Here's a 44 with a few small improvements to yours:

sub a{$_=sprintf"%.2f",@_;1while s/\B...\D/,$&/;$_}
[download]

It seems that the sprintf solutions fail sometimes in rounding, e.g. "123456789012345.125" becomes "123,456,789,012,345.12" instead of "123,456,789,012,345.13".

Update: I was just informed by thospel++ that IEEE rounding is toward even, not upward. That explains the behavior, though I'm not sure if it's the desired behavior for this problem.

-- Mike

--
just,my${.02}