![]() | |||||||||||||
Messages in this thread |
Guys, I observed a strange phenomemum. On a certain kind of sparc(32), the amount of page tables is accounted, and 2.5.66 seems to use about 4 times more page tables than 2.4.20. I am talking about these kinds of numbers. On a 28MB box, an idle 2.4 used about 57600 bytes of pagetables, which equals to 14400 PTEs, or about enough to map 50 MB of RAM. Obviously, there were partially filled tables, and also processes mapped some shared memory with non-shared page tables. On 2.5, the same system uses 273664 bytes of page tables. If they weren't shared, and were completely used (and not full of invalid PTEs, for instance), they would map 267MB of RAM. Sounds kinda wasteful, for a 32MB box. Truth to be told, I would not worry about a mere 270K of pagetables, but the sparc port as it is uses so-called "nocache" memory, which is preallocated on boot. OK, I know it's broken. Blame Anton. I plan to fix it some time in the future. But currently, my box runs out of nocache and panics. Developer's itch, you see... So, does anyone have a handy patch or tool or something, which prints page table statistics? Does anyone have ready numbers on other architectures? I would appreciate other insights, too. Thanks, -- Pete - To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html Please read the FAQ at http://www.tux.org/lkml/ | ||||||||||||
| Last update: 2005-03-22 12:34 [W:0.219 / U:0.040 seconds] ©2003-2008 Jasper Spaans | |||||||||||||