ARM THUMB: fundamental bug in handling of far jumps?

[Richard Earnshaw]

> On Tue, 2002-01-29 at 22:46, Richard Henderson wrote:
> 
> > Not-taken branches are normally better for performance.  Might not
> > apply to the thumb though.
> 
> afaict at least on arm7tdmi branches are indifferent 

It would be best to consider the arm7 (and arm9 and StrongARM) as 
predicting all branches as not-taken.  The cost in all cases of not taking 
the branch is 1 cycle (the branch insn itself).  The cost of taking the 
branch is a pipeline refill (2 to 5 cycles depending on the processor).

R.