| 1 |
On Sat, Jul 10, 2010 at 11:13:04PM -0400, Matt Turner wrote: |
| 2 |
> On Thu, Jul 8, 2010 at 12:22 PM, Matt Turner <mattst88@×××××.com> wrote: |
| 3 |
> > Hi, |
| 4 |
> > I'm going to be working on Gentoo/MIPS. I'm planning to provide N32 |
| 5 |
> > stages, and I want to know what kind of MIPS hardware people have to |
| 6 |
> > run Gentoo. With this info, I should be able to better prioritize the |
| 7 |
> > tasks. |
| 8 |
> > |
| 9 |
> > So please reply with what kind of hardware you've got. |
| 10 |
> > |
| 11 |
> > Thanks! |
| 12 |
> > Matt |
| 13 |
> |
| 14 |
> Thanks everybody for your responses. |
| 15 |
> |
| 16 |
> A bit of background: MIPS has 3 ABIs: o32, n32, n64. o32 is the common |
| 17 |
> denominator, but is pretty inefficient. See [1]. As I suspected, all |
| 18 |
> the hardware people care about can use the n32 ABI, which is quite a |
| 19 |
> bit better. |
| 20 |
> |
| 21 |
> To summarize the hardware collection |
| 22 |
> - 5 MIPS III/Loongson little-endian STMicroelectronics systems (Lemote, Gdium) |
| 23 |
> - 11 MIPS IV big-endian SGI systems |
| 24 |
> - 3 MIPS IV little-endian Cobalt systems |
| 25 |
> - 2 MIPS64 selectable-endian Broadcom systems |
| 26 |
|
| 27 |
I have some little-endian MIPS32 hardware which runs latest Gentoo. |
| 28 |
|
| 29 |
|
| 30 |
> I've got a working n32 big-endian installation, and am preparing an |
| 31 |
> n32 little-endian installation at the moment. Once these are set up, |
| 32 |
> I'll start preparing some n32 stages. |
| 33 |
|
| 34 |
> I think first priority should be mips4 big-endian. From there, I don't |
| 35 |
> know whether I should build a mips4-le stage for cobalt systems, or a |
| 36 |
> mips3-le stage that would work for cobalt and Loongson. Do we know of |
| 37 |
> any plain MIPS III little-endian hardware? |
| 38 |
|
| 39 |
I believe all the small MIPS32-based routers which run OpenWRT qualify. |
| 40 |
Hoever I get the impression you want to focus on the "large" SGI |
| 41 |
and 64bit Broadcom systems? |
| 42 |
|
| 43 |
|
| 44 |
> Also, the MIPS port is going to be pretty unmaintainable until we can |
| 45 |
> kill off ~mips from ACCEPT_KEYWORDS. Currently, attempting to upgrade |
| 46 |
> almost anything causes something to break. |
| 47 |
|
| 48 |
In my experience, lots of packages have problems with GCC-4.4 (removal |
| 49 |
of the "h" asm constraint), and above all, the toolchain itself is |
| 50 |
pretty problematic: Lots of linker errors with large libraries (e.g. |
| 51 |
xulrunner, qt-webkit, gcj, ...) |
| 52 |
|
| 53 |
|
| 54 |
Manuel |
Archives