1
00:00:00,480 --> 00:00:03,480
foreign

2
00:00:08,540 --> 00:00:11,940
there are stickers here and there's been

3
00:00:10,620 --> 00:00:14,099
stickers passed around please feel free

4
00:00:11,940 --> 00:00:16,080
to take them um the micros are mine

5
00:00:14,099 --> 00:00:18,300
please don't take them

6
00:00:16,080 --> 00:00:20,939
um by way of introduction my name is

7
00:00:18,300 --> 00:00:22,500
Matt hi uh I'm an engineer of software

8
00:00:20,939 --> 00:00:24,600
engineer for about 25 years which makes

9
00:00:22,500 --> 00:00:26,640
me feel excessively old

10
00:00:24,600 --> 00:00:28,980
um I I kind of describe myself as a full

11
00:00:26,640 --> 00:00:30,840
stack engineer which uh we'll talk about

12
00:00:28,980 --> 00:00:32,099
that in a minute but uh I've been

13
00:00:30,840 --> 00:00:33,960
exposed to most areas of software

14
00:00:32,099 --> 00:00:35,340
development all the way from low level

15
00:00:33,960 --> 00:00:37,800
assembly stuff all the way to

16
00:00:35,340 --> 00:00:39,180
cloud-based web stuff so but a lot of my

17
00:00:37,800 --> 00:00:42,960
career has been spent

18
00:00:39,180 --> 00:00:44,040
um developing on or close to hardware

19
00:00:42,960 --> 00:00:45,840
um which should kind of make me a bit

20
00:00:44,040 --> 00:00:48,079
jaded because in better development had

21
00:00:45,840 --> 00:00:50,280
sort of been stagnant for a long time

22
00:00:48,079 --> 00:00:51,239
and sort of watch all these wonderful

23
00:00:50,280 --> 00:00:54,660
developments and the rest of the

24
00:00:51,239 --> 00:00:57,000
software Fields uh but yeah sort of saw

25
00:00:54,660 --> 00:00:58,980
the embedded stage just sort of plateau

26
00:00:57,000 --> 00:01:01,800
for a long time until micro python came

27
00:00:58,980 --> 00:01:03,120
along and we'll talk about it today

28
00:01:01,800 --> 00:01:05,400
um I would describe myself as a

29
00:01:03,120 --> 00:01:08,159
micropath and Enthusiast I've

30
00:01:05,400 --> 00:01:10,260
contributed code and documentation uh I

31
00:01:08,159 --> 00:01:12,840
host the Melbourne Michael python Meetup

32
00:01:10,260 --> 00:01:15,960
and I've talked to the previous pycons

33
00:01:12,840 --> 00:01:17,100
about Michael python so uh yeah just

34
00:01:15,960 --> 00:01:18,420
knowing that full stack development I

35
00:01:17,100 --> 00:01:20,100
looked up the definition of what a full

36
00:01:18,420 --> 00:01:21,960
stack developer is it's like

37
00:01:20,100 --> 00:01:25,500
the front end and back end of a website

38
00:01:21,960 --> 00:01:27,420
and I'm like hey kids it's like I'm not

39
00:01:25,500 --> 00:01:29,939
sure what you're talking about

40
00:01:27,420 --> 00:01:31,920
anyway moving along uh can I have some

41
00:01:29,939 --> 00:01:33,720
hands up and I'm gonna start with if

42
00:01:31,920 --> 00:01:36,540
you're human

43
00:01:33,720 --> 00:01:37,680
good audience participation question I

44
00:01:36,540 --> 00:01:39,659
saw most people with their hands up

45
00:01:37,680 --> 00:01:41,280
that's good I suppose uh who knows what

46
00:01:39,659 --> 00:01:44,759
a microcontroller is

47
00:01:41,280 --> 00:01:46,740
also good most people uh how about who's

48
00:01:44,759 --> 00:01:49,320
written software for Market controller

49
00:01:46,740 --> 00:01:51,000
it's quite a lot I guess about

50
00:01:49,320 --> 00:01:53,759
huh I'm

51
00:01:51,000 --> 00:01:56,159
um who knows what micropython is who

52
00:01:53,759 --> 00:01:58,259
went to Damien's talk earlier Damien you

53
00:01:56,159 --> 00:02:00,600
don't get to put your hand up good okay

54
00:01:58,259 --> 00:02:02,220
that means I can sort of briefly go over

55
00:02:00,600 --> 00:02:03,240
these things it's a python conference I

56
00:02:02,220 --> 00:02:05,219
need to know how much of the background

57
00:02:03,240 --> 00:02:06,840
we need to cover so who's seen Avengers

58
00:02:05,219 --> 00:02:08,520
Infinity War

59
00:02:06,840 --> 00:02:10,259
good that's helpful

60
00:02:08,520 --> 00:02:12,180
um right

61
00:02:10,259 --> 00:02:13,800
what's a microcontroller it's just a

62
00:02:12,180 --> 00:02:15,239
small computer right so

63
00:02:13,800 --> 00:02:16,440
microcontrollers is less about

64
00:02:15,239 --> 00:02:19,020
processing power when you compare them

65
00:02:16,440 --> 00:02:22,080
to a PC it's more about precise

66
00:02:19,020 --> 00:02:23,520
dedicated sort of tasks they have a

67
00:02:22,080 --> 00:02:25,440
small amount of resources they're

68
00:02:23,520 --> 00:02:27,060
Limited in what they can do with ram and

69
00:02:25,440 --> 00:02:29,160
Flash because they have very little of

70
00:02:27,060 --> 00:02:31,080
it and they often interact with the real

71
00:02:29,160 --> 00:02:32,940
world so they control Motors pumps

72
00:02:31,080 --> 00:02:33,920
interactive sensors that kind of stuff

73
00:02:32,940 --> 00:02:36,000
so

74
00:02:33,920 --> 00:02:38,819
a key defining factor is that they're

75
00:02:36,000 --> 00:02:40,379
low cost from cents to Dollars typically

76
00:02:38,819 --> 00:02:41,519
uh and they've got very predictable

77
00:02:40,379 --> 00:02:42,780
Behavior because they don't have an

78
00:02:41,519 --> 00:02:45,540
operating system they tend to be just

79
00:02:42,780 --> 00:02:48,060
doing that one task you tell them to do

80
00:02:45,540 --> 00:02:49,860
uh many systems are often comprised of

81
00:02:48,060 --> 00:02:51,660
lots of microcontrollers cars are a good

82
00:02:49,860 --> 00:02:53,700
example you've got dozens or hundreds of

83
00:02:51,660 --> 00:02:55,140
microcontrollers in a single car

84
00:02:53,700 --> 00:02:56,760
and it kind of led me on to think about

85
00:02:55,140 --> 00:02:59,160
the size of the market

86
00:02:56,760 --> 00:03:00,660
um so it's like 30 billion dollar market

87
00:02:59,160 --> 00:03:03,060
and it's expected to get to 50 billion

88
00:03:00,660 --> 00:03:05,640
by 238 which was frankly a lot larger

89
00:03:03,060 --> 00:03:07,140
than I expected and it seems like the

90
00:03:05,640 --> 00:03:08,819
average microcontroller costs about a

91
00:03:07,140 --> 00:03:10,739
dollar because they sell 30 billion of

92
00:03:08,819 --> 00:03:13,379
these things a year which if you compare

93
00:03:10,739 --> 00:03:14,519
it to PC sales is kind of scary

94
00:03:13,379 --> 00:03:15,659
um so there's a lot of these things out

95
00:03:14,519 --> 00:03:18,540
there if you look around you'll see

96
00:03:15,659 --> 00:03:19,500
devices in the room with plenty of them

97
00:03:18,540 --> 00:03:21,480
around

98
00:03:19,500 --> 00:03:23,700
what's Market python if you look at

99
00:03:21,480 --> 00:03:25,560
Reddit it's an Abomination

100
00:03:23,700 --> 00:03:28,140
um quite a funny read actually you have

101
00:03:25,560 --> 00:03:31,260
a look at that it's worth a worth a look

102
00:03:28,140 --> 00:03:34,260
what what it is really is uh

103
00:03:31,260 --> 00:03:34,260
sorry

104
00:03:35,940 --> 00:03:39,000
tough tough

105
00:03:38,280 --> 00:03:41,040
um

106
00:03:39,000 --> 00:03:42,540
what is it really it's it's python three

107
00:03:41,040 --> 00:03:44,099
re-implementation of Python 3 for

108
00:03:42,540 --> 00:03:45,780
microcontrollers so a lot of effort's

109
00:03:44,099 --> 00:03:49,019
been put into making it work on a

110
00:03:45,780 --> 00:03:51,120
constrained device written in C the

111
00:03:49,019 --> 00:03:52,500
Baseline was python34 but a lot of other

112
00:03:51,120 --> 00:03:54,120
features been added since then we still

113
00:03:52,500 --> 00:03:56,580
claim it's python 3-4 but there's

114
00:03:54,120 --> 00:03:59,280
features from all over the shop we've

115
00:03:56,580 --> 00:03:59,940
added as resources allow really

116
00:03:59,280 --> 00:04:01,860
um

117
00:03:59,940 --> 00:04:03,900
yeah so why should we keep updating it

118
00:04:01,860 --> 00:04:06,000
to newer versions of python new features

119
00:04:03,900 --> 00:04:10,700
especially so we can Port libraries more

120
00:04:06,000 --> 00:04:10,700
easily from the C python world yeah

121
00:04:10,799 --> 00:04:14,099
we've got excellent support for the

122
00:04:12,239 --> 00:04:17,699
language it really just feels like using

123
00:04:14,099 --> 00:04:19,040
big python and yes we call it big python

124
00:04:17,699 --> 00:04:21,299
um

125
00:04:19,040 --> 00:04:22,380
and we've got exceptional tests if you

126
00:04:21,299 --> 00:04:23,759
entertainment thought you'll know we've

127
00:04:22,380 --> 00:04:26,759
got exceptional test coverage across the

128
00:04:23,759 --> 00:04:28,440
language above 98 so you can be pretty

129
00:04:26,759 --> 00:04:30,440
confident that language itself is really

130
00:04:28,440 --> 00:04:32,880
robust and stable

131
00:04:30,440 --> 00:04:34,979
if you find a bug in Marketplace we will

132
00:04:32,880 --> 00:04:36,780
give you a sticker yeah

133
00:04:34,979 --> 00:04:39,720
standard library is a little hard harder

134
00:04:36,780 --> 00:04:40,979
it's pretty vast as we all know we've

135
00:04:39,720 --> 00:04:42,720
got a lot of things implemented but not

136
00:04:40,979 --> 00:04:44,100
everything and a lot of the libraries

137
00:04:42,720 --> 00:04:45,900
are written such that they're not so

138
00:04:44,100 --> 00:04:47,580
conducive to work on a microcontroller

139
00:04:45,900 --> 00:04:48,660
so we've got to be careful about what we

140
00:04:47,580 --> 00:04:50,699
put in

141
00:04:48,660 --> 00:04:51,720
it's a similar story for third-party

142
00:04:50,699 --> 00:04:53,460
libraries

143
00:04:51,720 --> 00:04:56,820
really tough to just pick up pandas and

144
00:04:53,460 --> 00:04:58,320
have that run on micro python and some

145
00:04:56,820 --> 00:05:01,500
libraries are built right in the

146
00:04:58,320 --> 00:05:03,600
micropython and others are built in in a

147
00:05:01,500 --> 00:05:05,400
separate repo or micropython lib between

148
00:05:03,600 --> 00:05:06,620
install with a tool called MIP which is

149
00:05:05,400 --> 00:05:09,540
like pip

150
00:05:06,620 --> 00:05:11,639
yeah it's kind of fun

151
00:05:09,540 --> 00:05:13,080
uh in addition to all this stuff we've

152
00:05:11,639 --> 00:05:15,900
got the machine Library which gives you

153
00:05:13,080 --> 00:05:17,040
access to the embedded side of things so

154
00:05:15,900 --> 00:05:19,020
you can Twitter with pins you can

155
00:05:17,040 --> 00:05:22,139
connect to peripherals it's very

156
00:05:19,020 --> 00:05:23,639
pythonic so it gives you a nice familiar

157
00:05:22,139 --> 00:05:25,259
sort of API if you if you're used to the

158
00:05:23,639 --> 00:05:26,759
python world to interact with stuff

159
00:05:25,259 --> 00:05:28,979
that's quite low level and typically

160
00:05:26,759 --> 00:05:31,380
done in uh much more of a Bose and

161
00:05:28,979 --> 00:05:33,600
low-level way

162
00:05:31,380 --> 00:05:36,240
so quick example just setting an LED

163
00:05:33,600 --> 00:05:37,560
turning it on using pwm which is a

164
00:05:36,240 --> 00:05:39,660
technique called pulse width modulation

165
00:05:37,560 --> 00:05:42,300
to flicker an LED on and off at

166
00:05:39,660 --> 00:05:43,500
different rates this is much easier and

167
00:05:42,300 --> 00:05:45,360
if you've done this if you've been in

168
00:05:43,500 --> 00:05:46,919
the Sea World before you'll be aware

169
00:05:45,360 --> 00:05:48,680
that's a lot easier than coding it and

170
00:05:46,919 --> 00:05:51,120
see

171
00:05:48,680 --> 00:05:53,160
why micropython and whenever I hear that

172
00:05:51,120 --> 00:05:55,020
term I think of Avengers Infinity War

173
00:05:53,160 --> 00:05:57,240
there's the throwback right where he

174
00:05:55,020 --> 00:05:58,919
went you know why is Gamora that's what

175
00:05:57,240 --> 00:06:01,500
I hear in my head when people ask me why

176
00:05:58,919 --> 00:06:02,220
do you use micropath you know why

177
00:06:01,500 --> 00:06:04,080
um

178
00:06:02,220 --> 00:06:06,539
so similar reasons to C python it's it's

179
00:06:04,080 --> 00:06:09,720
a readable easy language to learn and to

180
00:06:06,539 --> 00:06:10,740
use it's easy to test

181
00:06:09,720 --> 00:06:12,900
um it's very flexible it's a great

182
00:06:10,740 --> 00:06:14,400
language right but the main better the

183
00:06:12,900 --> 00:06:15,600
main benefit for me at least in the

184
00:06:14,400 --> 00:06:17,039
embedded domain especially when you're

185
00:06:15,600 --> 00:06:19,560
comparing it to CNC plus plus is 30

186
00:06:17,039 --> 00:06:21,180
right 30 for me is an important number

187
00:06:19,560 --> 00:06:22,800
because that's what I think the

188
00:06:21,180 --> 00:06:24,780
development time is reduced by by using

189
00:06:22,800 --> 00:06:26,039
micro python typically

190
00:06:24,780 --> 00:06:29,220
um a lot of caveats around that you can

191
00:06:26,039 --> 00:06:30,840
put around that but um uh I'd also think

192
00:06:29,220 --> 00:06:33,000
we get happier customers but that's

193
00:06:30,840 --> 00:06:34,440
harder to quantify

194
00:06:33,000 --> 00:06:36,479
it does obviously depend on the project

195
00:06:34,440 --> 00:06:38,940
but we've I work for a company uh

196
00:06:36,479 --> 00:06:42,060
planned Innovation that uses micropython

197
00:06:38,940 --> 00:06:43,380
for some um some projects and we've had

198
00:06:42,060 --> 00:06:45,060
handful projects now and they've all

199
00:06:43,380 --> 00:06:47,580
come in roughly through the estimates

200
00:06:45,060 --> 00:06:49,319
that we said and we've often compared

201
00:06:47,580 --> 00:06:52,500
them to the estimates in the CEC Plus

202
00:06:49,319 --> 00:06:53,520
World and 20 to 30 is pretty typical I

203
00:06:52,500 --> 00:06:55,319
also expect that number to actually

204
00:06:53,520 --> 00:06:56,520
improve or at least stabilize towards

205
00:06:55,319 --> 00:06:58,680
the higher end of that range because

206
00:06:56,520 --> 00:07:00,720
micro python is still maturing and the

207
00:06:58,680 --> 00:07:04,319
tooling is getting better and the um

208
00:07:00,720 --> 00:07:06,660
libraries are getting more robust so

209
00:07:04,319 --> 00:07:08,880
um so how do we get that

210
00:07:06,660 --> 00:07:10,680
um it's high level than CNC plus plus

211
00:07:08,880 --> 00:07:12,600
you're all aware of that if you're

212
00:07:10,680 --> 00:07:13,919
familiar with C python but you don't

213
00:07:12,600 --> 00:07:15,360
have to worry as much about memory

214
00:07:13,919 --> 00:07:16,380
management and undefined behavior and

215
00:07:15,360 --> 00:07:17,940
all that kind of stuff that you have to

216
00:07:16,380 --> 00:07:21,000
even see

217
00:07:17,940 --> 00:07:22,979
it's hard to overstate the live reple so

218
00:07:21,000 --> 00:07:24,840
there's a way to get access to a running

219
00:07:22,979 --> 00:07:26,580
device

220
00:07:24,840 --> 00:07:27,960
um and you use the python Rebel on the

221
00:07:26,580 --> 00:07:29,580
device so you can actually communicate

222
00:07:27,960 --> 00:07:31,740
to your peripherals and do a lot of work

223
00:07:29,580 --> 00:07:33,240
on the device itself and that's a huge

224
00:07:31,740 --> 00:07:34,860
Boon to development we can really

225
00:07:33,240 --> 00:07:36,780
explore

226
00:07:34,860 --> 00:07:37,560
um uh the development environment that

227
00:07:36,780 --> 00:07:39,060
way

228
00:07:37,560 --> 00:07:41,039
we've got the Unix support for testing

229
00:07:39,060 --> 00:07:43,199
so we can do a lot of things on x86

230
00:07:41,039 --> 00:07:45,720
which is really helpful I've got really

231
00:07:43,199 --> 00:07:48,539
fast workflow so I need to show you MP

232
00:07:45,720 --> 00:07:50,639
remote which is a tool that we use to

233
00:07:48,539 --> 00:07:53,039
connect to our devices

234
00:07:50,639 --> 00:07:55,440
into our micro python devices you can

235
00:07:53,039 --> 00:07:57,539
see here at this top here there's four

236
00:07:55,440 --> 00:08:00,479
files that that's on my computer at the

237
00:07:57,539 --> 00:08:04,380
top there and then MP remote Mount dot

238
00:08:00,479 --> 00:08:07,380
actually mounts my PC file system at the

239
00:08:04,380 --> 00:08:10,080
at the dot point on the device so it's

240
00:08:07,380 --> 00:08:11,639
just like my device sees the files that

241
00:08:10,080 --> 00:08:13,620
are on the PC

242
00:08:11,639 --> 00:08:16,080
so you can see here if I do an OS Lister

243
00:08:13,620 --> 00:08:18,060
that's running on my microcontroller and

244
00:08:16,080 --> 00:08:19,680
it's just like I'm seeing the files that

245
00:08:18,060 --> 00:08:21,660
um that are on my PC

246
00:08:19,680 --> 00:08:24,360
it's fantastic it's hard to explain how

247
00:08:21,660 --> 00:08:25,919
good this is because it feels a bit like

248
00:08:24,360 --> 00:08:28,139
magic having worked in the embedded

249
00:08:25,919 --> 00:08:30,419
space for some time it means that I can

250
00:08:28,139 --> 00:08:34,080
develop on a PC I can use my regular

251
00:08:30,419 --> 00:08:36,419
tools to do my my coding I can use git

252
00:08:34,080 --> 00:08:38,399
and when I'm ready I can test in the

253
00:08:36,419 --> 00:08:40,919
Unix port and then I can just mount it

254
00:08:38,399 --> 00:08:43,320
and try it on the device it's really low

255
00:08:40,919 --> 00:08:45,380
friction way to get moving

256
00:08:43,320 --> 00:08:45,380
um

257
00:08:45,420 --> 00:08:48,540
practice some of the other benefits

258
00:08:46,800 --> 00:08:50,940
we've seen at least the company is that

259
00:08:48,540 --> 00:08:53,160
the junior developers become superpowers

260
00:08:50,940 --> 00:08:55,920
really quickly like and it's been really

261
00:08:53,160 --> 00:08:58,620
really stark contrast to seeing people

262
00:08:55,920 --> 00:09:00,240
um to seeing Juniors start in the CNC by

263
00:08:58,620 --> 00:09:02,339
Foss world where it takes a long time

264
00:09:00,240 --> 00:09:03,300
for that ramp up to occur

265
00:09:02,339 --> 00:09:05,160
um it happens really quickly with

266
00:09:03,300 --> 00:09:07,560
Michael python um so that's been a

267
00:09:05,160 --> 00:09:09,600
surprise for me I didn't really expect

268
00:09:07,560 --> 00:09:11,339
that it was wonderful

269
00:09:09,600 --> 00:09:13,500
we've found that we're more responsive

270
00:09:11,339 --> 00:09:15,180
to customer requests and we can we can

271
00:09:13,500 --> 00:09:17,100
you know it's an overload a temple we're

272
00:09:15,180 --> 00:09:18,899
certainly a lot more agile and customers

273
00:09:17,100 --> 00:09:20,160
love it they can talk with us and work

274
00:09:18,899 --> 00:09:22,019
with us more

275
00:09:20,160 --> 00:09:24,000
and the other thing that for me at least

276
00:09:22,019 --> 00:09:25,620
was that there's a mental change like I

277
00:09:24,000 --> 00:09:27,360
tend to focus on the problem I'm trying

278
00:09:25,620 --> 00:09:28,560
to solve rather than the Nitty Gritty

279
00:09:27,360 --> 00:09:30,899
details of how to actually get it

280
00:09:28,560 --> 00:09:33,540
happening and how to get it done so for

281
00:09:30,899 --> 00:09:34,920
me that's a big win

282
00:09:33,540 --> 00:09:36,720
yeah we're like 10 minutes in we've got

283
00:09:34,920 --> 00:09:38,040
back around so right

284
00:09:36,720 --> 00:09:39,720
people come with me all the time and say

285
00:09:38,040 --> 00:09:41,820
you can't do that in micro python right

286
00:09:39,720 --> 00:09:43,080
when performance matters you can't do it

287
00:09:41,820 --> 00:09:44,459
market price and if you want to run from

288
00:09:43,080 --> 00:09:45,959
a battery and you can't do a

289
00:09:44,459 --> 00:09:47,940
microphizing if you're running if you

290
00:09:45,959 --> 00:09:48,839
want to worry about manufacturing and

291
00:09:47,940 --> 00:09:51,300
volume

292
00:09:48,839 --> 00:09:52,019
it has a kernel of treason all of these

293
00:09:51,300 --> 00:09:53,880
um

294
00:09:52,019 --> 00:09:56,040
but I think it's worth digging into

295
00:09:53,880 --> 00:09:57,779
right so

296
00:09:56,040 --> 00:10:00,660
the most common concern is Marketplace

297
00:09:57,779 --> 00:10:02,100
and slow and it's true to a degree right

298
00:10:00,660 --> 00:10:04,380
you're really comparing native to

299
00:10:02,100 --> 00:10:06,720
bytecode development so just a quick

300
00:10:04,380 --> 00:10:08,040
refresher with Native coding but this is

301
00:10:06,720 --> 00:10:09,660
you know if you're using C you can go

302
00:10:08,040 --> 00:10:10,380
any language that compiles down to the

303
00:10:09,660 --> 00:10:12,240
metal

304
00:10:10,380 --> 00:10:14,279
you take a code you compile it you're

305
00:10:12,240 --> 00:10:16,920
running code that the device can run

306
00:10:14,279 --> 00:10:18,899
directly right by code interpreters work

307
00:10:16,920 --> 00:10:20,820
a little differently they take code they

308
00:10:18,899 --> 00:10:22,320
compile it to bytecode by code gets run

309
00:10:20,820 --> 00:10:23,940
through the virtual machine and then the

310
00:10:22,320 --> 00:10:25,800
virtual machine runs that on the device

311
00:10:23,940 --> 00:10:27,899
directly

312
00:10:25,800 --> 00:10:28,860
uh and slower it's slower if you're

313
00:10:27,899 --> 00:10:30,420
spending time in the virtual machine

314
00:10:28,860 --> 00:10:32,399
there's benefits to it you've heard it's

315
00:10:30,420 --> 00:10:33,720
more portable there's a lot of things

316
00:10:32,399 --> 00:10:36,300
you can do with that but that's the

317
00:10:33,720 --> 00:10:37,920
difference the yellow bar there is just

318
00:10:36,300 --> 00:10:39,480
something that I only occurred to me

319
00:10:37,920 --> 00:10:40,440
later was worth pointing out is that all

320
00:10:39,480 --> 00:10:41,640
of the things that are next to that

321
00:10:40,440 --> 00:10:44,279
yellow bar actually occur on the device

322
00:10:41,640 --> 00:10:46,920
in micropython by default so the

323
00:10:44,279 --> 00:10:49,620
compiler actually works on the device

324
00:10:46,920 --> 00:10:50,519
um yeah all the way down and we'll talk

325
00:10:49,620 --> 00:10:51,540
about that a little further later

326
00:10:50,519 --> 00:10:53,579
because there's a way you can optimize

327
00:10:51,540 --> 00:10:56,220
some of that

328
00:10:53,579 --> 00:10:57,720
uh so techniques for performance there's

329
00:10:56,220 --> 00:10:59,700
a whole bunch of sophisticated methods

330
00:10:57,720 --> 00:11:00,959
to improve performance and micropism

331
00:10:59,700 --> 00:11:02,220
we're not going to cover them today

332
00:11:00,959 --> 00:11:04,320
Damien's talked about them before I've

333
00:11:02,220 --> 00:11:05,760
talked about them before and we'd need

334
00:11:04,320 --> 00:11:07,920
you need a fair bit of time to cover

335
00:11:05,760 --> 00:11:09,540
them uh but just for completeness uh

336
00:11:07,920 --> 00:11:12,180
hopefully these are court on the video

337
00:11:09,540 --> 00:11:14,279
you can go look these up later I'm going

338
00:11:12,180 --> 00:11:16,019
to look at one performance Hammer which

339
00:11:14,279 --> 00:11:17,160
is just using C um all the other

340
00:11:16,019 --> 00:11:18,660
techniques are really sophisticated

341
00:11:17,160 --> 00:11:21,360
elegant and have their place and they

342
00:11:18,660 --> 00:11:24,959
should be used but just in terms of

343
00:11:21,360 --> 00:11:27,120
um why you don't necessarily need to

344
00:11:24,959 --> 00:11:29,519
exclude micro python for performance is

345
00:11:27,120 --> 00:11:31,920
the fact that you can use c a chair

346
00:11:29,519 --> 00:11:33,720
microphize and shares see Python's Dirty

347
00:11:31,920 --> 00:11:36,600
Little Secret and that we offload our

348
00:11:33,720 --> 00:11:39,839
performance to see when we need to not

349
00:11:36,600 --> 00:11:42,420
so much Fortran on embedded devices but

350
00:11:39,839 --> 00:11:44,459
you know so yeah think about like numpy

351
00:11:42,420 --> 00:11:46,560
Hill sci-fi right they both got seen in

352
00:11:44,459 --> 00:11:48,000
the guts of them that makes them work

353
00:11:46,560 --> 00:11:49,500
faster

354
00:11:48,000 --> 00:11:52,079
there's an interesting

355
00:11:49,500 --> 00:11:53,519
um white paper recently that was talking

356
00:11:52,079 --> 00:11:55,079
about comparing micro pythons of other

357
00:11:53,519 --> 00:11:56,820
languages

358
00:11:55,079 --> 00:11:58,700
it's kind of weird

359
00:11:56,820 --> 00:12:02,760
um like this one for example talks about

360
00:11:58,700 --> 00:12:04,740
crc32 they actually implemented a crc32

361
00:12:02,760 --> 00:12:07,320
algorithm in Python and then tested that

362
00:12:04,740 --> 00:12:08,519
we already have the CSC 32 algorithm

363
00:12:07,320 --> 00:12:10,200
which works at the same performance of

364
00:12:08,519 --> 00:12:11,000
the C but they're trying to compare the

365
00:12:10,200 --> 00:12:13,440
language

366
00:12:11,000 --> 00:12:15,720
overheads and what they conclusively

367
00:12:13,440 --> 00:12:16,620
showed is that by code interpreted

368
00:12:15,720 --> 00:12:18,959
languages are slower than native

369
00:12:16,620 --> 00:12:21,120
languages which wasn't that surprising

370
00:12:18,959 --> 00:12:23,579
the problem I have is at the end they

371
00:12:21,120 --> 00:12:25,019
kind of make or they kind of that's to

372
00:12:23,579 --> 00:12:26,820
the conclusion that micro python is not

373
00:12:25,019 --> 00:12:29,579
appropriate for embedded use if

374
00:12:26,820 --> 00:12:30,899
performance is a concerned and I don't

375
00:12:29,579 --> 00:12:33,420
think that's correct it's kind of like

376
00:12:30,899 --> 00:12:35,040
saying you know big python is not

377
00:12:33,420 --> 00:12:37,800
appropriate for data science use because

378
00:12:35,040 --> 00:12:39,240
you know you can't uh you know python

379
00:12:37,800 --> 00:12:41,220
the language is too slow I don't believe

380
00:12:39,240 --> 00:12:43,620
that and so

381
00:12:41,220 --> 00:12:46,860
um I think you need to take this there's

382
00:12:43,620 --> 00:12:48,899
a bit of grain of salt to be taken here

383
00:12:46,860 --> 00:12:51,480
um so it is easy to extend marketplace

384
00:12:48,899 --> 00:12:53,880
with C you'll find the documentation

385
00:12:51,480 --> 00:12:55,560
talking about user modules

386
00:12:53,880 --> 00:12:56,880
um if there's one thing here we can

387
00:12:55,560 --> 00:12:58,500
improve it'll be around documentation

388
00:12:56,880 --> 00:12:59,820
we've got some good documentation but

389
00:12:58,500 --> 00:13:01,620
you can probably go a little further

390
00:12:59,820 --> 00:13:04,440
because it's a fairly complex topic

391
00:13:01,620 --> 00:13:06,959
about Bridging the the um the python to

392
00:13:04,440 --> 00:13:09,420
C boundary but

393
00:13:06,959 --> 00:13:12,360
the key point is that it's just C right

394
00:13:09,420 --> 00:13:13,680
like it's fast so

395
00:13:12,360 --> 00:13:15,839
um

396
00:13:13,680 --> 00:13:17,760
relevant quotes that I have also heard I

397
00:13:15,839 --> 00:13:19,079
might as well just write and see

398
00:13:17,760 --> 00:13:21,180
I think that comes back to that 30

399
00:13:19,079 --> 00:13:22,380
number before you're paying the cost of

400
00:13:21,180 --> 00:13:24,480
having develop and see if you've got to

401
00:13:22,380 --> 00:13:25,500
do that so my argument would be that you

402
00:13:24,480 --> 00:13:27,180
should spend as much time on micro

403
00:13:25,500 --> 00:13:28,920
python as you can and just optimize the

404
00:13:27,180 --> 00:13:30,779
bits that you need and the projects

405
00:13:28,920 --> 00:13:32,579
we've had at work

406
00:13:30,779 --> 00:13:34,500
there's been any minimal amounts of code

407
00:13:32,579 --> 00:13:35,820
that we've had to develop in C so it's a

408
00:13:34,500 --> 00:13:37,560
small fraction

409
00:13:35,820 --> 00:13:38,940
and the other one here is we're all

410
00:13:37,560 --> 00:13:41,279
human well most of you put your hand up

411
00:13:38,940 --> 00:13:42,720
as a human identify as a human so

412
00:13:41,279 --> 00:13:44,880
um but we're always bad at guessing

413
00:13:42,720 --> 00:13:47,540
where to optimize so

414
00:13:44,880 --> 00:13:50,279
um my I guess my statement would be

415
00:13:47,540 --> 00:13:52,639
Friday Michael python first important to

416
00:13:50,279 --> 00:13:55,320
see only when profiling shows a need

417
00:13:52,639 --> 00:13:57,180
and there's a small caveat there with it

418
00:13:55,320 --> 00:13:58,500
almost in Brackets but there's almost no

419
00:13:57,180 --> 00:14:00,060
performance ceiling like we can drop

420
00:13:58,500 --> 00:14:01,560
into C and you do whatever compute you

421
00:14:00,060 --> 00:14:04,100
need

422
00:14:01,560 --> 00:14:04,100
um yeah

423
00:14:04,680 --> 00:14:08,600
you can't use microphone when you're

424
00:14:06,060 --> 00:14:08,600
running for a battery

425
00:14:09,360 --> 00:14:13,200
so a little bit more background

426
00:14:11,220 --> 00:14:15,480
um modern my controllers have low power

427
00:14:13,200 --> 00:14:17,639
modes and this is kind of the way that

428
00:14:15,480 --> 00:14:18,779
they get good battery life

429
00:14:17,639 --> 00:14:20,579
um

430
00:14:18,779 --> 00:14:24,779
the power mode ruins that they're sort

431
00:14:20,579 --> 00:14:27,000
of uh trading off features for power

432
00:14:24,779 --> 00:14:29,040
consumption so they'll turn off the core

433
00:14:27,000 --> 00:14:30,720
and they'll just listen to some pins or

434
00:14:29,040 --> 00:14:33,120
whatever and these can have dramatic

435
00:14:30,720 --> 00:14:34,260
improvements in terms of power use

436
00:14:33,120 --> 00:14:36,420
and so the key thing if you're running

437
00:14:34,260 --> 00:14:38,040
from batteries is to drive your micro

438
00:14:36,420 --> 00:14:39,420
into the lowest power mode you can and

439
00:14:38,040 --> 00:14:42,120
keep it in there for as long as you can

440
00:14:39,420 --> 00:14:44,040
that's the key thing for low power

441
00:14:42,120 --> 00:14:45,300
applications

442
00:14:44,040 --> 00:14:47,220
so if you think about like temperature

443
00:14:45,300 --> 00:14:48,300
logging it's one of the nice easy

444
00:14:47,220 --> 00:14:49,920
example to think about temperature

445
00:14:48,300 --> 00:14:52,079
doesn't change frequently so we only

446
00:14:49,920 --> 00:14:53,699
need to pull it every so often

447
00:14:52,079 --> 00:14:55,079
so you might what does it read

448
00:14:53,699 --> 00:14:56,519
temperature every 30 seconds and you

449
00:14:55,079 --> 00:14:58,019
store it in a buffer

450
00:14:56,519 --> 00:15:00,899
and every 10 minutes you might offload

451
00:14:58,019 --> 00:15:02,579
it over Wi-Fi okay the key thing is to

452
00:15:00,899 --> 00:15:04,019
where you're not doing stuff driving

453
00:15:02,579 --> 00:15:07,079
into those low power modes and get the

454
00:15:04,019 --> 00:15:09,300
benefit of running at lower power

455
00:15:07,079 --> 00:15:12,240
so some quick charts hopefully it's not

456
00:15:09,300 --> 00:15:14,399
too overwhelming but uh

457
00:15:12,240 --> 00:15:16,079
that middle section there is showing

458
00:15:14,399 --> 00:15:17,579
that we're in a deep sleep mode and

459
00:15:16,079 --> 00:15:19,800
we're consuming one and a half milliamps

460
00:15:17,579 --> 00:15:21,420
for this particular microcontroller

461
00:15:19,800 --> 00:15:23,279
the other areas are just where the

462
00:15:21,420 --> 00:15:25,019
process is running as normal it's not

463
00:15:23,279 --> 00:15:29,339
normally responsive but yeah it's a 30

464
00:15:25,019 --> 00:15:30,300
times difference in power

465
00:15:29,339 --> 00:15:31,980
um

466
00:15:30,300 --> 00:15:33,779
this is with Wi-Fi this particular micro

467
00:15:31,980 --> 00:15:36,120
supports Wi-Fi as well and you can see

468
00:15:33,779 --> 00:15:37,620
that the power use goes right up that

469
00:15:36,120 --> 00:15:39,540
low bar there is actually the high bar

470
00:15:37,620 --> 00:15:42,240
from the previous slide

471
00:15:39,540 --> 00:15:43,800
and so we're up in 120 milliamps or

472
00:15:42,240 --> 00:15:46,680
higher if we're actually connected to

473
00:15:43,800 --> 00:15:48,600
Wi-Fi now it's also worth noting that's

474
00:15:46,680 --> 00:15:49,800
with Wi-Fi just connected not doing

475
00:15:48,600 --> 00:15:54,060
anything as soon as you start

476
00:15:49,800 --> 00:15:55,440
transferring stuff it cranks up again

477
00:15:54,060 --> 00:15:57,600
so

478
00:15:55,440 --> 00:15:59,699
um a key thing to be aware of is that

479
00:15:57,600 --> 00:16:01,500
these low power modes have identical

480
00:15:59,699 --> 00:16:02,880
costs regardless of what language you

481
00:16:01,500 --> 00:16:04,500
use the micro doesn't know that you're

482
00:16:02,880 --> 00:16:06,000
running micropath and you'll see it

483
00:16:04,500 --> 00:16:07,860
doesn't care

484
00:16:06,000 --> 00:16:10,699
um so the key thing is to keep in those

485
00:16:07,860 --> 00:16:12,660
low power modes as much as possible

486
00:16:10,699 --> 00:16:14,579
there is an additional cost to using

487
00:16:12,660 --> 00:16:16,079
micropython and that's because when the

488
00:16:14,579 --> 00:16:18,360
micro starts up

489
00:16:16,079 --> 00:16:19,860
we have to start an interpreter often we

490
00:16:18,360 --> 00:16:21,139
have to start a file system we have to

491
00:16:19,860 --> 00:16:24,240
start reading

492
00:16:21,139 --> 00:16:25,920
python compiling it and doing something

493
00:16:24,240 --> 00:16:28,199
now that takes about 40 or 15

494
00:16:25,920 --> 00:16:29,639
milliseconds varies a bit port to port

495
00:16:28,199 --> 00:16:31,980
but that's pretty typical

496
00:16:29,639 --> 00:16:34,199
so what you need to think about is if we

497
00:16:31,980 --> 00:16:36,540
go back to a deep sleep mode for example

498
00:16:34,199 --> 00:16:38,399
we want to come out of our deep sleep

499
00:16:36,540 --> 00:16:40,680
mode and stay in that higher power mode

500
00:16:38,399 --> 00:16:42,720
for as little time as possible so

501
00:16:40,680 --> 00:16:44,480
micropython adds 40 milliseconds is the

502
00:16:42,720 --> 00:16:46,980
way to think about it and then

503
00:16:44,480 --> 00:16:48,959
if you can live with that you can live

504
00:16:46,980 --> 00:16:51,300
with using micro python with battery

505
00:16:48,959 --> 00:16:52,019
powered applications

506
00:16:51,300 --> 00:16:53,639
um

507
00:16:52,019 --> 00:16:55,500
it's also possible to mitigate that cost

508
00:16:53,639 --> 00:16:57,839
as well there's a hook at the start when

509
00:16:55,500 --> 00:16:59,880
micro python starts to latch a little

510
00:16:57,839 --> 00:17:01,980
bit of C code if you want to so if you

511
00:16:59,880 --> 00:17:03,360
can do something like wake do a check to

512
00:17:01,980 --> 00:17:04,799
see if you should be woken up or if you

513
00:17:03,360 --> 00:17:07,620
need to and go back to sleep you

514
00:17:04,799 --> 00:17:08,160
actually mitigate that cost even so

515
00:17:07,620 --> 00:17:11,760
um

516
00:17:08,160 --> 00:17:13,799
yeah so yes there is a small cost to

517
00:17:11,760 --> 00:17:15,780
selling Marketplace and running in a VM

518
00:17:13,799 --> 00:17:18,299
but the key thing you should be doing is

519
00:17:15,780 --> 00:17:20,339
uh keeping down in those low power modes

520
00:17:18,299 --> 00:17:21,540
as much as possible and so the

521
00:17:20,339 --> 00:17:22,860
suggestion here is build a product in

522
00:17:21,540 --> 00:17:25,199
market price list you'll see a theme

523
00:17:22,860 --> 00:17:26,699
Here build a product in Marketplace and

524
00:17:25,199 --> 00:17:28,679
first where you'll work faster and

525
00:17:26,699 --> 00:17:30,419
you'll get to your end goal quicker

526
00:17:28,679 --> 00:17:33,140
and if you can't achieve your power

527
00:17:30,419 --> 00:17:35,220
budget put it to CNC plus plus then

528
00:17:33,140 --> 00:17:36,120
you'll have already learned a lot you'll

529
00:17:35,220 --> 00:17:37,620
have already been able to work with your

530
00:17:36,120 --> 00:17:39,539
customer more you've already been able

531
00:17:37,620 --> 00:17:41,520
to develop the thing that you want

532
00:17:39,539 --> 00:17:44,059
faster so that porting effort should be

533
00:17:41,520 --> 00:17:44,059
pretty small

534
00:17:45,740 --> 00:17:50,400
drink water but you can't use

535
00:17:48,900 --> 00:17:52,580
micropython if you're Manufacturing in

536
00:17:50,400 --> 00:17:52,580
volume

537
00:17:55,620 --> 00:18:01,200
again some truth here so Market python

538
00:17:58,380 --> 00:18:03,960
places a higher Demand on the resource

539
00:18:01,200 --> 00:18:05,460
uses of a of a microcontroller it needs

540
00:18:03,960 --> 00:18:07,860
higher specs because it needs more Flash

541
00:18:05,460 --> 00:18:09,960
and needs more RAM

542
00:18:07,860 --> 00:18:11,880
they're the key ones

543
00:18:09,960 --> 00:18:14,220
so if bomb costs if your bill of

544
00:18:11,880 --> 00:18:16,380
materials cost is critical my price may

545
00:18:14,220 --> 00:18:18,840
not be your best solution but it's worth

546
00:18:16,380 --> 00:18:19,980
having a look at how critical

547
00:18:18,840 --> 00:18:21,720
um so I put this in perspective

548
00:18:19,980 --> 00:18:23,160
microcontroller costs tend to be in the

549
00:18:21,720 --> 00:18:26,640
terms of like I said before like

550
00:18:23,160 --> 00:18:29,580
high-end is dollars tens of dollars low

551
00:18:26,640 --> 00:18:31,200
end is 20 30 cents

552
00:18:29,580 --> 00:18:34,559
um and still get a whole range in

553
00:18:31,200 --> 00:18:36,299
between now eight bits yeah just like

554
00:18:34,559 --> 00:18:37,919
everyone I'm wouldn't be surprised if

555
00:18:36,299 --> 00:18:39,120
you know we'll have stickers in a year

556
00:18:37,919 --> 00:18:40,740
or two with eight bit microcontrollers

557
00:18:39,120 --> 00:18:42,240
stuck to them

558
00:18:40,740 --> 00:18:43,260
um the microphone is not going to run

559
00:18:42,240 --> 00:18:45,840
the low end we just don't have the

560
00:18:43,260 --> 00:18:47,520
resources to do so

561
00:18:45,840 --> 00:18:49,320
um but how low can we go it's kind of

562
00:18:47,520 --> 00:18:50,520
important to look at because typically

563
00:18:49,320 --> 00:18:52,020
when you first come on a Marketplace and

564
00:18:50,520 --> 00:18:53,880
you'll be told that there's you the

565
00:18:52,020 --> 00:18:57,000
requirements are about about 200k a

566
00:18:53,880 --> 00:18:59,160
Flash and about 32k Ram give or take

567
00:18:57,000 --> 00:18:59,760
um it's not quite the whole story

568
00:18:59,160 --> 00:19:01,260
um

569
00:18:59,760 --> 00:19:02,940
it's kind of like when we package

570
00:19:01,260 --> 00:19:06,059
micropython we it's kind of a kitchen

571
00:19:02,940 --> 00:19:08,580
sink approach because micro python is a

572
00:19:06,059 --> 00:19:10,620
dynamic language runtime you need all of

573
00:19:08,580 --> 00:19:11,940
the features available to use them even

574
00:19:10,620 --> 00:19:13,740
if you're not going to so you have to

575
00:19:11,940 --> 00:19:15,660
package the whole thing

576
00:19:13,740 --> 00:19:17,760
but if you know what your application

577
00:19:15,660 --> 00:19:19,679
needs you can start carving it away it's

578
00:19:17,760 --> 00:19:21,660
pretty easy to build your own custom

579
00:19:19,679 --> 00:19:23,220
firmware of micro Python and just carve

580
00:19:21,660 --> 00:19:26,160
out features

581
00:19:23,220 --> 00:19:28,080
um so the claim here is I guess a

582
00:19:26,160 --> 00:19:30,179
kitchen sink micro python is around 200k

583
00:19:28,080 --> 00:19:31,740
a Flash and a minimal micro python is

584
00:19:30,179 --> 00:19:33,000
about 40K a flash I'm looking at gym

585
00:19:31,740 --> 00:19:36,600
because we're trying to get down there

586
00:19:33,000 --> 00:19:37,679
today yeah thumbs up good okay

587
00:19:36,600 --> 00:19:41,100
um

588
00:19:37,679 --> 00:19:43,200
so what can we trade off like features

589
00:19:41,100 --> 00:19:44,340
um f strings async SBI you can carve

590
00:19:43,200 --> 00:19:46,679
these things out at a pretty granular

591
00:19:44,340 --> 00:19:48,299
level you can drop the file system you

592
00:19:46,679 --> 00:19:49,500
can drop the rebel

593
00:19:48,299 --> 00:19:50,820
um you can drop the compiler so you

594
00:19:49,500 --> 00:19:53,160
start trading off these quality of life

595
00:19:50,820 --> 00:19:56,220
development features to aim at a smaller

596
00:19:53,160 --> 00:19:58,679
micro okay so I mentioned before that

597
00:19:56,220 --> 00:20:00,539
everything happens on the on the micro

598
00:19:58,679 --> 00:20:02,700
that's next to that yellow bar but that

599
00:20:00,539 --> 00:20:04,500
red bar is actually an alternative where

600
00:20:02,700 --> 00:20:06,480
we can actually do compilation and

601
00:20:04,500 --> 00:20:09,120
freeze our code into the into the

602
00:20:06,480 --> 00:20:10,260
firmware so it's kind of like in the sea

603
00:20:09,120 --> 00:20:12,720
world you would compile your application

604
00:20:10,260 --> 00:20:14,820
and deploy it in micropy you would

605
00:20:12,720 --> 00:20:17,640
compile your your source code and just

606
00:20:14,820 --> 00:20:19,260
like in Python you'll get a pi C file

607
00:20:17,640 --> 00:20:21,360
um you'll get a similar deal you get by

608
00:20:19,260 --> 00:20:23,520
code that you can put into your firmware

609
00:20:21,360 --> 00:20:25,580
which means that it's you lose that

610
00:20:23,520 --> 00:20:27,299
ability to do it dynamically at runtime

611
00:20:25,580 --> 00:20:28,500
you've got to build it into your

612
00:20:27,299 --> 00:20:30,780
firmware

613
00:20:28,500 --> 00:20:32,820
but you can Target a smaller firmware

614
00:20:30,780 --> 00:20:35,460
blob

615
00:20:32,820 --> 00:20:36,419
uh yeah

616
00:20:35,460 --> 00:20:38,400
so

617
00:20:36,419 --> 00:20:41,000
a possible workflow you can consider is

618
00:20:38,400 --> 00:20:43,380
to start with like a big monstrous

619
00:20:41,000 --> 00:20:45,480
microcontroller and just don't worry

620
00:20:43,380 --> 00:20:47,160
about your problems so if your volume

621
00:20:45,480 --> 00:20:48,960
manufacturing just get your application

622
00:20:47,160 --> 00:20:51,000
running use something that's got a ton

623
00:20:48,960 --> 00:20:53,340
of Slash a ton of RAM and develop as

624
00:20:51,000 --> 00:20:55,799
quick as possible okay and shift to a

625
00:20:53,340 --> 00:20:56,340
micro it's got smaller resources

626
00:20:55,799 --> 00:20:58,440
um

627
00:20:56,340 --> 00:21:00,360
you can start either customizing your

628
00:20:58,440 --> 00:21:02,160
firmware or maybe just customizing your

629
00:21:00,360 --> 00:21:04,280
application so it uses less less

630
00:21:02,160 --> 00:21:06,539
resources

631
00:21:04,280 --> 00:21:07,980
and then you can do that whole whole

632
00:21:06,539 --> 00:21:09,539
another step where you start really

633
00:21:07,980 --> 00:21:11,460
hacking into the firmware making it as

634
00:21:09,539 --> 00:21:12,780
small as possible and developing it so

635
00:21:11,460 --> 00:21:15,179
that it only supports the application

636
00:21:12,780 --> 00:21:16,860
that you write

637
00:21:15,179 --> 00:21:18,720
um and so the costs are important here

638
00:21:16,860 --> 00:21:19,799
these are real examples of boards that

639
00:21:18,720 --> 00:21:22,020
we've got

640
00:21:19,799 --> 00:21:24,120
um where the board to the left is the

641
00:21:22,020 --> 00:21:27,419
big beefy one we talked about it's an H5

642
00:21:24,120 --> 00:21:30,360
cortex m33 I'll gloss those the details

643
00:21:27,419 --> 00:21:33,419
but tons of flash tons of ram about 4.60

644
00:21:30,360 --> 00:21:36,659
and these numbers are all in units of

645
00:21:33,419 --> 00:21:39,480
10K which is St will give you these

646
00:21:36,659 --> 00:21:43,740
prices but they are relatively accurate

647
00:21:39,480 --> 00:21:45,480
I think a units and US dollars so 460

648
00:21:43,740 --> 00:21:47,880
for the big beefy one with tons of

649
00:21:45,480 --> 00:21:51,240
peripherals the middle one's interesting

650
00:21:47,880 --> 00:21:53,280
that's the same uh core it's 250 Meg

651
00:21:51,240 --> 00:21:54,360
cortex M3 but they've dropped a lot of

652
00:21:53,280 --> 00:21:56,700
peripherals they dropped a lot of flash

653
00:21:54,360 --> 00:21:58,260
dropped a lot of RAM and dropped a lot

654
00:21:56,700 --> 00:21:59,220
of cost and so you're suddenly looking

655
00:21:58,260 --> 00:22:00,780
at a dollar sixty for that

656
00:21:59,220 --> 00:22:03,240
microcontroller

657
00:22:00,780 --> 00:22:05,220
but it's a pretty easy Port because if

658
00:22:03,240 --> 00:22:06,720
your code works on the original H5 it's

659
00:22:05,220 --> 00:22:09,799
usually a pretty small jump to go from

660
00:22:06,720 --> 00:22:09,799
the big one to the medium one

661
00:22:09,840 --> 00:22:13,020
then you've got the small one there's a

662
00:22:11,400 --> 00:22:14,400
g zero that I've selected which is a

663
00:22:13,020 --> 00:22:16,980
dollar twenty

664
00:22:14,400 --> 00:22:19,799
um so you save 40 whole cents but you're

665
00:22:16,980 --> 00:22:20,820
going to a much smaller much lower spec

666
00:22:19,799 --> 00:22:23,760
micro

667
00:22:20,820 --> 00:22:25,500
now getting down to that size is more

668
00:22:23,760 --> 00:22:27,539
difficult

669
00:22:25,500 --> 00:22:30,200
um but it's not it's not out of the

670
00:22:27,539 --> 00:22:30,200
realm of possibility

671
00:22:30,539 --> 00:22:33,840
uh

672
00:22:32,400 --> 00:22:36,480
right so

673
00:22:33,840 --> 00:22:38,280
uh micro again there's a German truth

674
00:22:36,480 --> 00:22:39,740
here like Microsoft does Place High

675
00:22:38,280 --> 00:22:43,440
demands on

676
00:22:39,740 --> 00:22:45,780
a microcontroller resources but there

677
00:22:43,440 --> 00:22:47,760
are ways to offset some of that and so

678
00:22:45,780 --> 00:22:49,620
yeah I'd suggest like I have in all the

679
00:22:47,760 --> 00:22:53,539
other ones I've had micropython first

680
00:22:49,620 --> 00:22:53,539
and make changes if you have to later

681
00:22:54,480 --> 00:22:58,320
and that is my summary really uh

682
00:22:56,760 --> 00:23:00,360
development important to say if you

683
00:22:58,320 --> 00:23:02,340
really must right

684
00:23:00,360 --> 00:23:03,780
um there's no real the barriers here

685
00:23:02,340 --> 00:23:07,020
seem a lot more

686
00:23:03,780 --> 00:23:08,940
um Impressions than actual reality in my

687
00:23:07,020 --> 00:23:10,320
opinion and I guess I believe that

688
00:23:08,940 --> 00:23:11,580
thought was that

689
00:23:10,320 --> 00:23:13,620
I expected to find a lot more

690
00:23:11,580 --> 00:23:14,760
limitations developing with micro python

691
00:23:13,620 --> 00:23:16,620
I thought it was going to be useful for

692
00:23:14,760 --> 00:23:17,940
certain class of microcontrollers for

693
00:23:16,620 --> 00:23:19,980
certain applications

694
00:23:17,940 --> 00:23:22,320
but now it's my default choice I would

695
00:23:19,980 --> 00:23:24,720
need to be convinced to start a start an

696
00:23:22,320 --> 00:23:28,380
application in CNC plus plus and it's

697
00:23:24,720 --> 00:23:31,980
often the wrong choice it might be

698
00:23:28,380 --> 00:23:32,940
okay so uh actually just quickly uh for

699
00:23:31,980 --> 00:23:34,260
you to mentioned the stickers but we've

700
00:23:32,940 --> 00:23:37,500
got we're going to try and get to the

701
00:23:34,260 --> 00:23:38,880
python fair and uh Sprint day Monday

702
00:23:37,500 --> 00:23:40,980
um so if you guys are interested please

703
00:23:38,880 --> 00:23:42,780
stick around and thanks for listening

704
00:23:40,980 --> 00:23:44,700
and thanks to the organizers and

705
00:23:42,780 --> 00:23:47,340
Marketplace and community and especially

706
00:23:44,700 --> 00:23:48,659
to the uh my support team who helped out

707
00:23:47,340 --> 00:23:50,159
with making sure all these claims were

708
00:23:48,659 --> 00:23:51,240
actually possible which they more or

709
00:23:50,159 --> 00:23:51,960
less aren't

710
00:23:51,240 --> 00:23:55,580
um

711
00:23:51,960 --> 00:23:55,580
So yeah thank you

712
00:23:56,270 --> 00:23:59,520
[Music]

713
00:24:00,480 --> 00:24:06,360
all right we have time for a couple of

714
00:24:02,700 --> 00:24:08,880
questions if anyone has any

715
00:24:06,360 --> 00:24:10,140
sure that'll go right here

716
00:24:08,880 --> 00:24:11,640
hi

717
00:24:10,140 --> 00:24:15,120
um I was wondering if you tested and if

718
00:24:11,640 --> 00:24:15,840
so what uh the results were around the

719
00:24:15,120 --> 00:24:17,940
um

720
00:24:15,840 --> 00:24:20,520
what kind of performance gain did you

721
00:24:17,940 --> 00:24:22,860
see when you froze that code and and

722
00:24:20,520 --> 00:24:25,080
pre-compiled it uh did you mention that

723
00:24:22,860 --> 00:24:28,140
at all and see how that went we because

724
00:24:25,080 --> 00:24:30,299
it's a no it's worth talking about

725
00:24:28,140 --> 00:24:33,000
um so just for everyone's benefit like

726
00:24:30,299 --> 00:24:34,620
frozen code we um as we mentioned

727
00:24:33,000 --> 00:24:36,240
briefly we take python code we get by

728
00:24:34,620 --> 00:24:37,020
code out of it that gets stored in the

729
00:24:36,240 --> 00:24:39,900
flash

730
00:24:37,020 --> 00:24:42,240
um normally that would happen uh on the

731
00:24:39,900 --> 00:24:43,440
file system uh so what we do is we avoid

732
00:24:42,240 --> 00:24:45,659
a lot of memory use because we don't

733
00:24:43,440 --> 00:24:48,299
need to invoke the compiler and we don't

734
00:24:45,659 --> 00:24:50,520
need to compile it so

735
00:24:48,299 --> 00:24:52,860
um it varies wildly unfortunately is the

736
00:24:50,520 --> 00:24:55,320
way to say it I don't know if anyone's

737
00:24:52,860 --> 00:24:57,419
got any figures out there but um yeah

738
00:24:55,320 --> 00:24:59,340
yeah

739
00:24:57,419 --> 00:25:02,179
that thing is shaking his head Damien

740
00:24:59,340 --> 00:25:02,179
doesn't know it doesn't it

741
00:25:03,000 --> 00:25:06,419
yep I got a question so you mentioned

742
00:25:04,799 --> 00:25:07,919
that one of the ways to shrink down the

743
00:25:06,419 --> 00:25:09,720
footprint is to custom build that

744
00:25:07,919 --> 00:25:11,640
firmware is there any tools where I can

745
00:25:09,720 --> 00:25:13,440
just point it to my py file and it goes

746
00:25:11,640 --> 00:25:14,700
okay you're using these things I'm just

747
00:25:13,440 --> 00:25:16,980
going to build that firmware against

748
00:25:14,700 --> 00:25:18,480
that not yet yeah we should create that

749
00:25:16,980 --> 00:25:19,679
we should yeah let's do that that's a

750
00:25:18,480 --> 00:25:22,440
great idea we've got a Sprint day Monday

751
00:25:19,679 --> 00:25:25,080
right what are you up to okay

752
00:25:22,440 --> 00:25:27,419
it'd be good to do that analysis yeah it

753
00:25:25,080 --> 00:25:30,179
certainly came up during this discussion

754
00:25:27,419 --> 00:25:31,440
of this talking

755
00:25:30,179 --> 00:25:33,419
so it does sound like you're

756
00:25:31,440 --> 00:25:34,559
volunteering there Adam so are you

757
00:25:33,419 --> 00:25:37,460
showing up with the Sprints on Monday

758
00:25:34,559 --> 00:25:37,460
and building that

759
00:25:38,159 --> 00:25:41,360
if you Melvin come on tonight

760
00:25:41,720 --> 00:25:48,120
do we have any other questions yeah um

761
00:25:45,240 --> 00:25:51,059
I'm very I'm passingly familiar with uh

762
00:25:48,120 --> 00:25:52,860
writing C extensions for C python the

763
00:25:51,059 --> 00:25:54,120
the big python yep

764
00:25:52,860 --> 00:25:57,600
um thank you for using the correct term

765
00:25:54,120 --> 00:26:00,419
yeah sorry I realized as I said it

766
00:25:57,600 --> 00:26:02,580
um I'm just wondering if you have uh if

767
00:26:00,419 --> 00:26:05,460
you couldn't compare the experience of

768
00:26:02,580 --> 00:26:07,860
writing a c extension for big python

769
00:26:05,460 --> 00:26:10,140
versus microfizing uh it's completely

770
00:26:07,860 --> 00:26:11,880
different it's a short answer

771
00:26:10,140 --> 00:26:13,380
um I would say I found having an

772
00:26:11,880 --> 00:26:14,460
embedded background I found it easier to

773
00:26:13,380 --> 00:26:16,260
do the micro python side of things

774
00:26:14,460 --> 00:26:18,720
because it's kind of a it's kind of an

775
00:26:16,260 --> 00:26:20,820
obvious barrier to to jump

776
00:26:18,720 --> 00:26:22,500
um with C python there was more things I

777
00:26:20,820 --> 00:26:24,720
had to learn about how C python operated

778
00:26:22,500 --> 00:26:25,919
to create those extensions

779
00:26:24,720 --> 00:26:26,880
yeah so if you've gone in better

780
00:26:25,919 --> 00:26:28,380
background I think you'll find it

781
00:26:26,880 --> 00:26:29,220
relatively straightforward if you

782
00:26:28,380 --> 00:26:30,000
haven't

783
00:26:29,220 --> 00:26:33,000
um

784
00:26:30,000 --> 00:26:34,559
and probably in general like uh once you

785
00:26:33,000 --> 00:26:37,260
get beyond the sort of beginner range of

786
00:26:34,559 --> 00:26:38,279
micropython the most the biggest ramp up

787
00:26:37,260 --> 00:26:40,380
that you're going to have is becoming

788
00:26:38,279 --> 00:26:43,559
familiar with the embedded domain if

789
00:26:40,380 --> 00:26:45,179
you're not familiar with it so uh yeah I

790
00:26:43,559 --> 00:26:46,380
guess you'll find extensions relatively

791
00:26:45,179 --> 00:26:48,679
easy to do if you're sort of familiar

792
00:26:46,380 --> 00:26:48,679
with that

793
00:26:51,179 --> 00:26:55,260
anyone else

794
00:26:53,640 --> 00:26:57,919
oh cool

795
00:26:55,260 --> 00:26:57,919
um in that case

796
00:27:04,620 --> 00:27:09,419
uh thank you so much

797
00:27:07,380 --> 00:27:11,460
um for giving this talk it's been

798
00:27:09,419 --> 00:27:11,870
amazing thank you so much

799
00:27:11,460 --> 00:27:18,509
thank you

800
00:27:11,870 --> 00:27:18,509
[Applause]