Does using one obscure application of SIMD even count as benchmark? It just could be that the library is specifically optimized for x86. Run something like BLAS

Does using one obscure application of SIMD even count as benchmark? It just could be that the library is specifically optimized for x86. Run something like BLAS matrix multiplication which could use the full power of SIMD in both the platforms(I don't have access to M1 macbook).