Fix compute capability encoding for mathDx

[simonbyrne]

This was getting computed incorrectly (e.g. 809 instead of 890)

[copy-pr-bot]

This pull request requires additional validation before any workflows can run on NVIDIA's runners.

Pull request vetters can view their responsibilities here.

Contributors can view more details about this message here.

[greptile-apps]

Greptile Summary

This PR fixes an off-by-10× encoding bug in compute capability calculation where the minor version was added as a raw digit instead of being multiplied by 10 (e.g., SM 8.9 was encoded as 809 instead of the correct 890). The fix is applied consistently across all three affected sites.

• include/matx/core/utils.h: GetComputeCapability() now returns major * 100 + minor * 10, matching the __CUDA_ARCH__ encoding convention used throughout CUDA. The downstream consumer in tensor_impl.h (>= 1000 for SM 10.0) remains correct with both encodings since SM 10.0 has minor = 0.
• include/matx/operators/fft.h and include/matx/operators/matmul.h: The same fix is applied to the two local cc computations that feed into mathDx helper set_cc() calls, ensuring mathDx selects the correct code path for GPUs with non-zero minor versions (e.g., Ada Lovelace SM 8.9).

Confidence Score: 5/5

Safe to merge — all three encoding sites are corrected and no callers are broken by the fix.

The change is a minimal, targeted arithmetic correction applied consistently to every location that computes the CUDA compute capability value. The one downstream caller comparing against >= 1000 for SM 10.0 is unaffected because SM 10.0 has minor = 0, making both encodings produce identical results there.

No files require special attention.

Important Files Changed

Filename	Overview
include/matx/core/utils.h	Fixed GetComputeCapability() to multiply the minor version by 10, so SM 8.9 -> 890 instead of 809; also removes a trailing blank line.
include/matx/operators/matmul.h	Fixed local compute-capability encoding in the mathDx GEMM helper initialization path to use minor * 10.
include/matx/operators/fft.h	Fixed local compute-capability encoding in the mathDx FFT helper initialization path to use minor * 10; the previously-noted bug is now resolved.

_{Reviews (2): Last reviewed commit: "Fix compute capability encoding for math..." | Re-trigger Greptile}

[simonbyrne]

/build

[coveralls]

Coverage is 94.308% — fix/sm-encoding into main. No base build found for main.