"Richard Earnshaw (lists)" <richard.earns...@arm.com> writes:
[...] > + [(set (attr "length") > + (if_then_else > + (lt (minus (pc) (match_dup 0)) (const_int 1024)) > + (const_int 4) > + (const_int 6))) > + (set_attr "type" "branch")]) > > Shouldn't that be using "ltu" rather than "lt", so that if, for some > reason, the branch has been retargeted to come after the branch, then > the test will still fail and we'll get the comparison variant back. Right good point > Otherwise OK. > > R. Installed into trunk with the suggestion as 38c5703449c. Thanks! Andrea