On Wednesday, November 28, 2007 8:16 AM, Eric Saxe wrote: > Mark Haywood wrote: >> Thanks for bringing this up Aubrey. I've been looking into some of >> the reconstruction aspects myself. Specifically, I've been looking >> at how to remove the dependency the current implementation has on >> the Solaris PM framework. I was thinking that I'd want to do this >> work in the tesla-gate. Does it make sense for all of our CPU power >> management work to be done in the same repository? I'd favor doing >> it that way, but am open to alternatives. I'm also not sure how Eric >> intended that repository to be used. > The changes that decouple lbolt from clock() currently reside in the > tesla-gate repository. My initial intentions with tesla-gate were to > have a repository where various power management / efficiency changes > could co-exist for doing things like testing cumulative power savings > benefits...but given how extensive some of these changes are, > doing this > while things are in flux might be difficult to manage (I realized this > while doing a sync of the tesla-gate) with onnv_73. :) > > So having the CPU PM re-architecture changes in a separate > repository is > probably a good idea. :) I'm thinking it might make sense to have the > dispatcher changes co-exist in that same repo as well, if that > works... >
I think a separate repository is better. After the reconstruction is done, we can do the c-state and t-state driver in the same repo, ;-) Thanks, -Aubrey
