Re: MVME3100 FPU Support on RTEMS 4.10.2

[Joel Sherrill <joel@rtems.org>]

On Wed, Jun 6, 2018, 4:29 PM Michael Davidsaver <mdavidsaver@gmail.com> wrote:

On 06/06/2018 02:24 PM, Matt Rippa wrote:
> Thanks for the fast reply. Of all the mvme boards you listed,
> the 3100 hits a nice price point and we're attempting to find
> suitable replacement stock
> ​(quantity 30+) ​
> for the 2700 which are no longer
> manufactured.

Last I checked (~5 years ago) the newer mvme2500 had a price similar to the mvme3100.
The issue was software support.  At the time there didn't seem to be RTEMS board
support

I would expect the mvme2500 to just one of the many variants supported by the motorola_powerpc BSP family. Some of the board variants, you could just use another variant name. I don't know if this would require tweaks or not.

What about the mvme5500? It wasn't mentioned by anyone. Is that a price issue?

I suppose spinning off Motorola Computing Group and the processor group has resulted in few upgrade options.

> Are the 6100 boards on beatnik bsp the only remaining option for RTEMS?
> They cost nearly twice as much and have
> ​shown ​
> thermal regulation issues.
>
> -Matt
>
> On Wed, Jun 6, 2018 at 11:10 AM, Michael Davidsaver <mdavidsaver@ospreydcs.com <mailto:mdavidsaver@ospreydcs.com>> wrote:
>
>     Hi Matt,
>
>     The MVME3100 is based on the MPC8540 processor.  This chip doesn't have an FPU.
>     Other boards do (eg. mvme270x, mvme6100, or mvme2500) and RTEMS supports this.
>
>     I can certainly confirm that lack of an FPU really hurts performance, even
>     something as (apparently) simple as integer to float.
>
>     Michael
>
>
>     On 06/06/2018 02:03 PM, Matt Rippa wrote:
>     > It appears we've compiled our mvme3100 bsp without FPU support.
>     > We're on the mvme3100 running EPICS 3.14.12.7 and RTEMS 4.10.2.
>     >
>     > We're attempting to commission our astrometric kernel control system
>     > and we've run into trouble meeting some of our thread deadlines. What
>     > we've found is a transform routine that takes our mvme2700 ~200 us,
>     > takes the mvme3100 ~8 ms.  So not a surprise to see this with only "-msoft-float"?  
>     >
>     > ...
>     > 2018-05-31-tcs.log:May 31 11:41:15  E) PORT: tcs_vme, MSG: Welcome to rtems-4.10.2(PowerPC/Generic (*no FPU*)/mvme3100) 
>     >
>     > Is FPU support possible with RTEMS 4.10.2? If so, what would we need to do?
>     > It looks like there's a multilib on 4.11.2 https://docs.rtems.org/releases/rtems-docs-4.11.2/cpu-supplement/powerpc.html#multilibs(*Eg <https://docs.rtems.org/releases/rtems-docs-4.11.2/cpu-supplement/powerpc.html#multilibs(*Eg>., see #13* )
>     >
>     > Here's our bsp specs:
>     >
>     > beatnik:
>     > -specs bsp_specs -qrtems -fasm -c -mcpu=7400 -D__ppc_generic -DUNIX -O2 -g -g -Wall
>     > (mvme2700) mvme-2307:
>     > -specs bsp_specs -qrtems -fasm -c -mcpu=604 -mmultiple -mstring -mstrict-align -meabi -DUNIX -O2 -g -g -Wall
>     > mvme-3100
>     > -specs bsp_specs -qrtems -fasm -c -mcpu=powerpc *-msoft-float* -D__ppc_generic -DUNIX -O2 -g -g -Wall
>     >
>     > Thank you!
>     > -Matt
>     >
>
>
>
>
> --
>
_______________________________________________
users mailing list
users@rtems.org
http://lists.rtems.org/mailman/listinfo/users