On 10/07/2012 11:17 PM, Aurelien Jarno wrote: > For what it worth, I measure a 4% slow down booting a sparc64 guest on a > Core-i5 2500 machine. I guess the memcpy() of the register windows is > more expensive that the gain on the TCG side, though it should be > probably be confirmed using some profiling tools.
Confirmed with userland testing (running cc1plus on large input), and I see much more than 4% slow down -- more like 15%. I can reduce this by adding save/restore window inc/dec optimization within the cpu_set_cwp routine, but that only gets me back to 5% slow down. A bit disappointing; I guess other avenues will want exploring. Patch withdrawn. r~
