serial_no
int64
1
24.2k
cuda_source
stringlengths
11
9.01M
1,201
#include <cstdlib> #include <cstdio> #include <cmath> #include <cassert> #define SIZE 32 __global__ void matrix_add(float** d_A, float** d_B, float** d_C, size_t size) { size_t i = threadIdx.x + blockDim.x*blockIdx.x; size_t j = threadIdx.y + blockDim.y*blockIdx.y; printf("i: %d, j: %d, d_A[i][j]: %f\n", i, j, d_A[i][j]); /* if(i < size && j < size) */ /* d_C[i][j] = d_A[i][j] + d_B[i][j]; */ } void print_err_msg(cudaError_t err) { if(err != cudaSuccess) { printf("%s in %s at line %d\n", cudaGetErrorString(err), __FILE__, __LINE__); exit(EXIT_FAILURE); } } int main() { float** h_A; float** h_B; float** h_C; float** d_A; float** d_B; float** d_C; float** s_A; float** s_B; float** s_C; h_A = new float*[SIZE]; h_B = new float*[SIZE]; h_C = new float*[SIZE]; s_A = (float **)malloc(SIZE*sizeof(float*)); s_B = (float **)malloc(SIZE*sizeof(float*)); s_C = (float **)malloc(SIZE*sizeof(float*)); cudaError_t err; err = cudaMalloc((void**) &d_A, SIZE*sizeof(float*)); print_err_msg(err); err = cudaMalloc((void**) &d_B, SIZE*sizeof(float*)); print_err_msg(err); err = cudaMalloc((void**) &d_C, SIZE*sizeof(float*)); print_err_msg(err); err = cudaMemcpy(s_A, d_A, SIZE*sizeof(float*), cudaMemcpyDeviceToHost); err = cudaMemcpy(s_B, d_B, SIZE*sizeof(float*), cudaMemcpyDeviceToHost); err = cudaMemcpy(s_C, d_C, SIZE*sizeof(float*), cudaMemcpyDeviceToHost); for(size_t i = 0; i < SIZE; ++i) { h_A[i] = new float[SIZE]; h_B[i] = new float[SIZE]; h_C[i] = new float[SIZE]; err = cudaMalloc((void **) &s_A[i], SIZE*sizeof(float)); print_err_msg(err); err = cudaMalloc((void **) &s_B[i], SIZE*sizeof(float)); print_err_msg(err); err = cudaMalloc((void **) &s_C[i], SIZE*sizeof(float)); print_err_msg(err); } for(size_t i = 0; i < SIZE; ++i) { for(size_t j = 0; j < SIZE; ++j) { h_A[i][j] = 1.0f; h_B[i][j] = 2.0f; } } for(size_t i = 0; i < SIZE; ++i) { cudaMemcpy(s_A[i], h_A[i], SIZE*sizeof(float), cudaMemcpyHostToDevice); cudaMemcpy(s_B[i], h_B[i], SIZE*sizeof(float), cudaMemcpyHostToDevice); } int threads = 32; dim3 nthreads(threads, threads); int blocks = ceil(SIZE/threads); dim3 nblocks(blocks, blocks); matrix_add<<<nblocks, nthreads>>>(d_A, d_B, d_C, SIZE); for(size_t i = 0; i < SIZE; ++i) { cudaMemcpy(h_C[i], s_C[i], SIZE*sizeof(float), cudaMemcpyDeviceToHost); } /* for(size_t i = 0; i < SIZE; ++i) { */ /* for(size_t j = 0; j < SIZE; ++j) { */ /* /1* assert(h_C[i][j] == 3.0f); *1/ */ /* printf("%f\t", h_C[i][j]); */ /* } */ /* printf("\n"); */ /* } */ for(size_t i = 0; i < SIZE; ++i) { delete[] h_A[i]; delete[] h_B[i]; delete[] h_C[i]; err = cudaFree(s_A[i]); print_err_msg(err); err = cudaFree(s_B[i]); print_err_msg(err); err = cudaFree(s_C[i]); print_err_msg(err); } delete[] h_A; delete[] h_B; delete[] h_C; err = cudaFree(d_A); print_err_msg(err); err = cudaFree(d_B); print_err_msg(err); err = cudaFree(d_C); print_err_msg(err); return 0; }
1,202
#ifndef MATRIXMULTIPLICATIONKERNEL_CU #define MATRIXMULTIPLICATIONKERNEL_CU #include <curand.h> __global__ void matrixMul(float * g_C, float * g_A, float *g_B,int wa, int wb){ int x = blockIdx.x * blockDim.y + threadIdx.x; int y = blockIdx.y*blockDim.x + threadIdx.y; float result = 0; int i = 0; for(i = 0; i < wa; ++i){ float tempA = g_A[y*wa+i]; float tempB = g_B[i*wb+x]; result += tempA*tempB; } g_C[y*wa+x] = result; } __global__ void matrixMul2(float * g_C, float * g_A, float *g_B,int wa, int wb){ int TILE_SIZE = 16; int x = blockIdx.x*TILE_SIZE + threadIdx.x; int y = blockIdx.y*TILE_SIZE + threadIdx.y; float result = 0; int i = 0; for(i = 0; i < wa; ++i){ float tempA = g_A[y*wa+i]; float tempB = g_B[i*wb+x]; result += tempA*tempB; } g_C[y*wa+x] = result; } __global__ void matrixMul3(float * g_C, float * g_A, float *g_B,int wa, int wb){ const int TILE_WIDTH = 16; int tx = threadIdx.x; int ty = threadIdx.y; int bx = blockIdx.x; int by = blockIdx.y; __shared__ float s_a[TILE_WIDTH][TILE_WIDTH]; __shared__ float s_b[TILE_WIDTH][TILE_WIDTH]; int row = bx*blockDim.y + tx; int col = by*blockDim.x + ty; float result = 0; int i = 0; for(i = 0; i < wa/TILE_WIDTH; ++i){ s_a[tx][ty] = g_A[i*TILE_WIDTH + row*wa +ty]; s_b[tx][ty] = g_B[(i*TILE_WIDTH*wa)+tx*wa+ col]; __syncthreads(); int k =0; for(k=0;k<TILE_WIDTH;++k){ result += s_a[tx][k] * s_b[k][ty]; } __syncthreads(); } g_C[row*wa+col] = result; } #endif
1,203
#include <stdio.h> __global__ void VecAdd(float * A, float * B, float * C) { int i = blockIdx.x * blockDim.x + threadIdx.x; C[i] = A[i] + B[i]; } void VecPrint(float * V, int len) { int to_print = 10; if (to_print > len) to_print = len; for (int i=0; i<to_print; i++) { printf("%4.2f", V[i]); if (i<to_print-1) printf(", "); } if (to_print < len) printf("..."); printf("\n"); } int main() { int N = 1024; size_t size = N * sizeof(float); // Allocate input vectors h_A, h_B and h_C in host memory float* h_A = (float*)malloc(size); float* h_B = (float*)malloc(size); float* h_C = (float*)malloc(size); // Initialize input vectors for (int i=0; i<N; i++) { h_A[i] = 2*i+1; h_B[i] = 4*i+1; h_C[i] = 0; } // Print initialised vectors VecPrint(h_A, N); VecPrint(h_B, N); VecPrint(h_C, N); // Allocate vectors in device memory float* d_A; cudaMalloc(&d_A, size); float* d_B; cudaMalloc(&d_B, size); float* d_C; cudaMalloc(&d_C, size); // Copy vectors from host memory to device memory cudaMemcpy(d_A, h_A, size, cudaMemcpyHostToDevice); cudaMemcpy(d_B, h_B, size, cudaMemcpyHostToDevice); // Run the add kernel int blocks = 3; int threads_per_block = N/blocks; VecAdd<<<blocks, threads_per_block>>>(d_A, d_B, d_C); // Print result cudaMemcpy(h_C, d_C, size, cudaMemcpyDeviceToHost); VecPrint(h_C, N); // Free device memory cudaFree(d_A); cudaFree(d_B); cudaFree(d_C); // Free host memory delete[] h_A; delete[] h_B; return 0; }
1,204
__global__ void vecAdd(float *in1, float *in2, float *out, int len) { int i = blockIdx.x * blockDim.x + threadIdx.x; if( i<len ) { out[i] = in1[i]+in2[i]; } }
1,205
#include <cuda_runtime.h> #include <stdio.h> __global__ void kernel() { int tid = threadIdx.x; if (tid < 8) { printf("inside the kernel\n"); } else { printf("outside the kernel\n"); } } int cuda(int a, int b) { kernel<<<1, 10>>>(); cudaDeviceSynchronize(); return 0; }
1,206
__device__ volatile float BigData[1024*1024];
1,207
/* This is a automatically generated test. Do not modify */ #include <stdio.h> #include <stdlib.h> #include <math.h> __global__ void compute(float comp, int var_1,float var_2,float var_3,float var_4,float var_5,float var_6,float var_7,float var_8,float var_9,float var_10,float var_11,float var_12,float var_13,float var_14,float var_15,float var_16,float var_17,float var_18,float var_19,float var_20) { comp += var_2 - (var_3 + +1.4889E-43f); if (comp <= (+1.5026E25f + atanf(var_4 * expf(var_5 / var_6 / -1.8817E35f)))) { float tmp_1 = +1.8395E9f; float tmp_2 = -0.0f; comp += tmp_2 * tmp_1 * var_7 + (var_8 / asinf(var_9 / fabsf(-1.3023E-42f))); comp = (var_10 - sqrtf(-1.1907E35f)); } if (comp < +1.3739E-41f - var_11) { comp = (var_12 * (+1.6612E34f / -1.7853E8f / +1.4702E-43f)); comp += var_13 / var_14; comp += var_15 * -0.0f * (var_16 * var_17 / asinf(floorf((var_18 - +1.4509E-36f - -1.6745E2f)))); } for (int i=0; i < var_1; ++i) { comp += (+1.2415E20f - asinf(+0.0f / (var_19 - var_20 * -1.2148E-35f))); } printf("%.17g\n", comp); } float* initPointer(float v) { float *ret = (float*) malloc(sizeof(float)*10); for(int i=0; i < 10; ++i) ret[i] = v; return ret; } int main(int argc, char** argv) { /* Program variables */ float tmp_1 = atof(argv[1]); int tmp_2 = atoi(argv[2]); float tmp_3 = atof(argv[3]); float tmp_4 = atof(argv[4]); float tmp_5 = atof(argv[5]); float tmp_6 = atof(argv[6]); float tmp_7 = atof(argv[7]); float tmp_8 = atof(argv[8]); float tmp_9 = atof(argv[9]); float tmp_10 = atof(argv[10]); float tmp_11 = atof(argv[11]); float tmp_12 = atof(argv[12]); float tmp_13 = atof(argv[13]); float tmp_14 = atof(argv[14]); float tmp_15 = atof(argv[15]); float tmp_16 = atof(argv[16]); float tmp_17 = atof(argv[17]); float tmp_18 = atof(argv[18]); float tmp_19 = atof(argv[19]); float tmp_20 = atof(argv[20]); float tmp_21 = atof(argv[21]); compute<<<1,1>>>(tmp_1,tmp_2,tmp_3,tmp_4,tmp_5,tmp_6,tmp_7,tmp_8,tmp_9,tmp_10,tmp_11,tmp_12,tmp_13,tmp_14,tmp_15,tmp_16,tmp_17,tmp_18,tmp_19,tmp_20,tmp_21); cudaDeviceSynchronize(); return 0; }
1,208
#include "includes.h" __global__ void vecAdd(float* a, float* b, float* c, const int N) { const int i = blockIdx.x*blockDim.x + threadIdx.x; if(i<N) c[i] = a[i] + b[i]; }
1,209
#include "includes.h" __global__ void vector_add(float *out, float *a, float *b, int n) { int tid = blockIdx.x * blockDim.x + threadIdx.x; // Handling arbitrary vector size if (tid < n){ out[tid] = a[tid] + b[tid]; } }
1,210
#include <stdlib.h> #include <stdio.h> int main(void){ int num_elements = 16; int num_bytes = num_elements*sizeof(int); int *device_array = 0; int *host_array = 0; host_array = (int *)malloc(num_bytes); cudaMalloc((void**)&device_array, num_bytes); cudaMemset(device_array, 0, num_bytes); cudaMemcpy(host_array, device_array, num_bytes, cudaMemcpyDeviceToHost); int i; for(i = 0; i<num_elements; i++) printf("%d\n", host_array[i]); free(host_array); cudaFree(device_array); return 0; }
1,211
#include <stdio.h> #include <cuda_runtime.h> #include <cuda.h> #include <stdlib.h> #include "device_launch_parameters.h" #include <thrust/scan.h> #include <thrust/device_vector.h> #include <thrust/count.h> const int BASE1 = 10000 + 7; const int BASE2 = 100000 + 3; const int MOD1 = 1000000 + 3; const int MOD2 = 1000000 + 37; __global__ void findhash(int *d_qvert,int *d_qverc,int *d_qvid,int *d_qelist,bool *d_over,bool *d_qtree,int *d_hash1,int *d_hash2) { int i; int ver=threadIdx.x*blockDim.y+threadIdx.y+blockDim.x*blockDim.y*(blockIdx.x*gridDim.y+blockIdx.y); if(ver>=d_qverc[0]) return ; if(d_qvid[ver]!=0) return; int l=d_qvert[ver+1]; int hash1=1,hash2=1; int flag=0; for(i=d_qvert[ver];i<l;i++) { int m=d_qelist[i]; bool treeedge=d_qtree[i]; if(treeedge){ int tt=d_qvid[m]; if(tt==0) return; flag=1; hash1=(hash1*1L*BASE1)*tt % MOD1; hash2=(hash2*1L*BASE2)*tt % MOD2; } } if(flag==0) return; if(flag==1){ *d_over=false; d_hash1[hash1]=1; d_hash2[hash2]=1; } } __global__ void setdeg1(int *d_qvert,int *d_qverc,int *d_qvid,bool *d_qtree) { int i; int ver=threadIdx.x*blockDim.y+threadIdx.y+blockDim.x*blockDim.y*(blockIdx.x*gridDim.y+blockIdx.y); if(ver>=d_qverc[0]) return ; if(d_qvid[ver]!=0) return; int l=d_qvert[ver+1]; bool treeedge; for(i=d_qvert[ver];i<l;i++) { treeedge=d_qtree[i]; if(treeedge) return; } //printf("%d %d\n",ver,i); d_qvid[ver]=1; } /*__global__ void alignhash(bool *d_hash1,bool *d_hash2) { int ver=threadIdx.x*blockDim.y+threadIdx.y+blockDim.x*blockDim.y*(blockIdx.x*gridDim.y+blockIdx.y); if(ver>=1000038) return ; if(d_hash1[ver] || d_hash2[ver]){ d_hash1=true; } }*/ __global__ void puttoid(int *d_qvert,int *d_qverc,int *d_qvid,int *d_qelist,bool *d_qtree,int *d_loc,int * d_qidtov) { int i; int ver=threadIdx.x*blockDim.y+threadIdx.y+blockDim.x*blockDim.y*(blockIdx.x*gridDim.y+blockIdx.y); if(ver>=d_qverc[0]) return ; if(d_qvid[ver]!=0) return; int l=d_qvert[ver+1]; int hash1=1,hash2=1; int flag=0; for(i=d_qvert[ver];i<l;i++) { int m=d_qelist[i]; bool treeedge=d_qtree[i]; if(treeedge){ int tt=d_qvid[m]; if(tt==0) return; flag=1; hash1=(hash1*1L*BASE1)*tt % MOD1; hash2=(hash2*1L*BASE2)*tt % MOD2; } } //printf("%d %d %d \n",ver,flag,d_loc[hash1]); if(flag==0) return; int id=d_loc[hash1]; d_qvid[ver]=id; d_qidtov[id]=ver; } __device__ bool chechall(int ver,bool *check,int i,int dfrom,int dto,int *d_delist,int *d_qelist,int *d_qvid,int qfrom,int qto,int ** d_dcvslist){ //int ql=qfrom-qto; int ql=qto-qfrom; int j,k,l; //d_dcvslist[2][ql]=true; if(i==ql){ k=d_qelist[i+qfrom-1]; k=d_qvid[k]; if(k>=d_qvid[ver]) return true; for(j=dfrom;j<dto;j++){ l=d_delist[j]; if(check[l]) continue; if(!d_dcvslist[k][l]) continue; return true; } } else{ int res=false; k=d_qelist[i+qfrom-1]; k=d_qvid[k]; if(k>=d_qvid[ver]) return chechall(ver,check,i+1,dfrom,dto,d_delist,d_qelist,d_qvid,qfrom,qto,d_dcvslist); for(j=dfrom;j<dto;j++){ l=d_delist[j]; if(check[l]) continue; if(!d_dcvslist[k][l]) continue; check[l]=true; res|=chechall(ver,check,i+1,dfrom,dto,d_delist,d_qelist,d_qvid,qfrom,qto,d_dcvslist); if(res==true) return true; check[l]=false; } } return false; } __global__ void findcvs(int ver,int *d_dvert,int *d_dverc,int *d_delist,int *d_qvert,int *d_qelist,int *d_qvid,int ** d_dcvslist ) { //int i; int dver=threadIdx.x*blockDim.y+threadIdx.y+blockDim.x*blockDim.y*(blockIdx.x*gridDim.y+blockIdx.y); if(dver>=d_dverc[0]) return ; int ql=d_qvert[ver+1]-d_qvert[ver]; int dl=d_dvert[dver+1]-d_dvert[dver]; if(ql>dl) return; bool *checked=(bool*)malloc(sizeof(bool)*d_dverc[0]); //bool *checked=new bool[d_dverc[0]]; memset(checked,false,sizeof(bool)*d_dverc[0]); //chechall(bool *check,int i,int dfrom,int dto,int *d_delist,int *d_qelist,int *d_qvid,int qfrom,int qto,bool ** d_dcvslist) if(chechall(ver,checked,1,d_dvert[dver],d_dvert[dver+1],d_delist,d_qelist,d_qvid,d_qvert[ver],d_qvert[ver+1],d_dcvslist)) d_dcvslist[d_qvid[ver]][dver]=true; free(checked); } __global__ void puttolist(int *d_dverc,int *d_loc,int * d_dcvslist ) { int dver=threadIdx.x*blockDim.y+threadIdx.y+blockDim.x*blockDim.y*(blockIdx.x*gridDim.y+blockIdx.y); if(dver>=d_dverc[0]) return ; if(d_loc[dver]!=d_loc[dver+1]) d_dcvslist[d_loc[dver]]=dver; } __global__ void checkperm(bool *found,int * qdmap,int * d_qverc,int * d_qelist,int * d_qvert,int * d_dvert,int *d_delist){ int i; //found[0]=false; int ver=threadIdx.x*blockDim.y+threadIdx.y+blockDim.x*blockDim.y*(blockIdx.x*gridDim.y+blockIdx.y); if(ver>=d_qverc[0]) return ; int n,p,j,k,flag=0; //for(ver=0;ver<d_qverc[0];ver++){ int l=d_qvert[ver+1]; int dver=qdmap[ver]; n=d_dvert[dver+1]; for(i=d_qvert[ver];i<l;i++) { flag=0; j=d_qelist[i]; p=d_dvert[dver]; k=qdmap[j]; for(;p<n;p++){ if(k==d_delist[p]){ flag=1; break; } } if(!flag){ *found=false; return; } } //} } __global__ void findall(int *d_mapans,int *d_cans,int *d_qvid,int * d_qverc,int * d_qelist,int * d_qvert,int * d_dvert,int *d_delist,int ** d_cvsverlist,int * d_size_cvs) { bool found[1]={true}; long long int blockId = blockIdx.x + blockIdx.y * gridDim.x + gridDim.x * gridDim.y * blockIdx.z; long long int threadId = blockId * (blockDim.x * blockDim.y * blockDim.z) + (threadIdx.z * (blockDim.x * blockDim.y)) + (threadIdx.y * blockDim.x) + threadIdx.x; int i=0; long long int indexperm=threadId; for(i=0;i<d_qverc[0];i++){ int j=d_qvid[i]; indexperm/=d_size_cvs[j]; } if(indexperm) return; indexperm=threadId; int *d_qdmap=&d_mapans[d_qverc[0]*threadId];//new int[d_qverc[0]]; for(i=0;i<d_qverc[0];i++){ int j=d_qvid[i]; d_qdmap[i]=d_cvsverlist[j][indexperm%d_size_cvs[j]]; indexperm/=d_size_cvs[j]; } //dim3 blocks((max/16 )+ 1,(max/16)+1); //dim3 threads(16,16); //found[0]=true; //checkperm<<<blocks,threads>>> (found,d_qdmap,d_qverc,d_qelist,d_qvert,d_dvert,d_delist); int n,p,j,k,flag=0,ver; for(ver=0;ver<d_qverc[0];ver++){ int l=d_qvert[ver+1]; int dver=d_qdmap[ver]; n=d_dvert[dver+1]; for(i=d_qvert[ver];i<l;i++) { flag=0; j=d_qelist[i]; p=d_dvert[dver]; k=d_qdmap[j]; for(;p<n;p++){ if(k==d_delist[p]){ flag=1; break; } } if(!flag){ *found=false; return; } } } if(found[0]){ d_cans[threadId]=1; //printf("%d ", threadId); } //delete d_qdmap; } int * qdmap; int *d_qverc,*d_dverc; int *d_qvid,*d_qidtov,*h_qidtov,*h_qvid; int *d_qvert,*d_qelist,*d_dvert,*d_delist;//,*d_dvelist,*d_qvelist; bool *d_qtree,*d_over; int *d_qdmap; bool h_over; /*void callforallperm(bool * check,int ** cvslist,int i,int max,int dmax){ int j,k,l; l=h_qvid[i-1]; //printf("i%d %di",i,l); if(i==max){ for(j=0;j<dmax;j++) if(cvslist[l][j] && !check[j]){ qdmap[i-1]=j; dim3 blocks((max/16 )+ 1,(max/16)+1); dim3 threads(16,16); h_over=true; //for(k=0;k<max;k++) // printf("%d ",qdmap[k]); cudaMemcpy(d_over, &h_over, sizeof(bool), cudaMemcpyHostToDevice) ; cudaMemcpy(d_qdmap, qdmap, sizeof(int)*(max+1), cudaMemcpyHostToDevice); checkperm<<<blocks,threads>>> (d_over,d_qdmap,d_qverc,d_qelist,d_qvert,d_dvert,d_delist); //checkperm(bool *found,int * qdmap,int * d_qverc,int * d_qelist,int * d_qvert,int * d_dvert,int *d_delist) cudaError_t err = cudaGetLastError(); if(err!=cudaSuccess) { printf("Error: %s\n", cudaGetErrorString(err)); printf("Not Ok"); } cudaMemcpy(&h_over, d_over, sizeof(bool), cudaMemcpyDeviceToHost) ; if(h_over){ for(k=0;k<max;k++) printf("%d ",qdmap[k]); //printf("\n"); printf("OK\n"); } //printf("\n"); } } else{ for(j=0;j<dmax;j++){ //printf("%d %d %d\n",j,check[j],cvslist[l][j]); if(cvslist[l][j] && !check[j]){ check[j]=true; qdmap[i-1]=j; callforallperm(check,cvslist,i+1,max,dmax); check[j]=false; } } } }*/ int main(int argc, char **argv) { int deviceId = 4; cudaSetDevice(deviceId); int h_qverc,h_dverc; int *h_qvert,*h_qelist,*h_dvert,*h_delist;//,*h_dvelist,*h_qvelist; bool *h_qtree; int *d_hash1,*d_hash2; int i,j; int **h_cvslist,**d_cvslist,**h_tem; scanf("%d",&h_qverc); h_qvert=(int *)malloc(sizeof(int)*(h_qverc+1)); h_qvid=(int *)malloc(sizeof(int)*(h_qverc+1)); h_qidtov=(int *)malloc(sizeof(int)*(h_qverc+1)); h_tem=(int **)malloc(sizeof(int*)*(h_qverc+1)); h_cvslist=(int **)malloc(sizeof(int*)*(h_qverc+1)); for(i=0;i<=h_qverc;i++){ scanf("%d",&h_qvert[i]); } h_qelist=(int *)malloc(sizeof(int)*h_qvert[h_qverc]); for(i=0;i<h_qvert[h_qverc];i++) scanf("%d",&h_qelist[i]); h_qtree=(bool *)malloc(sizeof(bool)*h_qvert[h_qverc]); for(i=0;i<h_qvert[h_qverc];i++){ scanf("%d",&j); if(j==1) h_qtree[i]=true; else h_qtree[i]=false; } scanf("%d",&h_dverc); h_dvert=(int *)malloc(sizeof(int)*(h_dverc+1)); for(i=0;i<=h_dverc;i++){ scanf("%d",&h_dvert[i]); } for(i=0;i<=h_qverc;i++) h_cvslist[i]=(int *)malloc(sizeof(int)*(h_dverc+1)); h_delist=(int *)malloc(sizeof(int)*h_dvert[h_dverc]); for(i=0;i<h_dvert[h_dverc];i++) scanf("%d",&h_delist[i]); cudaMalloc(&d_qverc,sizeof(int)); cudaMalloc(&d_over,sizeof(bool)); cudaMalloc(&d_qvert,sizeof(int)*(h_qverc+1)); cudaMalloc(&d_qidtov,sizeof(int)*(h_qverc+1)); //cudaMalloc(&d_loc,sizeof(int)*(h_qverc+1)); cudaMalloc(&d_qelist,sizeof(int)*h_qvert[h_qverc]); cudaMalloc(&d_qtree,sizeof(bool)*h_qvert[h_qverc]); cudaMalloc(&d_hash1,sizeof(int)*1000038); cudaMalloc(&d_hash2,sizeof(int)*1000038); cudaMalloc(&d_qvid,sizeof(int)*(h_qverc+1)); cudaMemcpy(d_qverc,&h_qverc,sizeof(int),cudaMemcpyHostToDevice); cudaMemcpy(d_qvert,h_qvert,sizeof(int)*(h_qverc+1),cudaMemcpyHostToDevice); cudaMemcpy(d_qelist,h_qelist,sizeof(int)*h_qvert[h_qverc],cudaMemcpyHostToDevice); cudaMemcpy(d_qtree,h_qtree,sizeof(bool)*h_qvert[h_qverc],cudaMemcpyHostToDevice); cudaMemset(d_hash1,0,sizeof(int)*1000038); cudaMemset(d_hash2,0,sizeof(int)*1000038); //cudaMemset(d_loc,0,sizeof(int)*(h_qverc+1)); cudaMemset(d_qidtov,-1,sizeof(int)*(h_qverc+1)); cudaMemset(d_qvid,0,sizeof(int)*(h_qverc+1)); int *h_hash1=(int *)malloc(sizeof(int)*1000038); int *h_hash2=(int *)malloc(sizeof(int)*1000038); dim3 blocks((sqrt(h_qverc)/16 )+ 1,(sqrt(h_qverc)/16)+1); dim3 threads(16,16); //int *d_qvert,int *d_dverc,int *d_qvid,int *d_qelist,bool *d_over,bool *d_hash1,bool *d_hash2) h_over=true; //h_qvid[1]=1; //h_qvid[3]=1; //cudaMemcpy(d_qvid,h_qvid,sizeof(int)*(h_qverc+1),cudaMemcpyHostToDevice); //printf("qt%d %dqt\n",h_qtree[0],h_qtree[1]); setdeg1<<<blocks,threads>>>(d_qvert,d_qverc,d_qvid,d_qtree); h_over=false; int maxval=2; while(!h_over) { h_over=true; cudaMemcpy(d_over, &h_over, sizeof(bool), cudaMemcpyHostToDevice) ; cudaMemset(d_hash1,0,sizeof(int)*1000038); findhash <<<blocks,threads>>> (d_qvert,d_qverc,d_qvid,d_qelist,d_over,d_qtree,d_hash1,d_hash2); //(int *d_qvert,int *d_dverc,int *d_qvid,int *d_qelist,bool *d_over,bool *d_hash1,bool *d_qtree,bool *d_hash2) cudaError_t err = cudaGetLastError(); if(err!=cudaSuccess) { printf("Error: %s\n", cudaGetErrorString(err)); printf("Not Ok"); } cudaMemcpy(h_hash1,d_hash1,sizeof(int)*1000038,cudaMemcpyDeviceToHost); h_hash1[0]+=maxval; thrust::exclusive_scan(h_hash1,h_hash1+1000038,h_hash1); maxval=h_hash1[1000037]; cudaMemcpy(d_hash1,h_hash1,sizeof(int)*1000038,cudaMemcpyHostToDevice); puttoid<<<blocks,threads>>>(d_qvert,d_qverc,d_qvid,d_qelist,d_qtree,d_hash1,d_qidtov); /// cudaMemcpy(h_hash2,d_hash2,sizeof(bool)*1000038,cudaMemcpyDeviceToHost); cudaMemcpy(&h_over, d_over, sizeof(bool), cudaMemcpyDeviceToHost) ; //printf("over flag:%d ",h_over); /*for(i=0;i<h_qverc;i++){ //if() printf("%d ",h_qvid[i]); // if(h_hash2[i]) // printf("h2 %d ",i); // if(h_hash1[i] || h_hash2[i]) // printf("\n"); } printf("\n");*/ } cudaMemcpy(h_qvid,d_qvid,sizeof(int)*h_qverc,cudaMemcpyDeviceToHost); cudaMemcpy(h_qidtov,d_qidtov,sizeof(int)*(h_qverc+1),cudaMemcpyDeviceToHost); for(i=0;i<=h_qverc;i++){ printf("%d ",h_qidtov[i]); } printf("\n"); for(i=0;i<=h_qverc;i++){ printf("%d ",h_qvid[i]); } printf("\n"); cudaFree(d_qtree); cudaFree(d_hash1); cudaFree(d_hash2); free(h_hash1); free(h_hash2); free(h_qtree); cudaMalloc(&d_cvslist,sizeof(int*)*(h_qverc+1)); for(i=0;i<=h_qverc;i++){ cudaMalloc(&h_tem[i],sizeof(int)*(h_dverc+1)); cudaMemset(h_tem[i],0,sizeof(int)*(h_dverc+1)); } cudaMemset(h_tem[1],1,sizeof(int)*(h_dverc+1)); cudaMemcpy(d_cvslist,h_tem,sizeof(int*)*(h_qverc+1),cudaMemcpyHostToDevice); cudaMalloc(&d_dvert,sizeof(int)*(h_dverc+1)); cudaMalloc(&d_dverc,sizeof(int)); cudaMalloc(&d_delist,sizeof(int)*h_dvert[h_dverc]); cudaMemcpy(d_dverc,&h_dverc,sizeof(int),cudaMemcpyHostToDevice); cudaMemcpy(d_dvert,h_dvert,sizeof(int)*(h_dverc+1),cudaMemcpyHostToDevice); cudaMemcpy(d_delist,h_delist,sizeof(int)*h_dvert[h_dverc],cudaMemcpyHostToDevice); dim3 dblocks((sqrt(h_dverc)/16 )+ 1,(sqrt(h_dverc)/16)+1); dim3 dthreads(16,16); int **d_cvsverlist,**d_temverlist; int *d_size_cvs,*h_size_cvs; memset(h_cvslist[1],1,sizeof(int)*(h_dverc+1)); h_size_cvs=(int *)malloc(sizeof(int)*(h_qverc+1)); memset(h_size_cvs,0,sizeof(int)*(h_qverc+1)); cudaMalloc(&d_size_cvs,sizeof(int)*(h_qverc+1)); cudaMemset(d_size_cvs,0,sizeof(int)*(h_qverc+1)); cudaMalloc(&d_cvsverlist,sizeof(int*)*(h_qverc+1)); d_temverlist=(int **)malloc(sizeof(int*)*(h_qverc+1)); for(i=0;i<=h_qverc;i++){ cudaMalloc(&d_temverlist[i],sizeof(int)*(h_dverc+1)); cudaMemset(d_temverlist[i],0,sizeof(int)*(h_dverc+1)); } cudaMemcpy(d_cvsverlist,d_temverlist,sizeof(int*)*(h_qverc+1),cudaMemcpyHostToDevice); long long int totalthreads=1; for(i=0;i<h_dverc;i++) h_cvslist[1][i]=i; cudaMemcpy(d_temverlist[1],h_cvslist[1],sizeof(int)*(h_dverc+1),cudaMemcpyHostToDevice); h_size_cvs[1]=h_dverc; for(i=0;i<=h_qverc;i++) { if(h_qidtov[i]!=-1) { //findcvs(int ver,int *d_dvert,int *d_dverc,int *d_delist,int *d_qvert,int *d_qelist,int *d_qvid,bool ** d_dcvslist ) findcvs<<<dblocks,dthreads>>>(h_qidtov[i],d_dvert,d_dverc,d_delist,d_qvert,d_qelist,d_qvid,d_cvslist); cudaError_t err = cudaGetLastError(); cudaMemcpy(h_cvslist[i],h_tem[i],sizeof(int)*(h_dverc+1),cudaMemcpyDeviceToHost); for(j=0;j<=h_dverc;j++) if(h_cvslist[i][j]) printf("%d ",j); printf("\n"); //printf("%d ",h_qidtov[i]); thrust::exclusive_scan(h_cvslist[i],h_cvslist[i]+h_dverc+1,h_cvslist[i]); h_size_cvs[i]=h_cvslist[i][h_dverc]; cudaMemcpy(h_tem[i],h_cvslist[i],sizeof(int)*(h_dverc+1),cudaMemcpyHostToDevice); puttolist<<<dblocks,dthreads>>>(d_dverc,h_tem[i],d_temverlist[i]); // cudaMemcpy(h_cvslist[i],d_temverlist[i],sizeof(int)*(h_dverc+1),cudaMemcpyDeviceToHost); // for(j=0;j<=h_dverc;j++) // printf("%d ",h_cvslist[i][j]); } } // cudaMemcpy(h_delist,d_delist,sizeof(int)*(h_dvert[h_dverc]),cudaMemcpyDeviceToHost); // for(j=0;j<h_dvert[h_dverc];j++) // printf("%d ",h_delist[j]); for(i=0;i<h_qverc;i++) if(h_size_cvs[h_qvid[i]]) totalthreads*=h_size_cvs[h_qvid[i]]; printf("Start %lld\n",totalthreads); cudaMemcpy(d_size_cvs,h_size_cvs,sizeof(int)*(h_qverc+1),cudaMemcpyHostToDevice); //totalthreads=1000; dim3 dpblocks(((int)(sqrt(totalthreads)/16 )+ 1),((int)(sqrt(totalthreads)/16)+1)); dim3 dpthreads(16,16); int *d_mapans,*d_countans,*h_countans; cudaMalloc(&d_mapans,sizeof( int)*totalthreads*(h_qverc+1)); cudaMalloc(&d_countans,sizeof( int)*(totalthreads+1)); cudaMemset(d_countans,0,sizeof( int)*(totalthreads+1)); //h_countans=(int *)malloc(sizeof(int)*(totalthreads+1)); //h_countans=0; //cudaMemcpy(d_countans, &h_countans, sizeof( int), cudaMemcpyHostToDevice) ; //cudaMemcpy(&h_countans, d_qverc, sizeof(int), cudaMemcpyDeviceToHost) ; //printf("%d\n",h_countans); findall<<<dpblocks,dpthreads>>> (d_mapans,d_countans,d_qvid,d_qverc,d_qelist,d_qvert,d_dvert,d_delist,d_cvsverlist,d_size_cvs); thrust::device_ptr<int> cptr=thrust::device_pointer_cast(d_countans); //int sum=thrust::count(cptr,cptr+totalthreads,1); //cudaMemcpy(h_countans, d_countans, sizeof(int)*totalthreads, cudaMemcpyDeviceToHost) ; //thrust::exclusive_scan(h_countans,h_countans+totalthreads+1,h_countans); //printf("%d\n",h_countans[totalthreads-1] ); //printf("%d\n",h_countans); //printf("%d\n",sum); /*j=0; for(i=0;i<totalthreads;i++) if(h_countans[i]) j++; printf("%d\n",j); *///bool * check=(bool *)malloc(sizeof(bool)*(h_dverc+1)); //memset(check,false,sizeof(bool)*(h_dverc+1)); //qdmap=(int *)malloc(sizeof(int)*(h_qverc+1)); //cudaMalloc(&d_qdmap,sizeof(int)*(h_qverc+1)); //callforallperm(check,h_cvslist,1,h_qverc,h_dverc); cudaFree(d_over); cudaFree(d_qverc); cudaFree(d_qvert); cudaFree(d_qelist); cudaFree(d_qvid); cudaFree(d_qidtov); cudaFree(d_dvert); cudaFree(d_delist); cudaFree(d_dverc); cudaFree(d_cvslist); cudaFree(d_cvsverlist); cudaFree(d_size_cvs); /*free(h_qvid); free(h_qvert); //free(h_qelist); free(h_qidtov); free(h_cvslist); free(h_dvert); free(h_delist);*/ }
1,212
#include <stdio.h> #include <cuda.h> __global__ void vecmul(float *A, float* B, float *C, int size) { // Row and Column indexes: int row = blockIdx.y * blockDim.y + threadIdx.y; int col = blockIdx.x * blockDim.x + threadIdx.x; // Are they bellow the maximum? if (col < size && row < size) { float result = 0; for(int ix = 0; ix < size; ix++) { result += A[row*size + ix]*B[ix*size + col]; } C[row*size + col] = result; } } extern "C" { void maxmul(float *A, float* B, float *C, int size) { int total = size*size; // Allocate device memory: float* gpu_A; float* gpu_B; float* gpu_C; int msize = total * sizeof(float); cudaMalloc((void**)&gpu_A, msize); cudaMemcpy(gpu_A, A, msize, cudaMemcpyHostToDevice); cudaMalloc((void**)&gpu_B, msize); cudaMemcpy(gpu_B, B, msize, cudaMemcpyHostToDevice); cudaMalloc((void**)&gpu_C, msize); // Blocks & grids: dim3 blocks(size, size); dim3 grid(1, 1); // Call the kernel: vecmul<<<grid,blocks>>>(gpu_A, gpu_B, gpu_C, size); // Get the result Matrix: cudaMemcpy(C, gpu_C, msize, cudaMemcpyDeviceToHost); //Free device matrices cudaFree(gpu_A); cudaFree(gpu_B); cudaFree(gpu_C); } }
1,213
/* * Name: Nate Steawrt * Date: 04-04-16 * Description: Serial implementation of Matrix morphism */ #include <time.h> #include <stdio.h> #include <stdlib.h> #define RANDOM_VALUE_MIN 1.0 #define RANDOM_VALUE_MAX 2.0 #define NUM_ROWS 4097 #define NUM_COLS 4097 /* * Calculate and return a random value between min and max. */ double randDouble(double min, double max) { double range = max - min; double dist = RAND_MAX / range; return min + (rand() / dist); } /* * Output the matrix to fout */ void outputMatrix(FILE *fout, double *matrix, int rows, int cols) { int i, j; for (i = 0; i < rows; i++) { for (j = 0; j < cols; j++) { fprintf(fout, "%lf ", *(matrix + i * cols + j)); } fprintf(fout, "\n"); } } __global__ void computeMath(double *matrix) { int i; // Grab id of thread int threadId = blockDim.x * threadIdx.y + threadIdx.x + 1; // Declare pointers to the two arguments of the addition and the result pointer double *f_ptr, *first_ptr, *second_ptr; // Grab starting points for pointers f_ptr = matrix + threadId * NUM_COLS; first_ptr = matrix + (threadId - 1) * NUM_COLS + 1; second_ptr = f_ptr + 1; // Compute a single row for (i = 0; i < NUM_COLS - 1; i++, f_ptr++, first_ptr++, second_ptr++) { *f_ptr = *first_ptr + *second_ptr; } } /* * Check if an error occurred during the last CUDA command */ void checkError() { int errorCode = cudaGetLastError(); if (errorCode != 0) { printf("Error %d occurred during last operation.\n", errorCode); } } int main(void) { // Declare the needed variables int i, j; // Define thread hierarchy int nblocks = 64; int dimX = 64; int dimY = 1; // Declare the memory pointers double *h_matrix, *d_matrix; // Allocate memory for host and device size_t memSize = NUM_ROWS * NUM_COLS * sizeof(*h_matrix); // Create space on the host and device for matrix h_matrix = (double *)malloc(memSize); cudaMalloc( (void**) &d_matrix, memSize); checkError(); // Initialize the matrix and copy values into device double *f_ptr = h_matrix; // Setup a traversal pointer for (i = 0; i < NUM_ROWS; i++) { for (j = 0; j < NUM_COLS; j++, f_ptr++) { *f_ptr = randDouble(RANDOM_VALUE_MIN, RANDOM_VALUE_MAX); } } cudaMemcpy(d_matrix, h_matrix, memSize, cudaMemcpyHostToDevice); checkError(); // Set up grid and block structure dim3 dimGrid(nblocks); dim3 dimBlock(dimX, dimY); // Launch the kernel for (i = 0; i < 100; i++) { computeMath<<< dimGrid, dimBlock >>>(d_matrix); checkError(); } // Retrieve results and free memory cudaMemcpy(h_matrix, d_matrix, memSize, cudaMemcpyDeviceToHost); checkError(); free(h_matrix); cudaFree(d_matrix); checkError(); }
1,214
#include "includes.h" #define WEIGHTSUM 273 #define BLOCK_SIZE 16 int * heatmap; size_t heatmap_pitch; int * scaled_heatmap; size_t scaled_heatmap_pitch; int * blurred_heatmap; size_t blurred_heatmap_pitch; float* d_desiredPositionX; float* d_desiredPositionY; __global__ void computeScaledHeatmap(int* heatmap, size_t heatmap_pitch, int* scaled_heatmap, size_t scaled_heatmap_pitch) { // Block row and column int blockRow = blockIdx.y; int blockCol = blockIdx.x; // Thread row and column block int row = threadIdx.y; int col = threadIdx.x; // x, y coordinate int x = blockCol * blockDim.x + col; int y = blockRow * blockDim.y + row; // Scale the data for visual representation int value = *((int*)((char*)heatmap + y * heatmap_pitch) + x); for (int r = 0; r < CELLSIZE; r++) { int* row = (int*)((char*)scaled_heatmap + (r + y * CELLSIZE) * scaled_heatmap_pitch); for (int c = 0; c < CELLSIZE; c++) { row[x * CELLSIZE + c] = value; } } }
1,215
#include<stdio.h> #include<math.h> #include<cuda.h> #define N 256 __global__ void matrix_vector_multi_gpu_2_128(float *A_d,float *B_d,float *C_d){ int i,j; j=blockIdx.x*128+threadIdx.x; A_d[j]=0.0F; for(i=0;i<N;i++){ A_d[j]=A_d[j]+B_d[j*N+i]*C_d[i]; } } int main(){ int i,j; float A[N],B[N*N],C[N]; float *A_d,*B_d,*C_d; dim3 blocks(2,1,1); dim3 threads(128,1,1); for(j=0;j<N;j++){ for(i=0;i<N;i++){ B[j*N+i]=((float)j)/256.0; } } for(j=0;j<N;j++){ C[j]=1.0F; } cudaMalloc((void**)&A_d,N*sizeof(float)); cudaMalloc((void**)&B_d,N*N*sizeof(float)); cudaMalloc((void**)&C_d,N*sizeof(float)); cudaMemcpy(A_d,A,N*sizeof(float),cudaMemcpyHostToDevice); cudaMemcpy(B_d,B,N*N*sizeof(float),cudaMemcpyHostToDevice); cudaMemcpy(C_d,C,N*sizeof(float),cudaMemcpyHostToDevice); matrix_vector_multi_gpu_2_128<<<blocks,threads>>>(A_d,B_d,C_d); cudaMemcpy(A,A_d,N*sizeof(float),cudaMemcpyDeviceToHost); for(j=0;j<N;j++){ printf("A[ %d ]=%f \n",j,A[j]); } cudaFree(A_d); cudaFree(B_d); cudaFree(C_d); return 0; }
1,216
#include<stdio.h> #include<stdlib.h> #include<cuda.h> // #include <opencv2/opencv.hpp> #include<fstream> // #include <TooN/TooN.h> // #include <TooN/se3.h> // #include <TooN/GR_SVD.h> // using namespace cv; using namespace std; __device__ int get_pos(){ return threadIdx.x + blockIdx.x * blockDim.x; } struct Img{ int* data; int image_width; int image_height; int img_arr_size; Img(){ } Img(int w, int h){ image_height = h; image_width = w; img_arr_size = sizeof(int) * image_width*image_height; cudaMalloc((void**)&data, img_arr_size); } }; struct Matrix4 { float4 data[4]; inline __host__ __device__ float3 get_translation() const { return make_float3(data[0].w, data[1].w, data[2].w); } }; inline Matrix4 getCameraMatrix( const float4 & k ){ Matrix4 K; K.data[0] = make_float4(k.x, 0, k.z, 0); K.data[1] = make_float4(0, k.y, k.w, 0); K.data[2] = make_float4(0, 0, 1, 0); K.data[3] = make_float4(0, 0, 0, 1); return K; } inline Matrix4 getInverseCameraMatrix( const float4 & k ){ Matrix4 invK; invK.data[0] = make_float4(1.0f/k.x, 0, -k.z/k.x, 0); invK.data[1] = make_float4(0, 1.0f/k.y, -k.w/k.y, 0); invK.data[2] = make_float4(0, 0, 1, 0); invK.data[3] = make_float4(0, 0, 0, 1); return invK; } inline __host__ __device__ float3 make_float3(float4 a) { // printf("make_float3_start\n"); return make_float3(a.x, a.y, a.z); } inline __host__ __device__ float dot(float3 a, float3 b) { // printf("dot_start\n"); return a.x * b.x + a.y * b.y + a.z * b.z; } inline __host__ __device__ float3 rotate( const Matrix4 & M, const float3 & v){ // printf("start\n"); float3 ans = make_float3(dot(make_float3(M.data[0]), v),dot(make_float3(M.data[1]), v),dot(make_float3(M.data[2]), v)); // printf("end\n"); return ans; } std::ostream & operator<<( std::ostream & out, const Matrix4 & m ){ for(unsigned i = 0; i < 4; ++i) out << m.data[i].x << " " << m.data[i].y << " " << m.data[i].z << " " << m.data[i].w << "\n"; return out; } // __global__ void point2depthimage(Matrix4 r_inv, float* depth_image, float* point_sequence) { // printf("aaa\n"); // } __global__ void mykernel(float3* a, const Matrix4 r_inv, int max_num,Img* img){ // printf("%d %d %d %d \n", r_inv.data[0].x, r_inv.data[0].y, r_inv.data[0].z, r_inv.data[0].w); int index = get_pos(); if(index >= max_num) return ; // printf("index is %d x is %f, y is %f, z is %f\n", index,arr[index].x, arr[index].y, arr[index].z); float3 pixel_ = rotate(r_inv, a[index]); int depth = int(pixel_.z); int2 image_pos =make_int2(pixel_.x / pixel_.z, pixel_.y / pixel_.z); if(image_pos.x >= img->image_width || image_pos.y >= img->image_height ) return; else{ img->data[img->image_width * image_pos.y + image_pos.x] = depth; } printf("index is %d,x is %f, y is %f, z is %f, image_pos.x is %d, image_pos.z is %d\n",index,pixel_.x, pixel_.y, pixel_.z, image_pos.x, image_pos.y); } int num = 1000; float3* simulation_array; float3* test_gpu_array; int* img; int array_size = sizeof(float3)*num; float4 four_element = make_float4(1,1,1,1); int img_width = 100, img_height = 200; int* image_show; Img* img_show = new Img(img_width, img_height); // Matrix4 camera_M = getCameraMatrix(rotate_matrix); Matrix4 inv_camera_M = getInverseCameraMatrix(four_element); int main() { simulation_array = (float3*)malloc(array_size); image_show = (int*)malloc(img_show->img_arr_size); // mykernel<<<100,1000>>>(); for(int x = 0; x<num;x++){ simulation_array[x] = make_float3(rand()%10,rand()%10,rand()%10); // printf("%f, %f, %f\n", simulation_array[x].x, simulation_arr/ay[x].y,simulation_array[x].z); } cudaMalloc((void**)&test_gpu_array, array_size); cudaMemcpy(test_gpu_array,simulation_array,array_size, cudaMemcpyHostToDevice); // cout<<inv_camera_M; mykernel<<<100,10>>>(test_gpu_array, inv_camera_M, num, img_show); cudaMemcpy(image_show,img_show->data, img_show->img_arr_size, cudaMemcpyDeviceToHost); // printf("Hello World!\n"); cudaDeviceSynchronize(); ofstream ofile; ofile.open("image_show.txt"); for(int i=0; i<img_height;i++) for(int j=0;j<img_width;j++){ if(j == img_width-1) ofile<<image_show[i*img_width + j]<<'\n'; else ofile<<image_show[i*img_width + j]<<' '; } ofile.close(); cudaDeviceReset(); // uchar *ptmp = NULL; // Mat depth_image = Mat(img_height, img_width, CV_8UC1); // for(int i = 0 ;i<img_height;i++){ // // ptmp = depth_image.ptr<uchar>(i); // for(int ii = 0; ii < img_width; ii++){ // depth_image.at<uchar>(i,ii) = image_show[ii+i*img_show->image_width]; // } // } // // namedWindow("depth_image"); // imshow("depth_image", depth_image); // waitKey(0); }
1,217
void ocean() { int m = 1 << 7; int n = 1 << 7; }
1,218
#include <stdio.h> #include <cuda.h> //size of array #define N 4096 //vector addition kernel __global__ void vectorAddKernel(int *a, int *b, int *c) { int tdx = blockIdx.x * blockDim.x + threadIdx.x; if(tdx < N) { c[tdx] = a[tdx]+b[tdx]; } } int main() { //grid and block sizes dim3 grid(16, 1, 1); dim3 block(256, 1, 1); //host arrays int a_h[N]; int b_h[N]; int c_h[N]; //device memory pointers int *a_d; int *b_d; int *c_d; //load arrays with some numbers for(int i=0; i<N; i++) { a_h[i] = i; b_h[i] = i*1; } //allocate device memory cudaMalloc((void**)&a_d, N*sizeof(int)); cudaMalloc((void**)&b_d, N*sizeof(int)); cudaMalloc((void**)&c_d, N*sizeof(int)); //copy the host arrays to device cudaMemcpy(&a_d, &a_h, N*sizeof(int), cudaMemcpyHostToDevice); cudaMemcpy(&b_d, &b_h, N*sizeof(int), cudaMemcpyHostToDevice); cudaMemcpy(&c_d, &c_h, N*sizeof(int), cudaMemcpyHostToDevice); //CUDA events to measure time cudaEvent_t start; cudaEvent_t stop; float elapsedTime; //start timer cudaEventCreate(&start); cudaEventCreate(&stop); cudaEventRecord(start, 0); //launch kernel vectorAddKernel<<<grid, block>>>(a_d, b_d, c_d); //stop timer cudaEventRecord(stop, 0); cudaEventSynchronize(stop); cudaEventElapsedTime(&elapsedTime, start, stop); //copy the results to host cudaMemcpy(c_h, c_d, N*sizeof(int), cudaMemcpyDeviceToHost); //print the results for(int i=0; i<N; i++) { printf("%i+%i = %i\n",a_h[i], b_h[i], c_h[i]); } //print out execution time printf("Time to calculate results: %f ms.\n", elapsedTime); //clean up cudaFree(a_h); cudaFree(b_h); cudaFree(c_h); cudaEventDestroy(start); cudaEventDestroy(stop); return 0; }
1,219
#include <stdio.h> #include <time.h> #include <stdlib.h> #include <cuda.h> #define n 2 void fillMatrix(double *w, int li, int lj){ double count = 0; for(int i=0; i<li; i++){ for(int j=0; j<lj; j++){ w[i*lj+j] = count; count++; } } } void print(double *w, int li, int lj){ for(int i=0; i<li; i++){ for(int j=0; j<lj; j++){ printf("%.4lf ", w[i*lj+j]); } printf("\n"); } } __global__ void product(double *d_x, double *d_y, double *d_z){ int row = blockIdx.y*blockDim.y+threadIdx.y; int col = blockIdx.x*blockDim.x+threadIdx.x; double sum = 0; if ((row < n) && (col < n)){ for (int i = 0; i < n; i++) sum += d_x[n*row + i] * d_y[i*n+col]; d_z[row*n+col] = sum; } } int main(int argc, char const *argv[]) { int size1 = n*n*sizeof(double); int size2 = n*n*sizeof(double); int size3 = n*n*sizeof(double); double *x = (double*)malloc(size1); double *y = (double*)malloc(size2); double *z = (double*)malloc(size3); fillMatrix(x,n,n); fillMatrix(y,n,n); clock_t begin, end; double time_spent; begin = clock(); double *d_x = (double*)malloc(size1); double *d_y = (double*)malloc(size2); double *d_z = (double*)malloc(size3); cudaMalloc((void**)&d_x, size1); cudaMalloc((void**)&d_y, size2); cudaMalloc((void**)&d_z, size3); cudaMemcpy(d_x, x, size1, cudaMemcpyHostToDevice); cudaMemcpy(d_y, y, size2, cudaMemcpyHostToDevice); int threads = 16; dim3 dimBlock(threads,threads); dim3 dimGrid((n+dimBlock.x-1)/dimBlock.x, (n+dimBlock.y-1)/dimBlock.y); product<<<dimGrid,dimBlock>>>(d_x, d_y, d_z); cudaMemcpy(z,d_z,size3,cudaMemcpyDeviceToHost); print(x,n,n); printf("\n"); print(y,n,n); printf("\n"); print(z,n,n); end = clock(); time_spent = (double)(end - begin) / CLOCKS_PER_SEC; printf("%lf\n", time_spent); return 0; }
1,220
#include<stdio.h> #include<cuda_runtime.h> #include <stdlib.h> #define length 10 #define length_thread 256 #define test(a){\ for(int i =0;i<length;i++){\ printf("a[%d] = %d \n",i,a[i] );\ }\ } #define pr_array(a,start,end){\ for(int i=start;i<=end;i++){\ printf("a[%d] = %d\n",i,a[i]);\ }\ } //return b_end where the vuela a will be put in the new array //if a equel a vuale which is in the b array,a will in front of the vuale; #define insert0(a,b,b_start,b_end){\ while((b_end-b_start)>1){\ int point = (b_start+b_end)/2;\ if(a<=b[point])b_end = point;\ else b_start = point;\ }\ b_end += (a>b[b_end])-(a<=b[b_start]);\ }\ //if a equel a vuale which is in the b array,a will in back of the vuale; #define insert1(a,b,b_start,b_end){\ while((b_end-b_start)>1){\ int point = (b_start+b_end)/2;\ if(a<b[point])b_end = point;\ else b_start = point;\ }\ b_end += (a>=b[b_end])-(a<b[b_start]);\ }\ int cmp(const void *a,const void *b) { return *(int *)a-*(int *)b; } __global__ void pr(int *a){ int tid = threadIdx.x; __shared__ int a_s[length]; a_s[tid] = a[tid]; a[tid] = a_s[tid]*2; } __global__ void merger_thread(int *a,int len){ __shared__ int a_s[length_thread]; int tid = threadIdx.x + blockIdx.x * blockDim.x; int r1,r2; if(tid < len){ a_s[threadIdx.x] = a[tid]; } r1 = blockDim.x/2; int flag = (threadIdx.x>=r1); tid = threadIdx.x%r1; if((gridDim.x-1) == blockIdx.x)len %= blockDim.x; else len = blockDim.x; __syncthreads(); if(threadIdx.x<(len/2)){ r1 = threadIdx.x*2; if(a_s[r1]>a_s[r1+1]){ r2 = a_s[r1]; a_s[r1] = a_s[r1+1]; a_s[r1+1] = r2; } } int loop = 1; int x , start , end; x = len; x = (x/2) + (x%2); while(x>1){ x = (x/2) + (x%2); r1 = tid>>loop; r1 *= 2; start = r1+1-(int)flag; r1 += flag; end = 1<<loop; r1 *= end; start *= end; r2 = tid % end; r1 += r2; end += (start-1); __syncthreads(); if(end > len)end = len; if(r1 < len){ r1 = a_s[r1]; if(flag){ insert1(r1,a_s,start,end); } else{ insert0(r1,a_s,start,end); } end %= (1<<loop); r2 +=end; a_s[r2] = r1; } loop++; __syncthreads(); } if(threadIdx.x < len){ a[threadIdx.x + blockIdx.x * blockDim.x] = a_s[threadIdx.x]; } } int random(int range){ static int start = 444; int d = 233333; int k = 33352; start = ((start*k)+d)%range; return start; } int* gen(){ int *a_h; cudaHostAlloc( (void**)&a_h, length* sizeof(int),cudaHostAllocDefault ); for (int i = 0; i < length; ++i) { a_h[i] = random(52); printf("a_h[%d] = %d \n",i,a_h[i]); } return a_h; } void sort_int(int *a,int len){ int block_num = len/length_thread; if((len%length_thread)!=0)block_num++; merger_thread<<<block_num,length_thread>>>(a,len); } void msort(void *a,size_t num,size_t size,int ( * comparator ) ( const void *, const void * ) ){ } int main(){ /* int *a_h = gen(); test(a_h); qsort(a_h,length,sizeof(int),cmp); printf("sorted!!!!!!!\n"); test(a_h); a_h[43]=10; a_h[42]=10; pr_array(a_h,10,100); int start = 10,end = 50; insert0(10,a_h,start,end); printf("end %d\n",end ); start = 10; end=50; insert1(10,a_h,start,end); printf("end %d\n",end ); printf("used ipad pro maked\n"); */ int *a_h; cudaHostAlloc( (void**)&a_h, length* sizeof(int),cudaHostAllocDefault ); for (int i = 0;i<length;i++) { a_h[i] = length - i; printf("a_h[%d] = %d \n",i,a_h[i]); } int *a_d; cudaMalloc( (void**)&a_d, length*sizeof(int) ); cudaMemcpy(a_d,a_h, length*sizeof(int),cudaMemcpyHostToDevice); sort_int(a_d,length); cudaMemcpy(a_h,a_d,length*sizeof(int),cudaMemcpyDeviceToHost); for (int i = 0;i<length;i++) { printf("a_h[%d] = %d \n",i,a_h[i]); } }
1,221
/* #include "cuda_runtime.h" #include "device_launch_parameters.h" #include <stdlib.h> #include <stdio.h> #include <time.h> __global__ void addMultipleBlocks(float *d_a,float *d_b,float *d_c,int m,int n) { int i=blockIdx.x*blockDim.x+ threadIdx.x; if(i<(m*n)) d_c[i]=d_a[i]+d_b[i]; } __global__ void addSingleBlock(float *d_a,float *d_b,float *d_c,int m,int n) { int i=threadIdx.x; d_c[i]=d_a[i]+d_b[i]; } int main() { float *a,*b,*c; float *d_a,*d_b,*d_c; int size,m,n,i,j; time_t start; printf("Enter the number of rows and columns : \n"); scanf("%d%d",&m,&n); a=(float *)malloc((size=sizeof(float)*m*n)); b=(float *)malloc(size); c=(float *)malloc(size); for(i=0;i<m;i++) for(j=0;j<n;j++) { a[i*n+j]=i*2+j+0.8; b[i*n+j]=i*2+j+0.1; } cudaMalloc((void **)&d_a,size); cudaMalloc((void**)&d_b,size); cudaMalloc((void **)&d_c,size); cudaMemcpy(d_a,a,size,cudaMemcpyHostToDevice); cudaMemcpy(d_b,b,size,cudaMemcpyHostToDevice); start=time(0); addSingleBlock<<<1,(m*n)>>>(d_a,d_b,d_c,m,n); cudaMemcpy(c,d_c,size,cudaMemcpyDeviceToHost); printf("Result using a single block : \n"); for(i=0;i<m;i++) { for(j=0;j<n;j++) printf("%f ",c[i*n+j]); printf("\n"); } printf("\n\n\n TIME TAKEN TO COMPUTER THE ADDITION WITH SINGLE BLOCK - %f",(difftime(time(0),start))); cudaFree(d_c); free(c); cudaMalloc((void **)&d_c,size); c=(float *)malloc(size); start=time(0); addMultipleBlocks<<< (int)((m*n)/9.0+1),9>>>(d_a,d_b,d_c,m,n); cudaMemcpy(c,d_c,size,cudaMemcpyDeviceToHost); printf("Result using a multiple blocks(3x3) : \n"); for(i=0;i<m;i++) { for(j=0;j<n;j++) printf("%f ",c[i*n+j]); printf("\n"); } printf("\n\n\n TIME TAKEN TO COMPUTER THE ADDITION WITH MULTIPLE BLOCKS - %f",(difftime(time(0),start))); free(a); free(b); free(c); cudaFree(d_a); cudaFree(d_b); cudaFree(d_c); return 0; } */
1,222
#include <iostream> #include <thrust/device_vector.h> #include <thrust/host_vector.h> #include <thrust/reduce.h> using namespace std; int main(int argc, const char *argv[]) { string N; if (argc > 1) { N = string(argv[1]); } unsigned int n = atoi(N.c_str()); thrust::host_vector<int> H(n); for (unsigned int i = 0; i < n; i++) { H[i] = 1; } thrust::device_vector<int> D = H; cudaEvent_t start; cudaEvent_t stop; cudaEventCreate(&start); cudaEventCreate(&stop); cudaEventRecord(start); int sum = thrust::reduce(D.begin(), D.end()); cudaEventRecord(stop); cudaEventSynchronize(stop); // Get the elapsed time in milliseconds float ms; cudaEventElapsedTime(&ms, start, stop); cout << sum << endl; cout << ms << endl; return 0; }
1,223
#include<cuda_runtime.h> #include<device_launch_parameters.h> #include<stdio.h> #include<stdlib.h> #include<string.h> __global__ void q1(int* d_a,int* d_r,int *d_m) { int n = threadIdx.x; for(int i = 0;i<(*d_m);i++) { d_r[n*(*d_m)+i] = d_a[n*(*d_m)+i]; for(int j = 0;j<n;j++) d_r[n*(*d_m)+i] *= d_a[n*(*d_m)+i]; } } int main(void) { int *a,*r,m,n,i; int *d_a,*d_r,*d_m; printf("Enter m,n : "); scanf("%d %d",&m,&n); a = (int*)malloc(m*n*sizeof(int)); r = (int*)malloc(m*n*sizeof(int)); printf("Enter matrix:\n"); for(i=0;i<m*n;i++) { scanf("%d",&a[i]); } cudaMalloc((void **)&d_a,(m*n)*sizeof(int)); cudaMalloc((void **)&d_r,(m*n)*sizeof(int)); cudaMalloc((void **)&d_m,sizeof(int)); cudaMemcpy(d_a,a,(m*n)*sizeof(int),cudaMemcpyHostToDevice); cudaMemcpy(d_r,r,(m*n)*sizeof(int),cudaMemcpyHostToDevice); cudaMemcpy(d_m,&m,sizeof(int),cudaMemcpyHostToDevice); q1<<<1,n>>>(d_a,d_r,d_m); cudaError_t error = cudaGetLastError(); if(error!= cudaSuccess) { printf("%s\n",cudaGetErrorString(error)); } cudaMemcpy(r,d_r,(m*n)*sizeof(int),cudaMemcpyDeviceToHost); printf("Result matrix :\n"); for(i=0;i<m*n;i++) { printf("%d\t",r[i]); if((i+1)%m==0) printf("\n"); } }
1,224
#include "includes.h" __global__ void calculateDelaysAndPhases(double * gpuDelays, double lo, double sampletime, int fftsamples, int fftchannels, int samplegranularity, float * rotationPhaseInfo, int *sampleShifts, float* fractionalSampleDelays) { size_t ifft = threadIdx.x + blockIdx.x * blockDim.x; size_t iant = blockIdx.y; int numffts = blockDim.x * gridDim.x; double meandelay, deltadelay, netdelaysamples_f, startphase; double d0, d1, d2, a, b; double * interpolator = &(gpuDelays[iant*4]); double filestartoffset = gpuDelays[iant*4+3]; float fractionaldelay; int netdelaysamples; // evaluate the delay for the given FFT of the given antenna // calculate values at the beginning, middle, and end of this FFT d0 = interpolator[0]*ifft*ifft + interpolator[1]*ifft + interpolator[2]; d1 = interpolator[0]*(ifft+0.5)*(ifft+0.5) + interpolator[1]*(ifft+0.5) + interpolator[2]; d2 = interpolator[0]*(ifft+1.0)*(ifft+1.0) + interpolator[1]*(ifft+1.0) + interpolator[2]; // use these to calculate a linear interpolator across the FFT, as well as a mean value a = d2-d0; //this is the delay gradient across this FFT b = d0 + (d1 - (a*0.5 + d0))/3.0; //this is the delay at the start of the FFT meandelay = a*0.5 + b; //this is the delay in the middle of the FFT deltadelay = a / fftsamples; // this is the change in delay per sample across this FFT window netdelaysamples_f = (meandelay - filestartoffset) / sampletime; netdelaysamples = __double2int_rn(netdelaysamples_f/samplegranularity) * samplegranularity; // Save the integer number of sample shifts sampleShifts[iant*numffts + ifft] = netdelaysamples; // Save the fractional delay fractionaldelay = (float)(-(netdelaysamples_f - netdelaysamples)*2*M_PI/fftsamples); // radians per FFT channel fractionalSampleDelays[iant*numffts + ifft] = fractionaldelay; // set the fringe rotation phase for the first sample of a given FFT of a given antenna startphase = b*lo; rotationPhaseInfo[iant*numffts*2 + ifft*2] = (float)(startphase - int(startphase))*2*M_PI; rotationPhaseInfo[iant*numffts*2 + ifft*2 + 1] = (float)(deltadelay * lo)*2*M_PI; }
1,225
/* This is a automatically generated test. Do not modify */ #include <stdio.h> #include <stdlib.h> #include <math.h> __global__ void compute(float comp, int var_1,float var_2,int var_3,int var_4,float var_5,float var_6,float var_7,float var_8,float var_9,float var_10,float var_11,float* var_12,float var_13,float var_14,float var_15,float var_16,float var_17,float var_18,float var_19,float var_20,float var_21,float* var_22,float* var_23,float var_24,float var_25,float var_26,float var_27,float var_28,float var_29,float var_30,float var_31,float var_32,float var_33) { for (int i=0; i < var_1; ++i) { if (comp > (var_2 + -1.5747E-35f)) { comp += coshf((+0.0f / +1.3473E-43f)); if (comp >= -1.3494E36f - (-0.0f * (+1.7416E-44f - var_5))) { comp += asinf((var_6 + var_7)); float tmp_1 = -0.0f + (+0.0f * var_8); float tmp_2 = -1.9823E34f / +0.0f - var_9; comp = tmp_2 * tmp_1 + (var_10 * var_11); } for (int i=0; i < var_3; ++i) { var_12[i] = var_13 + var_14 * fabsf((var_15 - var_16 * ldexpf(floorf(tanhf((-1.9774E-35f * var_17 + var_18))), 2))); float tmp_3 = +1.5409E26f; comp = tmp_3 / var_12[i] / -1.2900E-35f + var_19 / (var_20 / (-1.3379E-15f - -1.8224E-35f - var_21)); } for (int i=0; i < var_4; ++i) { var_22[i] = sinhf(acosf(var_24 + +1.4771E-35f)); var_23[i] = +1.8351E-35f * (-1.0405E10f * (var_25 / (var_26 * fmodf((-1.5535E35f + -1.3174E-24f * -1.2706E-41f - (-1.4619E-35f - var_27)), (var_28 + (var_29 - (var_30 / ceilf(+1.0588E-43f)))))))); comp += var_23[i] + var_22[i] + powf(var_31 - (var_32 / -1.1886E34f * -0.0f - (var_33 - +0.0f)), -1.0284E-35f); } } } printf("%.17g\n", comp); } float* initPointer(float v) { float *ret = (float*) malloc(sizeof(float)*10); for(int i=0; i < 10; ++i) ret[i] = v; return ret; } int main(int argc, char** argv) { /* Program variables */ float tmp_1 = atof(argv[1]); int tmp_2 = atoi(argv[2]); float tmp_3 = atof(argv[3]); int tmp_4 = atoi(argv[4]); int tmp_5 = atoi(argv[5]); float tmp_6 = atof(argv[6]); float tmp_7 = atof(argv[7]); float tmp_8 = atof(argv[8]); float tmp_9 = atof(argv[9]); float tmp_10 = atof(argv[10]); float tmp_11 = atof(argv[11]); float tmp_12 = atof(argv[12]); float* tmp_13 = initPointer( atof(argv[13]) ); float tmp_14 = atof(argv[14]); float tmp_15 = atof(argv[15]); float tmp_16 = atof(argv[16]); float tmp_17 = atof(argv[17]); float tmp_18 = atof(argv[18]); float tmp_19 = atof(argv[19]); float tmp_20 = atof(argv[20]); float tmp_21 = atof(argv[21]); float tmp_22 = atof(argv[22]); float* tmp_23 = initPointer( atof(argv[23]) ); float* tmp_24 = initPointer( atof(argv[24]) ); float tmp_25 = atof(argv[25]); float tmp_26 = atof(argv[26]); float tmp_27 = atof(argv[27]); float tmp_28 = atof(argv[28]); float tmp_29 = atof(argv[29]); float tmp_30 = atof(argv[30]); float tmp_31 = atof(argv[31]); float tmp_32 = atof(argv[32]); float tmp_33 = atof(argv[33]); float tmp_34 = atof(argv[34]); compute<<<1,1>>>(tmp_1,tmp_2,tmp_3,tmp_4,tmp_5,tmp_6,tmp_7,tmp_8,tmp_9,tmp_10,tmp_11,tmp_12,tmp_13,tmp_14,tmp_15,tmp_16,tmp_17,tmp_18,tmp_19,tmp_20,tmp_21,tmp_22,tmp_23,tmp_24,tmp_25,tmp_26,tmp_27,tmp_28,tmp_29,tmp_30,tmp_31,tmp_32,tmp_33,tmp_34); cudaDeviceSynchronize(); return 0; }
1,226
/** * Copyright 1993-2012 NVIDIA Corporation. All rights reserved. * * Please refer to the NVIDIA end user license agreement (EULA) associated * with this source code for terms and conditions that govern your use of * this software. Any use, reproduction, disclosure, or distribution of * this software and related documentation outside the terms of the EULA * is strictly prohibited. */ #include <stdio.h> #include <stdlib.h> #include <sys/time.h> static const int WORK_SIZE = 256; /** * This macro checks return value of the CUDA runtime call and exits * the application if the call failed. */ #define CUDA_CHECK_RETURN(value) { \ cudaError_t _m_cudaStat = value; \ if (_m_cudaStat != cudaSuccess) { \ fprintf(stderr, "Error %s at line %d in file %s\n", \ cudaGetErrorString(_m_cudaStat), __LINE__, __FILE__); \ exit(1); \ } } __device__ unsigned int bitreverse(unsigned int number) { number = ((0xf0f0f0f0 & number) >> 4) | ((0x0f0f0f0f & number) << 4); number = ((0xcccccccc & number) >> 2) | ((0x33333333 & number) << 2); number = ((0xaaaaaaaa & number) >> 1) | ((0x55555555 & number) << 1); return number; } /** * CUDA kernel function that reverses the order of bits in each element of the array. */ __global__ void bitreverse(void *data) { unsigned int *idata = (unsigned int*) data; idata[threadIdx.x] = bitreverse(idata[threadIdx.x]); } __global__ void add(int *a, int *b, int *c) { //c[blockIdx.x] = a[blockIdx.x] + b[blockIdx.x]; int index = threadIdx.x + blockIdx.x * blockDim.x; c[index] = a[index] + b[index]; } /** * Host function that prepares data array and passes it to the CUDA kernel. */ #define THREADS_PER_BLOCK 1024 int main(int argc, char *argv[]) { struct timeval t0; struct timeval t1; int N = atoi(argv[1]); int *a, *b, *c; // host copies of a, b, c int *d_a, *d_b, *d_c; // device copies of a, b, c int size = N * sizeof(int); // Allocate space for device copies of a, b, c cudaMalloc((void **)&d_a, size); cudaMalloc((void **)&d_b, size); cudaMalloc((void **)&d_c, size); // Setup input values a = (int *)malloc(size); b = (int *)malloc(size); c = (int *)malloc(size); int i = 0; for(i = 0; i < N; i++) { a[i] = i;//rand() % 100; b[i] = i;//rand() % 100; } gettimeofday(&t0,0); // Copy inputs to device cudaMemcpy(d_a, a, size, cudaMemcpyHostToDevice); cudaMemcpy(d_b, b, size, cudaMemcpyHostToDevice); // Launch add() kernel on GPU with N blocks add<<<N/THREADS_PER_BLOCK,THREADS_PER_BLOCK>>>(d_a, d_b, d_c); // Copy result back to host cudaMemcpy(c, d_c, size, cudaMemcpyDeviceToHost); gettimeofday(&t1,0); double time_spent = (t1.tv_sec-t0.tv_sec) + (double)(t1.tv_usec-t0.tv_usec)/1000000; for(i = 0; i < N; i++) { printf("%d\n",c[i]); } printf("\nTime Spent GPU: %f\n\n",time_spent); // Cleanup cudaFree(d_a); cudaFree(d_b); cudaFree(d_c); // CPU Part gettimeofday(&t0,0); for(i = 0; i < N; i++) { c[i] = a[i]+b[i]; } gettimeofday(&t1,0); time_spent = (t1.tv_sec-t0.tv_sec) + (double)(t1.tv_usec-t0.tv_usec)/1000000; printf("\nTime Spent CPU: %f\n\n",time_spent); return 0; }
1,227
extern "C" __device__ float vxd(const float m,float Vy, float w, float ZpTei, float Xr, float Kvx) { if (Vy > 0) { return ((m * Vy * w - Xr + ZpTei) * Kvx); } else { return ((m * Vy * w * 1.09f - Xr + ZpTei) * Kvx);//Vx*1.061 ;N0=3;k11=580.91f Ubuntu //return ((m * Vy * w - Xr + ZpTei) * Kvx); }//*1.1045; k11=5.91f windows ; 1.09; k11=580.91f } extern "C" __device__ float vyd(const float m,float Vx, float w, float ZpYri, float Yr, float Kvy) { if (Vx > 0) { return ((Yr - ZpYri - m * Vx * w * 1.0165f) * Kvy);//Vx*1.0179 //return ((Yr - ZpYri - m * Vx * w) * Kvy); } else {//*1.0165; k11=5.91f windows return ((Yr - ZpYri - m * Vx * w) * Kvy); } } extern "C" __device__ float wd(const float lr,float Mr, float ZpYri, float Yr, float Kw) { return ((Mr + lr * ZpYri) * Kw); } extern "C" __global__ void modelCalc(float* VX, float* VY, float* ww, float* X, float* Y, float* W, float* Xobs, float* Yobs, float T, float L, float B,float m,float lr,float Vx, float Vy, float v,float w, const int size,const int nShips) { int j = blockDim.x * blockIdx.x + threadIdx.x; if (j<nShips){ float k11, k22, k66, k26; const float delta = 0.7f;// float c1, c2, c3, m1, m2;// .. float w_ = 0.0f;// approx float betta_d = 0.0f;// (rad) float Cxr, Cyr, Cmr;// Cxr, , // float L1=B,T1=L/2;//, ???????????????????? const float p = 1000.0f;// float Xr, Yr, Mr;// float ZpTei = 0;// =R+Xa float Jz;// Gz float ZpYri;// const float a66 = 0.31f;// //float Ramp = 0.0f;// const float lyamdaR = 1.4f;// c 335 //0.5--1.4 //1 const float deltaR = 0.349f;// - 20% //0.349f // float Va;// float Yri, Ysi;// const float Ar = 6.0f;// //5 const float D = 1.5f;// //2 const float Cta = 10.0f, Ctt = 50.0f;//c 334 //Cta<=20, 0.5--30, Ctt<=20, 1--50 UP!!! //Ctt =30 float Kvx, Kvy, Kw;// const float No = 3.0f;// ////3 const float Re = 5000000.0f;// >5E8 float K_betta; float fit; const float xk = 1.0f; const float bettar = 0.9f; const float fik = 0.95f; float ld_;// c228 float betta_D;// float fiD;//c 338 float CyD1; float CyD; float A0;// float xD;//c 339 float viv = 0.0f;// 27 ch3_2 float Rmatr[3][3]={cosf(viv),-sinf(viv),0.0f,sinf(viv),cosf(viv),0.0f,0.0f,0.0f,1.0f}; //------------------------------------------------------------ //c 330 5.91- , 580.91 - k11 = (580.91f * (float) pow(B / L, 2.0f) + 7.76f * (B / L) - 0.259f) / (48.4f - 6.89f * (B / T) + 1.47f * (float) pow(B / T, 2.0f) - 0.0475f * (float) pow(B / T, 3.0f)); k22 = ((0.722f + 0.224f * delta) * (1.022f - (float) pow(B / L, 2.0f))) / (0.264f + 0.368f * (B / T)); k66 = (2.0f * T / B) * (2.59f + 0.781f * delta) * (0.357f - 1.77f * (float) pow(B / L, 2.0f)); k26 = k22; //k26=0; //c 323 c1 = 3.14f * (T / L) * (float) pow((0.63f / delta), (5.0f / 2.0f)) * (float) pow(L / (6.0f * B), (1.0f / 3.0f)) - 0.032f; c2 = -2.0f * k11 * delta * (B / L); c3 = 1.35f * (float) pow(T / B, (1.0f / 2.0f)) * (float) pow((0.63f / delta), (3.0f / 2.0f)) - 0.029f; m1 = 1.67f * (T / L) - 0.56f * delta + 0.43f; m2 = -0.44f * (T / L) - 0.0375f; // System.out.printf("c1=%f\tc2=%f\tc3=%f\tm1=%f\tm2=%f\n", c1, c2, c3, m1, m2); Jz = (m * (float) pow(L, 2.0f) / 12.4f) * (0.463f + 0.574f * (float) pow(delta, a66) + (float) pow(B / L, 2.0f));//c 330 Kvx = 1 / (m * (1 + k11)); Kvy = 1 / (m * (1 + k22)); Kw = 1 / (Jz * (1 + k66));//????? m //---------------------------------------------------------------- float k1, k2, k3, k4; float q1, q2, q3, q4; float z1, z2, z3, z4; float j1, j2, j3, j4; //t = 0.0f; // int t = 0; float h = 1.0f; for (int i = 0; i < size; i++) { //16550 v = (float) sqrt((float) pow(Vx, 2.0f) + (float) pow(Vy, 2.0f)); //assert(Vx==0); if (Vx != 0) { //c 353 ????????????????????????? w_ = w * L / v;//?????????????????????? betta_d = -(float) atan(Vy / Vx);//c 350 } else { w_ = w * L / v; //betta_d = 0; betta_d = -(float) atan(Vy / Vx);//c 350 } Cxr = 0.01f * (1.0f + 170.0f * (T / L));// c 119 Cyr = c1 * betta_d + c2 * w_ + c3 * betta_d * abs(betta_d);//c 323 Cmr = m1 * betta_d + m2 * w_; Xr = Cxr * L * T * (float) pow(v, 2.0f) * p / 2.0f;//c 320 Yr = Cyr * L * T * (float) pow(v, 2.0f) * p / 2.0f; Mr = Cmr * L * T * (float) pow(v, 2.0f) * p / 2.0f; K_betta = 0.43f * (float) pow(Ctt, -0.6f); fit = (float) pow(1.0f + Ctt, 0.508f); //IMPORTANT!!! deltaR Yri = 3.14f * (deltaR - K_betta * xk * (betta_d + lr * w_)) * p * Ar * (float) pow(v * fik * fit, 2.0f) / (1.0f + 2.2f / (float) pow(lyamdaR, 2.0f / 3.0f)); ld_ = 0.77f - 0.125f * (float) sqrt(Ctt) / (1.65f * (float) sqrt(Ctt) - 1.0f); betta_D = 1.22f - 0.0563f * (float) sqrt(Ctt) / (1.65f * (float) sqrt(Ctt) - 1.0f); fiD = 0.5f * ((float) sqrt(1.0f + 2.0f * Ctt / betta_D) + 1.0f); CyD1 = 12.0f * ld_ / (1.0f + 1.56f * ld_); CyD = CyD1 + 2.0f * betta_D * (float) pow(fiD, 2.0f); xD = xk * (CyD1 + 2.0f * betta_D * fiD) / (CyD1 + 2.0f * betta_D * (float) pow(fiD, 2.0f)); A0 = 3.14f * (float) pow(D, 2.0f) / 4.0f; Ysi = CyD * (xD - 0.02f * xk) * (betta_d + lr * w_) * (p / 2.0f) * A0 * (float) pow(v, 2.0f) * (float) pow(fik, 2.0f); ZpTei = 1000000.0f * (9.740f * (float) pow(No, 2.0f) - 2.23f * v); ////9.740f ZpYri = 2.0f * (Yri - Ysi);//2 k1 = h * vxd(m,Vy, w, ZpTei, Xr, Kvx); q1 = h * vyd(m,Vx, w, ZpYri, Yr, Kvy); z1 = h * wd(lr,Mr, ZpTei, Yr, Kw); k2 = h * vxd(m,Vy + q1 / 2.0f, w + z1 / 2.0f, ZpTei, Xr, Kvx); q2 = h * vyd(m,Vx + k1 / 2.0f, w + z1 / 2.0f, ZpYri, Yr, Kvy); z2 = h * wd(lr,Mr, ZpYri, Yr, Kw); k3 = h * vxd(m,Vy + q2 / 2.0f, w + z2 / 2.0f, ZpTei, Xr, Kvx); q3 = h * vyd(m,Vx + k2 / 2.0f, w + z2 / 2.0f, ZpYri, Yr, Kvy); z3 = h * wd(lr,Mr, ZpYri, Yr, Kw); k4 = h * vxd(m,Vy + q3, w + z3, ZpTei, Xr, Kvx); q4 = h * vyd(m,Vx + k3, w + z3, ZpYri, Yr, Kvy); z4 = h * wd(lr,Mr, ZpYri, Yr, Kw); Vx = Vx + (1.0f / 6.0f) * (k1 + 2.0f * k2 + 2.0f * k3 + k4); //VX[t] = Vx / 1.24f; VX[t] = Vx; Vy = Vy + (1.0f / 6.0f) * (q1 + 2.0f * q2 + 2.0f * q3 + q4); VY[t] = Vy; w = w + (1.0f / 6.0f) * (z1 + 2.0f * z2 + 2.0f * z3 + z4); ww[t] = w; //--- ----------------------------------------- k1 = h * vxd(m,Vy, w, ZpTei, Xr, Kvx); q1 = h * vyd(m,Vx, w, ZpYri, Yr, Kvy); z1 = h * wd(lr,Mr, ZpTei, Yr, Kw); k2 = h * vxd(m,Vy + q1 / 2.0f, w + z1 / 2.0f, ZpTei, Xr, Kvx); q2 = h * vyd(m,Vx + k1 / 2.0f, w + z1 / 2.0f, ZpYri, Yr, Kvy); z2 = h * wd(lr,Mr, ZpYri, Yr, Kw); k3 = h * vxd(m,Vy + q2 / 2.0f, w + z2 / 2.0f, ZpTei, Xr, Kvx); q3 = h * vyd(m,Vx + k2 / 2.0f, w + z2 / 2.0f, ZpYri, Yr, Kvy); z3 = h * wd(lr,Mr, ZpYri, Yr, Kw); k4 = h * vxd(m,Vy + q3, w + z3, ZpTei, Xr, Kvx); q4 = h * vyd(m,Vx + k3, w + z3, ZpYri, Yr, Kvy); z4 = h * wd(lr,Mr, ZpYri, Yr, Kw); X[t] = Vx + (1.0f / 6.0f) * (k1 + 2.0f * k2 + 2.0f * k3 + k4); Y[t] = Vy + (1.0f / 6.0f) * (q1 + 2.0f * q2 + 2.0f * q3 + q4); W[t] = w + (1.0f / 6.0f) * (z1 + 2.0f * z2 + 2.0f * z3 + z4); // viv = W[t]; //------- Rmatr[0][0] = (float)cos(viv); Rmatr[0][1] = -(float)sin(viv); Rmatr[1][0] = (float)sin(viv); Rmatr[1][1] = (float)cos(viv); Xobs[t] = Rmatr[0][0] * X[t] + Rmatr[0][1] * Y[t]; Yobs[t] = Rmatr[1][0] * X[t] + Rmatr[1][1] * Y[t]; //---------- t++; } } }
1,228
__global__ void mapping(double *point_cloud, const double *img1, const double *img2, const double *img3, const double *img4, const double *T, const double *P1, const double *P2, const double *P3, const double *P4, const double *ratiox, const double *ratioy, const double *ratioz, const int *img_width, const int *img_height) { int idx = blockIdx.x*blockDim.x + threadIdx.x; int idy = blockIdx.y*blockDim.y + threadIdx.y; int idz = threadIdx.z; int width = gridDim.x*blockDim.x; int height = gridDim.y*blockDim.y; int index = idz*width*height + idx*height + idy; //grid to coordinates double mx = idx*ratiox[0]; double my = idy*ratioy[0]; double mz = idz*ratioz[0]; //world coordinates double x = T[0]*mx + T[4]*my + T[8]*mz + T[12]; double y = T[1]*mx + T[5]*my + T[9]*mz + T[13]; double z = mz; //image coordinates double u1 = P1[0]*x + P1[3]*y + P1[6]*z + P1[9]; double v1 = P1[1]*x + P1[4]*y + P1[7]*z + P1[10]; double norm1 = P1[2]*x + P1[5]*y + P1[8]*z + P1[11]; u1/=norm1; v1/=norm1; double u2 = P2[0]*x + P2[3]*y + P2[6]*z + P2[9]; double v2 = P2[1]*x + P2[4]*y + P2[7]*z + P2[10]; double norm2 = P2[2]*x + P2[5]*y + P2[8]*z + P2[11]; u2/=norm2; v2/=norm2; double u3 = P3[0]*x + P3[3]*y + P3[6]*z + P3[9]; double v3 = P3[1]*x + P3[4]*y + P3[7]*z + P3[10]; double norm3 = P3[2]*x + P3[5]*y + P3[8]*z + P3[11]; u3/=norm3; v3/=norm3; double u4 = P4[0]*x + P4[3]*y + P4[6]*z + P4[9]; double v4 = P4[1]*x + P4[4]*y + P4[7]*z + P4[10]; double norm4 = P4[2]*x + P4[5]*y + P4[8]*z + P4[11]; u4/=norm4; v4/=norm4; int u11 = (u1); int u22 = (u2); int u33 = (u3); int u44 = (u4); int v11 = (v1); int v22 = (v2); int v33 = (v3); int v44 = (v4); int final_width = img_width[0]; int final_height = img_height[0]; int seen_record[4] = {0}; //decide the point cloud if((u11>0)&&(u11<final_width)&&(v11>0)&&(v11<final_height)) seen_record[0] = 1; if((u22>0)&&(u22<final_width)&&(v22>0)&&(v22<final_height)) seen_record[1] = 1; if((u33>0)&&(u33<final_width)&&(v33>0)&&(v33<final_height)) seen_record[2] = 1; if((u44>0)&&(u44<final_width)&&(v44>0)&&(v44<final_height)) seen_record[3] = 1; int sum = 0; if((seen_record[0]==1)&&(img1[u11*final_height+v11]!=0)) ++sum; if((seen_record[1]==1)&&(img2[u22*final_height+v22]!=0)) ++sum; if((seen_record[2]==1)&&(img3[u33*final_height+v33]!=0)) ++sum; if((seen_record[3]==1)&&(img4[u44*final_height+v44]!=0)) ++sum; point_cloud[index] = 0; if(sum>=3) { point_cloud[index] = 1; } }
1,229
#include "includes.h" using namespace std; // parameter describing the size of matrix A const int rows = 4096; const int cols = 4096; const int BLOCK_SIZE = 16; // transpose shared kernel // transpose kernel __global__ void transpose_naive(float* a, float*b) { int x = blockIdx.x * blockDim.x + threadIdx.x; int y = blockIdx.y * blockDim.y + threadIdx.y; int width = gridDim.x * blockDim.x; int height = gridDim.y * blockDim.y; // perform transpose if (x < height && y < width) { b[x*height + y] = a[y*width + x]; } }
1,230
#include <stdio.h> template <typename T> struct SinFunctor { __host__ __device__ T operator()(const T& x) const { return sinf(x); } }; template <typename T> struct CosFunctor { __host__ __device__ T operator()(const T& x) const { return cosf(x); } }; struct OrderFunctor { const float num_oscilators; OrderFunctor(float _num_oscilators) : num_oscilators(_num_oscilators) {} __host__ __device__ float operator()(const float& x, const float& y) const { return sqrtf(powf(x, 2)+powf(y, 2))/num_oscilators; } }; __global__ void kernel_heuns(int num_oscilators, int num_couplings, float *phases_new, float *phases_old, float *omegas,float *couplings, float dt, int *indices, int *ptr) { int i_thread = blockIdx.x * blockDim.x + threadIdx.x; int stride = blockDim.x * gridDim.x; for (int index =i_thread; index < num_oscilators*num_couplings; index += stride){ float findex = (float)index; float fnum_oscilators = (float)num_oscilators; float fi_section = (findex+1)/fnum_oscilators-1; fi_section=ceil(fi_section); int i_section = (int)fi_section; //int i_section = (int)ceil(((float)index+1)/(float)num_oscilators-1); int i_node = index - i_section*num_oscilators; float coupling = couplings[i_section]; float phase = phases_old[index]; float omega = omegas[i_node]; int i_nei = 0; float f = 0; float f_tild = 0; float phase_tild = 0.; for( int i_ptr = ptr[i_node]; i_ptr < ptr[i_node+1]; i_ptr = i_ptr + 1 ) { i_nei = indices[i_ptr]+ i_section*num_oscilators; f += coupling*sinf( phases_old[i_nei] -phase); }; f = f + omega; phase_tild = phase + dt*f; for( int i_ptr = ptr[i_node]; i_ptr < ptr[i_node+1]; i_ptr = i_ptr + 1 ) { i_nei = indices[i_ptr]+ i_section*num_oscilators; f_tild += coupling*sinf( phases_old[i_nei] - phase_tild); }; f_tild = f_tild + omega; phases_new[index] = phase + dt*(f+f_tild)/2.; } }
1,231
#include <float.h> #include <math.h> #include <stdio.h> __global__ void calculateAreas(const int recs, const double w, const int offset, double *areas) { const int index = threadIdx.x + offset; if (index >= recs) return; const double x = index * w; double h = 1 - x * x; //Detect a 0 by accounting for roundoff error. h = h < DBL_EPSILON ? 0 : sqrt(h); areas[index] = w * h; } void calculateArea(const int recs, double *area) { double *areas = (double*) malloc(recs * sizeof(double)); if (areas == NULL) { fprintf(stderr, "malloc failed!\n"); return; } double *w_areas; cudaError_t err = cudaMalloc((void**) &w_areas, (recs * sizeof(double))); if (err != cudaSuccess) { fprintf(stderr, "cudaMalloc failed: %s\n", cudaGetErrorString(err)); return; } const int threadCount = 512, loops = ceil((double) recs / threadCount); const double width = 1.0 / recs; for (int c = 0; c < loops; ++c) { //kernel<<<blocks, threads>>> calculateAreas<<<1, threadCount>>>(recs, width, c * threadCount, w_areas); } err = cudaMemcpy(areas, w_areas, recs * sizeof(double), cudaMemcpyDeviceToHost); if (err != cudaSuccess) { fprintf(stderr, "cudaMemcpy failed: %s\n", cudaGetErrorString(err)); return; } *area = 0; for (int c = 0; c < recs; ++c) { *area += areas[c]; } *area *= 4; cudaFree(w_areas); free(areas); }
1,232
__global__ void subtract_and_square(float *dest, float *a, float *b, int n) { // const int index = threadIdx.x * (threadIdx.y + 1); // dest[index] = ( a[index] - b[index] ) * ( a[index] - b[index] ); int index = blockDim.x * blockIdx.x + threadIdx.x; if (index < n) dest[index] = ( a[index] - b[index] ) * ( a[index] - b[index] ); }
1,233
#include <cuda.h> template<typename T> __device__ __forceinline__ T ldg(const T* ptr) { #if __CUDA_ARCH__ >= 350 return __ldg(ptr); #else return *ptr; #endif } extern "C" __global__ void transpose_constY( int nx , int ny , int nz , float * in , float * out // XYZ -> ZYX ) { int kx = blockIdx.x*blockDim.x + threadIdx.x; int ky = blockIdx.y*blockDim.y + threadIdx.y; if (kx < nx && ky < ny) { // input is ordered by {X,Y,Z} // output is ordered by {Z,Y,X} // out[z*num_y*num_x + y*num_x + x] := in[x*num_y*num_z + y*num_z + z] //float * in_xy = &in[kx*ny*nz + ky*nz]; int k_0 = kx + ky*nx; int k_1 = kx*ny*nz + ky*nz; int i = 0; for (; i < nz; i++, k_0+=ny*nx, k_1++) { out[k_0] = ldg(&in[k_1]); } } }
1,234
#include <stdio.h> #include <stdlib.h> #include <iostream> #include <math.h> #include <cuda.h> #include <cuda_runtime.h> #include <sys/time.h> #define NPB_VERSION "3.3.1" using namespace std; #define min(x,y) (x) <= (y) ? (x) : (y) #define max(x,y) (x) >= (y) ? (x) : (y) // block sizes for CUDA kernels #define NORM_BLOCK 32 #define SOLVE_BLOCK 32 #define ERHS_BLOCK 32 // timer constants #define t_total 0 #define t_rhsx 1 #define t_rhsy 2 #define t_rhsz 3 #define t_rhs 4 #define t_xsolve 5 #define t_ysolve 6 #define t_zsolve 7 #define t_rdis1 8 #define t_rdis2 9 #define t_txinvr 10 #define t_pinvr 11 #define t_ninvr 12 #define t_tzetar 13 #define t_add 14 #define t_last 15 //--------------------------------------------------------------------- // diffusion coefficients //--------------------------------------------------------------------- #define dx1 0.75 #define dx2 0.75 #define dx3 0.75 #define dx4 0.75 #define dx5 0.75 #define dy1 0.75 #define dy2 0.75 #define dy3 0.75 #define dy4 0.75 #define dy5 0.75 #define dz1 1.0 #define dz2 1.0 #define dz3 1.0 #define dz4 1.0 #define dz5 1.0 //#define dxmax max(dx3,dx4) //#define dymax max(dy2,dy4) //#define dzmax max(dz2,dz3) #define dxmax dx3 #define dymax dy2 #define dzmax dz2 //--------------------------------------------------------------------- // fourth difference dissipation //--------------------------------------------------------------------- #define dssp (max(max(dx1,dy1),dz1)*.25) #define c4dssp (4.0*dssp) #define c5dssp (5.0*dssp) #define c1 1.4 #define c2 0.4 #define c3 0.1 #define c4 1.0 #define c5 1.4 #define c1c2 (c1*c2) #define c1c5 (c1*c5) #define c3c4 (c3*c4) #define c1345 (c1c5*c3c4) #define conz1 (1.0-c1c5) #define c2iv 2.5 #define con43 (4.0/3.0) #define con16 (1.0/6.0) // macros to linearize multidimensional array accesses #define fu(m,i,j,k) fu[(i)+nx*((j)+ny*((k)+nz*(m)))] #define forcing(m,i,j,k) forcing[(i)+nx*((j)+ny*((k)+nz*(m)))] #define rhs(m,i,j,k) rhs[m+(i)*5+(j)*5*nx+(k)*5*nx*ny] #define rho_i(i,j,k) rho_i[i+(j)*nx+(k)*nx*ny] #define us(i,j,k) us[i+(j)*nx+(k)*nx*ny] #define vs(i,j,k) vs[i+(j)*nx+(k)*nx*ny] #define ws(i,j,k) ws[i+(j)*nx+(k)*nx*ny] #define square(i,j,k) square[i+(j)*nx+(k)*nx*ny] #define qs(i,j,k) qs[i+(j)*nx+(k)*nx*ny] #define speed(i,j,k) speed[i+(j)*nx+(k)*nx*ny] static void inline HandleError( cudaError_t err, const char *file, int line ) { if (err != cudaSuccess) { printf( "%s in %s at line %d\n", cudaGetErrorString( err ), file, line ); exit( EXIT_FAILURE ); } } #define HANDLE_ERROR( err ) (HandleError( err, __FILE__, __LINE__ )) __constant__ double tx1, tx2, tx3, ty1, ty2, ty3, tz1, tz2, tz3; __constant__ double bt, dt, dtdssp; __constant__ double dnxm1, dnym1, dnzm1; __constant__ double dtx1, dttx2, dty1, dtty2, dtz1, dttz2, c2dttx1, c2dtty1, c2dttz1; __constant__ double comz1, comz4, comz5, comz6, c3c4tx3, c3c4ty3, c3c4tz3; __constant__ double xxcon1, xxcon2, xxcon3, xxcon4, xxcon5, dx1tx1, dx2tx1, dx3tx1, dx4tx1, dx5tx1; __constant__ double yycon1, yycon2, yycon3, yycon4, yycon5, dy1ty1, dy2ty1, dy3ty1, dy4ty1, dy5ty1; __constant__ double zzcon1, zzcon2, zzcon3, zzcon4, zzcon5, dz1tz1, dz2tz1, dz3tz1, dz4tz1, dz5tz1; __constant__ double ce[13][5]; //--------------------------------------------------------------------- // exact_rhs computation //--------------------------------------------------------------------- __device__ static void exact_solution_kernel (const double xi, const double eta, const double zta, double *dtemp) { for (int m = 0; m < 5; m++) dtemp[m] = ce[0][m] + xi*(ce[1][m] + xi*(ce[4][m] + xi*(ce[7][m] + xi*ce[10][m]))) + eta*(ce[2][m] + eta*(ce[5][m] + eta*(ce[8][m] + eta*ce[11][m])))+ zta*(ce[3][m] + zta*(ce[6][m] + zta*(ce[9][m] + zta*ce[12][m]))); } __global__ static void exact_rhs_kernel_init (double *forcing, const int nx, const int ny, const int nz) { int i, j, k, m; k = blockIdx.y; j = blockIdx.x; i = threadIdx.x; for (m = 0; m < 5; m++) forcing(m,i,j,k) = (double)0.0; } __global__ static void exact_rhs_kernel_x (double *forcing, const int nx, const int ny, const int nz) { int i, j, k, m; double xi, eta, zta, dtemp[5], dtpp; double ue[5][5], buf[3][5], cuf[3], q[3]; k = blockIdx.x*blockDim.x+threadIdx.x+1; j = blockIdx.y*blockDim.y+threadIdx.y+1; if (k >= nz-1 || j >= ny-1) return; zta = (double)k * dnzm1; eta = (double)j * dnym1; //--------------------------------------------------------------------- // xi-direction flux differences //--------------------------------------------------------------------- for (i = 0; i < 3; i++) { xi = (double)i * dnxm1; exact_solution_kernel(xi, eta, zta, dtemp); for (m = 0; m < 5; m++) ue[i+1][m] = dtemp[m]; dtpp = 1.0/dtemp[0]; for (m = 1; m < 5; m++) buf[i][m] = dtpp*dtemp[m]; cuf[i] = buf[i][1] * buf[i][1]; buf[i][0] = cuf[i] + buf[i][2] * buf[i][2] + buf[i][3] * buf[i][3]; q[i] = 0.5 * (buf[i][1]*ue[i+1][1] + buf[i][2]*ue[i+1][2] + buf[i][3]*ue[i+1][3]); } for (i = 1; i < nx-1; i++) { if (i+2 < nx) { xi = (double)(i+2) * dnxm1; exact_solution_kernel(xi, eta, zta, dtemp); for (m = 0; m < 5; m++) ue[4][m] = dtemp[m]; } dtemp[0] = 0.0 - tx2*(ue[3][1]-ue[1][1])+ dx1tx1*(ue[3][0]-2.0*ue[2][0]+ue[1][0]); dtemp[1] = 0.0 - tx2*((ue[3][1]*buf[2][1]+c2*(ue[3][4]-q[2]))-(ue[1][1]*buf[0][1]+c2*(ue[1][4]-q[0])))+xxcon1*(buf[2][1]-2.0*buf[1][1]+buf[0][1])+dx2tx1*(ue[3][1]-2.0*ue[2][1]+ue[1][1]); dtemp[2] = 0.0 - tx2*(ue[3][2]*buf[2][1]-ue[1][2]*buf[0][1])+xxcon2*(buf[2][2]-2.0*buf[1][2]+buf[0][2])+dx3tx1*(ue[3][2]-2.0*ue[2][2]+ue[1][2]); dtemp[3] = 0.0 - tx2*(ue[3][3]*buf[2][1]-ue[1][3]*buf[0][1])+xxcon2*(buf[2][3]-2.0*buf[1][3]+buf[0][3])+dx4tx1*(ue[3][3]-2.0*ue[2][3]+ue[1][3]); dtemp[4] = 0.0 - tx2*(buf[2][1]*(c1*ue[3][4]-c2*q[2])-buf[0][1]*(c1*ue[1][4]-c2*q[0]))+0.5*xxcon3*(buf[2][0]-2.0*buf[1][0]+buf[0][0])+xxcon4*(cuf[2]-2.0*cuf[1]+cuf[0])+ xxcon5*(buf[2][4]-2.0*buf[1][4]+buf[0][4])+dx5tx1*(ue[3][4]-2.0*ue[2][4]+ ue[1][4]); //--------------------------------------------------------------------- // Fourth-order dissipation //--------------------------------------------------------------------- if (i == 1) { for (m = 0; m < 5; m++) forcing(m,i,j,k) = dtemp[m] - dssp*(5.0*ue[2][m] - 4.0*ue[3][m] + ue[4][m]); } else if (i == 2) { for (m = 0; m < 5; m++) forcing(m,i,j,k) = dtemp[m] - dssp*(-4.0*ue[1][m] + 6.0*ue[2][m] - 4.0*ue[3][m] + ue[4][m]); } else if (i >= 3 && i < nx-3) { for (m = 0; m < 5; m++) forcing(m,i,j,k) = dtemp[m] - dssp*(ue[0][m] - 4.0*ue[1][m]+6.0*ue[2][m] - 4.0*ue[3][m] + ue[4][m]); } else if (i == nx-3) { for (m = 0; m < 5; m++) forcing(m,i,j,k) = dtemp[m] - dssp*(ue[0][m] - 4.0*ue[1][m] +6.0*ue[2][m] - 4.0*ue[3][m]); } else if (i == nx-2) { for (m = 0; m < 5; m++) forcing(m,i,j,k) = dtemp[m] - dssp*(ue[0][m] - 4.0*ue[1][m] + 5.0*ue[2][m]); } for (m = 0; m < 5; m++) { ue[0][m] = ue[1][m]; ue[1][m] = ue[2][m]; ue[2][m] = ue[3][m]; ue[3][m] = ue[4][m]; buf[0][m] = buf[1][m]; buf[1][m] = buf[2][m]; } cuf[0] = cuf[1]; cuf[1] = cuf[2]; q[0] = q[1]; q[1] = q[2]; if (i < nx-2) { dtpp = 1.0/ue[3][0]; for (m = 1; m < 5; m++) buf[2][m] = dtpp*ue[3][m]; cuf[2] = buf[2][1] * buf[2][1]; buf[2][0] = cuf[2] + buf[2][2] * buf[2][2] + buf[2][3] * buf[2][3]; q[2] = 0.5 * (buf[2][1]*ue[3][1] + buf[2][2]*ue[3][2] + buf[2][3]*ue[3][3]); } } } __global__ static void exact_rhs_kernel_y (double *forcing, const int nx, const int ny, const int nz) { int i, j, k, m; double xi, eta, zta, dtemp[5], dtpp; double ue[5][5], buf[3][5], cuf[3], q[3]; k = blockIdx.x*blockDim.x+threadIdx.x+1; i = blockIdx.y*blockDim.y+threadIdx.y+1; if (k >= nz-1 || i >= nx-1) return; zta = (double)k * dnzm1; xi = (double)i * dnxm1; //--------------------------------------------------------------------- // eta-direction flux differences //--------------------------------------------------------------------- for (j = 0; j < 3; j++) { eta = (double)j * dnym1; exact_solution_kernel(xi, eta, zta, dtemp); for (m = 0; m < 5; m++) ue[j+1][m] = dtemp[m]; dtpp = 1.0/dtemp[0]; for (m = 1; m < 5; m++) buf[j][m] = dtpp * dtemp[m]; cuf[j] = buf[j][2] * buf[j][2]; buf[j][0] = cuf[j] + buf[j][1] * buf[j][1] + buf[j][3] * buf[j][3]; q[j] = 0.5*(buf[j][1]*ue[j+1][1] + buf[j][2]*ue[j+1][2] + buf[j][3]*ue[j+1][3]); } for (j = 1; j < ny-1; j++) { if (j+2 < ny) { eta = (double)(j+2) * dnym1; exact_solution_kernel(xi, eta, zta, dtemp); for (m = 0; m < 5; m++) ue[4][m] = dtemp[m]; } dtemp[0] = forcing(0,i,j,k) - ty2*(ue[3][2]-ue[1][2])+ dy1ty1*(ue[3][0]-2.0*ue[2][0]+ue[1][0]); dtemp[1] = forcing(1,i,j,k) - ty2*(ue[3][1]*buf[2][2]-ue[1][1]*buf[0][2])+yycon2*(buf[2][1]-2.0*buf[1][1]+buf[0][1])+dy2ty1*(ue[3][1]-2.0*ue[2][1]+ ue[1][1]); dtemp[2] = forcing(2,i,j,k) - ty2*((ue[3][2]*buf[2][2]+c2*(ue[3][4]-q[2]))-(ue[1][2]*buf[0][2]+c2*(ue[1][4]-q[0])))+yycon1*(buf[2][2]-2.0*buf[1][2]+buf[0][2])+dy3ty1*( ue[3][2]-2.0*ue[2][2] +ue[1][2]); dtemp[3] = forcing(3,i,j,k) - ty2*(ue[3][3]*buf[2][2]-ue[1][3]*buf[0][2])+yycon2*(buf[2][3]-2.0*buf[1][3]+buf[0][3])+dy4ty1*( ue[3][3]-2.0*ue[2][3]+ ue[1][3]); dtemp[4] = forcing(4,i,j,k) - ty2*(buf[2][2]*(c1*ue[3][4]-c2*q[2])-buf[0][2]*(c1*ue[1][4]-c2*q[0]))+0.5*yycon3*(buf[2][0]-2.0*buf[1][0]+buf[0][0])+yycon4*(cuf[2]-2.0*cuf[1]+cuf[0])+ yycon5*(buf[2][4]-2.0*buf[1][4]+buf[0][4])+dy5ty1*(ue[3][4]-2.0*ue[2][4]+ue[1][4]); //--------------------------------------------------------------------- // Fourth-order dissipation //--------------------------------------------------------------------- if (j == 1) { for (m = 0; m < 5; m++) forcing(m,i,j,k) = dtemp[m] - dssp * (5.0*ue[2][m] - 4.0*ue[3][m] +ue[4][m]); } else if (j == 2) { for (m = 0; m < 5; m++) forcing(m,i,j,k) = dtemp[m] - dssp * (-4.0*ue[1][m] + 6.0*ue[2][m] - 4.0*ue[3][m] + ue[4][m]); } else if (j >= 3 && j < ny-3) { for (m = 0; m < 5; m++) forcing(m,i,j,k) = dtemp[m] - dssp*(ue[0][m] - 4.0*ue[1][m] + 6.0*ue[2][m] - 4.0*ue[3][m] + ue[4][m]); } else if (j == ny-3) { for (m = 0; m < 5; m++) forcing(m,i,j,k) = dtemp[m] - dssp * (ue[0][m] - 4.0*ue[1][m] + 6.0*ue[2][m] - 4.0*ue[3][m]); } else if (j == ny-2) { for (m = 0; m < 5; m++) forcing(m,i,j,k) = dtemp[m] - dssp * (ue[0][m] - 4.0*ue[1][m] + 5.0*ue[2][m]); } for (m = 0; m < 5; m++) { ue[0][m] = ue[1][m]; ue[1][m] = ue[2][m]; ue[2][m] = ue[3][m]; ue[3][m] = ue[4][m]; buf[0][m] = buf[1][m]; buf[1][m] = buf[2][m]; } cuf[0] = cuf[1]; cuf[1] = cuf[2]; q[0] = q[1]; q[1] = q[2]; if (j < ny-2) { dtpp = 1.0/ue[3][0]; for (m = 1; m < 5; m++) buf[2][m] = dtpp * ue[3][m]; cuf[2] = buf[2][2] * buf[2][2]; buf[2][0] = cuf[2] + buf[2][1] * buf[2][1] + buf[2][3] * buf[2][3]; q[2] = 0.5*(buf[2][1]*ue[3][1] + buf[2][2]*ue[3][2] + buf[2][3]*ue[3][3]); } } } __global__ static void exact_rhs_kernel_z (double *forcing, const int nx, const int ny, const int nz) { int i, j, k, m; double xi, eta, zta, dtpp, dtemp[5]; double ue[5][5], buf[3][5], cuf[3], q[3]; j = blockIdx.x*blockDim.x+threadIdx.x+1; i = blockIdx.y*blockDim.y+threadIdx.y+1; if (j >= ny-1 || i >= nx-1) return; eta = (double)j * dnym1; xi = (double)i * dnxm1; //--------------------------------------------------------------------- // zeta-direction flux differences //--------------------------------------------------------------------- for (k = 0; k < 3; k++) { zta = (double)k * dnzm1; exact_solution_kernel(xi, eta, zta, dtemp); for (m = 0; m < 5; m++) ue[k+1][m] = dtemp[m]; dtpp = 1.0/dtemp[0]; for (m = 1; m < 5; m++) buf[k][m] = dtpp * dtemp[m]; cuf[k] = buf[k][3] * buf[k][3]; buf[k][0] = cuf[k] + buf[k][1] * buf[k][1] + buf[k][2] * buf[k][2]; q[k] = 0.5*(buf[k][1]*ue[k+1][1] + buf[k][2]*ue[k+1][2] + buf[k][3]*ue[k+1][3]); } for (k = 1; k < nz-1; k++) { if (k+2 < nz) { zta = (double)(k+2) * dnzm1; exact_solution_kernel(xi, eta, zta, dtemp); for (m = 0; m < 5; m++) ue[4][m] = dtemp[m]; } dtemp[0] = forcing(0,i,j,k) - tz2*(ue[3][3]-ue[1][3])+dz1tz1*(ue[3][0]-2.0*ue[2][0]+ue[1][0]); dtemp[1] = forcing(1,i,j,k) - tz2*(ue[3][1]*buf[2][3]-ue[1][1]*buf[0][3])+zzcon2*(buf[2][1]-2.0*buf[1][1]+buf[0][1])+dz2tz1*(ue[3][1]-2.0*ue[2][1]+ue[1][1]); dtemp[2] = forcing(2,i,j,k) - tz2*(ue[3][2]*buf[2][3]-ue[1][2]*buf[0][3])+zzcon2*(buf[2][2]-2.0*buf[1][2]+buf[0][2])+dz3tz1*(ue[3][2]-2.0*ue[2][2]+ue[1][2]); dtemp[3] = forcing(3,i,j,k) - tz2*((ue[3][3]*buf[2][3]+c2*(ue[3][4]-q[2]))-(ue[1][3]*buf[0][3]+c2*(ue[1][4]-q[0])))+zzcon1*(buf[2][3]-2.0*buf[1][3]+buf[0][3])+dz4tz1*(ue[3][3]-2.0*ue[2][3] +ue[1][3]); dtemp[4] = forcing(4,i,j,k) - tz2*(buf[2][3]*(c1*ue[3][4]-c2*q[2])-buf[0][3]*(c1*ue[1][4]-c2*q[0]))+0.5*zzcon3*(buf[2][0]-2.0*buf[1][0]+buf[0][0])+ zzcon4*(cuf[2]-2.0*cuf[1]+cuf[0])+zzcon5*(buf[2][4]-2.0*buf[1][4]+buf[0][4])+dz5tz1*(ue[3][4]-2.0*ue[2][4]+ue[1][4]); //--------------------------------------------------------------------- // Fourth-order dissipation //--------------------------------------------------------------------- if (k == 1) { for (m = 0; m < 5; m++) dtemp[m] = dtemp[m] - dssp*(5.0*ue[2][m]-4.0*ue[3][m]+ue[4][m]); } else if (k == 2) { for (m = 0; m < 5; m++) dtemp[m] = dtemp[m] - dssp*(-4.0*ue[1][m]+6.0*ue[2][m]-4.0*ue[3][m]+ue[4][m]); } else if (k >= 3 && k < nz-3) { for (m = 0; m < 5; m++) dtemp[m] = dtemp[m] - dssp*(ue[0][m]-4.0*ue[1][m]+6.0*ue[2][m]-4.0*ue[3][m]+ue[4][m]); } else if (k == nz-3) { for (m = 0; m < 5; m++) dtemp[m] = dtemp[m] - dssp*(ue[0][m]-4.0*ue[1][m] + 6.0*ue[2][m] - 4.0*ue[3][m]); } else if (k == nz-2) { for (m = 0; m < 5; m++) dtemp[m] = dtemp[m] - dssp*(ue[0][m]-4.0*ue[1][m]+5.0*ue[2][m]); } //--------------------------------------------------------------------- // now change the sign of the forcing function, //--------------------------------------------------------------------- for (m = 0; m < 5; m++) forcing(m,i,j,k) = -1.0 * dtemp[m]; for (m = 0; m < 5; m++) { ue[0][m] = ue[1][m]; ue[1][m] = ue[2][m]; ue[2][m] = ue[3][m]; ue[3][m] = ue[4][m]; buf[0][m] = buf[1][m]; buf[1][m] = buf[2][m]; } cuf[0] = cuf[1]; cuf[1] = cuf[2]; q[0] = q[1]; q[1] = q[2]; if (k < nz-2) { dtpp = 1.0/ue[3][0]; for (m = 1; m < 5; m++) buf[2][m] = dtpp * ue[3][m]; cuf[2] = buf[2][3] * buf[2][3]; buf[2][0] = cuf[2] + buf[2][1] * buf[2][1] + buf[2][2] * buf[2][2]; q[2] = 0.5*(buf[2][1]*ue[3][1] + buf[2][2]*ue[3][2] + buf[2][3]*ue[3][3]); } } } void exact_rhs (double* forcing, int nx, int ny, int nz) { dim3 gridinit(ny,nz); exact_rhs_kernel_init<<<gridinit,nx>>>(forcing, nx, ny, nz); int yblock = min(ERHS_BLOCK,ny); int ygrid = (ny+yblock-1)/yblock; int zblock_y = min(ERHS_BLOCK/yblock,nz); int zgrid_y = (nz+zblock_y-1)/zblock_y; dim3 grid_x(zgrid_y,ygrid), block_x(zblock_y,yblock); exact_rhs_kernel_x<<<grid_x,block_x>>>(forcing, nx, ny, nz); int xblock = min(ERHS_BLOCK,nx); int xgrid = (nx+xblock-1)/xblock; int zblock_x = min(ERHS_BLOCK/xblock,nz); int zgrid_x = (nz+zblock_x-1)/zblock_x; dim3 grid_y(zgrid_x,xgrid), block_y(zblock_x,xblock); exact_rhs_kernel_y<<<grid_y,block_y>>>(forcing, nx, ny, nz); int yblock_x = min(ERHS_BLOCK/xblock,ny); int ygrid_x = (ny+yblock_x-1)/yblock_x; dim3 grid_z(ygrid_x,xgrid), block_z(yblock_x,xblock); exact_rhs_kernel_z<<<grid_z,block_z>>>(forcing, nx, ny, nz); } //--------------------------------------------------------------------- // initialize_kernel //--------------------------------------------------------------------- __global__ static void initialize_kernel (double *fu, const int nx, const int ny, const int nz) { int i, j, k, m; double xi, eta, zta, temp[5]; double Pface11[5], Pface12[5], Pface21[5], Pface22[5], Pface31[5], Pface32[5]; double zero, one; k = blockIdx.x; j = blockIdx.y; i = threadIdx.x; //--------------------------------------------------------------------- // to compute the whole thing with a simple loop. Make sure those // values are nonzero by initializing the whole thing here. //--------------------------------------------------------------------- fu(0,i,j,k) = (double)1.0; fu(1,i,j,k) = (double)0.0; fu(2,i,j,k) = (double)0.0; fu(3,i,j,k) = (double)0.0; fu(4,i,j,k) = (double)1.0; zero = (double)0.0; one = (double)1.0; //--------------------------------------------------------------------- // first store the "interpolated" values everywhere on the zone //--------------------------------------------------------------------- zta = (double)k * dnzm1; eta = (double)j * dnym1; xi = (double)i * dnxm1; exact_solution_kernel (zero, eta, zta, Pface11); exact_solution_kernel (one, eta, zta, Pface12); exact_solution_kernel (xi, zero, zta, Pface21); exact_solution_kernel (xi, one, zta, Pface22); exact_solution_kernel (xi, eta, zero, Pface31); exact_solution_kernel (xi, eta, one, Pface32); for (m = 0; m < 5; m++) { double Pxi = xi * Pface12[m] + (1.0-xi)*Pface11[m]; double Peta = eta * Pface22[m] + (1.0-eta)*Pface21[m]; double Pzeta = zta * Pface32[m] + (1.0-zta)*Pface31[m]; fu(m,i,j,k) = Pxi + Peta + Pzeta - Pxi*Peta - Pxi*Pzeta - Peta*Pzeta + Pxi*Peta*Pzeta; } //--------------------------------------------------------------------- // now store the exact values on the boundaries //--------------------------------------------------------------------- //--------------------------------------------------------------------- // west face //--------------------------------------------------------------------- xi = (double)0.0; if (i == 0) { zta = (double)k * dnzm1; eta = (double)j * dnym1; exact_solution_kernel (xi, eta, zta, temp); for (m = 0; m < 5; m++) fu(m,i,j,k) = temp[m]; } //--------------------------------------------------------------------- // east face //--------------------------------------------------------------------- xi = (double)1.0; if (i == nx-1) { zta = (double)k * dnzm1; eta = (double)j * dnym1; exact_solution_kernel (xi, eta, zta, temp); for (m = 0; m < 5; m++) fu(m,i,j,k) = temp[m]; } //--------------------------------------------------------------------- // south face //--------------------------------------------------------------------- eta = (double)0.0; if (j == 0) { zta = (double)k * dnzm1; xi = (double)i * dnxm1; exact_solution_kernel (xi,eta,zta,temp); for (m = 0; m < 5; m++) fu(m,i,j,k) = temp[m]; } //--------------------------------------------------------------------- // north face //--------------------------------------------------------------------- eta = (double)1.0; if (j == ny-1) { zta = (double)k * dnzm1; xi = (double)i * dnxm1; exact_solution_kernel (xi,eta,zta,temp); for (m = 0; m < 5; m++) fu(m,i,j,k) = temp[m]; } //--------------------------------------------------------------------- // bottom face //--------------------------------------------------------------------- zta = (double)0.0; if (k == 0) { eta = (double)j * dnym1; xi = (double)i * dnxm1; exact_solution_kernel (xi, eta, zta, temp); for (m = 0; m < 5; m++) fu(m,i,j,k) = temp[m]; } //--------------------------------------------------------------------- // top face //--------------------------------------------------------------------- zta = (double)1.0; if (k == nz-1) { eta = (double)j * dnym1; xi = (double)i * dnxm1; exact_solution_kernel (xi, eta, zta, temp); for (m = 0; m < 5; m++) fu(m,i,j,k) = temp[m]; } } //--------------------------------------------------------------------- // adi: compute_rhs //--------------------------------------------------------------------- __global__ static void compute_rhs_kernel_1 (double *rho_i, double *us, double *vs, double *ws, double *speed, double *qs, double *square, double *fu, const int nx, const int ny, const int nz) { int i, j, k; k = blockIdx.y; j = blockIdx.x; i = threadIdx.x; //--------------------------------------------------------------------- // compute the reciprocal of density, and the kinetic energy, // and the speed of sound. //--------------------------------------------------------------------- double rho_nv = 1.0/fu(0,i,j,k); double square_ijk; rho_i(i,j,k) = rho_nv; us(i,j,k) = fu(1,i,j,k) * rho_nv; vs(i,j,k) = fu(2,i,j,k) * rho_nv; ws(i,j,k) = fu(3,i,j,k) * rho_nv; square_ijk = 0.5*(fu(1,i,j,k)*fu(1,i,j,k) + fu(2,i,j,k)*fu(2,i,j,k) + fu(3,i,j,k)*fu(3,i,j,k)) * rho_nv; square(i,j,k) = 0.5*(fu(1,i,j,k)*fu(1,i,j,k) + fu(2,i,j,k)*fu(2,i,j,k) + fu(3,i,j,k)*fu(3,i,j,k)) * rho_nv; qs(i,j,k) = square_ijk * rho_nv; //--------------------------------------------------------------------- // (don't need speed and ainx until the lhs computation) //--------------------------------------------------------------------- speed(i,j,k) = sqrt(c1c2*rho_nv*(fu(4,i,j,k) - square_ijk)); } __global__ static void compute_rhs_kernel_2 (double *rho_i, double *us, double *vs, double *ws, double *qs, double *square, double *rhs, double *forcing, double *fu, int nx, const int ny, const int nz) { int i, j, k, m; k = blockIdx.y; j = blockIdx.x; i = threadIdx.x; double rtmp[5]; //--------------------------------------------------------------------- // copy the exact forcing term to the right hand side; because // this forcing term is known, we can store it on the whole zone // including the boundary //--------------------------------------------------------------------- for (m = 0; m < 5; m++) rtmp[m] = forcing(m,i,j,k); //--------------------------------------------------------------------- // compute xi-direction fluxes //--------------------------------------------------------------------- if (k >= 1 && k < nz-1 && j >= 1 && j < ny-1 && i >= 1 && i < nx-1) { double uijk = us(i,j,k); double up1 = us(i+1,j,k); double um1 = us(i-1,j,k); rtmp[0] = rtmp[0] + dx1tx1*(fu(0,i+1,j,k) - 2.0*fu(0,i,j,k) + fu(0,i-1,j,k)) - tx2*(fu(1,i+1,j,k)-fu(1,i-1,j,k)); rtmp[1] = rtmp[1] + dx2tx1*(fu(1,i+1,j,k) - 2.0*fu(1,i,j,k) + fu(1,i-1,j,k)) + xxcon2*con43*(up1-2.0*uijk+um1) - tx2*(fu(1,i+1,j,k)*up1 - fu(1,i-1,j,k)*um1 + (fu(4,i+1,j,k)-square(i+1,j,k)-fu(4,i-1,j,k)+square(i-1,j,k))*c2); rtmp[2] = rtmp[2] + dx3tx1*(fu(2,i+1,j,k) - 2.0*fu(2,i,j,k) + fu(2,i-1,j,k)) + xxcon2*(vs(i+1,j,k)-2.0*vs(i,j,k)+vs(i-1,j,k)) - tx2*(fu(2,i+1,j,k)*up1 - fu(2,i-1,j,k)*um1); rtmp[3] = rtmp[3] + dx4tx1*(fu(3,i+1,j,k) - 2.0*fu(3,i,j,k) + fu(3,i-1,j,k)) + xxcon2*(ws(i+1,j,k)-2.0*ws(i,j,k)+ws(i-1,j,k)) - tx2*(fu(3,i+1,j,k)*up1 - fu(3,i-1,j,k)*um1); rtmp[4] = rtmp[4] + dx5tx1*(fu(4,i+1,j,k) - 2.0*fu(4,i,j,k) + fu(4,i-1,j,k)) + xxcon3*(qs(i+1,j,k)-2.0*qs(i,j,k)+qs(i-1,j,k))+ xxcon4*(up1*up1-2.0*uijk*uijk+um1*um1) + xxcon5*(fu(4,i+1,j,k)*rho_i(i+1,j,k) - 2.0*fu(4,i,j,k)*rho_i(i,j,k) + fu(4,i-1,j,k)*rho_i(i-1,j,k)) - tx2*((c1*fu(4,i+1,j,k) - c2*square(i+1,j,k))*up1 - (c1*fu(4,i-1,j,k) - c2*square(i-1,j,k))*um1 ); //--------------------------------------------------------------------- // add fourth order xi-direction dissipation //--------------------------------------------------------------------- if (i == 1) { for (m = 0; m < 5; m++) rtmp[m] = rtmp[m] - dssp * (5.0*fu(m,i,j,k)-4.0*fu(m,i+1,j,k)+fu(m,i+2,j,k)); } else if (i == 2) { for (m = 0; m < 5; m++) rtmp[m] = rtmp[m] - dssp * (-4.0*fu(m,i-1,j,k)+6.0*fu(m,i,j,k)-4.0*fu(m,i+1,j,k)+fu(m,i+2,j,k)); } else if (i >= 3 && i < nx-3) { for (m = 0; m < 5; m++) rtmp[m] = rtmp[m] - dssp * ( fu(m,i-2,j,k)-4.0*fu(m,i-1,j,k)+6.0*fu(m,i,j,k)-4.0*fu(m,i+1,j,k)+fu(m,i+2,j,k)); } else if (i == nx-3) { for (m = 0; m < 5; m++) rtmp[m] = rtmp[m] - dssp * (fu(m,i-2,j,k)-4.0*fu(m,i-1,j,k)+6.0*fu(m,i,j,k)-4.0*fu(m,i+1,j,k) ); } else if (i == nx-2) { for (m = 0; m < 5; m++) rtmp[m] = rtmp[m] - dssp * (fu(m,i-2,j,k)-4.0*fu(m,i-1,j,k) + 5.0*fu(m,i,j,k)); } //--------------------------------------------------------------------- // compute eta-direction fluxes //--------------------------------------------------------------------- double vijk = vs(i,j,k); double vp1 = vs(i,j+1,k); double vm1 = vs(i,j-1,k); rtmp[0] = rtmp[0] + dy1ty1*(fu(0,i,j+1,k) - 2.0*fu(0,i,j,k) + fu(0,i,j-1,k)) - ty2*(fu(2,i,j+1,k)-fu(2,i,j-1,k)); rtmp[1] = rtmp[1] + dy2ty1*(fu(1,i,j+1,k) - 2.0*fu(1,i,j,k) + fu(1,i,j-1,k)) + yycon2*(us(i,j+1,k)-2.0*us(i,j,k)+us(i,j-1,k)) - ty2*(fu(1,i,j+1,k)*vp1-fu(1,i,j-1,k)*vm1); rtmp[2] = rtmp[2] + dy3ty1*(fu(2,i,j+1,k) - 2.0*fu(2,i,j,k) + fu(2,i,j-1,k)) + yycon2*con43*(vp1-2.0*vijk+vm1) - ty2*(fu(2,i,j+1,k)*vp1-fu(2,i,j-1,k)*vm1+(fu(4,i,j+1,k)-square(i,j+1,k)-fu(4,i,j-1,k)+square(i,j-1,k))*c2); rtmp[3] = rtmp[3] + dy4ty1*(fu(3,i,j+1,k) - 2.0*fu(3,i,j,k) + fu(3,i,j-1,k)) + yycon2*(ws(i,j+1,k)-2.0*ws(i,j,k)+ws(i,j-1,k))-ty2*(fu(3,i,j+1,k)*vp1-fu(3,i,j-1,k)*vm1); rtmp[4] = rtmp[4] + dy5ty1*(fu(4,i,j+1,k) - 2.0*fu(4,i,j,k) + fu(4,i,j-1,k)) + yycon3*(qs(i,j+1,k)-2.0*qs(i,j,k)+qs(i,j-1,k)) + yycon4*(vp1*vp1-2.0*vijk*vijk+vm1*vm1) + yycon5*(fu(4,i,j+1,k)*rho_i(i,j+1,k)-2.0*fu(4,i,j,k)*rho_i(i,j,k)+fu(4,i,j-1,k)*rho_i(i,j-1,k)) - ty2*((c1*fu(4,i,j+1,k)-c2*square(i,j+1,k))*vp1 - (c1*fu(4,i,j-1,k)-c2*square(i,j-1,k))*vm1); //--------------------------------------------------------------------- // add fourth order eta-direction dissipation //--------------------------------------------------------------------- if (j == 1) { for (m = 0; m < 5; m++) rtmp[m] = rtmp[m] - dssp*(5.0*fu(m,i,j,k)-4.0*fu(m,i,j+1,k)+fu(m,i,j+2,k)); } else if (j == 2) { for (m = 0; m < 5; m++) rtmp[m] = rtmp[m] - dssp*(-4.0*fu(m,i,j-1,k)+6.0*fu(m,i,j,k)-4.0*fu(m,i,j+1,k)+fu(m,i,j+2,k)); } else if (j >= 3 && j < ny-3) { for (m = 0; m < 5; m++) rtmp[m] = rtmp[m] - dssp*(fu(m,i,j-2,k)-4.0*fu(m,i,j-1,k)+6.0*fu(m,i,j,k)-4.0*fu(m,i,j+1,k)+fu(m,i,j+2,k)); } else if (j == ny-3) { for (m = 0; m < 5; m++) rtmp[m] = rtmp[m] - dssp*(fu(m,i,j-2,k)-4.0*fu(m,i,j-1,k)+6.0*fu(m,i,j,k)-4.0*fu(m,i,j+1,k)); } else if (j == ny-2) { for (m = 0; m < 5; m++) rtmp[m] = rtmp[m] - dssp*(fu(m,i,j-2,k)-4.0*fu(m,i,j-1,k)+5.0*fu(m,i,j,k)); } //--------------------------------------------------------------------- // compute zeta-direction fluxes //--------------------------------------------------------------------- double wijk = ws(i,j,k); double wp1 = ws(i,j,k+1); double wm1 = ws(i,j,k-1); rtmp[0] = rtmp[0] + dz1tz1*(fu(0,i,j,k+1)-2.0*fu(0,i,j,k)+fu(0,i,j,k-1)) - tz2*(fu(3,i,j,k+1)-fu(3,i,j,k-1)); rtmp[1] = rtmp[1] + dz2tz1*(fu(1,i,j,k+1)-2.0*fu(1,i,j,k)+fu(1,i,j,k-1)) + zzcon2*(us(i,j,k+1)-2.0*us(i,j,k)+us(i,j,k-1)) - tz2*(fu(1,i,j,k+1)*wp1-fu(1,i,j,k-1)*wm1); rtmp[2] = rtmp[2] + dz3tz1*(fu(2,i,j,k+1)-2.0*fu(2,i,j,k)+fu(2,i,j,k-1)) + zzcon2*(vs(i,j,k+1)-2.0*vs(i,j,k)+vs(i,j,k-1)) - tz2*(fu(2,i,j,k+1)*wp1-fu(2,i,j,k-1)*wm1); rtmp[3] = rtmp[3] + dz4tz1*(fu(3,i,j,k+1)-2.0*fu(3,i,j,k)+fu(3,i,j,k-1)) + zzcon2*con43*(wp1-2.0*wijk+wm1) - tz2*(fu(3,i,j,k+1)*wp1-fu(3,i,j,k-1)*wm1+(fu(4,i,j,k+1)-square(i,j,k+1)-fu(4,i,j,k-1)+square(i,j,k-1))*c2); rtmp[4] = rtmp[4] + dz5tz1*(fu(4,i,j,k+1)-2.0*fu(4,i,j,k)+fu(4,i,j,k-1)) + zzcon3*(qs(i,j,k+1)-2.0*qs(i,j,k)+qs(i,j,k-1)) + zzcon4*(wp1*wp1-2.0*wijk*wijk+wm1*wm1) + zzcon5*(fu(4,i,j,k+1)*rho_i(i,j,k+1)-2.0*fu(4,i,j,k)*rho_i(i,j,k)+fu(4,i,j,k-1)*rho_i(i,j,k-1)) - tz2*((c1*fu(4,i,j,k+1)-c2*square(i,j,k+1))*wp1-(c1*fu(4,i,j,k-1)-c2*square(i,j,k-1))*wm1); //--------------------------------------------------------------------- // add fourth order zeta-direction dissipation //--------------------------------------------------------------------- if (k == 1) { for (m = 0; m < 5; m++) rtmp[m] = rtmp[m] - dssp*(5.0*fu(m,i,j,k)-4.0*fu(m,i,j,k+1)+fu(m,i,j,k+2)); } else if (k == 2) { for (m = 0; m < 5; m++) rtmp[m] = rtmp[m] - dssp*(-4.0*fu(m,i,j,k-1)+6.0*fu(m,i,j,k)-4.0*fu(m,i,j,k+1)+fu(m,i,j,k+2)); } else if (k >= 3 && k < nz-3) { for (m = 0; m < 5; m++) rtmp[m] = rtmp[m] - dssp*(fu(m,i,j,k-2)-4.0*fu(m,i,j,k-1)+6.0*fu(m,i,j,k)-4.0*fu(m,i,j,k+1)+fu(m,i,j,k+2)); } else if (k == nz-3) { for (m = 0; m < 5; m++) rtmp[m] = rtmp[m] - dssp*(fu(m,i,j,k-2)-4.0*fu(m,i,j,k-1)+6.0*fu(m,i,j,k)-4.0*fu(m,i,j,k+1)); } else if (k == nz-2) { for (m = 0; m < 5; m++) rtmp[m] = rtmp[m] - dssp*(fu(m,i,j,k-2)-4.0*fu(m,i,j,k-1)+5.0*fu(m,i,j,k)); } for (m = 0; m < 5; m++) rtmp[m] *= dt; } for (m = 0; m < 5; m++) rhs(m,i,j,k) = rtmp[m]; } //--------------------------------------------------------------------- // adi: txinvr //--------------------------------------------------------------------- __global__ static void txinvr_kernel ( double *rho_i, double *us, double *vs, double *ws, double *speed, double *qs, double *rhs, const int nx, const int ny, const int nz) { int i, j, k; k = blockIdx.y+1; j = blockIdx.x+1; i = threadIdx.x+1; double ru1 = rho_i(i,j,k); double uu = us(i,j,k); double vv = vs(i,j,k); double ww = ws(i,j,k); double ap = speed(i,j,k); double ac2inv = 1.0/( ap*ap ); double r1 = rhs(0,i,j,k); double r2 = rhs(1,i,j,k); double r3 = rhs(2,i,j,k); double r4 = rhs(3,i,j,k); double r5 = rhs(4,i,j,k); double t1 = c2*ac2inv*(qs(i,j,k)*r1 - uu*r2 - vv*r3 - ww*r4 + r5); double t2 = bt * ru1 * ( uu * r1 - r2 ); double t3 = ( bt * ru1 * ap ) * t1; rhs(0,i,j,k) = r1 - t1; rhs(1,i,j,k) = -ru1*(ww*r1-r4); rhs(2,i,j,k) = ru1*(vv*r1-r3); rhs(3,i,j,k) = -t2+t3; rhs(4,i,j,k) = t2+t3; } //--------------------------------------------------------------------- // adi: x_solve //--------------------------------------------------------------------- #define lhs(m,i,j,k) lhs[(j-1)+(ny-2)*((k-1)+(nz-2)*((i)+nx*(m-3)))] #define lhsp(m,i,j,k) lhs[(j-1)+(ny-2)*((k-1)+(nz-2)*((i)+nx*(m+4)))] #define lhsm(m,i,j,k) lhs[(j-1)+(ny-2)*((k-1)+(nz-2)*((i)+nx*(m-3+2)))] #define rtmp(m,i,j,k) rstmp[(j)+ny*((k)+nz*((i)+nx*(m)))] __global__ static void x_solve_kernel (double *rho_i, double *us, double *speed, double *rhs, double *lhs, double *rstmp, const int nx, const int ny, const int nz) { int i, j, k, m; double rhon[3], cv[3], _ls[3][5], _lp[3][5], _rs[3][5], fac1; double zero; k = blockIdx.x*blockDim.x+threadIdx.x+1; j = blockIdx.y*blockDim.y+threadIdx.y+1; if (k >= nz-1 || j >= ny-1) return; //--------------------------------------------------------------------- // Computes the left hand side for the three x-factors //--------------------------------------------------------------------- //--------------------------------------------------------------------- // zap the whole left hand side for starters //--------------------------------------------------------------------- _ls[0][0] = (double)0.0; _ls[0][1] = (double)0.0; _ls[0][2] = (double)1.0; _ls[0][3] = (double)0.0; _ls[0][4] = (double)0.0; lhsp(0,0,j,k) = (double)0.0; lhsp(1,0,j,k) = (double)0.0; lhsp(2,0,j,k) = (double)1.0; lhsp(3,0,j,k) = (double)0.0; lhsp(4,0,j,k) = (double)0.0; zero = (double)0.0; //--------------------------------------------------------------------- // first fill the lhs for the u-eigenvalue //-------------------------------------------------------------------- for (i = 0; i < 3; i++) { fac1 = c3c4*rho_i(i,j,k); //rhon[i] = max(max(max(dx2+con43*fac1, dx5+c1c5*fac1), dxmax+fac1), zero+dx1); if (dx2+con43*fac1>dx5+c1c5*fac1) rhon[i] = dx2+con43*fac1; else rhon[i] = dx5+c1c5*fac1; if (rhon[i]<dxmax+fac1) rhon[i] = dxmax+fac1; if (rhon[i]<zero+dx1) rhon[i] = zero+dx1; cv[i] = us(i,j,k); } _ls[1][0] = (double)0.0; _ls[1][1] = - dttx2 * cv[0] - dtx1 * rhon[0]; _ls[1][2] = 1.0 + c2dttx1 * rhon[1]; _ls[1][3] = dttx2 * cv[2] - dtx1 * rhon[2]; _ls[1][4] = (double)0.0; _ls[1][2] += comz5; _ls[1][3] -= comz4; _ls[1][4] += comz1; for (m = 0; m < 5; m++) lhsp(m,1,j,k) = _ls[1][m]; rhon[0] = rhon[1]; rhon[1] = rhon[2]; cv[0] = cv[1]; cv[1] = cv[2]; for (m = 0; m < 3; m++) { _rs[0][m] = rhs(m,0,j,k); _rs[1][m] = rhs(m,1,j,k); } //--------------------------------------------------------------------- // perform the Thomas algorithm; first, FORWARD ELIMINATION //--------------------------------------------------------------------- for (i = 0; i < nx-2; i++) { //--------------------------------------------------------------------- // first fill the lhs for the u-eigenvalue //--------------------------------------------------------------------- if (i+2 == nx-1) { _ls[2][0] = (double)0.0; _ls[2][1] = (double)0.0; _ls[2][2] = (double)1.0; _ls[2][3] = (double)0.0; _ls[2][4] = (double)0.0; lhsp(0,i+2,j,k) = (double)0.0; lhsp(1,i+2,j,k) = (double)0.0; lhsp(2,i+2,j,k) = (double)1.0; lhsp(3,i+2,j,k) = (double)0.0; lhsp(4,i+2,j,k) = (double)0.0; } else { fac1 = c3c4*rho_i(i+3,j,k); //rhon[2] = max(max(max(dx2+con43*fac1, dx5+c1c5*fac1), dxmax+fac1), zero+dx1); if (dx2+con43*fac1>dx5+c1c5*fac1) rhon[2] = dx2+con43*fac1; else rhon[2] = dx5+c1c5*fac1; if (rhon[2]<dxmax+fac1) rhon[2] = dxmax+fac1; if (rhon[2]<zero+dx1) rhon[2] = zero+dx1; cv[2] = us(i+3,j,k); _ls[2][0] = (double)0.0; _ls[2][1] = - dttx2 * cv[0] - dtx1 * rhon[0]; _ls[2][2] = 1.0 + c2dttx1 * rhon[1]; _ls[2][3] = dttx2 * cv[2] - dtx1 * rhon[2]; _ls[2][4] = (double)0.0; //--------------------------------------------------------------------- // add fourth order dissipation //--------------------------------------------------------------------- if (i+2 == 2) { _ls[2][1] -= comz4; _ls[2][2] += comz6; _ls[2][3] -= comz4; _ls[2][4] += comz1; } else if (i+2 >= 3 && i+2 < nx-3) { _ls[2][0] += comz1; _ls[2][1] -= comz4; _ls[2][2] += comz6; _ls[2][3] -= comz4; _ls[2][4] += comz1; } else if (i+2 == nx-3) { _ls[2][0] += comz1; _ls[2][1] -= comz4; _ls[2][2] += comz6; _ls[2][3] -= comz4; } else if (i+2 == nx-2) { _ls[2][0] += comz1; _ls[2][1] -= comz4; _ls[2][2] += comz5; } //--------------------------------------------------------------------- // store computed lhs for later reuse //--------------------------------------------------------------------- for (m = 0; m < 5; m++) lhsp(m,i+2,j,k) = _ls[2][m]; rhon[0] = rhon[1]; rhon[1] = rhon[2]; cv[0] = cv[1]; cv[1] = cv[2]; } //--------------------------------------------------------------------- // load rhs values for current iteration //--------------------------------------------------------------------- for (m = 0; m < 3; m++) _rs[2][m] = rhs(m,i+2,j,k); //--------------------------------------------------------------------- // perform current iteration //--------------------------------------------------------------------- fac1 = 1.0/_ls[0][2]; _ls[0][3] *= fac1; _ls[0][4] *= fac1; for (m = 0; m < 3; m++) _rs[0][m] *= fac1; _ls[1][2] -= _ls[1][1] * _ls[0][3]; _ls[1][3] -= _ls[1][1] * _ls[0][4]; for (m = 0; m < 3; m++) _rs[1][m] -= _ls[1][1] * _rs[0][m]; _ls[2][1] -= _ls[2][0] * _ls[0][3]; _ls[2][2] -= _ls[2][0] * _ls[0][4]; for (m = 0; m < 3; m++) _rs[2][m] -= _ls[2][0] * _rs[0][m]; //--------------------------------------------------------------------- // store computed lhs and prepare data for next iteration // rhs is stored in a temp array such that write accesses are coalesced //--------------------------------------------------------------------- lhs(3,i,j,k) = _ls[0][3]; lhs(4,i,j,k) = _ls[0][4]; for (m = 0; m < 5; m++) { _ls[0][m] = _ls[1][m]; _ls[1][m] = _ls[2][m]; } for (m = 0; m < 3; m++) { rtmp(m,i,j,k) = _rs[0][m]; _rs[0][m] = _rs[1][m]; _rs[1][m] = _rs[2][m]; } } //--------------------------------------------------------------------- // The last two rows in this zone are a bit different, // since they do not have two more rows available for the // elimination of off-diagonal entries //--------------------------------------------------------------------- i = nx-2; fac1 = 1.0/_ls[0][2]; _ls[0][3] *= fac1; _ls[0][4] *= fac1; for (m = 0; m < 3; m++) _rs[0][m] *= fac1; _ls[1][2] -= _ls[1][1] * _ls[0][3]; _ls[1][3] -= _ls[1][1] * _ls[0][4]; for (m = 0; m < 3; m++) _rs[1][m] -= _ls[1][1] * _rs[0][m]; //--------------------------------------------------------------------- // scale the last row immediately //--------------------------------------------------------------------- fac1 = 1.0/_ls[1][2]; for (m = 0; m < 3; m++) _rs[1][m] *= fac1; lhs(3,nx-2,j,k) = _ls[0][3]; lhs(4,nx-2,j,k) = _ls[0][4]; //--------------------------------------------------------------------- // subsequently, fill the other factors u+c, u-c //--------------------------------------------------------------------- for (i = 0; i < 3; i++) cv[i] = speed(i,j,k); for (m = 0; m < 5; m++) { _ls[0][m] = lhsp(m,0,j,k); _lp[0][m] = lhsp(m,0,j,k); _ls[1][m] = lhsp(m,1,j,k); _lp[1][m] = lhsp(m,1,j,k); } _lp[1][1] -= dttx2 * cv[0]; _lp[1][3] += dttx2 * cv[2]; _ls[1][1] += dttx2 * cv[0]; _ls[1][3] -= dttx2 * cv[2]; cv[0] = cv[1]; cv[1] = cv[2]; _rs[0][3] = rhs(3,0,j,k); _rs[0][4] = rhs(4,0,j,k); _rs[1][3] = rhs(3,1,j,k); _rs[1][4] = rhs(4,1,j,k); //--------------------------------------------------------------------- // do the u+c and the u-c factors //--------------------------------------------------------------------- for (i = 0; i < nx-2; i++) { //--------------------------------------------------------------------- // first, fill the other factors u+c, u-c //--------------------------------------------------------------------- for (m = 0; m < 5; m++) { _ls[2][m] = lhsp(m,i+2,j,k); _lp[2][m] = lhsp(m,i+2,j,k); } _rs[2][3] = rhs(3,i+2,j,k); _rs[2][4] = rhs(4,i+2,j,k); if (i+2 < nx-1) { cv[2] = speed(i+3,j,k); _lp[2][1] -= dttx2 * cv[0]; _lp[2][3] += dttx2 * cv[2]; _ls[2][1] += dttx2 * cv[0]; _ls[2][3] -= dttx2 * cv[2]; cv[0] = cv[1]; cv[1] = cv[2]; } m = 3; fac1 = 1.0/_lp[0][2]; _lp[0][3] *= fac1; _lp[0][4] *= fac1; _rs[0][m] *= fac1; _lp[1][2] -= _lp[1][1]*_lp[0][3]; _lp[1][3] -= _lp[1][1]*_lp[0][4]; _rs[1][m] -= _lp[1][1]*_rs[0][m]; _lp[2][1] -= _lp[2][0]*_lp[0][3]; _lp[2][2] -= _lp[2][0]*_lp[0][4]; _rs[2][m] -= _lp[2][0]*_rs[0][m]; m = 4; fac1 = 1.0/_ls[0][2]; _ls[0][3] *= fac1; _ls[0][4] *= fac1; _rs[0][m] *= fac1; _ls[1][2] -= _ls[1][1]*_ls[0][3]; _ls[1][3] -= _ls[1][1]*_ls[0][4]; _rs[1][m] -= _ls[1][1]*_rs[0][m]; _ls[2][1] -= _ls[2][0]*_ls[0][3]; _ls[2][2] -= _ls[2][0]*_ls[0][4]; _rs[2][m] -= _ls[2][0]*_rs[0][m]; //--------------------------------------------------------------------- // store computed lhs and prepare data for next iteration // rhs is stored in a temp array such that write accesses are coalesced //--------------------------------------------------------------------- for (m = 3; m < 5; m++) { lhsp(m,i,j,k) = _lp[0][m]; lhsm(m,i,j,k) = _ls[0][m]; rtmp(m,i,j,k) = _rs[0][m]; _rs[0][m] = _rs[1][m]; _rs[1][m] = _rs[2][m]; } for (m = 0; m < 5; m++) { _lp[0][m] = _lp[1][m]; _lp[1][m] = _lp[2][m]; _ls[0][m] = _ls[1][m]; _ls[1][m] = _ls[2][m]; } } //--------------------------------------------------------------------- // And again the last two rows separately //--------------------------------------------------------------------- i = nx-2; m = 3; fac1 = 1.0/_lp[0][2]; _lp[0][3] *= fac1; _lp[0][4] *= fac1; _rs[0][m] *= fac1; _lp[1][2] -= _lp[1][1]*_lp[0][3]; _lp[1][3] -= _lp[1][1]*_lp[0][4]; _rs[1][m] -= _lp[1][1]*_rs[0][m]; m = 4; fac1 = 1.0/_ls[0][2]; _ls[0][3] *= fac1; _ls[0][4] *= fac1; _rs[0][m] *= fac1; _ls[1][2] -= _ls[1][1]*_ls[0][3]; _ls[1][3] -= _ls[1][1]*_ls[0][4]; _rs[1][m] -= _ls[1][1]*_rs[0][m]; //--------------------------------------------------------------------- // Scale the last row immediately //--------------------------------------------------------------------- _rs[1][3] /= _lp[1][2]; _rs[1][4] /= _ls[1][2]; //--------------------------------------------------------------------- // BACKSUBSTITUTION //--------------------------------------------------------------------- for (m = 0; m < 3; m++) _rs[0][m] -= lhs(3,nx-2,j,k)*_rs[1][m]; _rs[0][3] -= _lp[0][3]*_rs[1][3]; _rs[0][4] -= _ls[0][3]*_rs[1][4]; for (m = 0; m < 5; m++) { _rs[2][m] = _rs[1][m]; _rs[1][m] = _rs[0][m]; } for (i = nx-3; i >= 0; i--) { //--------------------------------------------------------------------- // The first three factors //--------------------------------------------------------------------- for (m = 0; m < 3; m++) _rs[0][m] = rtmp(m,i,j,k) - lhs(3,i,j,k)*_rs[1][m] - lhs(4,i,j,k)*_rs[2][m]; //--------------------------------------------------------------------- // And the remaining two //--------------------------------------------------------------------- _rs[0][3] = rtmp(3,i,j,k) - lhsp(3,i,j,k)*_rs[1][3] - lhsp(4,i,j,k)*_rs[2][3]; _rs[0][4] = rtmp(4,i,j,k) - lhsm(3,i,j,k)*_rs[1][4] - lhsm(4,i,j,k)*_rs[2][4]; if (i+2 < nx-1) { //--------------------------------------------------------------------- // Do the block-diagonal inversion //--------------------------------------------------------------------- double r1 = _rs[2][0]; double r2 = _rs[2][1]; double r3 = _rs[2][2]; double r4 = _rs[2][3]; double r5 = _rs[2][4]; double t1 = bt * r3; double t2 = 0.5 * (r4+r5); _rs[2][0] = -r2; _rs[2][1] = r1; _rs[2][2] = bt * ( r4 - r5 ); _rs[2][3] = -t1 + t2; _rs[2][4] = t1 + t2; } for (m = 0; m < 5; m++) { rhs(m,i+2,j,k) = _rs[2][m]; _rs[2][m] = _rs[1][m]; _rs[1][m] = _rs[0][m]; } } //--------------------------------------------------------------------- // Do the block-diagonal inversion //--------------------------------------------------------------------- double tf1 = bt * _rs[2][2]; double tf2 = 0.5 * (_rs[2][3]+_rs[2][4]); rhs(0,1,j,k) = -_rs[2][1]; rhs(1,1,j,k) = _rs[2][0]; rhs(2,1,j,k) = bt * ( _rs[2][3] - _rs[2][4] ); rhs(3,1,j,k) = -tf1 + tf2; rhs(4,1,j,k) = tf1 + tf2; for (m = 0; m < 5; m++) rhs(m,0,j,k) = _rs[1][m]; } #undef lhs #undef lhsp #undef lhsm #undef rtmp //--------------------------------------------------------------------- // adi: y_solve //--------------------------------------------------------------------- #define lhs(m,i,j,k) lhs[(i-1)+(nx-2)*((k-1)+(nz-2)*((j)+ny*(m-3)))] #define lhsp(m,i,j,k) lhs[(i-1)+(nx-2)*((k-1)+(nz-2)*((j)+ny*(m+4)))] #define lhsm(m,i,j,k) lhs[(i-1)+(nx-2)*((k-1)+(nz-2)*((j)+ny*(m-3+2)))] #define rtmp(m,i,j,k) rstmp[(i)+nx*((k)+nz*((j)+ny*(m)))] __global__ static void y_solve_kernel (double *rho_i, double *vs, double *speed, double *rhs, double *lhs, double *rstmp, const int nx, const int ny, const int nz) { int i, j, k, m; double rhoq[3], cv[3], _ls[3][5], _lp[3][5], _rs[3][5], fac1; double zero; k = blockIdx.x*blockDim.x+threadIdx.x+1; i = blockIdx.y*blockDim.y+threadIdx.y+1; if (k >= nz-1 || i >= nx-1) return; //--------------------------------------------------------------------- // Computes the left hand side for the three y-factors //--------------------------------------------------------------------- //--------------------------------------------------------------------- // zap the whole left hand side for starters //--------------------------------------------------------------------- _ls[0][0] = (double)0.0; _ls[0][1] = (double)0.0; _ls[0][2] = (double)1.0; _ls[0][3] = (double)0.0; _ls[0][4] = (double)0.0; lhsp(0,i,0,k) = (double)0.0; lhsp(1,i,0,k) = (double)0.0; lhsp(2,i,0,k) = (double)1.0; lhsp(3,i,0,k) = (double)0.0; lhsp(4,i,0,k) = (double)0.0; zero = (double)0.0; //--------------------------------------------------------------------- // first fill the lhs for the u-eigenvalue //--------------------------------------------------------------------- for (j = 0; j < 3; j++) { fac1 = c3c4*rho_i(i,j,k); //rhoq[j] = max(max(max(dy3+con43*fac1, dy5+c1c5*fac1), dymax+fac1), zero+dy1); if (dy3+con43*fac1>dy5+c1c5*fac1) rhoq[j] = dy3+con43*fac1; else rhoq[j] = dy5+c1c5*fac1; if (rhoq[j]<dymax+fac1) rhoq[j] = dymax+fac1; if (rhoq[j]<zero+dy1) rhoq[j] = zero+dy1; cv[j] = vs(i,j,k); } _ls[1][0] = (double)0.0; _ls[1][1] = -dtty2*cv[0]-dty1 * rhoq[0]; _ls[1][2] = 1.0 + c2dtty1 * rhoq[1]; _ls[1][3] = dtty2*cv[2]-dty1 * rhoq[2]; _ls[1][4] = (double)0.0; _ls[1][2] += comz5; _ls[1][3] -= comz4; _ls[1][4] += comz1; for (m = 0; m < 5; m++) lhsp(m,i,1,k) = _ls[1][m]; rhoq[0] = rhoq[1]; rhoq[1] = rhoq[2]; cv[0] = cv[1]; cv[1] = cv[2]; for (m = 0; m < 3; m++) { _rs[0][m] = rhs(m,i,0,k); _rs[1][m] = rhs(m,i,1,k); } //--------------------------------------------------------------------- // FORWARD ELIMINATION //--------------------------------------------------------------------- for (j = 0; j < ny-2; j++) { //--------------------------------------------------------------------- // first fill the lhs for the u-eigenvalue //--------------------------------------------------------------------- if (j+2 == ny-1) { _ls[2][0] = (double)0.0; _ls[2][1] = (double)0.0; _ls[2][2] = (double)1.0; _ls[2][3] = (double)0.0; _ls[2][4] = (double)0.0; lhsp(0,i,j+2,k) = (double)0.0; lhsp(1,i,j+2,k) = (double)0.0; lhsp(2,i,j+2,k) = (double)1.0; lhsp(3,i,j+2,k) = (double)0.0; lhsp(4,i,j+2,k) = (double)0.0; } else { fac1 = c3c4*rho_i(i,j+3,k); //rhoq[2] = max(max(max(dy3+con43*fac1, dy5+c1c5*fac1), dymax+fac1), zero+dy1); if (dy3+con43*fac1>dy5+c1c5*fac1) rhoq[2] = dy3+con43*fac1; else rhoq[2] = dy5+c1c5*fac1; if (rhoq[2]<dymax+fac1) rhoq[2] = dymax+fac1; if (rhoq[2]<zero+dy1) rhoq[2] = zero+dy1; cv[2] = vs(i,j+3,k); _ls[2][0] = (double)0.0; _ls[2][1] = -dtty2*cv[0]-dty1 * rhoq[0]; _ls[2][2] = 1.0 + c2dtty1 * rhoq[1]; _ls[2][3] = dtty2*cv[2]-dty1 * rhoq[2]; _ls[2][4] = (double)0.0; //--------------------------------------------------------------------- // add fourth order dissipation //--------------------------------------------------------------------- if (j+2 == 2) { _ls[2][1] -= comz4; _ls[2][2] += comz6; _ls[2][3] -= comz4; _ls[2][4] += comz1; } else if (j+2 >= 3 && j+2 < ny-3) { _ls[2][0] += comz1; _ls[2][1] -= comz4; _ls[2][2] += comz6; _ls[2][3] -= comz4; _ls[2][4] += comz1; } else if (j+2 == ny-3) { _ls[2][0] += comz1; _ls[2][1] -= comz4; _ls[2][2] += comz6; _ls[2][3] -= comz4; } else if (j+2 == ny-2) { _ls[2][0] += comz1; _ls[2][1] -= comz4; _ls[2][2] += comz5; } //--------------------------------------------------------------------- // store computed lhs for later reuse //--------------------------------------------------------------------- for (m = 0; m < 5; m++) lhsp(m,i,j+2,k) = _ls[2][m]; rhoq[0] = rhoq[1]; rhoq[1] = rhoq[2]; cv[0] = cv[1]; cv[1] = cv[2]; } //--------------------------------------------------------------------- // load rhs values for current iteration //--------------------------------------------------------------------- for (m = 0; m < 3; m++) _rs[2][m] = rhs(m,i,j+2,k); //--------------------------------------------------------------------- // perform current iteration //--------------------------------------------------------------------- fac1 = 1.0/_ls[0][2]; _ls[0][3] *= fac1; _ls[0][4] *= fac1; for (m = 0; m < 3; m++) _rs[0][m] *= fac1; _ls[1][2] -= _ls[1][1] * _ls[0][3]; _ls[1][3] -= _ls[1][1] * _ls[0][4]; for (m = 0; m < 3; m++) _rs[1][m] -= _ls[1][1] * _rs[0][m]; _ls[2][1] -= _ls[2][0] * _ls[0][3]; _ls[2][2] -= _ls[2][0] * _ls[0][4]; for (m = 0; m < 3; m++) _rs[2][m] -= _ls[2][0] * _rs[0][m]; //--------------------------------------------------------------------- // store computed lhs and prepare data for next iteration // rhs is stored in a temp array such that write accesses are coalesced //--------------------------------------------------------------------- lhs(3,i,j,k) = _ls[0][3]; lhs(4,i,j,k) = _ls[0][4]; for (m = 0; m < 5; m++) { _ls[0][m] = _ls[1][m]; _ls[1][m] = _ls[2][m]; } for (m = 0; m < 3; m++) { rtmp(m,i,j,k) = _rs[0][m]; _rs[0][m] = _rs[1][m]; _rs[1][m] = _rs[2][m]; } } //--------------------------------------------------------------------- // The last two rows in this zone are a bit different, // since they do not have two more rows available for the // elimination of off-diagonal entries //--------------------------------------------------------------------- j = ny-2; fac1 = 1.0/_ls[0][2]; _ls[0][3] *= fac1; _ls[0][4] *= fac1; for (m = 0; m < 3; m++) _rs[0][m] *= fac1; _ls[1][2] -= _ls[1][1] * _ls[0][3]; _ls[1][3] -= _ls[1][1] * _ls[0][4]; for (m = 0; m < 3; m++) _rs[1][m] -= _ls[1][1] * _rs[0][m]; //--------------------------------------------------------------------- // scale the last row immediately //--------------------------------------------------------------------- fac1 = 1.0/_ls[1][2]; for (m = 0; m < 3; m++) _rs[1][m] *= fac1; lhs(3,i,ny-2,k) = _ls[0][3]; lhs(4,i,ny-2,k) = _ls[0][4]; //--------------------------------------------------------------------- // do the u+c and the u-c factors //--------------------------------------------------------------------- for (j = 0; j < 3; j++) cv[j] = speed(i,j,k); for (m = 0; m < 5; m++) { _ls[0][m] = lhsp(m,i,0,k); _lp[0][m] = lhsp(m,i,0,k); _ls[1][m] = lhsp(m,i,1,k); _lp[1][m] = lhsp(m,i,1,k); } _lp[1][1] -= dtty2*cv[0]; _lp[1][3] += dtty2*cv[2]; _ls[1][1] += dtty2*cv[0]; _ls[1][3] -= dtty2*cv[2]; cv[0] = cv[1]; cv[1] = cv[2]; _rs[0][3] = rhs(3,i,0,k); _rs[0][4] = rhs(4,i,0,k); _rs[1][3] = rhs(3,i,1,k); _rs[1][4] = rhs(4,i,1,k); for (j = 0; j < ny-2; j++) { for (m = 0; m < 5; m++) { _ls[2][m] = lhsp(m,i,j+2,k); _lp[2][m] = lhsp(m,i,j+2,k); } _rs[2][3] = rhs(3,i,j+2,k); _rs[2][4] = rhs(4,i,j+2,k); if (j+2 < ny-1) { cv[2] = speed(i,j+3,k); _lp[2][1] -= dtty2*cv[0]; _lp[2][3] += dtty2*cv[2]; _ls[2][1] += dtty2*cv[0]; _ls[2][3] -= dtty2*cv[2]; cv[0] = cv[1]; cv[1] = cv[2]; } fac1 = 1.0/_lp[0][2]; m = 3; _lp[0][3] *= fac1; _lp[0][4] *= fac1; _rs[0][m] *= fac1; _lp[1][2] -= _lp[1][1] * _lp[0][3]; _lp[1][3] -= _lp[1][1] * _lp[0][4]; _rs[1][m] -= _lp[1][1] * _rs[0][m]; _lp[2][1] -= _lp[2][0] * _lp[0][3]; _lp[2][2] -= _lp[2][0] * _lp[0][4]; _rs[2][m] -= _lp[2][0] * _rs[0][m]; m = 4; fac1 = 1.0/_ls[0][2]; _ls[0][3] *= fac1; _ls[0][4] *= fac1; _rs[0][m] *= fac1; _ls[1][2] -= _ls[1][1] * _ls[0][3]; _ls[1][3] -= _ls[1][1] * _ls[0][4]; _rs[1][m] -= _ls[1][1] * _rs[0][m]; _ls[2][1] -= _ls[2][0] * _ls[0][3]; _ls[2][2] -= _ls[2][0] * _ls[0][4]; _rs[2][m] -= _ls[2][0] * _rs[0][m]; //--------------------------------------------------------------------- // store computed lhs and prepare data for next iteration // rhs is stored in a temp array such that write accesses are coalesced //--------------------------------------------------------------------- for (m = 3; m < 5; m++) { lhsp(m,i,j,k) = _lp[0][m]; lhsm(m,i,j,k) = _ls[0][m]; rtmp(m,i,j,k) = _rs[0][m]; _rs[0][m] = _rs[1][m]; _rs[1][m] = _rs[2][m]; } for (m = 0; m < 5; m++) { _lp[0][m] = _lp[1][m]; _lp[1][m] = _lp[2][m]; _ls[0][m] = _ls[1][m]; _ls[1][m] = _ls[2][m]; } } //--------------------------------------------------------------------- // And again the last two rows separately //--------------------------------------------------------------------- j = ny-2; m = 3; fac1 = 1.0/_lp[0][2]; _lp[0][3] *= fac1; _lp[0][4] *= fac1; _rs[0][m] *= fac1; _lp[1][2] -= _lp[1][1] * _lp[0][3]; _lp[1][3] -= _lp[1][1] * _lp[0][4]; _rs[1][m] -= _lp[1][1] * _rs[0][m]; m = 4; fac1 = 1.0/_ls[0][2]; _ls[0][3] *= fac1; _ls[0][4] *= fac1; _rs[0][m] *= fac1; _ls[1][2] -= _ls[1][1] * _ls[0][3]; _ls[1][3] -= _ls[1][1] * _ls[0][4]; _rs[1][m] -= _ls[1][1] * _rs[0][m]; //--------------------------------------------------------------------- // Scale the last row immediately //--------------------------------------------------------------------- _rs[1][3] /= _lp[1][2]; _rs[1][4] /= _ls[1][2]; //--------------------------------------------------------------------- // BACKSUBSTITUTION //--------------------------------------------------------------------- for (m = 0; m < 3; m++) _rs[0][m] -= lhs(3,i,ny-2,k) * _rs[1][m]; _rs[0][3] -= _lp[0][3] * _rs[1][3]; _rs[0][4] -= _ls[0][3] * _rs[1][4]; for (m = 0; m < 5; m++) { _rs[2][m] = _rs[1][m]; _rs[1][m] = _rs[0][m]; } for (j = ny-3; j >= 0; j--) { //--------------------------------------------------------------------- // The first three factors //--------------------------------------------------------------------- for (m = 0; m < 3; m++) _rs[0][m] = rtmp(m,i,j,k) - lhs(3,i,j,k)*_rs[1][m] - lhs(4,i,j,k)*_rs[2][m]; //--------------------------------------------------------------------- // And the remaining two //--------------------------------------------------------------------- _rs[0][3] = rtmp(3,i,j,k) - lhsp(3,i,j,k)*_rs[1][3] - lhsp(4,i,j,k)*_rs[2][3]; _rs[0][4] = rtmp(4,i,j,k) - lhsm(3,i,j,k)*_rs[1][4] - lhsm(4,i,j,k)*_rs[2][4]; if (j+2 < ny-1) { //--------------------------------------------------------------------- // block-diagonal matrix-vector multiplication //--------------------------------------------------------------------- double r1 = _rs[2][0]; double r2 = _rs[2][1]; double r3 = _rs[2][2]; double r4 = _rs[2][3]; double r5 = _rs[2][4]; double t1 = bt * r1; double t2 = 0.5 * ( r4 + r5 ); _rs[2][0] = bt * ( r4 - r5 ); _rs[2][1] = -r3; _rs[2][2] = r2; _rs[2][3] = -t1 + t2; _rs[2][4] = t1 + t2; } for (m = 0; m < 5; m++) { rhs(m,i,j+2,k) = _rs[2][m]; _rs[2][m] = _rs[1][m]; _rs[1][m] = _rs[0][m]; } } //--------------------------------------------------------------------- // block-diagonal matrix-vector multiplication //--------------------------------------------------------------------- double tf1 = bt * _rs[2][0]; double tf2 = 0.5 * ( _rs[2][3] + _rs[2][4] ); rhs(0,i,1,k) = bt * ( _rs[2][3] - _rs[2][4] ); rhs(1,i,1,k) = -_rs[2][2]; rhs(2,i,1,k) = _rs[2][1]; rhs(3,i,1,k) = -tf1 + tf2; rhs(4,i,1,k) = tf1 + tf2; for (m = 0; m < 5; m++) rhs(m,i,0,k) = _rs[1][m]; } #undef lhs #undef lhsp #undef lhsm #undef rtmp //--------------------------------------------------------------------- // adi: z_solve //--------------------------------------------------------------------- #define lhs(m,i,j,k) lhs[(i-1)+(nx-2)*((j-1)+(ny-2)*((k)+nz*(m-3)))] #define lhsp(m,i,j,k) lhs[(i-1)+(nx-2)*((j-1)+(ny-2)*((k)+nz*(m+4)))] #define lhsm(m,i,j,k) lhs[(i-1)+(nx-2)*((j-1)+(ny-2)*((k)+nz*(m-3+2)))] #define rtmp(m,i,j,k) rstmp[(i)+nx*((j)+ny*((k)+nz*(m)))] __global__ static void z_solve_kernel (double *rho_i, double *us, double *vs, double *ws, double *speed, double *qs, double *fu, double *rhs, double *lhs, double *rstmp, const int nx, const int ny, const int nz) { int i, j, k, m; double rhos[3], cv[3], _ls[3][5], _lp[3][5], _rs[3][5], fac1; double zero; j = blockIdx.x*blockDim.x+threadIdx.x+1; i = blockIdx.y*blockDim.y+threadIdx.y+1; if (j >= ny-1 || i >= nx-1) return; //--------------------------------------------------------------------- // Computes the left hand side for the three z-factors //--------------------------------------------------------------------- //--------------------------------------------------------------------- // zap the whole left hand side for starters //--------------------------------------------------------------------- _ls[0][0] = (double)0.0; _ls[0][1] = (double)0.0; _ls[0][2] = (double)1.0; _ls[0][3] = (double)0.0; _ls[0][4] = (double)0.0; lhsp(0,i,j,0) = (double)0.0; lhsp(1,i,j,0) = (double)0.0; lhsp(2,i,j,0) = (double)1.0; lhsp(3,i,j,0) = (double)0.0; lhsp(4,i,j,0) = (double)0.0; zero = (double)0.0; //--------------------------------------------------------------------- // first fill the lhs for the u-eigenvalue //--------------------------------------------------------------------- for (k = 0; k < 3; k++) { fac1 = c3c4*rho_i(i,j,k); //rhos[k] = max(max(max(dz4+con43*fac1, dz5+c1c5*fac1), dzmax+fac1), zero+dz1); if (dz4+con43*fac1>dz5+c1c5*fac1) rhos[k] = dz4+con43*fac1; else rhos[k] = dz5+c1c5*fac1; if (rhos[k]<dzmax+fac1) rhos[k] = dzmax+fac1; if (rhos[k]<zero+dz1) rhos[k] = zero+dz1; cv[k] = ws(i,j,k); } _ls[1][0] = (double)0.0; _ls[1][1] = -dttz2*cv[0] - dtz1*rhos[0]; _ls[1][2] = 1.0 + c2dttz1 * rhos[1]; _ls[1][3] = dttz2*cv[2] - dtz1*rhos[2]; _ls[1][4]= (double)0.0; _ls[1][2] += comz5; _ls[1][3] -= comz4; _ls[1][4] += comz1; for (m = 0; m < 5; m++) lhsp(m,i,j,1) = _ls[1][m]; rhos[0] = rhos[1]; rhos[1] = rhos[2]; cv[0] = cv[1]; cv[1] = cv[2]; for (m = 0; m < 3; m++) { _rs[0][m] = rhs(m,i,j,0); _rs[1][m] = rhs(m,i,j,1); } //--------------------------------------------------------------------- // FORWARD ELIMINATION //--------------------------------------------------------------------- for (k = 0; k < nz-2; k++) { //--------------------------------------------------------------------- // first fill the lhs for the u-eigenvalue //--------------------------------------------------------------------- if (k+2 == nz-1) { _ls[2][0] = (double)0.0; _ls[2][1] = (double)0.0; _ls[2][2] = (double)1.0; _ls[2][3] = (double)0.0; _ls[2][4] = (double)0.0; lhsp(0,i,j,k+2) = (double)0.0; lhsp(1,i,j,k+2) = (double)0.0; lhsp(2,i,j,k+2) = (double)1.0; lhsp(3,i,j,k+2) = (double)0.0; lhsp(4,i,j,k+2) = (double)0.0; } else { fac1 = c3c4*rho_i(i,j,k+3); //rhos[2] = max(max(max(dz4+con43*fac1, dz5+c1c5*fac1), dzmax+fac1), zero+dz1); if (dz4+con43*fac1>dz5+c1c5*fac1) rhos[2] = dz4+con43*fac1; else rhos[2] = dz5+c1c5*fac1; if (rhos[2]<dzmax+fac1) rhos[2] = dzmax+fac1; if (rhos[2]<zero+dz1) rhos[2] = zero+dz1; cv[2] = ws(i,j,k+3); _ls[2][0] = (double)0.0; _ls[2][1] = -dttz2*cv[0] - dtz1*rhos[0]; _ls[2][2] = 1.0 + c2dttz1 * rhos[1]; _ls[2][3] = dttz2*cv[2] - dtz1*rhos[2]; _ls[2][4] = (double)0.0; //--------------------------------------------------------------------- // add fourth order dissipation //--------------------------------------------------------------------- if (k+2 == 2) { _ls[2][1] -= comz4; _ls[2][2] += comz6; _ls[2][3] -= comz4; _ls[2][4] += comz1; } else if (k+2 >= 3 && k+2 < nz-3) { _ls[2][0] += comz1; _ls[2][1] -= comz4; _ls[2][2] += comz6; _ls[2][3] -= comz4; _ls[2][4] += comz1; } else if (k+2 == nz-3) { _ls[2][0] += comz1; _ls[2][1] -= comz4; _ls[2][2] += comz6; _ls[2][3] -= comz4; } else if (k+2 == nz-2) { _ls[2][0] += comz1; _ls[2][1] -= comz4; _ls[2][2] += comz5; } //--------------------------------------------------------------------- // store computed lhs for later reuse //--------------------------------------------------------------------- for (m = 0; m < 5; m++) lhsp(m,i,j,k+2) = _ls[2][m]; rhos[0] = rhos[1]; rhos[1] = rhos[2]; cv[0] = cv[1]; cv[1] = cv[2]; } //--------------------------------------------------------------------- // load rhs values for current iteration //--------------------------------------------------------------------- for (m = 0; m < 3; m++) _rs[2][m] = rhs(m,i,j,k+2); //--------------------------------------------------------------------- // perform current iteration //--------------------------------------------------------------------- fac1 = 1.0/_ls[0][2]; _ls[0][3] *= fac1; _ls[0][4] *= fac1; for (m = 0; m < 3; m++) _rs[0][m] *= fac1; _ls[1][2] -= _ls[1][1] * _ls[0][3]; _ls[1][3] -= _ls[1][1] * _ls[0][4]; for (m = 0; m < 3; m++) _rs[1][m] -= _ls[1][1] * _rs[0][m]; _ls[2][1] -= _ls[2][0] * _ls[0][3]; _ls[2][2] -= _ls[2][0] * _ls[0][4]; for (m = 0; m < 3; m++) _rs[2][m] -= _ls[2][0] * _rs[0][m]; //--------------------------------------------------------------------- // store computed lhs and prepare data for next iteration // rhs is stored in a temp array such that write accesses are coalesced //--------------------------------------------------------------------- lhs(3,i,j,k) = _ls[0][3]; lhs(4,i,j,k) = _ls[0][4]; for (m = 0; m < 5; m++) { _ls[0][m] = _ls[1][m]; _ls[1][m] = _ls[2][m]; } for (m = 0; m < 3; m++) { rtmp(m,i,j,k) = _rs[0][m]; _rs[0][m] = _rs[1][m]; _rs[1][m] = _rs[2][m]; } } //--------------------------------------------------------------------- // The last two rows in this zone are a bit different, // since they do not have two more rows available for the // elimination of off-diagonal entries //--------------------------------------------------------------------- k = nz-2; fac1 = 1.0/_ls[0][2]; _ls[0][3] *= fac1; _ls[0][4] *= fac1; for (m = 0; m < 3; m++) _rs[0][m] *= fac1; _ls[1][2] -= _ls[1][1] * _ls[0][3]; _ls[1][3] -= _ls[1][1] * _ls[0][4]; for (m = 0; m < 3; m++) _rs[1][m] -= _ls[1][1] * _rs[0][m]; //--------------------------------------------------------------------- // scale the last row immediately //--------------------------------------------------------------------- fac1 = 1.0/_ls[1][2]; for (m = 0; m < 3; m++) _rs[1][m] *= fac1; lhs(3,i,j,k) = _ls[0][3]; lhs(4,i,j,k) = _ls[0][4]; //--------------------------------------------------------------------- // subsequently, fill the other factors u+c, u-c //--------------------------------------------------------------------- for (k = 0; k < 3; k++) cv[k] = speed(i,j,k); for (m = 0; m < 5; m++) { _ls[0][m] = lhsp(m,i,j,0); _lp[0][m] = lhsp(m,i,j,0); _ls[1][m] = lhsp(m,i,j,1); _lp[1][m] = lhsp(m,i,j,1); } _lp[1][1] -= dttz2*cv[0]; _lp[1][3] += dttz2*cv[2]; _ls[1][1] += dttz2*cv[0]; _ls[1][3] -= dttz2*cv[2]; cv[0] = cv[1]; cv[1] = cv[2]; _rs[0][3] = rhs(3,i,j,0); _rs[0][4] = rhs(4,i,j,0); _rs[1][3] = rhs(3,i,j,1); _rs[1][4] = rhs(4,i,j,1); //--------------------------------------------------------------------- // do the u+c and the u-c factors //--------------------------------------------------------------------- for (k = 0; k < nz-2; k++) { //--------------------------------------------------------------------- // first, fill the other factors u+c, u-c //--------------------------------------------------------------------- for (m = 0; m < 5; m++) { _ls[2][m] = lhsp(m,i,j,k+2); _lp[2][m] = lhsp(m,i,j,k+2); } _rs[2][3] = rhs(3,i,j,k+2); _rs[2][4] = rhs(4,i,j,k+2); if (k+2 < nz-1) { cv[2] = speed(i,j,k+3); _lp[2][1] -= dttz2*cv[0]; _lp[2][3] += dttz2*cv[2]; _ls[2][1] += dttz2*cv[0]; _ls[2][3] -= dttz2*cv[2]; cv[0] = cv[1]; cv[1] = cv[2]; } m = 3; fac1 = 1.0/_lp[0][2]; _lp[0][3] *= fac1; _lp[0][4] *= fac1; _rs[0][m] *= fac1; _lp[1][2] -= _lp[1][1] * _lp[0][3]; _lp[1][3] -= _lp[1][1] * _lp[0][4]; _rs[1][m] -= _lp[1][1] * _rs[0][m]; _lp[2][1] -= _lp[2][0] * _lp[0][3]; _lp[2][2] -= _lp[2][0] * _lp[0][4]; _rs[2][m] -= _lp[2][0] * _rs[0][m]; m = 4; fac1 = 1.0/_ls[0][2]; _ls[0][3] *= fac1; _ls[0][4] *= fac1; _rs[0][m] *= fac1; _ls[1][2] -= _ls[1][1] * _ls[0][3]; _ls[1][3] -= _ls[1][1] * _ls[0][4]; _rs[1][m] -= _ls[1][1] * _rs[0][m]; _ls[2][1] -= _ls[2][0] * _ls[0][3]; _ls[2][2] -= _ls[2][0] * _ls[0][4]; _rs[2][m] -= _ls[2][0] * _rs[0][m]; //--------------------------------------------------------------------- // store computed lhs and prepare data for next iteration // rhs is stored in a temp array such that write accesses are coalesced //--------------------------------------------------------------------- for (m = 3; m < 5; m++) { lhsp(m,i,j,k) = _lp[0][m]; lhsm(m,i,j,k) = _ls[0][m]; rtmp(m,i,j,k) = _rs[0][m]; _rs[0][m] = _rs[1][m]; _rs[1][m] = _rs[2][m]; } for (m = 0; m < 5; m++) { _lp[0][m] = _lp[1][m]; _lp[1][m] = _lp[2][m]; _ls[0][m] = _ls[1][m]; _ls[1][m] = _ls[2][m]; } } //--------------------------------------------------------------------- // And again the last two rows separately //--------------------------------------------------------------------- k = nz-2; m = 3; fac1 = 1.0/_lp[0][2]; _lp[0][3] *= fac1; _lp[0][4] *= fac1; _rs[0][m] *= fac1; _lp[1][2] -= _lp[1][1] * _lp[0][3]; _lp[1][3] -= _lp[1][1] * _lp[0][4]; _rs[1][m] -= _lp[1][1] * _rs[0][m]; m = 4; fac1 = 1.0/_ls[0][2]; _ls[0][3] *= fac1; _ls[0][4] *= fac1; _rs[0][m] *= fac1; _ls[1][2] -= _ls[1][1] * _ls[0][3]; _ls[1][3] -= _ls[1][1] * _ls[0][4]; _rs[1][m] -= _ls[1][1] * _rs[0][m]; //--------------------------------------------------------------------- // Scale the last row immediately some of this is overkill // if this is the last cell //--------------------------------------------------------------------- _rs[1][3] /= _lp[1][2]; _rs[1][4] /= _ls[1][2]; //--------------------------------------------------------------------- // BACKSUBSTITUTION //--------------------------------------------------------------------- for (m = 0; m < 3; m++) _rs[0][m] -= lhs(3,i,j,nz-2) * _rs[1][m]; _rs[0][3] -= _lp[0][3] * _rs[1][3]; _rs[0][4] -= _ls[0][3] * _rs[1][4]; for (m = 0; m < 5; m++) { _rs[2][m] = _rs[1][m]; _rs[1][m] = _rs[0][m]; } for (k = nz-3; k >= 0; k--) { //--------------------------------------------------------------------- // The first three factors //--------------------------------------------------------------------- for (m = 0; m < 3; m++) _rs[0][m] = rtmp(m,i,j,k) - lhs(3,i,j,k)*_rs[1][m] - lhs(4,i,j,k)*_rs[2][m]; //--------------------------------------------------------------------- // And the remaining two //--------------------------------------------------------------------- _rs[0][3] = rtmp(3,i,j,k) - lhsp(3,i,j,k)*_rs[1][3] - lhsp(4,i,j,k)*_rs[2][3]; _rs[0][4] = rtmp(4,i,j,k) - lhsm(3,i,j,k)*_rs[1][4] - lhsm(4,i,j,k)*_rs[2][4]; if (k+2 < nz-1) { //--------------------------------------------------------------------- // block-diagonal matrix-vector multiplication tzetar //--------------------------------------------------------------------- double xvel = us(i,j,k+2); double yvel = vs(i,j,k+2); double zvel = ws(i,j,k+2); double ac = speed(i,j,k+2); double uzik1 = fu(0,i,j,k+2); double t1 = (bt*uzik1)/ac * (_rs[2][3] + _rs[2][4]); double t2 = _rs[2][2] + t1; double t3 = bt*uzik1 * (_rs[2][3] - _rs[2][4]); _rs[2][4] = uzik1*(-xvel*_rs[2][1] + yvel*_rs[2][0]) + qs(i,j,k+2)*t2 + c2iv*(ac*ac)*t1 + zvel*t3; _rs[2][3] = zvel*t2 + t3; _rs[2][2] = uzik1*_rs[2][0] + yvel*t2; _rs[2][1] = -uzik1*_rs[2][1] + xvel*t2; _rs[2][0] = t2; } for (m = 0; m < 5; m++) { rhs(m,i,j,k+2) = _rs[2][m]; _rs[2][m] = _rs[1][m]; _rs[1][m] = _rs[0][m]; } } //--------------------------------------------------------------------- // block-diagonal matrix-vector multiplication tzetar //--------------------------------------------------------------------- double xfvel = us(i,j,1); double yfvel = vs(i,j,1); double zfvel = ws(i,j,1); double afc = speed(i,j,1); double ufzik1 = fu(0,i,j,1); double tf1 = (bt*ufzik1)/afc * (_rs[2][3] + _rs[2][4]); double tf2 = _rs[2][2] + tf1; double tf3 = bt*ufzik1 * (_rs[2][3] - _rs[2][4]); rhs(4,i,j,1) = ufzik1*(-xfvel*_rs[2][1] + yfvel*_rs[2][0]) + qs(i,j,1)*tf2 + c2iv*(afc*afc)*tf1 + zfvel*tf3; rhs(3,i,j,1) = zfvel*tf2 + tf3; rhs(2,i,j,1) = ufzik1*_rs[2][0] + yfvel*tf2; rhs(1,i,j,1) = -ufzik1*_rs[2][1] + xfvel*tf2; rhs(0,i,j,1) = tf2; for (m = 0; m < 5; m++) rhs(m,i,j,0) = _rs[1][m]; } #undef lhs #undef lhsp #undef lhsm #undef rtmp //--------------------------------------------------------------------- // addition of update to the vector u //--------------------------------------------------------------------- __global__ static void add_kernel (double *fu, double *rhs, const int nx, const int ny, const int nz) { int i, j, k, m; k = blockIdx.y+1; j = blockIdx.x+1; i = threadIdx.x+1; m = threadIdx.y; fu(m,i,j,k) += rhs(m,i,j,k); } //--------------------------------------------------------------------- // adi //--------------------------------------------------------------------- void adi(bool singlestep, int nx, int ny, int nz, int niter, double* rho_i, double* us, double* vs, double* ws, double* speed, double* qs, double* square, double* rhs, double* lhs, double* forcing, double* fu, double* rstmp) { HANDLE_ERROR(cudaDeviceSynchronize()); int itmax = singlestep ? 1 : niter; int xblock, xgrid, yblock, ygrid, zblock, zgrid; for (int step = 1; step <= itmax; step++) { if (step % 20 == 0 || step == 1 && !singlestep) printf(" Time step %4d\n", step); //compute_rhs(); dim3 grid1(ny,nz); compute_rhs_kernel_1<<<grid1,nx>>>(rho_i, us, vs, ws, speed, qs, square, fu, nx, ny, nz); compute_rhs_kernel_2<<<grid1,nx>>>(rho_i, us, vs, ws, qs, square, rhs, forcing, fu, nx, ny, nz); //txinvr(); dim3 grid2(ny-2,nz-2); txinvr_kernel<<<grid2,nx-2>>> (rho_i, us, vs, ws, speed, qs, rhs, nx, ny, nz); //x_solve(); yblock = min(SOLVE_BLOCK,ny); ygrid = (ny+yblock-1)/yblock; zblock = min(SOLVE_BLOCK/yblock,nz); zgrid = (nz+zblock-1)/zblock; dim3 grid3(zgrid,ygrid), block3(zblock,yblock); x_solve_kernel<<<grid3,block3>>>(rho_i, us, speed, rhs, lhs, rstmp, nx, ny, nz); //y_solve(); xblock = min(SOLVE_BLOCK,nx); xgrid = (nx+xblock-1)/xblock; zblock = min(SOLVE_BLOCK/xblock,nz); zgrid = (nz+zblock-1)/zblock; dim3 grid4(zgrid,xgrid), block4(zblock,xblock); y_solve_kernel<<<grid4,block4>>>(rho_i, vs, speed, rhs, lhs, rstmp, nx, ny, nz); //z_solve(); xblock = min(SOLVE_BLOCK,nx); xgrid = (nx+xblock-1)/xblock; yblock = min(SOLVE_BLOCK/xblock,ny); ygrid = (ny+yblock-1)/yblock; dim3 grid5(ygrid,xgrid), block5(yblock,xblock); z_solve_kernel<<<grid5,block5>>>(rho_i, us, vs, ws, speed, qs, fu, rhs, lhs, rstmp, nx, ny, nz); //add(); dim3 grid6(ny-2,nz-2); dim3 block6(nx-2,5); add_kernel<<<grid6,block6>>>(fu, rhs, nx, ny, nz); } HANDLE_ERROR(cudaDeviceSynchronize()); } //--------------------------------------------------------------------- // defaults from parameters //--------------------------------------------------------------------- void read_input(char benchclass, double* dd_td, int* nx, int* ny, int* nz, int* niter) { FILE *file; if ((file = fopen("inputsp.data", "rt")) != NULL) { char line[1024]; printf(" Reading from input file inputsp.data\n"); fgets(line, sizeof(line)-1, file); sscanf(line, "%i", niter); fgets(line, sizeof(line)-1, file); sscanf(line, "%lf", dd_td); fgets(line, sizeof(line)-1, file); sscanf(line, "%i %i %i", nx, ny, nz); fclose(file); } else { // printf(" No input file inputsp.data. Using compiled defaults\n"); int problem_size; switch (benchclass) { case 's': case 'S': problem_size = 12; *dd_td = 0.015; *niter = 100; break; case 'w': case 'W': problem_size = 36; *dd_td = 0.0015; *niter = 400; break; case 'a': case 'A': problem_size = 64; *dd_td = 0.0015; *niter = 400; break; case 'b': case 'B': problem_size = 102; *dd_td = 0.001; *niter = 400; break; case 'c': case 'C': problem_size = 162; *dd_td = 0.00067; *niter = 400; break; case 'd': case 'D': problem_size = 408; *dd_td = 0.00030; *niter = 500; break; case 'e': case 'E': problem_size = 1020; *dd_td = 0.0001; *niter = 500; break; default: printf("setparams: Internal error: invalid class %c\n", benchclass); exit(EXIT_FAILURE); } *nx = *ny = *nz = problem_size; } printf("\n\n NAS Parallel Benchmarks (NPB3.3-CUDA) - SP Benchmark\n\n"); printf(" Size: %4dx%4dx%4d\n", *nx, *ny, *nz); printf(" Iterations: %4d dt_d: %10.6F\n", *niter, *dd_td); printf("\n"); } int main(int argc, char **argv) { char benchclass = argc > 1 ? argv[1][0] : 'S'; struct timeval start_t; struct timeval end_t; struct timeval skt_t; struct timeval ske_t; int niter; int nx, ny, nz; double hdd; double dd_d; double *fu, *forcing, *rhs, *rho_i, *us, *vs, *ws, *qs, *speed, *square, *lhs, *rstmp; //double* rmsbuf; //double xce[5], xcr[5]; char CUDAname[256]; int CUDAmp, CUDAclock, CUDAmemclock, CUDAl2cache; size_t CUDAmem; //--------------------------------------------------------------------- // read input data //--------------------------------------------------------------------- read_input(benchclass, &hdd, &nx, &ny, &nz, &niter); dd_d = hdd; //--------------------------------------------------------------------- // allocate CUDA device memory //--------------------------------------------------------------------- int gridsize = nx*ny*nz; int facesize = max(max(nx*ny, nx*nz), ny*nz); gettimeofday(&start_t, NULL); HANDLE_ERROR(cudaMalloc((void **)&fu, 5*gridsize*sizeof(double))); HANDLE_ERROR(cudaMalloc((void **)&forcing, 5*gridsize*sizeof(double))); HANDLE_ERROR(cudaMalloc((void **)&rhs, 5*gridsize*sizeof(double))); HANDLE_ERROR(cudaMalloc((void **)&rho_i, gridsize*sizeof(double))); HANDLE_ERROR(cudaMalloc((void **)&us, gridsize*sizeof(double))); HANDLE_ERROR(cudaMalloc((void **)&vs, gridsize*sizeof(double))); HANDLE_ERROR(cudaMalloc((void **)&ws, gridsize*sizeof(double))); HANDLE_ERROR(cudaMalloc((void **)&qs, gridsize*sizeof(double))); HANDLE_ERROR(cudaMalloc((void **)&speed, gridsize*sizeof(double))); HANDLE_ERROR(cudaMalloc((void **)&square, gridsize*sizeof(double))); HANDLE_ERROR(cudaMalloc((void **)&lhs, 9*gridsize*sizeof(double))); HANDLE_ERROR(cudaMalloc((void **)&rstmp, 5*gridsize*sizeof(double))); //HANDLE_ERROR(cudaMalloc((void **)&rmsbuf, 5*facesize*sizeof(double))); double ce_d[13][5]; ce_d[0][0] = (double)2.0; ce_d[1][0] = (double)0.0; ce_d[2][0] = (double)0.0; ce_d[3][0] = (double)4.0; ce_d[4][0] = (double)5.0; ce_d[5][0] = (double)3.0; ce_d[6][0] = (double)0.5; ce_d[7][0] = (double)0.02; ce_d[8][0] = (double)0.01; ce_d[9][0] = (double)0.03; ce_d[10][0] = (double)0.5; ce_d[11][0] = (double)0.4; ce_d[12][0] = (double)0.3; ce_d[0][1] = (double)1.0; ce_d[1][1] = (double)0.0; ce_d[2][1] = (double)0.0; ce_d[3][1] = (double)0.0; ce_d[4][1] = (double)1.0; ce_d[5][1] = (double)2.0; ce_d[6][1] = (double)3.0; ce_d[7][1] = (double)0.01; ce_d[8][1] = (double)0.03; ce_d[9][1] = (double)0.02; ce_d[10][1] = (double)0.4; ce_d[11][1] = (double)0.3; ce_d[12][1] = (double)0.5; ce_d[0][2] = (double)2.0; ce_d[1][2] = (double)2.0; ce_d[2][2] = (double)0.0; ce_d[3][2] = (double)0.0; ce_d[4][2] = (double)0.0; ce_d[5][2] = (double)2.0; ce_d[6][2] = (double)3.0; ce_d[7][2] = (double)0.04; ce_d[8][2] = (double)0.03; ce_d[9][2] = (double)0.05; ce_d[10][2] = (double)0.3; ce_d[11][2] = (double)0.5; ce_d[12][2] = (double)0.4; ce_d[0][3] = (double)2.0; ce_d[1][3] = (double)2.0; ce_d[2][3] = (double)0.0; ce_d[3][3] = (double)0.0; ce_d[4][3] = (double)0.0; ce_d[5][3] = (double)2.0; ce_d[6][3] = (double)3.0; ce_d[7][3] = (double)0.03; ce_d[8][3] = (double)0.05; ce_d[9][3] = (double)0.04; ce_d[10][3] = (double)0.2; ce_d[11][3] = (double)0.1; ce_d[12][3] = (double)0.3; ce_d[0][4] = (double)5.0; ce_d[1][4] = (double)4.0; ce_d[2][4] = (double)3.0; ce_d[3][4] = (double)2.0; ce_d[4][4] = (double)0.1; ce_d[5][4] = (double)0.4; ce_d[6][4] = (double)0.3; ce_d[7][4] = (double)0.05; ce_d[8][4] = (double)0.04; ce_d[9][4] = (double)0.03; ce_d[10][4] = (double)0.1; ce_d[11][4] = (double)0.3; ce_d[12][4] = (double)0.2; double bt_d = sqrt(0.5); double dnxm1_d = 1.0/((double)nx-1.0); double dnym1_d = 1.0/((double)ny-1.0); double dnzm1_d = 1.0/((double)nz-1.0); double tx1_d = 1.0 / (dnxm1_d * dnxm1_d); double tx2_d = 1.0 / (2.0 * dnxm1_d); double tx3_d = 1.0 / dnxm1_d; double ty1_d = 1.0 / (dnym1_d * dnym1_d); double ty2_d = 1.0 / (2.0 * dnym1_d); double ty3_d = 1.0 / dnym1_d; double tz1_d = 1.0 / (dnzm1_d * dnzm1_d); double tz2_d = 1.0 / (2.0 * dnzm1_d); double tz3_d = 1.0 / dnzm1_d; double dtx1_d = dd_d*tx1_d; double dttx2_d = dd_d*tx2_d; double dty1_d = dd_d*ty1_d; double dtty2_d = dd_d*ty2_d; double dtz1_d = dd_d*tz1_d; double dttz2_d = dd_d*tz2_d; double c2dttx1_d = 2.0*dtx1_d; double c2dtty1_d = 2.0*dty1_d; double c2dttz1_d = 2.0*dtz1_d; double dtdssp_d = dd_d*dssp; double comz1_d = dtdssp_d; double comz4_d = 4.0*dtdssp_d; double comz5_d = 5.0*dtdssp_d; double comz6_d = 6.0*dtdssp_d; double c3c4tx3_d = c3c4*tx3_d; double c3c4ty3_d = c3c4*ty3_d; double c3c4tz3_d = c3c4*tz3_d; double dx1tx1_d = dx1*tx1_d; double dx2tx1_d = dx2*tx1_d; double dx3tx1_d = dx3*tx1_d; double dx4tx1_d = dx4*tx1_d; double dx5tx1_d = dx5*tx1_d; double dy1ty1_d = dy1*ty1_d; double dy2ty1_d = dy2*ty1_d; double dy3ty1_d = dy3*ty1_d; double dy4ty1_d = dy4*ty1_d; double dy5ty1_d = dy5*ty1_d; double dz1tz1_d = dz1*tz1_d; double dz2tz1_d = dz2*tz1_d; double dz3tz1_d = dz3*tz1_d; double dz4tz1_d = dz4*tz1_d; double dz5tz1_d = dz5*tz1_d; double xxcon1_d = c3c4tx3_d*con43*tx3_d; double xxcon2_d = c3c4tx3_d*tx3_d; double xxcon3_d = c3c4tx3_d*conz1*tx3_d; double xxcon4_d = c3c4tx3_d*con16*tx3_d; double xxcon5_d = c3c4tx3_d*c1c5*tx3_d; double yycon1_d = c3c4ty3_d*con43*ty3_d; double yycon2_d = c3c4ty3_d*ty3_d; double yycon3_d = c3c4ty3_d*conz1*ty3_d; double yycon4_d = c3c4ty3_d*con16*ty3_d; double yycon5_d = c3c4ty3_d*c1c5*ty3_d; double zzcon1_d = c3c4tz3_d*con43*tz3_d; double zzcon2_d = c3c4tz3_d*tz3_d; double zzcon3_d = c3c4tz3_d*conz1*tz3_d; double zzcon4_d = c3c4tz3_d*con16*tz3_d; double zzcon5_d = c3c4tz3_d*c1c5*tz3_d; HANDLE_ERROR (cudaMemcpyToSymbol (&ce, &ce_d, 13*5*sizeof(double))); HANDLE_ERROR (cudaMemcpyToSymbol (&bt, &bt_d, sizeof(double))); HANDLE_ERROR (cudaMemcpyToSymbol (&dnxm1, &dnxm1_d, sizeof(double))); HANDLE_ERROR (cudaMemcpyToSymbol (&dnym1, &dnym1_d, sizeof(double))); HANDLE_ERROR (cudaMemcpyToSymbol (&dnzm1, &dnzm1_d, sizeof(double))); HANDLE_ERROR (cudaMemcpyToSymbol (&tx1, &tx1_d, sizeof(double))); HANDLE_ERROR (cudaMemcpyToSymbol (&tx2, &tx2_d, sizeof(double))); HANDLE_ERROR (cudaMemcpyToSymbol (&tx3, &tx3_d, sizeof(double))); HANDLE_ERROR (cudaMemcpyToSymbol (&ty1, &ty1_d, sizeof(double))); HANDLE_ERROR (cudaMemcpyToSymbol (&ty2, &ty2_d, sizeof(double))); HANDLE_ERROR (cudaMemcpyToSymbol (&ty3, &ty3_d, sizeof(double))); HANDLE_ERROR (cudaMemcpyToSymbol (&tz1, &tz1_d, sizeof(double))); HANDLE_ERROR (cudaMemcpyToSymbol (&tz2, &tz2_d, sizeof(double))); HANDLE_ERROR (cudaMemcpyToSymbol (&tz3, &tz3_d, sizeof(double))); HANDLE_ERROR (cudaMemcpyToSymbol (&dtx1, &dtx1_d, sizeof(double))); HANDLE_ERROR (cudaMemcpyToSymbol (&dttx2, &dttx2_d, sizeof(double))); HANDLE_ERROR (cudaMemcpyToSymbol (&dty1, &dty1_d, sizeof(double))); HANDLE_ERROR (cudaMemcpyToSymbol (&dtty2, &dtty2_d, sizeof(double))); HANDLE_ERROR (cudaMemcpyToSymbol (&dtz1, &dtz1_d, sizeof(double))); HANDLE_ERROR (cudaMemcpyToSymbol (&dttz2, &dttz2_d, sizeof(double))); HANDLE_ERROR (cudaMemcpyToSymbol (&c2dttx1, &c2dttx1_d, sizeof(double))); HANDLE_ERROR (cudaMemcpyToSymbol (&c2dtty1, &c2dtty1_d, sizeof(double))); HANDLE_ERROR (cudaMemcpyToSymbol (&c2dttz1, &c2dttz1_d, sizeof(double))); HANDLE_ERROR (cudaMemcpyToSymbol (&dt, &dd_d, sizeof(double))); HANDLE_ERROR (cudaMemcpyToSymbol (&dtdssp, &dtdssp_d, sizeof(double))); HANDLE_ERROR (cudaMemcpyToSymbol (&comz1, &comz1_d, sizeof(double))); HANDLE_ERROR (cudaMemcpyToSymbol (&comz4, &comz4_d, sizeof(double))); HANDLE_ERROR (cudaMemcpyToSymbol (&comz5, &comz5_d, sizeof(double))); HANDLE_ERROR (cudaMemcpyToSymbol (&comz6, &comz6_d, sizeof(double))); HANDLE_ERROR (cudaMemcpyToSymbol (&c3c4tx3, &c3c4tx3_d, sizeof(double))); HANDLE_ERROR (cudaMemcpyToSymbol (&c3c4ty3, &c3c4ty3_d, sizeof(double))); HANDLE_ERROR (cudaMemcpyToSymbol (&c3c4tz3, &c3c4tz3_d, sizeof(double))); HANDLE_ERROR (cudaMemcpyToSymbol (&dx1tx1, &dx1tx1_d, sizeof(double))); HANDLE_ERROR (cudaMemcpyToSymbol (&dx2tx1, &dx2tx1_d, sizeof(double))); HANDLE_ERROR (cudaMemcpyToSymbol (&dx3tx1, &dx3tx1_d, sizeof(double))); HANDLE_ERROR (cudaMemcpyToSymbol (&dx4tx1, &dx4tx1_d, sizeof(double))); HANDLE_ERROR (cudaMemcpyToSymbol (&dx5tx1, &dx5tx1_d, sizeof(double))); HANDLE_ERROR (cudaMemcpyToSymbol (&dy1ty1, &dy1ty1_d, sizeof(double))); HANDLE_ERROR (cudaMemcpyToSymbol (&dy2ty1, &dy2ty1_d, sizeof(double))); HANDLE_ERROR (cudaMemcpyToSymbol (&dy3ty1, &dy3ty1_d, sizeof(double))); HANDLE_ERROR (cudaMemcpyToSymbol (&dy4ty1, &dy4ty1_d, sizeof(double))); HANDLE_ERROR (cudaMemcpyToSymbol (&dy5ty1, &dy5ty1_d, sizeof(double))); HANDLE_ERROR (cudaMemcpyToSymbol (&dz1tz1, &dz1tz1_d, sizeof(double))); HANDLE_ERROR (cudaMemcpyToSymbol (&dz2tz1, &dz2tz1_d, sizeof(double))); HANDLE_ERROR (cudaMemcpyToSymbol (&dz3tz1, &dz3tz1_d, sizeof(double))); HANDLE_ERROR (cudaMemcpyToSymbol (&dz4tz1, &dz4tz1_d, sizeof(double))); HANDLE_ERROR (cudaMemcpyToSymbol (&dz5tz1, &dz5tz1_d, sizeof(double))); HANDLE_ERROR (cudaMemcpyToSymbol (&xxcon1, &xxcon1_d, sizeof(double))); HANDLE_ERROR (cudaMemcpyToSymbol (&xxcon2, &xxcon2_d, sizeof(double))); HANDLE_ERROR (cudaMemcpyToSymbol (&xxcon3, &xxcon3_d, sizeof(double))); HANDLE_ERROR (cudaMemcpyToSymbol (&xxcon4, &xxcon4_d, sizeof(double))); HANDLE_ERROR (cudaMemcpyToSymbol (&xxcon5, &xxcon5_d, sizeof(double))); HANDLE_ERROR (cudaMemcpyToSymbol (&yycon1, &yycon1_d, sizeof(double))); HANDLE_ERROR (cudaMemcpyToSymbol (&yycon2, &yycon2_d, sizeof(double))); HANDLE_ERROR (cudaMemcpyToSymbol (&yycon3, &yycon3_d, sizeof(double))); HANDLE_ERROR (cudaMemcpyToSymbol (&yycon4, &yycon4_d, sizeof(double))); HANDLE_ERROR (cudaMemcpyToSymbol (&yycon5, &yycon5_d, sizeof(double))); HANDLE_ERROR (cudaMemcpyToSymbol (&zzcon1, &zzcon1_d, sizeof(double))); HANDLE_ERROR (cudaMemcpyToSymbol (&zzcon2, &zzcon2_d, sizeof(double))); HANDLE_ERROR (cudaMemcpyToSymbol (&zzcon3, &zzcon3_d, sizeof(double))); HANDLE_ERROR (cudaMemcpyToSymbol (&zzcon4, &zzcon4_d, sizeof(double))); HANDLE_ERROR (cudaMemcpyToSymbol (&zzcon5, &zzcon5_d, sizeof(double))); gettimeofday(&skt_t, NULL); exact_rhs(forcing, nx, ny, nz); //sp->initialize(); dim3 grid(nz,ny); initialize_kernel<<<grid,nx>>> (fu, nx, ny, nz); //--------------------------------------------------------------------- // do one time step to touch all code, and reinitialize //--------------------------------------------------------------------- adi(true, nx, ny, nz, niter, rho_i, us, vs, ws, speed, qs, square, rhs, lhs, forcing, fu, rstmp); //sp->initialize(); initialize_kernel<<<grid,nx>>> (fu, nx, ny, nz); //--------------------------------------------------------------------- // main time stepping loop //--------------------------------------------------------------------- //sp->adi(false); adi(false, nx, ny, nz, niter, rho_i, us, vs, ws, speed, qs, square, rhs, lhs, forcing, fu, rstmp); gettimeofday(&ske_t, NULL); gettimeofday(&end_t, NULL); std::cout << "time: "<<((end_t.tv_sec-start_t.tv_sec)+(end_t.tv_usec-start_t.tv_usec)*1e-6) << std::endl; std::cout << "kernel: "<<((ske_t.tv_sec-skt_t.tv_sec)+(ske_t.tv_usec-skt_t.tv_usec)*1e-6) << std::endl; //std::cout << (sdkGetAverageTimerValue(&timer)/1000.0) / iterations << " seconds per iteration" << std::endl; //--------------------------------------------------------------------- // verification test //--------------------------------------------------------------------- //char verifyclass; //bool verified; //--------------------------------------------------------------------- // More timers //--------------------------------------------------------------------- //sp->print_timers(); //delete sp; return EXIT_SUCCESS; }
1,235
#include "includes.h" __global__ void bcnn_grad_scales_kernel(float *x_norm, float *delta, int batch, int n, int size, float *scale_updates) { __shared__ float part[BCNN_CUDA_THREADS]; int i, b; int filter = blockIdx.x; int p = threadIdx.x; float sum = 0; for (b = 0; b < batch; ++b) { for (i = 0; i < size; i += BCNN_CUDA_THREADS) { int index = p + i + size * (filter + n * b); sum += (p + i < size) ? delta[index] * x_norm[index] : 0; } } part[p] = sum; __syncthreads(); if (p == 0) { for (i = 0; i < BCNN_CUDA_THREADS; ++i) scale_updates[filter] += part[i]; } }
1,236
#define ulong unsigned long long #define uint unsigned int #define MOD_P0 469762049LL #define MOD_P1 1811939329LL #define MOD_P2 2013265921LL //R‚݂̌ɑfP^̂ŁAꂼ̗]肩猳̒l𕜌 //̂ƂP͑SČŒȂ̂ŏ]vZ͑Sߑłł //E0`E2́AE3ɏo //Jオl //arrayLength2=arrayE3̔zTCY __global__ void GarnerGPU(uint *arrayE0,uint *arrayE1,uint *arrayE2,uint *arrayE3,uint arrayLength2 ) { int idx = threadIdx.x+blockIdx.x*256; ulong ar=arrayE0[idx]; ulong br=arrayE1[idx]; ulong cr=arrayE2[idx]; ulong x=ar; ulong brx=br-x+MOD_P1; if (brx>=MOD_P1)brx-=MOD_P1; x=x+(brx*1540148431)%MOD_P1*MOD_P0; //1540148431=modinv(MOD_P0,MOD_P1) //̎_xMOD_P1*MOD_P0ȉł邱Ƃۏ؂Ă ulong crx=cr+MOD_P2-x%MOD_P2; if (crx>=MOD_P2)crx-=MOD_P2; ulong w1=(crx*1050399624)%MOD_P2; //1050399624=modinv(MOD_P0,MOD_P2) *modinv(MOD_P1,MOD_P2)%MOD_P2 ulong w2=MOD_P0*MOD_P1; ulong out_lo=w1*w2; ulong out_hi=__umul64hi(w1,w2); if (out_lo>(out_lo+x)){ out_hi++; } out_lo+=x; //Jオ菈 uint ui00_32=(uint)(out_lo%(1ULL<<32ULL)); uint ui32_64=(uint)(out_lo/(1ULL<<32ULL)); uint ui64_96=(uint)(out_hi%(1ULL<<32ULL)); uint lastE3_0 = atomicAdd( &arrayE3[idx+0], ui00_32 ); if ((lastE3_0+ui00_32)<lastE3_0){//Jオl ui32_64++; if (ui32_64==0)ui64_96++; } if (ui32_64!=0){ uint lastE3_1 = atomicAdd( &arrayE3[idx+1], ui32_64 ); if ((lastE3_1+ui32_64)<lastE3_1){//Jオl ui64_96++;//‚I[o[t[邱Ƃ͐΂ɂȂ } } uint upflg=0; if (ui64_96!=0){ uint lastE3_2 = atomicAdd( &arrayE3[idx+2], ui64_96 ); if ((lastE3_2+ui64_96)<lastE3_2){//Jオl upflg++; } } uint lastE3_i; for(int i=idx+3;i<arrayLength2;i++){ //9999999+1݂ȂƂp if (upflg==0)break; lastE3_i = atomicAdd( &arrayE3[i], upflg ); if (lastE3_i==4294967295){ upflg=1; }else{ upflg=0; } } }
1,237
#include <stdio.h> #include <cuda.h> #include <stdlib.h> #define N 17 // size of arrays __global__ void transpose (int *a, int *b) { int col = blockIdx.x*blockDim.x+threadIdx.x; int row =blockIdx.y*blockDim.y+threadIdx.y; int index1 = col + row * N; int index2 = row + col * N; if ( col < N && row < N) b[index1]= a[index2]; } int main (int argc, char **argv ) { int i,j; int size = N * N *sizeof( int); int a[N][N], *devA, *devB; int gridsize = (int) ceil((double) N/16); for (i = 0; i < N; i++){ // put some numbers into array for (j= 0; j < N; j++) { a[i][j] = j + N * i; } } printf("Initial values"); for (i = 0; i < N; i++) { printf("\n"); for (j= 0; j < N; j++) { printf("%3d ",a[i][j]); } } printf("\nN = %d, grid size = %d\n",N,gridsize); dim3 block (16,16); dim3 grid (gridsize, gridsize); cudaMalloc( (void**)&devA, size ); cudaMalloc( (void**)&devB, size ); cudaMemcpy( devA, a, size, cudaMemcpyHostToDevice); transpose<<<grid, block>>>(devA, devB); cudaMemcpy( a, devB, size, cudaMemcpyDeviceToHost); printf("Results"); for (i = 0; i < N; i++) { printf("\n"); for (j= 0; j < N; j++) { printf("%3d ",a[i][j]); } } printf("\n"); cudaFree( devA); cudaFree( devB); return (0); }
1,238
#include "includes.h" __global__ void Product (float *a, float *b, float *c) { // Out of all the threads created each one computes 1 value of C and stores into cval float cval = 0.00; int R = blockIdx.y * blockDim.y + threadIdx.y; //Row of the matrix int C = blockIdx.x * blockDim.x + threadIdx.x; //Column of the matrix //Defining the size of the matrix// int N=1000; if(R> N || C > N ){ return; } for (int j = 0; j < N; j++) { cval += a[R * N+ j] *b[j * N + C]; } c[R * N + C]+= cval; }
1,239
#include <cuda_runtime.h> #include <stdio.h> __global__ void checkIndex(void) { printf("threadIdx:(%d, %d, %d) blockIdx:(%d, %d, %d) blockDim:(%d, %d, %d) " "gridDim:(%d, %d, %d)\n", threadIdx.x, threadIdx.y, threadIdx.z, blockIdx.x, blockIdx.y, blockIdx.z, blockDim.x, blockDim.y, blockDim.z, gridDim.x,gridDim.y,gridDim.z); } int main(int argc, char **argv) { int nX = 16; int nY = 16; // define total data element //int nEle = 16; //int iLen = 8; // define grid and block structure dim3 block (8, 8); dim3 grid (nX/block.x, nY/block.y); //dim3 block (iLen, iLen); //dim3 grid ((nEle + block.x-1)/block.x, (nEle + block.y-1)/block.y); // check grid and block dimension from host side printf("grid.x %d grid.y %d grid.z %d\n",grid.x, grid.y, grid.z); printf("block.x %d block.y %d block.z %d\n",block.x, block.y, block.z); // check grid and block dimension from device side checkIndex <<<grid, block>>> (); // reset device before you leave cudaDeviceReset(); return(0); }
1,240
// filename: ax.cu // a simple CUDA kernel to add two vectors extern "C" // ensure function name to be exactly "ax" { __global__ void ax(const int lengthC, const double a, const double *b, double *c) { int i = threadIdx.x + blockIdx.x * blockDim.x; if (i<lengthC) { c[i] = a*b[i]; // REMEMBER ZERO INDEXING IN C LANGUAGE!! } } }
1,241
#include "includes.h" __global__ void dot_cmp_kernaldm(const float* data1, const float* data2, const float* dm, float* device_soln, const int size, const int num_threads, const int offset) { float dot = 0.0f; float nnn = 0.0f; int idx = threadIdx.x + blockIdx.x*num_threads + offset; for(int i = 0; i < size; i++){ int index = i*size + idx % size + ((idx/size)*size*size); //for coalesing if(dm[index] > 0.5){ dot += data1[index]*data2[index]; nnn += 1.0f; } } device_soln[idx] = dot/nnn; }
1,242
/****************************************************************************** *cr *cr (C) Copyright 2010 The Board of Trustees of the *cr University of Illinois *cr All Rights Reserved *cr ******************************************************************************/ #define BLOCK_SIZE 512 __global__ void reduction(float *out, float *in, unsigned size) { /******************************************************************** Load a segment of the input vector into shared memory Traverse the reduction tree Write the computed sum to the output vector at the correct index ********************************************************************/ __shared__ float partialSum[2*BLOCK_SIZE]; unsigned int t = threadIdx.x; unsigned int start = 2*blockIdx.x*blockDim.x; partialSum[t] = in[start + t]; partialSum[blockDim.x + t] = in[start + blockDim.x + t]; for (unsigned int stride = BLOCK_SIZE; stride >= 1; stride /= 2) { __syncthreads(); if (t < stride && t + start + stride < size) partialSum[t]+=partialSum[t+stride]; } if (t == 0) out[blockIdx.x] = partialSum[0]; } // INSERT KERNEL CODE HERE
1,243
#include <cuda.h> #include <stdio.h> __global__ void K(int *p) { *p = 0; printf("%d\n", *p); } int main() { int *x, *y; cudaMalloc(&x, sizeof(int)); K<<<2, 10>>>(x); cudaDeviceSynchronize(); y = x; cudaFree(y); K<<<2, 10>>>(x); cudaDeviceSynchronize(); //cudaError_t err = cudaGetLastError(); //printf("error=%d, %s, %s\n", err, cudaGetErrorName(err), cudaGetErrorString(err)); return 0; }
1,244
#include "includes.h" char* concat(char *s1, char *s2); __global__ void r_final_sum_and_alpha_calculation(float * r_squared ,float * p_sum ,int size) { int index = threadIdx.x ; __shared__ float shared_r_squared[1024] ; __shared__ float shared_p_sum[1024] ; if (index < size) { shared_r_squared[index] = r_squared[index] ; shared_p_sum[index] = p_sum[index] ; } else { shared_r_squared[index] = 0 ; shared_p_sum[index] = 0 ; } __syncthreads() ; for (unsigned int s = blockDim.x/2 ; s> 0 ; s >>= 1) { if (index < s) { shared_r_squared[index] = shared_r_squared[index] + shared_r_squared[index +s] ; shared_p_sum[index] = shared_p_sum[index] + shared_p_sum[index +s] ; __syncthreads() ; } } if(threadIdx.x == 0) { //alpha r_squared[blockIdx.x] = shared_r_squared[0]/shared_p_sum[0] ; } }
1,245
#include "rgb_pixels_factory.cuh" int RgbPixelsFactory::random(int max) { return rand() % max; } unsigned char RgbPixelsFactory::randomChar() { return random(256); } Pixel * RgbPixelsFactory::generate(int count, int maxX, int maxY) { Pixel* pixels = new Pixel[count]; for (int i = 0; i < count; i++) { pixels[i].color.red = randomChar(); pixels[i].color.green = randomChar(); pixels[i].color.blue = randomChar(); pixels[i].point.x = random(maxX); pixels[i].point.y = random(maxY); } return pixels; }
1,246
// // CrossCorrelation.cu // CrossCorrelation // // Created by Vivek Sridhar on 29/06/17. // Copyright © 2017 Vivek Sridhar. All rights reserved. // #include <iostream> #include <fstream> #include <sstream> #include <vector> #include <algorithm> #include <stdio.h> #include <stdlib.h> #include <dirent.h> template <typename T> std::string to_string(const T& value) { std::stringstream ss; ss << value; return ss.str(); } long factorial(long val) { std::cout << val << "\n"; long result = 1; for (long i = 1; i <= val; ++i) { result *= i; } return result; } long combination(long n, long r) { return (factorial(n)) / ((factorial(n - r)) * factorial(r)); } __global__ void kernel(float *x1, float *y1, float *x2, float *y2, float *res, int tau, int na_frames, long nElements) { int index = blockIdx.x * blockDim.x + threadIdx.x; if (tau < 0) { if (index >= -tau+na_frames) { res[index] = x1[index] * x2[index + tau] + y1[index] * y2[index + tau]; } else res[index] = 0.0; } else { if (index < nElements - tau) { res[index] = x1[index] * x2[index + tau] + y1[index] * y2[index + tau]; } else res[index] = 0.0; } } // total measurement points in the time series is defined by nElements #define M 1024 // number of threads per block #define fps 10 // frames per second of input video (used to determine tau) #define time 5 // time in seconds within which time delayed cross correlation is calculated (tau ranges from -time*fps to time*fps) #define n_inds 10 int na_frames = 0; // number of frames in the start with nas int scale = 1; // time window for analysis in seconds; varying this allows us to examine dynamics of leadership across varying timescales; setting scale larger than the entire time series or -1 gives aggregated statistics across the entire duration (otherwise, timescale of analysis is scale*fps) //const int pairs = combination(n_inds, 2); const bool aggregate = false; // this boolean decides whether you output a dynamic time variable leadership network or a static time aggregated network; scale is set to -1 if aggregate is true std::ofstream outputFile1; int main () { DIR *dir; FILE *pFile_x1; FILE *pFile_y1; FILE *pFile_x2; FILE *pFile_y2; long lSize; long nElements; struct dirent *file; float *d_x1, *d_y1, *d_x2, *d_y2, *d_res; float *x1, *y1, *x2, *y2, *res; size_t result_x1, result_y1, result_x2, result_y2; if (aggregate) scale = -1; std::vector<std::string> files; std::string directory = "/home/user/Documents/Vivek/cuda/DirectionalCorrelation/Data/Input/pigeons/10_birds/ffA3/cross_correlation/"; dir = opendir(directory.c_str()); int idx = 0; while ((file = readdir(dir)) != NULL) { if (file->d_name[0] == 'd') { files.push_back(file->d_name); ++idx; } } std::sort(files.begin(), files.begin()+2*n_inds); closedir(dir); // Open output file std::string filename_cc; if (scale != -1) filename_cc = "cross_correlation_01.csv"; else filename_cc = "avgd_cross_correlation.csv"; outputFile1.open(filename_cc.c_str()); // Output file headers if (aggregate || scale == -1) outputFile1 << "id1"<< ", " << "id2" << ", " << "tau" << ", " << "cc" << "\n"; else outputFile1 << "time" << ", " << "id1" << ", " << "id2" << ", " << "tau" << ", " << "cc" << "\n"; //files = {"dir_x00", "dir_x01", "dir_y00", "dir_y01"} for (int a = 0; a < n_inds; ++a) { for (int b = 0; b < n_inds; ++b) { if (b != a) { pFile_x1 = fopen ((directory + files[a]).c_str(), "rb"); pFile_y1 = fopen ((directory + files[a+n_inds]).c_str(), "rb"); pFile_x2 = fopen ((directory + files[b]).c_str(), "rb"); pFile_y2 = fopen ((directory + files[b+n_inds]).c_str(), "rb"); if (pFile_x1==NULL || pFile_y1==NULL || pFile_x2==NULL || pFile_y2==NULL) { fputs ("File error",stderr); exit (1); } // obtain file size fseek (pFile_x1 , 0 , SEEK_END); lSize = ftell (pFile_x1); rewind (pFile_x1); nElements = lSize / sizeof(float); // allocate memory to contain the whole file // device memory cudaMalloc((void **) &d_x1, lSize); cudaMalloc((void **) &d_y1, lSize); cudaMalloc((void **) &d_x2, lSize); cudaMalloc((void **) &d_y2, lSize); cudaMalloc((void **) &d_res, lSize); // host memory x1 = (float*) malloc(lSize); y1 = (float*) malloc(lSize); x2 = (float*) malloc(lSize); y2 = (float*) malloc(lSize); res = (float*) malloc(lSize); if (x1 == NULL || y1==NULL || x2==NULL || y2==NULL || res==NULL) { fputs ("Memory error",stderr); exit (2); } // copy the file into the respective float pointers result_x1 = fread (x1, sizeof(float), nElements, pFile_x1); result_y1 = fread (y1, sizeof(float), nElements, pFile_y1); result_x2 = fread (x2, sizeof(float), nElements, pFile_x2); result_y2 = fread (y2, sizeof(float), nElements, pFile_y2); if (result_x1 != nElements || result_y1 != nElements || result_x2 != nElements || result_y2 != nElements) { fputs ("Reading error",stderr); exit (3); } // the whole files are now loaded in the memory x1, y1, x2 and y2 respectively cudaMemcpy(d_x1, x1, lSize, cudaMemcpyHostToDevice); cudaMemcpy(d_y1, y1, lSize, cudaMemcpyHostToDevice); cudaMemcpy(d_x2, x2, lSize, cudaMemcpyHostToDevice); cudaMemcpy(d_y2, y2, lSize, cudaMemcpyHostToDevice); if (scale*fps > nElements) scale = -1; int tau_max[nElements - scale*fps]; float res_tmp[nElements - scale*fps]; float res_max[nElements - scale*fps]; std::fill_n(tau_max, nElements - scale*fps, 0); std::fill_n(res_tmp, nElements - scale*fps, 0.0); std::fill_n(res_max, nElements - scale*fps, -1.0); for (int tau = -time*fps; tau <= time*fps; ++tau) { kernel<<<(nElements + M - 1) / M, M>>>(d_x1, d_y1, d_x2, d_y2, d_res, tau, na_frames, nElements); cudaMemcpy(res, d_res, lSize, cudaMemcpyDeviceToHost); if (scale == -1) { float res_now = -1.0f; for (int i = na_frames; i < nElements; ++i) { if (res[i] != res[i]) std::cout << x1[i] << " " << y1[i] << " " << i << " " << tau << "\n"; // if nans res_now += res[i]; } outputFile1 << (to_string(files[a][5])).c_str() << (to_string(files[a][6])).c_str() << (to_string(files[a][7])).c_str() << ", " << (to_string(files[b][5])).c_str() << (to_string(files[b][6])).c_str() << (to_string(files[b][7])).c_str() << ", " << tau << ", " << res_now / nElements << "\n"; } else { std::fill_n(res_tmp, nElements - scale*fps, 0.0); for (int i = na_frames; i < nElements - scale*fps; ++i) { for (int j = i; j < i + scale*fps; ++j) { res_tmp[i] += res[j]; if (j == i + scale*fps - 1 && res_max[i] < res_tmp[i]) { res_max[i] = res_tmp[i]; tau_max[i] = tau; } } } } } if (scale != -1) { for (int t = 0; t < nElements - scale*fps; ++t) { outputFile1 << t + scale*fps/2 << ", " << (to_string(files[a][5])).c_str() << (to_string(files[a][6])).c_str() << (to_string(files[a][7])).c_str() << ", " << (to_string(files[b][5])).c_str() << (to_string(files[b][6])).c_str() << (to_string(files[b][7])).c_str() << ", " << tau_max[t] << ", " << res_max[t] / (scale*fps) << "\n"; } } fclose(pFile_x1); fclose(pFile_x2); fclose(pFile_y1); fclose(pFile_y2); cudaFree(d_x1); cudaFree(d_y1); cudaFree(d_x2); cudaFree(d_y2); cudaFree(d_res); free(x1); free(y1); free(x2); free(y2); } } } // terminate fclose(pFile_x1); fclose(pFile_y1); fclose(pFile_x2); fclose(pFile_y2); return 0; }
1,247
#include "includes.h" /** * @brief cudaCreateBuffer Allocates a cuda buffer and stops the programm on error. * @param size * @return */ __global__ void kernelSetDoubleBuffer(float* gpuBuffPtr, float v, size_t size) { int index = threadIdx.x + blockIdx.x * blockDim.x; if (index < size) gpuBuffPtr[index] = v; }
1,248
#include "includes.h" __global__ void MatrixCopy_naive (const float * A , int Acount, int Acols, float * out0 , int out0count) { int id = blockDim.x*blockIdx.y*gridDim.x + blockDim.x*blockIdx.x + threadIdx.x; if (id<out0count) { out0[id] = A[id]; } }
1,249
// 3D convolution by CUDA __global__ void cu_conv(const float *A,const float *K,const float *B, int kw, int kh, int kn, int cw_rem, int ch_rem, float *C){ // A : input data, K : Kernel, B : bias int cx = threadIdx.x + blockIdx.x*blockDim.x; int cy = threadIdx.y + blockIdx.y*blockDim.y; int cz = blockIdx.z/int(kn*kh*kw); int n = (blockIdx.z%(kn*kh*kw)) / (kh*kw); int j = ((blockIdx.z%(kn*kh*kw)) % (kh*kw)) / kw; int i = ((blockIdx.z%(kn*kh*kw)) % (kh*kw)) % kw; int cw = blockDim.x*gridDim.x + cw_rem; int ch = blockDim.y*gridDim.y + ch_rem; int aw = cw + (kw-1); int ah = ch + (kh-1); int cidx = cx + cy*cw + cz*(cw*ch); int aidx = (cx+i) + (cy+j)*aw + (cz)*(aw*ah); int kidx = i + j*kw + n*(kw*kh); int bidx = n; if (cx < cw && cy < ch){ C[cidx] = A[aidx]*K[kidx] + B[bidx]/(kw*kh); } }
1,250
#include <iostream> using namespace std; int main(void) { cout << "Hello nvcc!" << endl; return 0; }
1,251
__global__ void print_values(const int* ints, const double* dbls, int* result) { int i = threadIdx.x; result[i] = ints[i] + (dbls[i] > 0.0); }
1,252
#include <stdio.h> #include <cuda_runtime_api.h> #include <cuda.h> #include <cstdlib> #include <ctime> #include <iostream> __global__ void matmul(float* matA, float* matB, float* matC, int width){ float pVal = 0; for(int i=0; i<width; ++i){ float elementMatA = matA[threadIdx.y*width+i]; float elementMatB = matB[i*width+threadIdx.x]; pVal += elementMatA * elementMatB; } matC[threadIdx.y*width+threadIdx.x] = pVal; } void matriksMul(float* mA, float* mB, float* mC, int width){ //Device pointer float* a_d, *b_d, *c_d; //Matriks size int size = width * width *sizeof(float) ; //allocate dan copy matriks a int cudaError = cudaMalloc((void**)&a_d, size); if (cudaError != cudaSuccess) { fprintf(stderr, "Error invoking cudaMemcpy (ERRCODE %d)\n", cudaError); } fprintf(stderr, "cudaMemcpy (ERRCODE %d)\n", cudaError); cudaMemcpy(a_d, mA, size , cudaMemcpyHostToDevice ); //allocate dan copy matriks b cudaMalloc((void**)&b_d, size); cudaMemcpy(b_d, mB, size , cudaMemcpyHostToDevice ); //allocate memory to device c cudaMalloc((void**)&c_d, size); dim3 dimGrid(1, 1); dim3 dimBlock(width, width); matmul<<<dimGrid,dimBlock>>>(a_d,b_d,c_d,width); cudaMemcpy(mC,c_d,size, cudaMemcpyDeviceToHost ); cudaFree(a_d); cudaFree(b_d); cudaFree(c_d); } int main(void){ void matriksMul(float *, float *, float *, int); const int width = 10; float* M, *N, *P; size_t size = width * width *sizeof(float); // allocate arrays on host M = (float *) malloc(size); N = (float *) malloc(size); P = (float *) malloc(size); // float M[width*width], N[width*width], P[width*width]; for(int i = 0; i < (width*width) ; i++) { M[i] = i; N[i] = width*width - i; P[i] = 0.f; // printf("%3f %3f %3f\n", M[i], N[i], P[i]); } matriksMul(M, N, P, width); for(int i = 0; i < (width*width) ; i++) { printf("%f", P[i]); if( i%width ==0){ printf("\n"); } } free(M); free(N); free(P); return 0; }
1,253
#include <stdio.h> #include <stdlib.h> #define NUM_BLOCKS 32 #define BLOCK_WIDTH 1 __global__ void hello() { printf("Hello world! I'm thread %d in block %d\n", threadIdx.x, blockIdx.x); } int main(int argc,char **argv) { int num_blocks = NUM_BLOCKS, block_width = BLOCK_WIDTH; if(argc>1){ num_blocks = atoi(argv[1]); block_width = atoi(argv[2]); } // launch the kernel hello<<<num_blocks, block_width>>>(); // force the printf()s to flush cudaDeviceSynchronize(); return 0; }
1,254
#include "includes.h" __global__ void add( double *a, double *b, double *c, int n ) { int tid = threadIdx.x + blockIdx.x * blockDim.x; // handle the data at this index while (tid < n) { c[tid] = a[tid] + b[tid]; tid += blockDim.x * gridDim.x; } //printf("Value of *ip variable: %f\n", a[tid] ); }
1,255
#include <stdio.h> #define BLOCK_SIZE 256 #define NUM_ELEMENTS (4096*100) // CUDA API error checking macro #define cudaCheck(error) \ if (error != cudaSuccess) { \ printf("Fatal error: %s at %s:%d\n", \ cudaGetErrorString(error), \ __FILE__, __LINE__); \ exit(1); \ } __global__ void reverse_1d(int *in, int *out) { __shared__ int temp[BLOCK_SIZE]; int gindex = threadIdx.x + (blockIdx.x * blockDim.x); int lindex = threadIdx.x; temp[BLOCK_SIZE - lindex - 1] = in[gindex]; __syncthreads(); out[BLOCK_SIZE * (gridDim.x - 1 - blockIdx.x) + threadIdx.x] = temp[lindex]; } __global__ void reverse_1Drection(int *in, int *out) { int gindex = threadIdx.x + (blockIdx.x * blockDim.x); out[NUM_ELEMENTS - gindex - 1] = in[gindex]; } int main() { unsigned int i; int h_in[NUM_ELEMENTS], h_out[NUM_ELEMENTS]; int *d_in, *d_out; // Initialize host data for( i = 0; i < (NUM_ELEMENTS); ++i ) h_in[i] = i; // Allocate space on the device cudaCheck( cudaMalloc( &d_in, (NUM_ELEMENTS) * sizeof(int)) ); cudaCheck( cudaMalloc( &d_out, NUM_ELEMENTS * sizeof(int)) ); // Copy input data to device cudaCheck( cudaMemcpy( d_in, h_in, (NUM_ELEMENTS) * sizeof(int), cudaMemcpyHostToDevice) ); /*cuda timing*/ cudaEvent_t start, stop; float time; cudaEventCreate(&start); cudaEventCreate(&stop); cudaEventRecord(start, 0); /* end start timing */ reverse_1d<<< (NUM_ELEMENTS + BLOCK_SIZE - 1)/BLOCK_SIZE, BLOCK_SIZE >>> (d_in, d_out); /*cuda timing stop*/ cudaEventRecord(stop, 0); cudaEventSynchronize(stop); cudaEventElapsedTime(&time, start, stop); cudaEventDestroy(start); cudaEventDestroy(stop); cudaCheck( cudaMemcpy( h_out, d_out, NUM_ELEMENTS * sizeof(int), cudaMemcpyDeviceToHost) ); // Verify every out value is 7 for( i = 0; i < NUM_ELEMENTS; ++i ) if (h_out[i] != h_in[NUM_ELEMENTS-i-1]) { printf("Element h_out[%d] == %d ERROR\n", i, h_out[i]); break; } if (i == NUM_ELEMENTS) printf("[SHARED] SUCCESS! %f\n", time); // Free out memory cudaFree(d_out); cudaCheck( cudaMalloc( &d_out, NUM_ELEMENTS * sizeof(int)) ); /*cuda timing*/ cudaEvent_t start1, stop1; cudaEventCreate(&start1); cudaEventCreate(&stop1); cudaEventRecord(start1, 0); /* end start timing */ reverse_1Drection<<< (NUM_ELEMENTS + BLOCK_SIZE - 1)/BLOCK_SIZE, BLOCK_SIZE >>> (d_in, d_out); /*cuda timing stop*/ cudaEventRecord(stop1, 0); cudaEventSynchronize(stop1); cudaEventElapsedTime(&time, start1, stop1); cudaEventDestroy(start1); cudaEventDestroy(stop1); cudaCheck( cudaMemcpy( h_out, d_out, NUM_ELEMENTS * sizeof(int), cudaMemcpyDeviceToHost) ); // Verify every out value is 7 for( i = 0; i < NUM_ELEMENTS; ++i ) if (h_out[i] != h_in[NUM_ELEMENTS-i-1]) { printf("Element h_out[%d] == %d ERROR\n", i, h_out[i]); break; } if (i == NUM_ELEMENTS) printf("[PRIVATE] SUCCESS! %f\n", time); // Free out memory cudaFree(d_in); cudaFree(d_out); return 0; }
1,256
/* File: vec_add.cu * Purpose: Implement vector addition on a gpu using cuda */ #include <stdio.h> #include <stdlib.h> #include <math.h> #include <iostream> #include <chrono> using namespace std; using namespace std::chrono; /* Kernel for vector addition */ __global__ void Vec_add(float x[], float y[], float z[], int n) { /* blockDim.x = threads_per_block */ /* First block gets first threads_per_block components. */ /* Second block gets next threads_per_block components, etc. */ int i = blockDim.x * blockIdx.x + threadIdx.x; /* block_count*threads_per_block may be >= n */ if (i < n) z[i] = x[i] + y[i]; } /* Vec_add */ /* Host code */ int main() { //create some vectors, run the test for each one and report the output. This will be good C++ and CUDA practice. int n, i; float *h_x, *h_y, *h_z; float *d_x, *d_y, *d_z; int threads_per_block; int block_count; size_t size; n = 10000000; // Number of elements in vector size = n*sizeof(float); // Vector size /* Allocate input vectors in host memory */ h_x = (float*) malloc(size); h_y = (float*) malloc(size); h_z = (float*) malloc(size); /* Initialize input vectors */ for (i = 0; i < n; i++) { h_x[i] = i+1; h_y[i] = n-i; } /* Allocate vectors in device memory */ // We use host pointers as a pointer to the on-device memory. cudaMalloc(&d_x, size); cudaMalloc(&d_y, size); cudaMalloc(&d_z, size); /* Copy vectors from host memory to device memory */ cudaMemcpy(d_x, h_x, size, cudaMemcpyHostToDevice); cudaMemcpy(d_y, h_y, size, cudaMemcpyHostToDevice); /* Define block size */ threads_per_block = 256; /* Define grid size. If we just computed n/threads_per_block */ /* we might get fewer threads than vector components. Using */ /* ceil(n/threads_per_block) guarantees at least one thread */ /* per vector component. The following formula is a kludge */ /* since it appears that the CUDA ceil function doesn't work */ /* correctly. */ block_count = (n + threads_per_block - 1)/threads_per_block; // just enough blocks that there is a thread per element /* Invoke kernel using block_count blocks, each of which */ /* contains threads_per_block threads */ Vec_add<<<block_count, threads_per_block>>>(d_x, d_y, d_z, n); cudaThreadSynchronize(); int numTests = 10000; high_resolution_clock::time_point t1 = high_resolution_clock::now(); for (int i =0; i<numTests; i++){ Vec_add<<<block_count, threads_per_block>>>(d_x, d_y, d_z, n); /* Wait for the kernel to complete */ } cudaThreadSynchronize(); high_resolution_clock::time_point t2 = high_resolution_clock::now(); auto duration = duration_cast<microseconds>( t2 - t1 ).count(); cout << (duration/1000) << " ms" << endl; float numSeconds = (float) duration / 1e6; long long numFlops = long(n) * long(numTests); float flopsPerSecond = (float)numFlops / numSeconds; float MflopsPerSecond = flopsPerSecond / 1e6; cout << "MFLOP/s = " << MflopsPerSecond << endl; /* Copy result from device memory to host memory */ /* h_z contains the result in host memory */ cudaMemcpy(h_z, d_z, size, cudaMemcpyDeviceToHost); float expectedSum = n + 1; printf("Testing...."); for (i = 0; i < n; i++){ if(h_z[i] != expectedSum){ printf("Failure at %i ", i); } } printf("\n"); /* Free device memory */ cudaFree(d_x); cudaFree(d_y); cudaFree(d_z); /* Free host memory */ free(h_x); free(h_y); free(h_z); return 0; } /* main */
1,257
#include <iostream> #include <cstdio> using namespace std; #include <cuda_runtime.h> #define TIMES 24 ////////////////////////////////////////////////////////////////////////////////////////////////////////////////// //////////////////////////////////////////////////HELP FUNCTIONS///////////////////////////////////////////////// void RandomInit(float* data, int n) { for (int i=0; i<n; i++) { data[i] = rand() / (float)RAND_MAX; } } void RandomInit(unsigned* data, int n) { for (int i=0; i<n; i++) { data[i] = rand() % n; } } #define checkCudaErrors(err) __checkCudaErrors (err, __FILE__, __LINE__) inline void __checkCudaErrors(cudaError err, const char *file, const int line ) { if(cudaSuccess != err) { fprintf(stderr, "%s(%i) : CUDA Runtime API error %d: %s.\n",file, line, (int)err, cudaGetErrorString( err ) ); exit(-1); } } // This will output the proper error string when calling cudaGetLastError #define getLastCudaError(msg) __getLastCudaError (msg, __FILE__, __LINE__) inline void __getLastCudaError(const char *errorMessage, const char *file, const int line ) { cudaError_t err = cudaGetLastError(); if (cudaSuccess != err) { fprintf(stderr, "%s(%i) : getLastCudaError() CUDA error : %s : (%d) %s.\n", file, line, errorMessage, (int)err, cudaGetErrorString( err ) ); exit(-1); } } //////////////////////////////////////////////////////////////////////////////////////////////////////////////////// /////////////////////////////////////////////_VECTOR_ADDITION_/////////////////////////////////////////////////////// // Device code __global__ void irreguler(const float* A, float* C, float* F) { int i = blockDim.x * blockIdx.x + threadIdx.x; if(i == 0) { C[i] = A[i]; __syncthreads(); //high latncy C[i] = C[i] + A[i]; __syncthreads(); F[i] = C[i+1]; } } __global__ void mb2(float* A, float* C) { int i = blockDim.x * blockIdx.x + threadIdx.x; if(i == 0) { C[i] = A[i]; //write to C[i] is a miss (cache line is missing) __syncthreads(); C[i+1] = A[i]; //write to C[i+1] is a hit (cache line is found) __syncthreads(); C[i] = C[i] + A[i]; //read of C[i] is a miss (entire sector is missing, fetch it from memory) __syncthreads(); A[i] = C[i] + C[i+1]; //read C[i] and C[i+1] are hits (entire sector exists) } } __global__ void l1(float* A, float* C) { int i = blockDim.x * blockIdx.x + threadIdx.x; if(i == 0) { C[i] = A[i] + A[i+32]; __syncthreads(); C[i] = A[i] + A[i+32] + C[i]; } } // Host code void VectorAddition(int N, int threadsPerBlock) { cout<<"Vector Addition for input size "<<N<<" :\n"; // Variables float* h_A; float* h_C; float* d_A; float* d_C; float total_time=0; size_t size = N * sizeof(float); // Allocate input vectors h_A and h_B in host memory h_A = (float*)malloc(size); h_C = (float*)malloc(size); // Initialize input vectors RandomInit(h_A, N); // Allocate vectors in device memory checkCudaErrors( cudaMalloc((void**)&d_A, size) ); checkCudaErrors( cudaMalloc((void**)&d_C, size) ); // Copy vectors from host memory to device memory checkCudaErrors( cudaMemcpy(d_A, h_A, size, cudaMemcpyHostToDevice) ); checkCudaErrors(cudaThreadSynchronize()); // Invoke kernel cout<<"Invoke Kernel\n"; //int threads = 128; int blocksPerGrid = ((N+ threadsPerBlock-1) / threadsPerBlock); for (int i = 0; i < 1; i++) { l1<<<blocksPerGrid, threadsPerBlock>>>(d_A, d_C); getLastCudaError("kernel launch failure"); checkCudaErrors(cudaThreadSynchronize()); } float dSeconds = total_time/((float)TIMES * 1000); float dNumOps = N; float gflops = 1.0e-9 * dNumOps/dSeconds; cout<<"Time = "<<dSeconds*1.0e3<< "msec"<<endl<<"gflops = "<<gflops<<endl; // Copy result from device memory to host memory // h_C contains the result in host memory checkCudaErrors( cudaMemcpy(h_C, d_C, size, cudaMemcpyDeviceToHost) ); // Verify result int i; for (i = 0; i < N; ++i) { float sum = h_A[i]; if (fabs(h_C[i] - sum) > 1e-5) break; } // Free device memory if (d_A) cudaFree(d_A); if (d_C) cudaFree(d_C); // Free host memory if (h_A) free(h_A); if (h_C) free(h_C); cudaDeviceReset(); if(i == N) cout<<"SUCCSESS"<<endl; else cout<<"FAILED"<<endl; } ////////////////////////////////////////////////////// int main(int argc,char *argv[]) { if(argc < 3) printf("Unsuffcient number of arguments!\n"); else { VectorAddition(atoi(argv[1]), atoi(argv[2])); } }
1,258
#include "includes.h" __global__ void stencil_1d(int *in, int *out) { // within a block, threads share data via shared memory ("global memory") // data is not visible to threads in other blocks // use __shared__ to declare a var/array in shared memory __shared__ int temp[BLOCK_SIZE + 2 * RADIUS]; // each thread processs one output element (blockDim.x elements per block) int gindex = threadIdx.x + (blockIdx.x * blockDim.x) + RADIUS; int lindex = threadIdx.x + RADIUS; // read input elements into shared memory temp[lindex] = in[gindex]; if (threadIdx.x < RADIUS) { temp[lindex - RADIUS] = in[gindex - RADIUS]; temp[lindex + BLOCK_SIZE] = in[gindex + BLOCK_SIZE]; } // synchronize all threads in the block : ensure all data is available __syncthreads(); // apply the stencil int result = 0; for (int offset = -RADIUS ; offset <= RADIUS ; offset++) { result += temp[lindex + offset]; } // store the result out[gindex-RADIUS] = result; }
1,259
#include <stdio.h> #include <stdlib.h> #include <stdbool.h> #include <limits> #include <sys/time.h> #define ARRAY_SIZE 10000 #define BLOCK_SIZE 256 #define MICROSECONDS(start, end) ((end.tv_sec - start.tv_sec) * 1000000LL + end.tv_usec - start.tv_usec) #define MILLISECONDS(start, end) MICROSECONDS(start, end) / 1000.0 #define SECONDS(start, end) MILLISECONDS(start, end) / 1000.0 void cpu_saxpy(const float *x, float *y, const float a) { for (unsigned int i = 0; i < ARRAY_SIZE; i++) { y[i] += a * x[i]; } } __global__ void gpu_saxpy(const float *x, float *y, const float a) { const unsigned int i = blockIdx.x * blockDim.x + threadIdx.x; if (i < ARRAY_SIZE) { y[i] += a * x[i]; } } double cpuSecond() { struct timeval tp; gettimeofday(&tp, NULL); return ((double)tp.tv_sec + (double)tp.tv_usec * 1.e-6); } int main(int argc, char **argv) { // Initialize data. struct timeval start, end; float *x = (float *)malloc(ARRAY_SIZE * sizeof(float)); float *y = (float *)malloc(ARRAY_SIZE * sizeof(float)); const float a = static_cast<float>(rand()) / static_cast<float>(RAND_MAX); for (unsigned int i = 0; i < ARRAY_SIZE; i++) { x[i] = static_cast<float>(rand()) / static_cast<float>(RAND_MAX); y[i] = static_cast<float>(rand()) / static_cast<float>(RAND_MAX); } // Copy data for CPU. float *cpu_x = (float *)malloc(ARRAY_SIZE * sizeof(float)); float *cpu_y = (float *)malloc(ARRAY_SIZE * sizeof(float)); memcpy(cpu_x, x, ARRAY_SIZE); memcpy(cpu_y, y, ARRAY_SIZE); // Run CPU SAXPY. printf("Computing SAXPY on the CPU... "); gettimeofday(&start, NULL); cpu_saxpy(cpu_x, cpu_y, a); gettimeofday(&end, NULL); printf("Done! Took %lfms.\n", MILLISECONDS(start, end)); // Copy data for GPU. float *gpu_x = (float *)malloc(ARRAY_SIZE * sizeof(float)); float *gpu_y = (float *)malloc(ARRAY_SIZE * sizeof(float)); memcpy(gpu_x, x, ARRAY_SIZE); memcpy(gpu_y, y, ARRAY_SIZE); float *cuda_x; float *cuda_y; cudaMalloc(&cuda_x, ARRAY_SIZE * sizeof(float)); cudaMalloc(&cuda_y, ARRAY_SIZE * sizeof(float)); cudaMemcpy(cuda_x, gpu_x, ARRAY_SIZE * sizeof(float), cudaMemcpyHostToDevice); cudaMemcpy(cuda_y, gpu_y, ARRAY_SIZE * sizeof(float), cudaMemcpyHostToDevice); // Run GPU SAXPY. printf("Computing SAXPY on the GPU... "); // Make sure that the grid size is enough to fit all elements. gettimeofday(&start, NULL); gpu_saxpy<<<(ARRAY_SIZE + BLOCK_SIZE - 1) / BLOCK_SIZE, BLOCK_SIZE>>>(cuda_x, cuda_y, a); cudaDeviceSynchronize(); gettimeofday(&end, NULL); printf("Done! Took %lfms.\n", MILLISECONDS(start, end)); cudaMemcpy(gpu_x, cuda_x, ARRAY_SIZE * sizeof(float), cudaMemcpyDeviceToHost); cudaMemcpy(gpu_y, cuda_y, ARRAY_SIZE * sizeof(float), cudaMemcpyDeviceToHost); // Compare CPU/GPU SAXPY. bool success = true; float epsilon = std::numeric_limits<float>::epsilon(); printf("Comparing the output for each implementation... "); for (unsigned int i = 0; i < 25; i++) { if (abs(cpu_y[i] - gpu_y[i]) > epsilon) { success = false; break; } } printf(success ? "Correct!\n" : "Incorrect!\n"); // Free resources. free(x); free(y); free(cpu_x); free(cpu_y); free(gpu_x); free(gpu_y); cudaFree(x); cudaFree(y); return 0; }
1,260
int main() { const unsigned int N = 1048576; const unsigned int bytes = N * sizeof(int); int *h_a = (int*)malloc(bytes); int *d_a; cudaMalloc((int**)&d_a, bytes); memset(h_a, 0, bytes); cudaMemcpy(d_a, h_a, bytes, cudaMemcpyHostToDevice); cudaMemcpy(h_a, d_a, bytes, cudaMemcpyDeviceToHost); return 0; } /* Run these next commands to run the program: $ nvcc profile.cu -o profile_test $ nvprof ./profile_test */
1,261
#include "Logger.cuh" #include <stdlib.h> #include <string.h> #include <stdarg.h> #include <stdio.h> #include <time.h> #define PREFIX_SIZE 256 // Do not change order. It matches the values of LOGGER_LEVEL_XXXXX static char* level_name_by_level[] = { "ERROR", "WARN", "INFO", "DEBUG" }; static void logger_print(const char *log_name, unsigned int level, unsigned int level_message, char* fmt, va_list args) { if (level < level_message) return; // Create a "prefixed" format string, by appending the log level and the log name char* prefixed_fmt = (char*) malloc(PREFIX_SIZE + (strlen(log_name) + 1) + (strlen(fmt) + 1)); char* log_level = level_name_by_level[level_message]; time_t t = time(NULL); struct tm tm = *localtime(&t); sprintf( prefixed_fmt, "%02d/%02d/%04d %02d:%02d:%02d - [%5s] - %s - %s \n", tm.tm_mday, // day tm.tm_mon + 1, // month tm.tm_year + 1900, // year tm.tm_hour, tm.tm_min, tm.tm_sec, // hour:minutes:seconds log_level, log_name, fmt ); vprintf(prefixed_fmt, args); free(prefixed_fmt); } Logger::Logger(char* name, unsigned int level) { unsigned int name_size = strlen(name) + 1; this->name = (char *) malloc(name_size * sizeof(char)); strcpy(this->name, name); this->level = level; } void Logger::info(char* fmt, ...) { va_list args; va_start(args, fmt); logger_print(this->name, this->level, LOGGER_LEVEL_INFO, fmt, args); va_end(args); } void Logger::warn(char* fmt, ...) { va_list args; va_start(args, fmt); logger_print(this->name, this->level, LOGGER_LEVEL_WARN, fmt, args); va_end(args); } void Logger::error(char* fmt, ...) { va_list args; va_start(args, fmt); logger_print(this->name, this->level, LOGGER_LEVEL_ERROR, fmt, args); va_end(args); } void Logger::debug(char* fmt, ...) { va_list args; va_start(args, fmt); logger_print(this->name, this->level, LOGGER_LEVEL_DEBUG, fmt, args); va_end(args); } Logger::~Logger() { free(this->name); }
1,262
#include <stdlib.h> #include <stdio.h> #include <time.h> //#include "cutil.h" void checkCUDAError(const char *msg) { cudaError_t err = cudaGetLastError(); if (cudaSuccess != err) { printf("Cuda error: %s: %s.\n", msg, cudaGetErrorString(err)); printf("\nPress ENTER to exit...\n"); getchar(); exit(-1); } } //name of the input file #define INPUT_FILE_NAME "input.txt" //name of the compressed file #define COMPRESSED_FILE_NAME "compressed.txt" #define COMPRESSED_FILE_NAME_GPU "compressed_gpu.txt" //name of the uncompressed file #define DECOMPRESSED_FILE_NAME "decompressed.txt" //name of the config file #define CONFIG_FILE_NAME "config.txt" //max number of characters #define MAX_CHAR 256 //#define MAX_CHAR 30 //max lenght of the number which can occur in char_frequency or char_huffman_table) #define MAX_LENGTH_OF_NUMBER 10 //lenght of the array in shared memory on device #define SHARED_MEMORY_SIZE 256 //lenght of the array in const memory on device #define CONST_MEMORY_SIZE 15000 //(MAX_CHAR*(MAX_CHAR-1)) //To fill and pass the file as an array to GPU #define MAX_FILE_CHARS 50000 #define BLOCK_SIZE 256 struct node { int val; int weight; struct node * right, * left; }; //keeps frequency of particular characters (index - symbof of the character, value - frequency of the character) int char_frequency[MAX_CHAR]; //keeps huffman table int char_huffman_table[MAX_CHAR][MAX_CHAR-1]; //keeps number which tells how many bits were unused in last byte (variable is set after call compress_file()) int last_byte_padding=0; //for writing gpu output int last_byte_padding_gpu = 0; //keeps number of characters in current input file - file has to have less than 2,147,483,647 characters (variable is set after call read_file()) int number_of_char=0 ; //To fill and pass the file as an array to GPU unsigned char *h_input=0,*d_input=0; // To read char_huffman_table at the GPU int *d_char_huffman_table=0; int copiedarray2[MAX_CHAR][MAX_CHAR-1]; __device__ int char_huffman_table_gpu[MAX_CHAR][MAX_CHAR-1]; //To write the output from compression in GPU //char *compressedfile_array=0; bool *compressedfile_array=0; bool *finalcompressed_array=0; // To keep track of how many characters each block wrote int *block_cntr_array=0; int *block_cntr_array_check=0; int *d_last_byte_padding=0; int *finalsize=0; int *orig_number_of_char=0; int *huffman_check = (int *)malloc((MAX_CHAR)*(MAX_CHAR-1) *sizeof(int)); bool *d_bool = 0; bool *h_bool = 0; __global__ void final_compression(int *block_cntr_array,bool *compressedfile_array,bool *finalcompressed_array,int number_of_char) //__device__ void final_compression(int *block_cntr_array,bool *compressedfile_array,bool *finalcompressed_array) { int index_blocks=blockIdx.x*blockDim.x+threadIdx.x; int index_file=(blockIdx.x*blockDim.x+threadIdx.x)*255; int final_index=0; if(index_blocks < number_of_char) { for(int i=0;i<index_blocks;i++) { final_index = final_index+ block_cntr_array[i]; } for(int i=0;i<block_cntr_array[index_blocks];i++) { finalcompressed_array[final_index+i]=compressedfile_array[index_file+i]; } } } //__global__ void computearray_size(int* block_cntr_array,int *finalsize,int *orig_number_of_char) __device__ void computearray_size(int* block_cntr_array,int *finalsize,int *orig_number_of_char) { *finalsize = 0; for(int i=0;i<*orig_number_of_char;i++) { (*finalsize)=(*finalsize) + block_cntr_array[i]; } } /*__global__ void compress_file_gpu(unsigned char *d_input,char *compressedfile_array,int *char_huffman_table2,int *block_cntr_array,int* d_last_byte_padding) { int write_counter=0,block_counter=0; //how many bits have been written in specific byte unsigned char input_char; unsigned char output_char = 0x0; unsigned char end_of_file = 255; unsigned char mask = 0x01; //00000001; int index_file=(blockIdx.x*blockDim.x+threadIdx.x)*255; int index_blocks=blockIdx.x*blockDim.x+threadIdx.x; //for(int i=0;i<MAX_CHAR;i++) //{ //int *row = (int*)((char*)char_huffman_table2 + i * pitch); //for (int c = 0; c < MAX_CHAR-1; ++c) { // char_huffman_table_gpu[i][c] = row[c]; //} //} input_char = d_input[index_blocks]; for(int i = 0 ; i < (MAX_CHAR - 1) ; i++) { if(char_huffman_table2[input_char*255+i] == 0) //detect if current character on particular position has 0 or 1 { output_char = output_char << 1; //if 0 then shift bits one position to left (last bit after shifting is 0) write_counter++; block_counter++; } else if(char_huffman_table2[input_char*255+i] == 1) { output_char = output_char << 1; //if 1 then shift bits one position to left... output_char = output_char | mask; //...and last bit change to: 1 write_counter++; block_counter++; } else //-1 { //if(input_char == end_of_file) //if EOF is detected then write current result to file //{ if(write_counter != 0) { output_char = output_char << (8-write_counter); compressedfile_array[index_file]=output_char; output_char = 0x0; } else //write_counter == 0 { compressedfile_array[index_file]=output_char; } //} break; } if(write_counter == 8) //if result achieved 8 (size of char) then write it to compressed_file { compressedfile_array[index_file]=output_char; output_char = 0x0; write_counter = 0; } } block_cntr_array[index_blocks]=block_counter; *d_last_byte_padding = write_counter; //to decompress file we have to know how many bits in last byte have been written //update_config(write_counter); //TODO to zakomentowac przy ostatecznych pomiarach }*/ //__global__ void compress_file_gpu(unsigned char *d_input,bool *compressedfile_array,int *char_huffman_table2,int *block_cntr_array,int* d_last_byte_padding) __global__ void compress_file_gpu(unsigned char *d_input,bool *compressedfile_array,int *char_huffman_table2,int *block_cntr_array,int* d_last_byte_padding,int *finalsize,int *orig_number_of_char,int number_of_char) { //int write_counter=0, int block_counter=0; //how many bits have been written in specific byte unsigned char input_char; //unsigned char output_char = 0x0; //unsigned char end_of_file = 255; //unsigned char mask = 0x01; //00000001; int index_file=(blockIdx.x*blockDim.x+threadIdx.x)*255; int index_blocks=blockIdx.x*blockDim.x+threadIdx.x; if(index_blocks < number_of_char) { //for(int i=0;i<MAX_CHAR;i++) //{ //int *row = (int*)((char*)char_huffman_table2 + i * pitch); //for (int c = 0; c < MAX_CHAR-1; ++c) { // char_huffman_table_gpu[i][c] = row[c]; //} //} input_char = d_input[index_blocks]; for(int i = 0 ; i < (MAX_CHAR - 1) ; i++) { if(char_huffman_table2[input_char*255+i] == 0) //detect if current character on particular position has 0 or 1 { //output_char = output_char << 1; //if 0 then shift bits one position to left (last bit after shifting is 0) compressedfile_array[index_file+i] = false; //write_counter++; block_counter++; } else if(char_huffman_table2[input_char*255+i] == 1) { //output_char = output_char << 1; //if 1 then shift bits one position to left... //output_char = output_char | mask; //...and last bit change to: 1 //write_counter++; compressedfile_array[index_file+i] = true; block_counter++; } else //-1 { /*if(input_char == end_of_file) //if EOF is detected then write current result to file { if(write_counter != 0) { output_char = output_char << (8-write_counter); compressedfile_array[index_file]=output_char; output_char = 0x0; } else //write_counter == 0 { compressedfile_array[index_file]=output_char; } }*/ break; } /*if(write_counter == 8) //if result achieved 8 (size of char) then write it to compressed_file { compressedfile_array[index_file]=output_char; output_char = 0x0; write_counter = 0; }*/ } block_cntr_array[index_blocks]=block_counter; //*d_last_byte_padding = write_counter; //to decompress file we have to know how many bits in last byte have been written //update_config(write_counter); //TODO to zakomentowac przy ostatecznych pomiarach computearray_size(block_cntr_array,finalsize,orig_number_of_char); //final_compression(block_cntr_array,compressedfile_array,finalcompressed_array); } } void write_GPU_compressed(bool *final_compressed_cpu,int *finalsize_cpu) { FILE *compressed_file; int write_counter=0; //how many bits have been written in specific byte //unsigned char input_char; unsigned char output_char = 0x0; //unsigned char end_of_file = 255; unsigned char mask = 0x01; //00000001; compressed_file = fopen(COMPRESSED_FILE_NAME_GPU, "wb"); if ((compressed_file==NULL)) { perror ("Error reading file"); } else { for(int i = 0 ; i < (*finalsize_cpu) ; i++) { if(int(final_compressed_cpu[i]) == 0) { output_char = output_char << 1; //if 0 then shift bits one position to left (last bit after shifting is 0) write_counter++; } else if (int(final_compressed_cpu[i]) == 1) { output_char = output_char << 1; //if 1 then shift bits one position to left... output_char = output_char | mask; //...and last bit change to: 1 write_counter++; } if(write_counter == 8) //if result achieved 8 (size of char) then write it to compressed_file { printf("Compressed char in decimal is %d \n", output_char); putc(output_char, compressed_file); output_char = 0x0; write_counter = 0; } } if(write_counter != 0) { output_char = output_char << (8-write_counter); printf("Compressed char in decimal is %d \n", output_char); putc(output_char, compressed_file); output_char = 0x0; } } fclose(compressed_file); last_byte_padding_gpu = write_counter; } void print_dchar_huffman_table() { printf("\n dchar huffman table "); getchar(); bool flag = false; printf("+++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++\n"); printf("Huffman table:\n"); printf("+++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++\n"); for(int i = 0 ; i < MAX_CHAR ; i++) { flag = false; for(int j = 0 ; j < (MAX_CHAR -1) ; j++) { if(copiedarray2[i][j] != -1) { if(!flag) { if(i == 10)//new line { printf("\\n:\t"); } else { printf("%c:\t",i); } } flag = true; printf("%d ", copiedarray2[i][j]); } } if(flag) printf("\n"); } printf("+++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++\n"); getchar(); } //Huffman table construction+++++++++++++++++++++++++++++++++++++++++++++++++++ void insertion_sort(node **forest, int length) { for(int i = 1; i < length ; i++) { node *tmp = forest[i]; int j = i - 1; bool done = false; do { if(forest[j]->weight < tmp->weight) //> ascending order; < descending order { forest[j+1] = forest[j]; j = j-1; if(j < 0) { done = true; } } else { done = true; } }while(!done); forest[j+1] = tmp; } } void print_char_frequency() { printf("+++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++\n"); printf("character frequency:\n"); printf("+++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++\n"); for(int i = 0 ; i < MAX_CHAR ; i++) { if(char_frequency[i] != 0) { if(i == 10)//new line { printf("%d)\tval: \\n\tfreq: %d\n",i, char_frequency[i]); } else { printf("%d)\tval: %c\tfreq: %d\n",i, i, char_frequency[i]); } } } printf("+++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++\n"); } void print_char_huffman_table() { bool flag = false; printf("+++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++\n"); printf("Huffman table:\n"); printf("+++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++\n"); for(int i = 0 ; i < MAX_CHAR ; i++) { flag = false; for(int j = 0 ; j < (MAX_CHAR -1) ; j++) { if(char_huffman_table[i][j] != -1) { if(!flag) { if(i == 10)//new line { printf("\\n:\t"); } else { printf("%c:\t",i); } } flag = true; printf("%d ", char_huffman_table[i][j]); } } if(flag) printf("\n"); } printf("+++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++\n"); } void printout_inorder(node * tree) { if(tree->left) printout_inorder(tree->left); if(tree->val != NULL) { if(tree->val == '\n') { printf("weight: %d\tvalue: \\n\n",tree->weight); } else { printf("weight: %d\tvalue: %c\n",tree->weight, tree->val); } } else { printf("weight: %d\tvalue: NULL\n",tree->weight); } if(tree->right) printout_inorder(tree->right); } void read_file() { FILE *file; unsigned char end_of_file = 255; unsigned char c; file = fopen(INPUT_FILE_NAME, "r"); if (file==NULL) { perror ("Error reading file"); } else { //storing the file contents into h_input h_input = (unsigned char *)malloc(MAX_FILE_CHARS*sizeof(char)); do { c = getc (file); //if(c == end_of_file) printf("\n Found EOF \n"); //printf("c before putting into array is %c\n",c); h_input[number_of_char]=c; number_of_char++; char_frequency[c]++; } while (c != end_of_file); fclose (file); } // h_input[number_of_char] = end_of_file; char_frequency[end_of_file] = 0; //to avoid problems with several EOF in one file //EOF is not needed ; so going to decrement number_of_char--; } void traverse_preorder(node *root, int *path) { if(root->val != NULL) { for(int i = 0 ; i < MAX_CHAR -1 ; i++) { char_huffman_table[root->val][i] = path[i]; } } if(root->left)//left 1 { int counter = 0; for(int i = 0 ; i < MAX_CHAR - 1 ; i++) { if(path[i] == -1) { break; } counter++; } path[counter] = 1; traverse_preorder(root->left, path); path[counter] = -1; } if(root->right)//right 0 { int counter = 0; for(int i = 0 ; i < MAX_CHAR - 1 ; i++) { if(path[i] == -1) { break; } counter++; } path[counter] = 0; traverse_preorder(root->right, path); path[counter] = -1; } } void construct_huffman_table(node *root) { int path[MAX_CHAR - 1]; for(int i = 0 ; i < MAX_CHAR - 1 ; i++) { path[i] = -1; } traverse_preorder(root, path); } void build_binary_tree() { int forest_counter = 0; node *forest[MAX_CHAR]; node *curr; for(int i = 0 ; i < MAX_CHAR ; i++) //initial forest { if(char_frequency[i] != 0) { curr = (node *)malloc(sizeof(node)); curr->left = curr->right = NULL; curr->val = i; curr->weight = char_frequency[i]; forest[forest_counter] = curr; forest_counter++; } } insertion_sort(forest, forest_counter);//sorted initial forest while(forest_counter > 1)//build final tree { node *parent; parent = (node *)malloc(sizeof(node)); parent->right = forest[forest_counter-1]; parent->left = forest[forest_counter-2]; parent->weight = forest[forest_counter-1]->weight + forest[forest_counter-2]->weight; parent->val = NULL; forest[forest_counter-1] = NULL; forest[forest_counter-2] = parent; forest_counter--; insertion_sort(forest, forest_counter); } printf("+++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++\n"); printf("Huffman tree (inorder traversal sequence):\n"); printf("+++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++\n"); printout_inorder(forest[0]); printf("+++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++\n"); construct_huffman_table(forest[0]); //delete_binary_tree_postorder(forest[0]); //after building Huffman table we do not need Huffman tree anymore } void array_initializer() { for(int i = 0 ; i < MAX_CHAR ; i++) { char_frequency[i] = 0; } for(int i = 0 ; i < MAX_CHAR ; i++) { for(int j = 0 ; j < (MAX_CHAR-1) ; j++) { char_huffman_table[i][j] = -1; } } } //+++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++ //Calculation on CPU+++++++++++++++++++++++++++++++++++++++++++++++++++++++++++ void compress_file() { FILE *input_file; FILE *compressed_file; unsigned char output_char; int write_counter; //how many bits have been written in specific byte unsigned char input_char; unsigned char end_of_file ; unsigned char mask ; //00000001; input_file = fopen(INPUT_FILE_NAME, "rb"); compressed_file = fopen(COMPRESSED_FILE_NAME, "wb"); // apend file (add text to a file or create a file if it does not exist) output_char= 0x0; write_counter = 0; end_of_file = 255; mask = 0x01; if ((input_file==NULL)||(compressed_file==NULL)) { perror ("Error reading file"); } else { do { input_char = getc (input_file); //read one character from input file for(int i = 0 ; i < (MAX_CHAR - 1) ; i++) { if(char_huffman_table[input_char][i] == 0) //detect if current character on particular position has 0 or 1 { output_char = output_char << 1; //if 0 then shift bits one position to left (last bit after shifting is 0) write_counter++; } else if(char_huffman_table[input_char][i] == 1) { output_char = output_char << 1; //if 1 then shift bits one position to left... output_char = output_char | mask; //...and last bit change to: 1 write_counter++; } else //-1 { if(input_char == end_of_file) //if EOF is detected then write current result to file { if(write_counter != 0) { output_char = output_char << (8-write_counter); printf("Compressed char in decimal is %d \n", output_char); putc(output_char, compressed_file); output_char = 0x0; } else //write_counter == 0 { printf("Compressed char in decimal is %d \n", output_char); putc(output_char, compressed_file); } } break; } if(write_counter == 8) //if result achieved 8 (size of char) then write it to compressed_file { printf("Compressed char in decimal is %d \n", output_char); putc(output_char, compressed_file); output_char = 0x0; write_counter = 0; } } } while (input_char != end_of_file); fclose (input_file); fclose(compressed_file); last_byte_padding = write_counter; //to decompress file we have to know how many bits in last byte have been written //update_config(write_counter); //TODO to zakomentowac przy ostatecznych pomiarach } } void print_gpu_compressed_file(char *final_compressed_cpu,int finalsize_cpu) { FILE *compressed_file; compressed_file = fopen(COMPRESSED_FILE_NAME_GPU, "wb"); for(int i=0;i<finalsize_cpu;i++) { char c=final_compressed_cpu[i]; printf("i is %d and c is %c",i,c); putc(c,compressed_file); } fclose(compressed_file); } void decompress_file() { FILE *compressed_file; FILE *decompressed_file; unsigned char end_of_file = 255; unsigned char mask = 0x7F; //01111111; unsigned char curr; unsigned char next; int written_char_counter=0; int pattern[MAX_CHAR - 1]; for(int i = 0 ; i < (MAX_CHAR - 1); i++) { pattern[i] = -1; } compressed_file = fopen(COMPRESSED_FILE_NAME, "rb"); decompressed_file = fopen(DECOMPRESSED_FILE_NAME, "wb"); if ((compressed_file==NULL)||(decompressed_file==NULL)) { perror ("Error reading file"); } else { int bit_counter=0; unsigned char first_bit; bool read_next = true; curr = getc (compressed_file); next = getc (compressed_file); //we have to read one byte in advance due to padding // for(int i = 0 ; i < (MAX_CHAR - 1) ; i++) //builds a pattern and chcecks if it matches to char_huffman_table int pattern_counter=-1; while(pattern_counter < (MAX_CHAR - 1)) { pattern_counter++; first_bit = curr | mask; //check if first bit is 0 or 1 curr = curr << 1; if(bit_counter == 7) { bit_counter = 0; curr = next; if(read_next) { next = getc (compressed_file); if(next == end_of_file) { if((number_of_char - written_char_counter) < 8) { read_next = false; bit_counter = 7 - last_byte_padding; } } } if((curr == end_of_file) && ((number_of_char - written_char_counter) < 8)) { break; } } else { bit_counter++; } if(first_bit == 255) { pattern[pattern_counter] = 1; } else { pattern[pattern_counter] = 0; } bool flag = true; for(int j = 0 ; j < MAX_CHAR ; j++) { flag = true; for(int k = 0 ; k < (MAX_CHAR - 1) ; k++) { if(char_huffman_table[j][k] != pattern[k]) { flag = false; break; } } if(flag == true) { written_char_counter++; putc(j, decompressed_file); for(int i = 0 ; i < (MAX_CHAR - 1); i++) { pattern[i] = -1; } pattern_counter = -1; break; } } } fclose (compressed_file); fclose (decompressed_file); } } void initialize() { array_initializer(); read_file(); print_char_frequency(); build_binary_tree(); print_char_huffman_table(); } /*__global__ void compress(int *d_input,int number_of_char,int *d_char_huffman_table,int MAX_CHAR) { int i=0; extern __shared__ int my2DArray[32][32]; //size need to be coded a development time though my2DArray[threadIdx.x][threadIdx.y] = flatArray[blockDim.x * threadIdx.y + threadIdx.x]; }*/ __global__ void read2darray(int *devPtr,int pitch) { int elements[2][2]; for (int r = 0; r < 2; ++r) { int* row = (int*)((char*)devPtr + r * pitch); for (int c = 0; c < 2; ++c) { elements[r][c] = row[c]; } } } __global__ void check_bool(bool *d_bool) { d_bool[0]=false; d_bool[1]=false; } void print_huffman() { printf(" \n Huffman after copying back \n " ); unsigned char input_char1; unsigned char input_char2; unsigned char input_char3; unsigned char input_char4; input_char1 = h_input[0]; input_char2 = h_input[1]; input_char3 = h_input[2]; input_char4 = h_input[3]; for (int i=0;i < MAX_CHAR-1;i++) { if(huffman_check[ input_char1*255+i]!= -1 ) printf ("\t%c code is %d \n", input_char1,huffman_check[ input_char1*255+i]); } for (int i=0;i < MAX_CHAR-1;i++) { if(huffman_check[ input_char2*255+i]!= -1) printf ("\t%c code is %d \n",input_char2, huffman_check[ input_char2*255+i]); } for (int i=0;i < MAX_CHAR-1;i++) { if(huffman_check[ input_char3*255+i]!= -1 ) printf ("\t%c code is %d \n", input_char3,huffman_check[ input_char3*255+i]); } for (int i=0;i < MAX_CHAR-1;i++) { if(huffman_check[ input_char4*255+i]!= -1 ) printf ("\t%c code is %d \n", input_char4,huffman_check[ input_char4*255+i]); } } int main(int argc, char* argv[]) { int *finalsize_cpu=0; unsigned char end_of_file = 255; printf("start\n"); initialize(); cudaEvent_t start, stop; // cuda events to measure time float elapsed_time,elapsed_time_Cont; cudaEventCreate(&start); // timing objects cudaEventCreate(&stop); unsigned int timer2=0; time_t seconds; // In initialize in cpu,we put the file chars into array, fill huffman table and char_freq_arrays //copy the input contents into an array cudaMalloc((void **)&d_input,number_of_char*sizeof(char)); checkCUDAError("Error in allocating d_input"); cudaMemcpy(d_input,h_input,number_of_char*sizeof(char),cudaMemcpyHostToDevice); checkCUDAError("Error in copying d_input"); // Allocate space for the compressed file to be used in GPU cudaMalloc((void **)&compressedfile_array,number_of_char*(MAX_CHAR -1)*sizeof(bool)); checkCUDAError("Error in allocating compressedfile_array"); // cudaMalloc((void **)&d_char_huffman_table,(MAX_CHAR)*(MAX_CHAR-1) * sizeof(int)); checkCUDAError("Error in allocating d_char_huffman_table"); cudaMemcpy(d_char_huffman_table,char_huffman_table,(MAX_CHAR)*(MAX_CHAR-1) * sizeof(int),cudaMemcpyHostToDevice); checkCUDAError("Error in copying d_char_huffman_table"); cudaMemcpy(huffman_check,d_char_huffman_table,(MAX_CHAR)*(MAX_CHAR-1) * sizeof(int),cudaMemcpyDeviceToHost); checkCUDAError("Error in copying back"); cudaMalloc((void **)&block_cntr_array,number_of_char*sizeof(int)); checkCUDAError("Error in allocating block_cntr_array"); cudaMalloc((void **)&d_last_byte_padding,sizeof(int)); checkCUDAError("Error in allocating d_last_byte_padding"); cudaMalloc((void **)&finalsize,sizeof(int)); checkCUDAError("Error in allocating finalsize"); cudaMalloc((void **)&orig_number_of_char,sizeof(int)); checkCUDAError("Error in allocating orig_number_of_char"); cudaMemcpy(orig_number_of_char,&number_of_char,sizeof(int),cudaMemcpyHostToDevice); checkCUDAError("Error in copying orig_number_of_char"); // check if i can make a boolean array h_bool=(bool *) malloc(2*sizeof(bool)); h_bool[0]=true; h_bool[1]=true; printf("bool1 is %d and bool2 is %d \n",h_bool[0],h_bool[1]); cudaMalloc((void **)&d_bool,2*sizeof(bool)); checkCUDAError("Error in d_bool"); cudaMemcpy(d_bool,h_bool,2*sizeof(bool),cudaMemcpyHostToDevice); checkCUDAError("Error in copying d_bool"); //check_bool<<<1,1>>>(d_bool); checkCUDAError("Error in kernel changing d_bool"); cudaThreadSynchronize(); checkCUDAError("Error in cudaThreadSynchronize"); cudaMemcpy(h_bool,d_bool,2*sizeof(bool),cudaMemcpyDeviceToHost); checkCUDAError("Error in copying d_bool back"); printf("Now bool1 is %d and bool2 is %d \n",h_bool[0],h_bool[1]); int checkhuff[2][3]= { {0, 0, 0}, {1, 1, 1} }; bool flag = true; for(int j=0;j<2;j++) { flag = true; for(int k=0;k<2;k++) { printf("h_bool is %d \t checkhuff is %d \n",int(h_bool[k]),checkhuff[j][k]); if(checkhuff[j][k] != int(h_bool[k])) { flag = false; break; } } if(flag == true) { printf("pattern for %d is found\n",checkhuff[j][0]); } } //copy and send the huffman table as a 2d array to GPU Device //int *darray=0; //size_t pitch; //cudaMallocPitch( (void**)&darray, &pitch, 2 * sizeof(int), 2); //cudaMemcpy2D(darray,pitch,harray,2*sizeof(int),2*sizeof(int),2,cudaMemcpyHostToDevice); //cudaMalloc((void **)&darray,4*sizeof(int)); //cudaMemcpy(darray,harray,4*sizeof(int),cudaMemcpyHostToDevice); //cudaMemcpy2D(copiedarray,2*sizeof(int),darray,pitch,pitch,2,cudaMemcpyDeviceToHost); // cudaMemcpy2D(copiedarray,2*sizeof(int),darray,pitch,2*sizeof(int),2,cudaMemcpyDeviceToHost); //printf("After copying back %d, \t %d, \t %d, \t %d \n",copiedarray[0][0],copiedarray[0][1],copiedarray[1][0],copiedarray[1][1]); //int *darray_2d=0; //cudaMalloc((void **)&darray_2d,4*sizeof(int)); //cudaMemcpy(darray,harray,4*sizeof(int),cudaMemcpyHostToDevice); // read2darray<<<1,1>>>(darray, pitch); //size_t pitch2; //cudaMallocPitch( (void**)&d_char_huffman_table, &pitch2, (MAX_CHAR-1) * sizeof(int), MAX_CHAR); //cudaMemcpy2D(d_char_huffman_table,pitch2,char_huffman_table,(MAX_CHAR-1) * sizeof(int),(MAX_CHAR-1) * sizeof(int),MAX_CHAR,cudaMemcpyHostToDevice); //cudaMemcpy2D(char_huffman_table_gpu,(MAX_CHAR-1) * sizeof(int),char_huffman_table,(MAX_CHAR-1) * sizeof(int),(MAX_CHAR-1) * sizeof(int),MAX_CHAR,cudaMemcpyHostToDevice); //checkCUDAError("Error in char_huffman_table_gpu"); //cudaMemcpy2D(copiedarray2,(MAX_CHAR-1)*sizeof(int),d_char_huffman_table,pitch2,(MAX_CHAR-1)*sizeof(int),MAX_CHAR,cudaMemcpyDeviceToHost); //cudaMemcpy2D(copiedarray2,(MAX_CHAR-1)*sizeof(int),char_huffman_table_gpu,(MAX_CHAR-1) * sizeof(int),(MAX_CHAR-1) * sizeof(int),MAX_CHAR,cudaMemcpyDeviceToHost); //cudaMemcpy(orig_number_of_char,&number_of_char,sizeof(int),cudaMemcpyHostToDevice); // checkCUDAError("Error in copiedarray2"); // print_dchar_huffman_table(); printf("\n the number of characters in the input file is %d \n",number_of_char); getchar(); /*for(int i=0;i<number_of_char;i++) { if( h_input[i] == end_of_file ) printf(" EOF \n"); printf(" Copying into array: i is %d and c is %c \n",i,h_input[i]); }*/ getchar(); print_huffman(); int no_of_blocks = (number_of_char + BLOCK_SIZE -1)/BLOCK_SIZE; printf("no_of_blocksis %d \n", no_of_blocks); if(no_of_blocks == 0) no_of_blocks =1; //compress_file_gpu<<<number_of_char,1>>>(d_input,compressedfile_array,d_char_huffman_table,block_cntr_array,d_last_byte_padding); cudaEventRecord(start, 0); // start time checkCUDAError("Error in cudaEventRecord start \n"); compress_file_gpu<<<no_of_blocks,BLOCK_SIZE>>>(d_input,compressedfile_array,d_char_huffman_table,block_cntr_array,d_last_byte_padding,finalsize,orig_number_of_char,number_of_char); checkCUDAError("Error in compress_file_gpu \n"); cudaThreadSynchronize(); //cudaMalloc((void **)&block_cntr_array_check,number_of_char*sizeof(int)); //checkCUDAError("Error in allocating block_cntr_array_check"); block_cntr_array_check = (int *) malloc(number_of_char*sizeof(int)); cudaMemcpy(block_cntr_array_check,block_cntr_array,number_of_char*sizeof(int),cudaMemcpyDeviceToHost); checkCUDAError("Error in copying back block_cntr_array_check"); for(int i=0; i < number_of_char; i++) { printf(" block size for i = %d is %d \n",i, block_cntr_array_check[i]); } // computearray_size<<<1,1>>>(block_cntr_array,finalsize,orig_number_of_char); checkCUDAError("Error in Compute array \n"); finalsize_cpu = (int *)malloc(sizeof(int)); cudaMemcpy(finalsize_cpu,finalsize,sizeof(int),cudaMemcpyDeviceToHost); printf("The final compressed array size is %d \n ", *finalsize_cpu); checkCUDAError("Error in finalsize_cpu"); int block = *finalsize_cpu; //allocate space for the final compressed array cudaMalloc((void **)&finalcompressed_array,((*finalsize_cpu)*sizeof(bool))); checkCUDAError("cudaMemcpyHostToDevice"); final_compression<<<no_of_blocks,BLOCK_SIZE>>>(block_cntr_array,compressedfile_array,finalcompressed_array,number_of_char); checkCUDAError("Error in final_compression call \n"); cudaThreadSynchronize(); checkCUDAError("Error in cudaThreadSynchronize \n"); cudaEventRecord(stop, 0); checkCUDAError("Error in cudaEventRecord stop \n"); cudaEventSynchronize(stop); cudaEventElapsedTime(&elapsed_time, start, stop); printf("Time to calculate results: %f ms.\n", elapsed_time); // print out execution time bool *final_compressed_cpu=0; final_compressed_cpu = (bool *)malloc((*finalsize_cpu)*sizeof(bool)); cudaMemcpy(final_compressed_cpu,finalcompressed_array,((*finalsize_cpu)*sizeof(bool)),cudaMemcpyDeviceToHost); checkCUDAError("Error in copying final_compressed_cpu\n"); //print_gpu_compressed_file(final_compressed_cpu,*finalsize_cpu); printf("The compressed value in binary is "); write_GPU_compressed(final_compressed_cpu,finalsize_cpu); for(int i=0;i<*finalsize_cpu;i++) // sprintf(compress_file+i,final_compressed_cpu[i]); //printf("i is %d and val is %d \n",i,final_compressed_cpu[i]); printf("\n"); // cudaMalloc((void **)&d_char_huffman_table,(MAX_CHAR)*(MAX_CHAR-1)*sizeof(int)); // cudaMemcpy(d_char_huffman_table,char_huffman_table,(MAX_CHAR)*(MAX_CHAR-1)*sizeof(int),cudaMemcpyHostToDevice); // printf("\n Going to compress on the GPU "); // compress<<<number_of_char,1>>>(d_input,number_of_char,d_char_huffman_table,MAX_CHAR); //test ends printf("compressing on CPU...\n"); //timer2=0; // CUT_SAFE_CALL(cutCreateTimer(&timer2)); //CUT_SAFE_CALL(cutStartTimer(timer2)); /*clock_t Linuxclock_start,Linuxclock_end; // clock return type cudaEvent_t CUDAevent_start, CUDAevent_end; float CUDAEvents_time; cudaEventRecord(CUDAevent_start, 0 ); cudaEventSynchronize(CUDAevent_start); Linuxclock_start = clock();*/ /*time_t before,after; before = time (NULL);*/ clock_t start1, stop1; start1 = clock(); compress_file(); stop1 = clock(); float elapsedTime = (float)(stop1 - start1) / (float)CLOCKS_PER_SEC * 1000.0f; printf( "Time in cpu : %3.1f ms\n", elapsedTime ); printf("Time to calculate results: %f ms.\n", elapsed_time); // print out execution time printf("Speedup achieved is %lf \n", elapsedTime/elapsed_time ); /*after = time (NULL); double dif; dif = difftime (after,before); printf ("It took you %.9lf seconds to type your name.\n", dif );*/ /*Linuxclock_end = clock(); cudaEventRecord(CUDAevent_end, 0 ); // instrument code to measure end time cudaEventSynchronize(CUDAevent_end); cudaEventElapsedTime(&CUDAEvents_time, CUDAevent_start, CUDAevent_end); printf("CPU Time using CUDA events: %f ms\n", CUDAEvents_time); // time_CUDAEvents is in ms printf("CPU Time using Linux clock: %f ms\n", ((double) (Linuxclock_end - Linuxclock_start) * 1000)/CLOCKS_PER_SEC); //Linuxclock in sec cudaEventDestroy(CUDAevent_start); cudaEventDestroy(CUDAevent_end);*/ //CUT_SAFE_CALL(cutStopTimer(timer2)); //float time2=cutGetAverageTimerValue(timer2); //printf(" Time on Host %f\n", time2); printf("decompressing on CPU...\n"); decompress_file(); getchar(); return 0; }
1,263
#include <stdio.h> #include <cuda_runtime.h> void printMatrix(int *C, const int nx, const int ny){ int *ic = C; printf("\n Matrix:(%d, %d)\n",nx,ny); for(int i =0; i < ny; i++){ for(int j =0; j < nx; j++){ printf("%3d",ic[j + i*nx]); } printf("\n"); } printf("\n"); } __global__ void printThreadIndex(int *A, const int nx, const int ny){ int ix = threadIdx.x + blockDim.x * blockIdx.x; int iy = threadIdx.y + blockDim.y * blockIdx.y; unsigned int idx = ix + iy*nx; printf("thread_id:(%d,%d),block_id:(%d,%d), coordinate(%d,%d) global index %2d ival %2d\n", threadIdx.x,threadIdx.y, blockIdx.x,blockIdx.y,ix,iy,idx,A[idx]); } void initInt(int *ip, int size){ for(int i =0; i < size; i++) ip[i] = i; } int main(){ printf("Starting ....\n"); // set device int dev = 0; cudaDeviceProp deviceProp; cudaSetDevice(dev); // set matrix dimension int nx = 8, ny = 6; int nxy = nx * ny, nBytes = nxy*sizeof(int); // malloc host memory int *h_A; h_A = (int *)malloc(nBytes); // init host matrix with integer initInt(h_A,nxy); printMatrix(h_A,nx,ny); // malloc device memory int *d_A; cudaMalloc((void **)&d_A, nBytes); // cp data from host to device cudaMemcpy(d_A,h_A,nBytes,cudaMemcpyHostToDevice); // set up execution configuration dim3 block(4,2); dim3 grid((nx + block.x - 1)/block.x, (ny + block.y -1)/block.y); // invoke the kernel printThreadIndex<<<grid,block>>>(d_A,nx,ny); cudaDeviceSynchronize(); // free host and device memory cudaFree(d_A); free(h_A); // reset device cudaDeviceReset(); return 0; }
1,264
#include <cuda.h> /* Size of a block */ #define BLOCK_X 32 #define BLOCK_Y 16 __global__ void kernadd (float* mout, float* min1, float *min2, int nx, int ny, size_t pitch) { int i, j, index; /* UP TO YOU edit line below so that the index is correctly evaluated */ i = blockDim.x * blockIdx.x +threadIdx.x; j = blockDim.y * blockIdx.y +threadIdx.y; index= i + j * pitch/sizeof(float); if ((i < nx) && (j < ny)) mout[index] = min1[index] + min2[index]; } /* extern "C" below is because this file follows C++ linking conventions */ /* whereas the companion C file (addition.c) follows C linking conventions */ /* which are different */ extern "C" void AddOnGpu(float* mat_out, float *mat_in1, float *mat_in2,\ int nx, int ny) { size_t pitch; /* Same pitch for all matrices, since they all have same size */ /* Matrix allocation on device */ float *mat_out_gpu, *mat_in1_gpu, *mat_in2_gpu; /* UP TO YOU : do the allocation below, using cudaMallocPitch ()*/ cudaMallocPitch(&mat_out_gpu, &pitch, sizeof(float)*nx, ny); cudaMallocPitch(&mat_in1_gpu, &pitch, sizeof(float)*nx, ny); cudaMallocPitch(&mat_in2_gpu, &pitch, sizeof(float)*nx, ny); /* The arguments mat_in1 and mat_in2 passed above are on the host. */ /* UP TO YOU : write below the instructions to copy it to the device */ /* You'll need to google the function cudaMemcpy2D () */ cudaMemcpy2D(mat_in1_gpu, pitch, mat_in1, nx*sizeof(float), nx*sizeof(float), ny, cudaMemcpyHostToDevice); cudaMemcpy2D(mat_in2_gpu, pitch, mat_in2, nx*sizeof(float), nx*sizeof(float), ny, cudaMemcpyHostToDevice); /* Grid topology below */ /* A block is BLOCK_X threads wide by BLOCK_Y threads high */ dim3 block (BLOCK_X, BLOCK_Y); /* UP TO YOU : complete the number of blocks below */ int n1 = (nx+BLOCK_X-1/BLOCK_X); int n2 = (nx+BLOCK_Y-1/BLOCK_Y); dim3 grid (n1,n2); /* UP TO YOU : kernel invocation */ kernadd <<< grid, block >>> (mat_out_gpu,mat_in1_gpu, mat_in2_gpu, nx, ny, pitch); cudaThreadSynchronize(); /* We now transfer back the matrix from the device to the host */ /* UP TO YOU : write cudaMemcpy2D () instruction below */ cudaMemcpy2D(mat_out, nx * sizeof(float), mat_out_gpu, pitch, nx * sizeof(float),ny,cudaMemcpyDeviceToHost); /* free memory */ cudaFree(mat_out_gpu); cudaFree(mat_in1_gpu); cudaFree(mat_in2_gpu); }
1,265
//2 layered neural network with LIF neurons //computing Vm in parallel, Computing Isyn //all-all connectivity between 2 layers //starting point of reading mnist set by 'start' #include<stdio.h> #include<math.h> #include<time.h> #include<stdlib.h> #include "device_launch_parameters.h" #include "cuda_runtime_api.h" #define C 300E-12 #define gL 30E-9 #define VT 20E-3 #define EL -70E-3 #define Rp 3E-3 #define dt 1E-4 #define decay 0.9802 #define decays 0.9231 #define decay1 0.9048 #define WT 5E-9 #define w_lat -1.0E-9 //inhibitory lateral strength //num of neurons in layer 1 and layer 2 #define p 28 #define q 28 #define N_imgs 50000 #define N2 10 //no. of neurons in 2nd layer //Convolution parameters #define Nw 3 #define L (p-Nw+1) //square matrix #define N1 (12*L*L) //no. of neurons in 1st layer #define Nthrds 1024 //use max no. of threads available per SM #define Nsyns N1*N2 //no. of connections #define T 0.1 #define M 1000 //Training parameters: #define r 20.0E-11 //Learning Rate for 100ms #define max_epochs 20 //1 complete presentation of all images //Variables for image reading unsigned char *pix_spks_d; int *d_imgs_lin, img_lin[N_imgs*(p*q+1)]; int test_set[N_imgs][p*q+1]; __device__ int d_imgs[N_imgs][p*q+1]; __device__ double img_spks[p*q][M]; __device__ double syn1[256], syn1s[256], syn[256][M]; __device__ unsigned char in_spk[N1]; __device__ double Isyn[N1][N2], weight[N1][N2]; __device__ double Isyn_tot[N2], Isyn_lat[N2]; __device__ double I_lat[N2]; //weight update variables: __device__ unsigned char D_op[N2][M]; __device__ signed char err[N2]; __device__ unsigned char Y_op[N2]; __device__ double del_w[N1][N2]; __device__ double ci[N1], d_hat[N1]; __device__ double cis[N1], norm_dh; __device__ double cil[N2], cils[N2]; __device__ double d_hat_sq[N1]; //Neuron variables: __device__ int ref_time1[N1],ref_time2[N2]; __device__ double Vm1[N1],Vm2[N2]; ////////////CUDA Kernels/////////// __global__ void img_readKernel(int *img) { for(int i=0; i<N_imgs; i++){ for(int j=0; j<(p*q+1); j++){ d_imgs[i][j]=img[i*(p*q+1)+j]; } } } __device__ unsigned char pix_spks[256][M]; unsigned char pix_spks_h[256*M]; //uniformly spaced spikes __global__ void convert2Spks(unsigned char *pix_spks_d) { for(int i=0; i<256; i++) { syn1[i]=0.0; syn1s[i]=0.0; for(int l=0; l<M; l++) { pix_spks[i][l]=pix_spks_d[i*M+l]; } } //Compute the synaptic kernels: for(int i=0; i<256; i++) { for(int l=0; l<M; l++) { syn1[i]*=decay; syn1s[i]*=decays; if(pix_spks[i][l]==1) { syn1[i]+=1.0; syn1s[i]+=1.0; } syn[i][l]=syn1[i]-syn1s[i]; } } } unsigned char pixspks[256*M]; __global__ void pix2spks(int k, int i) { int tx=threadIdx.x+blockIdx.x*blockDim.x; if(tx<(p*q)) { img_spks[tx][i]=syn[d_imgs[k][tx+1]][i]; __syncthreads(); } } __global__ void createDes(int k) { int tx=threadIdx.x+blockIdx.x*blockDim.x; if(tx<N2) { for(int i=0; i<M;i++) { D_op[tx][i]=0; } if(tx==d_imgs[k][0]) { for(int i=100; i<M; i+=35) D_op[tx][i]=1; } } } /*__global__ void pix2spks(int k) { //Load the spikes trains for the corresponding pixels: for(int i=1; i<(p*q+1); i++) { for(int l=0; l<M; l++) { img_spks[i-1][l]=syn[d_imgs[k][i]][l]; } } //Create the desired spike trains: for(int j=0; j<N2; j++) { for(int i=0; i<M; i++) { D_op[j][i]=0; } if(j==d_imgs[k][0]) { //for(int i=100; i<M; i+=50) for(int i=100; i<M; i+=35) D_op[j][i]=1; } } }*/ __device__ double I_in[N1]; __global__ void clear_vars() { int tx=threadIdx.x+blockIdx.x*blockDim.x; int ty=threadIdx.y+blockIdx.y*blockDim.y; if(tx<N1 && ty<N2) { Vm2[ty]=EL; Isyn_tot[ty]=0.0; ref_time2[ty]=0.0; Vm1[tx]=EL; ref_time1[tx]=0.0; I_in[tx]=0.0; del_w[tx][ty]=0.0; Isyn[tx][ty]=0.0; ci[tx]=0.0; cis[tx]=0.0; d_hat[tx]=0.0; cil[ty]=0.0; cils[ty]=0.0; Isyn_lat[ty]=0.0; } } __global__ void LifKernel1(int i) { int j = blockIdx.x * blockDim.x + threadIdx.x; double k1, k2; if(j<N1) { if(ref_time1[j]<i) ref_time1[j]=0; k1=(-gL*(Vm1[j]-EL)+I_in[j])/C; k2=(-gL*((Vm1[j]+dt*k1)-EL)+I_in[j])/C; Vm1[j]+=(dt*(k1+k2)/2)*(ref_time1[j]==0); if(Vm1[j]<EL) Vm1[j]=EL; if(Vm1[j]>=VT) { Vm1[j]=EL; ref_time1[j]=i+round(Rp/dt); in_spk[j]=1; } else { in_spk[j]=0; } __syncthreads(); } } __global__ void LifKernel2(int i) { int j = blockIdx.x * blockDim.x + threadIdx.x; double k1, k2; if(j<N2) { if(ref_time2[j]<i) ref_time2[j]=0; k1=(-gL*(Vm2[j]-EL)+Isyn_tot[j]+Isyn_lat[j])/C; k2=(-gL*((Vm2[j]+dt*k1)-EL)+Isyn_tot[j]+Isyn_lat[j])/C; Vm2[j]+=(dt*(k1+k2)/2)*(ref_time2[j]==0); if(Vm2[j]<EL) Vm2[j]=EL; if(Vm2[j]>=VT) { Vm2[j]=EL; ref_time2[j]=i+round(Rp/dt); Y_op[j]=1; } else { Y_op[j]=0; } //compute the error: err[j]=D_op[j][i]-Y_op[j]; __syncthreads(); } } //kernels for the total synapses in the network __global__ void SynKernel(int i) { int ix=blockIdx.x*blockDim.x + threadIdx.x; int iy=blockIdx.y*blockDim.y + threadIdx.y; if(ix<N1 && iy<N2) { if(iy==0) { ci[ix]=ci[ix]*decay; cis[ix]=cis[ix]*decays; if(in_spk[ix]==1) { ci[ix]+=1.0; cis[ix]+=1.0; } d_hat[ix]=d_hat[ix]*decay1 + ((ci[ix]-cis[ix])*dt)/C; d_hat_sq[ix]=d_hat[ix]*d_hat[ix]; } __syncthreads(); Isyn[ix][iy]=(ci[ix]-cis[ix])*weight[ix][iy]; } } __global__ void Lat_curr(int i) { int ix=blockIdx.x*blockDim.x+threadIdx.x; if(ix<N2) { cil[ix]=cil[ix]*decay; cils[ix]=cils[ix]*decays; if(Y_op[ix]==1) { cil[ix]+=1.0; cils[ix]+=1.0; } I_lat[ix]=w_lat*(cil[ix]-cils[ix]); Isyn_lat[ix]=0; for(int k=0; k<N2; k++) { if(k!=ix) { Isyn_lat[ix]+=I_lat[k]; } } } } __device__ double total_curr[8][N2]; __device__ double total_dhatsq[8]; //optimized version __global__ void IsynRedKernel(int i) { int ix=threadIdx.x+blockIdx.x*blockDim.x; int iy=threadIdx.y+blockIdx.y*blockDim.y; int tid=threadIdx.x; for(unsigned int s=blockDim.x/2;s>0;s>>=1) { if(iy<N2 && tid<s && (ix+s)<N1) { Isyn[ix][iy]+=Isyn[ix+s][iy]; } if(iy==0 && tid<s && (ix+s)<N1) { d_hat_sq[ix]+=d_hat_sq[ix+s]; } __syncthreads(); } if(tid==0 && iy<N2) { total_curr[blockIdx.x][iy]=Isyn[ix][iy]; if(iy==0) { total_dhatsq[blockIdx.x]=d_hat_sq[ix]; } } } __global__ void reduce1(int i) { int tx=threadIdx.x+blockIdx.x*blockDim.x; if(tx<N2) { double total1=0.0; for(int k=0; k<8; k++) { total1+=total_curr[k][tx]; } Isyn_tot[tx]=total1; if(tx==0) { double total2=0.0; for(int j=0; j<8; j++) total2+=total_dhatsq[j]; norm_dh=sqrt(total2); } } } __global__ void CalcUpdate(int i, double l_rate) { int ix=threadIdx.x+blockIdx.x*blockDim.x; int iy=threadIdx.y+blockIdx.y*blockDim.y; if(ix<N1 && iy<N2) { if(norm_dh!=0 && err[iy]!=0) { del_w[ix][iy]+=(err[iy]*l_rate*d_hat[ix]/norm_dh); } __syncthreads(); } } __global__ void WtUpdt() { int ix=threadIdx.x+blockIdx.x*blockDim.x; int iy=threadIdx.y+blockIdx.y*blockDim.y; if(ix<N1 && iy<N2) { weight[ix][iy]+=del_w[ix][iy]; } } __global__ void cpyWts(double *wts) { int ix=threadIdx.x+blockIdx.x*blockDim.x; int iy=threadIdx.y+blockIdx.y*blockDim.y; if(ix<N1 && iy<N2) { wts[ix*N2+iy]=weight[ix][iy]; } } __device__ double w_conv1[Nw][Nw], w_conv2[Nw][Nw]; __device__ double w_conv3[Nw][Nw], w_conv4[Nw][Nw]; __device__ double w_conv5[Nw][Nw], w_conv6[Nw][Nw]; __device__ double w_conv7[Nw][Nw], w_conv8[Nw][Nw]; __device__ double w_conv9[Nw][Nw], w_conv10[Nw][Nw]; __device__ double w_conv11[Nw][Nw], w_conv12[Nw][Nw]; __global__ void initialize2D(double *d_wts, double *c_wts) { for(int i=0; i<N1; i++) { for(int j=0; j<N2; j++) { weight[i][j]=d_wts[i*N2+j]; } } for(int i=0; i<(12*Nw); i++) { for(int j=0; j<Nw; j++) { if(i<Nw) { w_conv1[i][j]=c_wts[i*Nw+j]; } else if(i>=Nw && i<(2*Nw)) { w_conv2[i-Nw][j]=c_wts[i*Nw+j]; } else if(i>=(2*Nw) && i<(3*Nw)) { w_conv3[i-(2*Nw)][j]=c_wts[i*Nw+j]; } else if(i>=(3*Nw) && i<(4*Nw)){ w_conv4[i-(3*Nw)][j]=c_wts[i*Nw+j]; } else if(i>=(4*Nw) && i<(5*Nw)){ w_conv5[i-(4*Nw)][j]=c_wts[i*Nw+j]; } else if(i>=(5*Nw) && i<(6*Nw)){ w_conv6[i-(5*Nw)][j]=c_wts[i*Nw+j]; } else if(i>=(6*Nw) && i<(7*Nw)){ w_conv7[i-(6*Nw)][j]=c_wts[i*Nw+j]; } else if(i>=(7*Nw) && i<(8*Nw)){ w_conv8[i-(7*Nw)][j]=c_wts[i*Nw+j]; } else if(i>=(8*Nw) && i<(9*Nw)){ w_conv9[i-(8*Nw)][j]=c_wts[i*Nw+j]; } else if(i>=(9*Nw) && i<(10*Nw)){ w_conv10[i-(9*Nw)][j]=c_wts[i*Nw+j]; } else if(i>=(10*Nw) && i<(11*Nw)){ w_conv11[i-(10*Nw)][j]=c_wts[i*Nw+j]; } else if(i>=(11*Nw) && i<(12*Nw)){ w_conv12[i-(11*Nw)][j]=c_wts[i*Nw+j]; } } } } __global__ void convKernel1(int i) { int ix=threadIdx.x+blockIdx.x*blockDim.x; int iy=threadIdx.y+blockIdx.y*blockDim.y; int x, y; if(ix<L && iy<L) { double temp=0.0; for(x=0; x<Nw; x++) { for(y=0; y<Nw; y++) { temp+=WT*w_conv1[x][y]*img_spks[(ix+x)*q+iy+y][i]; } } I_in[ix*L+iy]=temp; } } __global__ void convKernel2(int i) { int ix=threadIdx.x+blockIdx.x*blockDim.x; int iy=threadIdx.y+blockIdx.y*blockDim.y; int x, y; if(ix<L && iy<L) { double temp=0.0; for(x=0; x<Nw; x++) { for(y=0; y<Nw; y++) { temp+=WT*w_conv2[x][y]*img_spks[(ix+x)*q+iy+y][i]; } } I_in[L*L+ix*L+iy]=temp; } } __global__ void convKernel3(int i) { int ix=threadIdx.x+blockIdx.x*blockDim.x; int iy=threadIdx.y+blockIdx.y*blockDim.y; int x, y; if(ix<L && iy<L) { double temp=0.0; for(x=0; x<Nw; x++) { for(y=0; y<Nw; y++) { temp+=WT*w_conv3[x][y]*img_spks[(ix+x)*q+iy+y][i]; } } I_in[2*L*L+ix*L+iy]=temp; } } __global__ void convKernel4(int i) { int ix=threadIdx.x+blockIdx.x*blockDim.x; int iy=threadIdx.y+blockIdx.y*blockDim.y; int x, y; if(ix<L && iy<L) { double temp=0.0; for(x=0; x<Nw; x++) { for(y=0; y<Nw; y++) { temp+=WT*w_conv4[x][y]*img_spks[(ix+x)*q+iy+y][i]; } } I_in[3*L*L+ix*L+iy]=temp; } } __global__ void convKernel5(int i) { int ix=threadIdx.x+blockIdx.x*blockDim.x; int iy=threadIdx.y+blockIdx.y*blockDim.y; int x, y; if(ix<L && iy<L) { double temp=0.0; for(x=0; x<Nw; x++) { for(y=0; y<Nw; y++) { temp+=WT*w_conv5[x][y]*img_spks[(ix+x)*q+iy+y][i]; } } I_in[4*L*L+ix*L+iy]=temp; } } __global__ void convKernel6(int i) { int ix=threadIdx.x+blockIdx.x*blockDim.x; int iy=threadIdx.y+blockIdx.y*blockDim.y; int x, y; if(ix<L && iy<L) { double temp=0.0; for(x=0; x<Nw; x++) { for(y=0; y<Nw; y++) { temp+=WT*w_conv6[x][y]*img_spks[(ix+x)*q+iy+y][i]; } } I_in[5*L*L+ix*L+iy]=temp; } } __global__ void convKernel7(int i) { int ix=threadIdx.x+blockIdx.x*blockDim.x; int iy=threadIdx.y+blockIdx.y*blockDim.y; int x, y; if(ix<L && iy<L) { double temp=0.0; for(x=0; x<Nw; x++) { for(y=0; y<Nw; y++) { temp+=WT*w_conv7[x][y]*img_spks[(ix+x)*q+iy+y][i]; } } I_in[6*L*L+ix*L+iy]=temp; } } __global__ void convKernel8(int i) { int ix=threadIdx.x+blockIdx.x*blockDim.x; int iy=threadIdx.y+blockIdx.y*blockDim.y; int x, y; if(ix<L && iy<L) { double temp=0.0; for(x=0; x<Nw; x++) { for(y=0; y<Nw; y++) { temp+=WT*w_conv8[x][y]*img_spks[(ix+x)*q+iy+y][i]; } } I_in[7*L*L+ix*L+iy]=temp; } } __global__ void convKernel9(int i) { int ix=threadIdx.x+blockIdx.x*blockDim.x; int iy=threadIdx.y+blockIdx.y*blockDim.y; int x, y; if(ix<L && iy<L) { double temp=0.0; for(x=0; x<Nw; x++) { for(y=0; y<Nw; y++) { temp+=WT*w_conv9[x][y]*img_spks[(ix+x)*q+iy+y][i]; } } I_in[8*L*L+ix*L+iy]=temp; } } __global__ void convKernel10(int i) { int ix=threadIdx.x+blockIdx.x*blockDim.x; int iy=threadIdx.y+blockIdx.y*blockDim.y; int x, y; if(ix<L && iy<L) { double temp=0.0; for(x=0; x<Nw; x++) { for(y=0; y<Nw; y++) { temp+=WT*w_conv10[x][y]*img_spks[(ix+x)*q+iy+y][i]; } } I_in[9*L*L+ix*L+iy]=temp; } } __global__ void convKernel11(int i) { int ix=threadIdx.x+blockIdx.x*blockDim.x; int iy=threadIdx.y+blockIdx.y*blockDim.y; int x, y; if(ix<L && iy<L) { double temp=0.0; for(x=0; x<Nw; x++) { for(y=0; y<Nw; y++) { temp+=WT*w_conv11[x][y]*img_spks[(ix+x)*q+iy+y][i]; } } I_in[10*L*L+ix*L+iy]=temp; } } __global__ void convKernel12(int i) { int ix=threadIdx.x+blockIdx.x*blockDim.x; int iy=threadIdx.y+blockIdx.y*blockDim.y; int x, y; if(ix<L && iy<L) { double temp=0.0; for(x=0; x<Nw; x++) { for(y=0; y<Nw; y++) { temp+=WT*w_conv12[x][y]*img_spks[(ix+x)*q+iy+y][i]; } } I_in[11*L*L+ix*L+iy]=temp; } } long timediff(clock_t t1, clock_t t2) { long elapsed; elapsed = ((double)t2 - t1) / CLOCKS_PER_SEC * 1000; return elapsed; } double h_wts[N1*N2], *d_wts; double *dcwts; double h_wts_saved[N1*N2], *d_wts_saved; double c_wts[12*Nw*Nw]; int main(int argc, char *argv[]) { int start=atoi(argv[1]); FILE *FW, *FWI; //to load initial wts and store final wts FILE *conv_wt; clock_t t1, t2; long elapsed=0; FILE *F_train, *F_lif_spks; //for concurrent execution of different kernels: cudaStream_t stream2, stream3, stream4, stream5, stream6; cudaStream_t stream7, stream8, stream9, stream10, stream11, stream12; cudaStream_t stream13, stream14,stream15; //set the gpu device: cudaSetDevice(4); cudaStreamCreate(&stream2); cudaStreamCreate(&stream3); cudaStreamCreate(&stream4); cudaStreamCreate(&stream5); cudaStreamCreate(&stream6); cudaStreamCreate(&stream7); cudaStreamCreate(&stream8); cudaStreamCreate(&stream9); cudaStreamCreate(&stream10); cudaStreamCreate(&stream11); cudaStreamCreate(&stream12); cudaStreamCreate(&stream13); cudaStreamCreate(&stream14); cudaStreamCreate(&stream15); F_lif_spks = fopen("pixels_spks.csv","r"); if(F_lif_spks == NULL) { perror("Error while opening file pixels_spks.csv\n"); exit(EXIT_FAILURE); } F_train = fopen("mnist_train.csv","r"); //F_test = fopen("mnist_test.csv","r"); if(F_train == NULL) { perror("Error while opening file mnist_train.csv\n"); exit(EXIT_FAILURE); } FWI = fopen("wts_initial.csv","r"); if(FWI == NULL) { perror("Error while opening file wts_trained.csv\n"); exit(EXIT_FAILURE); } conv_wt=fopen("kernels_3x3.csv","r"); if(conv_wt==NULL) { perror("Error while opening file kernel.csv\n"); exit(EXIT_FAILURE); } printf("Total no. of neurons=%d, no. of synapses to be trained=%d\n",(N1+N2), N1*N2); cudaMalloc((void**)&d_imgs_lin,sizeof(int)*(N_imgs*(p*q+1))); cudaMalloc((void**)&d_wts,N1*N2*sizeof(double)); cudaMalloc((void**)&dcwts,12*Nw*Nw*sizeof(double)); cudaMalloc((void**)&d_wts_saved,N1*N2*sizeof(double)); cudaMalloc((void**)&pix_spks_d,256*M*sizeof(unsigned char)); //Read the initial weights: //printf("Reading final trained weights from file\n"); for(int i=0; i<N1; i++) { for(int j=0; j<N2; j++) { fscanf(FWI,"%lf,",&h_wts[i*N2+j]); } } fclose(FWI); for(int i=0; i<(12*Nw); i++) { for(int j=0; j<Nw; j++) { fscanf(conv_wt,"%lf,",&c_wts[i*Nw+j]); } } fclose(conv_wt); for(int i=0; i<(256); i++) { for(int j=0; j<M; j++) { fscanf(F_lif_spks,"%d,",&pix_spks_h[i*M+j]); } } fclose(F_lif_spks); cudaMemcpy(pix_spks_d,pix_spks_h,256*M*sizeof(unsigned char),cudaMemcpyHostToDevice); cudaMemcpy(d_wts,h_wts,N1*N2*sizeof(double),cudaMemcpyHostToDevice); cudaMemcpy(dcwts,c_wts,12*Nw*Nw*sizeof(double),cudaMemcpyHostToDevice); initialize2D<<<1,1>>>(d_wts,dcwts); cudaDeviceSynchronize(); cudaFree(d_wts); cudaFree(dcwts); //Read the images from file: for(int n=0;n<N_imgs;n++) { for(int j=0;j<(p*q+1);j++) { fscanf(F_train,"%d,",&test_set[n][j]); } } fclose(F_train); //convert 2D matrix to 1D for transfer to device: for(int n=0; n<(N_imgs);n++) { for(int j=0;j<(p*q+1);j++) { img_lin[n*(p*q+1)+j]=test_set[n][j]; } } cudaMemcpy(d_imgs_lin,img_lin,sizeof(int)*(N_imgs*(p*q+1)),cudaMemcpyHostToDevice); //call cuda kernel to read in the images: img_readKernel<<<1,1>>>(d_imgs_lin); cudaThreadSynchronize(); cudaFree(d_imgs_lin); int NBlks=(N1/Nthrds)+1; //printf("blocks=%d, threads=%d\n",NBlks,Nthrds); dim3 dimGrid(82,1,1); dim3 dimBlock(100,10,1); dim3 grid_syn(NBlks,10,1); dim3 block_syn(Nthrds,1,1); //convert pixel values 0 to 255 into spike trains convert2Spks<<<1,1>>>(pix_spks_d); cudaDeviceSynchronize(); cudaFree(pix_spks_d); dim3 convGrid(1,1,1); dim3 convBlks(26,26,1); double learn=r; //CPU time required for computation t1 = clock(); int l=0; //image index for(int n=0; n<max_epochs; n++) { printf("Epoch=%d\n",n); if(n<3) learn=r; else if(n>=3 && n<6) learn=r/2; else if(n>=6 && n<9) learn=r/4; else if(n>=9 && n<12) learn=r/8; else if(n>=12 && n<15) learn=r/16; else if(n>=15 && n<18) learn=r/32; else learn=r/64; for(l=0; l<N_imgs; l++) { printf("l=%d image %d\n",l,test_set[l][0]); createDes<<<1,10>>>(l); clear_vars<<<dimGrid,dimBlock>>>(); cudaDeviceSynchronize(); //simulate for all time steps for(int i=0; i<M; i++) { pix2spks<<<1,784>>>(l,i); /////////////////////////////////////////////////// cudaDeviceSynchronize(); convKernel1<<<convGrid,convBlks,0,stream2>>>(i); convKernel2<<<convGrid,convBlks,0,stream3>>>(i); convKernel3<<<convGrid,convBlks,0,stream4>>>(i); convKernel4<<<convGrid,convBlks,0,stream5>>>(i); convKernel5<<<convGrid,convBlks,0,stream6>>>(i); convKernel6<<<convGrid,convBlks,0,stream7>>>(i); convKernel7<<<convGrid,convBlks,0,stream8>>>(i); convKernel8<<<convGrid,convBlks,0,stream9>>>(i); convKernel9<<<convGrid,convBlks,0,stream10>>>(i); convKernel10<<<convGrid,convBlks,0,stream11>>>(i); convKernel11<<<convGrid,convBlks,0,stream12>>>(i); convKernel12<<<convGrid,convBlks,0,stream13>>>(i); ///////////////////////////////////////////////////// cudaDeviceSynchronize(); LifKernel1<<<NBlks,Nthrds>>>(i); //////////////////////////////////////////////// cudaDeviceSynchronize(); SynKernel<<<dimGrid,dimBlock,0,stream14>>>(i); Lat_curr<<<1,10,0,stream15>>>(i); //////////////////////////////////////////////// cudaDeviceSynchronize(); IsynRedKernel<<<grid_syn,block_syn>>>(i); reduce1<<<10,1>>>(i); /////////////////////////////////////////////// cudaDeviceSynchronize(); LifKernel2<<<1,10>>>(i); //////////////////////////////////////////////////// cudaDeviceSynchronize(); CalcUpdate<<<dimGrid,dimBlock>>>(i,learn); } ////////////////////////////////////////////////////////////////////////////// WtUpdt<<<dimGrid,dimBlock>>>(); //cudaDeviceSynchronize(); } //end of loops over N_imgs t2 = clock(); elapsed += timediff(t1,t2); printf("Elapsed time: %ld ms\n", elapsed); char wts_file[25]; sprintf(wts_file,"wts_gpu%d.txt",n); if((FW=fopen(wts_file,"w"))==NULL) { printf("Failed to open file wts_gpu_sample%d.txt\n",n); } printf("Copying the trained weights to the host for analysis\n"); cpyWts<<<dimGrid,dimBlock>>>(d_wts_saved); cudaMemcpy(h_wts_saved,d_wts_saved,N1*N2*sizeof(double), cudaMemcpyDeviceToHost); //save the weights in a file: for(int i=0; i<N1; i++) { for(int j=0; j<N2; j++) { fprintf(FW,"%0.14f,",h_wts_saved[i*N2+j]); } fprintf(FW,"\n"); } fclose(FW); } //end of n epochs loops cudaFree(d_wts_saved); cudaDeviceReset(); return(0); }
1,266
#include <string.h> #include <stdio.h> #include <iostream> struct DataElement { char *name; int value; }; __global__ void Kernel(DataElement *elem) { printf("On device: name=%s, value=%d\n", elem->name, elem->value); elem->name[0] = 'd'; elem->value++; } void launch(DataElement *elem, cudaStream_t &stream) { Kernel<<< 1, 1, 0, stream >>>(elem); //cudaDeviceSynchronize(); } void iteration(cudaStream_t &stream) { DataElement *e; cudaMallocManaged((void**)&e, sizeof(DataElement)); e->value = 10; cudaMallocManaged((void**)&(e->name), sizeof(char) * (strlen("hello") + 1) ); strcpy(e->name, "hello"); launch(e, stream); printf("On host: name=%s, value=%d\n", e->name, e->value); cudaFree(e->name); cudaFree(e); } int main(void) { cudaError_t err; int count = 0; err = cudaGetDeviceCount(&count); std::cout << count << " devices found." << std::endl; for (int d=0;d<count;d++) { err = cudaSetDevice(d); if (err != cudaSuccess) { std::cout << "error setting device, #=" << cudaGetErrorString(err) << std::endl; } cudaDeviceProp deviceProp; err = cudaGetDeviceProperties(&deviceProp, d); if (err != cudaSuccess) { std::cout << "error getting device properties, #=" << cudaGetErrorString(err) << std::endl; } std::cout << "Using device " << d << ", name: " << deviceProp.name << std::endl; for (int s = 0 ; s < 10 ; s++) { cudaStream_t stream; err = cudaStreamCreate(&stream); if (err != cudaSuccess) { std::cout << "error in stream creation, #=" << cudaGetErrorString(err) << std::endl; } iteration(stream); cudaStreamDestroy(stream); } } }
1,267
#include "includes.h" __global__ void ElementwiseNorm(float * A, float *B, int size) { int id = blockDim.x*blockIdx.y*gridDim.x + blockDim.x*blockIdx.x + threadIdx.x; if (id < size) A[id] /= B[id]; }
1,268
#include<stdio.h> #include<stdlib.h> #define TPB 8 #define W 4 #define H 4 #define TX 1 #define TY 1 int N=H*W; __device__ float distance(float x1, float x2){ return sqrt ((x2-x1)*(x2-x1)); } __global__ void distanceKernel(float *d_out, float *d_in, float ref, int w){ const int c=blockIdx.x*blockDim.x+threadIdx.x; const int r=blockIdx.y*blockDim.y+threadIdx.y; const int i=r*w+c; const float x=d_in[i]; d_out[i]=distance(x, ref); printf("c=%d, r=%d, i=%d: the distance between %f to %f is %f. \n", c, r, i, ref, x, d_out[i]); //// } void distanceArray(float *out, float *in, float ref, int len){ float *d_in=0; float *d_out=0; //// cudaMalloc(&d_in, len*sizeof(float)); cudaMalloc(&d_out, len*sizeof(float)); cudaMemcpy(d_in, in, len*sizeof(float), cudaMemcpyHostToDevice); const dim3 blockSize(TX, TY); const int bx=(W+TX-1)/TX; const int by=(W+TY-1)/TY; const dim3 gridSize=dim3(bx, by); distanceKernel<<<gridSize, blockSize>>>(d_out, d_in, ref, W); cudaMemcpy(out, d_out, len*sizeof(float), cudaMemcpyDeviceToHost); cudaFree(d_in); cudaFree(d_out); } float scale(int i, int n){ return ((float) i)/(n-1); } int main(){ const float ref=0.5f; float *in=(float*) calloc(N,sizeof(float)); float *out=(float*) calloc(N, sizeof(float)); for(int i=0; i<N; ++i){ in[i]=scale(i,N); // } distanceArray(out, in, ref, N); printf("______________________________ \n"); for(int j=0; j<N; ++j){ printf("The distance, printed from the host, between %f to %f is %f. \n", ref, in[j], out[j]); } free(in); free(out); return 0; }
1,269
#include <thrust/device_vector.h> #include <thrust/host_vector.h> float* readData(char* filename) { FILE* handle = fopen(filename, "r"); if(handle == NULL) { printf("Error opening file: %s\n", filename); exit(0); } int num, i; fscanf(handle, "%d", &num); float data[num]; for(i=0; i<num; i++) fscanf(handle, "%f", &data[i]); //printf("%f %f %f\n", data[0], data[1], data[2]); return data; } int main(int argc, char *argv[]) { float *hostInput1 = NULL; float *hostInput2 = NULL; int i; /* parse the input arguments */ //@@ Insert code here if(argc != 11) { printf("\nUsage: ./ThrustVectorAdd_Template -e <expected.raw> -i <input0.raw> , <input1.raw> -o <output.raw> -t vector\n\n"); return 0; } char* input0_filename = argv[4]; char* input1_filename = argv[6]; char* output_filename = argv[8]; // Import host input data //@@ Read data from the raw files here //@@ Insert code here hostInput1 = readData(input0_filename); hostInput2 = readData(input1_filename); // Declare and allocate host output //@@ Insert code here int num = sizeof(hostInput1)/sizeof(float); thrust::host_vector<float> hostOutput(num); // Declare and allocate thrust device input and output vectors //@@ Insert code here thrust::device_vector<float> devInput1(num); thrust::device_vector<float> devInput2(num); thrust::device_vector<float> devOutput(num); // Copy to device //@@ Insert code here thrust::copy(hostInput1, hostInput1 + num, devInput1.begin()); thrust::copy(hostInput2, hostInput2 + num, devInput2.begin()); // Execute vector addition //@@ Insert Code here //printf("dev: %f %f\n", devInput1[1], devInput2[1]); thrust::transform(devInput1.begin(), devInput1.end(), devInput2.begin(), devOutput.begin(), thrust::plus<float>()); ///////////////////////////////////////////////////////// // Copy data back to host //@@ Insert code here thrust::copy(devOutput.begin(), devOutput.end(), hostOutput.begin()); //printf("%d %d %d\n", hostOutput[1], hostOutput[2], hostOutput[0]); //Cross-verification float* verifyData = readData(output_filename); if(num != sizeof(verifyData)/sizeof(float)) printf("Size not matching: Output size: %d\tExpected size: %d\n", num, sizeof(verifyData)/sizeof(float)); else for(i=0; i<num; i++) { if((float)verifyData[i] != (float)hostOutput[i]) printf("Data not matching: Location: %d\tOutput: %f\tExpected: %f\n", i+1, hostOutput[i], verifyData[i]); } return 0; }
1,270
// This example demonstrate how use the printf() function inside a kernel. // In order to do that the code must be generate to architetures with compute capability greater than 2.0 // Compile: // nvcc -gencode=arch=compute_30,code=sm_30 -g -o helloGPU helloGPU.cu #include <stdio.h> __global__ void helloCUDA(float f) { printf("Hello world from thread %d -- My value of f = %f\n", threadIdx.x, f); } int main() { helloCUDA<<<1, 5>>>(1.2345f); cudaDeviceReset(); return 0; }
1,271
/* ================================================================== Programmers: Kevin Wagner Elijah Malaby John Casey Omptimizing SDH histograms for input larger then global memory ================================================================== */ #include <stdio.h> #include <stdlib.h> #include <ctype.h> #include <math.h> #include <sys/time.h> #define BOX_SIZE 23000 /* size of the data box on one dimension */ /* descriptors for single atom in the tree */ typedef struct atomdesc { float x_pos; float y_pos; float z_pos; } atom; unsigned long long * histogram; /* list of all buckets in the histogram */ unsigned long long PDH_acnt; /* total number of data points */ int block_size; /* Number of threads per block */ int num_buckets; /* total number of buckets in the histogram */ float PDH_res; /* value of w */ atom * atom_list; /* list of all data points */ unsigned long long * histogram_GPU; unsigned long long * temp_histogram_GPU; atom * atom_list_GPU; __global__ void kernelSumHistogram( unsigned long long int *InputHists, unsigned long long int *hist, int num_atoms, int num_buckets, int block_size) { unsigned long long int tid = threadIdx.x + blockIdx.x * blockDim.x; int h_pos = tid; unsigned long long int NumberOfSumLoop = 0; NumberOfSumLoop = (num_atoms)/block_size + ((num_atoms%block_size) ? 1:0); while(h_pos < num_buckets) { unsigned long long int tmpAns = 0; for(int i=0;i<NumberOfSumLoop;i++){ tmpAns = tmpAns + *(InputHists+(i*num_buckets)+h_pos); } hist[h_pos] = tmpAns; h_pos += blockDim.x * gridDim.x; } __syncthreads(); } __device__ void block_to_block (atom * block_a, atom * block_b, int b_length, unsigned long long * histogram, float resolution) { atom me = block_a[threadIdx.x]; for(int i = 0; i < b_length; i++) atomicAdd(&(histogram[(int)(sqrt((me.x_pos - block_b[i].x_pos) * (me.x_pos - block_b[i].x_pos) + (me.y_pos - block_b[i].y_pos) * (me.y_pos - block_b[i].y_pos) + (me.z_pos - block_b[i].z_pos) * (me.z_pos - block_b[i].z_pos)) / resolution)]), 1); } __global__ void GPUKernelFunction (unsigned long long PDH_acnt, float PDH_res, atom * atom_list_GPU, unsigned long long * histogram_GPU, int num_buckets) { extern __shared__ unsigned long long SHist[]; /* assign register values */ int i, h_pos; float dist; atom * my_block = &atom_list_GPU[blockIdx.x * blockDim.x]; atom temp_atom_1 = my_block[threadIdx.x]; for(h_pos=threadIdx.x; h_pos < num_buckets; h_pos+=blockDim.x) SHist[h_pos] = 0; __syncthreads(); /* loop through all points in atom list calculating distance from current point to all further points */ for (i = threadIdx.x + 1; i < blockDim.x && i+blockIdx.x*blockDim.x < PDH_acnt; i++) { atom temp_atom_2 = my_block[i]; dist = sqrt((temp_atom_1.x_pos - temp_atom_2.x_pos) * (temp_atom_1.x_pos - temp_atom_2.x_pos) + (temp_atom_1.y_pos - temp_atom_2.y_pos) * (temp_atom_1.y_pos - temp_atom_2.y_pos) + (temp_atom_1.z_pos - temp_atom_2.z_pos) * (temp_atom_1.z_pos - temp_atom_2.z_pos)); h_pos = (int)(dist / PDH_res); atomicAdd(&(SHist[h_pos]), 1); } __syncthreads(); for(i=blockIdx.x+1; i < gridDim.x-1; i++) block_to_block(my_block, &atom_list_GPU[i*blockDim.x], blockDim.x, SHist, PDH_res); block_to_block(my_block, &atom_list_GPU[i*blockDim.x], PDH_acnt-i*blockDim.x, // Last block may be small SHist, PDH_res); __syncthreads(); for(h_pos = threadIdx.x; h_pos < num_buckets; h_pos += blockDim.x) *(histogram_GPU+(num_buckets*blockIdx.x)+h_pos) += SHist[h_pos]; } /* print the counts in all buckets of the histogram */ void output_histogram_GPU(){ int i; unsigned long long total_cnt = 0; for(i=0; i< num_buckets; i++) { if(i%5 == 0) /* we print 5 buckets in a row */ printf("\n%02d: ", i); printf("%15lld ", histogram[i]); total_cnt += histogram[i]; /* we also want to make sure the total distance count is correct */ if(i == num_buckets - 1) printf("\n T:%lld \n", total_cnt); else printf("| "); } } void GPU_baseline() { int num_blocks = ((PDH_acnt + block_size)/block_size); /* copy histogram to device memory */ cudaMalloc((void**) &histogram_GPU, sizeof(unsigned long long)*num_buckets); cudaMemset(histogram_GPU, 0, sizeof(unsigned long long)*num_buckets); cudaMalloc((void**) &temp_histogram_GPU, sizeof(unsigned long long)*num_buckets*num_blocks); cudaMemset(temp_histogram_GPU, 0, sizeof(unsigned long long)*num_buckets*num_blocks); /* copy atom list to device memory */ cudaMalloc((void**) &atom_list_GPU, sizeof(atom) * PDH_acnt); cudaMemcpy(atom_list_GPU, atom_list, sizeof(atom) * PDH_acnt, cudaMemcpyHostToDevice); /* start time keeping */ cudaEvent_t start, stop; cudaEventCreate(&start); cudaEventCreate(&stop); cudaEventRecord( start, 0 ); /* Run Kernel */ GPUKernelFunction <<<num_blocks, block_size, sizeof(unsigned long long)*num_buckets>>> (PDH_acnt, PDH_res, atom_list_GPU, temp_histogram_GPU, num_buckets); cudaDeviceSynchronize(); kernelSumHistogram<<<3, 512>>>(temp_histogram_GPU, histogram_GPU, PDH_acnt, num_buckets, block_size); /* stop time keeping */ cudaEventRecord( stop, 0 ); cudaEventSynchronize( stop ); float elapsedTime; cudaEventElapsedTime( &elapsedTime, start, stop ); /* transfer histogram to host memory */ cudaMemcpy(histogram, histogram_GPU, sizeof(unsigned long long)*num_buckets, cudaMemcpyDeviceToHost); /* print out the histogram */ output_histogram_GPU(); elapsedTime = elapsedTime/1000; printf( "******** Total Running Time of Kernel = %0.5f sec *******\n", elapsedTime ); /* free cuda timekeeping */ cudaEventDestroy( start ); cudaEventDestroy( stop ); cudaFree(temp_histogram_GPU); } /* Input Validation Function */ bool isNumber(char number[], bool floatingpoint) { for (int i = 0; number[i] != 0; i++) { //if (number[i] > '9' || number[i] < '0') if (!isdigit(number[i])) { if((number[i] == '.' && floatingpoint)) { floatingpoint = false; } else { return false; } } } return true; } /* Most of this input validation can probably be pulled whenever we hardcode our block size and if we hardcode our bucket width */ int main(int argc, char **argv) { /* input validation */ if((argc > 3)) { if(((isNumber(argv[1], false) && isNumber(argv[2], true)) && isNumber(argv[3], false))) { PDH_acnt = atoi(argv[1]); PDH_res = atof(argv[2]); block_size = atoi(argv[3]); } else { printf( "Invalid Input Error Invalid Arguments\n Valid input is ./program_name {#of_samples} {bucket_width} {block_size}\n"); return 0; } } else { printf( "Invalid Input Error Insufficient Arguments\n Valid input is ./program_name {#of_samples} {bucket_width} {block_size}\n"); return 0; } /* allocate memory */ num_buckets = (int)(BOX_SIZE * 1.732 / PDH_res) + 1; histogram = (unsigned long long *)malloc(sizeof(unsigned long long)*num_buckets); atom_list = (atom *)malloc(sizeof(atom)*PDH_acnt); srand(1); /* generate data following a uniform distribution */ for(int i = 0; i < PDH_acnt; i++) { atom_list[i].x_pos = ((float)(rand()) / RAND_MAX) * BOX_SIZE; atom_list[i].y_pos = ((float)(rand()) / RAND_MAX) * BOX_SIZE; atom_list[i].z_pos = ((float)(rand()) / RAND_MAX) * BOX_SIZE; } /* call GPU histrogram compute */ GPU_baseline(); /* free memory */ free(histogram); free(atom_list); return 0; }
1,272
#include "friction_update.cuh" __global__ void friction_update ( SimulationParameters sim_params, SolverParameters solver_params, real dt, AssembledSolution d_assem_sol ) { int x = blockIdx.x * blockDim.x + threadIdx.x; if (x < sim_params.cells + 2) { if (d_assem_sol.h_BC[x] > solver_params.tol_dry && abs(d_assem_sol.q_BC[x]) > solver_params.tol_dry) { real u = d_assem_sol.q_BC[x] / d_assem_sol.h_BC[x]; real Cf = solver_params.g * pow(sim_params.manning, C(2.0)) / pow(d_assem_sol.h_BC[x], C(1.0) / C(3.0)); real Sf = -Cf * abs(u) * u; real D = 1 + 2 * dt * Cf * abs(u) / d_assem_sol.h_BC[x]; // Update d_assem_sol.q_BC[x] += dt * Sf / D; } } }
1,273
#include "includes.h" __global__ void CalcAngMom(double *AngMomx_d, double *AngMomy_d, double *AngMomz_d, double *GlobalAMx_d, double *GlobalAMy_d, double *GlobalAMz_d, double *Mh_d, double *Rho_d, double A, double Omega, double *Altitude_d, double *Altitudeh_d, double *lonlat_d, double *areasT, double *func_r_d, int num, bool DeepModel) { int id = blockIdx.x * blockDim.x + threadIdx.x; int nv = gridDim.y; int lev = blockIdx.y; if (id < num) { double AMx, AMy, AMz; double rx, ry, rz, r; //calculate control volume double zup, zlow, Vol; zup = Altitudeh_d[lev + 1] + A; zlow = Altitudeh_d[lev] + A; if (DeepModel) { Vol = areasT[id] / pow(A, 2) * (pow(zup, 3) - pow(zlow, 3)) / 3; } else { Vol = areasT[id] * (zup - zlow); } //radius vector r = (A + Altitude_d[lev]); rx = r * func_r_d[id * 3 + 0]; ry = r * func_r_d[id * 3 + 1]; rz = r * func_r_d[id * 3 + 2]; //angular momentum r x p (total x and y over globe should ~ 0, z ~ const) AMx = ry * Mh_d[id * 3 * nv + lev * 3 + 2] - rz * Mh_d[id * 3 * nv + lev * 3 + 1] - Rho_d[id * nv + lev] * Omega * r * rz * cos(lonlat_d[id * 2 + 1]) * cos(lonlat_d[id * 2]); AMy = -rx * Mh_d[id * 3 * nv + lev * 3 + 2] + rz * Mh_d[id * 3 * nv + lev * 3 + 0] - Rho_d[id * nv + lev] * Omega * r * rz * cos(lonlat_d[id * 2 + 1]) * sin(lonlat_d[id * 2]); AMz = rx * Mh_d[id * 3 * nv + lev * 3 + 1] - ry * Mh_d[id * 3 * nv + lev * 3 + 0] + Rho_d[id * nv + lev] * Omega * r * r * cos(lonlat_d[id * 2 + 1]) * cos(lonlat_d[id * 2 + 1]); //AMx, AMy should go to zero when integrated over globe // (but in practice, are just much smaller than AMz) //total in control volume AngMomx_d[id * nv + lev] = AMx * Vol; AngMomy_d[id * nv + lev] = AMy * Vol; AngMomz_d[id * nv + lev] = AMz * Vol; } }
1,274
#include "includes.h" __global__ void box_encode_kernel(float * targets_dx, float * targets_dy, float * targets_dw, float * targets_dh, float4 * boxes, float4 * anchors, float wx, float wy, float ww, float wh, size_t gt, size_t idxJump) { int idx = blockIdx.x*blockDim.x + threadIdx.x; size_t row_offset; float anchors_x1,anchors_x2, anchors_y1, anchors_y2, boxes_x1, boxes_x2, boxes_y1, boxes_y2, ex_w, ex_h, ex_ctr_x, ex_ctr_y, gt_w, gt_h, gt_ctr_x, gt_ctr_y; for (int i=idx; i<gt; i+=idxJump){ row_offset = i; anchors_x1 = anchors[row_offset].x; anchors_y1 = anchors[row_offset].y; anchors_x2 = anchors[row_offset].z; anchors_y2 = anchors[row_offset].w; boxes_x1 = boxes[row_offset].x; boxes_y1 = boxes[row_offset].y; boxes_x2 = boxes[row_offset].z; boxes_y2 = boxes[row_offset].w; ex_w = anchors_x2 - anchors_x1 + 1; ex_h = anchors_y2 - anchors_y1 + 1; ex_ctr_x = anchors_x1 + 0.5 * ex_w; ex_ctr_y = anchors_y1 + 0.5 * ex_h; gt_w = boxes_x2 - boxes_x1 + 1; gt_h = boxes_y2 - boxes_y1 + 1; gt_ctr_x = boxes_x1 + 0.5 * gt_w; gt_ctr_y = boxes_y1 + 0.5 * gt_h; targets_dx[i] = wx * (gt_ctr_x-ex_ctr_x)/ex_w; targets_dy[i] = wy * (gt_ctr_y-ex_ctr_y)/ex_h; targets_dw[i] = ww * log(gt_w/ex_w); targets_dh[i] = wh * log(gt_h/ex_h); } }
1,275
#include <iostream> #include <math.h> #include <cuda.h> #include <stdio.h> // function to add the elements of two arrays #define checkCudaErrors(val) check( (val), #val, __FILE__, __LINE__) template<typename T> void check(T err, const char* const func, const char* const file, const int line) { if (err != cudaSuccess) { std::cerr << "CUDA error at: " << file << ":" << line << std::endl; std::cerr << cudaGetErrorString(err) << " " << func << std::endl; exit(1); } } __global__ void add(int n, float *x, float *y) { } int main(void) { int N = 1<<20; // 1M elements float *x = new float[N]; float *y = new float[N]; cudaMallocManaged(&x,N*sizeof(float)); cudaMallocManaged(&y,N*sizeof(float)); // Run kernel on 1M elements on the CPU const dim3 threadsPerBlock(16, 16); const dim3 numBlocks(1024/16, 768/16); add<<<numBlocks,threadsPerBlock>>>(N, x, y); // Wait for GPU to finish before accessing on host cudaDeviceSynchronize(); checkCudaErrors(cudaGetLastError()); // Free memory cudaFree(x); cudaFree(y); // Free memory return 0; }
1,276
#include "cuda.h" #define N 1000 __device__ float A[N][N]; __device__ float B[N][N]; __device__ float C[N][N]; __global__ void vectorAdd(float A[N][N], float B[N][N], float C[N][N]) { int i = threadIdx.x; int j = threadIdx.y; C[i][j] = A[i][j] + B[i][j]; } int main() { int bpg = 1; dim3 tpb(N, N); vectorAdd<<<bpg, tpb>>>(A, B, C); cudaFree(A); cudaFree(B); cudaFree(C); return 0; }
1,277
#include <stdio.h> #include <stdlib.h> #define CSC(call) \ do { \ cudaError_t res = call; \ if (res != cudaSuccess) { \ fprintf(stderr, "ERROR: file:%s line:%d message:%s\n", \ __FILE__, __LINE__, cudaGetErrorString(res)); \ exit(0); \ } \ } while (0) __global__ void kernel(double *devVec, double *revVec, int n) { int idx = threadIdx.x + blockIdx.x * blockDim.x; int offset = blockDim.x + gridDim.x; while (idx < n) { revVec[idx] = devVec[n - 1 - idx]; idx += offset; } } void printVector(double *vec, int size) { for (int i = 0; i < size; ++i) { printf("%f ", vec[i]); } printf("\n"); } int main() { int n; // Vector size double *vec; int blocks = 256, threads = 256; // Set up initial data scanf("%d", &n); vec = (double *)malloc(sizeof(double) * n); for (int i = 0; i < n; ++i) { scanf("%lf", &vec[i]); } // Create CUDA vectors and copy data to first double *devVec, *revVec; CSC(cudaMalloc(&devVec, sizeof(double) * n)); CSC(cudaMalloc(&revVec, sizeof(double) * n)); CSC(cudaMemcpy(devVec, vec, sizeof(double) * n, cudaMemcpyHostToDevice)); // Call kernel kernel<<<blocks, threads>>>(devVec, revVec, n); CSC(cudaGetLastError()); // Copy results from device to host memory CSC(cudaMemcpy(vec, revVec, sizeof(double) * n, cudaMemcpyDeviceToHost)); CSC(cudaFree(devVec)); CSC(cudaFree(revVec)); // Print results for (int i = 0; i < n; ++i) { printf("%.10e ", vec[i]); } free(vec); return 0; }
1,278
#include "includes.h" __global__ void packcoo_kernel(int num_entries, int* row_indices, int* column_indices, int* aggridx, int* partidx, int* partlabel) { int entryidx = blockIdx.x * blockDim.x + threadIdx.x; if(entryidx < num_entries) { int row = row_indices[entryidx]; int col = column_indices[entryidx]; int l = partlabel[row]; int partstart = aggridx[partidx[l]]; unsigned int newindex = row - partstart; newindex <<= 16; newindex += col - partstart; row_indices[entryidx] = newindex; } }
1,279
#include "includes.h" __global__ void glcm_calculation_nol(int *A,int *glcm, const int nx, const int ny,int maxx) { int ix = threadIdx.x + blockIdx.x * blockDim.x; int iy = threadIdx.y + blockIdx.y * blockDim.y; unsigned int idx = iy * nx + ix; //unsigned int idr = iy * (maxx+1) + ix; int k,l; int p; //Calculate GLCM if(idx < nx*ny ){ for(k=0;k<=maxx;k++){ for(l=0;l<=maxx;l++){ if((A[idx]==k) && (A[idx+1]==l)){ p=((maxx+1)*k) +l; atomicAdd(&glcm[p],1); } } } } }
1,280
#include <iostream> #include <stdio.h> #include <stdlib.h> #include "pinnedmem.cuh" cudaError mallocHost(void** h_mem ,uint memSize, memoryMode memMode, bool wc) { if( PINNED == memMode ) { #if CUDART_VERSION >= 2020 return cudaHostAlloc( h_mem, memSize, (wc) ? cudaHostAllocWriteCombined : 0 ); #else if (wc) {printf("Write-Combined unavailable on CUDART_VERSION less than 2020, running is: %d", CUDART_VERSION);} return cudaMallocHost( h_mem, memSize ); #endif } else { // PAGEABLE memory mode *h_mem = malloc( memSize ); } return cudaSuccess; } cudaError freeHost(void* h_mem, memoryMode memMode) { if( PINNED == memMode ) { return cudaFreeHost(h_mem); } else { free(h_mem); } return cudaSuccess; } /* * exitOnError: Show the error message and terminate the application. */ void exitOnError(const char *whereAt) { cudaError_t error = cudaGetLastError(); if(error != cudaSuccess) { // print the CUDA error message and exit printf("CUDA error at %s: %s\n", whereAt, cudaGetErrorString(error)); exit(-1); } }
1,281
#include <cuda_runtime_api.h> #include <device_launch_parameters.h> #include <stdio.h> #include <time.h> // CUDA kernel. Each thread takes care of one element of c __global__ void vecAdd(float *a, float *b, float *c, int n) { // Get our global thread ID int id = blockIdx.x*blockDim.x+threadIdx.x; // Make sure we do not go out of bounds if (id < n) c[id] = a[id] + b[id]; } #include <math.h> void API_add_v2(float *a, float *b, float *c, int data_num) { printf("In function: data nums = %d\n",data_num); // Number of threads in each thread block int blockSize = 1024; // Number of thread blocks in grid int gridSize = (int)ceil((float)data_num/blockSize); vecAdd<<<gridSize, blockSize>>>(a, b, c, data_num); }
1,282
// headers #include <stdio.h> #include <cuda.h> // for CUDA // global variables int inputLength=5; float *hostInput1=NULL; float *hostInput2=NULL; float *hostOutput=NULL; float *deviceInput1=NULL; float *deviceInput2=NULL; float *deviceOutput=NULL; // global kernel function definition __global__ void vecAdd(float *in1,float *in2,float *out,int len) { // variable declarations int i=blockIdx.x * blockDim.x + threadIdx.x; // code if(i < len) { out[i]=in1[i]+in2[i]; } } int main(int argc,char *argv[]) { // function declarations void cleanup(void); // code // allocate host-memory hostInput1=(float *)malloc(inputLength * sizeof(float)); if(hostInput1== NULL) { printf("CPU Memory Fatal Error = Can Not Allocate Enough Memory For Host Input Array 1.\nExitting ...\n"); cleanup(); exit(EXIT_FAILURE); } hostInput2=(float *)malloc(inputLength * sizeof(float)); if(hostInput2== NULL) { printf("CPU Memory Fatal Error = Can Not Allocate Enough Memory For Host Input Array 2.\nExitting ...\n"); cleanup(); exit(EXIT_FAILURE); } hostOutput=(float *)malloc(inputLength * sizeof(float)); if(hostOutput== NULL) { printf("CPU Memory Fatal Error = Can Not Allocate Enough Memory For Host Output Array.\nExitting ...\n"); cleanup(); exit(EXIT_FAILURE); } // fill above input host vectors with arbitary but hard-coded data hostInput1[0]=101.0; hostInput1[1]=102.0; hostInput1[2]=103.0; hostInput1[3]=104.0; hostInput1[4]=105.0; hostInput2[0]=201.0; hostInput2[1]=202.0; hostInput2[2]=203.0; hostInput2[3]=204.0; hostInput2[4]=205.0; // allocate device-memory int size=inputLength * sizeof(float); cudaError_t err=cudaSuccess; err=cudaMalloc((void **)&deviceInput1,size); if(err!=cudaSuccess) { printf("GPU Memory Fatal Error = %s In File Name %s At Line No. %d.\nExitting ...\n",cudaGetErrorString(err),__FILE__,__LINE__); cleanup(); exit(EXIT_FAILURE); } err=cudaMalloc((void **)&deviceInput2,size); if(err!=cudaSuccess) { printf("GPU Memory Fatal Error = %s In File Name %s At Line No. %d.\nExitting ...\n",cudaGetErrorString(err),__FILE__,__LINE__); cudaFree(deviceInput1); cleanup(); exit(EXIT_FAILURE); } err=cudaMalloc((void **)&deviceOutput,size); if(err!=cudaSuccess) { printf("GPU Memory Fatal Error = %s In File Name %s At Line No. %d.\nExitting ...\n",cudaGetErrorString(err),__FILE__,__LINE__); cleanup(); exit(EXIT_FAILURE); } // copy host memory contents to device memory err=cudaMemcpy(deviceInput1,hostInput1,size,cudaMemcpyHostToDevice); if(err!=cudaSuccess) { printf("GPU Memory Fatal Error = %s In File Name %s At Line No. %d.\nExitting ...\n",cudaGetErrorString(err),__FILE__,__LINE__); cleanup(); exit(EXIT_FAILURE); } err=cudaMemcpy(deviceInput2,hostInput2,size,cudaMemcpyHostToDevice); if(err!=cudaSuccess) { printf("GPU Memory Fatal Error = %s In File Name %s At Line No. %d.\nExitting ...\n",cudaGetErrorString(err),__FILE__,__LINE__); cleanup(); exit(EXIT_FAILURE); } // cuda kernel configuration dim3 DimGrid=dim3(ceil(inputLength/256.0),1,1); dim3 DimBlock=dim3(256,1,1); vecAdd<<<DimGrid,DimBlock>>>(deviceInput1,deviceInput2,deviceOutput,inputLength); // copy device memory to host memory err=cudaMemcpy(hostOutput,deviceOutput,size,cudaMemcpyDeviceToHost); if(err!=cudaSuccess) { printf("GPU Memory Fatal Error = %s In File Name %s At Line No. %d.\nExitting ...\n",cudaGetErrorString(err),__FILE__,__LINE__); cleanup(); exit(EXIT_FAILURE); } // results int i; for(i=0;i<inputLength;i++) { printf("%f + %f = %f\n",hostInput1[i],hostInput2[i],hostOutput[i]); } // total cleanup cleanup(); return(0); } void cleanup(void) { // code // free allocated device-memory if(deviceInput1) { cudaFree(deviceInput1); deviceInput1=NULL; } if(deviceInput2) { cudaFree(deviceInput2); deviceInput2=NULL; } if(deviceOutput) { cudaFree(deviceOutput); deviceOutput=NULL; } // free allocated host-memory if(hostInput1) { free(hostInput1); hostInput1=NULL; } if(hostInput2) { free(hostInput2); hostInput2=NULL; } if(hostOutput) { free(hostOutput); hostOutput=NULL; } }
1,283
#include <stdio.h> #include <cuda.h> __global__ void alloutputs(int *counter) { int oldc = atomicAdd(counter, 1); if (*counter == 34) printf("%d\n", oldc); } int main() { int *counter, hcounter = 0; cudaMalloc(&counter, sizeof(int)); cudaMemcpy(counter, &hcounter, sizeof(int), cudaMemcpyHostToDevice); alloutputs<<<1, 34>>>(counter); cudaDeviceSynchronize(); return 0; }
1,284
#include <stdio.h> __global__ void kicache_test4_2 (unsigned int *ts, unsigned int* out, int p1, int p2, int its); __global__ void kicache_test4_4 (unsigned int *ts, unsigned int* out, int p1, int p2, int its); __global__ void kicache_test4_6 (unsigned int *ts, unsigned int* out, int p1, int p2, int its); __global__ void kicache_test4_8 (unsigned int *ts, unsigned int* out, int p1, int p2, int its); __global__ void kicache_test4_10 (unsigned int *ts, unsigned int* out, int p1, int p2, int its); __global__ void kicache_test4_12 (unsigned int *ts, unsigned int* out, int p1, int p2, int its); __global__ void kicache_test4_14 (unsigned int *ts, unsigned int* out, int p1, int p2, int its); __global__ void kicache_test4_16 (unsigned int *ts, unsigned int* out, int p1, int p2, int its); __global__ void kicache_test4_18 (unsigned int *ts, unsigned int* out, int p1, int p2, int its); __global__ void kicache_test4_20 (unsigned int *ts, unsigned int* out, int p1, int p2, int its); __global__ void kicache_test4_22 (unsigned int *ts, unsigned int* out, int p1, int p2, int its); __global__ void kicache_test4_24 (unsigned int *ts, unsigned int* out, int p1, int p2, int its); __global__ void kicache_test4_26 (unsigned int *ts, unsigned int* out, int p1, int p2, int its); __global__ void kicache_test4_28 (unsigned int *ts, unsigned int* out, int p1, int p2, int its); __global__ void kicache_test4_30 (unsigned int *ts, unsigned int* out, int p1, int p2, int its); __global__ void kicache_test4_32 (unsigned int *ts, unsigned int* out, int p1, int p2, int its); __global__ void kicache_test4_34 (unsigned int *ts, unsigned int* out, int p1, int p2, int its); __global__ void kicache_test4_36 (unsigned int *ts, unsigned int* out, int p1, int p2, int its); __global__ void kicache_test4_38 (unsigned int *ts, unsigned int* out, int p1, int p2, int its); __global__ void kicache_test4_40 (unsigned int *ts, unsigned int* out, int p1, int p2, int its); __global__ void kicache_test4_42 (unsigned int *ts, unsigned int* out, int p1, int p2, int its); __global__ void kicache_test4_44 (unsigned int *ts, unsigned int* out, int p1, int p2, int its); __global__ void kicache_test4_46 (unsigned int *ts, unsigned int* out, int p1, int p2, int its); __global__ void kicache_test4_48 (unsigned int *ts, unsigned int* out, int p1, int p2, int its); __global__ void kicache_test4_50 (unsigned int *ts, unsigned int* out, int p1, int p2, int its); __global__ void kicache_test4_52 (unsigned int *ts, unsigned int* out, int p1, int p2, int its); __global__ void kicache_test4_54 (unsigned int *ts, unsigned int* out, int p1, int p2, int its); __global__ void kicache_test4_56 (unsigned int *ts, unsigned int* out, int p1, int p2, int its); __global__ void kicache_test4_58 (unsigned int *ts, unsigned int* out, int p1, int p2, int its); __global__ void kicache_test4_60 (unsigned int *ts, unsigned int* out, int p1, int p2, int its); __global__ void kicache_test4_62 (unsigned int *ts, unsigned int* out, int p1, int p2, int its); __global__ void kicache_test4_64 (unsigned int *ts, unsigned int* out, int p1, int p2, int its); __global__ void kicache_test4_72 (unsigned int *ts, unsigned int* out, int p1, int p2, int its); __global__ void kicache_test4_80 (unsigned int *ts, unsigned int* out, int p1, int p2, int its); __global__ void kicache_test4_88 (unsigned int *ts, unsigned int* out, int p1, int p2, int its); __global__ void kicache_test4_96 (unsigned int *ts, unsigned int* out, int p1, int p2, int its); __global__ void kicache_test4_104 (unsigned int *ts, unsigned int* out, int p1, int p2, int its); __global__ void kicache_test4_112 (unsigned int *ts, unsigned int* out, int p1, int p2, int its); __global__ void kicache_test4_120 (unsigned int *ts, unsigned int* out, int p1, int p2, int its); __global__ void kicache_test4_128 (unsigned int *ts, unsigned int* out, int p1, int p2, int its); __global__ void kicache_test4_136 (unsigned int *ts, unsigned int* out, int p1, int p2, int its); __global__ void kicache_test4_144 (unsigned int *ts, unsigned int* out, int p1, int p2, int its); __global__ void kicache_test4_152 (unsigned int *ts, unsigned int* out, int p1, int p2, int its); __global__ void kicache_test4_160 (unsigned int *ts, unsigned int* out, int p1, int p2, int its); __global__ void kicache_test4_168 (unsigned int *ts, unsigned int* out, int p1, int p2, int its); __global__ void kicache_test4_176 (unsigned int *ts, unsigned int* out, int p1, int p2, int its); __global__ void kicache_test4_184 (unsigned int *ts, unsigned int* out, int p1, int p2, int its); __global__ void kicache_test4_192 (unsigned int *ts, unsigned int* out, int p1, int p2, int its); __global__ void kicache_test4_200 (unsigned int *ts, unsigned int* out, int p1, int p2, int its); __global__ void kicache_test4_208 (unsigned int *ts, unsigned int* out, int p1, int p2, int its); __global__ void kicache_test4_216 (unsigned int *ts, unsigned int* out, int p1, int p2, int its); __global__ void kicache_test4_224 (unsigned int *ts, unsigned int* out, int p1, int p2, int its); __global__ void kicache_test4_232 (unsigned int *ts, unsigned int* out, int p1, int p2, int its); //__global__ void kicache_test4_240 (unsigned int *ts, unsigned int* out, int p1, int p2, int its); //__global__ void kicache_test4_248 (unsigned int *ts, unsigned int* out, int p1, int p2, int its); //__global__ void kicache_test4_256 (unsigned int *ts, unsigned int* out, int p1, int p2, int its); void measure_icache() { unsigned int ts[1024]; // ts, output from kernel. Two elements used per thread. unsigned int *d_ts; unsigned int *d_out; // Unused memory for storing output dim3 Db = dim3(1); dim3 Dg = dim3(1,1,1); // Allocate device array. if (cudaSuccess != cudaMalloc((void**)&d_ts, sizeof(ts))) { printf ("cudaMalloc failed %s:%d\n", __FILE__, __LINE__); return; } if (cudaSuccess != cudaMalloc((void**)&d_out, 4)) { printf ("cudaMalloc failed %s:%d\n", __FILE__, __LINE__); return; } fprintf (stderr, "Running icache test...\n"); // Measure instruction cache size printf ("Instruction cache size:\n"); unsigned int sum_times[32]; printf (" 0.5 KB steps: "); Db.x = 1; for (int p2 = 1; p2 <= 32; p2++) { unsigned int sum_time = 0; bool failed = false; for (int i=0;i<100 && !failed ;i++) { cudaGetLastError(); // Clear previous error code, if any switch (p2) { case 1: kicache_test4_2 <<<Dg, Db>>>(d_ts, d_out, 1, p2, 2); break; case 2: kicache_test4_4 <<<Dg, Db>>>(d_ts, d_out, 1, p2, 2); break; case 3: kicache_test4_6 <<<Dg, Db>>>(d_ts, d_out, 1, p2, 2); break; case 4: kicache_test4_8 <<<Dg, Db>>>(d_ts, d_out, 1, p2, 2); break; case 5: kicache_test4_10 <<<Dg, Db>>>(d_ts, d_out, 1, p2, 2); break; case 6: kicache_test4_12 <<<Dg, Db>>>(d_ts, d_out, 1, p2, 2); break; case 7: kicache_test4_14 <<<Dg, Db>>>(d_ts, d_out, 1, p2, 2); break; case 8: kicache_test4_16 <<<Dg, Db>>>(d_ts, d_out, 1, p2, 2); break; case 9: kicache_test4_18 <<<Dg, Db>>>(d_ts, d_out, 1, p2, 2); break; case 10: kicache_test4_20 <<<Dg, Db>>>(d_ts, d_out, 1, p2, 2); break; case 11: kicache_test4_22 <<<Dg, Db>>>(d_ts, d_out, 1, p2, 2); break; case 12: kicache_test4_24 <<<Dg, Db>>>(d_ts, d_out, 1, p2, 2); break; case 13: kicache_test4_26 <<<Dg, Db>>>(d_ts, d_out, 1, p2, 2); break; case 14: kicache_test4_28 <<<Dg, Db>>>(d_ts, d_out, 1, p2, 2); break; case 15: kicache_test4_30 <<<Dg, Db>>>(d_ts, d_out, 1, p2, 2); break; case 16: kicache_test4_32 <<<Dg, Db>>>(d_ts, d_out, 1, p2, 2); break; case 17: kicache_test4_34 <<<Dg, Db>>>(d_ts, d_out, 1, p2, 2); break; case 18: kicache_test4_36 <<<Dg, Db>>>(d_ts, d_out, 1, p2, 2); break; case 19: kicache_test4_38 <<<Dg, Db>>>(d_ts, d_out, 1, p2, 2); break; case 20: kicache_test4_40 <<<Dg, Db>>>(d_ts, d_out, 1, p2, 2); break; case 21: kicache_test4_42 <<<Dg, Db>>>(d_ts, d_out, 1, p2, 2); break; case 22: kicache_test4_44 <<<Dg, Db>>>(d_ts, d_out, 1, p2, 2); break; case 23: kicache_test4_46 <<<Dg, Db>>>(d_ts, d_out, 1, p2, 2); break; case 24: kicache_test4_48 <<<Dg, Db>>>(d_ts, d_out, 1, p2, 2); break; case 25: kicache_test4_50 <<<Dg, Db>>>(d_ts, d_out, 1, p2, 2); break; case 26: kicache_test4_52 <<<Dg, Db>>>(d_ts, d_out, 1, p2, 2); break; case 27: kicache_test4_54 <<<Dg, Db>>>(d_ts, d_out, 1, p2, 2); break; case 28: kicache_test4_56 <<<Dg, Db>>>(d_ts, d_out, 1, p2, 2); break; case 29: kicache_test4_58 <<<Dg, Db>>>(d_ts, d_out, 1, p2, 2); break; case 30: kicache_test4_60 <<<Dg, Db>>>(d_ts, d_out, 1, p2, 2); break; case 31: kicache_test4_62 <<<Dg, Db>>>(d_ts, d_out, 1, p2, 2); break; case 32: kicache_test4_64 <<<Dg, Db>>>(d_ts, d_out, 1, p2, 2); break; } if (cudaGetLastError() != cudaSuccess) { failed = true; break; } cudaThreadSynchronize(); cudaMemcpy(ts, d_ts, sizeof(ts), cudaMemcpyDeviceToHost); sum_time += (ts[1] - ts[0]); } if (failed) { printf ("xxxx "); } else { // Compute average latency over the lifetime of each warp (sum_time), and average throughput of the kernel (sum_max_time). printf ("%.5f ", sum_time/100.0/(p2*64)); fflush(stdout); } sum_times[p2-1] = sum_time; } printf (" (icache = "); for (int last_i=1, i=1;i<32;i++) { //printf("sum time %d, %d \n",sum_times[i]/(i+1),sum_times[last_i]/(last_i+1)*1.33); if (sum_times[i]/(i+1) > sum_times[last_i]/(last_i+1) /**1.33*/) { printf ("%.1fKB ", i*0.5); last_i = i; } } printf (")\n"); printf (" 2 KB steps: "); Db.x = 1; for (int p2 = 1; p2 <= 29/*32*/; p2++) { unsigned int sum_time = 0; bool failed = false; for (int i=0;i<100 && !failed ;i++) { cudaGetLastError(); // Clear previous error code, if any switch (p2) { case 1: kicache_test4_8 <<<Dg, Db>>>(d_ts, d_out, 1, p2, 2); break; case 2: kicache_test4_16 <<<Dg, Db>>>(d_ts, d_out, 1, p2, 2); break; case 3: kicache_test4_24 <<<Dg, Db>>>(d_ts, d_out, 1, p2, 2); break; case 4: kicache_test4_32 <<<Dg, Db>>>(d_ts, d_out, 1, p2, 2); break; case 5: kicache_test4_40 <<<Dg, Db>>>(d_ts, d_out, 1, p2, 2); break; case 6: kicache_test4_48 <<<Dg, Db>>>(d_ts, d_out, 1, p2, 2); break; case 7: kicache_test4_56 <<<Dg, Db>>>(d_ts, d_out, 1, p2, 2); break; case 8: kicache_test4_64 <<<Dg, Db>>>(d_ts, d_out, 1, p2, 2); break; case 9: kicache_test4_72 <<<Dg, Db>>>(d_ts, d_out, 1, p2, 2); break; case 10: kicache_test4_80 <<<Dg, Db>>>(d_ts, d_out, 1, p2, 2); break; case 11: kicache_test4_88 <<<Dg, Db>>>(d_ts, d_out, 1, p2, 2); break; case 12: kicache_test4_96 <<<Dg, Db>>>(d_ts, d_out, 1, p2, 2); break; case 13: kicache_test4_104 <<<Dg, Db>>>(d_ts, d_out, 1, p2, 2); break; case 14: kicache_test4_112 <<<Dg, Db>>>(d_ts, d_out, 1, p2, 2); break; case 15: kicache_test4_120 <<<Dg, Db>>>(d_ts, d_out, 1, p2, 2); break; case 16: kicache_test4_128 <<<Dg, Db>>>(d_ts, d_out, 1, p2, 2); break; case 17: kicache_test4_136 <<<Dg, Db>>>(d_ts, d_out, 1, p2, 2); break; case 18: kicache_test4_144 <<<Dg, Db>>>(d_ts, d_out, 1, p2, 2); break; case 19: kicache_test4_152 <<<Dg, Db>>>(d_ts, d_out, 1, p2, 2); break; case 20: kicache_test4_160 <<<Dg, Db>>>(d_ts, d_out, 1, p2, 2); break; case 21: kicache_test4_168 <<<Dg, Db>>>(d_ts, d_out, 1, p2, 2); break; case 22: kicache_test4_176 <<<Dg, Db>>>(d_ts, d_out, 1, p2, 2); break; case 23: kicache_test4_184 <<<Dg, Db>>>(d_ts, d_out, 1, p2, 2); break; case 24: kicache_test4_192 <<<Dg, Db>>>(d_ts, d_out, 1, p2, 2); break; case 25: kicache_test4_200 <<<Dg, Db>>>(d_ts, d_out, 1, p2, 2); break; case 26: kicache_test4_208 <<<Dg, Db>>>(d_ts, d_out, 1, p2, 2); break; case 27: kicache_test4_216 <<<Dg, Db>>>(d_ts, d_out, 1, p2, 2); break; case 28: kicache_test4_224 <<<Dg, Db>>>(d_ts, d_out, 1, p2, 2); break; case 29: kicache_test4_232 <<<Dg, Db>>>(d_ts, d_out, 1, p2, 2); break; //case 30: kicache_test4_240 <<<Dg, Db>>>(d_ts, d_out, 1, p2, 2); break; //case 31: kicache_test4_248 <<<Dg, Db>>>(d_ts, d_out, 1, p2, 2); break; //case 32: kicache_test4_256 <<<Dg, Db>>>(d_ts, d_out, 1, p2, 2); break; } if (cudaGetLastError() != cudaSuccess) { failed = true; break; } cudaThreadSynchronize(); cudaMemcpy(ts, d_ts, sizeof(ts), cudaMemcpyDeviceToHost); sum_time += (ts[1] - ts[0]); } if (failed) { printf ("xxxx "); } else { // Compute average latency over the lifetime of each warp (sum_time), and average throughput of the kernel (sum_max_time). printf ("%.1f ", sum_time/100.0/(p2*256)); fflush(stdout); } sum_times[p2-1] = sum_time; } printf (" (icache = "); for (int last_i=1, i=1;i<28/*32*/;i++) { if (sum_times[i]/(i+1) > sum_times[last_i]/(last_i+1) /**1.33*/) { printf ("%.1fKB ", i*2.0); last_i = i; } } printf (")\n"); printf ("\n Test instruction cache sharing by running two thread blocks concurently.\n"); Db.x = 1; dim3 Dg2 = dim3(31,1,1); for (int blk2 = 1; blk2 <31; blk2+= (blk2 == 1? 9 : 10)) { printf (" TPC 0,%d (2 KB steps): ", blk2); int mask = (1<<blk2) | 1; // Enable two blocks for execution for (int p2 = 1; p2 <= 28/*32*/; p2++) { unsigned int sum_time = 0; bool failed = false; for (int i=0;i<50 && !failed ;i++) { cudaGetLastError(); // Clear previous error code, if any switch (p2) { case 1: kicache_test4_8 <<<Dg2, Db>>>(d_ts, d_out, mask, p2, 2); break; case 2: kicache_test4_16 <<<Dg2, Db>>>(d_ts, d_out, mask, p2, 2); break; case 3: kicache_test4_24 <<<Dg2, Db>>>(d_ts, d_out, mask, p2, 2); break; case 4: kicache_test4_32 <<<Dg2, Db>>>(d_ts, d_out, mask, p2, 2); break; case 5: kicache_test4_40 <<<Dg2, Db>>>(d_ts, d_out, mask, p2, 2); break; case 6: kicache_test4_48 <<<Dg2, Db>>>(d_ts, d_out, mask, p2, 2); break; case 7: kicache_test4_56 <<<Dg2, Db>>>(d_ts, d_out, mask, p2, 2); break; case 8: kicache_test4_64 <<<Dg2, Db>>>(d_ts, d_out, mask, p2, 2); break; case 9: kicache_test4_72 <<<Dg2, Db>>>(d_ts, d_out, mask, p2, 2); break; case 10: kicache_test4_80 <<<Dg2, Db>>>(d_ts, d_out, mask, p2, 2); break; case 11: kicache_test4_88 <<<Dg2, Db>>>(d_ts, d_out, mask, p2, 2); break; case 12: kicache_test4_96 <<<Dg2, Db>>>(d_ts, d_out, mask, p2, 2); break; case 13: kicache_test4_104 <<<Dg2, Db>>>(d_ts, d_out, mask, p2, 2); break; case 14: kicache_test4_112 <<<Dg2, Db>>>(d_ts, d_out, mask, p2, 2); break; case 15: kicache_test4_120 <<<Dg2, Db>>>(d_ts, d_out, mask, p2, 2); break; case 16: kicache_test4_128 <<<Dg2, Db>>>(d_ts, d_out, mask, p2, 2); break; case 17: kicache_test4_136 <<<Dg2, Db>>>(d_ts, d_out, mask, p2, 2); break; case 18: kicache_test4_144 <<<Dg2, Db>>>(d_ts, d_out, mask, p2, 2); break; case 19: kicache_test4_152 <<<Dg2, Db>>>(d_ts, d_out, mask, p2, 2); break; case 20: kicache_test4_160 <<<Dg2, Db>>>(d_ts, d_out, mask, p2, 2); break; case 21: kicache_test4_168 <<<Dg2, Db>>>(d_ts, d_out, mask, p2, 2); break; case 22: kicache_test4_176 <<<Dg2, Db>>>(d_ts, d_out, mask, p2, 2); break; case 23: kicache_test4_184 <<<Dg2, Db>>>(d_ts, d_out, mask, p2, 2); break; case 24: kicache_test4_192 <<<Dg2, Db>>>(d_ts, d_out, mask, p2, 2); break; case 25: kicache_test4_200 <<<Dg2, Db>>>(d_ts, d_out, mask, p2, 2); break; case 26: kicache_test4_208 <<<Dg2, Db>>>(d_ts, d_out, mask, p2, 2); break; case 27: kicache_test4_216 <<<Dg2, Db>>>(d_ts, d_out, mask, p2, 2); break; case 28: kicache_test4_224 <<<Dg2, Db>>>(d_ts, d_out, mask, p2, 2); break; //case 29: kicache_test4_232 <<<Dg2, Db>>>(d_ts, d_out, mask, p2, 2); break; //case 30: kicache_test4_240 <<<Dg2, Db>>>(d_ts, d_out, mask, p2, 2); break; //case 31: kicache_test4_248 <<<Dg2, Db>>>(d_ts, d_out, mask, p2, 2); break; //case 32: kicache_test4_256 <<<Dg2, Db>>>(d_ts, d_out, mask, p2, 2); break; } if (cudaGetLastError() != cudaSuccess) { failed = true; break; } cudaThreadSynchronize(); cudaMemcpy(ts, d_ts, sizeof(ts), cudaMemcpyDeviceToHost); sum_time += (ts[1] - ts[0]); } if (failed) { printf ("xxxx "); } else { printf ("%.1f ", sum_time/50.0/(p2*256)); fflush(stdout); } sum_times[p2-1] = sum_time; } printf (" (apparent icache = "); for (int last_i=1, i=1;i<32;i++) { if (sum_times[i]/(i+1) > sum_times[last_i]/(last_i+1) *1.25) { printf ("%.1fKB ", i*2.0); last_i = i; } } printf (")\n"); } printf ("\n"); cudaFree(d_ts); cudaFree(d_out); } int main() { measure_icache(); }
1,285
#include "includes.h" __device__ size_t GIDX(size_t row, size_t col, int H, int W) { return row * W + col; } __global__ void kernel_blur(float* d_I, float* d_Ib, int H, int W) { size_t row = threadIdx.y + blockDim.y * blockIdx.y; size_t col = threadIdx.x + blockDim.x * blockIdx.x; size_t idx = GIDX(row, col, H, W); if (row >= H - KERN_RADIUS || row <= KERN_RADIUS || col >= W - KERN_RADIUS || col <= KERN_RADIUS) { return; } int count = 0; for (int i = -KERN_RADIUS; i <= KERN_RADIUS; i++) { for (int j = -KERN_RADIUS; j <= KERN_RADIUS; j++) { d_Ib[idx] += d_I[GIDX(row + i, col + j, H, W)] * gaussian_kernel[count]; count++; } } }
1,286
/** * @file markFilterEdge.cu * @date Spring 2020, revised Spring 2021 * @author Hugo De Moraes */ #include <stdio.h> #include <stdlib.h> __global__ void markFilterEdges_gpu(int * src, int * dst, int * matches, int * keepEdges, int numEdges) { // Get Thread ID const int NUM_THREADS = blockDim.x * gridDim.x; const int COL = blockIdx.x * blockDim.x + threadIdx.x; const int ROW = blockIdx.y * blockDim.y + threadIdx.y; const int FIRST_T_ID = COL + ROW * NUM_THREADS; for(int curTID = FIRST_T_ID; curTID <= numEdges; curTID += NUM_THREADS) { if(matches[src[curTID]] != -1 || matches[dst[curTID]] != -1) { keepEdges[curTID] = 0; } else { keepEdges[curTID] = 1; } } }
1,287
extern "C" { __global__ void DmeanSquareLoss(const int lengthx, const double pref, const double *gradc, const double *x,const double *y, double *gradn ) { int i = threadIdx.x + blockIdx.x * blockDim.x; if (i<lengthx) { gradn[i] += pref * gradc[0] * (x[i]-y[i]); } } }
1,288
#include "includes.h" __global__ void chol_kernel_cudaUFMG_division(float * U, int elem_per_thr) { // Get a thread identifier int tx = blockIdx.x * blockDim.x + threadIdx.x; int ty = blockIdx.y * blockDim.y + threadIdx.y; int tn = ty * blockDim.x * gridDim.x + tx; //#define DEBUGDIV #ifdef DEBUGDIV int dbg = 0; if(blockIdx.x == 4){ if(blockIdx.y == 5){ if(threadIdx.x == 2){ if(threadIdx.y == 1){ dbg = 1; printf("\n\n"); printf("\ntx=%d \nty=%d \ntn=%d", tx, ty, tn); } } } } #endif for(unsigned i=0;i<elem_per_thr;i++){ int iel = tn * elem_per_thr + i; int xval = iel % MATRIX_SIZE; int yval = iel / MATRIX_SIZE; if(xval == yval){ continue; } #ifdef DEBUGDIV if(dbg == 1){ if(i==37){ printf("\niel=%d \nxval=%d \nyval=%d", iel, xval, yval); } } #endif // if on the lower diagonal... if(yval > xval){ xval = MATRIX_SIZE - xval - 1; yval = MATRIX_SIZE - yval - 1; } int iU = xval + yval * MATRIX_SIZE; int iDiag = yval + yval * MATRIX_SIZE; #ifdef DEBUGDIV if(dbg == 1){ if(i==37){ printf("\nxtrans=%d \nytrans=%d \niU=%d \niDiag=%d", xval, yval, iU, iDiag); printf("\n\n"); } } #endif U[iU] /= U[iDiag]; } }
1,289
//Based on the work of Andrew Krepps #include <stdio.h> #include <stdlib.h> //srand and rand #include <math.h> // add function d __global__ void add(int *a, int *b, int *c, int n) { // Get our global thread ID int id = blockIdx.x*blockDim.x+threadIdx.x; // Make sure we do not go out of bounds if (id < n) c[id] = a[id] + b[id]; } __global__ void subtract(int *a, int *b, int *c, int n){ // Get our global thread ID int id = blockIdx.x*blockDim.x+threadIdx.x; // Make sure we do not go out of bounds if (id < n) c[id] = a[id] - b[id]; } __global__ void mult(int *a, int *b, int *c, int n){ // Get our global thread ID int id = blockIdx.x*blockDim.x+threadIdx.x; // Make sure we do not go out of bounds if (id < n) c[id] = a[id] * b[id]; } __global__ void mod(int *a, int *b, int *c, int n){ // Get our global thread ID int id = blockIdx.x*blockDim.x+threadIdx.x; // Make sure we do not go out of bounds if (id < n) c[id] = a[id] % b[id]; } int main(int argc, char** argv) { // read command line arguments int totalThreads = (1 << 20); int blockSize = 256; if (argc >= 2) { totalThreads = atoi(argv[1]); } if (argc >= 3) { blockSize = atoi(argv[2]); } int numBlocks = totalThreads/blockSize; printf("Using %d Threads and %d BlockSize\n",totalThreads, blockSize); // validate command line arguments if (totalThreads % blockSize != 0) { ++numBlocks; totalThreads = numBlocks*blockSize; printf("Warning: Total thread count is not evenly divisible by the block size\n"); printf("The total number of threads will be rounded up to %d\n", totalThreads); } // Host input vectors int *h_a, *h_b; //Host output vectors for different functions "h_c_func" int *h_c_add,*h_c_sub,*h_c_mult,*h_c_mod; // Device input vectors int *d_a, *d_b; //Device output vector int *d_c_add,*d_c_sub,*d_c_mult,*d_c_mod; // Size, in bytes, of each vector size_t bytes = totalThreads*sizeof(int); // Allocate memory for each vector on host h_a = (int*)malloc(bytes); h_b = (int*)malloc(bytes); h_c_add = (int*)malloc(bytes); h_c_sub = (int*)malloc(bytes); h_c_mult = (int*)malloc(bytes); h_c_mod = (int*)malloc(bytes); // Allocate memory for each vector on GPU cudaMalloc(&d_a, bytes); cudaMalloc(&d_b, bytes); cudaMalloc(&d_c_add, bytes); cudaMalloc(&d_c_sub, bytes); cudaMalloc(&d_c_mult, bytes); cudaMalloc(&d_c_mod, bytes); //initialize the input vectors for(int i = 0;i<totalThreads;i++){ //first array is 0 through number of threads h_a[i] = i; // second array is a random number between 0 and 3 h_b[i] = rand() % 4; } //copy both input arrays from host to device cudaMemcpy( d_a, h_a, bytes, cudaMemcpyHostToDevice); cudaMemcpy( d_b, h_b, bytes, cudaMemcpyHostToDevice); //performing add function printf("Performing Add function\n"); add<<<numBlocks, totalThreads>>>(d_a, d_b, d_c_add, totalThreads); //performing subtract function printf("Performing subtract function\n"); subtract<<<numBlocks, totalThreads>>>(d_a, d_b, d_c_sub, totalThreads); //performing mult function printf("Performing mult function\n"); mult<<<numBlocks, totalThreads>>>(d_a, d_b, d_c_mult, totalThreads); //performing mod fuction printf("Performing mod function\n"); mod<<<numBlocks, totalThreads>>>(d_a, d_b, d_c_mod, totalThreads); //copy the output arrays from device to host cudaMemcpy( h_c_add, d_c_add, bytes, cudaMemcpyDeviceToHost); cudaMemcpy( h_c_sub, d_c_sub, bytes, cudaMemcpyDeviceToHost); cudaMemcpy( h_c_mult, d_c_mult, bytes, cudaMemcpyDeviceToHost); cudaMemcpy( h_c_mod, d_c_mod, bytes, cudaMemcpyDeviceToHost); //free up space on our GPU cudaFree(d_c_add); cudaFree(d_c_sub); cudaFree(d_c_mult); cudaFree(d_c_add); return 0; }
1,290
// filename: gaxpy2.cu // a simple CUDA kernel to add two vectors extern "C" // ensure function name to be exactly "gaxpy2" { __global__ void gaxpy4(const int n, const double *a, const double *b, double *c) { int i = threadIdx.x + blockIdx.x*blockDim.x; if (i < n) { c[i] = (double) i; // REMEMBER ZERO INDEXING IN C LANGUAGE!! } } }
1,291
#include<cstdio> #define S 64 #define ZERO 0 #define PI 3.14159265 #define LOW 9 #define HIGH 18 #define QUEUE_SIZE 128 #define KERNEL_RADIUS 8 extern "C" { __device__ bool btwn(int a, int x, int y){ return (a>=x && a<y); } __device__ void load_to_shared(int* src, int cache[][S], int th_x, int th_y, int n, int m){ int val, pos, ind_x, ind_y; if(threadIdx.x==0 && threadIdx.y==0){ ind_x = th_x-1; ind_y = th_y-1; if(btwn(ind_x, 0, m) && btwn(ind_y, 0, n)){ pos = ind_y*m + ind_x; val = src[pos]; } cache[threadIdx.y][threadIdx.x] = val; } if(threadIdx.x==0 && threadIdx.y==31){ ind_x = th_x-1; ind_y = th_y+1; if(btwn(ind_x, 0, m) && btwn(ind_y, 0, n)){ pos = ind_y*m + ind_x; val = src[pos]; } cache[threadIdx.y+2][threadIdx.x] = val; } if(threadIdx.x==31 && threadIdx.y==0){ ind_x = th_x+1; ind_y = th_y-1; if(btwn(ind_x, 0, m) && btwn(ind_y, 0, n)){ pos = ind_y*m + ind_x; val = src[pos]; } cache[threadIdx.y][threadIdx.x+2] = val; } if(threadIdx.x==31 && threadIdx.y==31){ ind_x = th_x+1; ind_y = th_y-1; if(btwn(ind_x, 0, m) && btwn(ind_y, 0, n)){ pos = ind_y*m + ind_x; val = src[pos]; } cache[threadIdx.y+2][threadIdx.x+2] = val; } if(threadIdx.y==0){ ind_x = th_x; ind_y = th_y-1; val=ZERO; if(btwn(ind_x, 0, m) && btwn(ind_y, 0, n)){ pos = ind_y*m + ind_x; val = src[pos]; } cache[threadIdx.y][threadIdx.x+1] = val; } if(threadIdx.y==31){ ind_x = th_x; ind_y = th_y+1; val=ZERO; if(btwn(ind_x, 0, m) && btwn(ind_y, 0, n)){ pos = ind_y*m + ind_x; val = src[pos]; } cache[threadIdx.y+2][threadIdx.x+1] = val; } if(threadIdx.x==0){ ind_x = th_x-1; ind_y = th_y; val=ZERO; if(btwn(ind_x, 0, m) && btwn(ind_y, 0, n)){ pos = ind_y*m + ind_x; val = src[pos]; } cache[threadIdx.y+1][threadIdx.x] = val; } if(threadIdx.x==31){ ind_x = th_x+1; ind_y = th_y; val=ZERO; if(btwn(ind_x, 0, m) && btwn(ind_y, 0, n)){ pos = ind_y*m + ind_x; val = src[pos]; } cache[threadIdx.y+1][threadIdx.x+2] = val; } } __global__ void sobel(int* src, int* dstMagni, float * arcTangensOut){ __shared__ int cache[34][S]; int m = gridDim.x*32, n = gridDim.y*32, th_x = blockIdx.x * 32 + threadIdx.x, th_y = blockIdx.y * 32 + threadIdx.y, i_src = th_y*m + th_x, ind_x, ind_y, magn_x, magn_y, magnAbs_x, magnAbs_y; /*now we load to shared with a frame of thickness eq 1*/ cache[threadIdx.y+1][threadIdx.x+1] = src[i_src]; load_to_shared(src, cache, th_x, th_y, n, m); ind_y = threadIdx.y+1; ind_x = threadIdx.x+1; __syncthreads(); magn_x = cache[ind_y][ind_x-1] - cache[ind_y][ind_x+1]; magnAbs_x = ((magn_x>0) ? magn_x : -magn_x); magn_y = cache[ind_y+1][ind_x] - cache[ind_y-1][ind_x]; magnAbs_y = ((magn_y>0) ? magn_y : -magn_y); dstMagni[i_src] = magnAbs_x + magnAbs_y; arcTangensOut[i_src] = atan2((float) magn_y,(float) magn_x) * 180 / PI; } __global__ void nonMaximalSupression(int * magn, float * arcTangens, int * dest) { __shared__ int cacheMagn[34][S]; int m = gridDim.x*32, n = gridDim.y*32, th_x = blockIdx.x * 32 + threadIdx.x, th_y = blockIdx.y * 32 + threadIdx.y, i_src = th_y*m + th_x, ind_x, ind_y; float angle; cacheMagn[threadIdx.y+1][threadIdx.x+1] = magn[i_src]; load_to_shared(magn, cacheMagn, th_x, th_y, n, m); ind_y = threadIdx.y+1; ind_x = threadIdx.x+1; __syncthreads(); angle = arcTangens[i_src]; if (angle < 0) angle = 360 + angle; //north && south int centerCell = cacheMagn[ind_y][ind_x]; dest[i_src] = centerCell; if ((337.5 <= angle || angle < 22.5) || (157.25 <= angle && angle < 202.5)) { if (cacheMagn[ind_y][ind_x+1] > centerCell || cacheMagn[ind_y][ind_x-1] > centerCell) dest[i_src] = 0; } // north-east && south-west else if ((22.5 <= angle && angle < 67.5) || (202.5 <= angle && angle < 247.5)) { if (cacheMagn[ind_y-1][ind_x+1] > centerCell || cacheMagn[ind_y+1][ind_x-1] > centerCell) dest[i_src] = 0; } // west && east else if ((67.5 <= angle && angle < 112.5) || (247.5 <= angle && angle < 292.5)) { if (cacheMagn[ind_y+1][ind_x] > centerCell || cacheMagn[ind_y-1][ind_x] > centerCell) dest[i_src] = 0; } // west-north && east-south else if ((112.5 <= angle && angle < 157.5) || (292.5 <= angle || angle < 337.5)) { if (cacheMagn[ind_y-1][ind_x-1] > centerCell || cacheMagn[ind_y+1][ind_x+1] > centerCell) dest[i_src] = 0; } } __global__ void prepareBfs(int* src){ int m = gridDim.x*32; int th_x = blockIdx.x * 32 + threadIdx.x; int th_y = blockIdx.y * 32 + threadIdx.y; int i_src = th_y*m + th_x; int val = src[i_src]; if(val < LOW){ src[i_src] = 0; } else if(val >= HIGH){ src[i_src] = -2; } else{ src[i_src] = -1; } } __global__ void oneBfs(int* src, int* dst, int* changed){ __shared__ int cache[34][S]; int m = gridDim.x*32, n = gridDim.y*32, th_x = blockIdx.x * 32 + threadIdx.x, th_y = blockIdx.y * 32 + threadIdx.y, i_src = th_y*m + th_x, ind_y = threadIdx.y+1, ind_x = threadIdx.x+1, queue[QUEUE_SIZE], beg=0, end=0, val, procInd_x, procInd_y, x_new, y_new; cache[threadIdx.y+1][threadIdx.x+1] = src[i_src]; val = cache[ind_y][ind_x]; load_to_shared(src, cache, th_x, th_y, n, m); __syncthreads(); if(val==-1){ for(int i=-1; i<2; ++i){ for(int j=-1; j<2; ++j){ procInd_x = ind_x+i; procInd_y = ind_y+j; if(cache[procInd_y][procInd_x]==-2){ queue[end++] = ind_y; queue[end++] = ind_x; cache[ind_y][ind_x]=-2; *changed=1; i=2; j=2; } } } } while(beg!=end){ procInd_y = queue[beg++]; procInd_x = queue[beg++]; for(int i=-1; i<2; ++i){ for(int j=-1; j<2; ++j){ x_new = procInd_x+i; y_new = procInd_y+j; if(cache[y_new][x_new]==-1 && btwn(y_new, 1, 33) && btwn(x_new, 1, 33)){ queue[end++] = y_new; queue[end++] = x_new; cache[y_new][x_new]=-2; } } } } __syncthreads(); dst[i_src] = cache[ind_y][ind_x]; } __global__ void final_battle(int* src){ int m = gridDim.x*32; int th_x = blockIdx.x * 32 + threadIdx.x; int th_y = blockIdx.y * 32 + threadIdx.y; int i_src = th_y*m + th_x; int val = src[i_src]; if(val==-2){ src[i_src] = 255; } else{ src[i_src] = 0; } } __global__ void gaussianFilter(int * src, int * dest) { __shared__ int cache[34][S]; int n = gridDim.y*32; int m = gridDim.x*32; int th_x = blockIdx.x * 32 + threadIdx.x; int th_y = blockIdx.y * 32 + threadIdx.y; int i_src = th_y*m + th_x; int ind_y, ind_x; cache[threadIdx.y+1][threadIdx.x+1] = src[i_src]; load_to_shared(src, cache, th_x, th_y, n, m); ind_y = threadIdx.y+1+KERNEL_RADIUS; ind_x = threadIdx.x+1+KERNEL_RADIUS; __syncthreads(); int sum = 0; for (int i = -KERNEL_RADIUS; i <= KERNEL_RADIUS; i++) { for (int j = -KERNEL_RADIUS; j <= KERNEL_RADIUS; j++) { sum += cache[ind_y+i][ind_x+j]; //d_kernel[KERNEL_RADIUS + j] } } dest[i_src] = (int) sum; } }
1,292
#include <cstdio> #include <functional> #include <iostream> #include <random> #define BLOCKSIZE 256 void FillMatrix(float* matrix, int height, int width) { std::mt19937 gen(time(0)); std::uniform_real_distribution<float> distribution(-1.0f, 1.0f); auto generate = std::bind(distribution, gen); for (int i = 0; i < height * width; ++i) { matrix[i] = generate(); } } void PrintMatrix(float *matrix, int height, int width) { for (int i = 0; i < height; ++i) { for (int j = 0; j < width; ++j) { std::cout << matrix[i * width + j] << " "; } std::cout << std::endl; } } __global__ void kernel_mul(float *A, float *B, float *C, int mid_size) { int i = blockIdx.x * blockDim.x + threadIdx.x; int j = blockIdx.y * blockDim.y + threadIdx.y; int height = blockDim.x * gridDim.x; int width = blockDim.y * gridDim.y; C[i * width + j] = 0.0f; for (int k = 0; k < mid_size; ++k) { C[i * width + j] += A[i * mid_size + k] * B[k * width + j]; } } __global__ void kernel_my_mul(float *A, float *B, float *C, int mid_size) { int i = blockIdx.x * blockDim.x + threadIdx.x; int j = blockIdx.y * blockDim.y + threadIdx.y; int height = blockDim.x * gridDim.x; int width = blockDim.y * gridDim.y; __shared__ float block_a[BLOCKSIZE]; __shared__ float block_b[BLOCKSIZE]; block_a[threadIdx.y] = A[i * width + threadIdx.y]; block_b[threadIdx.x] = B[threadIdx.x * height + j]; __syncthreads(); C[i * width + j] = 0.0f; for (int k = 0; k < mid_size; ++k) { C[i * width + j] += block_a[k] * block_b[k]; } } void try_both_multiplications(float *h_A, float *h_B, float *h_C) { float* d_A; float* d_B; float* d_C; cudaMalloc(&d_A, sizeof(float) * 128 * 384); cudaMalloc(&d_B, sizeof(float) * 384 * 256); cudaMalloc(&d_C, sizeof(float) * 128 * 256); cudaMemcpy(d_A, h_A, sizeof(float) * 128 * 384, cudaMemcpyHostToDevice); cudaMemcpy(d_B, h_B, sizeof(float) * 384 * 256, cudaMemcpyHostToDevice); // kernel call dim3 num_blocks(8, 16); dim3 block_size(16, 16); cudaEvent_t start; cudaEvent_t stop; cudaEventCreate(&start); cudaEventCreate(&stop); cudaEventRecord(start); kernel_mul<<<num_blocks, block_size>>>(d_A, d_B, d_C, 384); cudaEventRecord(stop); cudaMemcpy(h_C, d_C, sizeof(float) * 128 * 256, cudaMemcpyDeviceToHost); cudaEventSynchronize(stop); float milliseconds = 0; cudaEventElapsedTime(&milliseconds, start, stop); std::cout << "banal elapsed in " << milliseconds << std::endl; PrintMatrix(h_C, 128, 256); cudaEvent_t start2; cudaEvent_t stop2; cudaEventCreate(&start2); cudaEventCreate(&stop2); cudaEventRecord(start2); kernel_my_mul<<<num_blocks, block_size>>>(d_A, d_B, d_C, 384); cudaEventRecord(stop2); cudaMemcpy(h_C, d_C, sizeof(float) * 128 * 256, cudaMemcpyDeviceToHost); cudaEventSynchronize(stop2); float milliseconds2 = 0; cudaEventElapsedTime(&milliseconds2, start2, stop2); std::cout << "shared memory elapsed in " << milliseconds2 << std::endl; PrintMatrix(h_C, 128, 256); cudaFree(d_A); cudaFree(d_B); cudaFree(d_C); } int main() { float *h_A; float *h_B; float *h_C; // h_A 128 * 384 // h_B 384 * 256 // h_C 128 * 256 h_A = new float[128 * 384]; h_B = new float[384 * 256]; h_C = new float[128 * 256]; FillMatrix(h_A, 128, 384); FillMatrix(h_B, 384, 256); try_both_multiplications(h_A, h_B, h_C); delete[] h_A; delete[] h_B; delete[] h_C; return 0; }
1,293
#include <stdio.h> #include <stdlib.h> #include <cuda.h> #include <cuda_runtime.h> static const int N = 10; #define CHECK_STATUS(status) \ if (status != cudaSuccess) \ fprintf(stderr, "File: %s\nLine:%d Function:%s>>>%s\n", __FILE__, __LINE__, __FUNCTION__,\ cudaGetErrorString(status)) // __global__ void VecAdd(float *A, float *B, float *C) { int i = threadIdx.x; C[i] = A[i] + B[i]; } int main(int argc, char **argv) { CHECK_STATUS(cudaSetDevice(0)); float a[N]; float b[N]; for(int i = 0;i<N;i++){ a[i] = i; b[i] = i; } float c[N]; float *d_a,*d_b,*d_c; //分配显存 CHECK_STATUS(cudaMalloc(&d_a, N*sizeof(float))); CHECK_STATUS(cudaMalloc(&d_b, N*sizeof(float))); CHECK_STATUS(cudaMalloc(&d_c, N*sizeof(float))); // 把数据从内存复制到显存 CHECK_STATUS(cudaMemcpy(d_a,a,N* sizeof(float),cudaMemcpyHostToDevice)); CHECK_STATUS(cudaMemcpy(d_b,b,N* sizeof(float),cudaMemcpyHostToDevice)); // 调用kernel VecAdd<<<1,N>>>(d_a,d_b,d_c); // 检查错误 CHECK_STATUS(cudaGetLastError()); // 从显存把数据复制到内存 CHECK_STATUS(cudaMemcpy(c,d_c,N* sizeof(float),cudaMemcpyDeviceToHost)); // 打印 for(int i=0;i<N;i++) printf("%f ",c[i]); printf("\n"); //释放显存 CHECK_STATUS(cudaFree(d_a)); CHECK_STATUS(cudaFree(d_b)); CHECK_STATUS(cudaFree(d_c)); return 0; }
1,294
#include "includes.h" __global__ void Ecalc2(float* out, const float* label) { int i = threadIdx.x; //4 //int j = blockDim.y*blockIdx.y + threadIdx.y; //Data.count out[i] = label[i] - out[i]; }
1,295
#include "stdio.h" #include <cuda.h> #include <cuda_runtime.h> #include <iostream> // Defining number of elements in Array #define N 50000 __global__ void gpuAdd(int *d_a, int *d_b, int *d_c) { int tid = threadIdx.x + blockIdx.x * blockDim.x; while (tid < N) { d_c[tid] = d_a[tid] + d_b[tid]; tid += blockDim.x * gridDim.x; } } int main(void) { int *h_a, *h_b, *h_c; int *d_a0, *d_b0, *d_c0; int *d_a1, *d_b1, *d_c1; cudaStream_t stream0, stream1; cudaStreamCreate(&stream0); cudaStreamCreate(&stream1); cudaEvent_t e_start, e_stop; cudaEventCreate(&e_start); cudaEventCreate(&e_stop); cudaEventRecord(e_start, 0); cudaHostAlloc((void **)&h_a, 2 * N * sizeof(int), cudaHostAllocDefault); cudaHostAlloc((void **)&h_b, 2 * N * sizeof(int), cudaHostAllocDefault); cudaHostAlloc((void **)&h_c, 2 * N * sizeof(int), cudaHostAllocDefault); cudaMalloc((void **)&d_a0, N * sizeof(int)); cudaMalloc((void **)&d_b0, N * sizeof(int)); cudaMalloc((void **)&d_c0, N * sizeof(int)); cudaMalloc((void **)&d_a1, N * sizeof(int)); cudaMalloc((void **)&d_b1, N * sizeof(int)); cudaMalloc((void **)&d_c1, N * sizeof(int)); for (int i = 0; i < N * 2; i++) { h_a[i] = 2 * i * i; h_b[i] = i; } cudaMemcpyAsync(d_a0, h_a, N * sizeof(int), cudaMemcpyHostToDevice, stream0); cudaMemcpyAsync(d_a1, h_a + N, N * sizeof(int), cudaMemcpyHostToDevice, stream1); cudaMemcpyAsync(d_b0, h_b, N * sizeof(int), cudaMemcpyHostToDevice, stream0); cudaMemcpyAsync(d_b1, h_b + N, N * sizeof(int), cudaMemcpyHostToDevice, stream1); gpuAdd<<<512, 512, 0, stream0>>>(d_a0, d_b0, d_c0); gpuAdd<<<512, 512, 0, stream1>>>(d_a1, d_b1, d_c1); cudaMemcpyAsync(h_c, d_c0, N * sizeof(int), cudaMemcpyDeviceToHost, stream0); cudaMemcpyAsync(h_c + N, d_c1, N * sizeof(int), cudaMemcpyDeviceToHost, stream0); cudaDeviceSynchronize(); cudaStreamSynchronize(stream0); cudaStreamSynchronize(stream1); cudaEventRecord(e_stop, 0); cudaEventSynchronize(e_stop); float elapsedTime; cudaEventElapsedTime(&elapsedTime, e_start, e_stop); printf("Time to add %d numbers: %3.1f ms\n", 2 * N, elapsedTime); int Correct = 1; printf("Vector addition on GPU \n"); // Printing result on console for (int i = 0; i < 2 * N; i++) { if ((h_a[i] + h_b[i] != h_c[i])) { Correct = 0; } } if (Correct == 1) { printf("GPU has computed Sum Correctly\n"); } else { printf("There is an Error in GPU Computation\n"); } // Free up memory cudaFree(d_a0); cudaFree(d_b0); cudaFree(d_c0); cudaFree(d_a0); cudaFree(d_b0); cudaFree(d_c0); cudaFreeHost(h_a); cudaFreeHost(h_b); cudaFreeHost(h_c); return 0; }
1,296
#include <stdio.h> __global__ void kadd(float *a, float *b, float *c, const unsigned int el_per_thread) { int i = blockIdx.x * blockDim.x + threadIdx.x; unsigned int offset = i * el_per_thread; for(unsigned int idx = 0; idx < el_per_thread; idx++) { c[offset+idx] = a[offset+idx] + b[offset+idx]; } }
1,297
#include <stdio.h> #include <stdlib.h> #include <sys/time.h> #define N 1000000 #define B 1024 __global__ void prescan( float *g_idata, float *INCR, int n); void scanCPU(float *f_out, float *f_in, int i_n); double myDiffTime(struct timeval &start, struct timeval &end) { double d_start, d_end; d_start = (double)(start.tv_sec + start.tv_usec/1000000.0); d_end = (double)(end.tv_sec + end.tv_usec/1000000.0); return (d_end - d_start); } int main(int argc, char **argv) { float a[N], c[N], g[N]; timeval start, end; float *dev_a, *dev_g, INCRR[B]; float size = N*sizeof(float); // cudaMallocManaged(&a, N*sizeof(float)); double d_gpuTime, d_cpuTime; cudaHostAlloc(&dev_a, size, cudaHostAllocDefault); cudaHostAlloc(&dev_g, size, cudaHostAllocDefault); dev_a = (float *)malloc(size); dev_g = (float *)malloc(size); cudaMalloc((void **)&dev_a, size); cudaMalloc((void **)&dev_g, size); for (int i = 0; i < N; i++) { // a[i] = (float)(rand() % 1000000)/1000.0; a[i] = i+1; } int BATCH = (N/B); gettimeofday(&start,NULL); cudaMemcpy(dev_a, a, size, cudaMemcpyHostToDevice); prescan <<< BATCH, B>>> ( dev_a ,INCRR , N); cudaDeviceSynchronize(); cudaMemcpy(g, dev_g, size, cudaMemcpyDeviceToHost); gettimeofday(&end, NULL); d_gpuTime = myDiffTime(start, end); gettimeofday(&start, NULL); scanCPU(c, a, N); gettimeofday(&end, NULL); d_cpuTime = myDiffTime(start, end); cudaFree(dev_a); cudaFree(dev_g); // for (int i = 0; i < N; i++) // { // printf("c[%i] = %0.3f, g[%i] = %0.3f\n", i, c[i], i, g[i]); // } printf("GPU Time for scan size %i: %f\n", N, d_gpuTime); printf("CPU Time for scan size %i: %f\n", N, d_cpuTime); } __global__ void prescan( float *g_idata, float *INCR, int n) { extern __shared__ float temp[], g_odata[], SUMS[]; // allocated on invocation int thid = threadIdx.x + (blockIdx.x * blockDim.x); int offset = 1; temp[2*thid] = g_idata[2*thid]; // load input into shared memory temp[2*thid+1] = g_idata[2*thid+1]; for (int d = B>>1; d > 0; d >>= 1) // build sum in place up the tree { __syncthreads(); if (thid < d) { int ai = offset*(2*thid+1)-1; int bi = offset*(2*thid+2)-1; temp[bi] += temp[ai]; } offset *= 2; for(int d =B>>1; d>0; d>>=1){ //build up sums __syncthreads(); if (thid < d) { int ai = offset*(2*thid+1)-1; int bi = offset*(2*thid+2)-1; } } if (SUMS && thid == 0) { SUMS[B] = temp[2*B-1]; temp[n - 1] = 0; } // clear the last element for (int d = 1; d < n; d *= 2) // traverse down tree & build scan { offset >>= 1; __syncthreads(); if (thid < d) { int ai = offset*(2*thid+1)-1; int bi = offset*(2*thid+2)-1; float t = SUMS[ai]; SUMS[ai] = SUMS[bi]; SUMS[bi] += INCR[bi]; } } __syncthreads(); temp[2*thid] = INCR[2*thid]; // write results to device memory // g_odata[2*thid+1] = temp[2*thid+1] // g_odata[2*thid+1] = INCR[thid]; } } void scanCPU(float *f_out, float *f_in, int i_n) { f_out[0] = 0; for (int i = 1; i < i_n; i++) f_out[i] = f_out[i-1] + f_in[i-1]; }
1,298
#include "includes.h" using namespace std; // this amazingly nice error checking function is stolen from: //https://stackoverflow.com/questions/14038589/what-is-the-canonical-way-to-check-for-errors-using-the-cuda-runtime-api __global__ void MatrixMulKernel(double *OutMat, double *Mat1, double *Mat2, int Arows, int Acols, int Bcols) { // row and column within submatrix int blockrow = blockIdx.y;//* int row = threadIdx.y; int blockcol = blockIdx.x; int col = threadIdx.x ; // allocate these arrays only once we can change the values in them later __shared__ double subAshared[BLOCKSIZE*BLOCKSIZE]; __shared__ double subBshared[BLOCKSIZE*BLOCKSIZE]; double Cvalue=0; for (int B = 0; B < ceil((double)(Acols / BLOCKSIZE)) + 1; B++) { // fetch from global memory // yes, these took a LONG time to figure out. Pencil and Paper FTW! /* notice: 1) how these indexes are actually offset a multiple of B, *not 1*. 2) threads are offset by col which will be 1 apart for each thread 3) which means that means all threads in the warp are hitting successive global memory cells */ int Mat1index = (row + blockrow*BLOCKSIZE)*Acols + col + B*BLOCKSIZE; int Mat2index = (B*BLOCKSIZE + row)*Bcols + BLOCKSIZE*blockcol + col; if (Mat1index < Arows*Acols) subAshared[row*BLOCKSIZE + col] = Mat1[Mat1index]; else subAshared[row*BLOCKSIZE + col] = 0; if (Mat2index < Acols*Bcols) subBshared[row*BLOCKSIZE + col] = Mat2[Mat2index]; else subBshared[row*BLOCKSIZE + col] = 0; __syncthreads(); // this computation is all using shared memory (fast) for (int j = 0; j < BLOCKSIZE; j++) if ((row*BLOCKSIZE + j < BLOCKSIZE*BLOCKSIZE) && (j*BLOCKSIZE + col < BLOCKSIZE*BLOCKSIZE)) Cvalue += subAshared[row*BLOCKSIZE + j]*subBshared[j*BLOCKSIZE + col]; __syncthreads(); } if ((row < Arows) && (col < Bcols)) { int finalmatrow = blockrow*BLOCKSIZE + row; int finalmatcol = blockcol*BLOCKSIZE + col; OutMat[finalmatrow*Bcols + finalmatcol] = Cvalue; } }
1,299
#include <stdio.h> #include <stdlib.h> #include <time.h> #include <sys/time.h> #include <cuda.h> //void Algorithm1(); void Algorithm4(int m, int n, int l); //for gemm 4 algorithm #define BLOCK_SIZE_x 16 #define BLOCK_SIZE_y 4 template<int block_size_x, int block_size_y> __global__ void device_Matrix_multi(const double* const device_matrix_A,const double* const device_matrix_B,double* device_matrix_C,const int m,const int n,const int l) { const int threadid_x = threadIdx.x; const int threadid_y = threadIdx.y; const int blockid_x = blockIdx.x; const int blockid_y = blockIdx.y; __shared__ double matrix_B_shared[block_size_x][block_size_x+1]; double c[block_size_x]; for (int i = 0; i< block_size_x; i++) { c[i] = 0.0; } int idx_A = blockid_x*block_size_x*block_size_y + threadid_x + threadid_y*block_size_x; int idx_B = threadid_x + (blockid_y*block_size_x + threadid_y)*n; int idx_B_last = idx_B + n; int col_A = 0; do { for(int i = 0; i < block_size_x; i += block_size_y) matrix_B_shared[threadid_x][threadid_y + i] = device_matrix_B[idx_B + i*n]; idx_B += block_size_x; __syncthreads(); int i_bound = min(block_size_x, n - col_A); for (int i = 0; i < i_bound; i++, idx_A+=m) { for (int j = 0; j < block_size_x; j++) { c[j] += device_matrix_A[idx_A]*matrix_B_shared[i][j]; } } col_A += block_size_x; __syncthreads(); }while (idx_B < idx_B_last); if (blockid_x*block_size_x*block_size_y + threadid_x + threadid_y*block_size_x < m) { int idx_D = blockid_x*block_size_x*block_size_y + (threadid_x + threadid_y*block_size_x) + blockid_y*block_size_x*m; int i_bound = min(block_size_x, l - blockid_y*block_size_x); for (int i = 0; i < i_bound; i++, idx_D += m) { device_matrix_C[idx_D] = c[i]; } } } int main() { Algorithm4(32, 32, 32); Algorithm4(64, 64, 64); Algorithm4(128,128, 128); Algorithm4(256, 256, 256); Algorithm4(512, 512, 512); Algorithm4(1024, 1024, 1024); Algorithm4(2048, 2048, 2048); Algorithm4(4096, 4096, 4096); } void Algorithm4(int m, int n, int l) { printf("inside function"); double* matrix_A; double* matrix_B; double* matrix_C; double *device_matrix_A; double *device_matrix_B; double *device_matrix_C; cudaEvent_t start, stop; cudaEventCreate(&start); cudaEventCreate(&stop); float milliseconds = 0; // Allocate the device memory matrix_A = (double*)malloc( m*n* sizeof(double)); matrix_B = (double*)malloc( m*l* sizeof(double)); matrix_C = (double*)malloc( m*l*sizeof(double)); cudaMalloc(&device_matrix_A, m*n*sizeof(double)); cudaMalloc(&device_matrix_B, n*l*sizeof(double)); cudaMalloc(&device_matrix_C, m*l*sizeof(double)); for(int i = 0; i < m; i++) { for(int j = 0; j <n; j++){ matrix_A[i *n + j] = rand()%10; matrix_B[i *n + j] = rand()%10; matrix_C[i *n + j] = 0; } } // Copy data from the host memory to the device memory cudaMemcpy(device_matrix_A, matrix_A, m*n*sizeof(double), cudaMemcpyHostToDevice); cudaMemcpy(device_matrix_B, matrix_B, n*l*sizeof(double), cudaMemcpyHostToDevice); dim3 nthreads(0, 0); dim3 nblocks(0, 0); // Launch the kernel nthreads.x = BLOCK_SIZE_x; nthreads.y = BLOCK_SIZE_y; nblocks.x = (m + nthreads.x*nthreads.y - 1)/(nthreads.x*nthreads.y); nblocks.y = (l + nthreads.x - 1)/nthreads.x; cudaEventRecord(start); printf("nuumber of blocks in x = %d\n", nblocks.x); printf("nuumber of blocks in y = %d\n", nblocks.y); printf("number of threads in x = %d\n", nthreads.x); printf("number of threads in y =%d\n", nthreads.y); printf("total threads = %d", nblocks.x*nblocks.y*nthreads.x*nthreads.y); device_Matrix_multi<BLOCK_SIZE_x, BLOCK_SIZE_y> <<<nblocks, nthreads>>> ( device_matrix_A,device_matrix_B, device_matrix_C,m,n,l ); // Copy data from the device memory to the host memory cudaMemcpy(matrix_C, device_matrix_C, m*l*sizeof(double), cudaMemcpyDeviceToHost); cudaEventRecord(stop); for(int i=0; i<m;i++){ for(int j =0; j<n; j++){ } } cudaEventElapsedTime(&milliseconds, start, stop); printf("elaspsed = %f ms\n\n\n", milliseconds); // Free the device memory cudaFree(device_matrix_A); cudaFree(device_matrix_B); cudaFree(device_matrix_C); free(matrix_A); free(matrix_B); free(matrix_C); }
1,300
#include <iostream> #include <sys/time.h> __global__ void saxpyDevice(int n, float a, float *x, float *y){ int i = blockIdx.x*blockDim.x + threadIdx.x; if (i < n) y[i] = a*x[i] + y[i]; } void saxpy(int n, float a, float *x, float *y){ float *d_x, *d_y; // allocate GPU memory, and upload data cudaMalloc(&d_x, n*sizeof(float)); cudaMalloc(&d_y, n*sizeof(float)); cudaMemcpy(d_x, x, n*sizeof(float), cudaMemcpyHostToDevice); cudaMemcpy(d_y, y, n*sizeof(float), cudaMemcpyHostToDevice); // send instructions to GPU saxpyDevice<<<(n+255)/256, 256>>>(n, 2.0f, d_x, d_y); // download data, and free GPU memory cudaMemcpy(y, d_y, n*sizeof(float), cudaMemcpyDeviceToHost); cudaFree(d_x); cudaFree(d_y); } int main(void){ int N = 1<<20; float *x, *y; x = (float*)malloc(N*sizeof(float)); y = (float*)malloc(N*sizeof(float)); struct timeval t0, t1; gettimeofday(&t0, NULL); for (int i=0; i<100; i++) saxpy(N, 2.0f, x, y); gettimeofday(&t1, NULL); std::cout<<"CUDA = "<<(t1.tv_sec - t0.tv_sec)*1000 + (t1.tv_usec-t0.tv_usec)/1000<<"ms"<<std::endl; return 0; }