CAV12 vs. Power

As the CAV’12 model is experimentally equivalent to our model, except for a few test which are not observed on hardware (see this document), the experimental comparison of our model with Power hardware applies.

We thus here only emphasize that the experiments that invalidate the PLDI’11 model do not invalidate the CAV’12 model.

	Kind	CAV12	Power	PowerG5	Power6	Power7
MP+sync+ctrl-detr	Allow	Allow	Ok, 12k/40G	No, 0/3.1G	—	Ok, 12k/37G
				Allow unseen
DETOUR0821	Allow	Allow	Ok, 26k/53G	Ok, 2/3.7G	No, 0/3.4G	Ok, 26k/46G
					Allow unseen
MP+lwsync+addr-po-detr	Allow	Allow	Ok, 21k/37G	Ok, 3/2.9G	—	Ok, 21k/34G
MP+sync+addr-po-detr	Allow	Allow	Ok, 17k/37G	Ok, 1/2.9G	—	Ok, 17k/34G

This document was translated from L^AT_EX by H^EV^EA.

Contents