Re: [Soekris] net6501 : CPU architecture ?
On Mon May 13, 2013 06:35:53 PM Kyle Brantley wrote: > On 5/13/2013 1:14 PM, Conrad Kostecki wrote: > > Am 13.05.2013 15:46, schrieb Stuart Henderson: > >> On 2013-05-10, Conrad Kostecki wrote: > >>> Am 10.05.2013 16:14, schrieb fabrice bessettes: > Hi list, > > [...] > > But I was wondering if the Soekris CPU E6xx is really a 64 bit CPU > ? or > if the openBSD kernel amd64 includes 32-bit in it. In this case, I > might > reinstall my system. > >> > >> OpenBSD/amd64 only runs in 64-bit mode. > >> > >>> Seems so. Ubuntu and Gentoo work perfectly fine with a 64-Bit kernel > >>> out > >>> of the box. Without any patching. So the CPU seems to be 64-Bit > >>> capable.. I don't know, why Intel says, that this is not possible. > >>> > >>> Cheers, > >>> Conrad > >> > >> The CPU is supposed to only support 32-bit mode but it appears that the > >> 6501 bios may not set this up correctly. > > > > I don't understand. How can a 32-Bit CPU work in 64-Bit Mode? I can't > > image, that the BIOS makes the difference. Linux shows the flags lm, > > which indicated, that it supports "AMD64/EM64T"? > > The CPU was produced as a 64-bit chip, but for reasons X or Y or Z, the > manufacturer is only going to support 32-bit. Maybe there are bugs with > respect to 64-bit, maybe it's a marketing thing. In the end, they market > the chips as 32-bit, nicely omitting any mention of 64-bit support, and > then ship the chips with notes to the people using them saying "we > support: 32-bit." The person using the chip (Soren in this case) is the > one obligated to only expose the 32-bit aspects of the chip. > > In short, as per Intel, using the 6501 in a 64-bit mode is unsupported. > With that said, I've been running my 6501s in 64-bit mode exclusively, > ever since I got one (two weeks after release?). I've not had any > problems with them, but then again, I may not be using them in a way > that triggers issues, should there be any. I think its up to intel to disable 64bit support if they really don't want it running in 64bit mode. like with other features it sells. Could be as simple as changing what the cpuid instruction returns. Or burning a trace somewhere that keeps the instruction decoder from decoding 64bit instructions. > > Cheers, > > Conrad > > ___ > > Soekris-tech mailing list > > Soekris-tech@lists.soekris.com > > http://lists.soekris.com/mailman/listinfo/soekris-tech > > --Kyle > ___ > Soekris-tech mailing list > Soekris-tech@lists.soekris.com > http://lists.soekris.com/mailman/listinfo/soekris-tech -- Thomas Fjellstrom tho...@fjellstrom.ca ___ Soekris-tech mailing list Soekris-tech@lists.soekris.com http://lists.soekris.com/mailman/listinfo/soekris-tech
Re: [Soekris] net6501 : CPU architecture ?
On 2013-05-13 18:35:53 -0600, Kyle Brantley wrote: > The CPU was produced as a 64-bit chip, but for reasons X or Y or Z, > the manufacturer is only going to support 32-bit. Maybe there are > bugs with respect to 64-bit, maybe it's a marketing thing. In the > end, they market the chips as 32-bit, nicely omitting any mention of > 64-bit support, and then ship the chips with notes to the people > using them saying "we support: 32-bit." The person using the chip > (Soren in this case) is the one obligated to only expose the 32-bit > aspects of the chip. Is the BIOS even able to hide the lm bit? Or are operating systems supposed to ignore it and use information from the BIOS (DMI?) instead? Otherwise I don't really see how the BIOS could prevent exposing that the CPU is a 64 bit chip. hp -- _ | Peter J. Holzer| Auf jedem Computer sollte der Satz Ludwigs II |_|_) | Sysadmin WSR | eingeprägt stehen: "Ein ewig Rätsel will ich | | | h...@wsr.ac.at | bleiben, mir und andern." __/ | http://www.hjp.at/ |-- Wolfram Heinrich in desd signature.asc Description: Digital signature ___ Soekris-tech mailing list Soekris-tech@lists.soekris.com http://lists.soekris.com/mailman/listinfo/soekris-tech
Re: [Soekris] net6501 : CPU architecture ?
On 14 May 2013, Kyle Brantley uttered the following: > On 5/13/2013 1:14 PM, Conrad Kostecki wrote: >> I don't understand. How can a 32-Bit CPU work in 64-Bit Mode? I can't >> image, that the BIOS makes the difference. Linux shows the flags lm, >> which indicated, that it supports "AMD64/EM64T"? >> > > The CPU was produced as a 64-bit chip, but for reasons X or Y or Z, > the manufacturer is only going to support 32-bit. Maybe there are bugs > with respect to 64-bit, maybe it's a marketing thing. I think it's more that it's slow. IIRC, 64-bit integer multiply is a *lot* slower than on other 64-bit parts: there may be other sloth. I wouldn't be surprised to find that it was slower than 32-bit even accounting for cache effects, where other 64-bit-capable CPUs are faster. > In short, as per Intel, using the 6501 in a 64-bit mode is > unsupported. ... so unsupported that Intel hackers have been adding Atom-specific 64-bit variants of the string routines into glibc of late. :) I suspect it's unsupported in the sense that "we want it to work, we rely on it working, but if you're not a geek don't bother." --- but anyone on this list has their geek credentials just on account of knowing that Soekris exists :) (I must say, having a houe move coming up I'm glad my firewall is a Soekris. It's the only machine I own that I can be *sure* will get to the far end of the car journey without risk of failure. Solid-state FTW!) -- NULL && (void) ___ Soekris-tech mailing list Soekris-tech@lists.soekris.com http://lists.soekris.com/mailman/listinfo/soekris-tech
Re: [Soekris] net6501 : CPU architecture ?
Peter J. Holzer wrote: > Is the BIOS even able to hide the lm bit? Or are operating systems > supposed to ignore it and use information from the BIOS (DMI?) instead? For a decade, Intel and AMD have provided patches that work around processor bugs and are implemented in microcode that is uploaded by the BIOS. Toggling CPU features in the same way isn't exactly magic. -- Christian "naddy" Weisgerber na...@mips.inka.de ___ Soekris-tech mailing list Soekris-tech@lists.soekris.com http://lists.soekris.com/mailman/listinfo/soekris-tech