> Actually some would argue that assembler should. For example with
>the MIPS version of as you can specify if you want it (as) to reorder for
>the branch delay slot or if you'll do it manually. The branch delay slot
>is required where as on both the MIPS and the StrongARM the load delay
>slot is to avoid a stall.
> I've also heard of optimising assemblers that try to reorder
>assembly code to avoid pipeline stalls and such.
Yes, optimising assemblers certainly do exist (as do optimising linkers).
But as regards GNU tools on ARM, all instruction scheduling is done in GCC
and GAS has no intelligence.
If you are seeing poor schedules you could try upgrading to the latest 2.95
version plus the patch in my directory on netwinder.org. If things are still
no good the next thing to try is using the latest development snapshot (see
http://gcc.gnu.org) which includes several more improvements.
Do you have particular testcases that are causing inefficient code to be
generated? If so, post them and I'll take a look.
p.
unsubscribe: body of `unsubscribe linux-arm' to [EMAIL PROTECTED]
++ Please use [EMAIL PROTECTED] for ++
++ kernel-related discussions. ++
