERMOD_DIR = $(patsubst %/,%,$(dir $(module))))\
$(info Including User C Module from $(USERMOD_DIR))\
$(eval include $(module))\
)
SRC_MOD += $(patsubst $(USER_C_MODULES)/%.c,%.c,$(SRC_USERMOD))
CFLAGS_MOD += $(CFLAGS_USERMOD)
LDFLAGS_MOD += $(LDFLAGS_USERMOD)
endif
# py object files
PY_CORE_O_BASENAME = $(addprefix py/,\
mpstate.o \
nlr.o \
nlrx86.o \
nlrx64.o \
nlrthumb.o \
nlrpowerpc.o \
nlrxtensa.o \
nlrsetjmp.o \
malloc.o \
gc.o \
pystack.o \
qstr.o \
vstr.o \
mpprint.o \
unicode.o \
mpz.o \
reader.o \
lexer.o \
parse.o \
scope.o \
compile.o \
emitcommon.o \
emitbc.o \
asmbase.o \
asmx64.o \
emitnx64.o \
asmx86.o \
emitnx86.o \
asmthumb.o \
emitnthumb.o \
emitinlinethumb.o \
asmarm.o \
emitnarm.o \
asmxtensa.o \
emitnxtensa.o \
emitinlinextensa.o \
emitnxtensawin.o \
formatfloat.o \
parsenumbase.o \
parsenum.o \
emitglue.o \
persistentcode.o \
runtime.o \
runtime_utils.o \
scheduler.o \
nativeglue.o \
pairheap.o \
ringbuf.o \
stackctrl.o \
argcheck.o \
warning.o \
profile.o \
map.o \
obj.o \
objarray.o \
objattrtuple.o \
objbool.o \
objboundmeth.o \
objcell.o \
objclosure.o \
objcomplex.o \
objdeque.o \
objdict.o \
objenumerate.o \
objexcept.o \
objfilter.o \
objfloat.o \
objfun.o \
objgenerator.o \
objgetitemiter.o \
objint.o \
objint_longlong.o \
objint_mpz.o \
objlist.o \
objmap.o \
objmodule.o \
objobject.o \
objpolyiter.o \
objproperty.o \
objnone.o \
objnamedtuple.o \
objrange.o \
objreversed.o \
objset.o \
objsingleton.o \
objslice.o \
objstr.o \
objstrunicode.o \
objstringio.o \
objtuple.o \
objtype.o \
objzip.o \
opmethods.o \
sequence.o \
stream.o \
binary.o \
builtinimport.o \
builtinevex.o \
builtinhelp.o \
modarray.o \
modbuiltins.o \
modcollections.o \
modgc.o \
modio.o \
modmath.o \
modcmath.o \
modmicropython.o \
modstruct.o \
modsys.o \
moduerrno.o \
modthread.o \
vm.o \
bc.o \
showbc.o \
repl.o \
smallint.o \
frozenmod.o \
)
PY_EXTMOD_O_BASENAME = \
extmod/moduasyncio.o \
extmod/moductypes.o \
extmod/modujson.o \
extmod/modure.o \
extmod/moduzlib.o \
extmod/moduheapq.o \
extmod/modutimeq.o \
extmod/moduhashlib.o \
extmod/moducryptolib.o \
extmod/modubinascii.o \
extmod/virtpin.o \
extmod/machine_mem.o \
extmod/machine_pinbase.o \
extmod/machine_signal.o \
extmod/machine_pulse.o \
extmod/machine_i2c.o \
extmod/machine_spi.o \
extmod/modbluetooth.o \
extmod/modussl_axtls.o \
extmod/modussl_mbedtls.o \
extmod/modurandom.o \
extmod/moduselect.o \
extmod/moduwebsocket.o \
extmod/modwebrepl.o \
extmod/modframebuf.o \
extmod/vfs.o \
extmod/vfs_blockdev.o \
extmod/vfs_reader.o \
extmod/vfs_posix.o \
extmod/vfs_posix_file.o \
extmod/vfs_fat.o \
extmod/vfs_fat_diskio.o \
extmod/vfs_fat_file.o \
extmod/vfs_lfs.o \
extmod/utime_mphal.o \
extmod/uos_dupterm.o \
lib/embed/abort_.o \
lib/utils/printf.o \
# prepend the build destination prefix to the py object files
PY_CORE_O = $(addprefix $(BUILD)/, $(PY_CORE_O_BASENAME))
PY_EXTMOD_O = $(addprefix $(BUILD)/, $(PY_EXTMOD_O_BASENAME))
# this is a convenience variable for ports that want core, extmod and frozen code
PY_O = $(PY_CORE_O) $(PY_EXTMOD_O)
# object file for frozen code specified via a manifest
ifneq ($(FROZEN_MANIFEST),)
PY_O += $(BUILD)/$(BUILD)/frozen_content.o
endif
# object file for frozen files
ifneq ($(FROZEN_DIR),)
PY_O += $(BUILD)/$(BUILD)/frozen.o
endif
# object file for frozen bytecode (frozen .mpy files)
ifneq ($(FROZEN_MPY_DIR),)
PY_O += $(BUILD)/$(BUILD)/frozen_mpy.o
endif
# Sources that may contain qstrings
SRC_QSTR_IGNORE = py/nlr%
SRC_QSTR += $(SRC_MOD) $(filter-out $(SRC_QSTR_IGNORE),$(PY_CORE_O_BASENAME:.o=.c)) $(PY_EXTMOD_O_BASENAME:.o=.c)
# Anything that depends on FORCE will be considered out-of-date
FORCE:
.PHONY: FORCE
$(HEADER_BUILD)/mpversion.h: FORCE | $(HEADER_BUILD)
$(Q)$(PYTHON) $(PY_SRC)/makeversionhdr.py $@
# mpconfigport.mk is optional, but changes to it may drastically change
# overall config, so they need to be caught
MPCONFIGPORT_MK = $(wildcard mpconfigport.mk)
# qstr data
# Adding an order only dependency on $(HEADER_BUILD) causes $(HEADER_BUILD) to get
# created before we run the script to generate the .h
# Note: we need to protect the qstr names from the preprocessor, so we wrap
# the lines in "" and then unwrap after the preprocessor is finished.
# See more information about this process in docs/develop/qstr.rst.
$(HEADER_BUILD)/qstrdefs.generated.h: $(PY_QSTR_DEFS) $(QSTR_DEFS) $(QSTR_DEFS_COLLECTED) $(PY_SRC)/makeqstrdata.py mpconfigport.h $(MPCONFIGPORT_MK) $(PY_SRC)/mpconfig.h | $(HEADER_BUILD)
$(ECHO) "GEN $@"
$(Q)$(CAT) $(PY_QSTR_DEFS) $(QSTR_DEFS) $(QSTR_DEFS_COLLECTED) | $(SED) 's/^Q(.*)/"&"/' | $(CPP) $(CFLAGS) - | $(SED) 's/^\"\(Q(.*)\)\"/\1/' > $(HEADER_BUILD)/qstrdefs.preprocessed.h
$(Q)$(PYTHON) $(PY_SRC)/makeqstrdata.py $(HEADER_BUILD)/qstrdefs.preprocessed.h > $@
$(HEADER_BUILD)/compressed.data.h: $(HEADER_BUILD)/compressed.collected
$(ECHO) "GEN $@"
$(Q)$(PYTHON) $(PY_SRC)/makecompresseddata.py $< > $@
# build a list of registered modules for py/objmodule.c.
$(HEADER_BUILD)/moduledefs.h: $(SRC_QSTR) $(QSTR_GLOBAL_DEPENDENCIES) | $(HEADER_BUILD)/mpversion.h
@$(ECHO) "GEN $@"
$(Q)$(PYTHON) $(PY_SRC)/makemoduledefs.py --vpath="., $(TOP), $(USER_C_MODULES)" $(SRC_QSTR) > $@
SRC_QSTR += $(HEADER_BUILD)/moduledefs.h
# Force nlr code to always be compiled with space-saving optimisation so
# that the function preludes are of a minimal and predictable form.
$(PY_BUILD)/nlr%.o: CFLAGS += -Os
# optimising gc for speed; 5ms down to 4ms on pybv2
$(PY_BUILD)/gc.o: CFLAGS += $(CSUPEROPT)
# optimising vm for speed, adds only a small amount to code size but makes a huge difference to speed (20% faster)
$(PY_BUILD)/vm.o: CFLAGS += $(CSUPEROPT)
# Optimizing vm.o for modern deeply pipelined CPUs with branch predictors
# may require disabling tail jump optimization. This will make sure that
# each opcode has its own dispatching jump which will improve branch
# branch predictor efficiency.
# https://marc.info/?l=lua-l&m=129778596120851
# http://hg.python.org/cpython/file/b127046831e2/Python/ceval.c#l828
# http://www.emulators.com/docs/nx25_nostradamus.htm
#-fno-crossjumping
# Include rules for extmod related code
include $(TOP)/extmod/extmod.mk
Reference in New Issue View Git Blame Copy Permalink