| To: | linux-mips@linux-mips.org |
|---|---|
| Subject: | Re: Using more than 256 MB of memory on SB1250 in 32-bit mode |
| From: | Daniel Walton <dwalton+mips@ddtsm.com> |
| Date: | Thu, 20 May 2004 16:27:04 -0400 |
| In-reply-to: | <20030831133434.GA23189@linux-mips.org> |
| Original-recipient: | rfc822;linux-mips@linux-mips.org |
| References: | <3F4FCCD5.1000604@tadpole.com> <20030831133434.GA23189@linux-mips.org> |
| Reply-to: | Daniel Walton <dwalton+mips@ddtsm.com> |
| Sender: | linux-mips-bounce@linux-mips.org |
| User-agent: | Mutt/1.5.5.1+cvs20040105i |
On Sun, Aug 31, 2003 at 03:34:34PM +0200, Ralf Baechle wrote: > The explanation you gave isn't exactly right. A 2GB/2GB split would normally > support 2GB of low memory. We don't on MIPS due to the very inconvenient and > unchangable mappings of KSEG0/KSEG1 - something that may have been sweet > in '85 when the address map was designed but not today when 32-bit address > spaces are beginning to be fairly tight. > > Highmem works ok in 2.4 as long as you have a reasonably low ratio of > highmem to lowmem. For typical loads that means going beyond 4:1 isn't > sensible but the actual number may vary much based on exact system > configuration or workload. > > Ralf > So what is a practical limit to total memory. Daniel |
| <Prev in Thread] | Current Thread | [Next in Thread> |
|---|---|---|
| ||
| Previous by Date: | fix 2.4 mips64 build (CONFIG_NEW_TIME_C), Atsushi Nemoto |
|---|---|
| Next by Date: | [zli4@cs.uiuc.edu: [OPERA] Another potential bug in /arch/mips/kernel/sysirix.c], Jan-Benedict Glaw |
| Previous by Thread: | fix 2.4 mips64 build (CONFIG_NEW_TIME_C), Atsushi Nemoto |
| Next by Thread: | [zli4@cs.uiuc.edu: [OPERA] Another potential bug in /arch/mips/kernel/sysirix.c], Jan-Benedict Glaw |
| Indexes: | [Date] [Thread] [Top] [All Lists] |