Julius Baxter
2014-06-04 13:58:43 UTC
On Thu, May 29, 2014 at 1:09 AM, Luís Vitório Cargnini
The current state of OR2k is that it's still vaporware. To be honest
it would be great if we could get someone to kick start the process on
this. I really think the OpenRISC project would be helped by
developing a newer architecture with a more dense instruction set, and
more suitable for deeply embedded use. It's a lot of work, though, and
my time is very limited at the moment, so I've not been involved at
all, however I'm excited by the prospect of the effort getting under
way.
However, if you're in a position to look at doing some work on it in
any capacity, it'd be useful. Probably the initial steps are to figure
out exactly what we're looking for. We'd need to think about what it
is we're aiming for. Probably something which derives from OR1K would
be nice, but not necessary?
Probably it's something along the lines of what we have now, in that
it's a configurable architecture which is capable of being stripped
bare, in terms of features, to run small bare metal apps and RTOSes,
and fully featured to run more complex operating systems such as those
based on the Linux kernel. Things like the floating point and SIMD
instructions are nice, but on FPGAs, of little use. Probably we should
focus on a tight, neat control-based instruction set, to begin with
anyway.
I think it might be worth revisiting this topic on the mailing lists,
so I'll CC them to see if anyone has any information or opinions.
In my opinion, something which is a lot like OR1K at the moment (a lot
of optional features such as caches, MMUs, parts of the instruction
set) but with better code density, and a better system of doing
conditional operations (branching, optional execution). Security is
probably also a concern. Some people have mentioned things about
virtualisation but I'm not too familiar with this, nor do I know if
it's of any use in deeply-embedded applications, which is where I
think OR2K could have its greatest potential.
So I look forward to reviving this topic a little, and seeing what
people come up with. As I say, I think if there's a proposal for an
ISA which doesn't preclude people from being able to scratch their
itch, add what they need and contribute it to the community (atomic
operations are a good, recent example of this) then that'd be a good
start.
Cheers
Julius
Hello Julius,
Please, which is the current state of the OR2K ?
I would like to be more involved since I would like to use it to my research
in my workplace.
Hi LuisPlease, which is the current state of the OR2K ?
I would like to be more involved since I would like to use it to my research
in my workplace.
The current state of OR2k is that it's still vaporware. To be honest
it would be great if we could get someone to kick start the process on
this. I really think the OpenRISC project would be helped by
developing a newer architecture with a more dense instruction set, and
more suitable for deeply embedded use. It's a lot of work, though, and
my time is very limited at the moment, so I've not been involved at
all, however I'm excited by the prospect of the effort getting under
way.
However, if you're in a position to look at doing some work on it in
any capacity, it'd be useful. Probably the initial steps are to figure
out exactly what we're looking for. We'd need to think about what it
is we're aiming for. Probably something which derives from OR1K would
be nice, but not necessary?
Probably it's something along the lines of what we have now, in that
it's a configurable architecture which is capable of being stripped
bare, in terms of features, to run small bare metal apps and RTOSes,
and fully featured to run more complex operating systems such as those
based on the Linux kernel. Things like the floating point and SIMD
instructions are nice, but on FPGAs, of little use. Probably we should
focus on a tight, neat control-based instruction set, to begin with
anyway.
I think it might be worth revisiting this topic on the mailing lists,
so I'll CC them to see if anyone has any information or opinions.
In my opinion, something which is a lot like OR1K at the moment (a lot
of optional features such as caches, MMUs, parts of the instruction
set) but with better code density, and a better system of doing
conditional operations (branching, optional execution). Security is
probably also a concern. Some people have mentioned things about
virtualisation but I'm not too familiar with this, nor do I know if
it's of any use in deeply-embedded applications, which is where I
think OR2K could have its greatest potential.
So I look forward to reviving this topic a little, and seeing what
people come up with. As I say, I think if there's a proposal for an
ISA which doesn't preclude people from being able to scratch their
itch, add what they need and contribute it to the community (atomic
operations are a good, recent example of this) then that'd be a good
start.
Cheers
Julius
Best Regards,
Luis Vitorio Cargnini
Luis Vitorio Cargnini