To view parent comment, click here.
To read all comments associated with this story, please click here.
The goal, to my understanding, is to execute legacy programs with competitive performance, not a Guinness records for the sake of it.
I didn't made such a performance claims. I'm talking about good enough performance. Even if it is just 50%.
As I said, Intel engineers did ARM binary translation without any "magic buzz" about it. It just works.
viton,
"I didn't made such a performance claims. I'm talking about good enough performance. Even if it is just 50%."
I've repeatedly said that efficiency is important to me, why do you think I'm so fixated on the optimiser? It's not just for the sake of complexity.
"As I said, Intel engineers did ARM binary translation without any 'magic buzz' about it. It just works."
Are you talking about running ARM/Android software on atoms?
http://www.anandtech.com/show/5365/intels-medfield-atom-z2460-arriv...
That's um, the opposite.
This following link however is closer to what I am talking about:
http://www.wired.com/wiredenterprise/2012/10/russian-hackers-arm/
"Elbrus Technology’s secret sauce is its binary translator with multiple layers of hand-tuned optimization."
They are getting 40% efficiency btw. Hopefully that will convince you that highly efficient translation is difficult and needs a target optimizer. If you still don't want to admit that, well what can I say, we'll just have to disagree.





Member since:
2011-01-28
viton,
"You're doing over-engineering here. Branch elimination could be helpful, but it will work fine without it."
That's because your aiming for lower branches than I am.
"ARM need more registers to implement load-op and op-in-memory x86 instructions. 16 registers is not much."
What your talking about isn't going to run as well as something optimized to use the ARM's registers natively. The x86 register limitation is a well documented bottleneck even on x86 itself where intel and amd have spent billions on making it perform well. It perplexes me why you think an ARM processor could run x86 code as well as intel can.
"Qemu is a full-system emulator. x86 on WinRT should not mess with MMU pages."
Yes, qemu manages the MMU, but even when everything is loaded into memory and running will have to incur overhead as a results of implementing x86 semantics & registers on a foreign architecture.
"The only I see here is a lack of understating of ARM arch and binary translation."
So everyone else is wrong, enlightening that.