Article: AMD's Mobile Strategy
By: Linus Torvalds (torvalds.delete@this.linux-foundation.org), December 21, 2011 8:18 pm
Room: Moderated Discussions
Exophase (exophase@gmail.com) on 12/20/11 wrote:
>
>http://board.flatassembler.net/topic.php?p=137802
>
>See the last post.
So it doesn't actually have any real pc-relative memops.
It has a 19-bit "load literal" operation (note: no store),
which is not actually useful for any variables. It's very
much just for loading constants that are embedded in the
instruction stream - a poor mans immediate.
To access actual globals, you have to use two instructions
and waste a register.
That's still better than most RISC machines tend to do, so
it's not a bad solution. It's probably as good as you
can do with a fixed-width instruction set. But it's an
example of why variable-length instruction sets are nice.
Linus
>
>http://board.flatassembler.net/topic.php?p=137802
>
>See the last post.
So it doesn't actually have any real pc-relative memops.
It has a 19-bit "load literal" operation (note: no store),
which is not actually useful for any variables. It's very
much just for loading constants that are embedded in the
instruction stream - a poor mans immediate.
To access actual globals, you have to use two instructions
and waste a register.
That's still better than most RISC machines tend to do, so
it's not a bad solution. It's probably as good as you
can do with a fixed-width instruction set. But it's an
example of why variable-length instruction sets are nice.
Linus
