| Author |
Message |
Guest
|
 Posted:
Sun Feb 27, 2005 12:21 am Post subject:
delayed branches |
|
|
hi,
Does delayed branches make the ISA compiler unfriendly? If so,
can someone explain me why?
regards
ganesh |
|
| Back to top |
|
 |
Tommy Thorn
Guest
|
| Posted:
Sun Feb 27, 2005 6:01 am Post subject:
Re: delayed branches |
|
|
gans1973@rediffmail.com wrote:
| Quote: | Does delayed branches make the ISA compiler unfriendly? If so,
can someone explain me why?
|
Homework? Have you considered the obvious questions of how to fill the
branch delay slot? With what? Which constraints?
Tommy |
|
| Back to top |
|
|
Peter Dickerson
Guest
|
| Posted:
Sun Feb 27, 2005 3:59 pm Post subject:
Re: delayed branches |
|
|
<gans1973@rediffmail.com> wrote in message
news:1109445685.736111.126300@o13g2000cwo.googlegroups.com...
| Quote: | hi,
Does delayed branches make the ISA compiler unfriendly? If so,
can someone explain me why?
regards
ganesh
|
Yes, of course, because now the compiler must run on a processor that
executes instructions out-of-order. While most processors do that these days
they also take a lot of trouble to act as if the instructions were in order.
The problem being that delayed branches don't work that way - the branch is
done after the following instruction. Getting the compiler to do this when
the CPU can't is hard work. Normally you need a JIT engine to build an
instruction stream with the branches and following instruction swapped.
Again, JIT technology is nothing new, its just one more issue the compiler
really didn't need.
For more details try Google "delayed branches ISA compiler-unfriendly"
Peter |
|
| Back to top |
|
|
|
|
|
|
FAQ
Memberlist
Register
Profile
Log in to check your private messages
Log in