In the video in the link, it was concluded that the GPU did not affect the test, and it seemed to be CPU-bound.
Actually, to me, the issue is more so that it's not truly verifiable that Photoshop is running the same instructions on Mac vs. PC. A better test would be to use GIMP, and use strace to confirm that the same code branches are being executed.
That said, your statement is correct. A 2017 CPU vs. a 2013 CPU isn't going to fare well, clock-for-clock, given that IPC usually get better. That difference in performance will be exacerbated if the software takes advantage of the newer instruction sets too.
If this filter is written in C/C++, then LLVM would probably have the advantage over MSVC. If this filter is written with Halide, then they should be running nearly identical code.
Also, strace is only really going to show you system calls, which should not be relevant to straight-line code.
Actually, to me, the issue is more so that it's not truly verifiable that Photoshop is running the same instructions on Mac vs. PC. A better test would be to use GIMP, and use strace to confirm that the same code branches are being executed.
That said, your statement is correct. A 2017 CPU vs. a 2013 CPU isn't going to fare well, clock-for-clock, given that IPC usually get better. That difference in performance will be exacerbated if the software takes advantage of the newer instruction sets too.