Port-alpha archive

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index][Old Index]

Re: MULTIPROCESSOR enabled by default on all Alpha kernels



My only SMP Alpha is a CS20 running DS20L firmware. I know there were issues at one time with the built-in NIC and SCSI controllers. Did those ever get resolved? If so, I’ll get it racked and powered up.

I had swapped the SCSI controller out with one for a DS20L to run Tru64 on it at one time, but I still have the original (if I remember correctly, NetBSD didn’t like the DS20L one). I know one was Adaptec, I forget what the other one was.

-Dustin
On Sep 19, 2020, 11:53 AM -0500, Jason Thorpe <thorpej%me.com@localhost>, wrote:


On Sep 19, 2020, at 9:39 AM, Izumi Tsutsui <tsutsui%ceres.dti.ne.jp@localhost> wrote:

The system has been very stable for me, and is feeling much snappier.

I'm afraid if timecounter(9) using multiple PCC processor clocks
works properly.

Yes, I think there are some known issues. Taylor (CC'd) also had some thoughts on it?


I've committed it back in 2007 to switch alpha port to MI timecounter(9),
https://mail-index.netbsd.org/source-changes/2007/07/21/0012.html
https://mail-index.netbsd.org/source-changes/2007/07/21/0013.html
but IIRC ad@ pointed out that x86 TSC implementation (that alpha's
PCC implementation was based on) was problematic.

There was also a report that sleep(1) or TOD etc. didn't work
as expected if timecounter(9) was pcc (and worked fine with
clockinterrupt).

---
Izumi Tsutsui

-- thorpej



Home | Main Index | Thread Index | Old Index