Port-i386 archive

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index][Old Index]

Re: MATH_EMULATE



On 12-Jan-08, at 1:43 PM, "Perry E. Metzger" <perry%piermont.com@localhost> 
wrote:


"Jared D. McNeill" <jmcneill%invisible.ca@localhost> writes:
Yes, I have looked at it before, many times. It is okay. Linus granted the right to use that file as we use it. I don't remember the details
any more, but it should be easy to find them.

Well that's good news. This should be noted in the file.

We'd have to find the original grant for that, but yes, it would be
reasonable to note that it is okay in the file.

How familiar are you with math_emulate.c? Would you be willing to
bring it back to a working state?

I could have a look. I used to be reasonably familiar with it, though
admittedly it has been ten years or more since I paid attention.

That would be great -- thanks!

First step, though -- no one has mentioned yet what exactly is broken
-- does anyone know exactly?  Andrew said he brought up a machine and
it then started to lose, but not exactly what was losing. Once we know
what's broken it should be possible to figure out what we want to do
-- it might be ten minutes work (in which case why not) or it might be
very bad in which case perhaps we let it die.

(BTW, I'll note that you almost always want to run soft float
executables on an FPUless machine, as the emulation is very, very slow
-- it is just that having the emulator means you don't completely lose
if you have an executable with x87 instructions.)

You should be able to fake an fpu-less config on a modern machine, either with an emulator or by disabling the real fpu.

Jared



Home | Main Index | Thread Index | Old Index