Rigel: Reverse-Engineering the Metal 4.1 Tensor Compute Path on the Apple M4 Max GPU

Researchers at an undisclosed institution reverse-engineered Apple's Metal 4.1 tensor compute path on the M4 Max GPU, revealing that the fp8 matmul2d operation is emulated rather than hardware-accelerated. The study, published as a preprint on arXiv, found the operation sustains only 0.94x the throughput of fp16 despite reading half the operand bytes, indicating it functions as a memory-footprint feature rather than a performance one. The findings, which also showed the operation executes entirely on GPU shader cores with no dedicated matrix datapath, enabled researchers to build a hand-fused GEMM kernel that outperformed the decomposed path by 6.5-12.9% in cache-resident regimes.

arXiv:2606.12765v1 Announce Type: new Abstract: Apple's Metal 4.1 exposes a tensor compute path: the Metal Performance Primitives MPP matmul2d operation over cooperative tensor fragments, whose interface is documented but whose hardware behavior is deliberately hidden. The specification states which data-type rows are supported, never whether they are hardware-accelerated, where the operation physically executes, what its accumulator width is, or how it partitions matrix fragments across threads. We present Rigel, an empirical characterization of this path on a single Apple M4 Max a pre-neural-accelerator generation . Using a checksum-gated, provenance-tracked microbenchmark harness, Rigel recovers eleven facts the v4.1 specification hides or contradicts. The headline finding: the Metal 4.1 fp8 E4M3 matmul2d is emulated, not accelerated: it sustains 0.94x the throughput of fp16 despite reading half the operand bytes, so on M4 it is a memory-footprint feature, not a performance feature. We further show, via a three-signal triangulation throughput ceiling, comparison against simdgroup matrix, and per-rail power attribution , that matmul2d executes entirely on the GPU shader cores with no dedicated matrix datapath and no evidence of Apple Neural Engine routing; that it accumulates in =fp32; and we reconstruct the opaque 8x8 cooperative tensor fragment layout Apple documents nowhere. Acting on the characterization, a hand-fused GEMM + bias + GELU kernel beats the decomposed path by +6.5-12.9% in the cache-resident regime. All findings are reproducible from committed MIT-licensed code and per-cell CSVs.