Previously speedy multi-threaded Fortran program slows dramatically on new M4 Max Mac Studio

[canecaster]

Output of brew config

HOMEBREW_VERSION: 4.4.26
ORIGIN: https://github.com/Homebrew/brew
HEAD: f1041842141f20bc701b522d4371d7843d905580
Last commit: 6 days ago
Branch: stable
Core tap JSON: 28 Mar 12:48 UTC
Core cask tap JSON: 28 Mar 12:48 UTC
HOMEBREW_PREFIX: /opt/homebrew
HOMEBREW_CASK_OPTS: []
HOMEBREW_MAKE_JOBS: 16
Homebrew Ruby: 3.3.7 => /opt/homebrew/Library/Homebrew/vendor/portable-ruby/3.3.7/bin/ruby
CPU: 16-core 64-bit arm_brava
Clang: 16.0.0 build 1600
Git: 2.39.5 => /Applications/Xcode.app/Contents/Developer/usr/bin/git
Curl: 8.7.1 => /usr/bin/curl
macOS: 15.3.2-arm64
CLT: 16.2.0.0.1.1733547573
Xcode: 16.2
Rosetta 2: false

Output of brew doctor

Your system is ready to brew.

Description of issue

I have a Fortran application, compiled using Homebrew's gcc, that can be executed either to run with 1 thread or multiple threads. My experience on my M1 Ultra Mac Studio and M2 Max MacBookPro is that it runs about 10-20% faster with 2 threads than 1. However, I just finished setting up an M4 Max Mac Studio, and on this machine the application runs MUCH slower with 2 threads than 1. I have a standard set of program inputs to test the timings.

When run with 1 thread, the M4 Max Mac Studio timing is the fastest machine among the three, as expected. But I can't figure out why the multi-thread performance of this program is suddenly so much degraded. It does not seem to matter on which of the Macs I compile the program on (the code and executables sync via iCloud so fresh compilations propagate to all three machines). My compile command looks like this:

gfortran -O2 -fopenmp -o verify_model verify_model.f ${lib_path}/verifymlib.o ${lib_path}/atcflib.o ${lib_path}/tclip2020lib.o ${lib_path}/sondelib.o ${lib_path}/pchiplib.o ${lib_path}/splinelib.o

I'm wondering if anyone has any insight into what might be going wrong with multi-threading on the new M4 Max Mac Studio or how I'd begin to diagnose the issue. Multi-thread performance on my older two machines continues to be fine. I haven't run any standardized benchmarks on the new machine, so perhaps someone could point me to the most appropriate one(s) to run.

Grateful for any suggestions.

James

[canecaster]

If it helps, I just ran the following GeekBench 6 benchmarks on the new Mac Studio: Single core score of 3999 and a multi-core score of 26338.