extern "C"
{
#include "sph/sph_blake.h"
#include "sph/sph_bmw.h"
#include "sph/sph_groestl.h"
#include "sph/sph_skein.h"
#include "sph/sph_jh.h"
#include "sph/sph_keccak.h"

#include "sph/sph_luffa.h"
#include "sph/sph_cubehash.h"
#include "sph/sph_shavite.h"
#include "sph/sph_simd.h"
#include "sph/sph_echo.h"
}

#include "miner.h"
#include "cuda_helper.h"

#include <stdio.h>
#include <memory.h>

static uint32_t *d_hash[MAX_GPUS];

extern void quark_blake512_cpu_init(int thr_id, uint32_t threads);
extern void quark_blake512_cpu_setBlock_80(int thr_id, uint32_t *pdata);
extern void quark_blake512_cpu_hash_80(int thr_id, uint32_t threads, uint32_t startNounce, uint32_t *d_hash);

extern void quark_bmw512_cpu_init(int thr_id, uint32_t threads);
extern void quark_bmw512_cpu_hash_64(int thr_id, uint32_t threads, uint32_t startNounce, uint32_t *d_nonceVector, uint32_t *d_hash, int order);

extern void quark_groestl512_cpu_init(int thr_id, uint32_t threads);
extern void quark_groestl512_cpu_hash_64(int thr_id, uint32_t threads, uint32_t startNounce, uint32_t *d_nonceVector, uint32_t *d_hash, int order);
extern void quark_doublegroestl512_cpu_hash_64(int thr_id, uint32_t threads, uint32_t startNounce, uint32_t *d_nonceVector, uint32_t *d_hash, int order);

extern void quark_skein512_cpu_init(int thr_id, uint32_t threads);
extern void quark_skein512_cpu_hash_64(int thr_id, uint32_t threads, uint32_t startNounce, uint32_t *d_nonceVector, uint32_t *d_hash, int order);

extern void quark_keccak512_cpu_init(int thr_id, uint32_t threads);
extern void quark_keccak512_cpu_hash_64(int thr_id, uint32_t threads, uint32_t startNounce, uint32_t *d_nonceVector, uint32_t *d_hash, int order);

extern void quark_jh512_cpu_init(int thr_id, uint32_t threads);
extern void quark_jh512_cpu_hash_64(int thr_id, uint32_t threads, uint32_t startNounce, uint32_t *d_nonceVector, uint32_t *d_hash, int order);

extern void x11_luffaCubehash512_cpu_init(int thr_id, uint32_t threads);
extern void x11_luffaCubehash512_cpu_hash_64(int thr_id, uint32_t threads, uint32_t startNounce, uint32_t *d_nonceVector, uint32_t *d_hash, int order);

extern void x11_shavite512_cpu_init(int thr_id, uint32_t threads);
extern void x11_shavite512_cpu_hash_64(int thr_id, uint32_t threads, uint32_t startNounce, uint32_t *d_nonceVector, uint32_t *d_hash, int order);

extern int  x11_simd512_cpu_init(int thr_id, uint32_t threads);
extern void x11_simd512_cpu_hash_64(int thr_id, uint32_t threads, uint32_t startNounce, uint32_t *d_nonceVector, uint32_t *d_hash, int order);

extern void x11_echo512_cpu_init(int thr_id, uint32_t threads);
extern void x11_echo512_cpu_hash_64(int thr_id, uint32_t threads, uint32_t startNounce, uint32_t *d_nonceVector, uint32_t *d_hash, int order);

extern void quark_compactTest_cpu_init(int thr_id, uint32_t threads);
extern void quark_compactTest_cpu_hash_64(int thr_id, uint32_t threads, uint32_t startNounce, uint32_t *inpHashes,
                                          uint32_t *d_noncesTrue, size_t *nrmTrue, uint32_t *d_noncesFalse, size_t *nrmFalse, int order);

// X11 CPU Hash
extern "C" void x11hash(void *output, const void *input)
{
	// blake1-bmw2-grs3-skein4-jh5-keccak6-luffa7-cubehash8-shavite9-simd10-echo11

	sph_blake512_context ctx_blake;
	sph_bmw512_context ctx_bmw;
	sph_groestl512_context ctx_groestl;
	sph_jh512_context ctx_jh;
	sph_keccak512_context ctx_keccak;
	sph_skein512_context ctx_skein;
	sph_luffa512_context ctx_luffa;
	sph_cubehash512_context ctx_cubehash;
	sph_shavite512_context ctx_shavite;
	sph_simd512_context ctx_simd;
	sph_echo512_context ctx_echo;

	unsigned char hash[128];
	memset(hash, 0, sizeof hash);

	sph_blake512_init(&ctx_blake);
	sph_blake512 (&ctx_blake, input, 80);
	sph_blake512_close(&ctx_blake, (void*) hash);

	sph_bmw512_init(&ctx_bmw);
	sph_bmw512 (&ctx_bmw, (const void*) hash, 64);
	sph_bmw512_close(&ctx_bmw, (void*) hash);

	sph_groestl512_init(&ctx_groestl);
	sph_groestl512 (&ctx_groestl, (const void*) hash, 64);
	sph_groestl512_close(&ctx_groestl, (void*) hash);

	sph_skein512_init(&ctx_skein);
	sph_skein512 (&ctx_skein, (const void*) hash, 64);
	sph_skein512_close(&ctx_skein, (void*) hash);

	sph_jh512_init(&ctx_jh);
	sph_jh512 (&ctx_jh, (const void*) hash, 64);
	sph_jh512_close(&ctx_jh, (void*) hash);

	sph_keccak512_init(&ctx_keccak);
	sph_keccak512 (&ctx_keccak, (const void*) hash, 64);
	sph_keccak512_close(&ctx_keccak, (void*) hash);

	sph_luffa512_init(&ctx_luffa);
	sph_luffa512 (&ctx_luffa, (const void*) hash, 64);
	sph_luffa512_close (&ctx_luffa, (void*) hash);

	sph_cubehash512_init(&ctx_cubehash);
	sph_cubehash512 (&ctx_cubehash, (const void*) hash, 64);
	sph_cubehash512_close(&ctx_cubehash, (void*) hash);

	sph_shavite512_init(&ctx_shavite);
	sph_shavite512 (&ctx_shavite, (const void*) hash, 64);
	sph_shavite512_close(&ctx_shavite, (void*) hash);

	sph_simd512_init(&ctx_simd);
	sph_simd512 (&ctx_simd, (const void*) hash, 64);
	sph_simd512_close(&ctx_simd, (void*) hash);

	sph_echo512_init(&ctx_echo);
	sph_echo512 (&ctx_echo, (const void*) hash, 64);
	sph_echo512_close(&ctx_echo, (void*) hash);

	memcpy(output, hash, 32);
}

#ifdef _DEBUG
#define TRACE(algo) { \
	if (max_nonce == 1 && pdata[19] <= 1) { \
		uint32_t* debugbuf = NULL; \
		cudaMallocHost(&debugbuf, 8*sizeof(uint32_t)); \
		cudaMemcpy(debugbuf, d_hash[thr_id], 8*sizeof(uint32_t), cudaMemcpyDeviceToHost); \
		printf("X11 %s %08x %08x %08x %08x...\n", algo, swab32(debugbuf[0]), swab32(debugbuf[1]), \
			swab32(debugbuf[2]), swab32(debugbuf[3])); \
		cudaFreeHost(debugbuf); \
	} \
}
#else
#define TRACE(algo) {}
#endif

static bool init[MAX_GPUS] = { 0 };

extern "C" int scanhash_x11(int thr_id, uint32_t *pdata,
    const uint32_t *ptarget, uint32_t max_nonce,
    unsigned long *hashes_done)
{
	const uint32_t first_nonce = pdata[19];
	int intensity = (device_sm[device_map[thr_id]] >= 500 && !is_windows()) ? 20 : 19;
	uint32_t throughput = device_intensity(thr_id, __func__, 1U << intensity); // 19=256*256*8;
	throughput = min(throughput, max_nonce - first_nonce);

	if (opt_benchmark)
		((uint32_t*)ptarget)[7] = 0x5;

	if (!init[thr_id])
	{
		cudaSetDevice(device_map[thr_id]);

		quark_blake512_cpu_init(thr_id, throughput);
		quark_bmw512_cpu_init(thr_id, throughput);
		quark_groestl512_cpu_init(thr_id, throughput);
		quark_skein512_cpu_init(thr_id, throughput);
		quark_keccak512_cpu_init(thr_id, throughput);
		quark_jh512_cpu_init(thr_id, throughput);
		x11_luffaCubehash512_cpu_init(thr_id, throughput);
		x11_shavite512_cpu_init(thr_id, throughput);
		x11_echo512_cpu_init(thr_id, throughput);
		if (x11_simd512_cpu_init(thr_id, throughput) != 0) {
			return 0;
		}
		CUDA_CALL_OR_RET_X(cudaMalloc(&d_hash[thr_id], 64 * throughput), 0); // why 64 ?

		cuda_check_cpu_init(thr_id, throughput);

		init[thr_id] = true;
	}

	uint32_t endiandata[20];
	for (int k=0; k < 20; k++)
		be32enc(&endiandata[k], pdata[k]);

	quark_blake512_cpu_setBlock_80(thr_id, endiandata);
	cuda_check_cpu_setTarget(ptarget);

	do {
		int order = 0;
		uint32_t foundNonce;

		// Hash with CUDA
		quark_blake512_cpu_hash_80(thr_id, throughput, pdata[19], d_hash[thr_id]); order++;
		TRACE("blake  :");
		quark_bmw512_cpu_hash_64(thr_id, throughput, pdata[19], NULL, d_hash[thr_id], order++);
		TRACE("bmw    :");
		quark_groestl512_cpu_hash_64(thr_id, throughput, pdata[19], NULL, d_hash[thr_id], order++);
		TRACE("groestl:");
		quark_skein512_cpu_hash_64(thr_id, throughput, pdata[19], NULL, d_hash[thr_id], order++);
		TRACE("skein  :");
		quark_jh512_cpu_hash_64(thr_id, throughput, pdata[19], NULL, d_hash[thr_id], order++);
		TRACE("jh512  :");
		quark_keccak512_cpu_hash_64(thr_id, throughput, pdata[19], NULL, d_hash[thr_id], order++);
		TRACE("keccak :");
		x11_luffaCubehash512_cpu_hash_64(thr_id, throughput, pdata[19], NULL, d_hash[thr_id], order++);
		TRACE("luffa+c:");
		x11_shavite512_cpu_hash_64(thr_id, throughput, pdata[19], NULL, d_hash[thr_id], order++);
		TRACE("shavite:");
		x11_simd512_cpu_hash_64(thr_id, throughput, pdata[19], NULL, d_hash[thr_id], order++);
		TRACE("simd   :");
		x11_echo512_cpu_hash_64(thr_id, throughput, pdata[19], NULL, d_hash[thr_id], order++);
		TRACE("echo => ");

		foundNonce = cuda_check_hash(thr_id, throughput, pdata[19], d_hash[thr_id]);
		if (foundNonce != UINT32_MAX)
		{
			const uint32_t Htarg = ptarget[7];
			uint32_t vhash64[8];
			be32enc(&endiandata[19], foundNonce);
			x11hash(vhash64, endiandata);

			if (vhash64[7] <= Htarg && fulltest(vhash64, ptarget)) {
				int res = 1;
				// check if there was some other ones...
				uint32_t secNonce = cuda_check_hash_suppl(thr_id, throughput, pdata[19], d_hash[thr_id], 1);
				*hashes_done = pdata[19] - first_nonce + throughput;
				if (secNonce != 0) {
					pdata[21] = secNonce;
					res++;
				}
				pdata[19] = foundNonce;
				return res;
			} else {
				applog(LOG_WARNING, "GPU #%d: result for %08x does not validate on CPU!", device_map[thr_id], foundNonce);
			}
		}

		pdata[19] += throughput;

	} while (pdata[19] < max_nonce && !work_restart[thr_id].restart);

	*hashes_done = pdata[19] - first_nonce + 1;
	return 0;
}