VOGONS


First post, by kjliew

User metadata
Rank Oldbie
Rank
Oldbie
wrote:

/EDIT by DOSFREAK DELETED non-related information

It is related as a comparison of dynamic recompilers. QEMU TCG is AArch64/ARMv8 portable.
I hate to say this, but you really acted in non-professional for academical and technical assessments.
What's wrong with assessment at comparing performance of different dynamic recompilers on Apple Silicon?

Edit by Dominus: split off from DOSBox SVN - Qemu on M1 benchmark comparison discussion because let's not derail the original thread too much

Last edited by Dominus on 2021-04-07, 19:27. Edited 1 time in total.

Reply 1 of 8, by DosFreak

User metadata
Rank l33t++
Rank
l33t++

The thread topic is "DOSBox SVN for macOS - custom patched version". This thread is not about you or QEMU.
As I've stated many times before, post your benchmarks in a dedicated thread. Throw the link in your sig and call it a day.

DOSBox Compilation Guides
DosBox Feature Request Thread
PC Game Compatibility List
How To Ask Questions The Smart Way
Running DRM games offline

Reply 3 of 8, by DosFreak

User metadata
Rank l33t++
Rank
l33t++
kjliew wrote on 2021-04-07, 19:08:

QEMU TCG running DOS PCPBENCH does not need any customized build.
Anyone can just download QEMU from anywhere for one's native OS, including macOS on M1.

Wonderful. More QEMU advertisement in an unrelated thread.

DOSBox Compilation Guides
DosBox Feature Request Thread
PC Game Compatibility List
How To Ask Questions The Smart Way
Running DRM games offline

Reply 4 of 8, by kjliew

User metadata
Rank Oldbie
Rank
Oldbie

I would rather call it as "knowledge sharing".
Only those who felt "... rubbing something to one's face ..." will take it as a form of advertisement.
No one ever needs to pay anyone for it, but again it is just as benchmark comparison.

Reply 5 of 8, by kjliew

User metadata
Rank Oldbie
Rank
Oldbie

Apple M1 Macbook Air 7-core GPU SKU
PCPBENCH 4101 640x4808bpp
QEMU 6.0.0 TCG - 315.4 fps

Screen Shot 2021-05-15 at 3.09.23 PM.png
Filename
Screen Shot 2021-05-15 at 3.09.23 PM.png
File size
257.39 KiB
Views
299 views
File comment
PCPBENCH QEMU Apple M1
File license
Fair use/fair dealing exception

MDK DOS Performance

Screen Shot 2021-05-15 at 3.10.35 PM.png
Filename
Screen Shot 2021-05-15 at 3.10.35 PM.png
File size
228.05 KiB
Views
299 views
File comment
MDK Perf DOS QEMU Apple M1
File license
Fair use/fair dealing exception

Notice the "X" logo on the screenshot 😁 Yes, it is XQuartz/SDL2. 😁
Though the TCG performance is so great (unseen on any prior ARM implementation), it still pales in comparison with any form the hardware virtualization support from x86 CPUs. Next is to start working on MESA GL/3Dfx Glide over XQuartz/OpenGL. I already got OpenGlide/XQuartz compiled without SDL1.2 but there are still bugs to be resolved.

Stay tuned!

Reply 6 of 8, by digger

User metadata
Rank Oldbie
Rank
Oldbie

That's pretty impressive performance! Those M1 chips really are amazing, even when passively cooled (!). Think of Apple what you will, but they seem to be single-handedly breaking the decades-long stranglehold that the x86 architecture has had on the desktop/laptop market. (Yes, I know, it's their own special flavor of ARM that has made this level of emulation performance possible, but still.) Apple became boring and started to be held back when they switched to Intel. It's good to see them innovating again.

Don't take me wrong, IMO Apple still still sucks for opposing Right-to-Repair, but I'm giving credit where credit is due.

And speaking of due credit, good job once again, kjliew! 🙂

Reply 7 of 8, by digger

User metadata
Rank Oldbie
Rank
Oldbie

Sure, of course emulation (even TCG) doesn't hold a candle to hardware-assisted virtualization, but my bigger point was that x86 won't reign supreme forever, so it's good to have reasonably performant emulation solutions on different architectures as well. As a matter of fact, that is exactly what will help us migrate to newer and arguably more superior architectures.

By the way, do any of you know if anybody is working on optimizing TCG performance for the (64-bit) RISC-V architecture as well? I found this document by Western Digital on the topic, which clarifies that TCG is indeed available for RISC-V hosts as well, but I wonder how well-optimized it currently is for x86 guests.

Reply 8 of 8, by Bruninho

User metadata
Rank Oldbie
Rank
Oldbie

What the hell... I knew it... The M1's are the great deal.

"Design isn't just what it looks like and feels like. Design is how it works."
JOBS, Steve.

List of ALL Android vulnerabilities