1
00:00:06,530 --> 00:00:08,440
- Productivity versus performance.

2
00:00:08,440 --> 00:00:10,140
This is a big one.

3
00:00:10,140 --> 00:00:13,677
As an industry, if I asked
people, for a long time,

4
00:00:13,677 --> 00:00:15,840
"Is performance your highest priority?"

5
00:00:15,840 --> 00:00:17,210
They would say yes.

6
00:00:17,210 --> 00:00:19,470
But the reality is that productivity

7
00:00:19,470 --> 00:00:20,680
has been our highest priority.

8
00:00:20,680 --> 00:00:22,970
It has been in just the
languages we've been designing,

9
00:00:22,970 --> 00:00:24,350
the languages we've been using,

10
00:00:24,350 --> 00:00:26,730
people wanna be productive
and get things done,

11
00:00:26,730 --> 00:00:29,390
and we've kind of looked at performance,

12
00:00:29,390 --> 00:00:30,500
it's secondary, though.

13
00:00:30,500 --> 00:00:33,010
If you ask anybody, they're
too afraid to tell you that,

14
00:00:33,010 --> 00:00:35,360
because then you don't
look like you're technical,

15
00:00:35,360 --> 00:00:36,480
or your chops are in place.

16
00:00:36,480 --> 00:00:38,010
We put people on a pedestal.

17
00:00:38,010 --> 00:00:39,470
We can get on stage and talk about

18
00:00:39,470 --> 00:00:41,350
how they can shave a few milliseconds

19
00:00:41,350 --> 00:00:43,300
or microseconds off of an algorithm,

20
00:00:43,300 --> 00:00:45,610
and I really kinda wanna
move away from that.

21
00:00:45,610 --> 00:00:48,980
Productivity really is, and
should be, our highest priority,

22
00:00:48,980 --> 00:00:50,910
but we do need performance,

23
00:00:50,910 --> 00:00:52,970
it does matter.

24
00:00:52,970 --> 00:00:56,480
And what I find interesting
about our industry,

25
00:00:56,480 --> 00:00:59,350
is that for the last 30 years

26
00:00:59,350 --> 00:01:01,190
we have kind of been on the same path

27
00:01:01,190 --> 00:01:04,290
that Niklaus Wirth has said, back in 1987,

28
00:01:04,290 --> 00:01:07,337
he said that "The hope is
that the progress in hardware

29
00:01:07,337 --> 00:01:09,590
"is gonna cure our software ills."

30
00:01:09,590 --> 00:01:11,990
In other words, he's saying
that we as software developers

31
00:01:11,990 --> 00:01:13,797
have always said "We don't have to worry

32
00:01:13,797 --> 00:01:15,957
"about performance, the
hardware will come in

33
00:01:15,957 --> 00:01:17,230
"and save us."

34
00:01:17,230 --> 00:01:19,870
We've continued to kind
of have this mantra

35
00:01:19,870 --> 00:01:21,270
for the last 30 years.

36
00:01:21,270 --> 00:01:24,250
Back in 1987, after I
graduated high school,

37
00:01:24,250 --> 00:01:25,760
I was on a plane to Italy

38
00:01:25,760 --> 00:01:27,850
and I sat next to a software developer

39
00:01:27,850 --> 00:01:30,420
and that software developer
told me, this is 1987,

40
00:01:30,420 --> 00:01:32,830
all of us, right now, watching this video,

41
00:01:32,830 --> 00:01:34,460
would be Basic developers.

42
00:01:34,460 --> 00:01:36,290
We'd all be coding in Basic today,

43
00:01:36,290 --> 00:01:38,240
because the hardware would be so fast

44
00:01:38,240 --> 00:01:40,300
we wouldn't need anything else.

45
00:01:40,300 --> 00:01:42,370
The reality is that has not happened.

46
00:01:42,370 --> 00:01:44,547
I mean, Henry, back in 2015, wrote,

47
00:01:44,547 --> 00:01:46,067
"The most amazing achievement

48
00:01:46,067 --> 00:01:47,367
"of the computer software industry

49
00:01:47,367 --> 00:01:49,377
"is its continuing cancellation

50
00:01:49,377 --> 00:01:50,697
"of the steady and staggering gains

51
00:01:50,697 --> 00:01:53,260
"made by the computer hardware industry."

52
00:01:53,260 --> 00:01:56,570
We're still talking about
this stuff, 30 years ago,

53
00:01:56,570 --> 00:01:58,080
remember I told you before,

54
00:01:58,080 --> 00:02:00,110
the hardware is the platform.

55
00:02:00,110 --> 00:02:01,580
If performance matters,

56
00:02:01,580 --> 00:02:04,410
then we have to be
sympathetic with the hardware,

57
00:02:04,410 --> 00:02:06,120
mechanical sympathy.

58
00:02:06,120 --> 00:02:07,260
And look, you know,

59
00:02:07,260 --> 00:02:09,180
the hardware folks really
haven't changed hardware

60
00:02:09,180 --> 00:02:10,570
in the last 15 plus years.

61
00:02:10,570 --> 00:02:11,660
They really haven't.

62
00:02:11,660 --> 00:02:13,720
No, we've had these multi-core processors,

63
00:02:13,720 --> 00:02:15,290
multiple hardware threads per core,

64
00:02:15,290 --> 00:02:17,600
and the reason why they're not doing it

65
00:02:17,600 --> 00:02:19,640
is because we haven't been
able to take full advantage

66
00:02:19,640 --> 00:02:20,710
of what's there now.

67
00:02:20,710 --> 00:02:22,880
We're starting to see now

68
00:02:22,880 --> 00:02:27,050
Intel talk about their
24, 36 core processors,

69
00:02:27,050 --> 00:02:29,470
and that's great and it sounds amazing,

70
00:02:29,470 --> 00:02:32,070
but if we don't engineer
for this hardware,

71
00:02:32,070 --> 00:02:34,480
if we don't understand how it works,

72
00:02:34,480 --> 00:02:36,910
those processors are
actually gonna slow us down

73
00:02:36,910 --> 00:02:38,340
not speed us up.

74
00:02:38,340 --> 00:02:39,950
So we're gonna spend
some time in this class

75
00:02:39,950 --> 00:02:41,970
talking about how caching works,

76
00:02:41,970 --> 00:02:43,380
and some of the mechanical sympathies

77
00:02:43,380 --> 00:02:46,570
the language drives
through us and allows us,

78
00:02:46,570 --> 00:02:49,330
so we can get to these levels
of performance that we want.

79
00:02:49,330 --> 00:02:50,610
And hopefully take advantage

80
00:02:50,610 --> 00:02:51,790
of the hardware that's there,

81
00:02:51,790 --> 00:02:53,760
so that the hardware
industry stops playing games,

82
00:02:53,760 --> 00:02:54,960
starts giving us more.

83
00:02:54,960 --> 00:02:56,710
But that's on us, it's not on them.

84
00:02:57,640 --> 00:02:59,770
But I think a core part of this class

85
00:02:59,770 --> 00:03:02,550
is gonna be this next
quote from Brian Kernighan,

86
00:03:02,550 --> 00:03:05,120
and it starts, him talking about C.

87
00:03:05,120 --> 00:03:08,027
He's asked a question about
the C programming language,

88
00:03:08,027 --> 00:03:10,780
but I think you can replace C with Go,

89
00:03:10,780 --> 00:03:13,967
and it absolutely mirrors it exactly.

90
00:03:13,967 --> 00:03:15,867
"Go is the best balance I've seen

91
00:03:15,867 --> 00:03:17,447
"between power and expressiveness.

92
00:03:17,447 --> 00:03:19,287
"You can do almost anything you want to do

93
00:03:19,287 --> 00:03:21,947
"by programming fairly straightforwardly,

94
00:03:21,947 --> 00:03:23,187
"and have a good mental model

95
00:03:23,187 --> 00:03:24,747
"of what's going to happen on the machine,

96
00:03:24,747 --> 00:03:26,707
"and you can predict reasonably well,

97
00:03:26,707 --> 00:03:27,867
"and how quickly it's gonna run,

98
00:03:27,867 --> 00:03:30,010
"and understand what's going on."

99
00:03:30,010 --> 00:03:32,280
This is Go in a nutshell.

100
00:03:32,280 --> 00:03:35,990
I'm going to teach you how to
read code in this language,

101
00:03:35,990 --> 00:03:38,320
so that holds true not just for me,

102
00:03:38,320 --> 00:03:39,200
but for you.

103
00:03:39,200 --> 00:03:41,400
And it's possible because Go's model

104
00:03:41,400 --> 00:03:42,970
is not a virtual machine,

105
00:03:42,970 --> 00:03:44,310
it is a real machine.

106
00:03:44,310 --> 00:03:46,050
It is the real machine,

107
00:03:46,050 --> 00:03:48,950
and once you're coding
against the real machine

108
00:03:48,950 --> 00:03:51,300
then you can do all of these things

109
00:03:51,300 --> 00:03:54,670
if you just know a little
bit about the mechanics

110
00:03:54,670 --> 00:03:57,360
and the semantics of the language.

111
00:03:57,360 --> 00:03:58,210
When I say mechanics,

112
00:03:58,210 --> 00:03:59,680
I'm talking about how things work,

113
00:03:59,680 --> 00:04:00,890
when I say semantics,

114
00:04:00,890 --> 00:04:02,620
I'm talking about how things behave.

115
00:04:02,620 --> 00:04:05,060
And I'm gonna teach you
throughout this class

116
00:04:05,060 --> 00:04:06,280
how to be able to do that,

117
00:04:06,280 --> 00:04:08,460
and that's going to drive you

118
00:04:08,460 --> 00:04:11,183
being able to make some
better engineering choices.