Thumb: pevent extraneous jumps around constant tables

[Richard Henderson]

On Wed, Feb 09, 2000 at 09:51:36AM -0800, Nick Clifton wrote:
> Because we are (no longer) getting barriers after the epilogue.

We aren't?  We're supposed to be.  Before I installed the recent
epilogue changes, I remember adding code to make absolutely sure
we got them.

I just re-checked -- I'm getting them on Alpha:

(jump_insn 65 64 66 (return) -1 (nil)
    (nil))
;; End of basic block 0

(barrier 66 65 41)

I can't imagine how ARM would be different for this case.


r~