Oh, sorry, poor choice of label. Compile time vs. run time is where
the decision about SE vs. FS mode is made. It might be informative to
look at build time too.
Gabe
Quoting Steve Reinhardt <[email protected]>:
Cool, thanks for doing this. What kind of machine is this on (particularly
how many cores)? I'd be interested in seeing the time it takes for a
from-scratch compile using a single core (say X86_FS/gem5.opt) just to get
a clearer measure of the additional work.
Thanks,
Steve
On Wed, Jan 18, 2012 at 6:07 AM, Gabe Black <[email protected]> wrote:
ARM_FS/tests/opt/long/10.linux-boot/arm/linux/realview-o3/
compile time:
host_seconds
840.68 # Real time elapsed on the host
run time:
host_seconds
840.02 # Real time elapsed on the host
ARM_FS/tests/opt/long/10.linux-boot/arm/linux/realview-o3/
compile time:
host_seconds
34.50 # Real time elapsed on the host
run time:
host_seconds
33.11 # Real time elapsed on the host
build/X86_SE/tests/opt/long/70.twolf/x86/linux/simple-atomic/
compile time:
host_seconds
87.55 # Real time elapsed on the host
host_seconds
93.78 # Real time elapsed on the host
run time:
host_seconds
103.96 # Real time elapsed on the host
host_seconds
100.22 # Real time elapsed on the host
One thing which surprised me is that choosing SE/FS mode at runtime
performed better than choosing it at compile time on ARM_FS, although
there was more overhead on X86_SE. I'm not really sure why that's
happening. It could be an SE vs. FS thing somehow.
Gabe
_______________________________________________
gem5-dev mailing list
[email protected]
http://m5sim.org/mailman/listinfo/gem5-dev
_______________________________________________
gem5-dev mailing list
[email protected]
http://m5sim.org/mailman/listinfo/gem5-dev
_______________________________________________
gem5-dev mailing list
[email protected]
http://m5sim.org/mailman/listinfo/gem5-dev
