Re: macOS metal rendering engine in mac port

[Stefan Monnier <monnier@iro.umontreal.ca>]

>> Mine are fairly consistent. I imagine hitting thermal would hurt
>> though and Intel macs love to do that.
>
> Yeah, I wonder if that's what's happening. I have to say that I don't
> think this Mac has ever been what I'd consider fast. My ancient first
> gen i7 absolutely wipes the floor with it in compilation speed
> despite being at least 7 years older. I guess Moores law really did
> run out of steam. ;)

Dennard scaling and Moore's law both ran out of steam, indeed.
For many modern machines there is a very large difference between the
peak single thread performance attainable in a short sprint and the
sustainable single thread performance, and yet more difference with the
sustainable performance of single thread running while other threads are
also running.
E.g. my Librem mini's "top speed" in GHz can vary between 4.7GHz
and 1.8GHz depending on those circumstances.

So, when I need stable results to compare the speed for 2 different
operations, I generally do it as follows:

- Write a script that runs A enough times for the duration to be in
  the "seconds" range.  Same for B.  Make that script print the
  time measurement.
- Write a loop whose body runs A followed by B.
- Run that loop and look at the measurements.

Then you should see that the measurements will stabilize as the thermal
constraints stabilize (of course, you'll also want to avoid other
activity on the machine during that time, as usual).

Of course, that doesn't guarantee that what you measure is
representative of what the user will see, but least it lets you compare
apples to apples.


        Stefan