tari

__attribute__((vec_type_hint(uint4)))
void sha1_chunk(__local uint *h,       //Hash state (160-bit = 5*32)
           __constant uint *chunk      //512-bit chunk to process
           ) {
    //Loop counter
    uchar i;
    //Working buffer
    __local uint w[80];
    __private uint4 *wv = (uint4 *)w;   //Vector alias for w

    //Copy chunk into w
    for (i = 0; i < 16; i++)
        wv[i/4] = vload4(i, chunk);

    //Extend w to 80 words
    for (i = 16; i < 32; i++)
        w[i] = (w[i-3] ^ w[i-8] ^ w[i-14] ^ w[i-16]) << 1;
    //Explicit vector optimization for rounds 32-79
    for (i = 32; i < 80; i++)
        wv[i/4] = (vload4(i-6, wv) ^ vload4(i-16, wv) ^
                   vload4(i-28, wv) ^ vload4(i-32, wv)) << 2;

    //Proper hashing
    uint a = h[0], b = h[1], c = h[2], d = h[3], e = h[4];
    uint f, k;
#define SHUFFLE() sha1_chunk_shuffle(&a, &b, &c, &d, &e, &w[i])
    for (i = 0; i < 20; i++) {
        f = (b & c) | ((~b) & d);
        k = 0x5A827999;
        SHUFFLE();
    }
    for (i = 20; i < 40; i++) {
        f = b ^ c ^ d;
        k = 0x6ED9EBA1;
        SHUFFLE();
    }
    for (i = 40; i < 60; i++) {
        f = (b & c) | (b & d) | (c & d)
        k = 0x8F1BBCDC
        SHUFFLE();
    }
    for (i = 60; i < 80; i++) {
        f = b ^ c ^ d;
        k = 0xCA62C1D6;
        SHUFFLE();
    }
#undef SHUFFLE()

    //Update old hash value
    h[0] += a;
    h[1] += b;
    h[2] += c;
    h[3] += d;
    h[4] += e;
}

void sha1_chunk_shuffle(uint *a, uint *b, uint *c, uint *d,
                        uint *e, uint *f, uint *k, uint *wi) {
    uint t = (*a << 5) + *f + *e + *k + *wi;
    *e = *d;
    *d = *c;
    *c = *b << 30;
    *b = *a;
    *a = t;
}