pub const W8A8_PREFILL_KERNELS: [(&str, &str); 2];Expand description
2026-09-25: The two (module, function) pairs the W8A8 block-scaled prefill
arm needs: the activation quantizer and fp8_gemm_t_blockscaled.
init.rs resolves the GEMM from this table, and
w8a8_prefill_kernels_loaded asks the backend for both, so serve load
can ask before any layer exists.