Ahmed W.

Hmmm I'll look into it, open to suggestions.

Thank you very much for taking the time to do this.

I completely removed the spinlock, it adds alloc overhead sadly but _in theory_, should work better on your intel box. If the changes don't add a noticeable improvement, can you...

@davecheney

@jonhoo I'm completely out of ideas :( my only theory is that using it on a true SMP system is different from a multicore processor, but I have 0 experience...

@jonhoo were you testing with go tip or go1.6?

@jonhoo I'm testing another theory to see if sharding the queue would help, if you have the time, can you please test again.

@jonhoo would you please give it another try. Also off topic, how do you generate those graphs?

run it with `-run NONE`, I broke the fifo test, not a big deal for now.

The results on the amd box looks promising with that last change.