/*
 * X14 algorithm
 * Added in ccminer by Tanguy Pruvot - 2014
 */

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 "sph/sph_hamsi.h"
#include "sph/sph_fugue.h"
#include "sph/sph_shabal.h"
}

#include "miner.h"

#include "cuda_helper.h"
#include "x11/cuda_x11.h"

// Memory for the hash functions
static uint32_t *d_hash[MAX_GPUS] = { 0 };

extern void x13_hamsi512_cpu_init(int thr_id, uint32_t threads);
extern void x13_hamsi512_cpu_hash_64(int thr_id, uint32_t threads, uint32_t startNounce, uint32_t *d_nonceVector, uint32_t *d_hash, int order);

extern void x13_fugue512_cpu_init(int thr_id, uint32_t threads);
extern void x13_fugue512_cpu_hash_64(int thr_id, uint32_t threads, uint32_t startNounce, uint32_t *d_nonceVector, uint32_t *d_hash, int order);
extern void x13_fugue512_cpu_free(int thr_id);

extern void x14_shabal512_cpu_init(int thr_id, uint32_t threads);
extern void x14_shabal512_cpu_hash_64(int thr_id, uint32_t threads, uint32_t startNounce, uint32_t *d_nonceVector, uint32_t *d_hash, int order);


// X14 CPU Hash function
extern "C" void x14hash(void *output, const void *input)
{
	unsigned char hash[128]; // uint32_t hashA[16], hashB[16];
	#define hashB hash+64

	memset(hash, 0, sizeof hash);

	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;
	sph_hamsi512_context ctx_hamsi;
	sph_fugue512_context ctx_fugue;
	sph_shabal512_context ctx_shabal;

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

	sph_bmw512_init(&ctx_bmw);
	sph_bmw512(&ctx_bmw, hash, 64);
	sph_bmw512_close(&ctx_bmw, hashB);

	sph_groestl512_init(&ctx_groestl);
	sph_groestl512(&ctx_groestl, hashB, 64);
	sph_groestl512_close(&ctx_groestl, hash);

	sph_skein512_init(&ctx_skein);
	sph_skein512(&ctx_skein, hash, 64);
	sph_skein512_close(&ctx_skein, hashB);

	sph_jh512_init(&ctx_jh);
	sph_jh512(&ctx_jh, hashB, 64);
	sph_jh512_close(&ctx_jh, hash);

	sph_keccak512_init(&ctx_keccak);
	sph_keccak512(&ctx_keccak, hash, 64);
	sph_keccak512_close(&ctx_keccak, hashB);

	sph_luffa512_init(&ctx_luffa);
	sph_luffa512(&ctx_luffa, hashB, 64);
	sph_luffa512_close(&ctx_luffa, hash);

	sph_cubehash512_init(&ctx_cubehash);
	sph_cubehash512(&ctx_cubehash, hash, 64);
	sph_cubehash512_close(&ctx_cubehash, hashB);

	sph_shavite512_init(&ctx_shavite);
	sph_shavite512(&ctx_shavite, hashB, 64);
	sph_shavite512_close(&ctx_shavite, hash);

	sph_simd512_init(&ctx_simd);
	sph_simd512(&ctx_simd, hash, 64);
	sph_simd512_close(&ctx_simd, hashB);

	sph_echo512_init(&ctx_echo);
	sph_echo512(&ctx_echo, hashB, 64);
	sph_echo512_close(&ctx_echo, hash);

	sph_hamsi512_init(&ctx_hamsi);
	sph_hamsi512(&ctx_hamsi, hash, 64);
	sph_hamsi512_close(&ctx_hamsi, hashB);

	sph_fugue512_init(&ctx_fugue);
	sph_fugue512(&ctx_fugue, hashB, 64);
	sph_fugue512_close(&ctx_fugue, hash);

	sph_shabal512_init(&ctx_shabal);
	sph_shabal512(&ctx_shabal, hash, 64);
	sph_shabal512_close(&ctx_shabal, hash);

	memcpy(output, hash, 32);
}

static bool init[MAX_GPUS] = { 0 };

extern "C" int scanhash_x14(int thr_id,  struct work* work, uint32_t max_nonce, unsigned long *hashes_done)
{
	uint32_t *pdata = work->data;
	uint32_t *ptarget = work->target;
	const uint32_t first_nonce = pdata[19];
	uint32_t endiandata[20];

	uint32_t throughput =  cuda_default_throughput(thr_id, 1U << 19); // 19=256*256*8;
	//if (init[thr_id]) throughput = min(throughput, max_nonce - first_nonce);

	if (opt_benchmark)
		ptarget[7] = 0x000f;

	if (!init[thr_id])
	{
		cudaSetDevice(device_map[thr_id]);
		if (opt_cudaschedule == -1 && gpu_threads == 1) {
			cudaDeviceReset();
			// reduce cpu usage
			cudaSetDeviceFlags(cudaDeviceScheduleBlockingSync);
			CUDA_LOG_ERROR();
		}
		gpulog(LOG_INFO, thr_id, "Intensity set to %g, %u cuda threads", throughput2intensity(throughput), throughput);

		quark_blake512_cpu_init(thr_id, throughput);
		quark_groestl512_cpu_init(thr_id, throughput);
		quark_skein512_cpu_init(thr_id, throughput);
		quark_bmw512_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_simd512_cpu_init(thr_id, throughput);
		x11_echo512_cpu_init(thr_id, throughput);
		x13_hamsi512_cpu_init(thr_id, throughput);
		x13_fugue512_cpu_init(thr_id, throughput);
		x14_shabal512_cpu_init(thr_id, throughput);

		cuda_check_cpu_init(thr_id, throughput);

		cudaMalloc(&d_hash[thr_id], (size_t) 64 * throughput);

		CUDA_LOG_ERROR();

		init[thr_id] = true;
	}

	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;
		quark_blake512_cpu_hash_80(thr_id, throughput, pdata[19], d_hash[thr_id]); order++;
		quark_bmw512_cpu_hash_64(thr_id, throughput, pdata[19], NULL, d_hash[thr_id], order++);
		quark_groestl512_cpu_hash_64(thr_id, throughput, pdata[19], NULL, d_hash[thr_id], order++);
		quark_skein512_cpu_hash_64(thr_id, throughput, pdata[19], NULL, d_hash[thr_id], order++);
		quark_jh512_cpu_hash_64(thr_id, throughput, pdata[19], NULL, d_hash[thr_id], order++);
		quark_keccak512_cpu_hash_64(thr_id, throughput, pdata[19], NULL, d_hash[thr_id], order++);
		x11_luffaCubehash512_cpu_hash_64(thr_id, throughput, d_hash[thr_id], order++);
		x11_shavite512_cpu_hash_64(thr_id, throughput, pdata[19], NULL, d_hash[thr_id], order++);
		x11_simd512_cpu_hash_64(thr_id, throughput, pdata[19], NULL, d_hash[thr_id], order++);
		x11_echo512_cpu_hash_64(thr_id, throughput, pdata[19], NULL, d_hash[thr_id], order++);
		x13_hamsi512_cpu_hash_64(thr_id, throughput, pdata[19], NULL, d_hash[thr_id], order++);
		x13_fugue512_cpu_hash_64(thr_id, throughput, pdata[19], NULL, d_hash[thr_id], order++);
		x14_shabal512_cpu_hash_64(thr_id, throughput, pdata[19], NULL, d_hash[thr_id], order++);

		CUDA_LOG_ERROR();

		*hashes_done = pdata[19] - first_nonce + throughput;

		work->nonces[0] = cuda_check_hash(thr_id, throughput, pdata[19], d_hash[thr_id]);

		if (work->nonces[0] != UINT32_MAX)
		{
			const uint32_t Htarg = ptarget[7];
			uint32_t _ALIGN(64) vhash[8];
			/* check now with the CPU to confirm */
			be32enc(&endiandata[19], work->nonces[0]);
			x14hash(vhash, endiandata);

			if (vhash[7] <= Htarg && fulltest(vhash, ptarget)) {
				work->valid_nonces = 1;
				work->nonces[1] = cuda_check_hash_suppl(thr_id, throughput, pdata[19], d_hash[thr_id], 1);
				work_set_target_ratio(work, vhash);
				if (work->nonces[1] != 0) {
					be32enc(&endiandata[19], work->nonces[1]);
					x14hash(vhash, endiandata);
					bn_set_target_ratio(work, vhash, 1);
					work->valid_nonces++;
					pdata[19] = max(work->nonces[0], work->nonces[1]) + 1;
				} else {
					pdata[19] = work->nonces[0] + 1; // cursor
				}
				return work->valid_nonces;
			}
			else if (vhash[7] > Htarg) {
				gpu_increment_reject(thr_id);
				if (!opt_quiet)
				gpulog(LOG_WARNING, thr_id, "result for %08x does not validate on CPU!", work->nonces[0]);
				pdata[19] = work->nonces[0] + 1;
				continue;
			}
		}

		if ((uint64_t)throughput + pdata[19] >= max_nonce) {
			pdata[19] = max_nonce;
			break;
		}

		pdata[19] += throughput;

	} while (!work_restart[thr_id].restart);

	CUDA_LOG_ERROR();

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

// cleanup
extern "C" void free_x14(int thr_id)
{
	if (!init[thr_id])
		return;

	cudaThreadSynchronize();

	quark_blake512_cpu_free(thr_id);
	quark_groestl512_cpu_free(thr_id);
	x11_simd512_cpu_free(thr_id);
	x13_fugue512_cpu_free(thr_id);

	cudaFree(d_hash[thr_id]);
	d_hash[thr_id] = NULL;

	cuda_check_cpu_free(thr_id);

	cudaDeviceSynchronize();
	init[thr_id] = false;
}