On 6/10/07, Carl-Daniel Hailfinger c-d.hailfinger.devel.2006@gmx.net wrote:
On 10.06.2007 06:40, RusH wrote:
Im asking about reinitialization because I want to implement it in Memtest (memtest.org). I want to be able to setup memory controller the way I like and continue the test. Memtest is big (>100KB) and Im afraid it wont fit in cache so i need some real ram. That's where the 'jump to CAR, reinitialize controller, jump back to ram' idea comes from. Is it doable?
Why not run Memtest from graphics card RAM? A few years ago, graphics card RAM was really low quality, so periodic checksumming of any software running there might be necessary.
Date: Wed, 30 May 2007 18:57:44 +0200 From: RusH citizenr@gmail.com To: linuxbios@linuxbios.org Subject: using Graphics card ram as actual ram? Message-ID: 3df49b7b0705300957l3d063c50p404ed2cb07b8bd9@mail.gmail.com
so yes, i was thinking about it :)
Does reinitializing memory controller er destroy all the data in ram?
Depends. Some embedded systems (like the OLPC laptop) put memory into self-refresh for suspend-to-RAM and reinitialize the memory controller on resume without losing any contents. However, I wouldn't count on that for any recent non-embedded x86 system.
What about AMD64 controller? I know I can change a lot of timings on it, but changing tCAS is not possible without reinitializing, that's why I want to be able to reinitialize on the fly.
I read somewhere that reinitializing (memory allready set up earlier) without touching the mappings would be ok, but I want to be sure before starting coding (easier to ask than sit on it a good week and wonder why it doesnt work).
I guess I'll have to check for myself. All I got now is two boards with KT133 chipset, and that doesnt look to be supported (but PLE133/CLE266 are similar, so maybe that code will do).