> What results does the Efika get with lame aksack.wav?

https://amigaworld.net/modules/newbb/viewtopic.php?topic_id=37054&forum=45#693183
http://documentation.aros-platform.de/html/benchmarks.html

Quote:

That won't help with a buggy and slow CPU emulation though.

Quote:

Call me stupid but which figure am I supposed to be looking at to compare it to this?

http://sintonen.fi/pics/lame_benchmark.png

Quote:

Better than nothing at all though and it can only improve.

Quote:

I guess it's encoding time in seconds. Efika does the job in 2 minutes and 20 seconds (140 s). PowerMac G5 2,7 GHz encodes this wav in 8 seconds. Now for the screenshot you posted the situation is interesting. The real time differs from estimated by a very high margin (and on a real machine - it's never this way, as the "real" equals to "estimated" ultimately). I guess something isn't right with the timer under emulation. I'd probably go for the ETA number and a stopwatch just to be sure.

>>> What results does the Efika get with lame aksack.wav?

>> https://amigaworld.net/modules/newbb/viewtopic.php?topic_id=37054&forum=45#693183
>> http://documentation.aros-platform.de/html/benchmarks.html

> which figure am I supposed to be looking at to compare it to this?
> http://sintonen.fi/pics/lame_benchmark.png

140 seconds.

>> http://sintonen.fi/pics/lame_benchmark.png

> PowerMac G5 2,7 GHz encodes this wav in 8 seconds.

Regression with newer lame version? The graph shows 8 seconds for the 2.0 GHz PowerMac G5 and 6 seconds for the 2.5 GHz PowerMac G5.

Quote:

Hmm... I looked at the numbers at AROS page. I guess it's time to do a final benchmark on my G5 machine to dispel the doubts.

guys is not simple do a benchmark because the lan dont work ... i will try to make an iso with some programs for morphos and mount this one for make something extra installed.

The Qemu is not the best in fpu but is good in integer .

qemu threadripper vs qemu kvm on quad g5 benchmark

https://scontent.fnap1-1.fna.fbcdn.net/v/t1.0-9/27540591_10211001296249952_3934778983080779619_n.jpg?_nc_cat=0&oh=523c1f6cddb62032b5879224613d210d&oe=5BBE5122

>>> PowerMac G5 2,7 GHz encodes this wav in 8 seconds.

>> Regression with newer lame version? The graph shows 8 seconds for
>> the 2.0 GHz PowerMac G5 and 6 seconds for the 2.5 GHz PowerMac G5.

> Hmm... I looked at the numbers at AROS page.

It's 7 seconds there ;-) I assume there have been different lame versions/binaries at play (apart from AltiVec vs. scalar versions).

Andreas can you please link your lame version in case i will download it for make some benchs?

Quote:

I guess this one was used. There are two versions in the archive: lame and lame_vmx (for AltiVec enabled machines).

> Andreas can you please link your lame version [...]?

Sorry, I haven't used any lame version myself, but merely discussed lame benchmark results published by others.

Quote:

I'd call it a lame excuse :rotfl:

Anyone know if KVM on an X1000 would work ?.

Quote:

Quote:

Not yet.

Quote:

It took around a minute roughly.

Will run it again to make sure and on much faster hardware too.

Quote:

Just boot from CD, go to prefs, network, add DNS addresses and save.

Download whatever you need to your work partition.

>> Anyone know if KVM on an X1000 would work ?.

> Not yet.

http://forum.hyperion-entertainment.biz/viewtopic.php?p=27150#p27150
http://forum.hyperion-entertainment.biz/viewtopic.php?p=29680#p29680

What am I missing?

Quote:

MorphOS and OS4 don't currently boot under virtualisation on a PPC CPU only under emulation.

Quote:

Yet another reason not to emulate the cpu, Jacek.
Power 9...not X64, that would take care of that issue.

Repetitive, I know, and very few have those systems, but it would run legacy PPC code.

>>> I'd be willing to bet that this could be done better on Power9.

>> QEMU-KVM doesn't work as yet so no running this on any PPC
>> or Power9 without emulating it same as on x64 for now.

> Power 9...not X64, that would take care of that issue. ;-)
> [...] it would run legacy PPC code.

Would? Does? Will? I'm confused :-)

Quote:

Again, not just yet. But soon KVM should be a possible option.
Interesting that I'd never considered the X1000, but that could serve as well.

What about the final G5s? I know the 970FX doesn't have a hypervisor (apparently disabled from the Power 4 subset it was taken from), but QEMU-KVM already appears to be working on these.

BTW - The low performance of PPC emulation on X64 doesn't surprise my, Mark told me it was borderline months ago and he wasn't sure if PPC emulation would make it into NG. Apparently primarily due to an issue related to available resources.

The question becomes what do we want to do with emulation, because it won't be supported outside MorphOS.
Then again, Nik's results are pretty remarkable, so maybe this is something the community can do on its own.

KVM will work for sure for Morphos vmachine and better than emulated one because the emulated cpu use emulated altivec instruction thing that for my experience is not good working because make in idle the vcpu.

for use the morphos in kvm do just sudo modprobe kvm and add --enable-kvm on Qemu command line.

I dint have time to test myself because i dont have a monitor attached on my g5 quad, this days summer make this machine unusable because add more celsious in the home that is warm in the summer period (two days ago was 42° external and 30° in the home)

PS: X1000 and G5,G4 are kvm compatible for Mac99 vhw

> What about the final G5s? I know the 970FX doesn't have a hypervisor
> (apparently disabled from the Power 4 subset it was taken from)

https://morph.zone/modules/newbb_plus/viewtopic.php?topic_id=9326&forum=11&start=16
https://morph.zone/modules/newbb_plus/viewtopic.php?topic_id=11993&forum=53&start=10

> QEMU-KVM already appears to be working on these.

Yes, that's because KVM-PR doesn't need a hypervisor to be present (or enabled) in the CPU.