Mac OS X 10.3 issues

About QEMU, a PPC Mac emulator for Windows, macOS and Linux that can run Mac OS 9.0 up to Mac OS X 10.5

Moderators: Cat_7, Ronald P. Regensburg

Post Reply
wavyinterlude
Space Cadet
Posts: 2
Joined: Sun Oct 25, 2020 1:34 pm

Mac OS X 10.3 issues

Post by wavyinterlude »

Hi everyone. I just tried installing Panther using QEMU and everything is all set up, but i'm having a huge problem with the graphics? Everything is unbearably slow, I can't find the right resolution, and there's no sound either. It's almost unusable which is really weird.

What should I do? I'm really stuck but if it helps, here's my qemu.command file.

Code: Select all

#!/bin/bash
cd "$(dirname "$0")"

./qemu-system-ppc -L pc-bios -boot c -M mac99,via=pmu -m 512 \
-prom-env 'auto-boot?=true' -prom-env 'boot-args=-v' -prom-env 'vga-ndrv?=true' \
-drive file=Panther_Disc2.toast,format=raw,media=cdrom \
-drive file=MacOSX.img,format=raw,media=disk \
-netdev user,id=network01 -device sungem,netdev=network01 \
-device VGA,edid=on \
-g 1280x768x32
User avatar
Cat_7
Expert User
Posts: 5100
Joined: Fri Feb 13, 2004 8:59 am
Location: Sittard, The Netherlands

Re: Mac OS X 10.3 issues

Post by Cat_7 »

Hi, and welcome,

There is nothing wrong with your command line. Qemu is not fast when emulating OSX. There is no graphic acceleration so things are bound to be slow. The only thing you can do is lower the color depth or resolution, but that will not help much.

To have sound support, you would need to run one of our experimental builds.
However, sound is only supported reasonably well with Mac OS 9.x guests, sound in OSX guest stutters.

So, all in all not much you can do.

Best,
Cat_7
wavyinterlude
Space Cadet
Posts: 2
Joined: Sun Oct 25, 2020 1:34 pm

Re: Mac OS X 10.3 issues

Post by wavyinterlude »

oh alright. it's a huge same. i'm going to look for alternatives; i'm glad i could at least use SheepShaver for OS9. the guide here was really helpful.

thank you.
User avatar
adespoton
Forum All-Star
Posts: 3102
Joined: Fri Nov 27, 2009 5:11 am
Location: Emaculation.com
Contact:

Re: Mac OS X 10.3 issues

Post by adespoton »

wavyinterlude wrote: Sun Oct 25, 2020 10:04 pm oh alright. it's a huge same. i'm going to look for alternatives; i'm glad i could at least use SheepShaver for OS9. the guide here was really helpful.

thank you.
Your only alternative is using PearPC, which doesn't have sound or graphics acceleration, and has known issues due to the shortcuts it takes to improve speed.
mcayland
Granny Smith
Posts: 111
Joined: Sun Nov 01, 2015 10:33 pm

Re: Mac OS X 10.3 issues

Post by mcayland »

adespoton wrote: Mon Oct 26, 2020 4:01 pm Your only alternative is using PearPC, which doesn't have sound or graphics acceleration, and has known issues due to the shortcuts it takes to improve speed.
Just out of curiosity, do you know what these shortcuts are?
User avatar
Cat_7
Expert User
Posts: 5100
Joined: Fri Feb 13, 2004 8:59 am
Location: Sittard, The Netherlands

Re: Mac OS X 10.3 issues

Post by Cat_7 »

At least one of them is to lower the screen refresh rate.

Best,
Cat_7
User avatar
adespoton
Forum All-Star
Posts: 3102
Joined: Fri Nov 27, 2009 5:11 am
Location: Emaculation.com
Contact:

Re: Mac OS X 10.3 issues

Post by adespoton »

Another is vastly simplified floating point emulation. I believe there's also some instructions that always produce dummy results, but it's been years since I've looked at the source.
mcayland
Granny Smith
Posts: 111
Joined: Sun Nov 01, 2015 10:33 pm

Re: Mac OS X 10.3 issues

Post by mcayland »

Does anyone have any numbers for comparison e.g. side-by-side benchmarks run in both QEMU and PearPC? If we can identify the bottlenecks then it may be possible to do something about them.
User avatar
Cat_7
Expert User
Posts: 5100
Joined: Fri Feb 13, 2004 8:59 am
Location: Sittard, The Netherlands

Re: Mac OS X 10.3 issues

Post by Cat_7 »

Some benchmarks from the Skidmarks test. Both PearPC and Qemu use the same 10.3 image, and as PearPC hangs when using the G4 cpu, both tests are with the G3. So no altivec scores. Memory 128Mb. Screen set to 1024x786x32. Screen redraw in PearPC seems faster, but the INT score is substantially lower compared to Qemu, while Qemu scores considerably lower on FPU performance. Changing the screen redraw in PearPC (from 10 to 80, whatever that means) did not change the actual INT and FPU performance.

PearPC general score:
https://surfdrive.surf.nl/files/index.p ... XiFzCXL7pP

PearPC specifics:
https://surfdrive.surf.nl/files/index.p ... Go27Stnmgi

Qemu general score:
https://surfdrive.surf.nl/files/index.p ... G2SsnpbHpS

Qemu specifics:
https://surfdrive.surf.nl/files/index.p ... FsXVt494Uz
mcayland
Granny Smith
Posts: 111
Joined: Sun Nov 01, 2015 10:33 pm

Re: Mac OS X 10.3 issues

Post by mcayland »

That's not bad at all - obviously benchmarks can be a bit artificial, but there are only 2 tests where QEMU is slower than PearPC:

FFT (FP): PearPC 61.4%, QEMU 6.7%
VolInt (FP): PearPC 102.3%, QEMU 14.5%

At some point it would be interesting to see which instructions are being used by the slower tests to see what scope there is for improvement.
User avatar
adespoton
Forum All-Star
Posts: 3102
Joined: Fri Nov 27, 2009 5:11 am
Location: Emaculation.com
Contact:

Re: Mac OS X 10.3 issues

Post by adespoton »

mcayland wrote: Sat Oct 31, 2020 11:54 am That's not bad at all - obviously benchmarks can be a bit artificial, but there are only 2 tests where QEMU is slower than PearPC:

FFT (FP): PearPC 61.4%, QEMU 6.7%
VolInt (FP): PearPC 102.3%, QEMU 14.5%

At some point it would be interesting to see which instructions are being used by the slower tests to see what scope there is for improvement.
Interesting... these are the two areas where we identified slowness in the QEMU FP before.

Cat_7, did you run this against mcayland's modified FPU build, or against the official build?
User avatar
Cat_7
Expert User
Posts: 5100
Joined: Fri Feb 13, 2004 8:59 am
Location: Sittard, The Netherlands

Re: Mac OS X 10.3 issues

Post by Cat_7 »

Against a build from the official code.
The fpu hardfloat patch is not by mcayland, but by zoltan balaton ;-) Soon it will no longer apply to the main code, I had to fix it once already.
Discussion on the qemu developer list suggested it will not be accepted into the main code as it introduces possibly inexact results. So why test it?

Best,
Cat_7
User avatar
Cat_7
Expert User
Posts: 5100
Joined: Fri Feb 13, 2004 8:59 am
Location: Sittard, The Netherlands

Re: Mac OS X 10.3 issues

Post by Cat_7 »

Some scores from xbench derived with the same settings as above.
I have no idea whether this tool does any scaling related to the different machines/bus speeds it appears to detect.
Pearpc scores on left, qemu scores on the right. The various memory, graphics and disk speed tests suggest qemu is lagging there.

Pearpc Qemu
Results 2.98 Results 3.88
System Info
Xbench Version 1.3 Xbench Version 1.3
System Version 10.3.9 (7W98) System Version 10.3.9 (7W98)
Physical RAM 128 MB Physical RAM 128 MB
Model PowerPC (PearPC 0.5) Model PowerMac3,1
Processor PowerPC G4 @ 2.36 GHz Processor PowerPC G3 @ 900 MHz
L1 Cache 32K (instruction), 32K (data) Version Arthur v3.1
Bus Frequency 40 MHz L1 Cache 32K (instruction), 32K (data)
Video Card PearPC,display Bus Frequency 100 MHz
Drive Type EIN GEBUESCH! Drive Type QEMU HARDDISK

CPU Test 0.56 0.87
GCD Loop 234.77 12.38 Mops/sec GCD Loop 166.57 8.78 Mops/sec
Floating Point Basic 2.06 49.06 Mflop/sec Floating Point Basic 1.8 42.87 Mflop/sec
vecLib FFT 0.15 5.00 Mflop/sec vecLib FFT 0.26 8.57 Mflop/sec
Floating Point Library 17.77 3.09 Mops/sec Floating Point Library 5.54 965.13 Kops/sec

Thread Test 30.77 36.23
Computation 132.68 2.69 Mops/sec, 4 threads Computation 85.69 1.74 Mops/sec, 4 threads
Lock Contention 17.4 748.73 Klocks/sec, 4 threads Lock Contention 22.97 988.16 Klocks/sec, 4 threads

Memory Test 47.21 30.64
System 44.98 System 23.59
Allocate 46.74 171.63 Kalloc/sec Allocate 9.78 35.91 Kalloc/sec
Fill 38.56 1874.85 MB/sec Fill 63.63 3093.66 MB/sec
Copy 51.64 1066.50 MB/sec Copy 109.04 2252.20 MB/sec

Stream 49.67 Stream 43.67
Copy 62.25 1285.71 MB/sec Copy 91.02 1879.91 MB/sec
Scale 43.92 907.29 MB/sec Scale 37.62 777.30 MB/sec
Add 49.18 1047.60 MB/sec Add 41.89 892.27 MB/sec
Triad 46.81 1001.46 MB/sec Triad 33.17 709.52 MB/sec

Quartz Graphics Test 24.12 11.31
Line 27.18 1.81 Klines/sec [50% alpha] Line 15.08 1.00 Klines/sec [50% alpha]
Rectangle 17.39 5.19 Krects/sec [50% alpha] Rectangle 7.73 2.31 Krects/sec [50% alpha]
Circle 19.06 1.55 Kcircles/sec [50% alpha] Circle 10.36 844.20 circles/sec [50% alpha]
Bezier 36.97 932.38 beziers/sec [50% alpha] Bezier 14.21 358.43 beziers/sec [50% alpha]
Text 29.84 1.87 Kchars/sec Text 12.61 788.85 chars/sec

OpenGL Graphics Test 3.33 4.86
Spinning Squares 3.33 4.22 frames/sec Spinning Squares 4.86 6.16 frames/sec

User Interface Test 6.2 3.54
Elements 6.2 28.44 refresh/sec Elements 3.54 16.25 refresh/sec

Disk Test 134.63 56.36
Sequential 80.25 Sequential 34.97
Uncached Write 32.94 20.22 MB/sec [4K blocks] Uncached Write 20.42 12.54 MB/sec [4K blocks]
Uncached Write 280.61 158.77 MB/sec [256K blocks] Uncached Write 113.96 64.48 MB/sec [256K blocks]
Uncached Read 78.06 22.85 MB/sec [4K blocks] Uncached Read 20.57 6.02 MB/sec [4K blocks]
Uncached Read 321.67 161.67 MB/sec [256K blocks] Uncached Read 125.06 62.85 MB/sec [256K blocks]

Random 417.57 Random 145.05
Uncached Write 166.46 17.62 MB/sec [4K blocks] Uncached Write 54.36 5.75 MB/sec [4K blocks]
Uncached Write 488.69 156.45 MB/sec [256K blocks] Uncached Write 197.26 63.15 MB/sec [256K blocks]
Uncached Read 2693.29 19.09 MB/sec [4K blocks] Uncached Read 823.49 5.84 MB/sec [4K blocks]
Uncached Read 866.3 160.75 MB/sec [256K blocks] Uncached Read 345.13 64.04 MB/sec [256K blocks]
User avatar
adespoton
Forum All-Star
Posts: 3102
Joined: Fri Nov 27, 2009 5:11 am
Location: Emaculation.com
Contact:

Re: Mac OS X 10.3 issues

Post by adespoton »

Interesting... there's some massive disk I/O differences in there too. Maybe I've avoided this by using raw rdisk pipes in both?
Post Reply