Casper.Dik at Sun.COM wrote: >> Then why does Windows work with it? It has to handle ticks too, and >> I'm sure it does it in a similar fashion at the lower level. Windows >> is architecturally just as nasty, if not more than Solaris, and is an >> evolution of NT 3.51, so I believe with enough manpower, it'd be >> possible, albeit highly difficult. >> > > "Given enough manpower" yes, it is possible, but it will cost quite a > bit of additional overhead to switch to a different timer base. > > (Solaris derives time from TSC and needds time to be the same across > all CPUs for obvious reasons.) > > Casper > Who's to blame? I still say Sun is here, you guys took forever to even fully support x86. I am aware that since Solaris 2.6 Sun had an x86 port, that didn't make it at feature parity with SPARC systems. Now, outside of the politics and general usability scope, what specific source files would I go about looking at, so I can figure this out a bit more. Excuse me for seeming a bit naive, I do not have experience with the actual OpenSolaris source, I have not had time in the past to look at it. But I will give it a look now, it'd be in my best interest to make a rough estimate in how many man hours it'd need to get done. I'm prematurely guessing it'd take a full year, based on your assertion that overdependence of "cheap ticks" is pervasive through the whole kernel architecture.
James
