Den 2023-09-24 kl. 21:48, skrev Johnny Billquist:
IIRC the 1.2 distribution took 11 days to compile, and 2 additional days to gzip the distribution files.On 2023-09-24 21:46, Anders Magnusson wrote:Den 2023-09-24 kl. 21:32, skrev Johnny Billquist:Change was back in 1998-08-21. Thanks for just reminding me how old I'm getting. :-)I find it interesting that this NetBSD doesn't use TBIS at all.Agreed. Seems like it would potentially be clever to just doTBIS when you update one page.True. It may be relevant that NetBSD/vax, 1.4T at least, uses4K pages in the getpagesize() sense; [...]The 4K page is a bit more recent. Back when this was messed around with, NetBSD used a 1K page size on VAX.Well, the 1.4T kernel I'm working with definitely uses 4K pages.Wow. It's just over 25 years ago.I remember when testing with 4k logical pages it gave a huge performance gain, especially when starting binaries. (most easy to time by doing "make build" on the 11/785 I used at that time. It took several days to compile everything).Well. That's better than now. Now a make build does not work at all, and haven't for quite a few years... :-(
But it worked :-)
But yeah. It made a big difference for performance. I still do think that there are times when TBIS could be motivated to use, but I'm not going to try and poke at it for now...Since (at least) 8 TBIS are needed for a kernel mapping, and even more for userspace, I realized it just wasn't worth it.
Don't think I ever timed it though. -- R