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

Re: new architecture wishlist

Hi, all.

2009/7/1 Simon Richter <sjr@debian.org>:
> On Tue, Jun 30, 2009 at 04:00:01PM +0200, Hector Oron wrote:
>> > Any other wishes for platforms to be added? The targets aren't required
>> > to be self-hosting, this is mainly so dpkg-architecture knows them, and
>> > can provide appropriate environment variables when invoked with -a.
>> SH arches if there is a formal port:
>>   sh3, sh4, sh4a, sh4al, shmedia,...
> sh3             sh3-linux-gnu
> sh3eb           sh3eb-linux-gnu
> sh4             sh4-linux-gnu
> sh4eb           sh4eb-linux-gnu
> These exist currently -- do we need all of these, or should any be dropped
> in the process?
> >From what I see, the main difference between sh4 and sh4a in userspace are
> the new interlocking functions and different instruction scheduling to
> avoid pipeline stalls, as I expect that the new address model can be
> abstracted away in kernel space?

At first a developer of SH include me which part the architecture in
ABI of glibc.
(I ignore shmedia and sh2, and other)
It has it of 2, and it is ABI with fpu and ABI which it does not use.
This is sh4 and sh3.
Because there is difference between an endian and this, I divide it
into sh4, sh4eb, sh3, sh3eb.

sh4a and sh4al are the architecture that point of focus to CPU-core.
sh4a is compatible with a high rank of sh4.
sh4a has fpu. Therefore, it is classified in the sh4 ABI architecture.

When the program for sh4a operates it on sh4, there is a problem, but
reverse case not to have any problem.
When it does static link of libgcc, it is an exception.

My sh4 port does not use an order of sh4a.

> shmedia would be sh5?
Yes. However, is there a person having a machine of sh5?
# I have .....

Best regards,

Nobuhiro Iwamatsu

Reply to: