The fastest PC emulator on Earth
#1
Posted 27 January 2007 - 08:33 PM
A friend of mine (josé feiteirinha) has worked intensively on this project for the last two years and now is finally reaching his initially proposed goals and getting support by other developers from the University of Coimbra (Portugal) and other research groups from Europe - they are now reaching 60 MIPS from common desktop machines, a good leap when compared to previous maximum record set on 15 MIPS from similar development clusters.
This link is an extract from the TV interview - unfortunately it's not in english - it narrates the global presentation of this project and overall performance of the emulated machine.
This software is written to run under linux and won't likely be available for open source distribution.
Here's a brief description on (M)IPS:
http://en.wikipedia....ions_per_second
As I'm posting this topic, new available processors are already reaching higher values but it's still fun to think on the possible uses for such technology applied on less "powered" machines and wouldn't mind to see it applied on a tweaked qemu either..
Hope you like it!
#2
Posted 27 January 2007 - 09:19 PM
The only part I understood is the MIPS part towards the end.
Ill take your word that its fast. Now we can have a window emulator that has a linux emulated then run the faster emulator under it.
#3
Posted 27 January 2007 - 09:26 PM
It might just work, who knows?
#4
Posted 28 January 2007 - 05:46 PM
Now we can have a window emulator that has a linux emulated then run the faster emulator under it.
Hmmm:
This software is written to run under linux and won't likely be available for open source distribution.
it seems to me like "Now we are shown that something exists, but we cannot have it"
jaclaz
#5
Posted 28 January 2007 - 06:21 PM
#6
Posted 24 February 2007 - 01:39 AM
Just my opinion, but I doubt they have any interest on releasing the source to public - it won't hurt to ask in any case. Also wonder if the original developer would be interested on tweaking our qemu..
Actually opening the source was our idea from the beginning but ESA's policies didn't allowed it.
However we did wrote a paper on it, and you can found it in (english) at http://eden.dei.uc.p...es_SESP2006.pdf
About the tweaking QEMU's source, I've been around that code a while and can only say that it would take me a long while just to get the hang of it. Mainly that's why I decided to code our emulator from scratch.
But the reason I dropped-by was not just to talk about emulators, but also to say Hi..
Hi Nuno!
Hi Carla!
Hi Everyone!
#7
Posted 24 February 2007 - 12:34 PM
Welcome to the boot-land.net community!
#8
Posted 24 February 2007 - 03:13 PM
Hi Nuno!
Hi Carla!
Hi Everyone!
Hello Zé!
It's nice to see you!
Keep on with the good work!
#9
Posted 04 March 2008 - 04:07 AM
Actually opening the source was our idea from the beginning but ESA's policies didn't allowed it.
However we did wrote a paper on it, and you can found it in (english) at <a href="http://eden.dei.uc.p...s_SESP2006.pdf" target="_blank"><a href="http://eden.dei.uc.p...s_SESP2006.pdf" target="_blank">http://eden.dei.uc.pt/~pmarques/papers/pmarques_SESP2006.pdf</a></a>
About the tweaking QEMU's source, I've been around that code a while and can only say that it would take me a long while just to get the hang of it. Mainly that's why I decided to code our emulator from scratch.
But the reason I dropped-by was not just to talk about emulators, but also to say Hi..
Hi Nuno!
Hi Carla!
Hi Everyone!
José,
I work for an AV company and was just wondering are you using dynamic translation to get such good speed?
Thanks,
Eric Sites, CTO
Sunbelt Software
#10
Posted 04 March 2008 - 05:11 AM
José,
I work for an AV company and was just wondering are you using dynamic translation to get such good speed?
Thanks,
Eric Sites, CTO
Sunbelt Software
José,
Yes you do, sorry, just read your paper, I do have a question on your Optimizer module. How much of a speed (MIPS) improvement did you get by adding that on top of basic block Dynamic Translation?
Thanks,
Eric Sites
#11
Posted 09 March 2008 - 08:19 PM
José,
Yes you do, sorry, just read your paper, I do have a question on your Optimizer module. How much of a speed (MIPS) improvement did you get by adding that on top of basic block Dynamic Translation?
Thanks,
Eric Sites
Hi Eric, nice to see you're interested.
I don't remember the exact numbers, but it I think it was around 40 MIPS.
You may also be interested to know that, that particular optimization was highly heavy on cache usage. We improved performance in 60% when switching from AMD CPUs (512KBs cache) to ~same-speed CPUs from Intel but with more cache.
Please feel free in any future doubts.
Best regards,
José
#12
Posted 11 March 2008 - 04:40 AM
We improved performance in 60% when switching from AMD CPUs (512KBs cache) to ~same-speed CPUs from Intel but with more cache
As an AMD user, I'm a little shocked. It's very interesting to know.
But I wonder now how much cache was allocated to the intel unit.
Thanks.
-t
#13
Posted 12 March 2008 - 10:55 AM
Could somebody benchmark it with ucLinux 2.0 and post results here?
0 user(s) are reading this topic
0 members, 0 guests, 0 anonymous users