path: root/2020/subtitles/emacsconf-2020--27-state-of-retro-gaming-in-emacs-chip8--vasilij-wasamasa-schneidermann-autogen.vtt

WEBVTT

00:00:00.880 --> 00:00:02.879
hello everyone and welcome to my talk

00:00:02.879 --> 00:00:04.520
state of retrogaming and Emacs

00:00:04.520 --> 00:00:06.960
[Music]

00:00:06.960 --> 00:00:08.639
first of all a little bit about myself

00:00:08.639 --> 00:00:12.000
my name is neilman I'm 28 years old

00:00:12.000 --> 00:00:13.599
I work as a cyber security consultant

00:00:13.599 --> 00:00:15.200
msg systems and

00:00:15.200 --> 00:00:17.440
test other people's web applications and

00:00:17.440 --> 00:00:19.359
review the source code for security

00:00:19.359 --> 00:00:20.160
problems

00:00:20.160 --> 00:00:22.960
you can reach me by email I have my own

00:00:22.960 --> 00:00:25.039
self-hosted git repositories

00:00:25.039 --> 00:00:26.480
and I have a blog where you can

00:00:26.480 --> 00:00:28.480
occasionally find new posts by me on all

00:00:28.480 --> 00:00:32.160
kinds of things not just emix things

00:00:32.160 --> 00:00:34.800
so but the motivation about this one I

00:00:34.800 --> 00:00:36.480
found that Emacs is the ultimate

00:00:36.480 --> 00:00:38.079
procrastination machine and there are

00:00:38.079 --> 00:00:40.160
lots of fun demonstrations I'll go over

00:00:40.160 --> 00:00:41.200
a few of them

00:00:41.200 --> 00:00:44.160
for example someone made a thing to

00:00:44.160 --> 00:00:46.079
order sell it for himself online so it

00:00:46.079 --> 00:00:48.239
doesn't have to walk over to the shop

00:00:48.239 --> 00:00:50.879
there's plenty rc bots there's some game

00:00:50.879 --> 00:00:51.760
things

00:00:51.760 --> 00:00:53.520
there's an emulator for the z machine

00:00:53.520 --> 00:00:55.600
which you can use to play zorg

00:00:55.600 --> 00:00:57.600
and so I asked myself at this point can

00:00:57.600 --> 00:00:59.039
you actually emulate retro games at

00:00:59.039 --> 00:01:01.039
60fps and it looked around a bit

00:01:01.039 --> 00:01:02.800
and found some projects but none that

00:01:02.800 --> 00:01:04.479
were actually able to

00:01:04.479 --> 00:01:07.360
do it at 60fps so I set out to do my own

00:01:07.360 --> 00:01:08.000
one

00:01:08.000 --> 00:01:09.439
and looked out for a console that you

00:01:09.439 --> 00:01:11.119
can actually emulate at that speed

00:01:11.119 --> 00:01:13.439
using emax with its very very limited

00:01:13.439 --> 00:01:14.690
rendering

00:01:14.690 --> 00:01:16.320
[Music]

00:01:16.320 --> 00:01:19.520
and here's the project chip8.el it's

00:01:19.520 --> 00:01:20.560
pretty much finished

00:01:20.560 --> 00:01:22.880
it clocks into under 1000 sourced lines

00:01:22.880 --> 00:01:24.000
of code

00:01:24.000 --> 00:01:26.159
it supports the superchip 8 extensions

00:01:26.159 --> 00:01:28.080
it runs at full speed all games behave

00:01:28.080 --> 00:01:30.159
okay as far as I'm concerned and

00:01:30.159 --> 00:01:32.320
yeah I'm pretty happy with it it's very

00:01:32.320 --> 00:01:34.479
much the hell world of emulation

00:01:34.479 --> 00:01:37.040
and I might maybe do some other

00:01:37.040 --> 00:01:40.880
emulation projects in the future

00:01:40.880 --> 00:01:43.360
now for the section which is the longest

00:01:43.360 --> 00:01:45.439
bunch of fun facts about ship a dot el

00:01:45.439 --> 00:01:46.320
which I've learned

00:01:46.320 --> 00:01:49.759
during this project so

00:01:49.759 --> 00:01:52.240
what the hell is debate anyway first of

00:01:52.240 --> 00:01:54.640
all unlike many other emulation game

00:01:54.640 --> 00:01:56.799
things it's not a console but a vm

00:01:56.799 --> 00:01:58.560
it was designed for easy parting of home

00:01:58.560 --> 00:02:00.000
computer games

00:02:00.000 --> 00:02:02.560
it wasn't terribly successful and but

00:02:02.560 --> 00:02:03.680
there's still a small community of

00:02:03.680 --> 00:02:04.320
enthusiasts

00:02:04.320 --> 00:02:06.079
writing games for it and there are even

00:02:06.079 --> 00:02:09.119
a few demos

00:02:09.119 --> 00:02:11.920
this vm has system specs it has a very

00:02:11.920 --> 00:02:14.720
very simple 8-bit cpu with 16 registers

00:02:14.720 --> 00:02:14.959
and

00:02:14.959 --> 00:02:17.760
36 fixed size instructions you have a

00:02:17.760 --> 00:02:18.160
whole

00:02:18.160 --> 00:02:20.560
4 kilobyte of ram you have a stack with

00:02:20.560 --> 00:02:22.080
16 return addresses

00:02:22.080 --> 00:02:24.480
the resolution is 64 by 32 black white

00:02:24.480 --> 00:02:25.760
pixels

00:02:25.760 --> 00:02:28.000
rendering is done by drawing sprites

00:02:28.000 --> 00:02:29.440
these are drawn in excel mode meaning

00:02:29.440 --> 00:02:30.160
that if you

00:02:30.160 --> 00:02:32.239
draw a sprite and set a bit it just

00:02:32.239 --> 00:02:33.840
flips over from black to white or white

00:02:33.840 --> 00:02:35.040
to black

00:02:35.040 --> 00:02:36.560
first one you have a modern buzz that

00:02:36.560 --> 00:02:38.239
can just beep at one

00:02:38.239 --> 00:02:40.640
frequency and most unusually there's a

00:02:40.640 --> 00:02:43.120
hexadecimal keypad as input

00:02:43.120 --> 00:02:45.360
so the keys are basically zero to nine

00:02:45.360 --> 00:02:48.480
and a to f

00:02:48.480 --> 00:02:50.879
so how does this whole thing work it

00:02:50.879 --> 00:02:52.400
runs an unspecified speed

00:02:52.400 --> 00:02:53.599
you'll probably have to do some fine

00:02:53.599 --> 00:02:54.879
tune you find the speed you're happy

00:02:54.879 --> 00:02:56.080
with

00:02:56.080 --> 00:02:58.560
sound and delay timers exist they count

00:02:58.560 --> 00:03:01.120
down at 60fps down to zero

00:03:01.120 --> 00:03:02.879
this is done so that you can play a

00:03:02.879 --> 00:03:05.120
sound at some specific time

00:03:05.120 --> 00:03:06.640
the game itself is loaded with a fixed

00:03:06.640 --> 00:03:08.640
offset into ram the program account is

00:03:08.640 --> 00:03:10.480
set to exactly that offset

00:03:10.480 --> 00:03:11.920
and from there it enters the game loop

00:03:11.920 --> 00:03:13.840
where decodes and instruction executes

00:03:13.840 --> 00:03:15.519
it for the side effects and just

00:03:15.519 --> 00:03:18.130
loops and does this at infinitum

00:03:18.130 --> 00:03:19.599
[Music]

00:03:19.599 --> 00:03:21.920
so the game was the first thing where

00:03:21.920 --> 00:03:23.920
into problems the usual game approach is

00:03:23.920 --> 00:03:24.239
to

00:03:24.239 --> 00:03:26.640
do stuff figure out how long to eight

00:03:26.640 --> 00:03:28.239
wait for exactly that much and

00:03:28.239 --> 00:03:30.640
repeat this doesn't work well in imax at

00:03:30.640 --> 00:03:31.680
all because well

00:03:31.680 --> 00:03:34.959
user input basically and

00:03:34.959 --> 00:03:37.280
Emacs is designed to just do whatever it

00:03:37.280 --> 00:03:39.040
needs to do whenever you enter use input

00:03:39.040 --> 00:03:40.080
instead of

00:03:40.080 --> 00:03:42.799
doing things at one specific time if you

00:03:42.799 --> 00:03:43.440
try to do

00:03:43.440 --> 00:03:45.040
interruptable sleep well you get

00:03:45.040 --> 00:03:46.640
unpredictable behavior

00:03:46.640 --> 00:03:49.440
for example can be the timer doesn't run

00:03:49.440 --> 00:03:50.959
at all at next time because you've

00:03:50.959 --> 00:03:52.560
accidentally cancelled it

00:03:52.560 --> 00:03:54.400
if you do uninterruptable sleep it's

00:03:54.400 --> 00:03:55.760
freezes instead which isn't what you

00:03:55.760 --> 00:03:56.720
want either

00:03:56.720 --> 00:03:59.360
so I went for timers which forced me to

00:03:59.360 --> 00:04:00.560
do inversion of control

00:04:00.560 --> 00:04:02.159
meaning that I have to write code in the

00:04:02.159 --> 00:04:04.080
style where it's just call it

00:04:04.080 --> 00:04:06.159
time and this allows this input to

00:04:06.159 --> 00:04:07.200
happen and to

00:04:07.200 --> 00:04:09.120
for things to progress at roughly the

00:04:09.120 --> 00:04:11.040
speed I want to

00:04:11.040 --> 00:04:12.879
so there's the skydiver function which

00:04:12.879 --> 00:04:14.159
is called a 60fps

00:04:14.159 --> 00:04:16.000
and I have to be very careful to not do

00:04:16.000 --> 00:04:17.359
too much in it

00:04:17.359 --> 00:04:20.479
and say this function execute cpu cycles

00:04:20.479 --> 00:04:22.960
decrypt the sound delay registers and

00:04:22.960 --> 00:04:23.680
redraw

00:04:23.680 --> 00:04:27.759
the screen so to map this whole system

00:04:27.759 --> 00:04:28.800
to mx lisp

00:04:28.800 --> 00:04:31.199
I've used just integers and vectors

00:04:31.199 --> 00:04:33.120
which contain even more integers

00:04:33.120 --> 00:04:35.040
this is used for the ram registers

00:04:35.040 --> 00:04:37.759
return stack key state screen and so on

00:04:37.759 --> 00:04:39.120
and so forth basically

00:04:39.120 --> 00:04:41.520
what you would do if you were writing c

00:04:41.520 --> 00:04:42.800
all of this is stored in global

00:04:42.800 --> 00:04:44.560
variables I'm not using any

00:04:44.560 --> 00:04:46.479
lists at all and as a side effect

00:04:46.479 --> 00:04:48.400
there's no constant going on at all

00:04:48.400 --> 00:04:50.320
there are no extra objects created which

00:04:50.320 --> 00:04:51.120
would trigger

00:04:51.120 --> 00:04:53.919
garbage collection pulses this getting

00:04:53.919 --> 00:04:55.840
this red was rather tricky actually and

00:04:55.840 --> 00:04:56.720
there were some

00:04:56.720 --> 00:04:58.560
in garbage collection problems which I

00:04:58.560 --> 00:05:01.759
had to resolve over time

00:05:01.759 --> 00:05:04.320
so the coding instructions for this you

00:05:04.320 --> 00:05:05.520
have to know that all instructions are

00:05:05.520 --> 00:05:06.800
two bytes long

00:05:06.800 --> 00:05:08.880
and the arguments encoded inside them

00:05:08.880 --> 00:05:10.240
for example the jump to address

00:05:10.240 --> 00:05:11.440
instruction

00:05:11.440 --> 00:05:15.120
is encoded as one and three hex digits

00:05:15.120 --> 00:05:18.400
the type is extracted masking with f000

00:05:18.400 --> 00:05:20.800
and then shifting it by 12 bits mask

00:05:20.800 --> 00:05:21.680
means the hd

00:05:21.680 --> 00:05:24.000
performance binary end you can do the

00:05:24.000 --> 00:05:25.440
same with the argument basement with

00:05:25.440 --> 00:05:26.639
zero fff

00:05:26.639 --> 00:05:29.520
and no shift if you do this long enough

00:05:29.520 --> 00:05:31.039
you'll find common patterns for example

00:05:31.039 --> 00:05:32.639
addresses are always encoded like this

00:05:32.639 --> 00:05:35.280
using the last three nibbles in the code

00:05:35.280 --> 00:05:36.880
you'll find a big count which dispatches

00:05:36.880 --> 00:05:38.400
on the type and executes it for the side

00:05:38.400 --> 00:05:40.070
effects

00:05:40.070 --> 00:05:41.440
[Music]

00:05:41.440 --> 00:05:43.440
for testing I've initially just accused

00:05:43.440 --> 00:05:45.919
the rom until I fit ctrl g

00:05:45.919 --> 00:05:47.280
and then use the debug command to run

00:05:47.280 --> 00:05:49.039
the screen to a buffer

00:05:49.039 --> 00:05:51.360
later on I found tiny roms that just

00:05:51.360 --> 00:05:52.320
display a static

00:05:52.320 --> 00:05:55.680
test screen for example logo and looked

00:05:55.680 --> 00:05:57.280
whether it looked right

00:05:57.280 --> 00:05:59.199
I added instructions as needed and went

00:05:59.199 --> 00:06:00.960
through more and more and more roms and

00:06:00.960 --> 00:06:03.360
later I wrote in unit test suite as a

00:06:03.360 --> 00:06:05.199
safety net and this unit test suite it

00:06:05.199 --> 00:06:06.000
just

00:06:06.000 --> 00:06:08.400
sets up an empty emulator state executes

00:06:08.400 --> 00:06:10.080
some instructions and then looks whether

00:06:10.080 --> 00:06:10.400
the

00:06:10.400 --> 00:06:14.880
expected side effects have happened

00:06:14.880 --> 00:06:17.120
for debugging I usually use e-debug but

00:06:17.120 --> 00:06:19.120
this was super ineffective because well

00:06:19.120 --> 00:06:20.880
you don't really want to step through

00:06:20.880 --> 00:06:22.960
big cons doing side effects for every

00:06:22.960 --> 00:06:24.960
single cycle when it can take like 100

00:06:24.960 --> 00:06:26.880
cycles for things to happen

00:06:26.880 --> 00:06:29.680
therefore I've set up logging and

00:06:29.680 --> 00:06:31.360
whenever I locked something

00:06:31.360 --> 00:06:32.720
and couldn't figure out the error I

00:06:32.720 --> 00:06:33.919
compared my lock output with

00:06:33.919 --> 00:06:37.039
instrumented version of another emulator

00:06:37.039 --> 00:06:39.199
and if the locks diverge then I have

00:06:39.199 --> 00:06:40.880
figured out where the bug lies and could

00:06:40.880 --> 00:06:42.720
look deeper into it

00:06:42.720 --> 00:06:44.479
future project idea might be a chip 8

00:06:44.479 --> 00:06:46.639
debugger but I doubt I'll ever

00:06:46.639 --> 00:06:50.720
go into it for analysis I initially

00:06:50.720 --> 00:06:52.639
wrote a disassembler which is a very

00:06:52.639 --> 00:06:54.400
simple thing but super tedious

00:06:54.400 --> 00:06:56.160
especially if you wanted to add advanced

00:06:56.160 --> 00:06:57.599
functionality for example analysis or

00:06:57.599 --> 00:06:59.120
thinking of what part is data what had

00:06:59.120 --> 00:07:00.000
this code

00:07:00.000 --> 00:07:01.840
and I had this great idea for using the

00:07:01.840 --> 00:07:03.360
radari 2 framework

00:07:03.360 --> 00:07:05.120
and adding analysis and disassembly

00:07:05.120 --> 00:07:06.479
plug-in for it

00:07:06.479 --> 00:07:08.400
so I looked into this found okay here

00:07:08.400 --> 00:07:10.319
where you can write plugins in c

00:07:10.319 --> 00:07:12.160
but also in python so I wrote one in

00:07:12.160 --> 00:07:13.520
python and then the scout there's

00:07:13.520 --> 00:07:15.039
actually existing one in core which you

00:07:15.039 --> 00:07:17.440
have to enable explicitly by passing its

00:07:17.440 --> 00:07:19.599
argument so I've tried it and found it's

00:07:19.599 --> 00:07:21.840
not exactly as good as my own one so

00:07:21.840 --> 00:07:23.680
improved this one and submitted pull

00:07:23.680 --> 00:07:26.610
requests until it was at the same level

00:07:26.610 --> 00:07:28.080
[Music]

00:07:28.080 --> 00:07:30.160
rendering was the trickiest part of this

00:07:30.160 --> 00:07:31.360
whole thing because

00:07:31.360 --> 00:07:34.319
well I decided against using a library

00:07:34.319 --> 00:07:35.759
not like there would have been any

00:07:35.759 --> 00:07:37.120
usable library for this

00:07:37.120 --> 00:07:39.599
my usual approach of accelerating svg

00:07:39.599 --> 00:07:41.680
file was too expensive it just created

00:07:41.680 --> 00:07:45.120
too much garbage and took too long time

00:07:45.120 --> 00:07:47.360
I then tried creating mutating strings

00:07:47.360 --> 00:07:49.599
this was either either too expensive

00:07:49.599 --> 00:07:52.639
just like svgs or too complicated I

00:07:52.639 --> 00:07:55.039
tried changing svg tiles which created

00:07:55.039 --> 00:07:57.280
gaps between the lines

00:07:57.280 --> 00:07:59.520
then I tried to create an xpm file which

00:07:59.520 --> 00:08:00.720
was backed by a bull vector

00:08:00.720 --> 00:08:02.400
administrating this bull vector

00:08:02.400 --> 00:08:04.400
but the image caching effect made it

00:08:04.400 --> 00:08:06.879
just every nth frame to appear which

00:08:06.879 --> 00:08:10.000
wasn't good either then I had the idea

00:08:10.000 --> 00:08:11.440
to just use plain text

00:08:11.440 --> 00:08:13.280
and paint the individual characters with

00:08:13.280 --> 00:08:14.800
a different background color this

00:08:14.800 --> 00:08:17.120
this had perfect perfect performance

00:08:17.120 --> 00:08:18.479
there were many optimization attempts

00:08:18.479 --> 00:08:20.000
until I got there and it was

00:08:20.000 --> 00:08:21.840
very very stressful I wasn't sure

00:08:21.840 --> 00:08:23.199
whether I would ever get to accept the

00:08:23.199 --> 00:08:26.160
performance at all

00:08:26.160 --> 00:08:28.560
for sound you only need to a single beep

00:08:28.560 --> 00:08:30.319
so technically it shouldn't be difficult

00:08:30.319 --> 00:08:31.280
to emulate it

00:08:31.280 --> 00:08:33.519
however doing this is hard because Emacs

00:08:33.519 --> 00:08:34.880
officially only supports synchronous

00:08:34.880 --> 00:08:37.200
playback of sounds

00:08:37.200 --> 00:08:39.039
but there's also emax process which you

00:08:39.039 --> 00:08:41.360
can launch in asynchronous way

00:08:41.360 --> 00:08:43.519
so I looked into it and found that

00:08:43.519 --> 00:08:45.279
employee has a slave mode and mpv

00:08:45.279 --> 00:08:46.640
supports listing on the

00:08:46.640 --> 00:08:50.880
fifo for commands so I've created a pipe

00:08:50.880 --> 00:08:54.000
started a past mpv in loop mode and

00:08:54.000 --> 00:08:55.519
always send in pause and pause command

00:08:55.519 --> 00:08:58.000
to the fifo and that way I could control

00:08:58.000 --> 00:09:02.640
when to start beeping and stop beeping

00:09:02.640 --> 00:09:05.760
so yeah that's it so far was a very

00:09:05.760 --> 00:09:07.200
educational experience

00:09:07.200 --> 00:09:09.279
I have tried out a bunch of games which

00:09:09.279 --> 00:09:10.320
were

00:09:10.320 --> 00:09:12.640
well I almost say the worst ports of

00:09:12.640 --> 00:09:14.320
classic games I've ever tried

00:09:14.320 --> 00:09:15.839
it wasn't terribly fun to play them but

00:09:15.839 --> 00:09:18.320
was fun to improve the emulator until

00:09:18.320 --> 00:09:21.760
well things worked good enough

00:09:21.760 --> 00:09:23.279
and I've learned a lot about how

00:09:23.279 --> 00:09:25.120
computers work at this level

00:09:25.120 --> 00:09:27.760
so maybe maybe I'll in the future make

00:09:27.760 --> 00:09:28.880
another emulator

00:09:28.880 --> 00:09:31.920
but uh I'm not sure whether anything

00:09:31.920 --> 00:09:34.000
more advanced like intel 8080 emulator

00:09:34.000 --> 00:09:36.560
will actually run mix fast enough

00:09:36.560 --> 00:09:37.839
but it's still an interesting idea

00:09:37.839 --> 00:09:39.200
because then you could actually have an

00:09:39.200 --> 00:09:39.600
os

00:09:39.600 --> 00:09:41.680
inside Emacs and fulfill that one

00:09:41.680 --> 00:09:43.120
specific meme

00:09:43.120 --> 00:09:45.440
but if I try to do most serious stuff

00:09:45.440 --> 00:09:47.279
I'll probably use chicken scheme which

00:09:47.279 --> 00:09:48.000
is my

00:09:48.000 --> 00:09:49.920
preferred language for serious projects

00:09:49.920 --> 00:09:53.279
and writing neso gamer emulator

00:09:53.279 --> 00:09:57.839
and that's it thank you