Messages in this thread | | | Date | Mon, 9 Mar 1998 07:41:30 -0800 (PST) | From | "B. James Phillippe" <> | Subject | Re: [PATCH] Cyrix setup |
| |
On Mon, 9 Mar 1998, Martin Mares wrote:
> So, what Cyrix-related things do we need to have in the kernel? > > - Suspend on halt: Do we really need to set it up during kernel > boot? The example posted earlier mentioning system which died of overheating > before it reached the rc scripts doesn't seem to be important from my > point of view as it's certainly a HW problem and a machine suffering from > such problems won't go reliably anyhow.
I agree that the reason cited is not particularly convincing; for this scenario, setting HLT-on-idle is only delaying the inevitable. However, I would strongly argue that this is a feature that all Cyrix users would want on by default, and there is no logical reason to make each person go through extra work to enable it (which is actually not an entirely safe operation in and of itself; a typo on the set6x86 command line can be doomsday).
A user should not need to identify the make and model of CPU before being able to enjoy radically reduced power consumption and extended CPU life. Also, packaged distributions should be able to install on off-the-shelf Pentium or Cyrix systems and boot into this mode without user-land black-magic CPU detection and register settings (which may not even be possible in some cases).
> - NO_LOCK (Coma bug workaround): Does anyone have a working solution > (Alan mentioned the original one broke some X servers)?
As I mentioned to Alan, unless someone else comes forward, I am the only originator of the X/NO_LOCK complaint, which I am prepared to retract. I have been running with 2.1.88 and NO_LOCK for 11 days with no problems, on the same system which I had problems with earlier. I have attributed my original failures to human error (I had other registers set with set6x86; yet another reason to not give people a loaded weapon).
> - VSPM: Probably not worth supporting as the performance gain is > almost unnoticeable. > > The rest of Cyrix features seems to be solved cleanly by userland > utilities.
I concur. I would decline arguing for any of the other Cyrix features in the kernel. I would of course enable many of them if they were present, but I think they're best left to user-space. The main difference is that HLT and NO_LOCK are in the category of "mission-critical" and the rest are in the category of "possible performance optimisation".
cheers, -bp -- B. James Phillippe <bryan@terran.org> Linux Software Engineer, WGT Inc. http://earth.terran.org/~bryan
- To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majordomo@vger.rutgers.edu
| |