A more general possible meltdown/spectre countermeasure

[Eric McCorkle]

On 01/05/2018 03:15, Dag-Erling Smørgrav wrote:
> Eric McCorkle <eric at metricspace.net> writes:
>> The obvious downside is that you take a performance hit storing things
>> in non-cacheable locations, especially if you plan on doing heavy
>> computation in that memory (say, encryption/decryption).  However, this
>> is almost certainly going to be less than the projected 30-50%
>> performance hit from other mitigations.
> 
> Where did you get those numbers?  Because the worst documented case for
> KPTI is ~20% for I/O-intensive workloads, and PCID is likely to bring
> this down to single digits if used correctly.  The KAISER paper claims a
> slowdown of < 1%, but that may have been the result of undisclosed
> features of the specific CPU they tested on.

Those were numbers being thrown around.  I'm not putting a lot of stake
in them.