[ORLinux] [OpenRISC] [ORToolChain][PATCH]'or32el', the little-endian support for gdb, no sim.

[Liu]

On Tue, Aug 30, 2011 at 7:48 PM, Jeremy Bennett
<jeremy.bennett at embecosm.com> wrote:
> On Sun, 2011-08-28 at 03:53 +0800, Liu wrote:
>> Hi all
>>
>> I'm working on an little-endian OpenRISC chip, and I have made
>> gdb support 'or32el' option.
>> That is, gdb now support big-endian OpenRISC and little-endian
>> OpenRISC at the same time!
>>
>> But I'm sorry for that or-gdb require or1ksim, I can NOT understand
>> or1ksim for now, so I can NOT make or1ksim support little-endian, it
>> makes little-endian-or-gdb couldn't support simulation in it:(
>
> Hi Liu,
>
> Thanks for all your work so far.
>
> Much more work will be needed. Your have provided the changes to add
> or32el to the configuration scripts.
>
> However you need to change the code functionality of or32-tdep.c, so it
> knows how to analyse little-endian instructions. For example breakpoints
> will not work if you do not provide a little endian l.trap. None of the
> prologue analysis will work if you do not unpick little-endian versions
> of the instructions.
>
>> I plan to provide a qemu port for OpenRISC, and I'm working on qemu
>> or32 porting, maybe weeks, maybe months, I'll make it work.
>> Then or32el-gdb will be more perfect:)
>
> Your top priority must be a simulator that can be used with the GNU tool
> chain and integrated into GDB. Until you can test your code it won't be
> able to be accepted.
>
> I am not sure QEMU is the correct approach. To get your binutils working
> you will need the CGEN description working, and it would be far simpler
> to add the simulation semantics to that.
>
> HTH,
>
>
> Jeremy
>
> --
> Tel:      +44 (1590) 610184
> Cell:     +44 (7970) 676050
> SkypeID: jeremybennett
> Email:   jeremy.bennett at embecosm.com
> Web:     www.embecosm.com
>
>

Thank you, Jeremy
I'll make it all right next time:)