papers/evmgpu-benchmark/sections/06-gpu-shader-inventory.tex

\section{GPU Shader Inventory}


The cevm GPU acceleration framework includes 35 Metal shaders and 11 WGSL shaders,
organized by function:

\begin{table}[h]
\centering
\small
\begin{tabular}{llp{5.5cm}}
\toprule
\textbf{Category} & \textbf{Count} & \textbf{Shaders} \\
\midrule
EVM Core & 5 & \texttt{evm\_kernel}, \texttt{state\_table},
               \texttt{block\_stm}, \texttt{tx\_validate}, \texttt{evm256} \\
\addlinespace
Crypto & 6 & \texttt{keccak256}, \texttt{blake3}, \texttt{secp256k1},
             \texttt{bls12\_381}, \texttt{ed25519}, \texttt{sr25519} \\
\addlinespace
Post-Quantum & 6 & \texttt{mldsa}, \texttt{mlkem}, \texttt{slhdsa},
                    \texttt{corona}, \texttt{frost}, \texttt{cggmp21} \\
\addlinespace
FHE & 7 & \texttt{fhe\_kernels}, \texttt{blind\_rotate} ($\times$2),
          \texttt{bsk\_prefetch}, \texttt{external\_product} ($\times$3) \\
\addlinespace
NTT & 10+ & \texttt{ntt} ($\times$6 variants), \texttt{poly\_mul},
            \texttt{twiddle\_cache} \\
\midrule
\textbf{Total} & \textbf{35 Metal + 11 WGSL} & All compile clean \\
\bottomrule
\end{tabular}
\caption{GPU shader inventory. Post-quantum shaders implement all three NIST FIPS
algorithms (ML-DSA, ML-KEM, SLH-DSA) plus Lux-specific Pulsar. FHE shaders
support TFHE blind rotation for encrypted EVM computation. NTT variants serve
as the shared primitive across PQ signatures and FHE.}
\label{tab:shaders}
\end{table}