native.intel.aes.aes_common_256b.h Maven / Gradle / Ivy

Go to download
Show more of this group Show more artifacts with this name
Show all versions of bcprov-lts8on Show documentation
The Long Term Stable (LTS) Bouncy Castle Crypto package is a Java implementation of cryptographic algorithms. This jar contains the JCA/JCE provider and low-level API for the BC LTS version 2.73.7 for Java 8 and later.
There is a newer version: 2.73.7
Show newest version
#ifndef BCN_AESCOMMON512_H
#define BCN_AESCOMMON512_H

#include 


static inline void aesenc_16_blocks_256b(
        __m256i *b1, __m256i *b2, __m256i *b3, __m256i *b4,
        __m256i *b5, __m256i *b6, __m256i *b7, __m256i *b8,
        const __m128i *round_keys, const int num_rounds,
        const int num_blocks) {

    const __m256i rk_first = _mm256_broadcastsi128_si256(round_keys[0]);


    *b1 = _mm256_xor_si256(*b1, rk_first);
    if (num_blocks > 2)
        *b2 = _mm256_xor_si256(*b2, rk_first);
    if (num_blocks > 4)
        *b3 = _mm256_xor_si256(*b3, rk_first);
    if (num_blocks > 6)
        *b4 = _mm256_xor_si256(*b4, rk_first);
    if (num_blocks > 8)
        *b5 = _mm256_xor_si256(*b5, rk_first);
    if (num_blocks > 10)
        *b6 = _mm256_xor_si256(*b6, rk_first);
    if (num_blocks > 12)
        *b7 = _mm256_xor_si256(*b7, rk_first);
    if (num_blocks > 14)
        *b8 = _mm256_xor_si256(*b8, rk_first);

    int round;
    for (round = 1; round < num_rounds; round++) {
        const __m256i rk = _mm256_broadcastsi128_si256(round_keys[round]);
        *b1 = _mm256_aesenc_epi128(*b1, rk);
        if (num_blocks > 2)
            *b2 = _mm256_aesenc_epi128(*b2, rk);
        if (num_blocks > 4)
            *b3 = _mm256_aesenc_epi128(*b3, rk);
        if (num_blocks > 6)
            *b4 = _mm256_aesenc_epi128(*b4, rk);
        if (num_blocks > 8)
            *b5 = _mm256_aesenc_epi128(*b5, rk);
        if (num_blocks > 10)
            *b6 = _mm256_aesenc_epi128(*b6, rk);
        if (num_blocks > 12)
            *b7 = _mm256_aesenc_epi128(*b7, rk);
        if (num_blocks > 14)
            *b8 = _mm256_aesenc_epi128(*b8, rk);

    }


    const __m256i rk_last = _mm256_broadcastsi128_si256(round_keys[round]);

    *b1 = _mm256_aesenclast_epi128(*b1, rk_last);

    if (num_blocks > 2)
        *b2 = _mm256_aesenclast_epi128(*b2, rk_last);
    if (num_blocks > 4)
        *b3 = _mm256_aesenclast_epi128(*b3, rk_last);
    if (num_blocks > 6)
        *b4 = _mm256_aesenclast_epi128(*b4, rk_last);
    if (num_blocks > 8)
        *b5 = _mm256_aesenclast_epi128(*b5, rk_last);
    if (num_blocks > 10)
        *b6 = _mm256_aesenclast_epi128(*b6, rk_last);
    if (num_blocks > 12)
        *b7 = _mm256_aesenclast_epi128(*b7, rk_last);
    if (num_blocks > 14)
        *b8 = _mm256_aesenclast_epi128(*b8, rk_last);

}


static inline void aesdec_16_blocks_256b(
        __m256i *b1, __m256i *b2, __m256i *b3, __m256i *b4,
        __m256i *b5, __m256i *b6, __m256i *b7, __m256i *b8,
        const __m128i *round_keys, const int num_rounds,
        const int num_blocks) {

    const __m256i rk_first = _mm256_broadcastsi128_si256(round_keys[0]);

    *b1 = _mm256_xor_si256(*b1, rk_first);

    if (num_blocks > 2)
        *b2 = _mm256_xor_si256(*b2, rk_first);
    if (num_blocks > 4)
        *b3 = _mm256_xor_si256(*b3, rk_first);
    if (num_blocks > 6)
        *b4 = _mm256_xor_si256(*b4, rk_first);
    if (num_blocks > 8)
        *b5 = _mm256_xor_si256(*b5, rk_first);
    if (num_blocks > 10)
        *b6 = _mm256_xor_si256(*b6, rk_first);
    if (num_blocks > 12)
        *b7 = _mm256_xor_si256(*b7, rk_first);
    if (num_blocks > 14)
        *b8 = _mm256_xor_si256(*b8, rk_first);

    int round;
    for (round = 1; round < num_rounds; round++) {
        const __m256i rk = _mm256_broadcastsi128_si256(round_keys[round]);
        *b1 = _mm256_aesdec_epi128(*b1, rk);
        if (num_blocks > 2)
            *b2 = _mm256_aesdec_epi128(*b2, rk);
        if (num_blocks > 4)
            *b3 = _mm256_aesdec_epi128(*b3, rk);
        if (num_blocks > 6)
            *b4 = _mm256_aesdec_epi128(*b4, rk);
        if (num_blocks > 8)
            *b5 = _mm256_aesdec_epi128(*b5, rk);
        if (num_blocks > 10)
            *b6 = _mm256_aesdec_epi128(*b6, rk);
        if (num_blocks > 12)
            *b7 = _mm256_aesdec_epi128(*b7, rk);
        if (num_blocks > 14)
            *b8 = _mm256_aesdec_epi128(*b8, rk);
    }


    const __m256i rk_last = _mm256_broadcastsi128_si256(round_keys[round]);

    *b1 = _mm256_aesdeclast_epi128(*b1, rk_last);

    if (num_blocks > 2)
        *b2 = _mm256_aesdeclast_epi128(*b2, rk_last);
    if (num_blocks > 4)
        *b3 = _mm256_aesdeclast_epi128(*b3, rk_last);
    if (num_blocks > 6)
        *b4 = _mm256_aesdeclast_epi128(*b4, rk_last);
    if (num_blocks > 8)
        *b5 = _mm256_aesdeclast_epi128(*b5, rk_last);
    if (num_blocks > 10)
        *b6 = _mm256_aesdeclast_epi128(*b6, rk_last);
    if (num_blocks > 12)
        *b7 = _mm256_aesdeclast_epi128(*b7, rk_last);
    if (num_blocks > 14)
        *b8 = _mm256_aesdeclast_epi128(*b8, rk_last);

}


#endif // BCN_AESCOMMON512_H