void intrin_sequential_enc8(const unsigned char* PT, unsigned char* CT, int n_aesiters, int nkeys, ROUND_KEYS* ks){ ROUND_KEYS *keyptr=(ROUND_KEYS *)ks; register __m128i keyA, keyB, keyC, keyD, keyE, keyF, keyG, keyH, con, mask, x2, keyA_aux, keyB_aux, keyC_aux, keyD_aux, globAux; unsigned char *ptptr, ctptr; int i, j, ptoffset, ctoffset; ctoffset = n_aesiters * 16; for (i=0;i<nkeys;i+=8){ for(j=0;j<n_aesiters; j++) { register __m128i block1 = _mm_loadu_si128((__m128i const*)(0*16+PT)); register __m128i block2 = _mm_loadu_si128((__m128i const*)(1*16+PT)); register __m128i block3 = _mm_loadu_si128((__m128i const*)(2*16+PT)); register __m128i block4 = _mm_loadu_si128((__m128i const*)(3*16+PT)); register __m128i block5 = _mm_loadu_si128((__m128i const*)(4*16+PT)); register __m128i block6 = _mm_loadu_si128((__m128i const*)(5*16+PT)); register __m128i block7 = _mm_loadu_si128((__m128i const*)(6*16+PT)); register __m128i block8 = _mm_loadu_si128((__m128i const*)(7*16+PT)); READ_KEYS(0) block1 = _mm_xor_si128(keyA, block1); block2 = _mm_xor_si128(keyB, block2); block3 = _mm_xor_si128(keyC, block3); block4 = _mm_xor_si128(keyD, block4); block5 = _mm_xor_si128(keyE, block5); block6 = _mm_xor_si128(keyF, block6); block7 = _mm_xor_si128(keyG, block7); block8 = _mm_xor_si128(keyH, block8); ENC_round(1) ENC_round(2) ENC_round(3) ENC_round(4) ENC_round(5) ENC_round(6) ENC_round(7) ENC_round(8) ENC_round(9) ENC_round_last(10) _mm_storeu_si128((__m128i *)(CT+0*16), block1); _mm_storeu_si128((__m128i *)(CT+1*16), block2); _mm_storeu_si128((__m128i *)(CT+2*16), block3); _mm_storeu_si128((__m128i *)(CT+3*16), block4); _mm_storeu_si128((__m128i *)(CT+4*16), block5); _mm_storeu_si128((__m128i *)(CT+5*16), block6); _mm_storeu_si128((__m128i *)(CT+6*16), block7); _mm_storeu_si128((__m128i *)(CT+7*16), block8); PT+=128; CT+=128; } keyptr+=8; } }
void intrin_sequential_gen_rnd8(unsigned char* ctr_buf, const unsigned long long ctr, unsigned char* CT, int n_aesiters, int nkeys, ROUND_KEYS* ks){ ROUND_KEYS *keyptr=(ROUND_KEYS *)ks; register __m128i keyA, keyB, keyC, keyD, keyE, keyF, keyG, keyH, con, mask, x2, keyA_aux, keyB_aux, keyC_aux, keyD_aux, globAux; unsigned char *ctptr; int i, j, ctoffset; unsigned long long* tmpctr = (unsigned long long*) ctr_buf; ctoffset = n_aesiters * 16; register __m128i inblock, block1, block2, block3, block4, block5, block6, block7, block8; for (i=0;i<nkeys;i+=8){ ctptr=CT + i*ctoffset; (*tmpctr) = ctr; for(j=0;j<n_aesiters; j++) { (*tmpctr)++; inblock = _mm_loadu_si128((__m128i const*)(ctr_buf)); READ_KEYS(0) block1 = _mm_xor_si128(keyA, inblock); block2 = _mm_xor_si128(keyB, inblock); block3 = _mm_xor_si128(keyC, inblock); block4 = _mm_xor_si128(keyD, inblock); block5 = _mm_xor_si128(keyE, inblock); block6 = _mm_xor_si128(keyF, inblock); block7 = _mm_xor_si128(keyG, inblock); block8 = _mm_xor_si128(keyH, inblock); ENC_round(1) ENC_round(2) ENC_round(3) ENC_round(4) ENC_round(5) ENC_round(6) ENC_round(7) ENC_round(8) ENC_round(9) ENC_round_last(10) _mm_storeu_si128((__m128i *)(ctptr+0*ctoffset), block1); _mm_storeu_si128((__m128i *)(ctptr+1*ctoffset), block2); _mm_storeu_si128((__m128i *)(ctptr+2*ctoffset), block3); _mm_storeu_si128((__m128i *)(ctptr+3*ctoffset), block4); _mm_storeu_si128((__m128i *)(ctptr+4*ctoffset), block5); _mm_storeu_si128((__m128i *)(ctptr+5*ctoffset), block6); _mm_storeu_si128((__m128i *)(ctptr+6*ctoffset), block7); _mm_storeu_si128((__m128i *)(ctptr+7*ctoffset), block8); ctptr+=16; } keyptr+=8; } }
void intrin_sequential_ks2_enc2(const unsigned char* PT, unsigned char* CT, int test_length, unsigned char* KEYS, unsigned char* first_key, unsigned char* TEMP_BUF){ ROUND_KEYS *keys=(ROUND_KEYS *)KEYS; register __m128i keyA, keyB, con, mask, x2, keyA_aux, keyB_aux, globAux; int i; int _con1[4]={1,1,1,1}; int _con2[4]={0x1b,0x1b,0x1b,0x1b}; int _mask[4]={0x0c0f0e0d,0x0c0f0e0d,0x0c0f0e0d,0x0c0f0e0d}; int _con3[4]={0x0ffffffff, 0x0ffffffff, 0x07060504, 0x07060504}; __m128i con3=_mm_loadu_si128((__m128i const*)_con3); for (i=0;i<test_length;i+=2){ keys[0].nr=10; keys[1].nr=10; keyA = _mm_loadu_si128((__m128i const*)(first_key)); keyB = _mm_loadu_si128((__m128i const*)(first_key+16)); _mm_storeu_si128((__m128i *)keys[0].KEY, keyA); _mm_storeu_si128((__m128i *)keys[1].KEY, keyB); con = _mm_loadu_si128((__m128i const*)_con1); mask = _mm_loadu_si128((__m128i const*)_mask); KS_round(1) KS_round(2) KS_round(3) KS_round(4) KS_round(5) KS_round(6) KS_round(7) KS_round(8) con = _mm_loadu_si128((__m128i const*)_con2); KS_round(9) KS_round_last(10) keys+=2; first_key+=32; } keys=(ROUND_KEYS *)KEYS; for (i=0;i<test_length;i+=2){ register __m128i block1 = _mm_loadu_si128((__m128i const*)(0*16+PT)); register __m128i block2 = _mm_loadu_si128((__m128i const*)(1*16+PT)); READ_KEYS(0) block1 = _mm_xor_si128(keyA, block1); block2 = _mm_xor_si128(keyB, block2); ENC_round(1) ENC_round(2) ENC_round(3) ENC_round(4) ENC_round(5) ENC_round(6) ENC_round(7) ENC_round(8) ENC_round(9) ENC_round_last(10) _mm_storeu_si128((__m128i *)(CT+0*16), block1); _mm_storeu_si128((__m128i *)(CT+1*16), block2); PT+=32; CT+=32; keys+=2; } }