| --- Log opened Sun Dec 14 00:00:10 2014 | ||
| -!- Netsplit *.net <-> *.split quits: jonmasters, freakazoid0223, poke53282, knz, pecastro, Pick | 05:32 | |
| -!- Netsplit *.net <-> *.split quits: Pick, knz, freakazoid0223 | 05:59 | |
| -!- Netsplit over, joins: freakazoid0223, Pick, knz | 06:01 | |
| wallento | stekern: i started to work on the accessed/dirty flags in TLBs | 10:16 |
|---|---|---|
| stekern | wallento: great, hopefully our changes won't clash too much ;) | 10:34 |
| wallento | yes, I hope the same. if you have a minute, can you please roughly summarize the huge vs. normal tlb thing? | 10:36 |
| wallento | I essentially understand that we can have Level 1 Pages, that are larger. the tlb can therefore hold such entries and we need dpram to check for both. The huge tables get precendence and tlb_huge is set when we found a level 1 tlb entry | 10:42 |
| wallento | is this correct? | 10:42 |
| wallento | but the huge entries are only written by the hw tlb reload? | 10:43 |
| wallento | ah, I see, I can of course also set the entry with spr | 10:44 |
| wallento | okay, that means the history flags need to be updated for both styles | 10:44 |
| stekern | mmm, but the dirty and accessed bits are only relevant for hw reload anyway | 11:37 |
| stekern | ...or is it...? | 11:45 |
| wallento | no, the OS can also use them | 11:52 |
| wallento | for swapping | 11:53 |
| wallento | what does the hw tlb reload do with them? | 11:53 |
| wallento | i think it is transparent to the reload | 11:53 |
| wallento | at the moment there is not field in xMMUCR or similar that contains the number of fields, correct? | 11:54 |
| stekern | yeah, I realised that the OS can explicitly clear the flags and then be interested if the page currently in cache was accessed again | 11:56 |
| stekern | what number of fields? | 11:59 |
| wallento | eh, ways, sorry | 12:01 |
| stekern | it's in the NTW field in xMMUCFGR | 12:09 |
| wallento | ah, okay, thanks! | 12:19 |
| -!- aou is now known as a_ou | 21:33 | |
| -!- a_ou is now known as aou | 21:34 | |
| olofk | stekern: Nice work with releasing the multicore stuff | 22:14 |
| olofk | Do you have any simulation setup you can push to orpsoc-cores? | 22:14 |
| olofk | For multicore | 22:15 |
| --- Log closed Mon Dec 15 00:00:12 2014 | ||
Generated by irclog2html.py 2.15.2 by Marius Gedminas - find it at mg.pov.lt!