SMSQmulator and Raspberry Pi 2

[stevepoole]

Hi Derek,

According to Wikipedia, the 'PLAN 9' OS for the PI, is 68000 series compatible...

Has anybody tried it ?
_______________________

[Derek_Stewart]

I like the the film?...

[pjw]

I dug out and booted up my old Raspberry PI model B+ for the first time in
9 or 10 years, to check whether Id remembered correctly in my reply some
posts back. Although the original questioner doesnt seem interested in the
answer, I didnt want to mislead any casual lurkers. So this is what I
found:

I dont know what a PI 2 is, but the B+ is a step up from the PI model A
(more memory, Ethernet, etc) but it uses the same single core Arm processor
as its predecessor, so it aint any faster. And speed is essential for
emulation!

My B+ installation was a huge mess and barely functional, thanks to
extensive experimentation with my first PI. I saw some half-hearted
attempts to install uQLx on it. I cant remember whether I succeeded in the
end (I have successfully compiled and installed it on other computers in
the past), but clearly, I did not think that RasPI B+ (and therefore any
model that went before it) was the way to go for me. It certainly wont
run SMSQmulator!

Enter the Raspberry PI 4 model B. This was my next PI, of about 2 years
ago, and a considerable step up. It has now been put to other use so I no
longer have access to the QL emulator installation I used, but this is the
one I was remembering when I hinted that SMSQmulator was usable on a PI
(see my post on it here: https://www.qlforum.co.uk/viewtopic.php?p=48565).
Im sure uQLx, sQLux and sQ68ux would do very well on it. It has plenty of
RAM, USB2 & 3, 1Gb/s ethernet, WiFi, Bluetooth, etc.

Once youre willing to consider a Raspberry PI 5, you have a lot of
alternative options. There are miniature computers as, or more powerful, at
similar prices and with similar or better specs, including some that run
versions of Window and could therefore, perhaps, run QPC2. However, I like
the Raspberry project with Open Source software etc. It seem less, er,
mercenary, so Ill stick with that for now.

As Ive banged on often enough here, I am currently using a PI5/SMSQmulator,
Logitec wireless keyboard and mouse, HDMI at 960x540 hicolor fullscreen.
As far as QL emulators go, it is perfectly adequate and could be used for real
QL stuff. Im not quite ready to jump ship yet, due to the host OS (Linux). If
something where to go wrong with the installation, I could waste hours
trying to fix it with no guarantee of success. Then there are those system
keypresses that, if you hit one by mistake, could kill your emulation or lock
you out of it. I use the PI5 for other things too, so am getting deeper into
Linux all the time. With more practise, perhaps..

[XorA]

For anyone slightly disheartened by the last post, uqlx, sQLux, and most likely sQ68ux (I have not specifically tested) work really well on old Pi machines. I did a lot of testing for sQLux on the original Pi0 which was a Pi1 squished a bit!

[pjw]

For anyone slightly disheartened by the last post, uqlx, sQLux, and most likely sQ68ux (I have not specifically tested) work really well on old Pi machines. I did a lot of testing for sQLux on the original Pi0 which was a Pi1 squished a bit!

Hi XorA,
It was not my intention to dishearten anyone. As anyone who has taken any notice of my opinions knows, since I first got my QL Ive been pushing for it to go faster, with more storage, more pixels, more colours, .. So when on trying something new I find it not much better in that respect than what went before, I may have felt a bit disheartened myself. However, I assumed that my opinions wouldnt affect the many retro-enthusiasts here, nor were they the intended audience for my reply.

I think the work youre doing is great, necessary and important, as is that of the hardware guys trying to improve or build a better "QL". Who knows what the future will bring? One thing is certain: if no seeds are planted, nothing will grow.

[tofro]

The point is simply: Modern, fast and SMSQ/E-based emulators (like QPC2, for example) tend to spoil the user base with speed (nothing wrong with that, just a fact. I sometimes feel spoilt as well ).

The original QL isn't exactly a fast machine, and even if an emulated one runs multiple times as fast on a RPI, QPC2 on a fast PC or Mac, for example, is just miles (rather: lightyears) away in speed and responsiveness and the RPi might actually feel slow.

[Derek_Stewart]

Hi,

To use QPC2 on a RPI needs an ARM to x86 emulator and then WINE.

Or just use sQLux, which runs Minerva very quickly on my Pi400.

sQ68ux runs nicely with SMSQ/E v3.41 from an SD Card image , but needs some development from more than one developers. I like to try but I am just an enthuistisic user.

Alternately install openjdk and run SMSQmulator. Runs okay in a RPI3B, but a little slow. On a Pi4 runs faster, on Pi5 no idea as I do not have one.

[XorA]

sQ68ux runs nicely with SMSQ/E v3.41 from an SD Card image , but needs some development from more than one developers. I like to try but I am just an enthuistisic user.

Yeah Im stuck on enableing write for SD, it crashes horribly at moment!

[pjw]

Benchmarks are such fickle tests they have to be taken with a ounce of salt. Usually the first thing I try on a newly commissioned system to give me a rough idea of how fast it is, is the following test:

Code: Select all

100 t = DATE
110 FOR i = 0 TO 99999
120  AT 0,0: PRINT i,
130 END FOR i
140 PRINT\\ DATE - t

Applied to some of my systems over the past two years, this gives the following:

All tests ran in full screen mode with 16-bit colours, and except where
obvious or stated, with SMSQmulator V2.33.

Code: Select all

RasPiB4,        1920x1080,  35
Q68,            1024x512,  185
Labda *1        960x540,    10
Theta *2        1920x1080,  20
QPC2 Theta      1920x1080,   5
RasPI5           960x540,    8(!) SMSQmulator V3.01, SMSQ/E V3.41

*1) Labda: i5-8250U, nominal 1.60GHz, running at ~3.30GHz, W11 Home
*2) Theta: i5-7500, nominal 3.40GHz, running at ~3.70GHz, W10 Pro

I was surprised that the Labda test ran so much faster than Theta. Theta is
rated the faster machine in raw power. Screen size? CPU optimisation from
G7 to G8?

Anyway, as seems to be indicated: The RPi5 is pretty damn fast!

For those who prefer a slightly more formal picture (albeit no more accurate):

Code: Select all

C68 Dhrystone 2.1,  GCC Dhrystone 2.1,   Bogomips 1.4; 1000,000 iterations
	C68                          	GCC
Dhr/s		MIPS		Dhr/s		MIPS		BogoMips
 83,333.3   	 47.429		109,890.1	 62.544		 71.58

RPi5 Quad-core Arm Cortex-A76 @ 2.4GHz
SMSQmulator V3.01, SMSQ-E V3.41, 16 bit 960x540 full screen

PS: One explanation for anomalies re subjective speed comparisons that dont appear to match the nominal CPU throughput may have to do with the emulators taking advantage of certain host OSes facilities, such as storage device drivers, screen acceleration and maths co-processors. For example, I notice that SMSQmulator has improved in various areas over the years (making some of the timings above moot) beyond what might be expected of a mere improvement in coding efficiency..

[pjw]

I know no one asked for this, and youre probably bored with it now, but I re-did some of the earlier tests, where possible:

Code: Select all

All tests ran in full screen mode with 16-bit colours, except where
obvious or stated.
1. SMSQmulator V2.33 and/or SMSQ/E V3.32
2. SMSQmulator V3.01 and/or SMSQ/E V3.41

                       1                2
==============  ==============   ==============  =======================
RasPiB4,        1920x1080,  35         ---
Q68,            1024x512,  185              193    (DISP_MODE 3)
Q68,                  ---        1024x768   103    (DISP_MODE 4)
Labda *1         960x540,   10    960x540    10
Theta *2        1920x1080   20    960x540     9
RasPI5                ---         960x540     8

Theta QPC2 4.05 1920x1080,   5        ---
Theta QPC2 5.01       ---         960x540,    4.5

*1) Labda: i5-8250U, nominal 1.60GHz, running at ~3.30GHz, W11 Home
*2) Theta: i5-7500, nominal 3.40GHz, running at ~3.70GHz, W10 Pro

Of course we're not comparing like with like here. There are different,
unknown, processes going on the in the host system and in the emulator.
Emulators have been configured, which has not been accounted for.
Some tests were conducted in significantly different environments.
The PICNIC issue..

All those caveats considered, I still think this test gives an idea of
comparative raw speed of the emulator engine.