ppc405 floating point emulation
Hi there,
Can someone tell me what they think the status of the mathemu on ppc405 currently is? I am faced with a situation in which we will probably require some kind of library or kernel emulation support and all around people suggest not to even go there :-).
Cheers,
Jon.
** Sent via the linuxppc-embedded mail list. See http://lists.linuxppc.org/
Quoted 5 lines · In message <[email protected]> you wrote:
Can someone tell me what they think the status of the mathemu on ppc405 currently is? I am faced with a situation in which we will probably require some kind of library or kernel emulation support and all around people suggest not to even go there :-).
Library support (i. e. using "-msoftfloat") works fine.
Kernel emulation is available, but makes little sense to use as it is awfully slow.
Best regards,
Wolfgang Denk
--
Software Engineering: Embedded and Realtime Systems, Embedded Linux
Phone: (+49)-8142-4596-87 Fax: (+49)-8142-4596-88 Email: [email protected]
panic: kernel trap (ignored)
** Sent via the linuxppc-embedded mail list. See http://lists.linuxppc.org/
Quoted 1 line · Wolfgang Denk
Library support (i. e. using "-msoftfloat") works fine.
Yes good. I had thought it would be fine to use that.
Quoted 2 lines
Kernel emulation is available, but makes little sense to use as it is awfully slow.
The problem I have with this kernel is that floating point emulated instructions which perform stores seem to be falling over. There is a bug in the fault handling code for stock 2.4.21 that I am using it would seem but am looking at whether it is worth fixing it.
By all accounts I might as well just use the floating point emulation in libc but am still interested in enlightened feedback.
Cheers,
Jon.
** Sent via the linuxppc-embedded mail list. See http://lists.linuxppc.org/
Op wo 26-11-2003, om 15:38 schreef Jon Masters:
Quoted 6 lines
Hi there,
Can someone tell me what they think the status of the mathemu on ppc405 currently is? I am faced with a situation in which we will probably require some kind of library or kernel emulation support and all around people suggest not to even go there :-).
That's what they told me too, but I did go there. And the result is still quite good.
I'm currently writing the last documentation on our laser-projector. This projects figures which are calculated by sin and cos together with a pow function. Off course, we used lookup-tables and approx-functions, but it works pretty well. Yesterday was the big show of it, and most people liked it.
Anyhow, I use it in combination with real-time, and you cannot rely on float calculations then. So you need a good implementation to get full result of it.
Calculation isn't very fast, but on half a second, my module calculates 360 X and Y coordinates with 2 sin, 2 cos and 3 pow functions each point.
So it is possible to use softfloat, but you MUST know that you cannot rely on speed then. So far my small experience with it :)
PS: next week, I hope the docs will be ready and available on my website under 'cases'/'our case'.
Regards,
--
Toni Van Remortel
Wetenschappelijk Medewerker - D-science lab
Real time Linux for embedded systems: http://linemb.d-sciencelab.com
Tel: +32 3 205 61 72 - Fax: +32 3 205 61 95
E-mail: [email protected]** Sent via the linuxppc-embedded mail list. See http://lists.linuxppc.org/