ccminer/m7/cuda_mul.cu

/*
 * tiger-192 djm34
 *
 */

/*
 * tiger-192 kernel implementation.
 *
 * ==========================(LICENSE BEGIN)============================
 *
 * Copyright (c) 2014  djm34
 *
 * Permission is hereby granted, free of charge, to any person obtaining
 * a copy of this software and associated documentation files (the
 * "Software"), to deal in the Software without restriction, including
 * without limitation the rights to use, copy, modify, merge, publish,
 * distribute, sublicense, and/or sell copies of the Software, and to
 * permit persons to whom the Software is furnished to do so, subject to
 * the following conditions:
 *
 * The above copyright notice and this permission notice shall be
 * included in all copies or substantial portions of the Software.
 *
 * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
 * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
 * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.
 * IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY
 * CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
 * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
 * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
 *
 * ===========================(LICENSE END)=============================
 *
 * @author   phm <phm@inbox.com>
 */
//#include <stdio.h>
#include <memory.h>

#include "cuda_helper.h"

#define HIWORD _HIWORD
#define LOWORD _LOWORD

#if 0
#define gpuErrchk(ans) { gpuAssert((ans), __FILE__, __LINE__); }
inline void gpuAssert(cudaError_t code, char *file, int line, bool abort=true)
{
   if (code != cudaSuccess)
   {
	  fprintf(stderr,"GPUassert: %s %s %d\n", cudaGetErrorString(code), file, line);
	  if (abort) exit(code);
   }
}
#endif

extern cudaError_t MyStreamSynchronize(cudaStream_t stream, int situation, int thr_id);

__device__ __forceinline__
void bigmul(uint64_t *w, uint64_t* am, uint64_t* bm, int sizea, int sizeb, int thread)
{
	int threads = 256*256*8*2;
	#pragma unroll
	for (int i=0;i<sizea+sizeb;i++) {w[i*threads+thread]=0;}
	#pragma unroll
	for (int i=0;i<sizeb;i++)
	{
		uint64_t c=0;
		uint64_t u=0,v=0;
		#pragma unroll
		for (int j=0;j<sizea;j++) {
			muladd128(u,v,am[j*threads+thread],bm[i*threads+thread],w[(i+j)*threads+thread],c);
			w[(i+j)*threads+thread]=v;
			c=u;
		}
		w[(i+sizea)*threads+thread]=u;
	}
}

__global__
void m7_bigmul1_gpu(int threads, int sizea, int sizeb, uint64_t* am, uint64_t* bm, uint64_t *w)
{
	int thread = (blockDim.x * blockIdx.x + threadIdx.x);
	if (thread < threads)
	{
		#pragma unroll
		for (int i=0;i<sizea+sizeb;i++) {w[i*threads+thread]=0;}
		#pragma unroll
		for (int i=0;i<sizeb;i++) {
			uint64_t c=0;
			uint64_t u=0,v=0;
			#pragma unroll
			for (int j=0;j<sizea;j++) {
				muladd128(u,v,am[j*threads+thread],bm[i*threads+thread],w[(i+j)*threads+thread],c);
				w[(i+j)*threads+thread]=v;
				c=u;
			}
			w[(i+sizea)*threads+thread]=u;
		}
	} // thread
}

__global__
void m7_bigmul_unroll1_gpu(int threads, uint64_t* am, uint64_t* bm, uint64_t *w)
{
	int thread = (blockDim.x * blockIdx.x + threadIdx.x);
	if (thread < threads)
	{
		#pragma unroll 32
		for (int i=0;i<32;i++) {
			w[i*threads + thread]=0;
		}

		#if __CUDA_ARCH__ < 500
		#pragma unroll 32
		#endif
		for (int i=0;i<32;i++)
		{
			uint64_t c=0;
			uint64_t u=0,v=0;
			#pragma unroll 3
			for (int j=0;j<3;j++) {
				muladd128(u,v,am[j*threads+thread],bm[i*threads+thread],w[(i+j)*threads+thread],c);
				w[(i+j)*threads+thread]=v;
				c=u;
			}
			w[(i+3)*threads+thread]=u;
		}
	} // threads
}

__global__
void m7_bigmul_unroll1_gpu_std(int threads, uint64_t* amg, uint64_t* bmg, uint64_t *wg)
{
	int thread = (blockDim.x * blockIdx.x + threadIdx.x);
	if (thread < threads)
	{
		uint64_t * am = amg + 8*thread;
		uint64_t * bm = bmg + 38*thread;
		uint64_t * w  = wg +  38*thread;

		#pragma unroll 32
		for (int i=0;i<32;i++) {
			w[i]=0;
		}

		#if __CUDA_ARCH__ < 500
		#pragma unroll 32
		#endif
		for (int i=0;i<32;i++)
		{
			uint64_t c=0;
			uint64_t u=0,v=0;
			#pragma unroll 3
			for (int j=0;j<3;j++) {
				muladd128(u,v,am[j],bm[i],w[(i+j)],c);
				w[(i+j)]=v;
				c=u;
			}
			w[(i+3)]=u;
		}
	} // threads
}

__global__
void m7_bigmul_unroll2_gpu(int threads, uint64_t* am, uint64_t* bm, uint64_t *w)
{
	int thread = (blockDim.x * blockIdx.x + threadIdx.x);
	if (thread < threads)
	{
		#if __CUDA_ARCH__ < 500
		#pragma unroll
		#endif
		for (int i=0;i<38;i++) {
			w[i*threads+thread]=0;
		}

		#if __CUDA_ARCH__ < 500
		#pragma unroll
		#endif
		for (int i=0;i<35;i++)
		{
			uint64_t c=0;
			uint64_t u=0,v=0;
			#if __CUDA_ARCH__ < 500
			#pragma unroll
			#endif
			for (int j=0;j<3;j++) {
				muladd128(u,v,am[j*threads+thread],bm[i*threads+thread],w[(i+j)*threads+thread],c);
				w[(i+j)*threads+thread]=v;
				c=u;
			}
			w[(i+3)*threads+thread]=u;
		}
	} // thread
}

__global__
void m7_bigmul_unroll2_gpu_std(int threads, uint64_t* amg, uint64_t* bmg, uint64_t *wg)
{
	int thread = (blockDim.x * blockIdx.x + threadIdx.x);

	if (thread < threads)
	{
		uint64_t * am = amg + 8*thread;
		uint64_t * bm = bmg + 38*thread;
		uint64_t * w  = wg +  38*thread;

		#if __CUDA_ARCH__ < 500
		#pragma unroll
		#endif
		for (int i=0;i<38;i++) {
			w[i]=0;
		}

		#if __CUDA_ARCH__ < 500
		#pragma unroll
		#endif
		for (int i=0;i<35;i++)
		{
			uint64_t c=0;
			uint64_t u=0,v=0;
			#if __CUDA_ARCH__ < 500
			#pragma unroll
			#endif
			for (int j=0;j<3;j++) {
				muladd128(u,v,am[j],bm[i],w[(i+j)],c);
				w[(i+j)]=v;
				c=u;
			}
			w[(i+3)]=u;
		}
	} // thread
}

__host__ void m7_bigmul1_cpu(int thr_id, int threads,int len1,int len2,uint64_t* Hash1, uint64_t* Hash2,uint64_t *finalHash,int order)
{
	const int threadsperblock = 256;

	dim3 grid((threads + threadsperblock-1)/threadsperblock);
	dim3 block(threadsperblock);

	size_t shared_size =0;
	m7_bigmul1_gpu<<<grid, block, shared_size>>>(threads,len1,len2,Hash1,Hash2,finalHash);

//	MyStreamSynchronize(NULL, order, thr_id);

//	gpuErrchk(cudaDeviceSynchronize());
//	gpuErrchk(cudaThreadSynchronize());
}

__host__ void m7_bigmul_unroll1_cpu(int thr_id, int threads,uint64_t* Hash1, uint64_t* Hash2,uint64_t *finalHash,int order)
{
	const int threadsperblock = 256;

	dim3 grid((threads + threadsperblock-1)/threadsperblock);
	dim3 block(threadsperblock);

	size_t shared_size =0;
	m7_bigmul_unroll1_gpu<<<grid, block, shared_size>>>(threads,Hash1,Hash2,finalHash);
}

__host__ void m7_bigmul_unroll2_cpu(int thr_id, int threads,uint64_t* Hash1, uint64_t* Hash2,uint64_t *finalHash,int order)
{
	const int threadsperblock = 256;

	dim3 grid((threads + threadsperblock-1)/threadsperblock);
	dim3 block(threadsperblock);

	size_t shared_size =0;
	m7_bigmul_unroll2_gpu<<<grid, block, shared_size>>>(threads,Hash1,Hash2,finalHash);
}

__host__ void m7_bigmul_init(int thr_id, int threads)
{
	// why I am here ?
}