Quote:
Originally Posted by ghostec
With a SSSE3 emulator, Phenom users would be able to run as good as bulldozer's?
|
It appears to be launchd that makes use of SSSE3, which is actualy open-source (I haven't yet looked it up) - but even if we could patch it, there might be other (closed source) parts of the OS that make use of that instructions set, so an emulator might be the way to go.
One disadvantage would of course be that using an emulation will decrease peformance a little...
Quote:
|
I took a look at the source of sse3 emulator for AnV's kernel (10.6.8) the procedure to inject the interrupt handler is the same in 10.7?
|
Yes these methods are still present.
Quote:
Originally Posted by ghostec
but I'm not sure yet how to make the osx kernel use them =/
|
For SSE3 Emulation INT 6h is overwritten with the adress of the interrupt handler, that passes the adress of the bad opcode to the actuall emulator that returns the emulation methods to the application and restores registers - principally the Kernel integration of a new emulator could be done the same way...
For details check it's whitepaper:
http://xnu-dev.googlecode.com/files/SSE3%20Emulator.pdf
Someone around here who has an enslaved asm guru locked-up in the basement?!^^
Quote:
Originally Posted by gils
"pmVersion=0" or other is useless, unless you're a sleepenabler somewhere!
|
The latest Kernel that I posted should no longer panic on version mismatch between Kernel & SleepEnabler.kext - no need to update that kext unless you want the warning to be removed from dmesg.