Finds the costs of functions/files in an mbed project and creates a spreadsheet

find-costs.py

#!/usr/bin/env python

import subprocess
import os
import re
import collections
import itertools
import csv
import argparse

PROJECT = os.path.basename(os.getcwd())
BUILD = 'PROFILE/BUILD'

def rmsuffix(s, ps):
    ps = ps if isinstance(ps, list) else [ps]
    while True:
        ns = s
        for p in ps:
            if ns.endswith(p):
                ns = ns[:-len(p)]
        if ns == s:
            return s
        s = ns

def rmprefix(s, ps):
    ps = ps if isinstance(ps, list) else [ps]
    while True:
        ns = s
        for p in ps:
            if ns.startswith(p):
                ns = ns[len(p):]
        if ns == s:
            return s
        s = ns

def main():
    parser = argparse.ArgumentParser(description='Binary sizes + PC sample combiner')
    parser.add_argument('--samples', help="csv file(s) containing samples", nargs="*", default=[])
    args = parser.parse_args()

    syms = {}

    # create build dir
    try:
        os.makedirs(BUILD)
    except OSError:
        pass
    with open(os.path.join(BUILD, '.mbedignore'), 'w') as f:
        f.write('*\n')

    subprocess.check_call(['mbed', 'compile',
            '--profile', 'release', '--build', BUILD])

    # Find static memory usage
    release_syms = subprocess.check_output(['arm-none-eabi-nm', '-p', '-S',
            os.path.join(BUILD, PROJECT+'_application.elf')])
    for m in re.finditer(
            '^([0-9a-fA-F]+)\s([0-9a-fA-F]+)\s(\w)\s(\w+)$',
            release_syms, re.M):
        off, size, type, name = m.groups()
        size = int(size, 16)
        type = type.upper()
        syms[name] = syms.get(name, {})
        if type in {'T', 'W', 'V'}:
            syms[name]['text'] = size
        elif type in {'D'}:
            syms[name]['data'] = size
        elif type in {'B'}:
            syms[name]['bss'] = size
        else:
            print 'Found unknown symbol type: %s (%s)' % (type, name)

    # Find samples
    sample_files = []
    for i, sample_file in enumerate(args.samples):
        try:
            with open(sample_file) as f:
                sample_file = i # rmsuffix(sample_file, '.csv')
                for _,_,_,name,_,_,_ in csv.reader(f):
                    if name and name in syms:
                        syms[name][('samples', sample_file)] = syms[name].get(('samples', sample_file), 0) + 1
        except OSError:
            pass

        sample_files.append(sample_file)

    # Find function dependencies in disassembly
    disas = subprocess.check_output(['arm-none-eabi-objdump', '-d',
            os.path.join(BUILD, PROJECT+'_application.elf')])
    currentsym = ''
    for line in itertools.chain(disas.splitlines(), ['']):
        m = re.match('^[0-9a-fA-F]+ <(\w+)>:', line)
        if m:
            currentsym = m.group(1)
        else:
            m = re.match('^ .*<(\w+)(\+[x0-9a-fA-F])?>', line)
            if m:
                targetsym = m.group(1)
                if currentsym in syms and targetsym in syms and currentsym != targetsym:
                    syms[currentsym]['children'] = (
                            syms[currentsym].get('children', set()) | {targetsym})
                    syms[targetsym]['parents'] = (
                            syms[targetsym].get('parents', set()) | {currentsym})

    # Find source file for each function
    file_list = subprocess.check_output([
            'find', BUILD, '-name', '*.o']).strip().split('\n')
    file_syms = subprocess.check_output(['arm-none-eabi-nm', '-p', '-S', '-A'] + file_list)
    for m in re.finditer(
            '^([^:]*):([0-9a-fA-F]+)\s([0-9a-fA-F]+)\s(\w)\s(\w+)$',
            file_syms, re.M):
        file, _, _, _, name = m.groups()
        file = rmprefix(file, [BUILD, '/'])
        if name in syms:
            syms[name]['file'] = file

    # Clean up missing values
    for s in syms.itervalues():
        s['file'] = s.get('file', '?')
        s['text'] = s.get('text', 0)
        s['data'] = s.get('data', 0)
        s['bss']  = s.get('bss', 0)
        for sample_file in sample_files:
            s[('samples', sample_file)]  = s.get(('samples', sample_file), 0)
        s['children'] = s.get('children', set())
        s['parents']  = s.get('parents', set())

    # Organize by file, write to csv
    files = {}
    for name, s in syms.iteritems():
        o = s['file']
        files[o] = files.get(o, {})
        files[o]['text'] = files[o].get('text', 0) + s['text']
        files[o]['data'] = files[o].get('data', 0) + s['data']
        files[o]['bss']  = files[o].get('bss', 0) + s['bss']
        for sample_file in sample_files:
            files[o][('samples', sample_file)] = files[o].get(('samples', sample_file), 0) + s[('samples', sample_file)]
        files[o]['children'] = files[o].get('children', set()) | set(
            syms[c].get('file', '?') for c in s['children'])
        files[o]['parents'] = files[o].get('parents', set()) | set(
            syms[c].get('file', '?') for c in s['parents'])

    print 'Writing PROFILE/results-files.csv...'
    with open('PROFILE/results-files.csv', 'w') as f:
        writer = csv.writer(f)
        writer.writerow(['file','text','data','bss'] +
                ['samples (%s)' % sf for sf in sample_files] +
                ['parents', 'children'])
        for o, s in sorted(files.iteritems()):
            if o != '?':
                for d in itertools.izip_longest(*(
                        [[o], [s['text']], [s['data']], [s['bss']]] +
                        [[s[('samples', sf)]] for sf in sample_files] +
                        [(u for u in s['parents']  if u != '?' and u != o)] +
                        [(u for u in s['children'] if u != '?' and u != o)]),
                        fillvalue=''):
                    writer.writerow(d)

    print 'Writing PROFILE/results-functions.csv...'
    with open('PROFILE/results-functions.csv', 'w') as f:
        writer = csv.writer(f)
        writer.writerow(['file','function','text','data','bss'] +
                ['samples (%s)' % sf for sf in sample_files] +
                ['parents', 'children'])
        for o in sorted(files.iterkeys()):
            for fn, s in sorted((fn, s) for fn, s in syms.iteritems()
                    if s['file'] == o):
                for d in itertools.izip_longest(*(
                        [[o], [fn], [s['text']], [s['data']], [s['bss']]] +
                        [[s[('samples', sf)]] for sf in sample_files] +
                        [s['parents'], s['children']]),
                        fillvalue=''):
                    writer.writerow(d)

    # Summary
    total_text    = sum(s['text']    for s in syms.itervalues())
    total_data    = sum(s['data']    for s in syms.itervalues())
    total_bss     = sum(s['bss']     for s in syms.itervalues())
    total_samples = [sum(s[('samples', sf)] for s in syms.itervalues())
            for sf in sample_files]

    print '--- totals ---'
    print ' '.join('%012s' % c for c in ['text', 'data', 'bss'] +
            ['samples (%s)' % sf for sf in sample_files])
    print ' '.join('%012s' % c for c in [total_text, total_data, total_bss] + total_samples)

if __name__ == "__main__":
    import sys
    sys.exit(main())