#include <stdio.h>
#include <cuda_runtime.h>
#include <curand_kernel.h>
#include <sha512.cuh>
#include <ed25519.cuh>
#include <edsign.cuh>
using Address = unsigned char[16];
using Key = unsigned char[32];
struct KeysBox {
    Key PublicKey;
    Key PrivateKey;
};
__device__ static unsigned high = 0x10;
struct ds64 {
    char data[65];
};
struct ds46 {
    char data[46];
};
__device__ ds64 KeyToString(const unsigned char* key) noexcept {
    ds64 str;
    const char* hexDigits = "0123456789abcdef";
    for (unsigned char i = 0; i < 32; i++) {
        str.data[2 * i] = hexDigits[key[i] >> 4];
        str.data[2 * i + 1] = hexDigits[key[i] & 0x0F];
    }
    str.data[64] = '\0';
    return str;
}
__device__ ds46 getAddress(const unsigned char rawAddr[16]) noexcept {
    ds46 addrStr;
    const char* hexDigits = "0123456789abcdef";
    unsigned pos = 0;
    for (unsigned char group = 0; group < 8; group++) {
        int idx = group * 2;
        addrStr.data[pos++] = hexDigits[rawAddr[idx] >> 4];
        addrStr.data[pos++] = hexDigits[rawAddr[idx] & 0x0F];
        addrStr.data[pos++] = hexDigits[rawAddr[idx + 1] >> 4];
        addrStr.data[pos++] = hexDigits[rawAddr[idx + 1] & 0x0F];
        if (group < 7) {
            addrStr.data[pos++] = ':';
        }
    }
    addrStr.data[pos] = '\0';
    return addrStr;
}
__device__ void getRawAddress(int lErase, Key& InvertedPublicKey, Address& rawAddr) noexcept {
    lErase++;
    const int bitsToShift = lErase % 8;
    const int start = lErase / 8;
    if (bitsToShift != 0) {
        for (int i = start; i < start + 15; i++) {
            InvertedPublicKey[i] = static_cast<unsigned char>((InvertedPublicKey[i] << bitsToShift) | (InvertedPublicKey[i + 1] >> (8 - bitsToShift)));
        }
    }
    rawAddr[0] = 0x02;
    rawAddr[1] = static_cast<unsigned char>(lErase - 1);
    memcpy(&rawAddr[2], &InvertedPublicKey[start], 14);
}
__device__ unsigned char zeroCounter(unsigned int x) {
    if (x == 0) return 32;
    return static_cast<unsigned char>(__builtin_clz(x));
}
__device__ unsigned char getZeros(const unsigned char* v) {
    unsigned char leadZeros = 0;
    for (int i = 0; i < 32; i += 4) {
        unsigned word = (static_cast<unsigned>(v[i]) << 24) | (static_cast<unsigned>(v[i + 1]) << 16) | (static_cast<unsigned>(v[i + 2]) << 8) | (static_cast<unsigned>(v[i + 3]));
        if (word == 0) {
            leadZeros += 32;
        } else {
            leadZeros += zeroCounter(word);
            break;
        }
    }
    return leadZeros;
}
__global__ void initRand(curandState* randStates) {
    int id = blockIdx.x * blockDim.x + threadIdx.x;
    curand_init((unsigned long long)clock64() + id, id, 0, &randStates[id]);
}
__device__ unsigned long long xorshift128plus(unsigned long long* state) noexcept {
    unsigned long long x = state[0];
    const unsigned long long y = state[1];
    state[0] = y;
    x ^= x << 23;
    x ^= x >> 17;
    x ^= y ^ (y >> 26);
    state[1] = x;
    return x + y;
}
__device__ void rmbytes(unsigned char* buf, unsigned long size, unsigned long long* state) {
    for (unsigned long i = 0; i < size; i++) {
        buf[i] = xorshift128plus(state) & 0xFF;
    }
}
__device__ void invertKey(const unsigned char* key, unsigned char* inverted) {
    for (int i = 0; i < 32; i++)
        inverted[i] = key[i] ^ 0xFF;
}
__global__ void minerKernel(curandState* randStates) {
    int thid = blockIdx.x * blockDim.x + threadIdx.x;
    curandState localState = randStates[thid];
    unsigned long long xorshiftState[2];
    xorshiftState[0] = curand(&localState);
    xorshiftState[1] = curand(&localState);
    Key seed;
    rmbytes(seed, sizeof(seed), xorshiftState);
    if (thid == 0) printf("Seed: %s\n", KeyToString(seed).data);
    while (true) {
        KeysBox keys;
        ed25519_keygen(keys.PrivateKey, keys.PublicKey, seed);
        if (unsigned zeros = getZeros(keys.PublicKey); zeros > atomicMax(&high, (unsigned)zeros)) {
            printf("\nIPv6:\t%x\nPK:\t%s\nSK:\t%s\n", zeros, KeyToString(keys.PublicKey).data, KeyToString(keys.PrivateKey).data);
        }
        rmbytes(seed, sizeof(seed), xorshiftState);
    }
}
int main() {
    const int threadsPerBlock = 256;
    cudaDeviceProp prop;
    cudaGetDeviceProperties_v2(&prop, 0);
    int mBpSM;
    cudaOccupancyMaxActiveBlocksPerMultiprocessor(&mBpSM, minerKernel, threadsPerBlock, 0);
    int SMs = prop.multiProcessorCount;
    int maxBlocks = mBpSM * SMs;
    const int totalThreads = maxBlocks * threadsPerBlock;
    printf("SMs: %d\n", SMs);
    printf("maxBlocks: %d\n", maxBlocks);
    printf("totalThreads: %d\n", totalThreads);
    printf("MaxBlocksPerSM: %d\n", mBpSM);
    curandState* rst;
    cudaMalloc(&rst, totalThreads * sizeof(curandState));
    initRand<<<100, threadsPerBlock >>>(rst);
    cudaDeviceSynchronize();
    minerKernel<<<100, threadsPerBlock>>>(rst);
    cudaDeviceSynchronize();
    cudaFree(rst);
    return 0;
}