circuitpython/py/makeqstrdefs.py

194 lines
5.4 KiB
Python
Raw Normal View History

"""
This script processes the output from the C preprocessor and extracts all
qstr. Each qstr is transformed into a qstr definition of the form 'Q(...)'.
This script works with Python 2.6, 2.7, 3.3 and 3.4.
"""
from __future__ import print_function
import re
import subprocess
import sys
import io
import os
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.
2019-09-26 08:19:29 -04:00
# Extract MP_QSTR_FOO macros.
_MODE_QSTR = "qstr"
# Extract MP_COMPRESSED_ROM_TEXT("") macros. (Which come from MP_ERROR_TEXT)
_MODE_COMPRESS = "compress"
def preprocess():
if any(src in args.dependencies for src in args.changed_sources):
sources = args.sources
elif any(args.changed_sources):
sources = args.changed_sources
else:
sources = args.sources
csources = []
cxxsources = []
for source in sources:
if source.endswith(".cpp"):
cxxsources.append(source)
else:
csources.append(source)
try:
os.makedirs(os.path.dirname(args.output[0]))
except OSError:
pass
with open(args.output[0], "w") as out_file:
if csources:
subprocess.check_call(args.pp + args.cflags + csources, stdout=out_file)
if cxxsources:
subprocess.check_call(args.pp + args.cxxflags + cxxsources, stdout=out_file)
def write_out(fname, output):
if output:
for m, r in [("/", "__"), ("\\", "__"), (":", "@"), ("..", "@@")]:
fname = fname.replace(m, r)
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.
2019-09-26 08:19:29 -04:00
with open(args.output_dir + "/" + fname + "." + args.mode, "w") as f:
f.write("\n".join(output) + "\n")
def process_file(f):
re_line = re.compile(r"#[line]*\s\d+\s\"([^\"]+)\"")
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.
2019-09-26 08:19:29 -04:00
if args.mode == _MODE_QSTR:
re_match = re.compile(r"MP_QSTR_[_a-zA-Z0-9]+")
elif args.mode == _MODE_COMPRESS:
re_match = re.compile(r'MP_COMPRESSED_ROM_TEXT\("([^"]*)"\)')
output = []
last_fname = None
for line in f:
if line.isspace():
continue
# match gcc-like output (# n "file") and msvc-like output (#line n "file")
if line.startswith(("# ", "#line")):
m = re_line.match(line)
assert m is not None
fname = m.group(1)
if os.path.splitext(fname)[1] not in [".c", ".cpp"]:
continue
if fname != last_fname:
write_out(last_fname, output)
output = []
last_fname = fname
continue
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.
2019-09-26 08:19:29 -04:00
for match in re_match.findall(line):
if args.mode == _MODE_QSTR:
name = match.replace("MP_QSTR_", "")
output.append("Q(" + name + ")")
elif args.mode == _MODE_COMPRESS:
output.append(match)
if last_fname:
write_out(last_fname, output)
return ""
def cat_together():
import glob
import hashlib
hasher = hashlib.md5()
all_lines = []
outf = open(args.output_dir + "/out", "wb")
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.
2019-09-26 08:19:29 -04:00
for fname in glob.glob(args.output_dir + "/*." + args.mode):
with open(fname, "rb") as f:
lines = f.readlines()
all_lines += lines
all_lines.sort()
all_lines = b"\n".join(all_lines)
outf.write(all_lines)
outf.close()
hasher.update(all_lines)
new_hash = hasher.hexdigest()
# print(new_hash)
old_hash = None
try:
with open(args.output_file + ".hash") as f:
old_hash = f.read()
except IOError:
pass
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.
2019-09-26 08:19:29 -04:00
mode_full = "QSTR"
if args.mode == _MODE_COMPRESS:
mode_full = "Compressed data"
if old_hash != new_hash:
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.
2019-09-26 08:19:29 -04:00
print(mode_full, "updated")
try:
# rename below might fail if file exists
os.remove(args.output_file)
except:
pass
os.rename(args.output_dir + "/out", args.output_file)
with open(args.output_file + ".hash", "w") as f:
f.write(new_hash)
else:
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.
2019-09-26 08:19:29 -04:00
print(mode_full, "not updated")
if __name__ == "__main__":
if len(sys.argv) < 6:
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.
2019-09-26 08:19:29 -04:00
print("usage: %s command mode input_filename output_dir output_file" % sys.argv[0])
sys.exit(2)
class Args:
pass
args = Args()
args.command = sys.argv[1]
if args.command == "pp":
named_args = {
s: []
for s in [
"pp",
"output",
"cflags",
"cxxflags",
"sources",
"changed_sources",
"dependencies",
]
}
for arg in sys.argv[1:]:
if arg in named_args:
current_tok = arg
else:
named_args[current_tok].append(arg)
if not named_args["pp"] or len(named_args["output"]) != 1:
print("usage: %s %s ..." % (sys.argv[0], " ... ".join(named_args)))
sys.exit(2)
for k, v in named_args.items():
setattr(args, k, v)
preprocess()
sys.exit(0)
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.
2019-09-26 08:19:29 -04:00
args.mode = sys.argv[2]
args.input_filename = sys.argv[3] # Unused for command=cat
args.output_dir = sys.argv[4]
args.output_file = None if len(sys.argv) == 5 else sys.argv[5] # Unused for command=split
if args.mode not in (_MODE_QSTR, _MODE_COMPRESS):
print("error: mode %s unrecognised" % sys.argv[2])
sys.exit(2)
try:
os.makedirs(args.output_dir)
except OSError:
pass
if args.command == "split":
with io.open(args.input_filename, encoding="utf-8") as infile:
process_file(infile)
if args.command == "cat":
cat_together()