[parisc-linux] User space locks -- what's wrong

[Grant Grundler]

On Sun, Jun 11, 2006 at 11:13:33PM -0400, Carlos O'Donell wrote:
> Reading the thread register is a single mfctl, I don't know if that is
> considered slow?

I believe it depends on the source of the "Move From".
But I can't find a table with timing info. (e.g. mfctl CR16).

I did find this bit though:

| 	Most control operations (the "light weight" control ops) are
| handled by a mechanism that is very similar to the way that C/B
| dependencies are handled. The major differences are that there
| are no control rename registers, and there is no bypassing.
| Thus, an instruction that is dependent on a control op will not
| be able to execute until the control op retires. 
...

"Heavy Weight" control ops are the ones where the instruction execution
is serialized completely.

hth,
grant