Date: Thu, 17 Nov 2005 15:50:21 -0500 From: Chuck Swiger <cswiger@mac.com> To: Dan Nelson <dnelson@allantgroup.com> Cc: "Marc G. Fournier" <scrappy@hub.org>, freebsd-questions@freebsd.org Subject: Re: EM64T supported? Message-ID: <437CED0D.2080800@mac.com> In-Reply-To: <20051117202734.GF62141@dan.emsphone.com> References: <20051117150323.U1019@ganymede.hub.org> <437CE254.3080701@mac.com> <20051117202734.GF62141@dan.emsphone.com>
next in thread | previous in thread | raw e-mail | index | archive | help
Dan Nelson wrote: > In the last episode (Nov 17), Chuck Swiger said: >>EM64T uses 64-bit wide registers and addressing, and can talk to >4GB >>of RAM natively. Older processors may still support >4GB of physical >>RAM using the PSE/PSE-36 CPU extensions, but are still using 32-bit >>registers. > > PAE/PAE36, right? Note that if you enable PAE, some drivers may not be > available. See the PAE kernel config file for a list. PAE is related, but I don't believe "PAE36" exists; cpuid lists these: PSE Page Size Extensions PAE Physical Address Extension PSE-36 36-bit Page Size Extension I believe PSE lets you choose whether your MMU uses a 4KB or a 4MB pagesize for virtual address translation. PAE was the first attempt at supporting more than 4GB of address space, but I gather it requires doing bank swapping or something fairly awkward that doesn't play too well with VM, whereas PSE-36 integrates more easily. The other point you've made is correct, that is, a fair number of drivers don't understand PAE/PSE36 yet, and will not work using it-- generally because the hardware associated with the driver has a DMA engine which is limited to 32-bit addressing. You end up having to double-buffer or use "DMA bounce buffers", whatever phrase you wish to use. :-) This link seems to have a more complete description: http://www.microsoft.com/whdc/system/platform/server/PAE/pae_os.mspx -- -Chuck
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?437CED0D.2080800>