On Mon, Sep 19, 2016 at 2:17 PM, Rick Jones <rick.jon...@hpe.com> wrote:
> > Are there better than epsilon odds of someone perhaps wanting to poke those > values as it gets exposure beyond Google? > This does not matter. A change would require patching net/ipv4/tcp_bbr.c , and the 'const' attribute being there or not does not prevent the change. I was simply asking to Stephen if the compiler would actually emit a very different code, worth doing a last minute change. The main BBR costs are divides and some multiplies , and they are using per socket fields.
