[ggorman]

Problem

get_platform() recognises Apple silicon only via 'm1' in brand (devito/arch/archinfo.py). A brand string such as "Apple M2 Pro" or "Apple M4 Max" falls through to ANYCPU, so @skipif('cpu64-arm') does not fire. tests/test_autotuner.py::test_nested_nthreads then runs with platform='knl7210', the generated code includes xmmintrin.h/pmmintrin.h, and Homebrew clang on arm64 rejects it:

xmmintrin.h:14:2: error: "This header is only meant to be used on x86 and x64 architecture"

Seen on #3047: that run had zero Arm doesn't support x86-specific instructions skips across the suite, while the main run the day before (same image, same LLVM 23.1.0) skipped the test.

macOS CI misses this bears on

The pytest-osx-py312-clang-omp-base job prints the detected platform in its header. On the same macos-26-arm64 image (20261013.0424):

Run	Header	Outcome
#3074, 2026-10-20 (run 39457547592)	platform: m1	green, 50 skipped
#3075 (run 39303223750)	platform: cpu64	test_nested_nthreads CompileError
#3074, 2026-10-21 (run 39622887638)	platform: cpu64	test_nested_nthreads CompileError and test_gradient.py::TestGradient::test_gradient_equivalence[...float64-1e-13...] (also skipif('cpu64-arm')) fails allclose

Each miss costs a 35-40 minute osx job plus a full rerun of the workflow. The job does not print the brand string, so the cause of these particular misses is unconfirmed: a host chip newer than M1 is fixed here; a brand missing because py-cpuinfo's helper subprocess failed is not, and still falls through to cpu64.

Fix

Map any apple m<N> brand to the registered AppleArm of that generation, falling back to the newest registered one for chips not yet listed; AppleArm.march derives the actual -mcpu at runtime from sysctl. The lookup is a small pure function, get_apple_platform, so it can be tested without stubbing get_cpu_info; tests/test_arch.py::test_apple_platform covers M1 to M5 brand strings, as asked in review.

Check

Besides the test, calling get_platform with get_cpu_info stubbed to "Apple M1 (Virtual)", "Apple M2 Pro", "Apple M3 Max", "Apple M4 Max" and "Apple M5" yields m1, m2, m3, m3, m3, all AppleArm; Xeon, EPYC and Neoverse brands are unchanged. Before the change only the M1 string resolves to AppleArm.

🤖 Generated with Claude Code

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 84.27%. Comparing base (a9e3f31) to head (2bc1bb6).

Additional details and impacted files

@@           Coverage Diff           @@
##             main    #3048   +/-   ##
=======================================
  Coverage   84.27%   84.27%           
=======================================
  Files         258      258           
  Lines       57421    57430    +9     
  Branches     4911     4911           
=======================================
+ Hits        48392    48401    +9     
  Misses       8205     8205           
  Partials      824      824           

Flag	Coverage Δ
pytest-gpu-aomp-amdgpuX	68.49% <20.00%> (+<0.01%)	⬆️
pytest-gpu-gcc-	78.89% <81.81%> (-0.01%)	⬇️
pytest-gpu-icx-	78.85% <81.81%> (+0.04%)	⬆️
pytest-gpu-nvc-nvidiaX	69.20% <20.00%> (-0.01%)	⬇️

Flags with carried forward coverage won't be shown. Click here to find out more.

☔ View full report in Codecov by Harness.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

[EdCaunt]

Not unreasonable, but lots of noise in here.

[ggorman]

Rebased onto main and squashed to the one archinfo.py change. Since your review the same miss hit #3075 and #3074 (runs 39303223750 and 39622887638, header platform: cpu64 on the same image that read m1 the day before); the PR description now has the table. Each hit costs the osx job plus a full rerun.

[FabioLuporini]

OK