for r := 0; r < c.rounds-1; r++ { t0 = xk[k+0] ^ td0[uint8(s0>>24)] ^ td1[uint8(s3>>16)] ^ td2[uint8(s2>>8)] ^ td3[uint8(s1)] t1 = xk[k+1] ^ td0[uint8(s1>>24)] ^ td1[uint8(s0>>16)] ^ td2[uint8(s3>>8)] ^ td3[uint8(s2)] t2 = xk[k+2] ^ td0[uint8(s2>>24)] ^ td1[uint8(s1>>16)] ^ td2[uint8(s0>>8)] ^ td3[uint8(s3)] t3 = xk[k+3] ^ td0[uint8(s3>>24)] ^ td1[uint8(s2>>16)] ^ td2[uint8(s1>>8)] ^ td3[uint8(s0)] k += 4 s0, s1, s2, s3 = t0, t1, t2, t3 } // Last round uses s-box directly and XORs to produce output....