Login
Username:

Password:

Remember me



Lost Password?

Register now!

Sections

Who's Online
50 user(s) are online (21 user(s) are browsing Forums)

Members: 0
Guests: 50

more...

Support us!

Headlines

 
  Register To Post  

« 1 ... 51 52 53 (54) 55 56 57 ... 75 »
Re: What the fastest possible x64 emulation way of OS4 today ?
Quite a regular
Quite a regular


See User information
@Maijestro
Yes, until you find a way to reproduce it it's hard to debug and find anything. Also I'm not sure it's not something in AmigaOS as QEMU should not behave differently between runs so it must be some difference in the guest. Maybe it's related to the order things start which I think is parallel and can be different in each boot but maybe there's some dependency which is not handled.

I've noticed that if I boot AmigaOS and start a new shell before the Dock, SFS showed up and the startup sound finished playing then programs started from the shell will crash (even if I start a new shell) but if I wait for it to fully boot I never had that problem so not sure this isn't something in AmigaOS rather than in QEMU.

Go to top
Re: What the fastest possible x64 emulation way of OS4 today ?
Quite a regular
Quite a regular


See User information
@balaton

I was wondering if these options could help in "sm501"
I see they are disabled

here the link:
https://manpages.ubuntu.com/manpages/f ... man4/siliconmotion.4.html

Option "PciBurst" "boolean"
will enable PCI burst mode. This should work on all but a few broken PCI chipsets,
and will increase performance. Default: on.

Option "PciRetry" "boolean"
will allow the driver to rely on PCI Retry to program the registers. PciBurst must
be enabled for this to work. This will increase performance, especially for small ones
fills/blits, because the driver does not have to poll the card before sending it
commands to make sure it is ready. It should work on most recent PCI chipsets.
Default: value of PciBurst option.


I tried to enable them and recompile qemu
It doesn't give me any errors but I don't understand if they are useful for the purpose
the "boot" happens without problems.

Go to top
Re: What the fastest possible x64 emulation way of OS4 today ?
Quite a regular
Quite a regular


See User information
@white
Don't know what are you talking about. These are options to the Xorg driver for Linux and have nothing to do with QEMU or AmigaOS nor they would likely have any effect on QEMU even if you enable it in a Linux guest running on an emulated machine. So I don't think these are useful at all.

Go to top
Re: What the fastest possible x64 emulation way of OS4 today ?
Quite a regular
Quite a regular


See User information
@balaton
ok, thanks for the reply.

Go to top
Re: What the fastest possible x64 emulation way of OS4 today ?
Just can't stay away
Just can't stay away


See User information
@balatonQuote:
balaton wrote:@Maijestro
I've noticed that if I boot AmigaOS and start a new shell before the Dock, SFS showed up and the startup sound finished playing then programs started from the shell will crash (even if I start a new shell) but if I wait for it to fully boot I never had that problem so not sure this isn't something in AmigaOS rather than in QEMU.


Tested not reproducible, system remains stable.

MacStudio ARM M1 Max Qemu//Pegasos2 AmigaOs4.1 FE / AmigaOne A1222plus AmigaOs4.1 FE
Go to top
Re: What the fastest possible x64 emulation way of OS4 today ?
Quite a regular
Quite a regular


See User information
@Maijestro
Maybe some updates solved it, I saw this with 4.1FE without updates juat booted from the CD but could be I had some wrong Kickstart module versions mixed in somewhere. At least this isn't a problem for others then.

Go to top
Re: What the fastest possible x64 emulation way of OS4 today ?
Just can't stay away
Just can't stay away


See User information
@balatonQuote:
balaton wrote:@Maijestro
Maybe some updates solved it, I saw this with 4.1FE without updates juat booted from the CD but could be I had some wrong Kickstart module versions mixed in somewhere. At least this isn't a problem for others then.


You should keep your system as up to date as possible and test with it, if it is not possible for you I am happy to get you an AmigaOs4.1 Pegasos 2 license that only applies to you, it is the least I can do for your great work.

Edit: In update 1 and 2 there were very big changes concerning AmigaOs4.1 and testing on a system which is not up to date will cause problems sooner or later because the differences are just too big.

MacStudio ARM M1 Max Qemu//Pegasos2 AmigaOs4.1 FE / AmigaOne A1222plus AmigaOs4.1 FE
Go to top
Re: What the fastest possible x64 emulation way of OS4 today ?
Quite a regular
Quite a regular


See User information
@Maijestro
Exactly.

@balaton it should work with a valid licensed system
and with a "hd" already configured well with that license.
to be able to test the various solutions.

he's obviously busy with other things.

but having a "hd" configured with all the applications configured is ready to use.
It would help him to understand the various problems better.

Which I have always said.

Go to top
Re: What the fastest possible x64 emulation way of OS4 today ?
Just can't stay away
Just can't stay away


See User information
@white

Can you run some FPS tests and maybe confirm that there is a performance drop within Qemu?

It is best to start 10 sessions and change the cpu each time and compare FPS tests.

-cpu 750cxe and -cpu7447

Then please post your results.

MacStudio ARM M1 Max Qemu//Pegasos2 AmigaOs4.1 FE / AmigaOne A1222plus AmigaOs4.1 FE
Go to top
Re: What the fastest possible x64 emulation way of OS4 today ?
Quite a regular
Quite a regular


See User information
@Maijestro

Yes, there is a performance hit
but only in test programs (but which have no meaning)
because those tests are not performed on "real machines"

but it's actually much faster using it in "linux" practice
E-UAE is much faster
Grim Fandango 5-6 fps more for example
Odyssey is much faster

for example in linux the "cursor" seems to move a little slower sometimes it is imperceptible but it does
in comparison to everything else that fluid windows applications etc. at least on my setup (linux).

instead in the version of qemu for "Windows"
speed increased practically double with "8.1.0-rc0"
it is almost the same as the linux version.

Go to top
Re: What the fastest possible x64 emulation way of OS4 today ?
Just can't stay away
Just can't stay away


See User information
@Maijestro
Quote:
It is best to start 10 sessions and change the cpu each time and compare FPS tests.

-cpu 750cxe and -cpu7447
It's to be expected that you get different results with 750CXe and 7447 CPUs, on real hardware the 7447 is much faster.
Emulation may be very different, but it still should result in different speeds/FPS. On emulation a 750CXe might even be faster than a 7447, but only for software not using AltiVec code.

Go to top
Re: What the fastest possible x64 emulation way of OS4 today ?
Just can't stay away
Just can't stay away


See User information
@joergQuote:
joerg wrote:@Maijestro
Quote:
It is best to start 10 sessions and change the cpu each time and compare FPS tests.

-cpu 750cxe and -cpu7447
It's to be expected that you get different results with 750CXe and 7447 CPUs, on real hardware the 7447 is much faster.
Emulation may be very different, but it still should result in different speeds/FPS. On emulation a 750CXe might even be faster than a 7447, but only for software not using AltiVec code.


The difference under Qemu doesn't matter with both CPUs, it's similarly fast. I have already tested it extensively, this only applies to real hardware. The only difference to the cpu 750CXe is that AmigaOs4.1 with CPU 7447 recognizes altivec and makes it usable.

CPU 750CXe Quake Timedemo 28.7 FPS on a good session.

CPU 7447 Quake Timedemo 32 FPS good session.

Since the performance drop with CPU 7447 and also CPU 750cxe is around 50% it seems to be a problem within Qemu.

If I switch back and forth between the two CPUs, I can reproduce the problem under Qemu, which has more power with the CPU 750CXe and also the CPU 7447. Since similar values are then achieved with both CPUs under Quake Timedemo.

So this should be checked under Qemu.

MacStudio ARM M1 Max Qemu//Pegasos2 AmigaOs4.1 FE / AmigaOne A1222plus AmigaOs4.1 FE
Go to top
Re: What the fastest possible x64 emulation way of OS4 today ?
Quite a regular
Quite a regular


See User information
@Maijestro

here for example qemu for Windows ("8.1.0 rc0")
750CXe CPU

taking into account that in the meantime the Browser on Windows is also open is the msi burning program for recording and 2-3 other programs.
You have to take into account that there are also programs that are obviously in the background.
therefore tests can be biased.

https://youtu.be/hSx_S8_T6Io

Go to top
Re: What the fastest possible x64 emulation way of OS4 today ?
Quite a regular
Quite a regular


See User information
@Maijestro
but I understand what you mean
test programs before were higher that's true.
but the speed actually increased in reality now

Go to top
Re: What the fastest possible x64 emulation way of OS4 today ?
Quite a regular
Quite a regular


See User information
@Maijestro
Quote:
So this should be checked under Qemu.


What to check though? If we can't reproduce it and also don't know what's different between the cases when it's fast vs. when it's slow then what should we look for? We would need something to be able to find anything. So you would try to find out what's different between the two cases, if it can be also reproduced with other guest OSes (Linux, MorphOS) or something similar to be able to get an idea what could cause it. Otherwise we can only guess and have to be very lucky to find a problem that way.

AS this seems to be dependent on some changes between boots maybe it's related to memory layout but I still can't find out what could be different. Try at least gather some info from QEMU monitor about the two cases such as info jit info irq and see if there are some differences there or if you can get a memory map or other statisctics from AmigaOS that can be compared then maybe some difference can be identified from those. I can't help with debugging these but I also can't find and fix anyhing in QEMU until the issue is identified a bit more exaclty.

Go to top
Re: What the fastest possible x64 emulation way of OS4 today ?
Not too shy to talk
Not too shy to talk


See User information
I'm just catching up after the weekend. Going back to the last thing that I was doing it looks like I need to experiment in bboot with what we are patching and what we are leaving alone. At the moment I'm struggling to get a working PPC cros compiler for the distro I picked (Arch linux). It doesn't have any pre-built packages that I can see and it's failing to build from source.

@balaton - to make life easier, what distro are you running?


Amiga x5040 ı 16GB ı RX580
GB-A1000 060@100,
A1200 PiStorm32-Lite CM4
Go to top
Re: What the fastest possible x64 emulation way of OS4 today ?
Just can't stay away
Just can't stay away


See User information
@balaton

Quote:
balaton wrote:@Maijestro
Quote:
So this should be checked under Qemu.

AS this seems to be dependent on some changes between boots maybe it's related to memory layout but I still can't find out what could be different. Try at least gather some info from QEMU monitor about the two cases such as info jit info irq and see if there are some differences there or if you can get a memory map or other statisctics from AmigaOS that can be compared then maybe some difference can be identified from those. I can't help with debugging these but I also can't find and fix anyhing in QEMU until the issue is identified a bit more exaclty.


How do I use "info jit info irq" ?

I could examine both sessions where it runs sometimes faster and sometimes slower with the same CPU. That there is a big performance drop is already confirmed by my video also other benchmark tests inside AmigaOs4.1 show this behavior. It occurs with all CPU's that are available to us.

As already mentioned there are no other programs running on the host system that could affect the CPU performance.

It doesn't matter if Peg2.rom or BBoot is used I've observed this problem since Qemu 7.2 when I started to do benchmarks.

I would like to investigate this problem to get a more stable and clean Qemu emulation. I just need some help on what to do with Qemu to see the differences.

MacStudio ARM M1 Max Qemu//Pegasos2 AmigaOs4.1 FE / AmigaOne A1222plus AmigaOs4.1 FE
Go to top
Re: What the fastest possible x64 emulation way of OS4 today ?
Quite a regular
Quite a regular


See User information
@balaton

Rebuild QEMU 8.1.0 rc0 without patches.
Used bboot 0.4
This is what I get with a RX560

qemu-system-ppc: ../hw/pci/pcie.c:927pcie_find_capability_listAssertion `next <= PCIE_CONFIG_SPACE_SIZE - 8' failed.
Aborted (core dumped)

Go to top
Re: What the fastest possible x64 emulation way of OS4 today ?
Quite a regular
Quite a regular


See User information
@MartinW
Both RedHat and Debian have ppc cross compiler so any distro based on those would also likely have it. Arch only seems to have binutils for ppc, installing that is one prerequisite but gcc has some others then it takes a while to compile. There are some scripts to build a cross compiler that you could seearch for.

Maybe even the compiler included in the AmigaOS SDK should work so if you have that (even on AmigaOS) you could try. Only other tool used is make so that should be available anywhere.

Go to top
Re: What the fastest possible x64 emulation way of OS4 today ?
Quite a regular
Quite a regular


See User information
@Maijestro
info irq and info jit are QEMU monitor commands that print statistics about how many interrupts or events during jit happened that could impact performance. You should do about the same thing once when it's fast and once when it's slow and after the same code has been run in the guest check the stats to see if there's a difference that could show where the slowness could come from.

Go to top

  Register To Post
« 1 ... 51 52 53 (54) 55 56 57 ... 75 »

 




Currently Active Users Viewing This Thread: 1 ( 0 members and 1 Anonymous Users )




Powered by XOOPS 2.0 © 2001-2023 The XOOPS Project