On 2023-10-08 18:07, Alan Browne wrote:
Just brought home a Mac Mini (M2) that belongs to a client.
Copied over a file I compiled in x86 (command line) and it would not run
on the M2 - Rosetta did not get invoked.
Next installed the compiler (universal binary itself) and the installer
for the compiler (x86) induced Rosetta to be installed.
Once that was done then the x86 binary (compiled on old Mac) ran fine on
the M2. Nice.
Next re-compiled the program on the M2 Mac and it of course ran fine.
Then ... tried a simple multithreaded test program, ran fine.
However, where the M2 can Handbrake the i7 iMac 2.5x to 1 or better,
when I run a brute force search in 8 threads to find a, b, c where:
ab + c = 2020
a + bc = 2021
Single threaded test: i7: 1.468s M2: 2.369s 1:1.59
Multithreaded test (8 threads): i7: 0.387s M2: 1.055s 1:2.73
Multithreaded test (4 threads): i7: 0.408s M2: 1.159s 1:2.84
Proving (no surprise really) that even a 2012 i7 (4 core, HT) @ 3.8GHz
will beat Apple silicon on number bashing.
Hmm --- re-complied after updating the Mac Mini to Sonoma 14.0 ...
Single threaded test: i7: 1.468s M2: 1.740s 1:1.19
Multithreaded test (8 threads): i7: 0.387s M2: 0.376s 1:0.97
Multithreaded test (4 threads): i7: 0.408s M2: 0.585s 1:1.43
So the M2 is outdoing the i7 at 8 threads, not at 4.
Not sure what happened between Ventura and Sonoma - but seems to have
gotten out of the way of the processor.
-- “Markets can remain irrational longer than your can remain solvent.”
- John Maynard Keynes.
--- SoupGate-Win32 v1.05
* Origin: fsxNet Usenet Gateway (21:1/5)