5 Nov
2010
5 Nov
'10
1:33 p.m.
On Fri, Nov 05, 2010 at 10:22:31AM +0000, Dimitris Papastamos wrote:
Yes that's the idea. I was also thinking, that during sync() we can have another flag like cache_bypass which would be set so when we are writing out the cache to the hardware, we don't write back to the cache again.
Yeah, though if we do the thing with keeping the last page uncompressed that'd get the overwhelming bulk of that anyway.