Re: LRA best_reload_nregs

[Vladimir Makarov <vmakarov at redhat dot com>]

On 12-11-04 5:46 PM, David Miller wrote:
> Unlike the other variables that track the state of the current
> instruction being analyzed by the LRA constraints code, I don't
> see anything which initializes best_reload_nregs when we start
> looking at a new instruction.
It is actually not necessary because the first alternative is chosen 
only by comparison 'overall' and the code for chosen best alternative 
initialize this variable.  There are analogous variables best_reload_sum 
and best_small_class_operands_num.

But may be you are right, it is better to have them initialized for code 
clearness.