pub fn log_gemm_shape(gpu: &dyn GpuBackend, name: &str, m: u32, n: u32, k: u32)Expand description
Roofline instrumentation: log each unique (kernel, M, N, K) GEMM shape once,
gated by ATLAS_GEMM_SHAPE_LOG=1. Used to cross-reference nsys per-call
durations → achieved TFLOPS/bandwidth vs GB10 peak.