Which itself can be reduced to %a = (%a, %b);. Granted if you wanted to preserve the elements of either hash, you would have to use the above method. I presume the behavior is undefined, but on my machine:

%a = (%a, %b);  # common elements contain %b's values
%a = (%b, %a);  # common elements contain %a's values
[download]

Yes, it is simpler, but it seems that @a{keys %b} = values %b; is consistently faster. A quick mesurement shows that

perl -we 'use Time::HiRes "time"; %a = map rand, 1 .. (1<<18); %b = ma
+p rand, 1 .. (1<<18); warn($t = time()); %a = (%a, %b); warn time() -
+ $t; warn 0+keys(%a);'
[download]

runs in 2.8s, while

perl -we 'use Time::HiRes "time"; %a = map rand, 1 .. (1<<18); %b = ma
+p rand, 1 .. (1<<18); warn($t = time()); @a{keys %b} = values %b; war
+n time() - $t; warn 0+keys(%a);'
[download]

in 1.0s.