serial_no
int64
1
24.2k
cuda_source
stringlengths
11
9.01M
19,701
#include "cuda_runtime.h" #include "device_launch_parameters.h" #include <stdio.h> #include <stdlib.h> #include <math.h> #include <time.h> // Row size const int N = 3; // Column Size const int M = 4; const int SIZE = M * N; __global__ void matrixAdd(int *c, const int *a, const int *b, int cols, int rows) { int row = blockIdx.y * blockDim.y + threadIdx.y; int col = blockIdx.x * blockDim.x + threadIdx.x; // 0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11 int i = row * cols + col; if (row < rows && col < cols) { c[i] = a[i] + b[i]; } } int main() { srand(time(NULL)); int a[N][M], b[N][M], *c; int *dev_a, *dev_b, *dev_c; c = (int *)malloc(SIZE * sizeof(int)); cudaMalloc((void **)&dev_a, SIZE * sizeof(int)); cudaMalloc((void **)&dev_b, SIZE * sizeof(int)); cudaMalloc((void **)&dev_c, SIZE * sizeof(int)); for (int row = 0; row < N; ++row) for (int col = 0; col < M; ++col) { a[row][col] = rand() % 10; b[row][col] = rand() % 10; } cudaMemcpy(dev_a, a, SIZE * sizeof(int), cudaMemcpyHostToDevice); cudaMemcpy(dev_b, b, SIZE * sizeof(int), cudaMemcpyHostToDevice); dim3 dimBlock(M, N); dim3 dimGrid((int)ceil(M / dimBlock.x), (int)ceil(N / dimBlock.y)); matrixAdd <<< dimGrid, dimBlock >>>(dev_c, dev_a, dev_b, M, N); cudaMemcpy(c, dev_c, SIZE * sizeof(int), cudaMemcpyDeviceToHost); for (int i = 0; i < SIZE; ++i) printf("c[%d] = %d\n", i, c[i]); // Free the Host array memory free(c); // Free the Device array memory cudaFree(dev_a); cudaFree(dev_b); cudaFree(dev_c); return 0; }
19,702
/* 3D IoU Calculation and Rotated NMS(modified from 2D NMS written by others) Written by Shaoshuai Shi All Rights Reserved 2018. */ #include <stdio.h> #define THREADS_PER_BLOCK 16 #define DIVUP(m, n) ((m) / (n) + ((m) % (n) > 0)) #define DEBUG const int THREADS_PER_BLOCK_NMS = sizeof(unsigned long long) * 8; const float EPS = 1e-8; struct Point { float x, y; __device__ Point() {} __device__ Point(float _x, float _y){ x = _x, y = _y; } __device__ void set(float _x, float _y){ x = _x; y = _y; } __device__ Point operator +(const Point &b)const{ return Point(x + b.x, y + b.y); } __device__ Point operator -(const Point &b)const{ return Point(x - b.x, y - b.y); } }; __device__ inline float cross(const Point &a, const Point &b){ return a.x * b.y - a.y * b.x; } __device__ inline float cross(const Point &p1, const Point &p2, const Point &p0){ return (p1.x - p0.x) * (p2.y - p0.y) - (p2.x - p0.x) * (p1.y - p0.y); } __device__ int check_rect_cross(const Point &p1, const Point &p2, const Point &q1, const Point &q2){ int ret = min(p1.x,p2.x) <= max(q1.x,q2.x) && min(q1.x,q2.x) <= max(p1.x,p2.x) && min(p1.y,p2.y) <= max(q1.y,q2.y) && min(q1.y,q2.y) <= max(p1.y,p2.y); return ret; } __device__ inline int check_in_box2d(const float *box, const Point &p){ //params: box (5) [x1, y1, x2, y2, angle] const float MARGIN = 1e-5; float center_x = (box[0] + box[2]) / 2; float center_y = (box[1] + box[3]) / 2; float angle_cos = cos(box[4]), angle_sin = sin(box[4]); // rotate the point in the opposite direction of box float rot_x = (p.x - center_x) * angle_cos + (p.y - center_y) * angle_sin + center_x; float rot_y = -(p.x - center_x) * angle_sin + (p.y - center_y) * angle_cos + center_y; //#ifdef DEBUG // printf("box: (%.3f, %.3f, %.3f, %.3f, %.3f)\n", box[0], box[1], box[2], box[3], box[4]); // printf("center: (%.3f, %.3f), cossin(%.3f, %.3f), src(%.3f, %.3f), rot(%.3f, %.3f)\n", center_x, center_y, // angle_cos, angle_sin, p.x, p.y, (p.x - center_x) * angle_cos + (p.y - center_y) * angle_sin + center_x, rot_y); //#endif return (rot_x > box[0] - MARGIN && rot_x < box[2] + MARGIN && rot_y > box[1] - MARGIN && rot_y < box[3] + MARGIN); } __device__ inline int intersection(const Point &p1, const Point &p0, const Point &q1, const Point &q0, Point &ans){ // fast exclusion if (check_rect_cross(p0, p1, q0, q1) == 0) return 0; // check cross standing float s1 = cross(q0, p1, p0); float s2 = cross(p1, q1, p0); float s3 = cross(p0, q1, q0); float s4 = cross(q1, p1, q0); if (!(s1 * s2 > 0 && s3 * s4 > 0)) return 0; // calculate intersection of two lines float s5 = cross(q1, p1, p0); if(fabs(s5 - s1) > EPS){ ans.x = (s5 * q0.x - s1 * q1.x) / (s5 - s1); ans.y = (s5 * q0.y - s1 * q1.y) / (s5 - s1); } else{ float a0 = p0.y - p1.y, b0 = p1.x - p0.x, c0 = p0.x * p1.y - p1.x * p0.y; float a1 = q0.y - q1.y, b1 = q1.x - q0.x, c1 = q0.x * q1.y - q1.x * q0.y; float D = a0 * b1 - a1 * b0; ans.x = (b0 * c1 - b1 * c0) / D; ans.y = (a1 * c0 - a0 * c1) / D; } return 1; } __device__ inline void rotate_around_center(const Point &center, const float angle_cos, const float angle_sin, Point &p){ float new_x = (p.x - center.x) * angle_cos + (p.y - center.y) * angle_sin + center.x; float new_y = -(p.x - center.x) * angle_sin + (p.y - center.y) * angle_cos + center.y; p.set(new_x, new_y); } __device__ inline int point_cmp(const Point &a, const Point &b, const Point &center){ return atan2(a.y - center.y, a.x - center.x) > atan2(b.y - center.y, b.x - center.x); } __device__ inline float box_overlap(const float *box_a, const float *box_b){ // params: box_a (5) [x1, y1, x2, y2, angle] // params: box_b (5) [x1, y1, x2, y2, angle] float a_x1 = box_a[0], a_y1 = box_a[1], a_x2 = box_a[2], a_y2 = box_a[3], a_angle = -box_a[4]; float b_x1 = box_b[0], b_y1 = box_b[1], b_x2 = box_b[2], b_y2 = box_b[3], b_angle = -box_b[4]; Point center_a((a_x1 + a_x2) / 2, (a_y1 + a_y2) / 2); Point center_b((b_x1 + b_x2) / 2, (b_y1 + b_y2) / 2); //#ifdef DEBUG // printf("a: (%.3f, %.3f, %.3f, %.3f, %.3f), b: (%.3f, %.3f, %.3f, %.3f, %.3f)\n", a_x1, a_y1, a_x2, a_y2, a_angle, // b_x1, b_y1, b_x2, b_y2, b_angle); // printf("center a: (%.3f, %.3f), b: (%.3f, %.3f)\n", center_a.x, center_a.y, center_b.x, center_b.y); //#endif Point box_a_corners[5]; box_a_corners[0].set(a_x1, a_y1); box_a_corners[1].set(a_x2, a_y1); box_a_corners[2].set(a_x2, a_y2); box_a_corners[3].set(a_x1, a_y2); Point box_b_corners[5]; box_b_corners[0].set(b_x1, b_y1); box_b_corners[1].set(b_x2, b_y1); box_b_corners[2].set(b_x2, b_y2); box_b_corners[3].set(b_x1, b_y2); // get oriented corners float a_angle_cos = cos(a_angle), a_angle_sin = sin(a_angle); float b_angle_cos = cos(b_angle), b_angle_sin = sin(b_angle); for (int k = 0; k < 4; k++){ //#ifdef DEBUG // printf("before corner %d: a(%.3f, %.3f), b(%.3f, %.3f) \n", k, box_a_corners[k].x, box_a_corners[k].y, box_b_corners[k].x, box_b_corners[k].y); //#endif rotate_around_center(center_a, a_angle_cos, a_angle_sin, box_a_corners[k]); rotate_around_center(center_b, b_angle_cos, b_angle_sin, box_b_corners[k]); //#ifdef DEBUG // printf("corner %d: a(%.3f, %.3f), b(%.3f, %.3f) \n", k, box_a_corners[k].x, box_a_corners[k].y, box_b_corners[k].x, box_b_corners[k].y); //#endif } box_a_corners[4] = box_a_corners[0]; box_b_corners[4] = box_b_corners[0]; // get intersection of lines Point cross_points[16]; Point poly_center; int cnt = 0, flag = 0; poly_center.set(0, 0); for (int i = 0; i < 4; i++){ for (int j = 0; j < 4; j++){ flag = intersection(box_a_corners[i + 1], box_a_corners[i], box_b_corners[j + 1], box_b_corners[j], cross_points[cnt]); if (flag){ poly_center = poly_center + cross_points[cnt]; cnt++; } } } // check corners for (int k = 0; k < 4; k++){ if (check_in_box2d(box_a, box_b_corners[k])){ poly_center = poly_center + box_b_corners[k]; cross_points[cnt] = box_b_corners[k]; cnt++; } if (check_in_box2d(box_b, box_a_corners[k])){ poly_center = poly_center + box_a_corners[k]; cross_points[cnt] = box_a_corners[k]; cnt++; } } poly_center.x /= cnt; poly_center.y /= cnt; // sort the points of polygon Point temp; for (int j = 0; j < cnt - 1; j++){ for (int i = 0; i < cnt - j - 1; i++){ if (point_cmp(cross_points[i], cross_points[i + 1], poly_center)){ temp = cross_points[i]; cross_points[i] = cross_points[i + 1]; cross_points[i + 1] = temp; } } } //#ifdef DEBUG // printf("cnt=%d\n", cnt); // for (int i = 0; i < cnt; i++){ // printf("All cross point %d: (%.3f, %.3f)\n", i, cross_points[i].x, cross_points[i].y); // } //#endif // get the overlap areas float area = 0; for (int k = 0; k < cnt - 1; k++){ area += cross(cross_points[k] - cross_points[0], cross_points[k + 1] - cross_points[0]); } return fabs(area) / 2.0; } // iou_bev => iou3d __device__ inline float iou3d(const float *box_a, const float *box_b, const float* box_a_z, const float* box_b_z, bool ignore_height){ // params: box_a (5) [x1, y1, x2, y2, angle] // params: box_b (5) [x1, y1, x2, y2, angle] // params: box_a_z (2) [z1min, z1max] // params: box_b_z (2) [z2min, z2max] float sa = (box_a[2] - box_a[0]) * (box_a[3] - box_a[1]); float sb = (box_b[2] - box_b[0]) * (box_b[3] - box_b[1]); float s_overlap = box_overlap(box_a, box_b); // compute the height float intersection_height = fminf(box_a_z[1], box_b_z[1]) - fmaxf(box_a_z[0], box_b_z[0]); //compute the volume if (!ignore_height) { float vol_a = sa * (box_a_z[1] - box_a_z[0]); float vol_b = sb * (box_b_z[1] - box_b_z[0]); float vol_overlap = s_overlap * intersection_height; //#ifdef DEBUG // printf("sa, sb, s_overlap, vol_a, vol_b, vol_overlap: (%.3f, %.3f, %.3f, %.3f, %.3f, %.3f)\n", sa, sb, s_overlap, vol_a, vol_b, vol_overlap); //#endif return vol_overlap / fmaxf(vol_a + vol_b - vol_overlap, EPS); } else { return s_overlap / fmaxf(sa + sb, EPS); } } __global__ void boxes_iou_3d_kernel(bool ignore_height, const int num_a, const float *boxes_a, const int num_b, const float *boxes_b, float *ans_iou){ // params: num_a : number of boxes in boxes_a // params: boxes_a (M, 7) [x, y, z, w, l, h, angle] // params: num_b : number of boxes in boxes_b // params: boxes_b (N, 7) [x, y, z, w, l, h, angle] const int a_idx = blockIdx.y * THREADS_PER_BLOCK + threadIdx.y; const int b_idx = blockIdx.x * THREADS_PER_BLOCK + threadIdx.x; if (a_idx >= num_a || b_idx >= num_b){ return; } const float * cur_box_a = boxes_a + a_idx * 7; const float * cur_box_b = boxes_b + b_idx * 7; float box_a_tmp[5]; float box_b_tmp[5]; float box_a_z_tmp[2]; float box_b_z_tmp[2]; // [x, y, z, w, l ,h, r] box_a_tmp[0] = cur_box_a[3] - cur_box_a[0] / 2; // x1, box_a_tmp[1] = cur_box_a[4] - cur_box_a[1] / 2; // y1 box_a_tmp[2] = cur_box_a[3] + cur_box_a[0] / 2; // x2 box_a_tmp[3] = cur_box_a[4] + cur_box_a[1] / 2; // y2 box_a_tmp[4] = cur_box_a[6]; // ry box_a_z_tmp[0] = cur_box_a[5] - cur_box_a[2] / 2; // z1min box_a_z_tmp[1] = cur_box_a[5] + cur_box_a[2] / 2; // z1max box_b_tmp[0] = cur_box_b[3] - cur_box_b[0] / 2; // x1, box_b_tmp[1] = cur_box_b[4] - cur_box_b[1] / 2; // y1 box_b_tmp[2] = cur_box_b[3] + cur_box_b[0] / 2; // x2 box_b_tmp[3] = cur_box_b[4] + cur_box_b[1] / 2; // y2 box_b_tmp[4] = cur_box_b[6]; // ry box_b_z_tmp[0] = cur_box_b[5] - cur_box_b[2] / 2; // z1min box_b_z_tmp[1] = cur_box_b[5] + cur_box_b[2] / 2; // z1max float cur_iou_3d = iou3d( &box_a_tmp[0], &box_b_tmp[0], &box_a_z_tmp[0], &box_b_z_tmp[0], ignore_height); ans_iou[a_idx * num_b + b_idx] = cur_iou_3d; } __global__ void nms3d_kernel(const int boxes_num, const float nms_overlap_thresh, const float *boxes, unsigned long long *mask){ //params: boxes (N, 7) [x, y, z, w, l ,h, ry] z-up coordinate system //params: mask (N, N/THREADS_PER_BLOCK_NMS) const int row_start = blockIdx.y; const int col_start = blockIdx.x; // if (row_start > col_start) return; const int row_size = fminf(boxes_num - row_start * THREADS_PER_BLOCK_NMS, THREADS_PER_BLOCK_NMS); const int col_size = fminf(boxes_num - col_start * THREADS_PER_BLOCK_NMS, THREADS_PER_BLOCK_NMS); __shared__ float block_boxes[THREADS_PER_BLOCK_NMS * 7]; if (threadIdx.x < col_size) { block_boxes[threadIdx.x * 7 + 0] = boxes[(THREADS_PER_BLOCK_NMS * col_start + threadIdx.x) * 7 + 0]; block_boxes[threadIdx.x * 7 + 1] = boxes[(THREADS_PER_BLOCK_NMS * col_start + threadIdx.x) * 7 + 1]; block_boxes[threadIdx.x * 7 + 2] = boxes[(THREADS_PER_BLOCK_NMS * col_start + threadIdx.x) * 7 + 2]; block_boxes[threadIdx.x * 7 + 3] = boxes[(THREADS_PER_BLOCK_NMS * col_start + threadIdx.x) * 7 + 3]; block_boxes[threadIdx.x * 7 + 4] = boxes[(THREADS_PER_BLOCK_NMS * col_start + threadIdx.x) * 7 + 4]; block_boxes[threadIdx.x * 7 + 5] = boxes[(THREADS_PER_BLOCK_NMS * col_start + threadIdx.x) * 7 + 5]; block_boxes[threadIdx.x * 7 + 6] = boxes[(THREADS_PER_BLOCK_NMS * col_start + threadIdx.x) * 7 + 6]; } __syncthreads(); if (threadIdx.x < row_size) { const int cur_box_idx = THREADS_PER_BLOCK_NMS * row_start + threadIdx.x; const float *cur_box = boxes + cur_box_idx * 7; float cur_box_tmp[5]; float cur_box_z_tmp[2]; cur_box_tmp[0] = cur_box[3] - cur_box[0] / 2; // x1, cur_box_tmp[1] = cur_box[4] - cur_box[1] / 2; // y1 cur_box_tmp[2] = cur_box[3] + cur_box[0] / 2; // x2 cur_box_tmp[3] = cur_box[4] + cur_box[1] / 2; // y2 cur_box_tmp[4] = cur_box[6]; // ry cur_box_z_tmp[0] = cur_box[5] - cur_box[2] / 2; // z1min cur_box_z_tmp[1] = cur_box[5] + cur_box[2] / 2; // z1max int i = 0; unsigned long long t = 0; int start = 0; if (row_start == col_start) { start = threadIdx.x + 1; } for (i = start; i < col_size; i++) { const float *block_box_ptr = block_boxes + i * 7; float block_box_tmp[5]; float block_box_z_tmp[2]; block_box_tmp[0] = block_box_ptr[3] - block_box_ptr[0] / 2; // x1, block_box_tmp[1] = block_box_ptr[4] - block_box_ptr[1] / 2; // y1 block_box_tmp[2] = block_box_ptr[3] + block_box_ptr[0] / 2; // x2 block_box_tmp[3] = block_box_ptr[4] + block_box_ptr[1] / 2; // y2 block_box_tmp[4] = block_box_ptr[6]; // ry block_box_z_tmp[0] = block_box_ptr[5] - block_box_ptr[2] / 2; // z1min block_box_z_tmp[1] = block_box_ptr[5] + block_box_ptr[2] / 2; // z1max float cur_iou_3d = iou3d(&cur_box_tmp[0], &block_box_tmp[0], &cur_box_z_tmp[0], &block_box_z_tmp[0], false); if (cur_iou_3d > nms_overlap_thresh){ t |= 1ULL << i; } } const int col_blocks = DIVUP(boxes_num, THREADS_PER_BLOCK_NMS); mask[cur_box_idx * col_blocks + col_start] = t; } } void boxesIouGPUKernelLauncher(bool ignore_height, const int num_a, const float *boxes_a, const int num_b, const float *boxes_b, float *ans_iou){ dim3 blocks(DIVUP(num_b, THREADS_PER_BLOCK), DIVUP(num_a, THREADS_PER_BLOCK)); // blockIdx.x(col), blockIdx.y(row) dim3 threads(THREADS_PER_BLOCK, THREADS_PER_BLOCK); // (256, 256) boxes_iou_3d_kernel<<<blocks, threads>>>(ignore_height, num_a, boxes_a, num_b, boxes_b, ans_iou); } void nms3dGPUKernelLauncher(const int boxes_num, const float nms_overlap_thresh, const float *boxes, unsigned long long * mask ){ dim3 blocks(DIVUP(boxes_num, THREADS_PER_BLOCK_NMS), DIVUP(boxes_num, THREADS_PER_BLOCK_NMS)); dim3 threads(THREADS_PER_BLOCK_NMS); nms3d_kernel<<<blocks, threads>>>(boxes_num, nms_overlap_thresh, boxes, mask); }
19,703
#include "includes.h" __device__ float gamma_correction(float f_stop, float gamma, float val) { return powf((val*powf(2,f_stop)),(1.0/gamma)); } __global__ void tonemap_gamma(float* imageIn, float* imageOut, int width, int height, int channels, int depth, float f_stop, float gamma) { int Row = blockDim.y * blockIdx.y + threadIdx.y; int Col = blockDim.x * blockIdx.x + threadIdx.x; if(Row < height && Col < width) { imageOut[(Row*width+Col)*3+BLUE] = gamma_correction(f_stop, gamma, imageIn[(Row*width+Col)*3+BLUE]); imageOut[(Row*width+Col)*3+GREEN] = gamma_correction(f_stop, gamma, imageIn[(Row*width+Col)*3+GREEN]); imageOut[(Row*width+Col)*3+RED] = gamma_correction(f_stop, gamma, imageIn[(Row*width+Col)*3+RED]); } }
19,704
/* For DIRECTED GRAPH */ #include <cuda.h> #include <stdio.h> #include <stdlib.h> #include <string.h> #define MAX_NODE 100000000 #define DEBUG 0 #define gpuErrchk(ans) { gpuAssert((ans), __FILE__, __LINE__); } inline void gpuAssert(cudaError_t code, const char *file, int line, bool abort=true) { if (code != cudaSuccess) { fprintf(stderr,"GPUassert: %s %s %d\n", cudaGetErrorString(code), file, line); if (abort) exit(code); } } __device__ volatile int Cx[MAX_NODE]; __device__ volatile int PQ[MAX_NODE]; //K in parallel __global__ void extractMin(int* PQ_size, int* expandNodes,int* expandNodes_size,int* openList,int N,int K){ int id = blockIdx.x*blockDim.x+threadIdx.x; if(id<K && PQ_size[id]>0){ //extract min from PQ int front = id* ( (N+K-1)/K ); int node = PQ[front]; // printf("extract min %d %d\n",id,node); // restructure the heap PQ[front]=PQ[front+PQ_size[id]-1]; PQ_size[id]-=1; int pqIndex = 0; while(2*pqIndex+1 < PQ_size[id]){ if(2*pqIndex+2 >= PQ_size[id]){ if( Cx[PQ[front+pqIndex]] > Cx[PQ[front+2*pqIndex+1]]){ int swap = PQ[front + 2*pqIndex+1]; PQ[front + 2*pqIndex+1] = PQ[front +pqIndex]; PQ[front + pqIndex] = swap; pqIndex = 2*pqIndex+1; } else break; } else{ if( Cx[PQ[front+pqIndex]] > Cx[PQ[front+2*pqIndex+1]] && Cx[PQ[front+2*pqIndex+1]] <= Cx[PQ[front+2*pqIndex+2]] ){ int swap = PQ[front + 2*pqIndex+1]; PQ[front + 2*pqIndex+1] = PQ[front +pqIndex]; PQ[front + pqIndex] = swap; pqIndex = 2*pqIndex+1; } else if(Cx[PQ[front+pqIndex]] > Cx[PQ[front+2*pqIndex+2]] && Cx[PQ[front+2*pqIndex+2]] <= Cx[PQ[front+2*pqIndex+1]] ){ int swap = PQ[front + 2*pqIndex+2]; PQ[front + 2*pqIndex+2] = PQ[front +pqIndex]; PQ[front + pqIndex] = swap; pqIndex = 2*pqIndex+2; } else{ break; } } } //removed from openList openList[node] = -1; //added to expand next int len = atomicAdd(expandNodes_size,1); expandNodes[len]=node; } } //for K in parallel __global__ void A_star_expand(int* off,int* edge,unsigned int* W,int* Hx,int* parent, int* expandNodes,int* expandNodes_size, int* lock ,int* flagfound,int* openList, int N,int E, int K,int dest,int* nVFlag,int* PQ_size, int flagDiff,int* diff_off,int* diff_edge,int* diff_weight ){ int id = blockIdx.x*blockDim.x+threadIdx.x; if(id< *expandNodes_size ){ int node = expandNodes[id]; //reach dest if(node == dest){ atomicOr(flagfound,1); // *flagfound = 1; printf("found %d\n",id); } // expand int start = off[node]; int end = E; if(node!=N-1) end = off[node+1]; if(DEBUG) printf("%d$expand %d:\n",id,start); while(start < end){ int child = edge[start]; //deleted edges if(child<0){ start++; continue; } //array L initilaized with 0 //get the lock for child to update C(x) //loop till acquire the lock bool leaveLoop = false; while(leaveLoop==false){ if(atomicCAS(&lock[child],0,1)==0){ //critical section if( Cx[child] > (Cx[node] - Hx[node])+ W[start]+ Hx[child] ){ Cx[child] = (Cx[node] - Hx[node])+ W[start]+ Hx[child]; __threadfence(); parent[child] = node; if(DEBUG) printf("exp: %d %d\n",node,child); if(openList[child]>=0){ //update operating on one thread if(DEBUG) printf("upd: %d %d\n",node,child); int Kq = openList[child]; int front = Kq*( (N+K-1)/K ); int index = -1; for(int i=front;i<front+PQ_size[Kq];i++){ if(PQ[i]==child){ index = i; } } if(index > 0){ int i = index; while(i > front){ if( Cx[PQ[(i-1)/2]] > Cx[PQ[i]] ){ int swap = PQ[i]; PQ[i] = PQ[(i-1)/2]; PQ[(i-1)/2] = swap; i = (i-1)/2; } else break; } } if(DEBUG) printf("out of while\n"); }else{ nVFlag[child]=1; //add only once } } //end critical section leaveLoop = true; atomicCAS(&lock[child],1,0); // if(DEBUG) // printf("im not deadlocked\n"); } __syncthreads(); if(DEBUG) printf("%d$ stuck here\n",id); } start++; } if(DEBUG) printf("%d outside while\n",id); } } //N threads __global__ void setNV(int* nextFlag,int* nextV,int* nvSize,int N){ int id = blockIdx.x*blockDim.x+threadIdx.x; if(id < N){ //printf("2: %d %d\n",id,nextFlag[id]); if(nextFlag[id]==1){ int index = atomicAdd(nvSize,1); nextV[index]=id; // printf("2: %d\n",id); } } } //for K in parallel __global__ void insertPQ(int* PQS,int* nextV,int* nVsize,int K,int N,int* openList){ int id = blockIdx.x*blockDim.x+threadIdx.x; if(id < K){ // printf("id: %d\n",id); int front = id*( (N+K-1)/K ); int i = id; if(id==0) printf("insert: %d\n",*nVsize); while(i<*nVsize){ PQ[front+PQS[id]]= nextV[i]; PQS[id]+=1; //add in openList openList[nextV[i]] = id; //printf("insert: %d, %d\n",nextV[i],PQS[id]); if(PQS[id]>1){ int index = PQS[id]-1; while(index>0){ if(Cx[PQ[front+ (index-1)/2]] > Cx[PQ[front+index]]){ int swap = PQ[front+index]; PQ[front+index]=PQ[front+ (index-1)/2]; PQ[front+ (index-1)/2] = swap; index = (index-1)/2; } else break; } } i += K; } } } //for K in parallel __global__ void checkMIN(int* PQ_size,int* flagEnd,int dest,int N,int K){ int id = blockIdx.x*blockDim.x+threadIdx.x; if(id < K && PQ_size[id] > 0 ){ int front = id* ( (N+K-1)/K ); int node = PQ[front]; //check if atleast one min, dont end the a* printf("%d ",Cx[node]); if(Cx[dest] > Cx[node] ){ atomicAnd(flagEnd,0); } } } __global__ void printCX(int dest){ int id = blockIdx.x*blockDim.x+threadIdx.x; if(id==0){ printf("cost: %d\n",Cx[dest]); } } int main(){ //the K PQ int K ; scanf("%d\n",&K); int startNode,endNode; scanf("%d %d",&startNode,&endNode); FILE* fgraph = fopen("graph.txt","r"); int N,E; fscanf(fgraph,"%d %d\n",&N,&E); int* H_offset = (int*)malloc(sizeof(int)*N); int* H_edges = (int*)malloc(sizeof(int)*E); unsigned int* H_weight = (unsigned int*)malloc(sizeof(unsigned int)*E); int* H_hx = (int*)malloc(sizeof(int)*N); int* H_cx = (int*)malloc(sizeof(int)*N); int* H_parent = (int*)malloc(sizeof(int)*N); int* H_PQ = (int*)malloc(sizeof(int)*N); int* H_openList = (int*)malloc(sizeof(int)*N); int* H_PQ_size = (int*)malloc(sizeof(int)*K); //for diff int* H_diff_edges = (int*)malloc(sizeof(int)*E); int* H_diff_offset = (int*)malloc(sizeof(int)*N); int* H_diff_weight = (int*)malloc(sizeof(int)*E); memset(H_PQ_size,0,sizeof(int)*K); memset(H_parent,-1,sizeof(int)*N); memset(H_openList,-1,sizeof(int)*N); //init cx for(int i=0;i<N;i++){ H_cx[i]=INT_MAX; } for(int i=0;i<E;i++){ fscanf(fgraph,"%d",&H_edges[i]); } for(int i=0;i<N;i++){ fscanf(fgraph,"%d",&H_offset[i]); } for(int i=0;i<E;i++){ fscanf(fgraph,"%u",&H_weight[i]); } FILE* fhx = fopen("Hx.txt","r"); for(int i=0;i<N;i++){ fscanf(fhx,"%d",&H_hx[i]); } fclose(fgraph); fclose(fhx); printf("completed input\n"); //init Host var int* H_flagEnd = (int*)malloc(sizeof(int)); int* H_flagfound = (int*)malloc(sizeof(int)); int* H_a0 = (int*)malloc(sizeof(int)); //required coz if many tries to add same in diff threads high low lower int* H_nVFlag = (int*)malloc(sizeof(int)*N); memset(H_nVFlag,-1,sizeof(int)*N); *H_flagEnd = 0; *H_flagfound = 0; *H_a0 = 0; //insert startNode in PQ[0] H_cx[startNode]=H_hx[startNode]; H_PQ[0]=startNode; H_PQ_size[0]=1; H_openList[startNode]=0; //graph struture int* D_offset; int* D_edges ; unsigned int* D_weight; int* D_hx; int* D_parent; //Priority queue size int* D_PQ_size; //flag if in openList(contains which PQ) int* D_openList; //lock for nodes int* D_lock; //Diff structure int* D_diff_edges; int* D_diff_offset; int* D_diff_weight; //next nodes flag int* D_nVFlag; //next nodes array to insert PQ int* D_nV; int* D_nV_size; //nodes to be expanded ( extracted from PQ ) int* D_expandNodes; int* D_expandNodes_size; //flag to end while loop and found the destination int* D_flagEnd; int* D_flagfound; gpuErrchk ( cudaMalloc(&D_offset,sizeof(int)*N) ); gpuErrchk ( cudaMalloc(&D_edges,sizeof(int)*E) ); gpuErrchk ( cudaMalloc(&D_weight,sizeof(unsigned int)*E) ); gpuErrchk ( cudaMalloc(&D_hx,sizeof(int)*N) ); gpuErrchk ( cudaMalloc(&D_parent,sizeof(int)*N) ); gpuErrchk ( cudaMalloc(&D_PQ_size,sizeof(int)*K) ); gpuErrchk ( cudaMalloc(&D_openList,sizeof(int)*N) ); gpuErrchk ( cudaMalloc(&D_lock,sizeof(int)*N) ); //diff csr gpuErrchk ( cudaMalloc(&D_diff_edges,sizeof(int)*E) ); gpuErrchk ( cudaMalloc(&D_diff_offset,sizeof(int)*N) ); gpuErrchk ( cudaMalloc(&D_diff_weight,sizeof(int)*E) ); //for next set of vertices to add in PQ gpuErrchk ( cudaMalloc(&D_nV,sizeof(int)*N) ); gpuErrchk ( cudaMalloc(&D_nV_size,sizeof(int)) ); gpuErrchk ( cudaMalloc(&D_nVFlag,sizeof(int)*N) ); //next nodes to expand gpuErrchk ( cudaMalloc(&D_expandNodes,sizeof(int)*K) ); //changed to K gpuErrchk ( cudaMalloc(&D_expandNodes_size,sizeof(int)) ); //flag to end search gpuErrchk( cudaMalloc(&D_flagEnd,sizeof(int)) ); gpuErrchk( cudaMalloc(&D_flagfound,sizeof(int)) ); gpuErrchk ( cudaMemcpy(D_offset,H_offset,sizeof(int)*N,cudaMemcpyHostToDevice) ); gpuErrchk ( cudaMemcpy(D_edges,H_edges,sizeof(int)*E,cudaMemcpyHostToDevice) ); gpuErrchk ( cudaMemcpy(D_weight,H_weight,sizeof(unsigned int)*E,cudaMemcpyHostToDevice) ); gpuErrchk ( cudaMemcpy(D_hx,H_hx,sizeof(int)*N,cudaMemcpyHostToDevice) ); gpuErrchk ( cudaMemcpy(D_parent,H_parent,sizeof(int)*N,cudaMemcpyHostToDevice) ); gpuErrchk ( cudaMemcpy(D_openList,H_openList,sizeof(int)*N,cudaMemcpyHostToDevice) ); gpuErrchk ( cudaMemcpy(D_diff_edges,H_diff_edges,sizeof(int)*E,cudaMemcpyHostToDevice) ); gpuErrchk ( cudaMemcpy(D_diff_offset,H_diff_offset,sizeof(int)*N,cudaMemcpyHostToDevice) ); gpuErrchk ( cudaMemcpy(D_diff_weight,H_diff_weight,sizeof(int)*E,cudaMemcpyHostToDevice) ); gpuErrchk ( cudaMemcpy(D_PQ_size,H_PQ_size,sizeof(int)*K,cudaMemcpyHostToDevice) ); gpuErrchk ( cudaMemcpyToSymbol(Cx,H_cx, sizeof(int)*N, 0, cudaMemcpyHostToDevice) ); gpuErrchk ( cudaMemcpyToSymbol(PQ,H_PQ, sizeof(int)*N, 0, cudaMemcpyHostToDevice) ); gpuErrchk ( cudaMemcpy(D_flagEnd,H_flagEnd,sizeof(int),cudaMemcpyHostToDevice) ); gpuErrchk ( cudaMemcpy(D_flagfound,H_flagfound,sizeof(int),cudaMemcpyHostToDevice) ); gpuErrchk ( cudaMemcpy(D_nVFlag,H_nVFlag,sizeof(int)*N,cudaMemcpyHostToDevice) ); gpuErrchk ( cudaMemcpy(D_nV_size,H_a0,sizeof(int),cudaMemcpyHostToDevice) ); gpuErrchk ( cudaMemcpy(D_expandNodes_size,H_a0,sizeof(int),cudaMemcpyHostToDevice) ); gpuErrchk ( cudaMemset(D_lock,0,sizeof(int)*N) ); int flag_PQ_not_empty = 0; for(int i=0;i<K;i++){ if(H_PQ_size[i]>0) flag_PQ_not_empty=1; } int numThreads = 512; int numBlocks = (K+numThreads-1)/numThreads; int N_numBlocks = (N+numThreads-1)/numThreads; //DO A* initailly on whole graph while(*H_flagEnd==0 && flag_PQ_not_empty==1){ //extract min extractMin<<<numBlocks,numThreads>>>(D_PQ_size, D_expandNodes,D_expandNodes_size,D_openList,N,K); gpuErrchk(cudaPeekAtLastError() ); cudaDeviceSynchronize(); A_star_expand<<<numBlocks,numThreads>>>(D_offset,D_edges,D_weight,D_hx,D_parent, D_expandNodes,D_expandNodes_size, D_lock ,D_flagfound,D_openList, N,E,K,endNode,D_nVFlag,D_PQ_size, false,D_diff_offset,D_diff_edges,D_diff_offset ); gpuErrchk(cudaPeekAtLastError() ); cudaDeviceSynchronize(); if(DEBUG) printf("expand over\n"); //gen from flag D_nV //for N in parallel setNV<<<N_numBlocks,numThreads>>>(D_nVFlag,D_nV,D_nV_size,N); gpuErrchk(cudaPeekAtLastError() ); cudaDeviceSynchronize(); insertPQ<<<numBlocks,numThreads>>>(D_PQ_size,D_nV,D_nV_size,K,N,D_openList); gpuErrchk(cudaPeekAtLastError() ); cudaDeviceSynchronize(); //cpy flagend and flagEmpty gpuErrchk( cudaMemcpy(H_flagfound,D_flagfound, sizeof(int),cudaMemcpyDeviceToHost) ); gpuErrchk( cudaMemcpy(H_PQ_size,D_PQ_size, sizeof(int)*K,cudaMemcpyDeviceToHost) ); //reset nVFlag gpuErrchk( cudaMemcpy(D_nVFlag,H_nVFlag,sizeof(int)*N,cudaMemcpyHostToDevice) ); //reset next insert array gpuErrchk( cudaMemcpy(D_nV_size,H_a0,sizeof(int),cudaMemcpyHostToDevice) ); gpuErrchk( cudaMemcpy(D_expandNodes_size,H_a0,sizeof(int),cudaMemcpyHostToDevice) ); flag_PQ_not_empty = 0; for(int i=0;i<K;i++){ if(H_PQ_size[i]>0) flag_PQ_not_empty=1; } //check for mins if( *H_flagfound==1 && flag_PQ_not_empty==1){ //end gpuErrchk( cudaMemcpy(D_flagEnd,H_flagfound,sizeof(int),cudaMemcpyHostToDevice) ); checkMIN<<< numBlocks,numThreads >>>(D_PQ_size,D_flagEnd,endNode,N,K); gpuErrchk(cudaPeekAtLastError() ); cudaDeviceSynchronize(); gpuErrchk( cudaMemcpy(H_flagEnd,D_flagEnd, sizeof(int),cudaMemcpyDeviceToHost) ); printf("\ninside MIN\n"); } } printCX<<<1,1>>>(endNode); gpuErrchk( cudaMemcpy(H_parent,D_parent, sizeof(int)*N,cudaMemcpyDeviceToHost) ); if(*H_flagfound==1){ int p = endNode; while(H_parent[p]!=-1){ printf("%d ",p); p = H_parent[p]; } printf("%d\n",startNode); } else{ printf("not found\n"); } }
19,705
#include "graph.hh" #include "node.hh" #include <algorithm> #include <cassert> #include <map> #include <set> #include <string> #include <iostream> namespace rt { Graph::~Graph() { for (auto n : nodes_) delete n; } void Graph::add(Node* node) { assert(node); nodes_.push_back(node); } void Graph::remove(Node* node) { assert(node); auto it = std::find(nodes_.begin(), nodes_.end(), node); assert(it != nodes_.end()); nodes_.erase(it); for (auto x : node->preds) { auto pred_it = std::find(x->succs.begin(), x->succs.end(), node); assert(pred_it != x->succs.end()); x->succs.erase(pred_it); } for (auto x : node->succs) remove(x); delete node; } const std::vector<Node*> Graph::nodes() const { return nodes_; } namespace { void add_preds(Node* node, std::set<Node*>& set) { if (!set.insert(node).second) return; for (auto x : node->preds) add_preds(x, set); } std::vector<Node*> get_preds(Node* node, std::set<Node*>& graph) { std::vector<Node*> res; for (auto x : node->preds) if (graph.find(x) != graph.end()) res.push_back(x); return res; } std::vector<Node*> get_succs(Node* node, std::set<Node*>& graph) { std::vector<Node*> res; for (auto x : node->succs) if (graph.find(x) != graph.end()) res.push_back(x); return res; } } /** * L ← Empty list that will contain the sorted elements * S ← Set of all nodes with no incoming edge * while S is non-empty do * remove a node n from S * add n to tail of L * for each node m with an edge e from n to m do * remove edge e from the graph * if m has no other incoming edges then * insert m into S * if graph has edges then * return error (graph has at least one cycle) * else * return L (a topologically sorted order) */ std::vector<Node*> Graph::topological_sort(const std::vector<Node*>& vals) { std::vector<Node*> res; std::set<Node*> graph; for (auto n : vals) add_preds(n, graph); std::vector<Node*> s; for (auto x : graph) if (get_preds(x, graph).empty()) s.push_back(x); while (!s.empty()) { Node* next = s.back(); s.pop_back(); res.push_back(next); for (auto succ : get_succs(next, graph)) { if (get_preds(succ, graph).size() == 1) s.push_back(succ); } graph.erase(next); } if (!graph.empty()) throw std::runtime_error {"Topological sort failed"}; return res; } namespace { std::string op_name(const Node* node, std::map<const Node*, std::string>& names) { auto it = names.find(node); if (it != names.end()) return it->second; std::string res = node->type == Node::OP_NOP ? "nop" : Node::OP_NAMES[node->type]; res += ":" + std::to_string(names.size()); names[node] = res; return res; } } utils::DotGraph Graph::to_dot_graph() const { std::map<const Node*, std::string> names; utils::DotGraph g; for (auto n : nodes_) for (auto succ : n->succs) g.add_edge(op_name(n, names), op_name(succ, names)); return g; } }
19,706
#include<stdlib.h> #include<stdio.h> #include<iostream> #include<string> #include<math.h> #include<fstream> #include<sstream> #include<cuda.h> #include<time.h> #define SIZE 82000 using namespace std; size_t threadsPerBlock; size_t numberOfBlocks; int deviceId; enum color { white, black, grew }; typedef struct edge_node { int vex; edge_node *next; } edge_node; typedef struct vex_node { int vex_num; color col; edge_node *edges; edge_node *tail; } vex_node; int graph_size; __global__ void search_kernel(int N,vex_node *g,bool *findx,bool *findy,int *tmpx,int *tmpy,int x,int y) { int index= threadIdx.x + blockIdx.x * blockDim.x; int gridStride = gridDim.x * blockDim.x; for (int i = index; i < N; i += gridStride) { if(g[i].vex_num==y) { *findy=true; *tmpy=i; } if(g[i].vex_num==x) { *findx=true; *tmpx=i; } //if(findx&&findy) break; } } void printgraph(vex_node *g) { int i; int total=0,max=0; for(i=0;i<graph_size;i++) { printf("No.%d: num=%d,edges:",i,g[i].vex_num); edge_node *p=g[i].edges; total=0; while(p!=NULL) { printf("%d ",g[p->vex].vex_num); total++; p=p->next; } if(max<total) max=total; printf("\n"); } } void creategraph(char *filename,vex_node *g) { int i; int x=1,y; int sta; //int tmpx=0,tmpy=0; //bool findx,findy; //int *d_x,*d_y; int *d_tmpx,*d_tmpy; bool *d_findx,*d_findy; //cudaMallocManaged(&d_x,sizeof(int)); //cudaMallocManaged(&d_y,sizeof(int)); cudaMallocManaged(&d_tmpx,sizeof(int)); cudaMallocManaged(&d_tmpy,sizeof(int)); cudaMallocManaged(&d_findx,sizeof(bool)); cudaMallocManaged(&d_findy,sizeof(bool)); *d_tmpx=0; *d_tmpy=0; sta=0; edge_node *q; ifstream in(filename); //printf("here2\n"); while(!in.eof()) { // new input in>>x>>y; // map old point *d_findx=false; *d_findy=false; /*cudaMemcpy(d_x,&x,sizeof(int),cudaMemcpyHostToDevice); cudaMemcpy(d_y,&y,sizeof(int),cudaMemcpyHostToDevice); cudaMemcpy(d_findx,&findx,sizeof(bool),cudaMemcpyHostToDevice); cudaMemcpy(d_findy,&findy,sizeof(bool),cudaMemcpyHostToDevice);*/ numberOfBlocks=sta/threadsPerBlock+1; search_kernel<<<numberOfBlocks, threadsPerBlock>>>(sta,g,d_findx,d_findy,d_tmpx,d_tmpy,x,y); cudaDeviceSynchronize(); /*cudaMemcpy(&tmpx,d_tmpx,sizeof(int),cudaMemcpyDeviceToHost); cudaMemcpy(&tmpy,d_tmpy,sizeof(int),cudaMemcpyDeviceToHost); cudaMemcpy(&findx,d_findx,sizeof(bool),cudaMemcpyDeviceToHost); cudaMemcpy(&findy,d_findy,sizeof(bool),cudaMemcpyDeviceToHost);*/ //add new node if(!*d_findx) { *d_tmpx=sta; g[*d_tmpx].vex_num=x; g[*d_tmpx].col=white; g[*d_tmpx].edges=NULL; g[*d_tmpx].tail=NULL; sta++; } //printf("here4\n"); if(!*d_findy) { *d_tmpy=sta; g[*d_tmpy].vex_num=y; g[*d_tmpy].col=white; g[*d_tmpy].edges=NULL; g[*d_tmpx].tail=NULL; sta++; } // update node edges table edge_node *p=g[*d_tmpx].tail; cudaMallocManaged(&q, sizeof(edge_node)); q->vex=*d_tmpy; q->next=NULL; if(p==NULL) { g[*d_tmpx].edges=q; g[*d_tmpx].tail=q; } else { p->next=q; g[*d_tmpx].tail=q; } // cheak map if(y!=g[*d_tmpy].vex_num) printf("Error:map wrong!\n"); //graph_size=*sta; //printgraph(g); printf("current vex number: %d\n",sta); } graph_size=sta; } __global__ void bfs_kernel(vex_node *g,int N,bool *change) { int index= threadIdx.x + blockIdx.x * blockDim.x; int gridStride = gridDim.x * blockDim.x; for (int i = index; i < N; i += gridStride) { if(g[i].col==grew) { edge_node *p=g[i].edges; while(p!=NULL) { if(g[p->vex].col==white) { *change=true; //printf("get point\n"); } g[p->vex].col=grew; p=p->next; } g[i].col==black; } } } void bfs(vex_node *g) { int i; bool *change; int total=0; cudaMallocManaged(&change,sizeof(bool)); for(i=0;i<graph_size;i++) { if(g[i].col==white) { g[i].col=grew; *change=true; while(*change) { *change=false; //if(!*change) printf("clear\n"); total++; cudaDeviceSynchronize(); bfs_kernel<<<numberOfBlocks, threadsPerBlock>>>(g,graph_size,change); cudaDeviceSynchronize(); //if(*change) printf("next bfs\n"); } } } printf("\ntotal bfs times:%d\n",total); } int main(int argc,char *argv[]) { int numberOfSMs; cudaGetDevice(&deviceId); cudaDeviceGetAttribute(&numberOfSMs, cudaDevAttrMultiProcessorCount, deviceId); threadsPerBlock = 256; numberOfBlocks = 32 * numberOfSMs; clock_t begin,end; double time1,time2; vex_node *g; size_t size = SIZE * sizeof(vex_node); cudaMallocManaged(&g, size); begin=clock(); creategraph(argv[1],g); end=clock(); time1=(double)(end-begin)/CLOCKS_PER_SEC; printgraph(g); begin=clock(); bfs(g); end=clock(); time2=(double)(end-begin)/CLOCKS_PER_SEC; printf("\nread time:%.8fs",time1); printf("\nbfs time:%.8fs\n",time2); }
19,707
#include <iostream> #include <getopt.h> #define no_argument 0 #define required_argument 1 #define optional_argument 2 int main(int argc, char * argv[]) { std::cout << "Hello" << std::endl; const struct option long_opts[] = { {"version", no_argument, 0, 'v'}, {"help", no_argument, 0, 'h'}, {"metric_set", required_argument, 0, 'm'}, {"eval", required_argument, 0, 'e'}, {0,0,0,0}, }; int index; int iarg=0; //turn off getopt error message opterr=1; while(iarg != -1) { iarg = getopt_long_only(argc, argv, "e:m:h", long_opts, &index); switch (iarg) { case 'e': std::cout << "You hit eval" << std::endl; break; case 'm': std::cout << "You hit metric_set: " << optarg << std::endl; break; case 'h': std::cout << "You hit help" << std::endl; break; } } std::cout << "GoodBye!" << std::endl; return 0; }
19,708
#include <stdio.h> #include <algorithm> #include <cmath> __global__ void mish(int n, float* tx, float* aten_mul) { for (int i = (threadIdx.x + blockDim.x * blockIdx.x) * 4; i < n; i += gridDim.x * blockDim.x * 4) { float4 tx4 = __ldg(reinterpret_cast<float4*>(tx + i)); tx4.x = tx4.x * tanh(log1p(exp(tx4.x))); tx4.y = tx4.y * tanh(log1p(exp(tx4.y))); tx4.z = tx4.z * tanh(log1p(exp(tx4.z))); tx4.w = tx4.w * tanh(log1p(exp(tx4.w))); *reinterpret_cast<float4*>(aten_mul + i) = tx4; } } template<typename T, typename U> constexpr T ceildiv(T t, U u) { return (t + u - 1) / u; } int main() { constexpr int N = 1 << 28; float *x, *y, *d_x, *d_y; x = (float*)malloc(N * sizeof(float)); y = (float*)malloc(N * sizeof(float)); cudaMalloc(&d_x, N * sizeof(float)); cudaMalloc(&d_y, N * sizeof(float)); for (int i = 0; i < N; i++) { x[i] = 3.0f; y[i] = 2.0f; } cudaEvent_t start, stop; cudaEventCreate(&start); cudaEventCreate(&stop); cudaMemcpy(d_x, x, N * sizeof(float), cudaMemcpyHostToDevice); cudaMemcpy(d_y, y, N * sizeof(float), cudaMemcpyHostToDevice); // constexpr int blockSize = 512; // constexpr int nBlocks = ceildiv(N, blockSize); constexpr int blockSize = 512; constexpr int nBlocks = ceildiv(N, blockSize) / 4; float millis = 0.0f; float temp = 0.0f; for (int i = 0; i < 500; i++) { cudaEventRecord(start); mish<<<nBlocks, blockSize>>>(N, d_x, d_y); cudaEventRecord(stop); cudaEventSynchronize(stop); cudaEventElapsedTime(&temp, start, stop); millis += temp; } millis = millis / 500; cudaMemcpy(y, d_y, N * sizeof(float), cudaMemcpyDeviceToHost); float maxError = 0.0f; for (int i = 0; i < N; i++) { float mv = 3.0f * tanhf(std::log1p(std::exp(3.0))); maxError = std::max(maxError, std::abs(mv - y[i])); } printf("max error: %f\n", maxError); printf("duration (ms): %f\n", millis); printf("effective bandwidth (gb/s): %f\n", (float)N * sizeof(float) * 3 / millis / 1e6); cudaFree(d_x); cudaFree(d_y); free(x); free(y); return 0; }
19,709
#include <iostream> #include <string> #include <vector> #include <fstream> #include <sstream> #include <cuda.h> #define THREAD_COUNT 1024 // Max device memory : 4 GB #define MAX_MEMORY ((long long)4e9) void read_graph(std::string fname, int *&row_ptr, int *&col_ind, int &num_nodes, int &num_edges, bool zero_based = false) { std::ifstream input(fname.c_str()); if (input.fail()) throw "No file is found in the current path!"; // read graph std::string line = "%"; while (line.find("%") != std::string::npos) { getline(input, line); } std::istringstream ss(line.c_str()); ss >> num_nodes >> num_nodes >> num_edges; int edge_cnt = 0; int v1, v2; std::vector< std::vector<int> > adj_list(num_nodes); for (int i = 0; i < num_edges; i++) { getline(input, line); std::istringstream inp(line.c_str()); inp >> v1 >> v2; if (!zero_based) { v1--; // make it 0 based v2--; } if (v1 != v2) { adj_list[v1].push_back(v2); // add the edge v1->v2 adj_list[v2].push_back(v1); // add the edge v2->v1 edge_cnt++; } } input.close(); num_edges = edge_cnt; cudaMallocHost((void **)&row_ptr, sizeof(int) * (num_nodes + 1)); cudaMallocHost((void **)&col_ind, sizeof(int) * (2 * num_edges)); row_ptr[0] = 0; int index = 0; for (int v = 0; v < num_nodes; v++) { row_ptr[v + 1] = adj_list[v].size(); // assign number of edges going from node v for (int i = 0; i < (int)adj_list[v].size(); i++) { col_ind[index] = adj_list[v][i]; // put all edges in order wrt row_ptr index++; } } for (int v = 1; v < num_nodes + 1; v++) { // cumulative sum row_ptr[v] += row_ptr[v - 1]; } } __global__ void cent_kernel(float *results, int *dist, int *sigma, float *delta, int *rp, int *ci, int n) { __shared__ int level; __shared__ int improved; for(int s = blockIdx.x; s < n; s += gridDim.x) { if(threadIdx.x == 0) { results[s] = rp[s + 1] - rp[s]; // degree 1 level = 0; improved = 1; dist[blockIdx.x * n + s] = 0; sigma[blockIdx.x * n + s] = 1; } __syncthreads(); // BFS while(improved == 1) { if(threadIdx.x == 0) improved = 0; for(int node = threadIdx.x; node < n; node += blockDim.x) { for(int edge = rp[node]; edge < rp[node + 1]; edge++) { int &adj = ci[edge]; if(dist[(blockIdx.x * n) + adj] == level && dist[(blockIdx.x * n) + node] == -1) { dist[(blockIdx.x * n) + node] = level + 1; improved = 1; } if(dist[(blockIdx.x * n) + adj] == level && dist[(blockIdx.x * n) + node] == level + 1) { sigma[(blockIdx.x * n) + node] += (float) sigma[(blockIdx.x * n) + adj]; } } } if(threadIdx.x == 0) level++; __syncthreads(); } int dist_sum = 0; int dist2_cnt = 0; // DISTANCE ADDER if(threadIdx.x == 0) { for(int i = 0; i < n; i++) { if(dist[(blockIdx.x * n) + i] > 0) { if(dist[(blockIdx.x * n) + i] <= 2) dist2_cnt++; dist_sum += dist[(blockIdx.x * n) + i]; } } results[n + s] = dist2_cnt; // degree 2 results[2 * n + s] = (float) n / dist_sum; // closeness cent. } while(level > 0) { for(int node = threadIdx.x; node < n; node += blockDim.x) { if(dist[blockIdx.x * n + node] == level){ for(int edge = rp[node]; edge < rp[node + 1]; edge++) { int adj = ci[edge]; if(dist[(blockIdx.x * n) + adj] + 1 == dist[(blockIdx.x * n) + node]) { atomicAdd(&delta[(blockIdx.x * n) + adj], (sigma[(blockIdx.x * n) + adj] * 1.0) / sigma[(blockIdx.x * n) + node] * (1 + delta[(blockIdx.x * n) + node])); } } atomicAdd(&results[3 * n + node], delta[(blockIdx.x * n) + node] / 2); } } if(threadIdx.x == 0) level--; __syncthreads(); } for(int i = 0; i < n; i++) { dist[(blockIdx.x * n) + i] = -1; sigma[(blockIdx.x * n) + i] = 0; delta[(blockIdx.x * n) + i] = 0; } } } float* compute_centralities(int *rp, int *ci, int n, float &time_taken) { const int BLOCK_COUNT = MAX_MEMORY / (4 * 3 * n); int *sigma, *dist; float *delta, *d_results; cudaMalloc((void **)&d_results, sizeof(float) * n * 4); cudaMalloc((void **)&sigma, sizeof(int) * n * BLOCK_COUNT); cudaMalloc((void **)&dist, sizeof(int) * n * BLOCK_COUNT); cudaMalloc((void **)&delta, sizeof(float) * n * BLOCK_COUNT); cudaMemset(dist, -1, sizeof(int) * n * BLOCK_COUNT); cudaMemset(sigma, 0, sizeof(int) * n * BLOCK_COUNT); cudaMemset(delta, 0, sizeof(float) * n * BLOCK_COUNT); cudaMemset(d_results, 0, sizeof(float) * 4 * n); cudaEvent_t start, end; cudaEventCreate(&start); cudaEventCreate(&end); cudaEventRecord(start); cent_kernel<<<BLOCK_COUNT, THREAD_COUNT>>>(d_results, dist, sigma, delta, rp, ci, n); cudaDeviceSynchronize(); cudaEventRecord(end); cudaEventSynchronize(end); cudaEventElapsedTime(&time_taken, start, end); float *results; cudaMallocHost((void **)&results, sizeof(float) * n * 4); cudaMemcpy(results, d_results, sizeof(float) * n * 4, cudaMemcpyDeviceToHost); cudaFree(sigma); cudaFree(dist); cudaFree(delta); cudaFree(d_results); cudaDeviceSynchronize(); return results; } int main() { cudaSetDevice(0); std::string filename = "../data/wave.mtx"; int *row_ptr, *col_ind; int num_nodes, num_edges; read_graph(filename, row_ptr, col_ind, num_nodes, num_edges); printf("[INFO] Graph is read: %s.\n", filename.c_str()); int *rp; int *ci; cudaMalloc((void **)&rp, sizeof(int) * (num_nodes + 1)); cudaMalloc((void **)&ci, sizeof(int) * (num_edges * 2)); printf("[INFO] CUDA memory parameters are allocated for kernel function.\n"); cudaMemcpy(rp, row_ptr, sizeof(int) * (num_nodes + 1), cudaMemcpyHostToDevice); cudaMemcpy(ci, col_ind, sizeof(int) * (num_edges * 2), cudaMemcpyHostToDevice); printf("[INFO] CUDA memory parameters are set for kernel function.\n"); float time_taken; float *results = compute_centralities(rp, ci, num_nodes, time_taken); printf("[INFO] Kernel function is finished.\n"); printf("Centrality Results:\n"); for (int i = 0; i < num_nodes; i++) { printf("%.5f; %.5f; %.5f; %.5f\n ", results[i], results[num_nodes + i], results[2 * num_nodes + i], results[3 * num_nodes + i]); } printf("[INFO] Kernel call is ended in: %.5f ms.\n", time_taken); cudaFreeHost(results); cudaFreeHost(row_ptr); cudaFreeHost(col_ind); cudaFree(rp); cudaFree(ci); return 0; }
19,710
#include "includes.h" __global__ void RemoveEdgesKernel( int *connection, int *age, int maxAge, int *activityFlag, float *winningFraction, int *winningCount, float *utility, float *localError, int *neuronAge, int maxCells ) { int threadId = blockDim.x*blockIdx.y*gridDim.x //rows preceeding current row in grid + blockDim.x*blockIdx.x //blocks preceeding current block + threadIdx.x; if(threadId < maxCells) { if(activityFlag[threadId] == 1) { neuronAge[threadId] = neuronAge[threadId] + 1; // TO DO : GET RID OF IFs & ELSEs int activeConnections = 0; int connId; for(int c = 0; c < maxCells; c++) { connId = threadId * maxCells + c; if(connection[connId] == 1) { if(age[connId] <= maxAge) { activeConnections++; } else { connection[connId] = 0; age[connId] = 0; } } } if(activeConnections == 0) { activityFlag[threadId] = 0; localError[threadId] = 0.00f; neuronAge[threadId] = 0; winningFraction[threadId] = 0.00f; winningCount[threadId] = 0; utility[threadId] = 0.00f; } } } }
19,711
# include <bits/stdc++.h> # include <cuda.h> #define SIZE 60000000// Global Size #define BLOCK_SIZE 1024 using namespace std; //::::::::::::::::::::::::::::::::::::::::::GPU:::::::::::::::::::::::::::::::: // :::: Kernel __global__ void KernelNormalVec(double *g_idata,double *g_odata,int l){ // Sequential Addressing technique __shared__ double sdata[BLOCK_SIZE]; // each thread loads one element from global to shared mem unsigned int tid = threadIdx.x; unsigned int i = blockIdx.x*blockDim.x + threadIdx.x; if(i<l){ // bad thing -> severely punished performance. sdata[tid] = g_idata[i]; }else{ sdata[tid] = 0.0; } __syncthreads(); // do reduction in shared mem for(unsigned int s=blockDim.x/2; s > 0; s >>= 1) { if(tid < s){ sdata[tid] += sdata[tid + s]; } __syncthreads(); } // write result for this block to global mem if (tid == 0) g_odata[blockIdx.x] = sdata[0]; } // :::: Calls void d_VectorMult(double *Vec1,double *Total){ double * d_Vec1; double * d_Total; double Blocksize=BLOCK_SIZE; // Block of 1Dim cudaMalloc((void**)&d_Vec1,SIZE*sizeof(double)); cudaMalloc((void**)&d_Total,SIZE*sizeof(double)); cudaMemcpy(d_Vec1, Vec1,SIZE*sizeof(double),cudaMemcpyHostToDevice); cudaMemcpy(d_Total,Total,SIZE*sizeof(double),cudaMemcpyHostToDevice); int temp=SIZE; while(temp>1){ dim3 dimBlock(Blocksize,1,1); int grid=ceil(temp/Blocksize); dim3 dimGrid(grid,1,1); KernelNormalVec<<<dimGrid,dimBlock>>>(d_Vec1,d_Total,temp); cudaDeviceSynchronize(); cudaMemcpy(d_Vec1,d_Total,SIZE*sizeof(double),cudaMemcpyDeviceToDevice); temp=ceil(temp/Blocksize); } cudaMemcpy(Total,d_Total,SIZE*sizeof(double),cudaMemcpyDeviceToHost); cudaFree(d_Vec1); cudaFree(d_Total); } //::::::::::::::::::::::::::::::::::::::::::CPU:::::::::::::::::::::::::::::::: void h_sum_vec(double *Vec1, double *all){ for(int i=0;i<SIZE;i++) *all+=Vec1[i]; } //:::::::::::::::::::::::::::: Rutinary Functions void Fill_vec(double *Vec,double Value){ for(int i =0 ; i<SIZE ; i++) Vec[i]=Value; } void Show_vec(double *Vec){ for (int i=0;i<SIZE;i++){ if(i%10==0 && i!=0){ cout<<endl; } cout<<"["<<Vec[i]<<"] "; } cout<<endl; } void Checksum(double *Answer1 , double *Answer2){ if(fabs(Answer1[0]-Answer2[0]) < 0.1) cout<<"Nice Work Guy"<<endl; else cout<<"BAD Work Guy"<<endl; } // :::::::::::::::::::::::::::::::::::Clock Function:::::::::::::::::::::::::::: double diffclock(clock_t clock1,clock_t clock2){ double diffticks=clock2-clock1; double diffms=(diffticks)/(CLOCKS_PER_SEC/1); // /1000 mili return diffms; } // :::::::::::::::::::::::::::::::::::::::Main::::::::::::::::::::::::::::::::. int main(){ double T1,T2; // Time flags double *Vec1 = (double*)malloc((SIZE)*sizeof(double)); // Elements to compute. CPU way double *Total2 = (double*)malloc((SIZE)*sizeof(double)); // GPU double *Total1 = (double*)malloc(sizeof(double)); // Total Variables. // Fill the containers vectors of data Fill_vec(Vec1,1.0); Fill_vec(Total2,0.0); // Register time to finish the algorithm // Secuential clock_t start = clock(); h_sum_vec(Vec1,Total1); clock_t end = clock(); T1=diffclock(start,end); cout<<"Serial Result: "<<*Total1<<" At "<<T1<<",Seconds"<<endl; // Parallel start = clock(); d_VectorMult(Vec1,Total2); end = clock(); T2=diffclock(start,end); cout<<"Parallel Result: "<<Total2[0]<<" At "<<T2<<",Seconds"<<endl; cout<<"Total Acceleration: "<<T1/T2<<",X"<<endl; Checksum(Total1,Total2); // releasing Memory free(Vec1); free(Total2); return 0; }
19,712
/* To compile: nvcc --arch=sm_60 -O3 -o mandelbrot mandelbrot.c -lm To create an image with 4096 x 4096 pixels ./mandelbrot 4096 4096 */ #include <math.h> #include <stdio.h> #include <stdlib.h> #include <time.h> int writeMandelbrot(const char *fileName, int width, int height, float *img, int minI, int maxI); // Q2a: add include for CUDA header file here: #define MXITER 1000 typedef struct { double r; double i; }complex_t; // return iterations before z leaves mandelbrot set for given c __device__ int testpoint(complex_t c){ int iter; complex_t z; double temp; z = c; for(iter=0; iter<MXITER; iter++){ temp = (z.r*z.r) - (z.i*z.i) + c.r; z.i = z.r*z.i*2. + c.i; z.r = temp; if((z.r*z.r+z.i*z.i)>4.0){ return iter; } } return iter; } // perform Mandelbrot iteration on a grid of numbers in the complex plane // record the iteration counts in the count array __global__ void mandelbrotKernel(int Nre, int Nim, complex_t cmin, complex_t dc, float *count){ // Q2c: replace this loop with a CUDA kernel // for(int n=0;n<Nim;++n){ // for(int m=0;m<Nre;++m){ int n = threadIdx.y + blockIdx.y*blockDim.y; int m = threadIdx.x + blockIdx.x*blockDim.x; if(n<Nim && m<Nre){ complex_t c; c.r = cmin.r + dc.r*m; c.i = cmin.i + dc.i*n; count[m+n*Nre] = (float) testpoint(c); } } int main(int argc, char **argv){ // to create a 4096x4096 pixel image // usage: ./mandelbrot 4096 4096 int Nre = (argc==3) ? atoi(argv[1]): 4096; int Nim = (argc==3) ? atoi(argv[2]): 4096; // Q2b: set the number of threads per block and the number of blocks here: // storage for the iteration counts float *count; count = (float*) malloc(Nre*Nim*sizeof(float)); // Parameters for a bounding box for "c" that generates an interesting image const float centRe = -.759856, centIm= .125547; const float diam = 0.151579; complex_t cmin; complex_t cmax; complex_t dc; cmin.r = centRe - 0.5*diam; cmax.r = centRe + 0.5*diam; cmin.i = centIm - 0.5*diam; cmax.i = centIm + 0.5*diam; //set step sizes dc.r = (cmax.r-cmin.r)/(Nre-1); dc.i = (cmax.i-cmin.i)/(Nim-1); cudaEvent_t start, end; cudaEventCreate(&start); cudaEventCreate(&end); float *c_count; cudaMalloc(&c_count, Nre*Nim*sizeof(float)); int BX = 16, BY = 16; int GX = (Nre+BX-1)/BX, GY = (Nim+BY-1)/BY; dim3 B(BX,BY,1); dim3 G(GX,GY,1); // mark event cudaEventRecord(start); // compute mandelbrot set mandelbrotKernel <<< G, B >>> (Nre, Nim, cmin, dc, c_count); // mark event cudaEventRecord(end); // copy from the GPU back to the host here cudaMemcpy(count, c_count, Nre*Nim*sizeof(float), cudaMemcpyDeviceToHost); // print elapsed time float elapsed; cudaEventSynchronize(end); cudaEventElapsedTime(&elapsed, start, end); elapsed /= 1000.; printf("elapsed = %f\n", elapsed); // output mandelbrot to ppm format image printf("Printing mandelbrot.ppm..."); writeMandelbrot("mandelbrot.ppm", Nre, Nim, count, 0, 80); printf("done.\n"); free(count); exit(0); return 0; } /* Output data as PPM file */ void saveppm(const char *filename, unsigned char *img, int width, int height){ /* FILE pointer */ FILE *f; /* Open file for writing */ f = fopen(filename, "wb"); /* PPM header info, including the size of the image */ fprintf(f, "P6 %d %d %d\n", width, height, 255); /* Write the image data to the file - remember 3 byte per pixel */ fwrite(img, 3, width*height, f); /* Make sure you close the file */ fclose(f); } int writeMandelbrot(const char *fileName, int width, int height, float *img, int minI, int maxI){ int n, m; unsigned char *rgb = (unsigned char*) calloc(3*width*height, sizeof(unsigned char)); for(n=0;n<height;++n){ for(m=0;m<width;++m){ int id = m+n*width; int I = (int) (768*sqrt((double)(img[id]-minI)/(maxI-minI))); // change this to change palette if(I<256) rgb[3*id+2] = 255-I; else if(I<512) rgb[3*id+1] = 511-I; else if(I<768) rgb[3*id+0] = 767-I; else if(I<1024) rgb[3*id+0] = 1023-I; else if(I<1536) rgb[3*id+1] = 1535-I; else if(I<2048) rgb[3*id+2] = 2047-I; } } saveppm(fileName, rgb, width, height); free(rgb); return 1; }
19,713
#include "includes.h" __device__ void convolution(int conv_col, int conv_row, float *d_kernel, int k_size, float *d_matrix, int size_x, int size_y, float *d_conv, int max_row, int max_col){ int conv_index = conv_col+ conv_row*max_col; d_conv[conv_index] = 0; for(int k_row = 0; k_row < k_size; k_row ++){ for(int k_col = 0; k_col < k_size ; k_col ++){ d_conv[conv_index] += d_kernel[k_col + (k_row*k_size)] * d_matrix[(conv_col+k_col) + (conv_row+k_row)*size_x]; // printf("row %i col %i d_conv[] = %f \n", row, col, d_conv[col+ row*max_col]); } } } __global__ void valid_convolution(float *d_kernel, int k_size, float *d_matrix, int size_x, int size_y, float *d_conv, int max_row, int max_col){ int col = threadIdx.x + blockIdx.x * blockDim.x; int row = threadIdx.y + blockIdx.y * blockDim.y; if(max_row > row && max_col > col){ convolution(col, row, d_kernel, k_size, d_matrix, size_x, size_y, d_conv, max_row, max_col); } }
19,714
/* * Connected Components in the GPU * Paper Source: An Optimized Union-Find Algorithm for Connected * Components Labeling Using GPUs * Adapted from: https://github.com/victormatheus/CCL-GPU * Modified by: Imanol Luengo <imaluengo@gmail.com> */ typedef unsigned int uint32; #define MAX_UINT32 0xFFFFFFFF __device__ uint32 find(uint32* buf, uint32 x) { uint32 p = x; while ( x != buf[x] ) { x = buf[x]; } buf[p] = x; // fast linking return x; } __device__ void findAndUnion(uint32* buf, uint32 g1, uint32 g2) { bool done; uint32 old; do { g1 = find(buf, g1); g2 = find(buf, g2); if (g1 < g2) { old = atomicMin(&buf[g2], g1); done = (old == g2); g2 = old; } else if (g2 < g1) { old = atomicMin(&buf[g1], g2); done = (old == g1); g1 = old; } else { done = true; } } while ( !done ); } __global__ void uf_local(const uint32* in, uint32* out, int3 shape) { int3 p; p.z = blockIdx.z * blockDim.z + threadIdx.z; p.y = blockIdx.y * blockDim.y + threadIdx.y; p.x = blockIdx.x * blockDim.x + threadIdx.x; long image_plane = (shape.y * shape.x); long block_plane = (blockDim.y * blockDim.x); uint32 gidx = p.z * image_plane + p.y * shape.x + p.x; uint32 bidx = threadIdx.z * block_plane + \ threadIdx.y * blockDim.x + \ threadIdx.x; long bsize = blockDim.z * blockDim.y * blockDim.x; extern __shared__ uint32 s_buffer[]; bool in_limits = p.z < shape.z && p.y < shape.y && p.x < shape.x; s_buffer[bidx] = bidx; s_buffer[bsize + bidx] = in_limits? in[p.z * image_plane + p.y * shape.x + p.x] : 0; __syncthreads(); if ( !in_limits ) {return;} uint32 v = s_buffer[bsize + bidx]; if ( threadIdx.x > 0 && s_buffer[bsize + bidx - 1] == v ) { findAndUnion(s_buffer, bidx, bidx - 1); } __syncthreads(); if ( threadIdx.y > 0 && s_buffer[bsize + bidx - blockDim.x] == v ) { findAndUnion(s_buffer, bidx, bidx - blockDim.x); } __syncthreads(); if ( threadIdx.z > 0 && s_buffer[bsize + bidx - block_plane] == v ) { findAndUnion(s_buffer, bidx, bidx - block_plane); } __syncthreads(); uint32 f = find(s_buffer, bidx); uint32 aux = f % block_plane; uint32 fz = f / block_plane; uint32 fy = aux / blockDim.x; uint32 fx = aux % blockDim.x; out[gidx] = (blockIdx.z * blockDim.z + fz) * image_plane + \ (blockIdx.y * blockDim.y + fy) * shape.x + \ (blockIdx.x * blockDim.x + fx); } __global__ void uf_global(const uint32* in, uint32* out, int3 shape) { int3 p; p.z = blockIdx.z * blockDim.z + threadIdx.z; p.y = blockIdx.y * blockDim.y + threadIdx.y; p.x = blockIdx.x * blockDim.x + threadIdx.x; long image_plane = (shape.y * shape.x); uint32 gidx = p.z * image_plane + p.y * shape.x + p.x; if ( p.z >= shape.z || p.y >= shape.y || p.x >= shape.x ) { return; } uint32 v = in[gidx]; if ( p.z > 0 && threadIdx.z == 0 && in[gidx - image_plane] == v ) { findAndUnion(out, gidx, gidx - image_plane); } if ( p.y > 0 && threadIdx.y == 0 && in[gidx - shape.x] == v ) { findAndUnion(out, gidx, gidx - shape.x); } if ( p.x > 0 && threadIdx.x == 0 && in[gidx - 1] == v ) { findAndUnion(out, gidx, gidx - 1); } } __global__ void uf_final(uint32* labels, int3 shape) { int3 p; p.z = blockIdx.z * blockDim.z + threadIdx.z; p.y = blockIdx.y * blockDim.y + threadIdx.y; p.x = blockIdx.x * blockDim.x + threadIdx.x; long gidx = p.z * shape.y * shape.x + p.y * shape.x + p.x; if ( p.z < shape.z && p.y < shape.y && p.x < shape.x ) { labels[gidx] = find(labels, gidx); } }
19,715
#include <iostream> #include <cstdio> #include <cstdlib> // #include <helper_cuda.h> // #include <helper_string.h> /* Run with only HOST code *\ // Say goodbye to the universe int main(void) { printf("Heat death boogaloo!\n"); return 0; } */ /* Run with DEVICEEEEEEE code */ __global__ void mykernel(void) { } int main(void) { mykernel<<<1,1>>>(); printf("HEat death boogaloo Universe!\n"); return 0; } /* blah blah comment not conflicting __global__ indicates function that 1) runs on DEVICE , 2) called from the HOST <<< >>> indicates function that runs on DEVICE and is called from the HOST *////////
19,716
/* Program to add 2 matrics of size M * N in CUDA C++ Using grid of one block Block contains M*N threads arranged in M rows and N columns */ #include<iostream> #include "cuda.h" #include "cuda_runtime.h" #define M 2 #define N 9 __global__ void matAdd(int* a, int* b, int* c) { int idx = threadIdx.x * blockDim.y + threadIdx.y; c[idx] = a[idx] + b[idx]; } __host__ void print_matrix(int* matrix) { for(int i=0; i<M; ++i) { for(int j=0; j<N; ++j) { std::cout<<matrix[i*N+j]<<' '; } std::cout<<"\n"; } std::cout<<"\n"; } int main() { int size = M * N * sizeof(int); int* a = new int[size]; int* b = new int[size]; int* c = new int[size]; for(int i=0; i<M; ++i) { for(int j=0; j<N; ++j) { a[i*N + j] = i; //Fill your own values here b[i*N + j] = j; //Fill your own values here } } std::cout<<"MATRIX A =\n"; print_matrix(a); std::cout<<"MATRIX B =\n"; print_matrix(b); /* Setting up variables on device. i.e. GPU */ int *d_a, *d_b, *d_c; cudaMalloc((void**)&d_a, size); cudaMalloc((void**)&d_b, size); cudaMalloc((void**)&d_c, size); /* Copy data from host to device */ cudaMemcpy(d_a, a, size, cudaMemcpyHostToDevice); cudaMemcpy(d_b, b, size, cudaMemcpyHostToDevice); /* Kernel Launch Grid contains 1 block Each block has M*N threads arranged in M*N rectangle Hence index of vector is idx = threadIdx.x * blockDim.y + threadIdx.y; */ dim3 blockSize(M, N); matAdd<<<1, blockSize>>>(d_a, d_b, d_c); cudaDeviceSynchronize(); /* Copy result from GPU device to host */ cudaMemcpy(c, d_c, size, cudaMemcpyDeviceToHost); /* Print result */ std::cout<<"A + B =\n"; print_matrix(c); std::cout<<'\n'; /* Cleanup device and host memory */ cudaFree(d_a); cudaFree(d_b); cudaFree(d_c); delete a; delete b; delete c; return 0; }
19,717
#include <thrust/device_vector.h> #include <thrust/host_vector.h> #include <thrust/sequence.h> #include <thrust/transform.h> #include <math.h> #include <stdio.h> #define N 64 using namespace thrust::placeholders; // Define transformation SqrtOf()(x) -> sqrt(x) struct SqrtOf { __host__ __device__ float operator()(float x) { return sqrt(x); } }; int main() { const float ref = 0.5; thrust::device_vector<float> dvec_x(N); thrust::device_vector<float> dvec_dist(N); thrust::sequence(dvec_x.begin(), dvec_x.end()); thrust::transform(dvec_x.begin(), dvec_x.end(), dvec_x.begin(), _1 / (N - 1)); thrust::transform(dvec_x.begin(), dvec_x.end(), dvec_dist.begin(), (_1 - ref)*(_1 - ref)); thrust::transform(dvec_dist.begin(), dvec_dist.end(), dvec_dist.begin(), SqrtOf()); thrust::host_vector<float> h_x = dvec_x; thrust::host_vector<float> h_dist = dvec_dist; for (int i = 0; i < N; ++i) { printf("x=%3.3f, dist=%3.3f\n", h_x[i], h_dist[i]); } return 0; }
19,718
#include <stdio.h> __global__ void cuda_hello_world() { printf("Hello World from GPU! [ThreadID = %d, BlockID = %d]\n", threadIdx.x, blockIdx.x); } int main() { cuda_hello_world<<<1, 256>>>(); cudaDeviceSynchronize(); return 0; }
19,719
#include "includes.h" __global__ void naive_backward_cross_entropy(float *in, int *one_hot_classes, float batches, int size, float *out) { int bid = blockIdx.x * blockDim.x + threadIdx.x; if (!(bid < size)) return; out[bid] = (in[bid] - one_hot_classes[bid]) / batches; }
19,720
#include "includes.h" /* Program Parameters */ #define MAXN 8000 /* Max value of N */ int N; /* Matrix size */ // Thread block size #define BLOCK_SIZE 16 /* Matrices */ float A[MAXN][MAXN], B[MAXN][MAXN]; /* junk */ #define randm() 4|2[uid]&3 /* Prototype */ /* ------------------ Cuda Code --------------------- */ /****** You will replace this routine with your own parallel version *******/ /* Provided global variables are MAXN, N, A[][] and B[][], * defined in the beginning of this code. B[][] is initialized to zeros. */ /* returns a seed for srand based on the time */ __global__ void matrixSD(float* d_in, float* d_mean, float* d_sd, int N) { extern __shared__ float sdata1[]; //each thread loads one element from global to shared mem int idx_x = blockIdx.x * blockDim.x + threadIdx.x; int idx_y = blockIdx.y * blockDim.y + threadIdx.y; unsigned int tid = threadIdx.y; unsigned int i = idx_y * N + idx_x; sdata1[tid] = powf(d_in[i] - d_mean[blockIdx.x], 2.0); __syncthreads(); // do reduction in shared mem for(unsigned int s=1; s < blockDim.y; s *= 2) { if(tid +s < N) { if(tid % (2*s) == 0) { sdata1[tid] += sdata1[tid + s]; } } __syncthreads(); } // write result for this block to global mem if(tid == 0) d_sd[blockIdx.x] = sqrtf(sdata1[0]/(float) N); }
19,721
//Alfred Shaker //10/23/2015 //Homework 2 #include <stdio.h> //function to get and print device properties void printDeviceProperties(cudaDeviceProp devProp) { //get the cuda driver version int driverVersion = 0; cudaDriverGetVersion(&driverVersion); printf("Version Number: %d\n",driverVersion/1000 ); //get the number of multiprocessors int mp = devProp.multiProcessorCount; printf("Number of MultipPrcessors: %d\n", mp); //check the computation capacity and calculate cores based on that int cores = 0; switch(devProp.major) { //Fermi case 2: if (devProp.minor == 1){ cores = mp * 48; printf("Number of cores: %d\n", cores); } else{ cores = mp * 32; printf("Number of cores: %d\n", cores); } break; //Kepler case 3: cores = mp * 192; printf("Number of cores: %d\n", cores); break; //Maxwell case 5: cores = mp * 128; printf("Number of cores: %d\n", cores); break; default: printf("Unknown device type\n"); break; } int kb = 1024; int mb = kb * kb; //get the total global memory in megabytes int globalMemory = devProp.totalGlobalMem / mb; printf("Total Global Memory: %d mb\n", globalMemory); //get the shared memory per block in kilobytes int sharedMemory = devProp.sharedMemPerBlock / kb; printf("Shared Memory Per Block: %d kb\n", sharedMemory); //get the maximum number of threads per block int maxThreads = devProp.maxThreadsPerBlock; printf("Maximum Threads Per Block: %d\n", maxThreads); //get the maximum dimension size of the block int maxBlockDim = devProp.maxThreadsDim[0]; printf("Maximum Size of Block Dimensions: %d\n", maxBlockDim); //get the maximum dimension size for the grid int maxGridDim = devProp.maxGridSize[0]; printf("Maximum Size of Grid Dimensions: %d\n", maxGridDim); } int main() { //get number of devices int deviceCount; cudaGetDeviceCount(&deviceCount); printf("Total number of Devices: %d\n", deviceCount); //iterate through each device and get properties of each one for(int i = 0; i < deviceCount; ++i) { printf("Showing info for device number %d\n", i+1); cudaDeviceProp devProp; cudaGetDeviceProperties(&devProp, i); printDeviceProperties(devProp); } return 0; }
19,722
#include <stdlib.h> #include <stdio.h> #include <vector> #include <numeric> #include <iostream> #include<chrono> // Here you can set the device ID that was assigned to you #define MYDEVICE 0 constexpr bool DEBUG = false; constexpr size_t BLOCK_SIZE = 512; constexpr size_t SHARE_BLOCK_SIZE = 2*512; double random_double(void) { return static_cast<double>(rand()) / RAND_MAX; } // Simple utility function to check for CUDA runtime errors void checkCUDAError(const char *msg); // Part 1 of 6: implement the kernel __global__ void block_sum(const double *input, double *per_block_results, const size_t n) { //fill me __shared__ double sdata[SHARE_BLOCK_SIZE]; int g_index = threadIdx.x + blockIdx.x * blockDim.x; int s_index = threadIdx.x; sdata[s_index] = (g_index < n) ? input[g_index] : 0.; // Synchronize (ensure all the data is available) __syncthreads(); int sdata_size = SHARE_BLOCK_SIZE; bool loop = (n == 1) ? false : true; while (__syncthreads_or(loop) ) { int r_index = sdata_size-1-s_index; // reduction origami if(r_index >= (sdata_size+1)/2 && r_index <= sdata_size) { sdata[s_index] += sdata[r_index]; // sdata[r_index] = 0; } sdata_size = (sdata_size+1) / 2; if(sdata_size==1) loop = false; } if (s_index==0) per_block_results[blockIdx.x] = sdata[s_index]; if (DEBUG) { if (blockIdx.x < 20 && (g_index < n) && s_index < 3 ) printf("per_block_results[%d]] %f %d\n",blockIdx.x,per_block_results[blockIdx.x],s_index); } } //////////////////////////////////////////////////////////////////////////////// // Program main //////////////////////////////////////////////////////////////////////////////// int main(int argc, char* argv[]) { // create array of 256ki elements // 1<<N is the equivalent to 2^N const int num_elements = (argc > 1) ? std::atoi(argv[1]) : 1<<18; // bool DEBUG = (num_elements < 1000 ? true : false); srand(time(NULL)); // generate random input on the host std::vector<double> h_input(num_elements); for(int i = 0; i < h_input.size(); ++i) { // h_input[i] = random_double(); h_input[i] = double(i); if (DEBUG) std::cout << " " << h_input[i]; } if (DEBUG) std::cout << std::endl; { auto t0 = std::chrono::high_resolution_clock::now(); const double host_result = std::accumulate(h_input.begin(), h_input.end(), 0.); auto t1 = std::chrono::high_resolution_clock::now(); std::chrono::duration<float> d = t1 - t0; std::cerr << "Host sum: " << host_result << " in " << d.count() << " s" << std::endl; } // WITH ONLY 1 STREAM IS NOT NECESSARY // cudaEvent_t cuda_start, cuda_stop; // cudaEventCreate(&cuda_start); // cudaEventCreate(&cuda_stop); { auto t0 = std::chrono::high_resolution_clock::now(); //Part 1 of 6: move input to device memory size_t memSize = num_elements*sizeof(double); double *d_input = 0; cudaMalloc( &d_input,memSize ); cudaMemcpyAsync( d_input, &h_input[0], memSize, cudaMemcpyHostToDevice ); auto t = std::chrono::high_resolution_clock::now(); { auto t1 = std::chrono::high_resolution_clock::now(); std::chrono::duration<float> d = t1 - t0; std::cout << "h2d cudaMemcpyAsync " << d.count() << " s" << std::endl; } dim3 num_blocks, block_size; block_size.x = BLOCK_SIZE; num_blocks.x = ( num_elements + block_size.x - 1 )/block_size.x; // Part 2 of 6: allocate the partial sums: How much space does it need? double *d_partial_sums_and_total = 0; memSize = num_blocks.x*sizeof(double); cudaMalloc( &d_partial_sums_and_total,memSize ); t = std::chrono::high_resolution_clock::now(); { // cudaEventRecord(cuda_start); // Part 3 of 6: launch one kernel to compute, per-block, a partial sum. How much shared memory does it need? block_sum<<<num_blocks,block_size>>>(d_input, d_partial_sums_and_total, num_elements); auto t1 = std::chrono::high_resolution_clock::now(); std::chrono::duration<float> d = t1 - t0; std::cout << "block_sum<<<num_blocks,block_size>>> " << d.count() << " s " << std::chrono::duration<float>(t1-t).count() << " s" << std::endl; } t = std::chrono::high_resolution_clock::now(); { // Part 4 of 6: compute the sum of the partial sums block_sum<<<1,block_size>>>(d_partial_sums_and_total, &d_partial_sums_and_total[0], num_blocks.x); auto t1 = std::chrono::high_resolution_clock::now(); std::chrono::duration<float> d = t1 - t0; std::cout << "block_sum<<<1,block_size>>> " << d.count() << " s " << std::chrono::duration<float>(t1-t).count() << " s" << std::endl; } // check if kernel execution generated an error checkCUDAError("kernel execution"); t = std::chrono::high_resolution_clock::now(); { // Part 5 of 6: copy the result back to the host double device_result = 0; memSize = sizeof(double); cudaMemcpyAsync(&device_result, &d_partial_sums_and_total[0], memSize, cudaMemcpyDeviceToHost); // Part 6 of 6: deallocate device memory cudaFree(d_input); cudaFree(d_partial_sums_and_total); auto t1 =std::chrono::high_resolution_clock::now(); std::chrono::duration<float> d = t1 - t0; // float milliseconds = 0; // cudaEventElapsedTime(&milliseconds, cuda_start, cuda_stop); std::cout << "Device sum: " << device_result << " in " << d.count() << " s " << std::chrono::duration<float>(t1-t).count() << " s" << std::endl; } } return 0; } void checkCUDAError(const char *msg) { cudaError_t err = cudaGetLastError(); if( cudaSuccess != err) { std::cerr << "Cuda error: " << msg << " " << cudaGetErrorString(err) << std::endl; exit(-1); } }
19,723
#include <iostream> __global__ void vectorAdd(int *a, int *b, int *c, int n){ int i = blockIdx.x*blockDim.x+threadIdx.x; if(i<n) for(int j=0;j<100;j++) c[i] = a[i] + b[i]; } int main(void){ int * a, * b, * c; int * d_a, * d_b, * d_c; int * temp; int n = 1<<24; a = new int[n*sizeof(int)]; b = new int[n*sizeof(int)]; c = new int[n*sizeof(int)]; temp = new int[n*sizeof(int)]; cudaMalloc(&d_a, n*sizeof(int)); cudaMalloc(&d_b, n*sizeof(int)); cudaMalloc(&d_c, n*sizeof(int)); for(int i=0;i<n;i++){ a[i] = 3; b[i] = 5; } int blockSize = 256; int numBlocks = n/256; cudaMemcpy(d_a, a, n*sizeof(int), cudaMemcpyHostToDevice); cudaMemcpy(d_b, b, n*sizeof(int), cudaMemcpyHostToDevice); vectorAdd<<<numBlocks,blockSize>>>(d_a,d_b,d_c,n); cudaDeviceSynchronize(); temp[0] = a[0]; for(int i=1;i<n;i++) temp[i] = temp[i-1]+a[i]; vectorAdd<<<numBlocks, blockSize>>>(d_a,d_b,d_c,n); cudaDeviceSynchronize(); temp[0] = a[0]; for(int i=1;i<n;i++) temp[i] = temp[i-1]+a[i]; cudaDeviceSynchronize(); temp[0] = b[0]; for(int i=1;i<n;i++) temp[i] = temp[i-1]+b[i]; vectorAdd<<<numBlocks, blockSize>>>(d_a,d_b,d_c,n); cudaDeviceSynchronize(); cudaMemcpy(c, d_c, n*sizeof(int), cudaMemcpyDeviceToHost); temp[0] = c[0]; for(int i=1;i<n;i++) temp[i] = temp[i-1]+c[i]; vectorAdd<<<numBlocks, blockSize>>>(d_a,d_b,d_c,n); cudaDeviceSynchronize(); vectorAdd<<<numBlocks, blockSize>>>(d_a,d_b,d_c,n); cudaDeviceSynchronize(); cudaFree(a); cudaFree(b); cudaFree(c); delete temp; return 0; }
19,724
#include <iostream> using namespace std; #include <thrust/reduce.h> #include <thrust/sequence.h> #include <thrust/host_vector.h> #include <thrust/device_vector.h> __global__ void fillKernel(int *a, int n) { int tid = blockIdx.x * blockDim.x + threadIdx.x; if (tid < n) a[tid] = tid; } void fill(int *d_a, int n) { int nThreadsPerBlock = 512; int nBlocks = n / nThreadsPerBlock + ((n % nThreadsPerBlock) ? 1 : 0); fillKernel <<< nBlocks, nThreadsPerBlock >>> (d_a, n); } int main() { const int N = 50000; thrust::device_vector<int> a(N); fill(thrust::raw_pointer_cast(&a[0]), N); int sumA = thrust::reduce(a.begin(), a.end(), 0); int sumCheck = 0; for(int i = 0; i < N; i++) sumCheck += i; if(sumA == sumCheck) cout << "ok" << endl; else cout << "fail" << endl; return 0; }
19,725
#include <stdio.h> #include <stdlib.h> #include <stdint.h> #include <math.h> #include <complex.h> #include <cuda_runtime.h> #include <utility> #include <sys/time.h> #define K 3 #define BLCH 8 #define BLCW 32 __constant__ float filter[K*K]; int compute_tiled_naive(float *img, float *f, float * out, int bh, int bw, int imgH, int imgW, int imgN, int nF, int convH, int convW) { //compute tile num int bbw = imgW / bw; int bbh = imgH / bh; for (int i = 0; i < imgN; i++){ int con = i * convW * convH; int imgg = i * imgW * imgH; //compute center tiles for (int j = 0; j < bbh-1; j++){ for (int k = 0; k < bbw-1; k++){ for (int mi = 0; mi < bh; mi++){ int inm = imgW * (j*bh + mi) + k*bw; int ind = convW * (j*bh + mi) + k*bw; for (int mj = 0; mj < bw; mj++){ inm += mj; ind += mj; for (int fi = 0; fi < nF; fi++){ int inf = fi * convW; inm += fi * imgW; for (int fj = 0; fj < nF; fj++){ inf += fj; inm += fj; out[ind + con] += img[inm + imgg] * f[inf]; } } } } } } //compute right most tiles for (int j = 0; j < bbh-1; j++){ for (int mi = 0; mi < bh; mi++){ int inm = imgW * (j*bh + mi) + (bbw-1)*bw; int ind = convW * (j*bh + mi) + (bbw-1)*bw; for (int mj = 0; mj < (bw - nF + 1); mj++){ inm += mj; ind += mj; for (int fi = 0; fi < nF; fi++){ int inf = fi * nF; inm += fi * imgW; for (int fj = 0; fj < nF; fj++){ inf += fj; inm += fj; out[ind + con] += img[inm + imgg] * f[inf]; } } } } } //compute bottom tiles for (int j = 0; j < bbw-1; j++){ for (int mi = 0; mi < (bh -nF + 1); mi++){ int inm = imgW * ((bbh-1)*bh + mi) + (j-1)*bw; int ind = convW * ((bbh-1)*bh + mi) + (j-1)*bw; for (int mj = 0; mj < bw; mj++){ inm += mj; ind += mj; for (int fi = 0; fi < nF; fi++){ int inf = fi * nF; inm += fi * imgW; for (int fj = 0; fj < nF; fj++){ inf += fj; inm += fj; out[ind + con] += img[inm + imgg] * f[inf]; } } } } } //compute the final tile for (int mi = 0; mi < (bh -nF + 1); mi++){ int inm = imgW * ((bbh-1)*bh + mi) + (bbw-1)*bw; int ind = convW * ((bbh-1)*bh + mi) + (bbw-1)*bw; for (int mj = 0; mj < (bw - nF + 1); mj++){ inm += mj; ind += mj; for (int fi = 0; fi < nF; fi++){ int inf = fi * nF; inm += fi * imgW; for (int fj = 0; fj < nF; fj++){ inf += fj; inm += fj; out[ind + con] += img[inm + imgg] * f[inf]; } } } } } return 0; } __global__ void compute_gpu_naive(float *img, float *out, int bh, int bw, int imgH, int imgW, int imgN, int nF, int convH, int convW){ int idX = blockDim.x * blockIdx.x + threadIdx.x; int idY = blockDim.y * blockIdx.y + threadIdx.y; for (int i = 0; i < imgN; i++){ int con = i * convW * convH; int imgg = i * imgW * imgH; if (idX < convH && idY < convW){ int ind = idY * convW + idX + con; int inm = idY * imgW + idX + imgg; for (int fi = 0; fi < nF; fi++){ inm += fi * imgW; int inf = fi*nF; for (int fj = 0; fj < nF; fj++){ out[ind] += img[fj + inm] * filter[inf + fj]; } } } } } __global__ void compute_gpu_tiled(float *img, float *out, int bh, int bw, int imgH, int imgW, int imgN, int nF, int convH, int convW){ int tx = threadIdx.x; int ty = threadIdx.y; int bx = blockIdx.x; int by = blockIdx.y; int nT = blockDim.x; int nB = gridDim.x; int bbw = imgW / bw; int bbh = imgH / bh; for (int i = 0; i < imgN; i++){ int con = i * convW * convH; int imgg = i * imgW * imgH; //compute center tiles if(bx < (bbh-1) && by < (bbw-1)){ if(tx < bh && ty < bw){ int inm = imgW * (bx*bh + tx) + by*bw + ty; int ind = convW * (bx*bh + ty) + by*bw + ty; for (int fi = 0; fi < nF; fi++){ int inf = fi * convW; inm += fi * imgW; for (int fj = 0; fj < nF; fj++){ inf += fj; inm += fj; out[ind + con] += img[inm + imgg] * filter[inf]; } } } __syncthreads(); } //compute right most tiles if(bx < (bbh-1) && by == (bbw-1)){ if(tx < bh && ty < (bw - nF + 1)){ int inm = imgW * (bx*bh + tx) + by*bw + ty; int ind = convW * (bx*bh + ty) + by*bw + ty; for (int fi = 0; fi < nF; fi++){ int inf = fi * convW; inm += fi * imgW; for (int fj = 0; fj < nF; fj++){ inf += fj; inm += fj; out[ind + con] += img[inm + imgg] * filter[inf]; } } } __syncthreads(); } //compute the bottom tiles if(bx == (bbh-1) && by < (bbw-1)){ if(tx < (bh -nF + 1) && ty < bw){ int inm = imgW * (bx*bh + tx) + by*bw + ty; int ind = convW * (bx*bh + ty) + by*bw + ty; for (int fi = 0; fi < nF; fi++){ int inf = fi * convW; inm += fi * imgW; for (int fj = 0; fj < nF; fj++){ inf += fj; inm += fj; out[ind + con] += img[inm + imgg] * filter[inf]; } } } __syncthreads(); } //compute the final tile if(bx == (bbh-1) && by == (bbw-1)){ if(tx < (bh - nF + 1) && ty < (bw - nF + 1)){ int inm = imgW * (bx*bh + tx) + by*bw + ty; int ind = convW * (bx*bh + ty) + by*bw + ty; for (int fi = 0; fi < nF; fi++){ int inf = fi * convW; inm += fi * imgW; for (int fj = 0; fj < nF; fj++){ inf += fj; inm += fj; out[ind + con] += img[inm + imgg] * filter[inf]; } } } __syncthreads(); } } } //remember the threads number to be larger than tile size __global__ void compute_gpu_sm(float *img, float *out, int bh, int bw, int imgH, int imgW, int imgN, int nF, int convH, int convW){ int tx = threadIdx.x; int ty = threadIdx.y; int bx = blockIdx.x; int by = blockIdx.y; int nT = blockDim.x; int nB = gridDim.x; int bbw = imgW / bw; int bbh = imgH / bh; __shared__ float sm[ (BLCH + K - 1) * (BLCW + K - 1) ]; for (int i = 0; i < imgN; i++){ int con = i * convW * convH; int imgg = i * imgW * imgH; //compute center tiles if(bx < (bbh-1) && by < (bbw-1)){ if(tx < (bh+nF-1) && ty < (bw+nF-1)){ int inm = imgW * (bx*bh + tx) + by*bw + ty; //set shared memory sm[tx*bw+ty] = img[inm]; } __syncthreads(); if(tx < bh && ty < bw){ int ind = convW * (bx*bh + ty) + by*bw + ty; for (int fi = 0; fi < nF; fi++){ int inf = fi * convW; int ins = (tx + fi)*bw + ty; for (int fj = 0; fj < nF; fj++){ inf += fj; ins += fj ; out[ind + con] += sm[ins] * filter[inf]; } } } __syncthreads(); } //compute right most tiles if(bx < (bbh-1) && by == (bbw-1)){ if(tx < (bh+nF-1) && ty < bw){ int inm = imgW * (bx*bh + tx) + by*bw + ty; //set shared memory sm[tx*bw+ty] = img[inm]; } __syncthreads(); if(tx < bh && ty < (bw - nF + 1)){ int ind = convW * (bx*bh + ty) + by*bw + ty; for (int fi = 0; fi < nF; fi++){ int inf = fi * convW; int ins = (tx + fi)*bw + ty; for (int fj = 0; fj < nF; fj++){ inf += fj; ins += fj; out[ind + con] += sm[ins] * filter[inf]; } } } __syncthreads(); } //compute the bottom tiles if(bx == (bbh-1) && by < (bbw-1)){ if(tx < bh && ty < (bw+nF-1)){ int inm = imgW * (bx*bh + tx) + by*bw + ty; //set shared memory sm[tx*bw+ty] = img[inm]; } __syncthreads(); if(tx < (bh -nF + 1) && ty < bw){ int ind = convW * (bx*bh + ty) + by*bw + ty; for (int fi = 0; fi < nF; fi++){ int inf = fi * convW; int ins = (tx + fi)*bw + ty; for (int fj = 0; fj < nF; fj++){ inf += fj; ins += fj; out[ind + con] += sm[ins] * filter[inf]; } } } __syncthreads(); } //compute the final tile if(bx == (bbh-1) && by == (bbw-1)){ if(tx < bh && ty < bw){ int inm = imgW * (bx*bh + tx) + by*bw + ty; //set shared memory sm[tx*bw+ty] = img[inm]; } __syncthreads(); if(tx < (bh - nF + 1) && ty < (bw - nF + 1)){ int ind = convW * (bx*bh + ty) + by*bw + ty; for (int fi = 0; fi < nF; fi++){ int inf = fi * convW; int ins = (tx + fi)*bw + ty; for (int fj = 0; fj < nF; fj++){ inf += fj; ins += fj; out[ind + con] += sm[ins] * filter[inf]; } } } __syncthreads(); } } } int main(int argc, char **argv){ //create parameters int imgH = 2048; int imgW = 2048; int imgN = 10; int blcH = BLCH; int blcW = BLCW; int k = K; int s = 1; int nB = (imgH * imgW) / (blcH * blcW); //int nT = (blcW+k) * (blcH+k); int nT = blcW * blcH; int imgDims = imgH * imgW * imgN; int imgSize = imgDims * sizeof(float); srand (time(NULL)); // create host array that can hold pixel intensity values float *h_img = new float[imgDims]; for(int i=0; i<imgDims; i++){ h_img[i] = (float)(rand()%10485)/10485; } // create filter and copy to constant memory int filterDims = k * k; int filterSize = filterDims * sizeof(float); float *f = new float[filterDims]; for(int i=0; i<filterDims; i++){ f[i] = (float)(rand()%10485)/10485; } // create host and device array that holds the convoluted matrix int convH = ( (imgH - k) / s ) + 1; int convW = ( (imgW - k) / s ) + 1; int convDims = convH * convW; int convSize = convDims * sizeof(float); float *h_convolved = new float[convDims]; for(int i=0; i<convDims; i++){ h_convolved[i] = 0.0; } // create device array that can hold pixel intensity values in GPU GM float *d_img; float *d_convolved; cudaMalloc((void **) &d_img, imgSize); cudaMemcpy(d_img, h_img, imgSize, cudaMemcpyHostToDevice); cudaMemcpyToSymbol(filter, f, filterSize); cudaMalloc((void **) &d_convolved, convSize); cudaMemcpy(d_convolved, h_convolved, convSize, cudaMemcpyHostToDevice); struct timeval starttime, endtime; double elapsed = 0.0; for (int i = 0; i<10000; i++){ gettimeofday(&starttime,NULL); // call the kernel compute_gpu_tiled<<<nB, nT>>>(d_img, d_convolved, blcH, blcW, imgH, imgW, imgN, k, convH, convW); gettimeofday(&endtime,NULL); elapsed += ((endtime.tv_sec-starttime.tv_sec)*1000000 + endtime.tv_usec-starttime.tv_usec)/1000000.0; } cudaMemcpy(h_convolved, d_convolved, convSize, cudaMemcpyDeviceToHost); cudaDeviceReset(); printf("Input imgH: %d imgW: %d imgN: %d\n", imgH, imgW, imgN); printf("Tile width: %d height: %d\n", blcW, blcH); printf("Block number: %d, block size: %d \n", nB, nT); printf("time: %f \n", elapsed); delete h_img; delete h_convolved; return 0; }
19,726
// fermi /* * Copyright 2018 Vrije Universiteit Amsterdam, The Netherlands * * Licensed under the Apache License, Version 2.0 (the "License"); * you may not use this file except in compliance with the License. * You may obtain a copy of the License at * * http://www.apache.org/licenses/LICENSE-2.0 * Unless required by applicable law or agreed to in writing, software * distributed under the License is distributed on an "AS IS" BASIS, * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. * See the License for the specific language governing permissions and * limitations under the License. */ extern "C" { __global__ void zeromeanVerticallyKernel(const int h, const int w, float* output, const float* input); } __global__ void zeromeanVerticallyKernel(const int h, const int w, float* output, const float* input) { const int bj = blockIdx.x; const int wtj = threadIdx.y; const int ttj = threadIdx.x; const int nrThreadsW = min(1024, w); const int nrThreadsNrThreadsW = min(32, nrThreadsW); const int tj = wtj * (1 * nrThreadsNrThreadsW) + ttj; if (tj < nrThreadsW) { const int j = bj * (1 * nrThreadsW) + tj; if (j < w) { float sumEven = 0.0; float sumOdd = 0.0; for (int i = 0; i < h - 1; i += 2) { sumEven += input[j + i * (1 * w)]; sumOdd += input[j + (i + 1) * (1 * w)]; } const float meanEven = sumEven / ((h + 1) / 2); const float meanOdd = sumOdd / (h / 2); for (int i = 0; i < h - 1; i += 2) { output[j + i * (1 * w)] = input[j + i * (1 * w)] - meanEven; output[j + (i + 1) * (1 * w)] = input[j + (i + 1) * (1 * w)] - meanOdd; } } } }
19,727
/* * The MIT License (MIT) * This file is part of waifu2x-converter-cpp * * Permission is hereby granted, free of charge, to any person obtaining a copy * of this software and associated documentation files (the "Software"), to deal * in the Software without restriction, including without limitation the rights * to use, copy, modify, merge, publish, distribute, sublicense, and/or sell * copies of the Software, and to permit persons to whom the Software is * furnished to do so, subject to the following conditions: * * The above copyright notice and this permission notice shall be included in all * copies or substantial portions of the Software. * * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE * AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE * SOFTWARE. */ /* -*- mode: c++ -*- */ #define UNROLL9(F) \ F(0); \ F(1); \ F(2); \ F(3); \ F(4); \ F(5); \ F(6); \ F(7); \ F(8); \ #define UNROLL8x3x3(F) \ F(0,0,0); \ F(0,0,1); \ F(0,0,2); \ F(0,1,0); \ F(0,1,1); \ F(0,1,2); \ F(0,2,0); \ F(0,2,1); \ F(0,2,2); \ \ F(1,0,0); \ F(1,0,1); \ F(1,0,2); \ F(1,1,0); \ F(1,1,1); \ F(1,1,2); \ F(1,2,0); \ F(1,2,1); \ F(1,2,2); \ \ F(2,0,0); \ F(2,0,1); \ F(2,0,2); \ F(2,1,0); \ F(2,1,1); \ F(2,1,2); \ F(2,2,0); \ F(2,2,1); \ F(2,2,2); \ \ F(3,0,0); \ F(3,0,1); \ F(3,0,2); \ F(3,1,0); \ F(3,1,1); \ F(3,1,2); \ F(3,2,0); \ F(3,2,1); \ F(3,2,2); \ \ F(4,0,0); \ F(4,0,1); \ F(4,0,2); \ F(4,1,0); \ F(4,1,1); \ F(4,1,2); \ F(4,2,0); \ F(4,2,1); \ F(4,2,2); \ \ F(5,0,0); \ F(5,0,1); \ F(5,0,2); \ F(5,1,0); \ F(5,1,1); \ F(5,1,2); \ F(5,2,0); \ F(5,2,1); \ F(5,2,2); \ \ F(6,0,0); \ F(6,0,1); \ F(6,0,2); \ F(6,1,0); \ F(6,1,1); \ F(6,1,2); \ F(6,2,0); \ F(6,2,1); \ F(6,2,2); \ \ F(7,0,0); \ F(7,0,1); \ F(7,0,2); \ F(7,1,0); \ F(7,1,1); \ F(7,1,2); \ F(7,2,0); \ F(7,2,1); \ F(7,2,2); \ #define UNROLL8(F) \ F(0); \ F(1); \ F(2); \ F(3); \ F(4); \ F(5); \ F(6); \ F(7); \ #define UNROLL8x3(F) \ F(0,0); \ F(0,1); \ F(0,2); \ F(0,3); \ F(0,4); \ F(0,5); \ F(0,6); \ F(0,7); \ \ F(1,0); \ F(1,1); \ F(1,2); \ F(1,3); \ F(1,4); \ F(1,5); \ F(1,6); \ F(1,7); \ \ F(2,0); \ F(2,1); \ F(2,2); \ F(2,3); \ F(2,4); \ F(2,5); \ F(2,6); \ F(2,7); \ #define UNROLL10x3(F) \ F(0,0); \ F(0,1); \ F(0,2); \ F(0,3); \ F(0,4); \ F(0,5); \ F(0,6); \ F(0,7); \ F(0,8); \ F(0,9); \ \ F(1,0); \ F(1,1); \ F(1,2); \ F(1,3); \ F(1,4); \ F(1,5); \ F(1,6); \ F(1,7); \ F(1,8); \ F(1,9); \ \ F(2,0); \ F(2,1); \ F(2,2); \ F(2,3); \ F(2,4); \ F(2,5); \ F(2,6); \ F(2,7); \ F(2,8); \ F(2,9); \ #define BLOCK_SIZE 8 extern __shared__ float shared_buf[]; template <int nInputPlanes> __device__ void filter(const float * __restrict__ packed_input, float * __restrict__ packed_output, int nOutputPlanes, const float * __restrict__ biases, unsigned int hsz, unsigned int wsz, const float * __restrict__ weight) { unsigned int yi = blockIdx.x; size_t in_step = wsz * nInputPlanes; const float *inp = packed_input; inp += yi * in_step; const float *in0p = inp - in_step; if (yi == 0) { in0p = inp; } const float *in1p = inp; const float *in2p = inp + in_step; if (yi == hsz-1) { in2p = in1p; } const float *in01 = in0p; const float *in11 = in1p; const float *in21 = in2p; float *shared_ptr = shared_buf; float *in_block0_base = shared_ptr; shared_ptr += nInputPlanes*(BLOCK_SIZE+2); float *in_block1_base = shared_ptr; shared_ptr += nInputPlanes*(BLOCK_SIZE+2); float *in_block2_base = shared_ptr; shared_ptr += nInputPlanes*(BLOCK_SIZE+2); float *in_block0 = in_block0_base + nInputPlanes; float *in_block1 = in_block1_base + nInputPlanes; float *in_block2 = in_block2_base + nInputPlanes; int lid = threadIdx.x; float bv = biases[lid]; for (int xi0=0; xi0<wsz; xi0+=BLOCK_SIZE) { /*for (unsigned int op=0; op<nOutputPlanes; op++) thread */ { int op = lid; int rem = wsz - xi0; __syncthreads(); if (lid < nInputPlanes/2) { int bi; int lid2 = lid*2; for (bi=0; bi<BLOCK_SIZE; bi++) { int xi = xi0 + bi; if (xi == wsz) { break; } /* load to shared */ *(float2*)&in_block0[bi*nInputPlanes + lid2] = *(float2*)&in01[xi*nInputPlanes + lid2]; *(float2*)&in_block1[bi*nInputPlanes + lid2] = *(float2*)&in11[xi*nInputPlanes + lid2]; *(float2*)&in_block2[bi*nInputPlanes + lid2] = *(float2*)&in21[xi*nInputPlanes + lid2]; } { int xi = xi0 + bi; if (xi == wsz) { *(float2*)&in_block0[bi*(int)nInputPlanes + lid2] = *(float2*)&in01[(xi-1)*(int)nInputPlanes + lid2]; *(float2*)&in_block1[bi*(int)nInputPlanes + lid2] = *(float2*)&in11[(xi-1)*(int)nInputPlanes + lid2]; *(float2*)&in_block2[bi*(int)nInputPlanes + lid2] = *(float2*)&in21[(xi-1)*(int)nInputPlanes + lid2]; } else { *(float2*)&in_block0[bi*(int)nInputPlanes + lid2] = *(float2*)&in01[xi*(int)nInputPlanes + lid2]; *(float2*)&in_block1[bi*(int)nInputPlanes + lid2] = *(float2*)&in11[xi*(int)nInputPlanes + lid2]; *(float2*)&in_block2[bi*(int)nInputPlanes + lid2] = *(float2*)&in21[xi*(int)nInputPlanes + lid2]; } } { int xi = xi0-1; if (xi == -1) { *(float2*)&in_block0[-1*(int)nInputPlanes + (int)lid2] = *(float2*)&in01[lid2]; *(float2*)&in_block1[-1*(int)nInputPlanes + (int)lid2] = *(float2*)&in11[lid2]; *(float2*)&in_block2[-1*(int)nInputPlanes + (int)lid2] = *(float2*)&in21[lid2]; } else { *(float2*)&in_block0[-1*(int)nInputPlanes + (int)lid2] = *(float2*)&in01[xi*(int)nInputPlanes + lid2]; *(float2*)&in_block1[-1*(int)nInputPlanes + (int)lid2] = *(float2*)&in11[xi*(int)nInputPlanes + lid2]; *(float2*)&in_block2[-1*(int)nInputPlanes + (int)lid2] = *(float2*)&in21[xi*(int)nInputPlanes + lid2]; } } } __syncthreads(); if (rem >= BLOCK_SIZE) { #define DECL_PTR(y,x) float *p##y##x = &in_block##y[nInputPlanes * (x-1)]; UNROLL10x3(DECL_PTR); float sum0 = 0; float sum1 = 0; float sum2 = 0; float sum3 = 0; float sum4 = 0; float sum5 = 0; float sum6 = 0; float sum7 = 0; { const float *w0 = weight + lid; for (int ip = 0; ip < nInputPlanes; ip++) { #define LOAD_INPUT2(y,x) float2 i##y##x##_2 = *(float2*)&p##y##x[ip]; UNROLL10x3(LOAD_INPUT2); #define LOAD_COEF(X) float w_##X = w[X * 128]; #define CALC(IDX,Y,I0,I1,I2,I3,I4,I5,I6,I7) \ sum0 += w_##IDX * i##Y##I0; \ sum1 += w_##IDX * i##Y##I1; \ sum2 += w_##IDX * i##Y##I2; \ sum3 += w_##IDX * i##Y##I3; \ sum4 += w_##IDX * i##Y##I4; \ sum5 += w_##IDX * i##Y##I5; \ sum6 += w_##IDX * i##Y##I6; \ sum7 += w_##IDX * i##Y##I7; { #define LOAD_INPUT1X(Y,X) float i##Y##X = i##Y##X##_2.x; UNROLL10x3(LOAD_INPUT1X); const float *w = (w0 + (ip * 128) * 9); UNROLL9(LOAD_COEF); { CALC(0,0,0,1,2,3,4,5,6,7); CALC(1,0,1,2,3,4,5,6,7,8); CALC(2,0,2,3,4,5,6,7,8,9); CALC(3,1,0,1,2,3,4,5,6,7); CALC(4,1,1,2,3,4,5,6,7,8); CALC(5,1,2,3,4,5,6,7,8,9); CALC(6,2,0,1,2,3,4,5,6,7); CALC(7,2,1,2,3,4,5,6,7,8); CALC(8,2,2,3,4,5,6,7,8,9); } } ip++; { #define LOAD_INPUT1Y(Y,X) float i##Y##X = i##Y##X##_2.y; UNROLL10x3(LOAD_INPUT1Y); const float *w = (w0 + (ip * 128) * 9); UNROLL9(LOAD_COEF); { CALC(0,0,0,1,2,3,4,5,6,7); CALC(1,0,1,2,3,4,5,6,7,8); CALC(2,0,2,3,4,5,6,7,8,9); CALC(3,1,0,1,2,3,4,5,6,7); CALC(4,1,1,2,3,4,5,6,7,8); CALC(5,1,2,3,4,5,6,7,8,9); CALC(6,2,0,1,2,3,4,5,6,7); CALC(7,2,1,2,3,4,5,6,7,8); CALC(8,2,2,3,4,5,6,7,8,9); } } } #define RELU(BI) \ { \ float *out = packed_output + (yi*wsz + (xi0+BI))*nOutputPlanes; \ \ { \ int opIndex = lid; \ float v = sum##BI; \ v += bv; \ \ float mtz = max(v, 0.0f); \ float ltz = min(v, 0.0f); \ \ v = ltz * 0.1f + mtz; \ \ out[opIndex] = v; \ } \ } UNROLL8(RELU); #undef DECL_PTR #undef LOAD_COEF #undef CALC #undef LOAD_INPUT2 #undef LOAD_INPUT1X #undef LOAD_INPUT1Y #undef RELU } } else { for (int bi=0; bi<BLOCK_SIZE; bi++) { int xi = xi0+bi; if (xi == wsz) { break; } const float *w0 = weight + lid; float sum = 0; for (int ip=0; ip<nInputPlanes; ip++) { float i00, i01, i02; float i10, i11, i12; float i20, i21, i22; i00 = in_block0[(bi-1)*nInputPlanes+ip]; i10 = in_block1[(bi-1)*nInputPlanes+ip]; i20 = in_block2[(bi-1)*nInputPlanes+ip]; i01 = in_block0[bi*nInputPlanes+ip]; i11 = in_block1[bi*nInputPlanes+ip]; i21 = in_block2[bi*nInputPlanes+ip]; i02 = in_block0[(bi+1)*nInputPlanes+ip]; i12 = in_block1[(bi+1)*nInputPlanes+ip]; i22 = in_block2[(bi+1)*nInputPlanes+ip]; const float *w = w0; sum += w[(9*ip+0) * 128]*i00; sum += w[(9*ip+1) * 128]*i01; sum += w[(9*ip+2) * 128]*i02; sum += w[(9*ip+3) * 128]*i10; sum += w[(9*ip+4) * 128]*i11; sum += w[(9*ip+5) * 128]*i12; sum += w[(9*ip+6) * 128]*i20; sum += w[(9*ip+7) * 128]*i21; sum += w[(9*ip+8) * 128]*i22; } float *out = packed_output + (yi*wsz + xi)*nOutputPlanes; { float v = sum; v += bv; float mtz = max(v, 0.0f); float ltz = min(v, 0.0f); v = ltz * 0.1f + mtz; out[op] = v; } } } } } } extern "C" __global__ void filter_i32(const float * __restrict__ packed_input, float * __restrict__ packed_output, int nOutputPlanes, const float * __restrict__ biases, unsigned int hsz, unsigned int wsz, const float * __restrict__ weight) { filter<32>(packed_input, packed_output, nOutputPlanes, biases, hsz, wsz, weight); } extern "C" __global__ void filter_i64(const float * __restrict__ packed_input, float * __restrict__ packed_output, int nOutputPlanes, const float * __restrict__ biases, unsigned int hsz, unsigned int wsz, const float * __restrict__ weight) { filter<64>(packed_input, packed_output, nOutputPlanes, biases, hsz, wsz, weight); } extern "C" __global__ void filter_i128(const float * __restrict__ packed_input, float * __restrict__ packed_output, int nOutputPlanes, const float * __restrict__ biases, unsigned int hsz, unsigned int wsz, const float * __restrict__ weight) { filter<128>(packed_input, packed_output, nOutputPlanes, biases, hsz, wsz, weight); } #if __CUDA_ARCH__ >= 300 static inline __device__ float warp_sum(float v) { v += __shfl_down_sync(0xFFFFFFFF, v, 1); v += __shfl_down_sync(0xFFFFFFFF, v, 2); v += __shfl_down_sync(0xFFFFFFFF, v, 4); v += __shfl_down_sync(0xFFFFFFFF, v, 8); v += __shfl_down_sync(0xFFFFFFFF, v, 16); return v; } #endif template <int nInputPlanes, int nOutputPlanes> void __device__ filter_weight_blocking(const float * __restrict__ packed_input, float * __restrict__ packed_output, const float * __restrict__ biases, unsigned int hsz, unsigned int wsz, const float * __restrict__ weight, int ib0, int ob0) { #define INPUT_BLOCK_SIZE 32 #define OUTPUT_BLOCK_SIZE 64 // == blockDim.x #define X_BLOCK_SIZE 8 unsigned int yi = blockIdx.x; size_t in_step = wsz * nInputPlanes; const float *inp = packed_input; inp += yi * in_step; const float *in0p = inp - in_step; if (yi == 0) { in0p = inp; } const float *in1p = inp; const float *in2p = inp + in_step; if (yi == hsz-1) { in2p = in1p; } const float *in01 = in0p; const float *in11 = in1p; const float *in21 = in2p; __shared__ float shared_buf_base[INPUT_BLOCK_SIZE * (X_BLOCK_SIZE+2) * 3]; float *in_block0_base = shared_buf_base + INPUT_BLOCK_SIZE * (BLOCK_SIZE+2) * 0; float *in_block1_base = shared_buf_base + INPUT_BLOCK_SIZE * (BLOCK_SIZE+2) * 1; float *in_block2_base = shared_buf_base + INPUT_BLOCK_SIZE * (BLOCK_SIZE+2) * 2; float *in_block0 = in_block0_base + INPUT_BLOCK_SIZE; float *in_block1 = in_block1_base + INPUT_BLOCK_SIZE; float *in_block2 = in_block2_base + INPUT_BLOCK_SIZE; int lid = threadIdx.x; { // ib0 { // ob0 int op = lid + ob0; float bv = biases[op]; for (int xi0=0; xi0<wsz; xi0+=BLOCK_SIZE) { float *out_base = packed_output + (yi*wsz + xi0)*nOutputPlanes + op; float *linp0 = in_block0 + lid; float *linp1 = in_block1 + lid; float *linp2 = in_block2 + lid; __syncthreads(); int rem = wsz - xi0; const float *inb0 = in01 + ib0+lid; const float *inb1 = in11 + ib0+lid; const float *inb2 = in21 + ib0+lid; if (rem > 8 && xi0 != 0) { if (lid < INPUT_BLOCK_SIZE) { linp0[-1*INPUT_BLOCK_SIZE] = linp0[7*INPUT_BLOCK_SIZE]; linp1[-1*INPUT_BLOCK_SIZE] = linp1[7*INPUT_BLOCK_SIZE]; linp2[-1*INPUT_BLOCK_SIZE] = linp2[7*INPUT_BLOCK_SIZE]; linp0[0*INPUT_BLOCK_SIZE] = linp0[8*INPUT_BLOCK_SIZE]; linp1[0*INPUT_BLOCK_SIZE] = linp1[8*INPUT_BLOCK_SIZE]; linp2[0*INPUT_BLOCK_SIZE] = linp2[8*INPUT_BLOCK_SIZE]; } __syncthreads(); if (lid < INPUT_BLOCK_SIZE) { int bi; #pragma unroll for (bi=1; bi<X_BLOCK_SIZE+1; bi++) { int xi = xi0 + bi; /* load to shared */ linp0[bi*INPUT_BLOCK_SIZE] = inb0[xi*nInputPlanes]; linp1[bi*INPUT_BLOCK_SIZE] = inb1[xi*nInputPlanes]; linp2[bi*INPUT_BLOCK_SIZE] = inb2[xi*nInputPlanes]; } } } else { if (lid < INPUT_BLOCK_SIZE) { int bi; for (bi=0; bi<X_BLOCK_SIZE; bi++) { int xi = xi0 + bi; if (xi == wsz) { break; } /* load to shared */ linp0[bi*INPUT_BLOCK_SIZE] = inb0[xi*nInputPlanes]; linp1[bi*INPUT_BLOCK_SIZE] = inb1[xi*nInputPlanes]; linp2[bi*INPUT_BLOCK_SIZE] = inb2[xi*nInputPlanes]; } { int xi = xi0 + bi; if (xi == wsz) { linp0[bi*(int)INPUT_BLOCK_SIZE] = inb0[(xi-1)*(int)nInputPlanes]; linp1[bi*(int)INPUT_BLOCK_SIZE] = inb1[(xi-1)*(int)nInputPlanes]; linp2[bi*(int)INPUT_BLOCK_SIZE] = inb2[(xi-1)*(int)nInputPlanes]; } else { linp0[bi*(int)INPUT_BLOCK_SIZE] = inb0[xi*(int)nInputPlanes]; linp1[bi*(int)INPUT_BLOCK_SIZE] = inb1[xi*(int)nInputPlanes]; linp2[bi*(int)INPUT_BLOCK_SIZE] = inb2[xi*(int)nInputPlanes]; } } { int xi = xi0-1; if (xi == -1) { linp0[-1*(int)INPUT_BLOCK_SIZE] = inb0[0]; linp1[-1*(int)INPUT_BLOCK_SIZE] = inb1[0]; linp2[-1*(int)INPUT_BLOCK_SIZE] = inb2[0]; } else { linp0[-1*(int)INPUT_BLOCK_SIZE] = inb0[xi*(int)nInputPlanes]; linp1[-1*(int)INPUT_BLOCK_SIZE] = inb1[xi*(int)nInputPlanes]; linp2[-1*(int)INPUT_BLOCK_SIZE] = inb2[xi*(int)nInputPlanes]; } } } } __syncthreads(); const float *w0 = weight + op; if (rem >= BLOCK_SIZE) { #define DECL_PTR(y,x) float *p##y##x = &in_block##y[INPUT_BLOCK_SIZE * (x-1)]; UNROLL10x3(DECL_PTR); float sum0 = 0; float sum1 = 0; float sum2 = 0; float sum3 = 0; float sum4 = 0; float sum5 = 0; float sum6 = 0; float sum7 = 0; for (int ip1 = 0; ip1 < INPUT_BLOCK_SIZE; ip1+=2) { int ip = ip1 + ib0; #define LOAD_INPUT2(y,x) float2 i##y##x##_2 = *(float2*)&p##y##x[ip1]; UNROLL10x3(LOAD_INPUT2); #define LOAD_COEF(X) float w_##X = w[X * 128]; #define CALC(SYM,IDX,Y,I0,I1,I2,I3,I4,I5,I6,I7) \ sum0 += w_##IDX * i##Y##I0##_2.SYM; \ sum1 += w_##IDX * i##Y##I1##_2.SYM; \ sum2 += w_##IDX * i##Y##I2##_2.SYM; \ sum3 += w_##IDX * i##Y##I3##_2.SYM; \ sum4 += w_##IDX * i##Y##I4##_2.SYM; \ sum5 += w_##IDX * i##Y##I5##_2.SYM; \ sum6 += w_##IDX * i##Y##I6##_2.SYM; \ sum7 += w_##IDX * i##Y##I7##_2.SYM; { const float *w = (w0 + (ip * 128) * 9); UNROLL9(LOAD_COEF); { CALC(x, 0,0,0,1,2,3,4,5,6,7); CALC(x, 1,0,1,2,3,4,5,6,7,8); CALC(x, 2,0,2,3,4,5,6,7,8,9); CALC(x, 3,1,0,1,2,3,4,5,6,7); CALC(x, 4,1,1,2,3,4,5,6,7,8); CALC(x, 5,1,2,3,4,5,6,7,8,9); CALC(x, 6,2,0,1,2,3,4,5,6,7); CALC(x, 7,2,1,2,3,4,5,6,7,8); CALC(x, 8,2,2,3,4,5,6,7,8,9); } } ip++; { const float *w = (w0 + (ip * 128) * 9); UNROLL9(LOAD_COEF); { CALC(y, 0,0,0,1,2,3,4,5,6,7); CALC(y, 1,0,1,2,3,4,5,6,7,8); CALC(y, 2,0,2,3,4,5,6,7,8,9); CALC(y, 3,1,0,1,2,3,4,5,6,7); CALC(y, 4,1,1,2,3,4,5,6,7,8); CALC(y, 5,1,2,3,4,5,6,7,8,9); CALC(y, 6,2,0,1,2,3,4,5,6,7); CALC(y, 7,2,1,2,3,4,5,6,7,8); CALC(y, 8,2,2,3,4,5,6,7,8,9); } } } #define RELU(BI) \ { \ \ { \ float v = sum##BI + out_base[BI*nOutputPlanes]; \ v += bv; \ \ float mtz = max(v, 0.0f); \ float ltz = min(v, 0.0f); \ \ v = ltz * 0.1f + mtz; \ \ out_base[BI*nOutputPlanes] = v; \ } \ } if ((ib0+INPUT_BLOCK_SIZE) == nInputPlanes) { UNROLL8(RELU); } else if (ib0 == 0) { out_base[nOutputPlanes*0] = sum0; out_base[nOutputPlanes*1] = sum1; out_base[nOutputPlanes*2] = sum2; out_base[nOutputPlanes*3] = sum3; out_base[nOutputPlanes*4] = sum4; out_base[nOutputPlanes*5] = sum5; out_base[nOutputPlanes*6] = sum6; out_base[nOutputPlanes*7] = sum7; } else { out_base[nOutputPlanes*0] += sum0; out_base[nOutputPlanes*1] += sum1; out_base[nOutputPlanes*2] += sum2; out_base[nOutputPlanes*3] += sum3; out_base[nOutputPlanes*4] += sum4; out_base[nOutputPlanes*5] += sum5; out_base[nOutputPlanes*6] += sum6; out_base[nOutputPlanes*7] += sum7; } } else { for (int bi=0; bi<X_BLOCK_SIZE; bi++) { int xi = xi0+bi; if (xi == wsz) { break; } float sum = 0; for (int ip1=0; ip1<INPUT_BLOCK_SIZE; ip1++) { int ip = ib0 + ip1; float i00, i01, i02; float i10, i11, i12; float i20, i21, i22; i00 = in_block0[(bi-1)*INPUT_BLOCK_SIZE+ip1]; i10 = in_block1[(bi-1)*INPUT_BLOCK_SIZE+ip1]; i20 = in_block2[(bi-1)*INPUT_BLOCK_SIZE+ip1]; i01 = in_block0[bi*INPUT_BLOCK_SIZE+ip1]; i11 = in_block1[bi*INPUT_BLOCK_SIZE+ip1]; i21 = in_block2[bi*INPUT_BLOCK_SIZE+ip1]; i02 = in_block0[(bi+1)*INPUT_BLOCK_SIZE+ip1]; i12 = in_block1[(bi+1)*INPUT_BLOCK_SIZE+ip1]; i22 = in_block2[(bi+1)*INPUT_BLOCK_SIZE+ip1]; sum += w0[(9*ip+0) * 128]*i00; sum += w0[(9*ip+1) * 128]*i01; sum += w0[(9*ip+2) * 128]*i02; sum += w0[(9*ip+3) * 128]*i10; sum += w0[(9*ip+4) * 128]*i11; sum += w0[(9*ip+5) * 128]*i12; sum += w0[(9*ip+6) * 128]*i20; sum += w0[(9*ip+7) * 128]*i21; sum += w0[(9*ip+8) * 128]*i22; } float *out = packed_output + (yi*wsz + xi)*nOutputPlanes; if ((ib0+INPUT_BLOCK_SIZE) == nInputPlanes) { /* last */ float v = sum + out[op]; v += bv; float mtz = max(v, 0.0f); float ltz = min(v, 0.0f); v = ltz * 0.1f + mtz; out[op] = v; } else if (ib0 == 0) { out[op] = sum; } else { out[op] += sum; } } } } } } } extern "C" __global__ void filter_i128_o128(const float * __restrict__ packed_input, float * __restrict__ packed_output, const float * __restrict__ biases, unsigned int hsz, unsigned int wsz, const float * __restrict__ weight, int ib0, int ob0) { filter_weight_blocking<128,128>(packed_input, packed_output, biases, hsz, wsz, weight, ib0, ob0); } extern "C" __global__ void filter_i64_o128(const float * __restrict__ packed_input, float * __restrict__ packed_output, const float * __restrict__ biases, unsigned int hsz, unsigned int wsz, const float * __restrict__ weight, int ib0, int ob0) { filter_weight_blocking<64,128>(packed_input, packed_output, biases, hsz, wsz, weight, ib0, ob0); } extern "C" __global__ void filter_i64_o64(const float * __restrict__ packed_input, float * __restrict__ packed_output, const float * __restrict__ biases, unsigned int hsz, unsigned int wsz, const float * __restrict__ weight, int ib0, int ob0) { filter_weight_blocking<64,64>(packed_input, packed_output, biases, hsz, wsz, weight, ib0, ob0); } extern "C" __global__ void filter_i128_o1(const float * __restrict__ packed_input, float * __restrict__ packed_output, float * __restrict__ biases, unsigned int hsz, unsigned int wsz, float * __restrict__ weight) { int nInputPlanes = 128; int nOutputPlanes = 1; { unsigned int yi = blockIdx.x; size_t in_step = wsz * nInputPlanes; const float *inp = packed_input; inp += yi * in_step; const float *in0p = inp - in_step; if (yi == 0) { in0p = inp; } const float *in1p = inp; const float *in2p = inp + in_step; if (yi == hsz-1) { in2p = in1p; } const float *in01 = in0p; const float *in11 = in1p; const float *in21 = in2p; unsigned int lid = threadIdx.x; float bv0 = biases[0]; /* 128 item */ /* x : (1width/group) */ /* y : (2height/group) */ /* iplane : 1plane / 1item * 128plane */ __shared__ float shared_buf[128 * 10]; float lin00; float lin01; float lin02; float lin10; float lin11; float lin12; float lin20; float lin21; float lin22; float *sum_buffer = shared_buf + 128*9; #define OUT1_LOAD_WEIGHT(I,Y,X) float w##I##Y##X = weight[(I*16 + lid)*9 + Y*3 + X]; float w00 = weight[lid*9 + 0]; float w01 = weight[lid*9 + 1]; float w02 = weight[lid*9 + 2]; float w10 = weight[lid*9 + 3]; float w11 = weight[lid*9 + 4]; float w12 = weight[lid*9 + 5]; float w20 = weight[lid*9 + 6]; float w21 = weight[lid*9 + 7]; float w22 = weight[lid*9 + 8]; const float *pin01 = in01 + lid; const float *pin02 = in01 + nInputPlanes + lid; const float *pin11 = in11 + lid; const float *pin12 = in11 + nInputPlanes + lid; const float *pin21 = in21 + lid; const float *pin22 = in21 + nInputPlanes + lid; lin01 = pin01[0]; lin02 = pin01[0]; lin11 = pin11[0]; lin12 = pin11[0]; lin21 = pin21[0]; lin22 = pin21[0]; #define OUT1_BODY(LEDGE,REDGE,SUM_RELU) \ { \ float sum = 0; \ { \ lin00 = lin01; \ lin01 = lin02; \ \ lin10 = lin11; \ lin11 = lin12; \ \ lin20 = lin21; \ lin21 = lin22; \ \ if (REDGE) { \ lin02 = lin01; \ lin12 = lin11; \ lin22 = lin21; \ } else { \ lin02 = pin02[xi*128]; \ lin12 = pin12[xi*128]; \ lin22 = pin22[xi*128]; \ } \ \ sum += w00 * lin00; \ sum += w10 * lin10; \ sum += w20 * lin20; \ \ sum += w01 * lin01; \ sum += w11 * lin11; \ sum += w21 * lin21; \ \ sum += w02 * lin02; \ sum += w12 * lin12; \ sum += w22 * lin22; \ \ } \ __syncthreads(); \ sum_buffer[lid] = sum; \ __syncthreads(); \ if (lid < 64) { \ float v2 = sum_buffer[lid+64]; \ sum_buffer[lid] += v2; \ } \ __syncthreads(); \ SUM_RELU(0); \ } #if __CUDA_ARCH__ >= 300 #define SUM_RELU(OI) \ if (lid < 32) { \ float v0 = sum_buffer[lid] + sum_buffer[lid+32]; \ float sum = warp_sum(v0); \ \ if (lid == 0) { \ float v = sum; \ float *out = packed_output + (yi*wsz + xi)*nOutputPlanes; \ v += bv##OI; \ float mtz = max(v, 0.0f); \ float ltz = min(v, 0.0f); \ v = ltz * 0.1f + mtz; \ out[OI] = v; \ } \ } \ #else #define SUM_RELU(OI) \ if (lid < 32) { \ sum_buffer[lid] += sum_buffer[lid+32]; \ } \ __syncthreads(); \ if (lid < 16) { \ sum_buffer[lid] += sum_buffer[lid+16]; \ } \ __syncthreads(); \ if (lid < 8) { \ sum_buffer[lid] += sum_buffer[lid+8]; \ } \ __syncthreads(); \ if (lid < 4) { \ sum_buffer[lid] += sum_buffer[lid+4]; \ } \ __syncthreads(); \ if (lid < 2) { \ sum_buffer[lid] += sum_buffer[lid+2]; \ } \ __syncthreads(); \ if (lid == 0) { \ float sum = sum_buffer[0] + sum_buffer[1]; \ float v = sum; \ float *out = packed_output + (yi*wsz + xi)*nOutputPlanes; \ v += bv##OI; \ float mtz = max(v, 0.0f); \ float ltz = min(v, 0.0f); \ v = ltz * 0.1f + mtz; \ out[OI] = v; \ } #endif for (int xi=0; xi<wsz-1; xi++) { OUT1_BODY(0,0,SUM_RELU); } { int xi = wsz-1; OUT1_BODY(0,1,SUM_RELU); } } } extern "C" __global__ void filter_i1_o32(const float * __restrict__ packed_input, float * __restrict__ packed_output, float * __restrict__ biases, unsigned int hsz, unsigned int wsz, float * __restrict__ weight) { //int nInputPlanes = 1; int nOutputPlanes = 32; unsigned int yi = blockIdx.x; unsigned int lid = threadIdx.x; size_t in_step = wsz; const float *inp = packed_input; inp += in_step * yi; const float *in0p = inp - in_step; if (yi == 0) { in0p = inp; } const float *in1p = inp; const float *in2p = inp + in_step; if (yi == hsz-1) { in2p = in1p; } const float *in01 = in0p; const float *in11 = in1p; const float *in21 = in2p; __shared__ float in_block0_base[256+2]; __shared__ float in_block1_base[256+2]; __shared__ float in_block2_base[256+2]; float *in_block0 = in_block0_base + 1; float *in_block1 = in_block1_base + 1; float *in_block2 = in_block2_base + 1; /* 256 item / group */ /* x : (64width/group) */ /* 32 oplane : (8weight/item * 4item)*/ unsigned int xoff = lid / 4U; unsigned int ooff = (lid % 4U) * 8; #define IN1_LOAD_COEF(O,Y,X) \ float w##O##Y##X = weight[9 * (O + ooff) + (Y*3) + X]; UNROLL8x3x3(IN1_LOAD_COEF); for (int xi0=0; xi0<wsz; xi0+=256) { /* load */ __syncthreads(); { int xi = xi0 + lid; if (xi < wsz) { in_block0[lid] = in01[xi0 + lid]; in_block1[lid] = in11[xi0 + lid]; in_block2[lid] = in21[xi0 + lid]; } if (lid == 0) { if (xi == 0) { in_block0[-1] = in01[0]; in_block1[-1] = in11[0]; in_block2[-1] = in21[0]; } else { in_block0[-1] = in01[xi-1]; in_block1[-1] = in11[xi-1]; in_block2[-1] = in21[xi-1]; } } if (xi == wsz-1) { in_block0[lid+1] = in01[xi]; in_block1[lid+1] = in11[xi]; in_block2[lid+1] = in21[xi]; } if ((lid == 255) && (xi < wsz-1)) { in_block0[256] = in01[xi+1]; in_block1[256] = in11[xi+1]; in_block2[256] = in21[xi+1]; } } __syncthreads(); for (int xi1_base=0; xi1_base<4; xi1_base++) { { int xi1 = xi1_base*64 + xoff; int xi = xi0 + xi1; if (xi < wsz) { #define IN1_DECLSUM(O) float sum##O = 0; #define IN1_CALC(O,Y,X) sum##O += in_block##Y[xi1+X-1] * w##O##Y##X; #define IN1_RELU(O) { \ float v = sum##O; \ int opIndex = ooff + O; \ float bv = biases[opIndex]; \ v += bv; \ float mtz = max(v, 0.0f); \ float ltz = min(v, 0.0f); \ v = ltz * 0.1f + mtz; \ out[opIndex] = v; \ } UNROLL8(IN1_DECLSUM); UNROLL8x3x3(IN1_CALC); float *out = packed_output + (yi*wsz + xi) * nOutputPlanes; UNROLL8(IN1_RELU); } } } } } /* blockDim.x == 192 */ extern "C" __global__ void filter_i3_o32(const float * __restrict__ packed_input, float * __restrict__ packed_output, float * __restrict__ biases, unsigned int hsz, unsigned int wsz, float * __restrict__ weight) { int nInputPlanes = 3; int nOutputPlanes = 32; unsigned int yi = blockIdx.x; unsigned int lid = threadIdx.x; size_t in_step = wsz * nInputPlanes; const float *inp = packed_input; inp += in_step * yi; const float *in0p = inp - in_step; if (yi == 0) { in0p = inp; } const float *in1p = inp; const float *in2p = inp + in_step; if (yi == hsz-1) { in2p = in1p; } const float *in01 = in0p; const float *in11 = in1p; const float *in21 = in2p; __shared__ float in_block0_base[(64+2)*3]; __shared__ float in_block1_base[(64+2)*3]; __shared__ float in_block2_base[(64+2)*3]; __shared__ float sum_buffer[192]; float *in_block0 = in_block0_base + 3; float *in_block1 = in_block1_base + 3; float *in_block2 = in_block2_base + 3; /* 192 item / group */ /* load 192 item */ /* 3 iplane : */ /* x : (64width/group) */ /* 32 oplane : (8weight/item * 4item)*/ unsigned int ioff = lid / 32U; unsigned int ooff = lid % 32U; #define I3_O32_LOAD_COEF(I) \ float w##I = weight[9*nOutputPlanes*ioff+ooff+I*nOutputPlanes]; UNROLL9(I3_O32_LOAD_COEF); for (int xi0=0; xi0<wsz; xi0+=64) { /* load */ int nelem = min(wsz - xi0, 64); int nload = nelem * 3; if (lid < nload) { int xi = xi0*3 + lid; in_block0[lid] = in01[xi]; in_block1[lid] = in11[xi]; in_block2[lid] = in21[xi]; if (lid < 3) { if (xi <= 2) { /* left edge */ in_block0[-3+(int)lid] = in01[lid]; in_block1[-3+(int)lid] = in11[lid]; in_block2[-3+(int)lid] = in21[lid]; } else { /* 0, 1, 2 */ in_block0[-3+(int)lid] = in01[-3+(int)xi]; in_block1[-3+(int)lid] = in11[-3+(int)xi]; in_block2[-3+(int)lid] = in21[-3+(int)xi]; } } if (xi >= wsz*3-3) { /* right edge */ in_block0[lid+3] = in01[xi]; in_block1[lid+3] = in11[xi]; in_block2[lid+3] = in21[xi]; } else if (lid >= 189) { /* 189, 190, 191 */ in_block0[lid+3] = in01[xi+3]; in_block1[lid+3] = in11[xi+3]; in_block2[lid+3] = in21[xi+3]; } } __syncthreads(); for (int xi1=0; xi1<nelem; xi1++) { int xi = xi0 + xi1; if (lid < 96) { // 3input x 32output float sum = 0; sum += w0 * in_block0[(xi1 - 1)*3+(int)ioff]; sum += w1 * in_block0[(xi1 )*3+(int)ioff]; sum += w2 * in_block0[(xi1 + 1)*3+(int)ioff]; sum += w3 * in_block1[(xi1 - 1)*3+(int)ioff]; sum += w4 * in_block1[(xi1 )*3+(int)ioff]; sum += w5 * in_block1[(xi1 + 1)*3+(int)ioff]; sum += w6 * in_block2[(xi1 - 1)*3+(int)ioff]; sum += w7 * in_block2[(xi1 )*3+(int)ioff]; sum += w8 * in_block2[(xi1 + 1)*3+(int)ioff]; sum_buffer[lid] = sum; } __syncthreads(); if (lid < 32) { int oi = lid; float v = 0; float *out = packed_output + (yi*wsz + xi) * nOutputPlanes; /* 96 to 32 reduction */ v += sum_buffer[32 * 0 + lid]; v += sum_buffer[32 * 1 + lid]; v += sum_buffer[32 * 2 + lid]; float bv = biases[oi]; v += bv; float mtz = max(v, 0.0f); float ltz = min(v, 0.0f); v = ltz * 0.1f + mtz; out[oi] = v; } __syncthreads(); } } } /* blockDim.x == 128 */ extern "C" __global__ void filter_i128_o3(const float * __restrict__ packed_input, float * __restrict__ packed_output, float * __restrict__ biases, unsigned int hsz, unsigned int wsz, float * __restrict__ weight) { int nInputPlanes = 128; int nOutputPlanes = 3; unsigned int yi = blockIdx.x; unsigned int lid = threadIdx.x; size_t in_step = wsz * nInputPlanes; const float *inp = packed_input; inp += in_step * yi; const float *in0p = inp - in_step; if (yi == 0) { in0p = inp; } const float *in1p = inp; const float *in2p = inp + in_step; if (yi == hsz-1) { in2p = in1p; } const float *in01 = in0p; const float *in11 = in1p; const float *in21 = in2p; float lin00, lin01, lin02; float lin10, lin11, lin12; float lin20, lin21, lin22; __shared__ float sum_buffer[128]; /* 128 item / group */ /* load 128 item (load 3elem/item) */ /* 128 iplane * 1 input * 3 output (27coeff) */ int ioff = lid; float bv0 = biases[0]; float bv1 = biases[1]; float bv2 = biases[2]; #define I128_O3_LOAD_COEF(I) \ float w0##I = weight[9*0*nInputPlanes + I*nInputPlanes + ioff]; \ float w1##I = weight[9*1*nInputPlanes + I*nInputPlanes + ioff]; \ float w2##I = weight[9*2*nInputPlanes + I*nInputPlanes + ioff]; UNROLL9(I128_O3_LOAD_COEF); lin01 = lin02 = in01[lid]; lin11 = lin12 = in11[lid]; lin21 = lin22 = in21[lid]; int addroff = 0; char *p0 = (char*)(in01 + lid + nInputPlanes); char *p1 = (char*)(in11 + lid + nInputPlanes); char *p2 = (char*)(in21 + lid + nInputPlanes); for (int xi=0; xi<wsz; xi++) { lin00 = lin01; lin01 = lin02; lin10 = lin11; lin11 = lin12; lin20 = lin21; lin21 = lin22; if (xi == wsz-1) { /* nop */ } else { lin02 = *(float *)(p0 + addroff); lin12 = *(float *)(p1 + addroff); lin22 = *(float *)(p2 + addroff); } addroff += nInputPlanes * sizeof(float); #define I128_O3(OI) \ { \ float sum = 0; \ sum += w##OI##0 * lin00; \ sum += w##OI##1 * lin01; \ sum += w##OI##2 * lin02; \ \ sum += w##OI##3 * lin10; \ sum += w##OI##4 * lin11; \ sum += w##OI##5 * lin12; \ \ sum += w##OI##6 * lin20; \ sum += w##OI##7 * lin21; \ sum += w##OI##8 * lin22; \ \ __syncthreads(); \ sum_buffer[lid] = sum; \ \ /* 128 to 1 */ \ __syncthreads(); \ if (lid < 64) { \ sum_buffer[lid] += sum_buffer[lid + 64]; \ } \ __syncthreads(); \ \ SUM_RELU(OI); \ } I128_O3(0); I128_O3(1); I128_O3(2); } }
19,728
/* * Copyright 1993-2015 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 <cstdio> #include <vector> using namespace std; const char *sSampleName = "P2P (Peer-to-Peer) GPU Bandwidth Latency Test"; //Macro for checking cuda errors following a cuda launch or api call #define cudaCheckError() { \ cudaError_t e=cudaGetLastError(); \ if(e!=cudaSuccess) { \ printf("Cuda failure %s:%d: '%s'\n",__FILE__,__LINE__,cudaGetErrorString(e)); \ exit(EXIT_FAILURE); \ } \ } void checkP2Paccess(int numGPUs) { for (int i=0; i<numGPUs; i++) { cudaSetDevice(i); for (int j=0; j<numGPUs; j++) { int access; if (i!=j) { cudaDeviceCanAccessPeer(&access,i,j); printf("Device=%d %s Access Peer Device=%d\n", i, access ? "CAN" : "CANNOT", j); } } } printf("\n***NOTE: In case a device doesn't have P2P access to other one, it falls back to normal memcopy procedure.\nSo you can see lesser Bandwidth (GB/s) in those cases.\n\n"); } void enableP2P(int numGPUs) { for (int i=0; i<numGPUs; i++) { cudaSetDevice(i); for (int j=0; j<numGPUs; j++) { int access; cudaDeviceCanAccessPeer(&access,i,j); if (access) { cudaDeviceEnablePeerAccess(j,0); cudaCheckError(); } } } } void disableP2P(int numGPUs) { for (int i=0; i<numGPUs; i++) { cudaSetDevice(i); for (int j=0; j<numGPUs; j++) { int access; cudaDeviceCanAccessPeer(&access, i, j); if (access) { cudaDeviceDisablePeerAccess(j); cudaGetLastError(); } } } } void outputBandwidthMatrix(int numGPUs) { int numElems=10000000; int repeat=5; vector<int *> buffers(numGPUs); vector<cudaEvent_t> start(numGPUs); vector<cudaEvent_t> stop(numGPUs); for (int d=0; d<numGPUs; d++) { cudaSetDevice(d); cudaMalloc(&buffers[d],numElems*sizeof(int)); cudaCheckError(); cudaEventCreate(&start[d]); cudaCheckError(); cudaEventCreate(&stop[d]); cudaCheckError(); } vector<double> bandwidthMatrix(numGPUs*numGPUs); for (int i=0; i<numGPUs; i++) { cudaSetDevice(i); for (int j=0; j<numGPUs; j++) { cudaDeviceSynchronize(); cudaCheckError(); cudaEventRecord(start[i]); for (int r=0; r<repeat; r++) { cudaMemcpyPeerAsync(buffers[i],i,buffers[j],j,sizeof(int)*numElems); } cudaEventRecord(stop[i]); cudaDeviceSynchronize(); cudaCheckError(); float time_ms; cudaEventElapsedTime(&time_ms,start[i],stop[i]); double time_s=time_ms/1e3; double gb=numElems*sizeof(int)*repeat/(double)1e9; bandwidthMatrix[i*numGPUs+j]=gb/time_s; } } printf(" D\\D"); for (int j=0; j<numGPUs; j++) { printf("%6d ", j); } printf("\n"); for (int i=0; i<numGPUs; i++) { printf("%6d ",i); for (int j=0; j<numGPUs; j++) { printf("%6.02f ", bandwidthMatrix[i*numGPUs+j]); } printf("\n"); } for (int d=0; d<numGPUs; d++) { cudaSetDevice(d); cudaFree(buffers[d]); cudaCheckError(); cudaEventDestroy(start[d]); cudaCheckError(); cudaEventDestroy(stop[d]); cudaCheckError(); } } void outputBidirectionalBandwidthMatrix(int numGPUs) { int numElems=10000000; int repeat=5; vector<int *> buffers(numGPUs); vector<cudaEvent_t> start(numGPUs); vector<cudaEvent_t> stop(numGPUs); vector<cudaStream_t> stream0(numGPUs); vector<cudaStream_t> stream1(numGPUs); for (int d=0; d<numGPUs; d++) { cudaSetDevice(d); cudaMalloc(&buffers[d],numElems*sizeof(int)); cudaCheckError(); cudaEventCreate(&start[d]); cudaCheckError(); cudaEventCreate(&stop[d]); cudaCheckError(); cudaStreamCreate(&stream0[d]); cudaCheckError(); cudaStreamCreate(&stream1[d]); cudaCheckError(); } vector<double> bandwidthMatrix(numGPUs*numGPUs); for (int i=0; i<numGPUs; i++) { cudaSetDevice(i); for (int j=0; j<numGPUs; j++) { cudaDeviceSynchronize(); cudaCheckError(); cudaEventRecord(start[i]); for (int r=0; r<repeat; r++) { cudaMemcpyPeerAsync(buffers[i],i,buffers[j],j,sizeof(int)*numElems,stream0[i]); cudaMemcpyPeerAsync(buffers[j],j,buffers[i],i,sizeof(int)*numElems,stream1[i]); } cudaEventRecord(stop[i]); cudaDeviceSynchronize(); cudaCheckError(); float time_ms; cudaEventElapsedTime(&time_ms,start[i],stop[i]); double time_s=time_ms/1e3; double gb=2.0*numElems*sizeof(int)*repeat/(double)1e9; bandwidthMatrix[i*numGPUs+j]=gb/time_s; } } printf(" D\\D"); for (int j=0; j<numGPUs; j++) { printf("%6d ", j); } printf("\n"); for (int i=0; i<numGPUs; i++) { printf("%6d ",i); for (int j=0; j<numGPUs; j++) { printf("%6.02f ", bandwidthMatrix[i*numGPUs+j]); } printf("\n"); } for (int d=0; d<numGPUs; d++) { cudaSetDevice(d); cudaFree(buffers[d]); cudaCheckError(); cudaEventDestroy(start[d]); cudaCheckError(); cudaEventDestroy(stop[d]); cudaCheckError(); cudaStreamDestroy(stream0[d]); cudaCheckError(); cudaStreamDestroy(stream1[d]); cudaCheckError(); } } void outputLatencyMatrix(int numGPUs) { int repeat=10000; vector<int *> buffers(numGPUs); vector<cudaEvent_t> start(numGPUs); vector<cudaEvent_t> stop(numGPUs); for (int d=0; d<numGPUs; d++) { cudaSetDevice(d); cudaMalloc(&buffers[d],1); cudaCheckError(); cudaEventCreate(&start[d]); cudaCheckError(); cudaEventCreate(&stop[d]); cudaCheckError(); } vector<double> latencyMatrix(numGPUs*numGPUs); for (int i=0; i<numGPUs; i++) { cudaSetDevice(i); for (int j=0; j<numGPUs; j++) { cudaDeviceSynchronize(); cudaCheckError(); cudaEventRecord(start[i]); for (int r=0; r<repeat; r++) { cudaMemcpyPeerAsync(buffers[i],i,buffers[j],j,1); } cudaEventRecord(stop[i]); cudaDeviceSynchronize(); cudaCheckError(); float time_ms; cudaEventElapsedTime(&time_ms,start[i],stop[i]); latencyMatrix[i*numGPUs+j]=time_ms*1e3/repeat; } } printf(" D\\D"); for (int j=0; j<numGPUs; j++) { printf("%6d ", j); } printf("\n"); for (int i=0; i<numGPUs; i++) { printf("%6d ",i); for (int j=0; j<numGPUs; j++) { printf("%6.02f ", latencyMatrix[i*numGPUs+j]); } printf("\n"); } for (int d=0; d<numGPUs; d++) { cudaSetDevice(d); cudaFree(buffers[d]); cudaCheckError(); cudaEventDestroy(start[d]); cudaCheckError(); cudaEventDestroy(stop[d]); cudaCheckError(); } } int main(int argc, char **argv) { int numGPUs; cudaGetDeviceCount(&numGPUs); printf("[%s]\n", sSampleName); //output devices for (int i=0; i<numGPUs; i++) { cudaDeviceProp prop; cudaGetDeviceProperties(&prop,i); printf("Device: %d, %s, pciBusID: %x, pciDeviceID: %x, pciDomainID:%x\n",i,prop.name, prop.pciBusID, prop.pciDeviceID, prop.pciDomainID); } checkP2Paccess(numGPUs); //compute cliques vector<vector<int> > cliques; vector<bool> added(numGPUs,false); for (int i=0; i<numGPUs; i++) { if (added[i]==true) continue; //already processed //create new clique with i vector<int> clique; added[i]=true; clique.push_back(i); for (int j=i+1; j<numGPUs; j++) { int access; cudaDeviceCanAccessPeer(&access,i,j); if (access) { clique.push_back(j); added[j]=true; } } cliques.push_back(clique); } printf("P2P Cliques: \n"); for (int c=0; c<(int)cliques.size(); c++) { printf("["); for (int j=0; j<(int)cliques[c].size()-1; j++) { printf("%d ",cliques[c][j]); } printf("%d]\n",cliques[c][cliques[c].size()-1]); } printf("Unidirectional P2P=Disabled Bandwidth Matrix (GB/s)\n"); outputBandwidthMatrix(numGPUs); enableP2P(numGPUs); printf("Unidirectional P2P=Enabled Bandwidth Matrix (GB/s)\n"); outputBandwidthMatrix(numGPUs); disableP2P(numGPUs); printf("Bidirectional P2P=Disabled Bandwidth Matrix (GB/s)\n"); outputBidirectionalBandwidthMatrix(numGPUs); enableP2P(numGPUs); printf("Bidirectional P2P=Enabled Bandwidth Matrix (GB/s)\n"); outputBidirectionalBandwidthMatrix(numGPUs); disableP2P(numGPUs); printf("P2P=Disabled Latency Matrix (us)\n"); outputLatencyMatrix(numGPUs); enableP2P(numGPUs); printf("P2P=Enabled Latency Matrix (us)\n"); outputLatencyMatrix(numGPUs); // cudaDeviceReset causes the driver to clean up all state. While // not mandatory in normal operation, it is good practice. It is also // needed to ensure correct operation when the application is being // profiled. Calling cudaDeviceReset causes all profile data to be // flushed before the application exits cudaDeviceReset(); printf("\nNOTE: The CUDA Samples are not meant for performance measurements. Results may vary when GPU Boost is enabled.\n"); exit(EXIT_SUCCESS); }
19,729
#define LOG_2_PI 1.83787706640935f #define LOG_PI 1.144729885849400f __device__ int d_next_multiple(int k, int mult) { if (k % mult) return k + (mult - k % mult); else return k; } __device__ void copy_chunks(float* in_buf, float* out_buf, unsigned int tid, unsigned int total) { for (unsigned int chunk = 0; chunk + tid < total; chunk += blockDim.x) { out_buf[chunk + tid] = in_buf[chunk + tid]; } } __device__ void copy_chunks_strided(float* in_buf, float* out_buf, unsigned int tid, unsigned int ncols, unsigned int nrows, unsigned int stride) { unsigned int outind = 0; unsigned int total = ncols*nrows; for (unsigned int chunk = 0; chunk + tid < total; chunk += blockDim.x) { outind = ((chunk + tid)/ncols)*stride + (chunk + tid) % ncols; out_buf[outind] = in_buf[chunk + tid]; } } __device__ inline void atomic_add(float* address, float value){ #if __CUDA_ARCH__ >= 200 // for Fermi, atomicAdd supports floats atomicAdd(address, value); #elif __CUDA_ARCH__ >= 110 float old = value; while ((old = atomicExch(address, atomicExch(address, 0.0f)+old))!=0.0f); #endif }
19,730
// ---------------------------------------------------------------------------- // CUDA code to compute minimun distance between n points // #include <stdio.h> #include <stdlib.h> #include <math.h> #include <time.h> #include<limits> #include<float.h> #define MAX_POINTS 1048576 #define block_size 1024 // ---------------------------------------------------------------------------- // Kernel Function to compute distance between all pairs of points // Input: // X: X[i] = x-coordinate of the ith point // Y: Y[i] = y-coordinate of the ith point // n: number of points // Output: // D: D[0] = minimum distance // // variable to keep track of number of blocks for which minimum calculation is done on the device __device__ unsigned int block_count = 0; __global__ void minimum_distance(float * X, float * Y, volatile float * D, int n) { // current thread operating = block_x * blickdim_x * threadid_x unsigned int curr_i = blockIdx.x * blockDim.x + threadIdx.x; float dx,dy, curr_dist; //intializing minimum distance to a float max value float minimum_dist = FLT_MAX; // to check if last block is completed or not bool last_blk; // sharing a list of local minimum distances calculated by all threads in a block __shared__ float block_local_mins[block_size]; // to make sure current thread lies withing the range of points in the grid if(curr_i < n-1) { //calculating distance for ith point pointed by curr_i to all other points for(int j = curr_i+1;j<n;j++) { dx = X[curr_i] - X[j]; //xi- xj dy = Y[curr_i] - Y[j]; //yi- yj curr_dist = sqrtf(dx*dx + dy*dy) ;// dij = √(xi − xj)^2+ (yi − yj)^2 //update minimum distance computed by this thread if(curr_dist < minimum_dist) { minimum_dist = curr_dist; } } //updating block mins list with the thread's minimum block_local_mins[threadIdx.x] = minimum_dist; //wait for all threads in that block to update this list __syncthreads(); // compute block local minimum with logarithmic reduction (binary tree based approach) int stop_index = (n%block_size); if(stop_index == 0){ stop_index = block_size; } else{ if(blockIdx.x != n/block_size) { stop_index = block_size; } } // traverse through the block adn update minimum //step of 2 for(int i=1;i<stop_index;i *=2){ if(threadIdx.x % (2*i) == 0 && (threadIdx.x +i) < stop_index-1){ if(block_local_mins[threadIdx.x] > block_local_mins[threadIdx.x + i]){ block_local_mins[threadIdx.x] = block_local_mins[threadIdx.x + i]; } __syncthreads(); } } //move the minimum computed from the first thread in every block to the correct spot in D if(threadIdx.x == 0){ D[blockIdx.x] = block_local_mins[0]; //also increment block_count int inc_val = atomicInc(&block_count,gridDim.x); last_blk = (inc_val == (gridDim.x -1)); } //make last thread in list compute global minimum and put it in D[0] if(last_blk && threadIdx.x ==0){ int num_blocks = n/block_size + (n % block_size != 0); for(int i=1;i<num_blocks;i++){ if(D[0] > D[i]){ D[0] = D[i]; } } } } } // ---------------------------------------------------------------------------- // Host function to compute minimum distance between points // Input: // X: X[i] = x-coordinate of the ith point // Y: Y[i] = y-coordinate of the ith point // n: number of points // Output: // D: minimum distance // float minimum_distance_host(float * X, float * Y, int n) { float dx, dy, Dij, min_distance, min_distance_i; int i, j; dx = X[1]-X[0]; dy = Y[1]-Y[0]; min_distance = sqrtf(dx*dx+dy*dy); for (i = 0; i < n-1; i++) { for (j = i+1; j < i+2; j++) { dx = X[j]-X[i]; dy = Y[j]-Y[i]; min_distance_i = sqrtf(dx*dx+dy*dy); } for (j = i+1; j < n; j++) { dx = X[j]-X[i]; dy = Y[j]-Y[i]; Dij = sqrtf(dx*dx+dy*dy); if (min_distance_i > Dij) min_distance_i = Dij; } if (min_distance > min_distance_i) min_distance = min_distance_i; } return min_distance; } // ---------------------------------------------------------------------------- // Print device properties void print_device_properties() { int i, deviceCount; cudaDeviceProp deviceProp; cudaGetDeviceCount(&deviceCount); printf("------------------------------------------------------------\n"); printf("Number of GPU devices found = %d\n", deviceCount); for ( i = 0; i < deviceCount; ++i ) { cudaGetDeviceProperties(&deviceProp, i); printf("[Device: %1d] Compute Capability %d.%d.\n", i, deviceProp.major, deviceProp.minor); printf(" ... multiprocessor count = %d\n", deviceProp.multiProcessorCount); printf(" ... max threads per multiprocessor = %d\n", deviceProp.maxThreadsPerMultiProcessor); printf(" ... max threads per block = %d\n", deviceProp.maxThreadsPerBlock); printf(" ... max block dimension = %d, %d, %d (along x, y, z)\n", deviceProp.maxThreadsDim[0], deviceProp.maxThreadsDim[1], deviceProp.maxThreadsDim[2]); printf(" ... max grid size = %d, %d, %d (along x, y, z)\n", deviceProp.maxGridSize[0], deviceProp.maxGridSize[1], deviceProp.maxGridSize[2]); printf(" ... warp size = %d\n", deviceProp.warpSize); printf(" ... clock rate = %d MHz\n", deviceProp.clockRate/1000); } printf("------------------------------------------------------------\n"); } // ---------------------------------------------------------------------------- // Main program - initializes points and computes minimum distance // between the points // int main(int argc, char* argv[]) { // Host Data float * hVx; // host x-coordinate array float * hVy; // host y-coordinate array float * hmin_dist; // minimum value on host // Device Data float * dVx; // device x-coordinate array float * dVy; // device x-coordinate array float * dmin_dist; // minimum value on device // Device parameters int MAX_BLOCK_SIZE; // Maximum number of threads allowed on the device int blocks; // Number of blocks in grid int threads_per_block; // Number of threads per block // Timing variables cudaEvent_t start, stop; // GPU timing variables struct timespec cpu_start, cpu_stop; // CPU timing variables float time_array[10]; // Other variables int i, size, num_points; float min_distance, sqrtn; int seed = 0; // Print device properties print_device_properties(); // Get device information and set device to use int deviceCount; cudaDeviceProp deviceProp; cudaGetDeviceCount(&deviceCount); if (deviceCount > 0) { cudaSetDevice(0); cudaGetDeviceProperties(&deviceProp, 0); MAX_BLOCK_SIZE = deviceProp.maxThreadsPerBlock; } else { printf("Warning: No GPU device found ... results may be incorrect\n"); } // Timing initializations cudaEventCreate(&start); cudaEventCreate(&stop); // Check input if (argc != 2) { printf("Use: %s <number of points>\n", argv[0]); exit(0); } if ((num_points = atoi(argv[argc-1])) < 2) { printf("Minimum number of points allowed: 2\n"); exit(0); } if ((num_points = atoi(argv[argc-1])) > MAX_POINTS) { printf("Maximum number of points allowed: %d\n", MAX_POINTS); exit(0); } // Allocate host coordinate arrays size = num_points * sizeof(float); hVx = (float *) malloc(size); hVy = (float *) malloc(size); hmin_dist = (float *) malloc(size); // Initialize points srand48(seed); sqrtn = (float) sqrt(num_points); for (i = 0; i < num_points; i++) { hVx[i] = sqrtn * (float)drand48(); hVy[i] = sqrtn * (float)drand48(); } // Allocate device coordinate arrays cudaMalloc(&dVx, size); cudaMalloc(&dVy, size); cudaMalloc(&dmin_dist, size); // Copy coordinate arrays from host memory to device memory cudaEventRecord( start, 0 ); cudaMemcpy(dVx, hVx, size, cudaMemcpyHostToDevice); cudaMemcpy(dVy, hVy, size, cudaMemcpyHostToDevice); cudaEventRecord(stop, 0); cudaEventSynchronize(stop); cudaEventElapsedTime(&(time_array[0]), start, stop); // Invoke kernel cudaEventRecord( start, 0 ); // ------------------------------------------------------------ int num_blocks = num_points/(block_size) + ((num_points % (block_size)) !=0); minimum_distance<<<num_blocks,block_size>>>(dVx,dVy,dmin_dist,num_points); // ------------------------------------------------------------ cudaEventRecord(stop, 0); cudaEventSynchronize(stop); cudaEventElapsedTime(&(time_array[1]), start, stop); // Copy result from device memory to host memory cudaEventRecord( start, 0 ); cudaMemcpy(hmin_dist, dmin_dist, sizeof(float), cudaMemcpyDeviceToHost); cudaEventRecord(stop, 0); cudaEventSynchronize(stop); cudaEventElapsedTime(&(time_array[2]), start, stop); // Compute minimum distance on host to check device computation clock_gettime(CLOCK_REALTIME, &cpu_start); min_distance = minimum_distance_host(hVx, hVy, num_points); clock_gettime(CLOCK_REALTIME, &cpu_stop); time_array[3] = 1000*((cpu_stop.tv_sec-cpu_start.tv_sec) +0.000000001*(cpu_stop.tv_nsec-cpu_start.tv_nsec)); // Print results printf("Number of Points = %d\n", num_points); printf("GPU Host-to-device = %f ms \n", time_array[0]); printf("GPU Device-to-host = %f ms \n", time_array[2]); printf("GPU execution time = %f ms \n", time_array[1]); printf("CPU execution time = %f ms\n", time_array[3]); printf("Min. distance (GPU) = %e\n", hmin_dist); printf("Min. distance (CPU) = %e\n", min_distance); printf("Relative error = %e\n", fabs(min_distance-hmin_dist[0])/min_distance); // Free device memory cudaFree(dVx); cudaFree(dVy); cudaFree(dmin_dist); // Free host memory free(hVx); free(hVy); }
19,731
#include<stdio.h> extern "C" void GPUDeviceInfo(const int gpu_device) { int deviceCount; cudaGetDeviceCount(&deviceCount); int device; if (deviceCount>0) { printf("#########\n"); for (device = 0; device < deviceCount; ++device) { cudaDeviceProp deviceProp; cudaGetDeviceProperties(&deviceProp, device); printf("Found CUDA device %d : %s\n",device, deviceProp.name); printf("Compute capability : %d.%d\n",deviceProp.major, deviceProp.minor); printf("#########\n"); } } else { printf("No CUDA device found \n"); } printf("Selecting device %d\n",gpu_device); cudaSetDevice(gpu_device); }
19,732
//#include "scale.h" #include <iostream> #include <fstream> #include <vector> #include <algorithm> #include <math.h> /* fabs */ #include <string.h> #include <stdlib.h> #include <sstream> #include <unordered_map> using namespace std; #define THREADS_PER_BLOCK 256 #define STREAM_COUNT 4 #define gpuErrchk(ans) { gpuAssert((ans), __FILE__, __LINE__); } int flag; inline void gpuAssert(cudaError_t code, const char *file, int line, bool abort=true) { if (code != cudaSuccess) { fprintf(stderr,"GPUassert: %s %s %d\n", cudaGetErrorString(code), file, line); if (abort) exit(code); } } void printArray(int *arr, int nov){ for (int j=0;j<nov;j++) { cout << j << " " << arr[j] << endl; } } __global__ void kernel3(int* adj, int* xadj, int* output, int nov){ int index = threadIdx.x + (blockIdx.x * blockDim.x); if(index < nov){ //int *marked = new int[n]; //memset(marked, -1, n * sizeof(int)); // bu belki silinebilir int localcount = 0; // int round = 0; // 0--> int s0 = xadj[index]; int e0 = xadj[index+1]; for(int i=s0; i < e0; i++){ // 0 --> 1 int neighbour_1 = adj[i]; int s1 = xadj[neighbour_1]; int e1 = xadj[neighbour_1+1]; for(int j=s1;j < e1; j++){ // 0 --> 1 --> 2 int neighbour_2 = adj[j]; if (neighbour_2 == index) continue; int s2 = xadj[neighbour_2]; int e2 = xadj[neighbour_2+1]; for(int k=s2; k < e2; k++){ // 0 --> 1 --> 2 --> 3 int neighbour_3 = adj[k]; if (neighbour_3 == index){ localcount+=1; break; } } } } output[index] = localcount; } } __global__ void kernel4(int* adj, int* xadj, int* output, int nov){ int index = threadIdx.x + (blockIdx.x * blockDim.x); if(index < nov){ //int *marked = new int[n]; //memset(marked, -1, n * sizeof(int)); // bu belki silinebilir int localcount = 0; // int round = 0; // 0--> int s0 = xadj[index]; int e0 = xadj[index+1]; for(int i=s0; i < e0; i++){ // 0 --> 1 int neighbour_1 = adj[i]; int s1 = xadj[neighbour_1]; int e1 = xadj[neighbour_1+1]; for(int j=s1;j < e1; j++){ // 0 --> 1 --> 2 int neighbour_2 = adj[j]; //eliminate 0 == 2 if (neighbour_2 == index) continue; int s2 = xadj[neighbour_2]; int e2 = xadj[neighbour_2+1]; for(int k=s2; k < e2; k++){ // 0 --> 1 --> 2 --> 3 int neighbour_3 = adj[k]; //eliminate 3 == 0 if (neighbour_3 == index) continue; // eliminate 3 ==1 if (neighbour_3 == neighbour_1) continue; int s3 = xadj[neighbour_3]; int e3 = xadj[neighbour_3+1]; for(int n=s3; n < e3; n++){ //0 -->1 -->2 -->3 -->4 int neighbour_4 = adj[n]; if (neighbour_4 == index){ localcount+=1; break; } } } } } output[index] = localcount; } } __global__ void kernel5(int* adj, int* xadj, int* output, int nov){ int index = threadIdx.x + (blockIdx.x * blockDim.x); if(index < nov){ //int *marked = new int[n]; //memset(marked, -1, n * sizeof(int)); // bu belki silinebilir int localcount = 0; // int round = 0; // 0--> int s0 = xadj[index]; int e0 = xadj[index+1]; for(int i=s0; i < e0; i++){ // 0 --> 1 int neighbour_1 = adj[i]; int s1 = xadj[neighbour_1]; int e1 = xadj[neighbour_1+1]; for(int j=s1;j < e1; j++){ // 0 --> 1 --> 2 int neighbour_2 = adj[j]; //eliminate 0 == 2 if (neighbour_2 == index) continue; int s2 = xadj[neighbour_2]; int e2 = xadj[neighbour_2+1]; for(int k=s2; k < e2; k++){ // 0 --> 1 --> 2 --> 3 int neighbour_3 = adj[k]; //eliminate 3 == 0 if (neighbour_3 == index) continue; // eliminate 3 ==1 if (neighbour_3 == neighbour_1) continue; int s3 = xadj[neighbour_3]; int e3 = xadj[neighbour_3+1]; for(int n=s3; n < e3; n++){ //0 -->1 -->2 -->3 -->4 int neighbour_4 = adj[n]; //eliminate 4 == 0 if (neighbour_4 == index) continue; // eliminate 4 ==1 if (neighbour_4 == neighbour_1) continue; // eliminate 4 ==2 if (neighbour_4 == neighbour_2) continue; int s4 = xadj[neighbour_4]; int e4 = xadj[neighbour_4+1]; for(int o=s4; o < e4; o++){ //0 -->1 -->2 -->3 -->4--> 5 int neighbour_5 = adj[o]; if (neighbour_5 == index){ localcount+=1; break; } } } } } } output[index] = localcount; } } void wrapper(int *xadj, int *adj, int n, int nov, int nnz){ // int X = nov; // int Y = maxSize; // int Z = maxSize; // // dim3 threadsPerBlock(8, 8, 8); // dim3 numBlocks(X/threadsPerBlock.x, /* for instance 512/8 = 64 // Y /threadsPerBlock.y, // Z/threadsPerBlock.z); // cudaSetDevice(0); int *adj_d; int *xadj_d; int *output_d; int *output_h = new int[nov]; int numBlock = (nov + THREADS_PER_BLOCK - 1) / THREADS_PER_BLOCK; cudaEvent_t start, stop; float elapsedTime; /* int novForThread = (nov+STREAM_COUNT-1)/STREAM_COUNT; int novStart = novForThread * threadId; int novEnd = novForThread * (threadId+1); if (novEnd> nov) novEnd = nov; int numBlock = (novEnd-novStart + THREADS_PER_BLOCK-1) / THREADS_PER_BLOCK; */ gpuErrchk(cudaMalloc((void**)&adj_d, (nnz) * sizeof(int))); gpuErrchk(cudaMalloc((void**)&xadj_d, (nov + 1) * sizeof(int))); gpuErrchk(cudaMalloc((void**)&output_d, (nov) * sizeof(int))); //gpuErrchk(cudaMallocHost((void **)&output_h, (nov) * sizeof(int))); gpuErrchk(cudaMemcpy(adj_d, adj, (nnz) * sizeof(int), cudaMemcpyHostToDevice)); gpuErrchk(cudaMemcpy(xadj_d, xadj, (nov + 1) * sizeof(int), cudaMemcpyHostToDevice)); cudaEventCreate(&start); cudaEventRecord(start, 0); if (n==3) kernel3<<<numBlock, THREADS_PER_BLOCK>>>(adj_d, xadj_d, output_d, nov); else if (n==4) kernel4<<<numBlock, THREADS_PER_BLOCK>>>(adj_d, xadj_d, output_d, nov); else if (n==5) kernel5<<<numBlock, THREADS_PER_BLOCK>>>(adj_d, xadj_d, output_d, nov); //combination<<<numBlocks, threadsPerBlock>>>(adj_d, xadj_d, output_d, n, nov); gpuErrchk(cudaDeviceSynchronize()); gpuErrchk(cudaMemcpy(output_h, output_d, (nov) * sizeof(int), cudaMemcpyDeviceToHost)); cudaEventCreate(&stop); cudaEventRecord(stop, 0); cudaEventSynchronize(stop); cudaEventElapsedTime(&elapsedTime, start, stop); if(flag == 0) printArray(output_h,nov); cudaFree(adj_d); cudaFree(xadj_d); if(flag == 1) printf("GPU scale took: %f s\n", elapsedTime/1000); } void read_mtxbin(string fname, int k){ //cout << "fname: " << fname << endl; ifstream infile(fname); int a, b; int nnv = 0; unordered_map<int, vector<int> > hashmap; int maxElement = -1; while (infile >> a >> b) { nnv+=2; hashmap[a].push_back(b); hashmap[b].push_back(a); if(b > maxElement){ maxElement = b; } } //cout << end1-start1 << " -- ILK OKUMA SU (s).\n"; int nov = maxElement +1; //cout <<"nov " << nov << endl; //cout <<"nnv " << nnv << endl; int * adj = new int[nnv]; int * xadj = new int[nov+1]; xadj[0]=0; int j = 0; int maxSize = -1; for(int i=0; i < nov ; i++ ){ auto current = hashmap.find(i); if (current == hashmap.end()){ xadj[i+1] = xadj[i]; } else{ int size = current->second.size(); maxSize = max(size,maxSize); xadj[i+1] = xadj[i] + size; for(auto val : current->second) { adj[j] = val; j++; } } } // cout << "maxSize: "<<maxSize<<endl; // cout << end-start << " -- OKUMA SURE (s).\n"; wrapper(xadj,adj,k,nov,nnv); //cout<<"CYCLES: --> "<<countCycles_sparse(xadj, adj,k,nov)<<endl; /*double end2 = omp_get_wtime(); cout << end2-start << " -- TOTAL SURE (s).\n";*/ } int main(int argc, char *argv[]){ char* fname = argv[1]; int k = atoi(argv[2]); flag = atoi(argv[3]); read_mtxbin(fname,k); return 0; }
19,733
#include "includes.h" __global__ void Mask_Difference_Kernel( int* A, int* B, int* devOut) { const int idx = blockDim.x*blockIdx.x + threadIdx.x; if(A[idx] == B[idx]) devOut[idx] = 0; else devOut[idx] = 1; // Should test if the extra algebra ops are worth removing the if-statement // Convert to {-1, +1} //int aval = A[idx]*2 - 1; //int bval = B[idx]*2 - 1; //devOut[idx] = (aval*bval+1)/2; }
19,734
#include <iostream> #include <cuda.h> #include <cuda_runtime.h> #include <stdlib.h> #include <ctime> using namespace std; __global__ void AddInts(int * a, int* b, int count) { int id = blockIdx.x * blockDim.x * threadIdx.x; if (id < count) { a[id]+=b[id]; } } int main(int argc, char const *argv[]) { srand (time(NULL)); int count = 100; int * h_a = new int[count]; int * h_b = new int[count]; for (int i = 0; i < count; i++) { h_a[i] = rand() % 1000; h_b[i] = rand() % 1000; } cout << "Prior to addition:" << endl; for (int i = 0; i < 5; i++) { cout << h_a[i] << " " << h_b[i] << endl; } int *d_a, *d_b; if (cudaMalloc(&d_a, sizeof(int)*count) != cudaSuccess) { cout<< "Could not allocate d_a" << endl; return 1; } if (cudaMalloc(&d_b, sizeof(int)*count) != cudaSuccess) { cout<< "Could not allocate d_b" << endl; cudaFree(d_a); return 1; } if (cudaMemcpy(d_a, h_a, sizeof(int) * count, cudaMemcpyHostToDevice)!=cudaSuccess) { cout<< "Could not copy d_a" << endl; cudaFree(d_a); cudaFree(d_b); return 1; } if (cudaMemcpy(d_b, h_b, sizeof(int) * count, cudaMemcpyHostToDevice)!=cudaSuccess) { cout<< "Could not copy d_b" << endl; cudaFree(d_a); cudaFree(d_b); return 1; } AddInts <<<count/256+1, 256>>>(d_a,d_b,count); if (cudaMemcpy(h_a, d_a, sizeof(int) * count, cudaMemcpyDeviceToHost)!=cudaSuccess) { cout<< "Could not copy back from device" << endl; cudaFree(d_a); cudaFree(d_b); delete[] h_a; delete[] h_b; return 1; } for (int i = 0; i < 5; i++) { cout <<"It is: " << h_a[i] << endl; } delete[] h_a; delete[] h_b; return 0; }
19,735
#include "cuda_runtime.h" //#include "device_launch_parameters.h" #include <stdio.h> #include <stdlib.h> #include <math.h> // CUDA kernel __global__ void vecCompare(int *R, int *G, int *B, int *result, int n) //A is for the green array { int i = blockIdx.x * blockDim.x + threadIdx.x; if(i < n) { if(G[i] > 2 * R[i] || G[i] > 2 * B[i]) { result[i] = 1; } else { result[i] = 0; } } } extern "C" void compareMatrices(int height, int width, int*r, int*g, int*b, int*green) { // Size of vectors int n = height * width; // Device input vectors int *d_r; int *d_g; int *d_b; //Device output vector int *d_green; // Size, in bytes, of each vector size_t bytes = n*sizeof(int); // Allocate memory for each vector on GPU cudaMalloc(&d_r, bytes); cudaMalloc(&d_g, bytes); cudaMalloc(&d_b, bytes); cudaMalloc(&d_green, bytes); // Copy host vectors to device cudaMemcpy( d_r, r, bytes, cudaMemcpyHostToDevice); cudaMemcpy( d_g, g, bytes, cudaMemcpyHostToDevice); cudaMemcpy( d_b, b, bytes, cudaMemcpyHostToDevice); int blockSize, gridSize; // Number of threads in each thread block blockSize = 1024; // Number of thread blocks in grid gridSize = (int)ceil((float)n/blockSize); // Execute the kernel vecCompare<<<gridSize, blockSize>>>(d_r, d_g, d_b, d_green, n); // Copy array back to host cudaMemcpy( green, d_green, bytes, cudaMemcpyDeviceToHost ); // Release device memory cudaFree(d_r); cudaFree(d_g); cudaFree(d_b); cudaFree(d_green); // Release host memory free(r); free(g); free(b); }
19,736
#include <cstdio> #include <iostream> #include <vector> #include <cmath> #include <thrust/device_vector.h> #include <thrust/extrema.h> using namespace std; #define CSC(call) do { cudaError_t res = call; if (res != cudaSuccess) { fprintf(stderr, "CUDA Error in %s:%d: %s\n", __FILE__, __LINE__, cudaGetErrorString(res)); exit(0); } } while (0) __global__ void kernelSwap(int n, double *cols, double *vec, int step, int maxPos) { int idx = blockDim.x * blockIdx.x + threadIdx.x; int offsetx = blockDim.x * gridDim.x; double tmp; for (int col = idx; col <= n; col += offsetx) { if (col < step) { continue; } if (col == n) { tmp = vec[step]; vec[step] = vec[maxPos]; vec[maxPos] = tmp; } else { tmp = cols[col * n + step]; cols[col * n + step] = cols[col * n + maxPos]; cols[col * n + maxPos] = tmp; } } } __global__ void kernelModify(int n, double *cols, double *vec, int step) { int idx = blockDim.x * blockIdx.x + threadIdx.x; int idy = blockDim.y * blockIdx.y + threadIdx.y; int offsetx = blockDim.x * gridDim.x; int offsety = blockDim.y * gridDim.y; double coef; for (int row = idx; row < n; row += offsetx) { for (int col = idy; col <= n; col += offsety) { if (row <= step || col <= step) { continue; } coef = cols[step * n + row] / cols[step * n + step]; if (col == n) { vec[row] = vec[row] - coef * vec[step]; } else { cols[col * n + row] = cols[col * n + row] - coef * cols[col * n + step]; } } } } struct compareKeyValue { __host__ __device__ bool operator()(double a, double b) { return fabs(a) < fabs(b); } }; int main() { ios_base::sync_with_stdio(false); int n; cin >> n; double *cols = new double[n * n]; double *vec = new double[n]; for (int row = 0; row < n; row++) { for (int col = 0; col < n; col++) { cin >> cols[col * n + row]; } } for (int row = 0; row < n; row++) { cin >> vec[row]; } double *devCols, *devVec; CSC(cudaMalloc(&devCols, sizeof(double) * n * n)); CSC(cudaMalloc(&devVec, sizeof(double) * n)); CSC(cudaMemcpy(devCols, cols, sizeof(double) * n * n, cudaMemcpyHostToDevice)); CSC(cudaMemcpy(devVec, vec, sizeof(double) * n, cudaMemcpyHostToDevice)); int maxPos; for (int step = 0; step < n - 1; step++) { thrust::device_ptr<double> devPtr = thrust::device_pointer_cast(devCols + step * (n + 1)); thrust::device_ptr<double> maxPtr = thrust::max_element(devPtr, devPtr + (n - step), compareKeyValue()); maxPos = &maxPtr[0] - &devPtr[0] + step; kernelSwap<<<256, 256>>>(n, devCols, devVec, step, maxPos); CSC(cudaGetLastError()); kernelModify<<<dim3(16, 16), dim3(16, 16)>>>(n, devCols, devVec, step); CSC(cudaGetLastError()); } CSC(cudaMemcpy(cols, devCols, sizeof(double) * n * n, cudaMemcpyDeviceToHost)); CSC(cudaMemcpy(vec, devVec, sizeof(double) * n, cudaMemcpyDeviceToHost)); CSC(cudaFree(devCols)); CSC(cudaFree(devVec)); vector<double> res(n); double tmp; for (int row = n - 1; row >= 0; row--) { tmp = 0; for (int col = row + 1; col < n; col++) { tmp += cols[col * n + row] * res[col]; } res[row] = (vec[row] - tmp) / cols[row * n + row]; } cout.precision(10); cout.setf(ios::scientific); for (int i = 0; i < n; i++) { cout << res[i] << " "; } cout << endl; delete[] cols; delete[] vec; return 0; }
19,737
#include <stdio.h> #include <stdlib.h> #include <time.h> //CUDA RunTime API #include <cuda_runtime.h> #include <device_launch_parameters.h> #include "device_functions.h" #define THREAD_NUM 256 #define raw_row 512*680 #define raw_column 224 #define MATRIX_SIZE raw_row*raw_column const int blocks_num = 16; //// __global__ м˷ //__global__ static void matMultCUDA(Imagdata* imagdata, CovData* imagecov, cBYTE* datatemp, clock_t* time) //{ // // //ʾĿǰ thread ǵڼ thread 0 ʼ㣩 // const int tid = threadIdx.x; // // //ʾĿǰ thread ڵڼ block 0 ʼ㣩 // const int bid = blockIdx.x; // // //// ֻ÷ // //// bid tid thread Ӧü row column // //const int idx = bid * THREAD_NUM + tid; // //const int row = idx / n; // //const int column = idx % n; // // //¼㿪ʼʱ // // //ֻ thread 0 threadIdx.x = 0 ʱ򣩽м¼ÿ block ¼ʼʱ估ʱ // if (tid == 0) time[bid] = clock(); // if (tid == 0) // printf("Hello thread %d, f=%f\n", tid); // // /* // Э̿ʼ // */ // // long count; // long count2; // // imagecov->row = imagdata->Dim; // imagecov->column = imagdata->Dim; // // for (int i = 0; i < imagdata->Dim; ++i) { // count = 0; // int j = 0; // for (j = 0; j < imagdata->imgSize; ++j) { // count += imagdata->data[i * imagdata->imgSize + j]; // } // int mean = count / imagdata->imgSize; // for (j = 0; j < imagdata->imgSize; ++j) { // datatemp[i * imagdata->imgSize + j] = imagdata->data[i * imagdata->imgSize + j] - mean; // } // } // //cout << "(imgcompute.cpp)ؾֵ\n" << endl; // printf("(imgcompute.cpp)ؾֵ...\n"); // //----------------------------------------------------------------------------------------------------------- // //˲ʱΪ45롣Ϊ21롣ѭμ룬ʱΪ12s // // for (int i = 0; i < imagdata->Dim; ++i) // { // int j; // for (j = i; j < imagdata->Dim; ++j) // { // count2 = 0; // int k; // //for (k = 0; k < imagdata->imgSize; ++k) { // // // count2 += datatemp[i * imagdata->imgSize + k] * datatemp[j * imagdata->imgSize + k]; // //} // for (k = 0; k < imagdata->imgSize; ++k) { // // count2 += datatemp[i * imagdata->imgSize + k] * datatemp[j * imagdata->imgSize + k]; // } // imagecov->data[i * (imagecov->row) + j] = count2 / (imagdata->imgSize - 1); // if (i != j) { // imagecov->data[j * (imagecov->row) + i] = count2 / (imagdata->imgSize - 1); // } // } // } // // delete[]datatemp; // printf("(imgcompute.cpp)Э\n"); // // /* // Э̽ // */ // // // //ʱ,¼ֻ thread 0 threadIdx.x = 0 ʱ򣩽Уÿ block ¼ʼʱ估ʱ // if (tid == 0) // { // time[bid + blocks_num] = clock(); // } // //} // // //__global__ void cudatest() { // printf("success!"); //} // // //extern "C" void cuda_juzhen(RawImag * myraw, Imagdata * imagdata, CovData * imagecov) //{ // clock_t* time; // // printf("1\n"); // // //CPUGPUڴ ͬһṹҪֿʵ // //CPUڴ // imagecov->data = new covBYTE[sizeof(covBYTE) * imagdata->Dim * imagdata->Dim]; // cBYTE* datatemp = new cBYTE[sizeof(cBYTE) * imagdata->rawSize]; // rawSize 512x680x224,8000W // // printf("2\n"); // // // //cudaMalloc ȡһԿڴ // //ṹڴҪṹ屾Լṹÿһ // // // Imagdata* cuda_imagdata; // CovData* cuda_imagecov; // cBYTE* cuda_datatemp; // // cuda_imagdata = new Imagdata[sizeof(Imagdata)]; // ṹ飨ָ룩ڴ // cuda_imagecov = new CovData[sizeof(CovData)]; // cuda_datatemp = new cBYTE[sizeof(cBYTE)]; // // // //memset(&cuda_imagdata, 0, sizeof(Imagdata)); // //memset(&cuda_imagecov, 0, sizeof(CovData)); // //memset(&cuda_datatemp, 0, sizeof(cBYTE)); // // // cuda_imagdata->Col = imagdata->Col; // cuda_imagdata->Row = imagdata->Row; // cuda_imagdata->Dim = imagdata->Dim; // cuda_imagdata->rawSize = imagdata->rawSize; // cuda_imagdata->imgSize = imagdata->imgSize; // // cuda_imagecov->row = imagdata->Dim; // cuda_imagecov->column = imagdata->Dim; // // // cuda_imagdata->data = new covBYTE[sizeof(covBYTE) * imagdata->Dim * imagdata->Dim];; // cuda_datatemp = new cBYTE[sizeof(cBYTE) * imagdata->rawSize]; // rawSize 512x680x224,8000W; // // //cudaMalloc((void**)&cuda_imagdata->Row, sizeof(int)); // //cudaMalloc((void**)&cuda_imagdata->Col, sizeof(int)); // //cudaMalloc((void**)&cuda_imagdata->Dim, sizeof(int)); // //cudaMalloc((void**)&cuda_imagdata->rawSize, sizeof(int)); // //cudaMalloc((void**)&cuda_imagdata->imgSize, sizeof(int)); // // //cudaMalloc((void**)&cuda_imagdata, sizeof(Imagdata)); // cudaMalloc((void**)&cuda_imagdata->data, sizeof(cBYTE) * imagdata->rawSize); // // printf("2.1\n"); // // //cudaMalloc((void**)&cuda_imagecov->row, sizeof(int)); // //cudaMalloc((void**)&cuda_imagecov->column, sizeof(int)); // // // //cudaMalloc((void**)&cuda_imagecov, sizeof(CovData)); // // cuda_imagecov->data = new cBYTE[sizeof(cBYTE) * imagdata->rawSize]; // cudaMalloc((void**)&cuda_imagecov->data, sizeof(covBYTE) * imagdata->Dim * imagdata->Dim); // // printf("2.3\n"); // // cudaMalloc((void**)&cuda_datatemp, sizeof(cBYTE) * imagdata->rawSize); // // cudaMalloc((void**)&time, sizeof(clock_t) * blocks_num * 2); // // printf("3\n"); // // // //cudaMemcpy ľƵԿڴ // //cudaMemcpyHostToDevice - ڴ渴ƵԿڴ // //cudaMemcpyDeviceToHost - Կڴ渴Ƶڴ // // //cudaMemcpy(cuda_imagdata, imagdata, sizeof(cBYTE) * imagdata->rawSize, cudaMemcpyHostToDevice); // cudaMemcpy(cuda_imagdata->data, imagdata->data, sizeof(cBYTE) * imagdata->rawSize, cudaMemcpyHostToDevice); // // //cudaMemcpy(cuda_imagecov, imagecov, sizeof(covBYTE) * imagdata->rawSize, cudaMemcpyHostToDevice); // cudaMemcpy(cuda_imagecov->data, imagecov->data, sizeof(covBYTE) * imagdata->rawSize, cudaMemcpyHostToDevice); // // cudaMemcpy(cuda_datatemp, datatemp, sizeof(cBYTE) * imagdata->rawSize, cudaMemcpyHostToDevice); // // printf("4\n"); // // // CUDA ִк ﷨<<<block Ŀ, thread Ŀ, shared memory С>>>(...); // //matMultCUDA <<< blocks_num, THREAD_NUM>>> (cuda_imag;data, cuda_imagecov, cuda_datatemp, time); // cudatest << < 1, 64 >> > ; // // cudaError_t cudaStatus = cudaGetLastError(); // if (cudaStatus != cudaSuccess) // { // fprintf(stderr, "addKernel launch failed: %s\n", cudaGetErrorString(cudaStatus)); // } // // // printf("5\n"); // // /*ѽʾоƬƻڴ*/ // // clock_t time_use[blocks_num * 2]; // // //cudaMemcpy Դиƻڴ // Imagdata* new_imagdata; // CovData* new_imagecov; // // new_imagdata = new Imagdata[sizeof(Imagdata)]; // ṹ飨ָ룩ڴ // new_imagecov = new CovData[sizeof(CovData)]; // // //cudaMemcpy(new_imagdata, cuda_imagdata, sizeof(cBYTE) * imagdata->rawSize, cudaMemcpyDeviceToHost); // cudaMemcpy(new_imagdata->data, cuda_imagdata->data, sizeof(cBYTE) * imagdata->rawSize, cudaMemcpyDeviceToHost); // //cudaMemcpy(new_imagecov, cuda_imagecov, sizeof(covBYTE) * imagdata->rawSize, cudaMemcpyDeviceToHost); // cudaMemcpy(new_imagecov->data, cuda_imagecov->data, sizeof(covBYTE) * imagdata->rawSize, cudaMemcpyDeviceToHost); // // cudaMemcpy(&time_use, time, sizeof(clock_t) * blocks_num * 2, cudaMemcpyDeviceToHost); // // printf("6\n"); // // //Free // cudaFree(cuda_imagdata->data); // //cudaFree(cuda_imagdata); // // cudaFree(cuda_imagecov->data); // //cudaFree(cuda_imagdata); // // cudaFree(cuda_datatemp); // cudaFree(time); // // //ÿ block Ŀʼʱ䣬Ľʱȡʱ // clock_t min_start, max_end; // // min_start = time_use[0]; // // max_end = time_use[blocks_num]; // // for (int i = 1; i < blocks_num; i++) // { // if (min_start > time_use[i]) min_start = time_use[i]; // // if (max_end < time_use[i + blocks_num]) max_end = time_use[i + blocks_num]; // } // // //˺ʱ // clock_t final_time = max_end - min_start; // // printf("gputime: %d\n", final_time); // //}
19,738
/* * * 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 <iostream> #include <fstream> using namespace std; int patient_count; int icd_count; int* patients_host; int* icds_host; /** * 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); \ } } void free_all() { CUDA_CHECK_RETURN(cudaFreeHost(patients_host)); CUDA_CHECK_RETURN(cudaFreeHost(icds_host)); } void read_patients() { patient_count = 0; string line; ifstream file("csv_data/patients_sorted_short.csv"); if (file.is_open()) { // skip first line getline(file, line); while (getline(file, line)) { patient_count ++; } } printf("%d patients read\n", patient_count); CUDA_CHECK_RETURN(cudaHostAlloc((void**) &patients_host, patient_count * 4 * sizeof(int), cudaHostAllocDefault)); patient_count = 0; file.clear(); file.seekg(0, ios::beg); if (file.is_open()) { // skip first line getline(file, line); while (getline(file, line)) { char * dup = strdup(line.c_str()); char * record = strtok(dup, ","); for (int i = 0; i < 4; i++) { patients_host[patient_count * 4 + i] = atoi(record); record = strtok(NULL, ","); } patient_count++; } } } void read_icds() { icd_count = 0; string line; ifstream file("csv_data/icds.csv"); if (file.is_open()) { // skip first line getline(file, line); while (getline(file, line)) { icd_count ++; } } printf("%d ICDs read\n", icd_count); CUDA_CHECK_RETURN(cudaHostAlloc((void**) &icds_host, icd_count * sizeof(int), cudaHostAllocDefault)); icd_count = 0; file.clear(); file.seekg(0, ios::beg); if (file.is_open()) { // skip first line getline(file, line); while (getline(file, line)) { char * dup = strdup(line.c_str()); char * record = strtok(dup, ","); icds_host[icd_count] = atoi(record); icd_count++; } } } int main(int argc, char* argv[]) { read_patients(); read_icds(); free_all(); }
19,739
#include "includes.h" __global__ void MarkSplits(int size, bool force, int minPartSize, int maxPartSize, int* partSizes, int* splitsToMake) { int idx = blockIdx.x * blockDim.x + threadIdx.x; if (idx < size) { int currentSize = partSizes[idx]; bool shouldSplit = currentSize > maxPartSize && (force || currentSize > minPartSize * 2); splitsToMake[idx] = shouldSplit ? 1 : 0; } }
19,740
#include <cuda.h> #include <iostream> #include <cstdlib> using namespace std; #define THREADSPERBLOCK 1024 // __global__ void mandel(char *d_vet, int MAX_N, int MAX_COL, int MAX_ROW) { __global__ void mandel(char *d_vet, int MAX_ROW, int MAX_COL, int MAX_NUM) { int i = blockIdx.x * blockDim.x + threadIdx.x; if (i < MAX_ROW * MAX_COL) { int c = i % MAX_COL; int r = i / MAX_COL; int n = 0; float x=0, y=0, tmp; while ((x*x) + (y*y) < 4 && ++n < MAX_NUM) { tmp = x*x - y*y + ((float)c * 2 / MAX_COL - 1.5); y = x * y * 2 + ((float)r * 2 / MAX_ROW - 1); x = tmp; } d_vet[i] = (n == MAX_NUM ? '#' : '.'); } } int main(int argc, char **argv) { int MAX_ROW; int MAX_COL; int MAX_NUM; if (argc == 4) { MAX_ROW = atoi(argv[1]); MAX_COL = atoi(argv[2]); MAX_NUM = atoi(argv[3]); } else exit(1); int MAX_TAM = MAX_ROW * MAX_COL; size_t i, size = MAX_TAM * sizeof(char); int nBlocks = ( MAX_TAM + THREADSPERBLOCK - 1 ) / THREADSPERBLOCK; char *h_vet = (char *)malloc(size); char *d_vet = (char *)malloc(size); cudaSetDevice(0); cudaMalloc((void**)&d_vet, size); mandel<<<nBlocks, THREADSPERBLOCK>>>(d_vet, MAX_ROW, MAX_COL, MAX_NUM); cudaMemcpy(h_vet, d_vet, size, cudaMemcpyDeviceToHost); cudaFree(d_vet); for(i = 0; i < MAX_TAM; ++i) { cout << h_vet[i]; if (i % MAX_COL == MAX_COL-1) cout << endl; } free(h_vet); }
19,741
#include <stdio.h> __global__ void report(){ int i = blockIdx.x; int j = threadIdx.x; printf("My group id is %d, and my thread id is %d\n",i,j); } __global__ void report_in_detail(){ int ix = blockIdx.x; int iy = blockIdx.y; int iz = blockIdx.z; int jx = threadIdx.x; int jy = threadIdx.y; int jz = threadIdx.z; printf("My group id is (%d,%d,%d), and my thread id is (%d,%d,%d)\n",ix,iy,iz,jx,jy,jz); } int main(){ report_in_detail<<<3,2>>>(); cudaThreadSynchronize(); // 同步标识。让 CPU 等待 GPU 运行结束 printf("-----------\n"); report_in_detail<<<dim3(1,1,3),dim3(1,2,1)>>>(); return 0; }
19,742
#include "includes.h" __global__ void gpu_transpose(const float* src, float* dst, int colssrc, int colsdst, int n) { int tid = threadIdx.x + blockIdx.x * blockDim.x; int stride = blockDim.x * gridDim.x; while (tid < n) { int cdst = tid % colsdst; int rdst = tid / colsdst; int rsrc = cdst; int csrc = rdst; dst[tid] = src[rsrc * colssrc + csrc]; tid += stride; } }
19,743
#include <cuda_runtime.h> int main() { int* a; cudaMalloc(&a, 100); cudaFree(a); return 0; }
19,744
#include "includes.h" __global__ void poli_warp(float* poli, const int N) { int idx = blockIdx.x * blockDim.x + threadIdx.x; float x; if (idx < N) { x = poli[idx]; poli[idx] = 5 + x * ( 7 - x * (9 + x * (5 + x * (5 + x))))- 1.0f/x + 3.0f/(x*x) + x/5.0f; } poli[idx] = x; }
19,745
__global__ void vecAdd(float * in1, float * in2, float * out, int len) { //@@ Insert code to implement vector addition here int i = blockDim.x*blockIdx.x+threadIdx.x; if( i < len ) out[i] = in1[i] + in2[i]; }
19,746
#include <thrust/device_vector.h> #include <thrust/sort.h> #include <thrust/functional.h> int main(void) { thrust::device_vector<int> data(8); data[0] = 6; data[1] = 3; data[2] = 7; data[3] = 5; data[4] = 9; data[5] = 0; data[6] = 8; data[7] = 1; thrust::sort(data.begin(), data.end()); std::cout<<"ascending"<<std::endl; for (int i = 0; i < data.size(); i++) std::cout<<data[i]<<std::endl; thrust::sort(data.begin(), data.end(), thrust::greater<int>()); std::cout<<"descending"<<std::endl; for (int i = 0; i < data.size(); i++) std::cout<<data[i]<<std::endl; }
19,747
#include <stdio.h> #include <stdlib.h> #include <math.h> #define BLOCKSIZE 4 // Number of threads in each thread block // CUDA kernel. Each thread takes care of one element of a __global__ void diffKernel( float *in, float *out, int n ) { int i = blockIdx.x * blockDim.x + threadIdx.x; if(i < n - 1) out[i] = in[i + 1] - in[i]; } int main( int argc, char* argv[] ) { // Size of vectors int i; float input[] = {4, 5, 6, 7, 19, 10, 0, 4, 2, 3, 1, 7, 9, 11, 45, 23, 99, 29}; int n = sizeof(input) / sizeof(float); //careful, this usage only works with statically allocated arrays, NOT dynamic arrays // Host input vectors float *h_in = input; //Host output vector float *h_out = (float *) malloc((n - 1) * sizeof(float)); // Device input vectors float *d_in; //Device output vector float *d_out; // Size, in bytes, of each vector size_t bytes = n * sizeof(float); // Allocate memory for each vector on GPU cudaMalloc(&d_in, bytes); cudaMalloc(&d_out, bytes - sizeof(float)); // Copy host data to device cudaMemcpy( d_in, h_in, bytes, cudaMemcpyHostToDevice); // TODO: setup the blocksize and gridsize and launch the kernel below. // Number of threads in each thread block int blockSize = BLOCKSIZE; // Number of thread blocks in grid int threadCount = ceil( n / (float) BLOCKSIZE); // Execute the kernel diffKernel<<<blockSize, threadCount>>>(d_in, d_out, n); // Copy array back to host cudaMemcpy( h_out, d_out, bytes - sizeof(float), cudaMemcpyDeviceToHost ); // Show the result printf("The original array is: "); for(i = 0; i < n; i ++) printf("%4.0f,", h_in[i] ); printf("\n\nThe diff array is: "); for(i = 0; i < n - 1; i++) printf("%4.0f,", h_out[i] ); puts(""); // Release device memory cudaFree(d_in); cudaFree(d_out); // Release host memory free(h_out); return 0; }
19,748
#include <iostream> using namespace std; int main() { int ver; cudaDriverGetVersion(&ver); cout << "DRIVER VERSION: " << ver << "\n"; cudaRuntimeGetVersion (&ver); cout << "RUNTIME VERSION: " << ver << "\n"; cout << "\n"; size_t total_mem, free_mem; cudaMemGetInfo(&free_mem, &total_mem); cout << "TOTAL MEMORY: " << total_mem * 1e-6f << "MB\n"; cout << "FREE MEMORY: " << free_mem * 1e-6f << "MB\n"; cout << "\n"; cudaDeviceProp prop; int numberOfDevices; cudaGetDeviceCount(&numberOfDevices); cout << "Number of devices: " << numberOfDevices << "\n"; for (int i = 0; i < numberOfDevices; i++) { cout << "\n"; cudaGetDeviceProperties(&prop, i); cout << "Name: " << prop.name << endl; cout << "Clock rate: " << prop.clockRate * 1e-6f << "Ghz\n"; cout << "Device compute capability: " << prop.major << "." << prop.minor << "\n"; cout << "Multiprocessor count: " << prop.multiProcessorCount << "\n"; cout << "Total global mem: " << prop.totalGlobalMem * 1e-6f << "MB\n"; cout << "Max threads per MProcessor: " << prop.maxThreadsPerMultiProcessor << "\n"; cout << "Shared memory per Block: " << prop.sharedMemPerBlock * 1e-6f << "MB\n"; cout << "Warp size: " << prop.warpSize << "\n"; cout << "Max threads pen Block: " << prop.maxThreadsPerBlock << "\n"; cout << "Max block dimension: (" << prop.maxThreadsDim[0] << ", " << prop.maxThreadsDim[1] << ", " << prop.maxThreadsDim[2] << ")\n"; cout << "Max grid dimension: (" << prop.maxGridSize[0] << ", " << prop.maxGridSize[1] << ", " << prop.maxGridSize[2] << ")\n"; cout << "Kernel execution timeout is " << (prop.kernelExecTimeoutEnabled ? "on" : "off") << "\n"; cout << "Device overlap is " << (prop.deviceOverlap ? "on" : "off") << "\n"; } cout << endl; return 0; }
19,749
#include <thrust/device_vector.h> #include <thrust/host_vector.h> #include <iostream> #include <chrono> #include <thrust/extrema.h> #include <thrust/execution_policy.h> #include <thrust/functional.h> int main() { double stocks; std::cin >> stocks; thrust::host_vector<double> host; for(int i =0; i < 2517; i++){ std::cin >> stocks; host.push_back(stocks); } /* na linha abaixo os dados são copiados para GPU */ thrust::device_vector<double> dev(host); double val = thrust::reduce(dev.begin(), dev.end()); double mean = val/2517; std::cout << "Média: " << mean << "\n"; double val2 = thrust::reduce(dev.begin()+2152, dev.end()); double mean365 = val2/365; std::cout << "Média último ano: " << mean365 << "\n"; // auto result = thrust::minmax_element(dev.begin(), dev.end()); auto result = thrust::minmax_element(dev.begin(), dev.end()); auto result2 = thrust::minmax_element(dev.end() - 365, dev.end()); std::cout << "Mínimo 10 anos: " << *(result.first) << "\n"; std::cout << "Máximo 10 anos: " << *(result.second) << "\n"; std::cout << "Mínimo último ano: " << *(result2.first) << "\n"; std::cout << "Máximo último ano: " << *(result2.second)<< "\n"; // printf("Host vector: "); // for (auto i = host.begin(); i != host.end(); i++) { // std::cout << *i << " "; // este acesso é rápido -- CPU // } // printf("\n"); // printf("Device vector: "); // for (auto i = dev.begin(); i != dev.end(); i++) { // std::cout << *i << " "; // este acesso é lento! -- GPU // } // printf("\n"); }
19,750
#define t_max 1 #define t 1 /* (u[0][0][0][1][0]=(a*((((u[-3][0][0][0][0]+(u[0][-3][0][0][0]+u[0][0][-3][0][0]))*-2.0)+(((u[-2][0][0][0][0]+(u[0][-2][0][0][0]+u[0][0][-2][0][0]))*15.0)+((u[-1][0][0][0][0]+(u[0][-1][0][0][0]+u[0][0][-1][0][0]))*-60.0)))+((u[0][0][0][0][0]*20.0)+(((u[1][0][0][0][0]+(u[0][1][0][0][0]+u[0][0][1][0][0]))*30.0)+((u[2][0][0][0][0]+(u[0][2][0][0][0]+u[0][0][2][0][0]))*-3.0)))))) */ __global__ void upstream_5_3d(double * * u_0_1_out, double * u_0_0, double * u_0_1, double a, int x_max, int y_max, int z_max, int tbx, int tby, int tbz, int c) { //double * const u__u_0[16] = { u_0_0, u_0_1 } ; int _idx0; int _idx1; int _idx10; int _idx11; int _idx12; int _idx13; int _idx14; int _idx15; int _idx2; int _idx3; int _idx4; int _idx5; int _idx6; int _idx7; int _idx8; int _idx9; int chunk_idx_x; int chunk_idx_x_max; int chunk_idx_y; int chunk_idx_y_max; int chunk_idx_z; int chunk_idx_z_max; int idx_1_2; int size_1_1; int size_1_2; //int t; int thd_idx_x; int thd_idx_y; int thd_idx_z; int thdblks_idx_x; int thdblks_idx_x_max; int thdblks_idx_y; int thdblks_idx_y_max; int thdblks_idx_z; int thdblks_idx_z_max; int tmp; /* Initializations */ size_1_1=(y_max/blockDim.y); size_1_2=(z_max/blockDim.z); idx_1_2=(blockIdx.y/size_1_2); tmp=(blockIdx.y-(idx_1_2*size_1_2)); chunk_idx_x=(c*(threadIdx.x+(blockDim.x*blockIdx.x))); chunk_idx_x_max=(chunk_idx_x+c); chunk_idx_y=(threadIdx.y+(tmp*blockDim.y)); chunk_idx_y_max=(chunk_idx_y+1); chunk_idx_z=(threadIdx.z+(idx_1_2*blockDim.z)); chunk_idx_z_max=(chunk_idx_z+1); thdblks_idx_x=(tbx*(threadIdx.x+(blockDim.x*blockIdx.x))); thdblks_idx_x_max=(thdblks_idx_x+tbx); thdblks_idx_y=(tby*(threadIdx.y+(tmp*blockDim.y))); thdblks_idx_y_max=(thdblks_idx_y+tby); thdblks_idx_z=(tbz*(threadIdx.z+(idx_1_2*blockDim.z))); thdblks_idx_z_max=(thdblks_idx_z+tbz); /* Implementation */ /* for t = 1..t_max by 1 parallel 1 <level 0> schedule { ... } */ //for (t=1; t<=t_max; t+=1) { /* Index bounds calculations for iterators in thdblks[t=t, s=(tbx, tby, tbz)][0] */ /* Index bounds calculations for iterators in chunk[t=t, s=(c, 1, 1)][0] */ /* for POINT thd[t=t, s=(1, 1, 1)][0] of size [1, 1, 1] in chunk[t=t, s=(:, :, :)][0] parallel 1 <level 2> schedule default { ... } */ { /* Index bounds calculations for iterators in thd[t=t, s=(1, 1, 1)][0] */ thd_idx_z=chunk_idx_z; thd_idx_y=chunk_idx_y; for (thd_idx_x=chunk_idx_x; thd_idx_x<(chunk_idx_x_max-0); thd_idx_x+=1) { /* Index bounds calculations for iterators in thd[t=t, s=(1, 1, 1)][0] */ /* u[t=(t+1), s=thd[t=?, s=?][0]][0]=stencil(u[t=t, s=thd[t=?, s=?][0]][0]) */ /* _idx0 = (((((((((((thd_idx_z+3)*x_max)+((5*t)*thd_idx_z))+(15*t))*y_max)+((((((5*t)*thd_idx_z)+thd_idx_y)+(15*t))+3)*x_max))+((25*(t*t))*thd_idx_z))+((5*t)*thd_idx_y))+thd_idx_x)+(75*(t*t)))+(15*t)) */ _idx0=(((((((((((thd_idx_z+3)*x_max)+((5*t)*thd_idx_z))+(15*t))*y_max)+((((((5*t)*thd_idx_z)+thd_idx_y)+(15*t))+3)*x_max))+((25*(t*t))*thd_idx_z))+((5*t)*thd_idx_y))+thd_idx_x)+(75*(t*t)))+(15*t)); /* _idx1 = (((((((((((thd_idx_z+3)*x_max)+((5*t)*thd_idx_z))+(15*t))*y_max)+(((((5*t)*thd_idx_z)+thd_idx_y)+(15*t))*x_max))+((25*(t*t))*thd_idx_z))+((5*t)*thd_idx_y))+thd_idx_x)+(75*(t*t)))+3) */ _idx1=(((_idx0-(3*x_max))-(15*t))+3); /* _idx2 = (((((((((thd_idx_z*x_max)+((5*t)*thd_idx_z))*y_max)+(((((5*t)*thd_idx_z)+thd_idx_y)+3)*x_max))+((25*(t*t))*thd_idx_z))+((5*t)*thd_idx_y))+thd_idx_x)+(15*t))+3) */ _idx2=((((_idx0+(((-3*x_max)-(15*t))*y_max))-((15*t)*x_max))-(75*(t*t)))+3); /* _idx3 = ((((((((((((thd_idx_z+3)*x_max)+((5*t)*thd_idx_z))+(15*t))*y_max)+((((((5*t)*thd_idx_z)+thd_idx_y)+(15*t))+3)*x_max))+((25*(t*t))*thd_idx_z))+((5*t)*thd_idx_y))+thd_idx_x)+(75*(t*t)))+(15*t))+1) */ _idx3=(_idx0+1); /* _idx4 = ((((((((((((thd_idx_z+3)*x_max)+((5*t)*thd_idx_z))+(15*t))*y_max)+((((((5*t)*thd_idx_z)+thd_idx_y)+(15*t))+1)*x_max))+((25*(t*t))*thd_idx_z))+((5*t)*thd_idx_y))+thd_idx_x)+(75*(t*t)))+(5*t))+3) */ _idx4=((_idx1+x_max)+(5*t)); /* _idx5 = ((((((((((((thd_idx_z+1)*x_max)+((5*t)*thd_idx_z))+(5*t))*y_max)+((((((5*t)*thd_idx_z)+thd_idx_y)+(5*t))+3)*x_max))+((25*(t*t))*thd_idx_z))+((5*t)*thd_idx_y))+thd_idx_x)+(25*(t*t)))+(15*t))+3) */ _idx5=(((_idx2+((x_max+(5*t))*y_max))+((5*t)*x_max))+(25*(t*t))); /* _idx6 = ((((((((((((thd_idx_z+3)*x_max)+((5*t)*thd_idx_z))+(15*t))*y_max)+((((((5*t)*thd_idx_z)+thd_idx_y)+(15*t))+3)*x_max))+((25*(t*t))*thd_idx_z))+((5*t)*thd_idx_y))+thd_idx_x)+(75*(t*t)))+(15*t))+2) */ _idx6=(_idx3+1); /* _idx7 = ((((((((((((thd_idx_z+3)*x_max)+((5*t)*thd_idx_z))+(15*t))*y_max)+((((((5*t)*thd_idx_z)+thd_idx_y)+(15*t))+2)*x_max))+((25*(t*t))*thd_idx_z))+((5*t)*thd_idx_y))+thd_idx_x)+(75*(t*t)))+(10*t))+3) */ _idx7=((_idx4+x_max)+(5*t)); /* _idx8 = ((((((((((((thd_idx_z+2)*x_max)+((5*t)*thd_idx_z))+(10*t))*y_max)+((((((5*t)*thd_idx_z)+thd_idx_y)+(10*t))+3)*x_max))+((25*(t*t))*thd_idx_z))+((5*t)*thd_idx_y))+thd_idx_x)+(50*(t*t)))+(15*t))+3) */ _idx8=(((_idx5+((x_max+(5*t))*y_max))+((5*t)*x_max))+(25*(t*t))); /* _idx9 = ((((((((((((thd_idx_z+3)*x_max)+((5*t)*thd_idx_z))+(15*t))*y_max)+((((((5*t)*thd_idx_z)+thd_idx_y)+(15*t))+3)*x_max))+((25*(t*t))*thd_idx_z))+((5*t)*thd_idx_y))+thd_idx_x)+(75*(t*t)))+(15*t))+3) */ _idx9=(_idx3+2); /* _idx10 = ((((((((((((thd_idx_z+3)*x_max)+((5*t)*thd_idx_z))+(15*t))*y_max)+((((((5*t)*thd_idx_z)+thd_idx_y)+(15*t))+3)*x_max))+((25*(t*t))*thd_idx_z))+((5*t)*thd_idx_y))+thd_idx_x)+(75*(t*t)))+(15*t))+4) */ _idx10=(_idx3+3); /* _idx11 = ((((((((((((thd_idx_z+3)*x_max)+((5*t)*thd_idx_z))+(15*t))*y_max)+((((((5*t)*thd_idx_z)+thd_idx_y)+(15*t))+4)*x_max))+((25*(t*t))*thd_idx_z))+((5*t)*thd_idx_y))+thd_idx_x)+(75*(t*t)))+(20*t))+3) */ _idx11=((_idx9+x_max)+(5*t)); /* _idx12 = ((((((((((((thd_idx_z+4)*x_max)+((5*t)*thd_idx_z))+(20*t))*y_max)+((((((5*t)*thd_idx_z)+thd_idx_y)+(20*t))+3)*x_max))+((25*(t*t))*thd_idx_z))+((5*t)*thd_idx_y))+thd_idx_x)+(100*(t*t)))+(15*t))+3) */ _idx12=(((_idx9+((x_max+(5*t))*y_max))+((5*t)*x_max))+(25*(t*t))); /* _idx13 = ((((((((((((thd_idx_z+3)*x_max)+((5*t)*thd_idx_z))+(15*t))*y_max)+((((((5*t)*thd_idx_z)+thd_idx_y)+(15*t))+3)*x_max))+((25*(t*t))*thd_idx_z))+((5*t)*thd_idx_y))+thd_idx_x)+(75*(t*t)))+(15*t))+5) */ _idx13=(_idx3+4); /* _idx14 = ((((((((((((thd_idx_z+3)*x_max)+((5*t)*thd_idx_z))+(15*t))*y_max)+((((((5*t)*thd_idx_z)+thd_idx_y)+(15*t))+5)*x_max))+((25*(t*t))*thd_idx_z))+((5*t)*thd_idx_y))+thd_idx_x)+(75*(t*t)))+(25*t))+3) */ _idx14=((_idx11+x_max)+(5*t)); /* _idx15 = ((((((((((((thd_idx_z+5)*x_max)+((5*t)*thd_idx_z))+(25*t))*y_max)+((((((5*t)*thd_idx_z)+thd_idx_y)+(25*t))+3)*x_max))+((25*(t*t))*thd_idx_z))+((5*t)*thd_idx_y))+thd_idx_x)+(125*(t*t)))+(15*t))+3) */ _idx15=(((_idx12+((x_max+(5*t))*y_max))+((5*t)*x_max))+(25*(t*t))); u_0_1[_idx9]=(a*((((u_0_0[_idx0]+(u_0_0[_idx1]+u_0_0[_idx2]))*-2.0)+(((u_0_0[_idx3]+(u_0_0[_idx4]+u_0_0[_idx5]))*15.0)+((u_0_0[_idx6]+(u_0_0[_idx7]+u_0_0[_idx8]))*-60.0)))+((u_0_0[_idx9]*20.0)+(((u_0_0[_idx10]+(u_0_0[_idx11]+u_0_0[_idx12]))*30.0)+((u_0_0[_idx13]+(u_0_0[_idx14]+u_0_0[_idx15]))*-3.0))))); } } } } __global__ void initialize(double * u_0_0, double * u_0_1, double a, int x_max, int y_max, int z_max, int tbx, int tby, int tbz, int c) { double * const u__u_0[16] = { u_0_0, u_0_1 } ; int _idx0; int _idx1; int _idx10; int _idx11; int _idx12; int _idx13; int _idx14; int _idx15; int _idx2; int _idx3; int _idx4; int _idx5; int _idx6; int _idx7; int _idx8; int _idx9; int chunk_idx_x; int chunk_idx_x_max; int chunk_idx_y; int chunk_idx_y_max; int chunk_idx_z; int chunk_idx_z_max; int idx_1_2; int size_1_1; int size_1_2; //int t; int thd_idx_x; int thd_idx_y; int thd_idx_z; int thdblks_idx_x; int thdblks_idx_x_max; int thdblks_idx_y; int thdblks_idx_y_max; int thdblks_idx_z; int thdblks_idx_z_max; int tmp; /* Initializations */ size_1_1=(y_max/blockDim.y); size_1_2=(z_max/blockDim.z); idx_1_2=(blockIdx.y/size_1_2); tmp=(blockIdx.y-(idx_1_2*size_1_2)); chunk_idx_x=(c*(threadIdx.x+(blockDim.x*blockIdx.x))); chunk_idx_x_max=(chunk_idx_x+c); chunk_idx_y=(threadIdx.y+(tmp*blockDim.y)); chunk_idx_y_max=(chunk_idx_y+1); chunk_idx_z=(threadIdx.z+(idx_1_2*blockDim.z)); chunk_idx_z_max=(chunk_idx_z+1); thdblks_idx_x=(tbx*(threadIdx.x+(blockDim.x*blockIdx.x))); thdblks_idx_x_max=(thdblks_idx_x+tbx); thdblks_idx_y=(tby*(threadIdx.y+(tmp*blockDim.y))); thdblks_idx_y_max=(thdblks_idx_y+tby); thdblks_idx_z=(tbz*(threadIdx.z+(idx_1_2*blockDim.z))); thdblks_idx_z_max=(thdblks_idx_z+tbz); /* Implementation */ /* for t = 1..t_max by 1 parallel 1 <level 0> schedule { ... } */ //for (t=1; t<=t_max; t+=1) { /* Index bounds calculations for iterators in thdblks[t=t, s=(tbx, tby, tbz)][0] */ /* Index bounds calculations for iterators in chunk[t=t, s=(c, 1, 1)][0] */ /* for POINT thd[t=t, s=(1, 1, 1)][0] of size [1, 1, 1] in chunk[t=t, s=(:, :, :)][0] parallel 1 <level 2> schedule default { ... } */ { /* Index bounds calculations for iterators in thd[t=t, s=(1, 1, 1)][0] */ thd_idx_z=chunk_idx_z; thd_idx_y=chunk_idx_y; for (thd_idx_x=chunk_idx_x; thd_idx_x<(chunk_idx_x_max-0); thd_idx_x+=1) { /* Index bounds calculations for iterators in thd[t=t, s=(1, 1, 1)][0] */ /* u[t=(t+1), s=thd[t=?, s=?][0]][0]=stencil(u[t=t, s=thd[t=?, s=?][0]][0]) */ /* _idx0 = (((((((((((thd_idx_z+3)*x_max)+((5*t)*thd_idx_z))+(15*t))*y_max)+((((((5*t)*thd_idx_z)+thd_idx_y)+(15*t))+3)*x_max))+((25*(t*t))*thd_idx_z))+((5*t)*thd_idx_y))+thd_idx_x)+(75*(t*t)))+(15*t)) */ _idx0=(((((((((((thd_idx_z+3)*x_max)+((5*t)*thd_idx_z))+(15*t))*y_max)+((((((5*t)*thd_idx_z)+thd_idx_y)+(15*t))+3)*x_max))+((25*(t*t))*thd_idx_z))+((5*t)*thd_idx_y))+thd_idx_x)+(75*(t*t)))+(15*t)); u__u_0[(t-1)][_idx0]=0.1; /* _idx1 = ((((((((((((thd_idx_z+3)*x_max)+((5*t)*thd_idx_z))+(15*t))*y_max)+((((((5*t)*thd_idx_z)+thd_idx_y)+(15*t))+3)*x_max))+((25*(t*t))*thd_idx_z))+((5*t)*thd_idx_y))+thd_idx_x)+(75*(t*t)))+(15*t))+1) */ _idx1=(_idx0+1); u__u_0[(t-1)][_idx1]=0.1; /* _idx2 = ((((((((((((thd_idx_z+3)*x_max)+((5*t)*thd_idx_z))+(15*t))*y_max)+((((((5*t)*thd_idx_z)+thd_idx_y)+(15*t))+3)*x_max))+((25*(t*t))*thd_idx_z))+((5*t)*thd_idx_y))+thd_idx_x)+(75*(t*t)))+(15*t))+2) */ _idx2=(_idx1+1); u__u_0[(t-1)][_idx2]=0.1; /* _idx3 = (((((((((((thd_idx_z+3)*x_max)+((5*t)*thd_idx_z))+(15*t))*y_max)+(((((5*t)*thd_idx_z)+thd_idx_y)+(15*t))*x_max))+((25*(t*t))*thd_idx_z))+((5*t)*thd_idx_y))+thd_idx_x)+(75*(t*t)))+3) */ _idx3=(((_idx1-(3*x_max))-(15*t))+2); u__u_0[(t-1)][_idx3]=0.1; /* _idx4 = ((((((((((((thd_idx_z+3)*x_max)+((5*t)*thd_idx_z))+(15*t))*y_max)+((((((5*t)*thd_idx_z)+thd_idx_y)+(15*t))+1)*x_max))+((25*(t*t))*thd_idx_z))+((5*t)*thd_idx_y))+thd_idx_x)+(75*(t*t)))+(5*t))+3) */ _idx4=((_idx3+x_max)+(5*t)); u__u_0[(t-1)][_idx4]=0.1; /* _idx5 = ((((((((((((thd_idx_z+3)*x_max)+((5*t)*thd_idx_z))+(15*t))*y_max)+((((((5*t)*thd_idx_z)+thd_idx_y)+(15*t))+2)*x_max))+((25*(t*t))*thd_idx_z))+((5*t)*thd_idx_y))+thd_idx_x)+(75*(t*t)))+(10*t))+3) */ _idx5=((_idx4+x_max)+(5*t)); u__u_0[(t-1)][_idx5]=0.1; /* _idx6 = (((((((((thd_idx_z*x_max)+((5*t)*thd_idx_z))*y_max)+(((((5*t)*thd_idx_z)+thd_idx_y)+3)*x_max))+((25*(t*t))*thd_idx_z))+((5*t)*thd_idx_y))+thd_idx_x)+(15*t))+3) */ _idx6=((((_idx1+(((-3*x_max)-(15*t))*y_max))-((15*t)*x_max))-(75*(t*t)))+2); u__u_0[(t-1)][_idx6]=0.1; /* _idx7 = ((((((((((((thd_idx_z+1)*x_max)+((5*t)*thd_idx_z))+(5*t))*y_max)+((((((5*t)*thd_idx_z)+thd_idx_y)+(5*t))+3)*x_max))+((25*(t*t))*thd_idx_z))+((5*t)*thd_idx_y))+thd_idx_x)+(25*(t*t)))+(15*t))+3) */ _idx7=(((_idx6+((x_max+(5*t))*y_max))+((5*t)*x_max))+(25*(t*t))); u__u_0[(t-1)][_idx7]=0.1; /* _idx8 = ((((((((((((thd_idx_z+2)*x_max)+((5*t)*thd_idx_z))+(10*t))*y_max)+((((((5*t)*thd_idx_z)+thd_idx_y)+(10*t))+3)*x_max))+((25*(t*t))*thd_idx_z))+((5*t)*thd_idx_y))+thd_idx_x)+(50*(t*t)))+(15*t))+3) */ _idx8=(((_idx7+((x_max+(5*t))*y_max))+((5*t)*x_max))+(25*(t*t))); u__u_0[(t-1)][_idx8]=0.1; /* _idx9 = ((((((((((((thd_idx_z+3)*x_max)+((5*t)*thd_idx_z))+(15*t))*y_max)+((((((5*t)*thd_idx_z)+thd_idx_y)+(15*t))+3)*x_max))+((25*(t*t))*thd_idx_z))+((5*t)*thd_idx_y))+thd_idx_x)+(75*(t*t)))+(15*t))+3) */ _idx9=(_idx1+2); u__u_0[(t-1)][_idx9]=0.1; /* _idx10 = ((((((((((((thd_idx_z+4)*x_max)+((5*t)*thd_idx_z))+(20*t))*y_max)+((((((5*t)*thd_idx_z)+thd_idx_y)+(20*t))+3)*x_max))+((25*(t*t))*thd_idx_z))+((5*t)*thd_idx_y))+thd_idx_x)+(100*(t*t)))+(15*t))+3) */ _idx10=(((_idx9+((x_max+(5*t))*y_max))+((5*t)*x_max))+(25*(t*t))); u__u_0[(t-1)][_idx10]=0.1; /* _idx11 = ((((((((((((thd_idx_z+5)*x_max)+((5*t)*thd_idx_z))+(25*t))*y_max)+((((((5*t)*thd_idx_z)+thd_idx_y)+(25*t))+3)*x_max))+((25*(t*t))*thd_idx_z))+((5*t)*thd_idx_y))+thd_idx_x)+(125*(t*t)))+(15*t))+3) */ _idx11=(((_idx10+((x_max+(5*t))*y_max))+((5*t)*x_max))+(25*(t*t))); u__u_0[(t-1)][_idx11]=0.1; /* _idx12 = ((((((((((((thd_idx_z+3)*x_max)+((5*t)*thd_idx_z))+(15*t))*y_max)+((((((5*t)*thd_idx_z)+thd_idx_y)+(15*t))+4)*x_max))+((25*(t*t))*thd_idx_z))+((5*t)*thd_idx_y))+thd_idx_x)+(75*(t*t)))+(20*t))+3) */ _idx12=((_idx9+x_max)+(5*t)); u__u_0[(t-1)][_idx12]=0.1; /* _idx13 = ((((((((((((thd_idx_z+3)*x_max)+((5*t)*thd_idx_z))+(15*t))*y_max)+((((((5*t)*thd_idx_z)+thd_idx_y)+(15*t))+5)*x_max))+((25*(t*t))*thd_idx_z))+((5*t)*thd_idx_y))+thd_idx_x)+(75*(t*t)))+(25*t))+3) */ _idx13=((_idx12+x_max)+(5*t)); u__u_0[(t-1)][_idx13]=0.1; /* _idx14 = ((((((((((((thd_idx_z+3)*x_max)+((5*t)*thd_idx_z))+(15*t))*y_max)+((((((5*t)*thd_idx_z)+thd_idx_y)+(15*t))+3)*x_max))+((25*(t*t))*thd_idx_z))+((5*t)*thd_idx_y))+thd_idx_x)+(75*(t*t)))+(15*t))+4) */ _idx14=(_idx1+3); u__u_0[(t-1)][_idx14]=0.1; /* _idx15 = ((((((((((((thd_idx_z+3)*x_max)+((5*t)*thd_idx_z))+(15*t))*y_max)+((((((5*t)*thd_idx_z)+thd_idx_y)+(15*t))+3)*x_max))+((25*(t*t))*thd_idx_z))+((5*t)*thd_idx_y))+thd_idx_x)+(75*(t*t)))+(15*t))+5) */ _idx15=(_idx1+4); u__u_0[(t-1)][_idx15]=0.1; u__u_0[t][_idx9]=1.1; } } } }
19,751
#include "cuda_runtime.h" int main(){ }
19,752
#include <stdio.h> #define N 64 #define TPB 32 // A scaling function to convert integers 0,1,...,N-1 to evenly spaced floats __device__ float scale(int i, int n) { return ((float)i) / (n - 1); } // Compute the distance between 2 points on a line. __device__ float distance(float x1, float x2) { return sqrt((x2 - x1) * (x2 - x1)); } __global__ void distanceKernel(float* d_out, float ref, int len) { const int i = blockIdx.x * blockDim.x + threadIdx.x; const float x = scale(i, len); d_out[i] = distance(x, ref); printf("i = %2d: dist from %f to %f is %f.\n", i, ref, x, d_out[i]); } int main() { // Choose a reference value from which distances are measured. const float ref = 0.5; // Declare a pointer for an array of floats float* d_out = 0; // Allocate device memory for d_out cudaMalloc(&d_out, N * sizeof(float)); // Launch kernel to compute, NOTE: it is advicable to replace N/TPB with // (N+TPB-1)/TPB to make sure the number of blocks needed is rounded up. distanceKernel<<<N/TPB, TPB>>>(d_out, ref, N); // Wait for device to finish before exiting cudaDeviceSynchronize(); // Free the memory (Don't forget!!) cudaFree(d_out); return 0; }
19,753
#include "includes.h" __global__ void divideByCSCColSums(const float *values, const int *colPointers, float *pixels, const size_t n) { const size_t idx = blockIdx.x * blockDim.x + threadIdx.x; if (idx >= n) return; float weight = 0.0f; for (size_t ridx = colPointers[idx]; ridx < colPointers[idx+1]; ++ridx) { weight += values[ridx]; } pixels[idx] /= weight + 1e-6f; }
19,754
#include "includes.h" __global__ void sobelEdgeDetectionSharedMem2(int *input, int *output, int width, int height, int thresh) { int regArr[4][4]; int i = (blockIdx.x * blockDim.x + threadIdx.x) * 2; int j = (blockIdx.y * blockDim.y + threadIdx.y) * 2; if ( i > 0 && j > 0 && i < width - 1 && j < height - 1) { regArr[0][0] = input[width * (j-1) + i - 1]; regArr[0][1] = input[width * (j-1) + i ]; regArr[0][2] = input[width * (j-1) + i + 1]; regArr[0][3] = input[width * (j-1) + i + 2]; regArr[1][0] = input[width * (j) + i - 1]; regArr[1][1] = input[width * (j) + i ]; regArr[1][2] = input[width * (j) + i + 1]; regArr[1][3] = input[width * (j) + i + 2]; regArr[2][0] = input[width * (j+1) + i - 1]; regArr[2][1] = input[width * (j+1) + i ]; regArr[2][2] = input[width * (j+1) + i + 1]; regArr[2][3] = input[width * (j+1) + i + 2]; regArr[3][0] = input[width * (j+2) + i - 1]; regArr[3][1] = input[width * (j+2) + i ]; regArr[3][2] = input[width * (j+2) + i + 1]; regArr[3][3] = input[width * (j+2) + i + 2]; __syncthreads(); int sum1 = 0, sum2 = 0, magnitude; int num = 3; for(int xind = 1; xind < num; xind++) { for(int yind = 1; yind < num; yind++) { sum1 = regArr[xind+1][yind-1] - regArr[xind-1][yind-1] + 2 * regArr[xind+1][yind ] - 2 * regArr[xind-1][yind ] + regArr[xind+1][yind+1] - regArr[xind-1][yind+1]; sum2 = regArr[xind-1][yind-1] + 2 * regArr[xind][yind-1] + regArr[xind+1][yind-1] - regArr[xind-1][yind+1] - 2 * regArr[xind][yind+1] - regArr[xind+1][yind+1]; magnitude = sum1 * sum1 + sum2 * sum2; if(magnitude > thresh) output[(j + yind - 1) * width + (i + xind - 1)] = 255; else output[(j + yind - 1) * width + (i + xind - 1)] = 0; } } } }
19,755
#include <stdio.h> #include <stdlib.h> #include <iostream> __global__ void add_vec(int *a,int *b, int offset,int N) { int i = offset + threadIdx.x + blockIdx.x*blockDim.x; if(i < N){ a[i] = a[i] + b[i]; } } template <typename T> void fill_arr(T *data,T val,int N){ for(int i=0;i<N;++i){ data[i] = val; } } int main(){ int N = 1<<28; int *a,*b,*d_a,*d_b; int int_size = N * sizeof(int); //allocate host memory a = (int *)malloc(int_size); fill_arr(a,1,N); b = (int *)malloc(int_size); fill_arr(b,2,N); cudaMalloc((void **)&d_a,int_size); cudaMalloc((void **)&d_b,int_size); int threads = 256; int blocks = 16; cudaStream_t *i_stream; int num_stream = 16; int StreamSize = N/num_stream; size_t StreamBytes = StreamSize * sizeof(float); i_stream = (cudaStream_t*) new cudaStream_t[num_stream]; for (int i = 0; i < num_stream; i++){ cudaStreamCreate(&i_stream[i]); } for (int i = 0; i < num_stream; i++){ int offset = i * StreamSize; cudaMemcpyAsync(&d_a[offset],&a[offset],StreamBytes,cudaMemcpyHostToDevice,i_stream[i]); cudaMemcpyAsync(&d_b[offset],&b[offset],StreamBytes,cudaMemcpyHostToDevice,i_stream[i]); add_vec<<< threads, blocks, 0, i_stream[i] >>>(d_a,d_b,offset,N); cudaMemcpyAsync(&a[offset],&d_a[offset],StreamBytes,cudaMemcpyDeviceToHost,i_stream[i]); } for (int i = 0; i < num_stream; i++){ int offset = i * StreamSize; cudaMemcpyAsync(&a[offset],&d_a[offset],StreamBytes,cudaMemcpyDeviceToHost,i_stream[i]); } cudaDeviceSynchronize(); for (int i = 0; i < num_stream; i++){ cudaStreamDestroy(i_stream[i]); } delete [] i_stream; free(a); free(b); cudaFree(d_a); cudaFree(d_b); }
19,756
//#include <data_types/timeseries.hpp> //#include <data_types/fourierseries.hpp> //#include <data_types/candidates.hpp> //#include <data_types/filterbank.hpp> //#include <transforms/dedisperser.hpp> //#include <transforms/resampler.hpp> //#include <transforms/folder.hpp> //#include <transforms/ffter.hpp> //#include <transforms/dereddener.hpp> //#include <transforms/spectrumformer.hpp> //#include <transforms/birdiezapper.hpp> //#include <transforms/peakfinder.hpp> //#include <transforms/distiller.hpp> //#include <transforms/harmonicfolder.hpp> //#include <transforms/scorer.hpp> //#include <utils/exceptions.hpp> //#include <utils/utils.hpp> //#include <utils/stats.hpp> //#include <utils/stopwatch.hpp> //#include <utils/progress_bar.hpp> //#include <utils/cmdline.hpp> //#include <utils/output_stats.hpp> //#include <string> //#include <iostream> //#include <stdio.h> //#include <unistd.h> //#include "cuda.h" //#include "cufft.h" //#include "pthread.h" //#include <cmath> //#include <map> // // //int main4(int argc, char **argv) //{ // std::map<std::string,Stopwatch> timers; // timers["reading"] = Stopwatch(); // timers["dedispersion"] = Stopwatch(); // timers["searching"] = Stopwatch(); // timers["folding"] = Stopwatch(); // timers["total"] = Stopwatch(); // timers["total"].start(); // // CmdLineOptions args; // if (!read_cmdline_options(args,argc,argv)) // ErrorChecker::throw_error("Failed to parse command line arguments."); // // int nthreads = std::min(Utils::gpu_count(),args.max_num_threads); // nthreads = std::max(1,nthreads); // // if (args.verbose) // std::cout << "Using file: " << args.infilename << std::endl; // std::string filename(args.infilename); // // //Stopwatch timer; // if (args.progress_bar) // printf("Reading data from %s\n",args.infilename.c_str()); // // timers["reading"].start(); // SigprocFilterbank filobj(filename); // timers["reading"].stop(); // // if (args.progress_bar){ // printf("Complete (execution time %.2f s)\n",timers["reading"].getTime()); // } // // Dedisperser dedisperser(filobj,nthreads); // if (args.killfilename!=""){ // if (args.verbose) // std::cout << "Using killfile: " << args.killfilename << std::endl; // dedisperser.set_killmask(args.killfilename); // } // // if (args.verbose) // std::cout << "Generating DM list" << std::endl; // dedisperser.generate_dm_list(args.dm_start,args.dm_end,args.dm_pulse_width,args.dm_tol); // std::vector<float> dm_list = dedisperser.get_dm_list(); // // if (args.verbose){ // std::cout << dm_list.size() << " DM trials" << std::endl; // for (int ii=0;ii<dm_list.size();ii++) // std::cout << dm_list[ii] << std::endl; // std::cout << "Executing dedispersion" << std::endl; // } // // if (args.progress_bar) // printf("Starting dedispersion...\n"); // // timers["dedispersion"].start(); // PUSH_NVTX_RANGE("Dedisperse",3) // DispersionTrials<unsigned char> trials = dedisperser.dedisperse(); // POP_NVTX_RANGE // timers["dedispersion"].stop(); // // if (args.progress_bar) // printf("Complete (execution time %.2f s)\n",timers["dedispersion"].getTime()); // // unsigned int size; // if (args.size==0) // size = Utils::prev_power_of_two(filobj.get_nsamps()); // else // //size = std::min(args.size,filobj.get_nsamps()); // size = args.size; // if (args.verbose) // std::cout << "Setting transform length to " << size << " points" << std::endl; // // AccelerationPlan acc_plan(args.acc_start, args.acc_end, args.acc_tol, // args.acc_pulse_width, size, filobj.get_tsamp(), // filobj.get_cfreq(), filobj.get_foff()); // // // //Multithreading commands // timers["searching"].start(); // std::vector<Worker*> workers(nthreads); // std::vector<pthread_t> threads(nthreads); // DMDispenser dispenser(trials); // if (args.progress_bar) // dispenser.enable_progress_bar(); // // for (int ii=0;ii<nthreads;ii++){ // workers[ii] = (new Worker(trials,dispenser,acc_plan,args,size,ii)); // pthread_create(&threads[ii], NULL, launch_worker_thread, (void*) workers[ii]); // } // // DMDistiller dm_still(args.freq_tol,true); // HarmonicDistiller harm_still(args.freq_tol,args.max_harm,true,false); // CandidateCollection dm_cands; // for (int ii=0; ii<nthreads; ii++){ // pthread_join(threads[ii],NULL); // dm_cands.append(workers[ii]->dm_trial_cands.cands); // } // timers["searching"].stop(); // // if (args.verbose) // std::cout << "Distilling DMs" << std::endl; // dm_cands.cands = dm_still.distill(dm_cands.cands); // dm_cands.cands = harm_still.distill(dm_cands.cands); // // CandidateScorer cand_scorer(filobj.get_tsamp(),filobj.get_cfreq(), filobj.get_foff(), // fabs(filobj.get_foff())*filobj.get_nchans()); // cand_scorer.score_all(dm_cands.cands); // // if (args.verbose) // std::cout << "Setting up time series folder" << std::endl; // // MultiFolder folder(dm_cands.cands,trials); // timers["folding"].start(); // if (args.progress_bar) // folder.enable_progress_bar(); // // if (args.npdmp > 0){ // if (args.verbose) // std::cout << "Folding top "<< args.npdmp <<" cands" << std::endl; // folder.fold_n(args.npdmp); // } // timers["folding"].stop(); // // if (args.verbose) // std::cout << "Writing output files" << std::endl; // //dm_cands.write_candidate_file("./old_cands.txt"); // // int new_size = std::min(args.limit,(int) dm_cands.cands.size()); // dm_cands.cands.resize(new_size); // // CandidateFileWriter cand_files(args.outdir); // cand_files.write_binary(dm_cands.cands,"candidates.peasoup"); // // OutputFileWriter stats; // stats.add_misc_info(); // stats.add_header(filename); // stats.add_search_parameters(args); // stats.add_dm_list(dm_list); // // std::vector<float> acc_list; // acc_plan.generate_accel_list(0.0,acc_list); // stats.add_acc_list(acc_list); // // std::vector<int> device_idxs; // for (int device_idx=0;device_idx<nthreads;device_idx++) // device_idxs.push_back(device_idx); // stats.add_gpu_info(device_idxs); // stats.add_candidates(dm_cands.cands,cand_files.byte_mapping); // timers["total"].stop(); // stats.add_timing_info(timers); // // std::stringstream xml_filepath; // xml_filepath << args.outdir << "/" << "overview.xml"; // stats.to_file(xml_filepath.str()); // // return 0; //}
19,757
#include<stdio.h> #include<stdlib.h> #include<cuda.h> #define N 3 #define BLOCK_DIM 3 __global__ void matrixAdd(int *a,int *b,int *c) { int col=blockIdx.x*blockDim.x+threadIdx.x; int row=blockIdx.y*blockDim.y+threadIdx.y; int index=col+row*N; printf("\n%d\t%d",threadIdx.x,threadIdx.y); printf("\nIndex val:%d\n",index); if(col<N && row<N) { c[index]=a[index]+b[index]; } } int main() { int x[3][3]={1,2,3,4,5,6,7,8,9}; int y[3][3]={0,1,2,3,4,5,6,7,8}; int z[3][3]; int i=0,j=0; int *dev_a,*dev_b,*dev_c; cudaMalloc((void**)&dev_a,sizeof(x)); cudaMalloc((void**)&dev_b,sizeof(y)); cudaMalloc((void**)&dev_c,sizeof(z)); cudaMemcpy(dev_a,x,sizeof(x),cudaMemcpyHostToDevice); cudaMemcpy(dev_b,y,sizeof(y),cudaMemcpyHostToDevice); cudaMemcpy(dev_c,z,sizeof(z),cudaMemcpyHostToDevice); dim3 dimBlock(BLOCK_DIM,BLOCK_DIM ); dim3 dimGrid((int)ceil(N/dimBlock.x),(int)ceil(N/dimBlock.y)); matrixAdd<<<dimGrid,dimBlock>>>(dev_a,dev_b,dev_c); cudaMemcpy(z,dev_c,sizeof(z),cudaMemcpyDeviceToHost); printf("\noutput\n"); for(i=0;i<3;i++) { for(j=0;j<3;j++) { printf("\n%d",z[i][j]); } } cudaFree(dev_a); cudaFree(dev_b); cudaFree(dev_c); return 0; }
19,758
#include <stdio.h> #include <stdlib.h> #include <time.h> #include <inttypes.h> #include <math.h> #define gettime(t) clock_gettime(CLOCK_MONOTONIC_RAW, t) #define get_sub_seconde(t) (1e-9*(double)t.tv_nsec) /** return time in second */ __host__ double get_elapsedtime(void) { struct timespec st; int err = gettime(&st); if (err !=0) return 0; return (double)st.tv_sec + get_sub_seconde(st); } __host__ void init(double* A, double* B, double* C, int size) { int i = 0, j = 0; srand(2019); for(i = 0; i < size; i++) { for(j = 0; j < size; j++) { A[i * size + j] = rand(); B[i * size + j] = rand(); C[i * size + j] = 0.0; } } } void mult(double* A, double* B, double* C, int size) { int i = 0, j = 0, k = 0; for(i = 0; i < size; i++) { for(j = 0; j < size; j++) { double sum = 0.; for(k = 0; k < size; k++) { sum += A[i * size + k] * B[k * size + j]; } C[i * size + j] = sum; } } } __global__ void kernel(double* A, double* B, double* C, int N) { int x, y; x = blockIdx.x * blockDim.x + threadIdx.x; y = blockIdx.y * blockDim.y + threadIdx.y; int id = x * N + y; C[0] = 9999999; double sum = 0.; for (int i = 0; i < N; ++i) { sum += A[x * N + i] * B[i * N + y]; } C[id] = sum; } int main(int argc, char** argv){ int N = 0; double *A_h = NULL; double *B_h = NULL; double *C_h = NULL; //double t0 = 0., t1 = 0., duration = 0.; N = (argc < 2)?1000:atoi(argv[1]); fprintf(stdout, "Matrix Multiplication\n Size: %dx%d\n", N, N); dim3 DimBlock(32, 32, 1); dim3 DimGrid(ceil(N/32.), ceil(N/32.), 1); // Memory allocation A_h = (double*) malloc(sizeof(double) * N * N); B_h = (double*) malloc(sizeof(double) * N * N); C_h = (double*) malloc(sizeof(double) * N * N); // Value initialization init(A_h, B_h, C_h, N); //allocation des vecteurs sur le GPU avec gestion d'erreur : abort cudaError_t error; double *A_d = NULL; double *B_d = NULL; double *C_d = NULL; error = cudaMalloc((void**)&A_d, N*N*sizeof(double)); if(error != cudaSuccess) abort(); error = cudaMalloc((void**)&B_d, N*N*sizeof(double)); if(error != cudaSuccess) abort(); error = cudaMalloc((void**)&C_d, N*N*sizeof(double)); if(error != cudaSuccess) abort(); //transfert des données de host a device (avec gestion d'erreur, bien sur) error = cudaMemcpy(A_d, A_h, N*N*sizeof(double), cudaMemcpyHostToDevice); if(error != cudaSuccess) abort(); error = cudaMemcpy(B_d, B_h, N*N*sizeof(double), cudaMemcpyHostToDevice); if(error != cudaSuccess) abort(); error = cudaMemcpy(C_d, C_h, N*N*sizeof(double), cudaMemcpyHostToDevice); if(error != cudaSuccess) abort(); for(int i=0; i<N; i++) { printf("%lf ", C_h[i]); } float time; cudaEvent_t start, stop; cudaEventCreate(&start); cudaEventCreate(&stop); cudaEventRecord(start, 0); printf("%d %d %d\n%d %d %d", DimGrid.x, DimGrid.y, DimGrid.z, DimBlock.x, DimBlock.y, DimBlock.z); kernel<<<DimGrid, DimBlock>>>(A_d, B_d, C_d, N); cudaMemcpy(C_h, C_d, N*N*sizeof(double), cudaMemcpyDeviceToHost); cudaEventRecord(stop, 0); cudaEventSynchronize(stop); cudaEventElapsedTime(&time, start, stop); printf("Time to generate: %3.1f ms \n", time); cudaFree(A_d); cudaFree(B_d); cudaFree(C_d); for(int i = 0;i < N; i++) { printf("%lf ", C_h[i]); } // Compute multiplication //t0 = get_elapsedtime(); //mult(A_h, B_h, C_h, N); //t1 = get_elapsedtime(); // Pretty print //duration = (t1 - t0); //uint64_t nb_op = N * N * N; //fprintf(stdout, "Performance results: \n"); //fprintf(stdout, " Time: %lf s\n", duration); //fprintf(stdout, " MFlops: %.2f\n", (nb_op / duration)*1E-6); return 0; }
19,759
#include "includes.h" __global__ void differenceImg_gpu() { }
19,760
#include <stdio.h> __global__ void kernel0(void) { printf("kernel0\n"); } int main() { kernel0 <<<1,1>>> (); return 0; }
19,761
#include <stdio.h> #include <stdlib.h> #include <sys/time.h> #include <math.h> #include <cuda.h> // function to integrate #define F(x) (x*x) #define CUDA_CALL(F) if( (F) != cudaSuccess ) \ {printf("Error %s at %s:%d\n", cudaGetErrorString(cudaGetLastError()), __FILE__,__LINE__); exit(-1);} const long blocks_per_grid = 64; const long threads_per_block = 128; // Must be a power of 2 for reduction __device__ double rn(unsigned long * seed) { double ret; unsigned long n1; unsigned long a = 16807; unsigned long m = 2147483647; n1 = ( a * (*seed) ) % m; *seed = n1; ret = (double) n1 / m; return ret; } __global__ void lookup(double *F_vals, long F_len, double interval, long total_lookups, double *sums) { // A per-block cache. Each thread i writes to sum_cache[i] __shared__ double sum_cache[threads_per_block]; long i,j,k; double x, f; unsigned long seed; int thread_id = blockDim.x * blockIdx.x + threadIdx.x; int cache_id = threadIdx.x; seed = 10000*threadIdx.x + 10* blockIdx.x + threadIdx.x; for (i=thread_id; i < total_lookups; i += gridDim.x*blockDim.x) { // Randomly sample a continous value for x x = (double) rn(&seed); // Find the indices that bound x on the grid j = x / interval; k = j+1; // Calculate interpolation factor f = (k*interval - x) / (k*interval - j*interval); // Interpolate and accumulate result sum_cache[cache_id] += F_vals[j+1] - f * (F_vals[j+1] - F_vals[j]); } __syncthreads(); // Naive reduction for (i=blockDim.x/2; i != 0; i /= 2) { if (cache_id < i) sum_cache[cache_id] += sum_cache[cache_id + i]; __syncthreads(); } if (cache_id == 0) sums[blockIdx.x] = sum_cache[0]; } int main(int argc, char* argv[]) { // number of lookups long n_lookups = (argc < 2) ? 10000000 : atol(argv[1]); // number of gridpoints long F_len = (argc < 3) ? 250000000 : atol(argv[2]); // Discrete values for F(x) double *F_vals, *dev_F_vals; // interval for linearly-spaced grid double interval = (double) 1 / (F_len - 1); // Sum of random lookups on F_vals double sum = 0; // Vectors for sums of F(x_i). Dimensions will be sums[0:blocks_per_grid]. // Each block j will reduce is results to sum[i]. double *sums, *dev_sums; // Timing cudaEvent_t start, stop; float elapsed_time; // Loop control long i; printf("Running %0.2e lookups with %0.2e gridpoints in a %0.2f MB array...\n", (double) n_lookups, (double) F_len, (double) F_len*sizeof(double)/1e6); CUDA_CALL( cudaEventCreate( &start ) ); CUDA_CALL( cudaEventCreate( &stop ) ); sums = (double *) malloc( blocks_per_grid*sizeof(double) ); F_vals = (double *) malloc(F_len*sizeof(double)); // Populate values for F(x) on grid for (i=0; i<F_len; i++) { F_vals[i] = F(i*interval); } // Malloc and set dev_sums CUDA_CALL( cudaMalloc( (void**)&dev_sums, blocks_per_grid*sizeof(double) ) ); CUDA_CALL( cudaMemset( (void*) dev_sums, 0, blocks_per_grid*sizeof(double) ) ); // Malloc and copyto dev_F_vals CUDA_CALL( cudaMalloc( (void**)&dev_F_vals, F_len*sizeof(double)) ); CUDA_CALL( cudaMemcpy( dev_F_vals, F_vals, F_len*sizeof(double), cudaMemcpyHostToDevice ) ); CUDA_CALL( cudaEventRecord( start, 0 ) ); lookup<<<blocks_per_grid,threads_per_block>>>(dev_F_vals, F_len, interval, n_lookups, dev_sums); CUDA_CALL( cudaEventRecord( stop, 0 ) ); CUDA_CALL( cudaEventSynchronize( stop ) ); // Copy dev_sums to sums CUDA_CALL( cudaMemcpy( sums, dev_sums, blocks_per_grid*sizeof(double), cudaMemcpyDeviceToHost )); // Get cumulative sum for (i=0; i<blocks_per_grid; i++) { sum += sums[i]; } // Get timings CUDA_CALL( cudaEventElapsedTime( &elapsed_time, start, stop ) ); printf("Result: %0.6f\n", sum / n_lookups); printf("Time: %0.2e s\n", elapsed_time); printf("Rate: %0.2e lookups/s\n", n_lookups / elapsed_time); // Cleanup CUDA_CALL( cudaEventDestroy( start ) ); CUDA_CALL( cudaEventDestroy( stop ) ); CUDA_CALL( cudaFree( dev_F_vals ) ); CUDA_CALL( cudaFree( dev_sums ) ); free(F_vals); free(sums); return 0; }
19,762
#include <stdio.h> #include <cuda.h> __device__ int sumg = 0; __global__ void K(int num) { num += num; ++num; atomicAdd(&sumg, num); __shared__ int sum; sum = 0; __syncthreads(); sum += num; } int main() { for (unsigned ii = 0; ii < 100; ++ii) { K<<<5, 32>>>(ii); cudaDeviceSynchronize(); } return 0; }
19,763
#include <stdio.h> #include <stdlib.h> /** * =============== Comparação entre os tempos de execução: =============== * Sequencial: 1m18.116s * Paralelo: 0m25.027s * Paralelo (GPU - OpenMP): 0m15.604s * Paralelo (GPU - CUDA): 0m1.534s * * =============== Métricas relacionas as versões em GPU =============== * OpenMP: * warps_launched: 535592 * warp_execution_efficiency: 98.95% * * CUDA: * warps_launched: 127008 * warp_execution_efficiency: 99.21% */ __global__ void mm(double* a, double* b, double* c, int width) { int i = blockIdx.y * blockDim.y + threadIdx.y; int j = blockIdx.x * blockDim.x + threadIdx.x; if (i < width && j < width) { double sum = 0; for (int k = 0; k < width; k++) { double x = a[i * width + k]; double y = b[k * width + j]; sum += x * y; } c[i * width + j] = sum; } } int main() { int width = 2000; double *a = (double*) malloc (width * width * sizeof(double)); double *b = (double*) malloc (width * width * sizeof(double)); double *c = (double*) malloc (width * width * sizeof(double)); for(int i = 0; i < width; i++) { for(int j = 0; j < width; j++) { a[i * width + j] = i; b[i * width + j] = j; c[i * width + j] = 0; } } int size = width * width * sizeof(double); double *cuda_a, *cuda_b, *cuda_c; cudaMalloc((void **) &cuda_a, size); cudaMemcpy(cuda_a, a, size, cudaMemcpyHostToDevice); cudaMalloc((void **) &cuda_b, size); cudaMemcpy(cuda_b, b, size, cudaMemcpyHostToDevice); cudaMalloc((void **) &cuda_c, size); int block_size = 32; int dim = (width - 1) / block_size + 1; dim3 dimGrid(dim, dim, 1); dim3 dimBlock(block_size, block_size, 1); mm<<<dimGrid, dimBlock>>>(cuda_a, cuda_b, cuda_c, width); cudaMemcpy(c, cuda_c, size, cudaMemcpyDeviceToHost); // for(int i = 0; i < width; i++) { // for(int j = 0; j < width; j++) { // printf("\n c[%d][%d] = %f", i, j, c[i * width + j]); // } // } cudaFree(cuda_a); cudaFree(cuda_b); cudaFree(cuda_c); }
19,764
#include "includes.h" __global__ void exclusive_scan(unsigned int *in,unsigned int *out, int n) { unsigned int i = blockIdx.x * blockDim.x + threadIdx.x; if (i < n) { out[i] -= in[i]; } }
19,765
__global__ void mat_dot(float *a, float *b, float *c, int a_rows, int a_columns, int b_rows, int b_columns) { const int i = blockDim.y * blockIdx.y + threadIdx.y, j = blockDim.x * blockIdx.x + threadIdx.x; if (i < a_rows && j < b_columns) { float c_at_ij = 0; for (int k = 0; k < a_columns; k++) c_at_ij += a[i * a_columns + k] * b[k * b_columns + j]; c[i * b_columns + j] = c_at_ij; } }
19,766
#include <stdio.h> #include <stdlib.h> #include <cuda_runtime_api.h> #define BASE_TYPE float __global__ void mult(const BASE_TYPE *a, const BASE_TYPE *b, BASE_TYPE *c, const int N, const int M) { int i = N * (blockDim.y * blockIdx.y + threadIdx.y); int j = blockDim.x * blockIdx.x + threadIdx.x; BASE_TYPE sum = 0; for (int k = 0; k < N; k++) { sum += a[i + k] * b[k * M + j]; } int id = M * (blockDim.y * blockIdx.y + threadIdx.y) + blockDim.x * blockIdx.x + threadIdx.x; c[id] = sum; } BASE_TYPE* gen_array(const int N) { BASE_TYPE *a = new BASE_TYPE[N * N]; for (int i = 0; i < N; i++) { for (int j = 0; j < N; j++) a[i * N + j] = i * N + j; } return a; } void print_array(BASE_TYPE *a, const int N) { for (int i = 0; i < N; i++) { for (int j = 0; j < N; j++) printf("%5.0f ", a[i *N + j]); printf("\n"); } printf("\n"); } void cuda_init_array(BASE_TYPE **dev, const BASE_TYPE *host, const size_t size) { cudaError_t err; err = cudaMalloc((void **)dev, size); if (err != cudaSuccess) throw err; if (host != NULL) { err = cudaMemcpy(*dev, host, size, cudaMemcpyHostToDevice); if (err != cudaSuccess) throw err; } } void cuda_init_grid_and_block(dim3 *grid, dim3 *block, const int N) { *grid = dim3(1); *block = dim3(N, N, 1); printf("Block %d %d %d\n", block->x, block->y, block->z); printf("Grid %d %d %d\n", grid->x, grid->y, grid->z); } int main() { const int N = 10; const size_t size = N * N * sizeof(BASE_TYPE); cudaError_t err; dim3 threadsPerBlock, blocksPerGrid; cuda_init_grid_and_block(&blocksPerGrid, &threadsPerBlock, N); BASE_TYPE *host_a = gen_array(N), *host_b = gen_array(N); BASE_TYPE *dev_a, *dev_b, *dev_c; if (host_a == NULL) { fprintf(stderr, "Failed to allocate host vectors!\n"); exit(EXIT_FAILURE); } print_array(host_a, N); print_array(host_b, N); try { cuda_init_array(&dev_a, host_a, size); cuda_init_array(&dev_b, host_b, size); cuda_init_array(&dev_c, NULL, size); } catch (cudaError_t err) { fprintf(stderr, "Failed to allocate device (error code: %s)!\n", cudaGetErrorString(err)); exit(EXIT_FAILURE); } mult<<<blocksPerGrid, threadsPerBlock>>>(dev_a, dev_b, dev_c, N, N); err = cudaMemcpy(host_a, dev_c, size, cudaMemcpyDeviceToHost); if (err != cudaSuccess) { fprintf(stderr, "Failed to allocate device (error code: %s)!\n", cudaGetErrorString(err)); exit(EXIT_FAILURE); } print_array(host_a, N); cudaFree(dev_a); cudaFree(dev_b); delete[] host_a; return 0; }
19,767
# include <stdlib.h> # include <cuda.h> #include<stdio.h> const int N = 1024; __global__ void f(long long int *dev_a) { unsigned int tid = threadIdx.x; long long int temp = dev_a[(tid+1)%N]; __syncthreads(); dev_a[tid] = temp; } int main(void) { long long int host_a[N]; long long int *dev_a; cudaMalloc((void**)&dev_a, N * sizeof(long long int)); for(int i = 0 ; i < N ; i++) { host_a[i] = i; } cudaMemcpy(dev_a, host_a, N * sizeof(long long int), cudaMemcpyHostToDevice); f<<<1, N>>>(dev_a); cudaMemcpy(host_a, dev_a, N * sizeof(long long int), cudaMemcpyDeviceToHost); for(int i = 0 ; i < N ; i++) { printf("%d ", host_a[i]); } printf("\n"); }
19,768
/* Daniel Parker * University of Reading * 215 Parallel algorithms for Bioinformatics * * random.cu - generate some random strings for testing */ #include <stdlib.h> #include <time.h> void generate(char string[], int len) { int i; for (i = 0; i < len; i++) { string[i] = 'a' + (rand() % 4 + 1); } }
19,769
// nnCount: B*M // nnDist: B*M*nnSample // Weight: B*M*nnSample __global__ void cal_weight(int B, int M, int nnSample, int weightType, float radius, const int* nnCount, const float* nnDist, float* Weight) { // get the neighbor indices for(int i=blockIdx.x;i<B;i+=gridDim.x) { for(int j=threadIdx.x;j<M;j+=blockDim.x) { int K = nnCount[i*M+j]; for(int k=0;k<K;k++) { float dist = max(nnDist[i*M*nnSample+j*nnSample+k],1e-15); if (weightType==0) { Weight[i*M*nnSample+j*nnSample+k] = float(1)/dist; // inverse sqrt distance } else { Weight[i*M*nnSample+j*nnSample+k] = max(0.0, 1 - dist/radius); // bilinear like } } } } } void buildSampleWeightLauncher(int B, int M, int nnSample, int weightType, float radius, const int* nnCount, const float* nnDist, float* Weight) { cal_weight<<<B,1024>>>(B, M, nnSample, weightType, radius, nnCount, nnDist, Weight); }
19,770
#include "includes.h" __global__ void combineSourceAndBackground ( const int nwl, const int n, const float scale, float *src, const float *bkg ) { int i = threadIdx.x + blockDim.x * blockIdx.x; int j = threadIdx.y + blockDim.y * blockIdx.y; if ( i < n && j < nwl ) { src[i+j*n] = src[i+j*n] + scale * bkg[i+j*n]; } }
19,771
#include <cuda.h> #include <stdio.h> #include <time.h> #include <iostream> #include <fstream> #include <string> #include <vector> //#define BLOCK_WIDTH 512 __global__ void printMatrix(float **d_matrix, int size) { int i = (blockIdx.x * blockDim.x) + threadIdx.x; int j = (blockIdx.y * blockDim.y) + threadIdx.y; if (i < size && i >= 0) { if (j < size + 1 && j >=0) { printf("i is %d, j is %d, %f \n", i, j, d_matrix[i][j]); } } } __global__ void changeFirstElementToOne(float **d_matrix, int pivot, int size, int firstElement) { int i = (blockIdx.x * blockDim.x) + threadIdx.x; int j = (blockIdx.y * blockDim.y) + threadIdx.y; if (i == pivot) { if (j >= pivot && j < size+1) { d_matrix[i][j] = d_matrix[i][j] / firstElement; } } } __global__ void eliminationKernel(float **d_matrix, int pivot, int size) { int i = (blockIdx.x * blockDim.x) + threadIdx.x; int j = (blockIdx.y * blockDim.y) + threadIdx.y; if (i > pivot && i < size) { if (j > pivot && j < size+1) { d_matrix[i][j] = d_matrix[i][j] - d_matrix[i][pivot] * d_matrix[pivot][j]; } } } __global__ void setPivotColumnToZero(float **d_matrix, int pivot, int size) { int i = (blockIdx.x * blockDim.x) + threadIdx.x; int j = (blockIdx.y * blockDim.y) + threadIdx.y; if (i > pivot && i < size) { if (j == pivot) { d_matrix[i][j] = 0.0; } } } __global__ void backSubstitution(float **d_matrix, int subLine, int size) { int i = (blockIdx.x * blockDim.x) + threadIdx.x; int j = (blockIdx.y * blockDim.y) + threadIdx.y; if (i < subLine && i >= 0) { if (j == size) { d_matrix[i][j] = d_matrix[i][j] - d_matrix[i][subLine] * d_matrix[subLine][size]; } } } __global__ void setSubColToZero(float **d_matrix, int subLine, int size) { int i = (blockIdx.x * blockDim.x) + threadIdx.x; int j = (blockIdx.y * blockDim.y) + threadIdx.y; if (i < subLine && i >= 0) { if (j == subLine) { d_matrix[i][j] = 0.0; } } } __global__ void writeToDResult(float **d_matrix, int size, float *d_result) { int i = (blockIdx.x * blockDim.x) + threadIdx.x; int j = (blockIdx.y * blockDim.y) + threadIdx.y; if (i < size && i >= 0) { if (j == size) { d_result[i] = d_matrix[i][j]; } } } int main(void) { // read in data std::vector<int> int_list; std::string line_; std::ifstream file_("test100.txt"); if (!file_) { std::cout << "Cannot open file.\n"; return 0; } int size; // size of the matrix and vectors file_ >> size; float **matrix; // matrix of the linear system matrix = new float*[size]; for (int i = 0; i < size; i++) { matrix[i] = new float[size+1]; } for (int i = 0; i < size; i++) { for (int j = 0; j < size + 1; j++) { file_ >> matrix[i][j]; } } // initialize variable float * result, * d_result; // result vector float **d_matrix; float **d_matrix_h; d_matrix_h = (float**)malloc(size * sizeof(float *)); // alloc space for device copies of a cudaMalloc((void **) &d_result, size * sizeof(float)); cudaMalloc((void **) &d_matrix, size * sizeof(float*)); // alloc space for host copies of result result = (float *)malloc(size * sizeof(float)); // copy from host to device for (int i = 0; i < size; i++) { cudaMalloc((void**)&(d_matrix_h[i]), (size+1) * sizeof(float)); cudaMemcpy(d_matrix_h[i], matrix[i], (size+1) * sizeof(float), cudaMemcpyHostToDevice); } cudaMemcpy(d_matrix, d_matrix_h, size * sizeof(float *), cudaMemcpyHostToDevice); dim3 threadsPerBlock(8, 8); dim3 numBlocks((size - 1 + threadsPerBlock.x)/threadsPerBlock.x, (size + 1 - 1 + threadsPerBlock.y)/threadsPerBlock.y); struct timespec cudalustart = {0,0}; //time of constructing GF struct timespec cudaluend = {0,0}; clock_gettime(CLOCK_REALTIME,&cudalustart); // gaussian elimination for (int i = 0; i < size; i++) { // i is the pivot line here. // change first element of the pivot line to 1 float firstElement; cudaMemcpy(&firstElement, &d_matrix_h[i][i], sizeof(float), cudaMemcpyDeviceToHost); // std::cout << firstElement << std::endl; changeFirstElementToOne<<<numBlocks, threadsPerBlock>>>(d_matrix, i, size, firstElement); // the line under pivot line minus pivot line eliminationKernel<<<numBlocks, threadsPerBlock>>>(d_matrix, i, size); setPivotColumnToZero<<<numBlocks, threadsPerBlock>>>(d_matrix, i, size); } // back substitution for (int i = size - 1; i > 0; i--) { // form the last line to first line // current line is i. every line i 's "b" backSubstitution<<<numBlocks, threadsPerBlock>>>(d_matrix, i, size); setSubColToZero<<<numBlocks, threadsPerBlock>>>(d_matrix, i, size); } // write result from d_matrix to d_result writeToDResult<<<numBlocks, threadsPerBlock>>>(d_matrix, size, d_result); // copy result back to host cudaMemcpy(result, d_result, size * sizeof(float), cudaMemcpyDeviceToHost); clock_gettime(CLOCK_REALTIME,&cudaluend); std::cout<<"The time is "<<(cudaluend.tv_sec-cudalustart.tv_sec)*1000+(cudaluend.tv_nsec-cudalustart.tv_nsec)/1000000<<"ms\n"; /* // print the result for (int x = 0; x < size; x++) { std::cout << result[x] << std::endl; } */ // clean up free(matrix); free(result); free(d_matrix_h); cudaFree(d_matrix); cudaFree(d_result); return 0; }
19,772
#include <stdio.h> #include <stdlib.h> #define DEBUG __global__ void convol2D (float *a, float *h, float *c, int a_rows, int a_cols, int h_rows, int h_cols) { //Calculating indices along x and y directions int index_x = blockIdx.x * blockDim.x + threadIdx.x; int index_y = blockIdx.y * blockDim.y + threadIdx.y; //Calculating Grid width in terms of number of threads int grid_width = gridDim.x * blockDim.x; int sum = 0; for(int i = 0; i<h_rows; i++) { for(int j = 0; j<h_cols;j++) { if((index_y-i)>=0 & (index_y-i)<=(a_rows-1) & (index_x-j)>=0 & (index_x-j)<=(a_cols-1)) sum = sum + (a[(index_y-i)*a_cols+(index_x-j)] * h[i*h_cols + j]); } } c[index_y*grid_width + index_x] = sum; } int main (int argc, char *argv[]) { if(argc != 2) { fprintf(stderr,"Incorrect arguments passed.\nUse \"./2dconvol.o <input_file>\"\n"); exit(1); } FILE *f; int a_cols = 1; int h_cols = 1; int a_rows = 0; int h_rows = 0; int c_cols = 0; int c_rows = 0; float *a_h = 0; float *a_d = 0; float *hinv_h = 0; float *hinv_d = 0; float *c_h = 0; float *c_d = 0; size_t a_size = 0; size_t h_size = 0; size_t c_size = 0; dim3 block_size; dim3 grid_size; int i=0,j=0; char junk,junk_old; //Opening File f = fopen(argv[1],"r"); //First pass to find out size of the matrices junk = fgetc(f); while (junk != EOF) { if(junk == '\n') { a_rows++; } else if(junk == 0x20 & a_rows == 0) { a_cols++; } junk_old = junk; junk = fgetc(f); if(junk == '\n' & junk == junk_old) { break; } } junk = fgetc(f); while (junk != EOF) { if(junk == '\n') { h_rows++; } else if(junk == 0x20 & h_rows == 0) { h_cols++; } junk = fgetc(f); } //Calculating op dimensions c_rows = a_rows + h_rows - 1; block_size.y = c_rows > 32 ? 32 : c_rows; c_cols = a_cols + h_cols - 1; block_size.x = c_cols > 16 ? 16 : c_cols; grid_size.y = (c_rows/32)+1; grid_size.x = (c_cols/16)+1; #ifdef DEBUG printf("Size of A: %dx%d\n",a_rows,a_cols); printf("Size of H: %dx%d\n",h_rows,h_cols); printf("Size of C: %dx%d\n",c_rows,c_cols); printf("Size of grid: %dx%d\n",grid_size.y,grid_size.x); printf("Size of block: %dx%d\n",block_size.y,block_size.x); #endif //Calculating the sizes of all the involved matrices a_size = a_rows * a_cols *sizeof(float); h_size = h_rows * h_cols *sizeof(float); c_size = c_rows * c_cols *sizeof(float); //Allocating memory on host a_h = (float *) malloc(a_size); hinv_h = (float *) malloc(h_size); c_h = (float *) malloc(c_size); //Rewinding file to read the actual data rewind(f); //Reading all the data matrices for(i = 0;i<a_rows;i++) { for (j = 0; j<a_cols;j++) fscanf(f,"%f",&a_h[i*a_cols + j]); } for(i = 0 ; i<h_rows;i++) { for (j = 0; j<h_cols ;j++) { fscanf(f,"%f",&hinv_h[i*h_cols + j]); } } #ifdef DEBUG for(i = 0;i<a_rows;i++) { for (j = 0; j<a_cols;j++) printf("%f ",a_h[i*a_cols + j]); printf("\n"); } for(i = 0;i<h_rows;i++) { for (j = 0; j<h_cols;j++) { printf("%f ",hinv_h[i*h_cols + j]); } printf("\n"); } printf("Completed Loading Matrices...\n"); #endif //cudaMalloc to allocate required matrices on the device cudaMalloc((void **)&a_d,a_size); cudaMalloc((void **)&hinv_d,h_size); cudaMalloc((void **)&c_d,c_size); //Copying input data from the Host to Device cudaMemcpy(a_d,a_h,a_size,cudaMemcpyHostToDevice); cudaMemcpy(hinv_d,hinv_h,h_size,cudaMemcpyHostToDevice); //Setting Op matrix to all zeros cudaMemset(c_d,0,c_size); //Convolution function convol2D<<<grid_size,block_size>>>(a_d,hinv_d,c_d,a_rows,a_cols,h_rows,h_cols); //Synchronize to wait for the kernel to complete exectution cudaThreadSynchronize(); //Copy the output matrix from the Device to host cudaMemcpy(c_h,c_d,c_size,cudaMemcpyDeviceToHost); //Print Output for(i=0;i<c_rows;i++) { for(j=0;j<c_cols;j++) { printf("%f ",c_h[i*c_cols + j]); } printf("\n"); } //Freeing all the allocated memory from the device cudaFree(a_d); cudaFree(hinv_d); cudaFree(c_d); //Freeing all the allocated memory from the host free(a_h); free(hinv_h); free(c_h); fclose(f); return 0; }
19,773
/* CPU Based Wallsolver nvcc wallsolverCPU.cu -o testCPU */ #include <stdio.h> #include <stdbool.h> #include <stdlib.h> #include <math.h> #include <sys/time.h> #define SPACE_LENGTH 5 // Spaces Size of rows / columns #define SPACE_WIDTH 5 #define NUM_SPACES 25 #define WALL_LENGTH 4 // Walls size of rows/colums #define WALL_WIDTH 4 #define NUM_WALLS 16 #define POSSIBLE_DIRECTIONS 4 // Possible directions for traversing/finding neighbors typedef enum wall { UP, DOWN, LEFT, RIGHT } wall; typedef enum status { UNEXPLORED, VISITED, EXPLORED } status; typedef struct space { bool up, down, left, right, start, finish; int parent; int distance; status state; } space; typedef struct nextSpace { int index; int distance; } nextSpace; typedef struct nextMove { int space; int playerScore; int oppScore; int wallIdx; wall newDir; } nextMove; /* These are for an old way of tracking time */ struct timezone Idunno; struct timeval startTime, endTime; /* set a checkpoint and show the (natural) running time in seconds */ double report_running_time() { long sec_diff, usec_diff; gettimeofday(&endTime, &Idunno); sec_diff = endTime.tv_sec - startTime.tv_sec; usec_diff= endTime.tv_usec-startTime.tv_usec; if(usec_diff < 0) { sec_diff --; usec_diff += 1000000; } printf("Running time for CPU version: %ld.%06ld\n", sec_diff, usec_diff); return (double)(sec_diff*1.0 + usec_diff/1000000.0); } bool checkWallCollisions(wall *walls, int idx) { /* Make sure no walls overlap For each wall, identify the neighboring walls if they exist determine if a neighbor caused a conflict Return TRUE if there is a collision */ int i = idx / WALL_LENGTH; int j = idx % WALL_WIDTH; bool colUp = false; bool colDown = false; bool colLeft = false; bool colRight = false; wall up, down, left, right; if (j < 4) { right = walls[idx + 1]; colRight = (walls[idx] == RIGHT) && (right == LEFT); } if (j > 0) { left = walls[idx - 1]; colLeft = (walls[idx] == LEFT) && (left == RIGHT); } if (i < 4) { down = walls[idx + WALL_WIDTH]; colDown = (walls[idx] == DOWN) && (down == UP); } if (i > 0) { up = walls[idx - WALL_LENGTH]; colUp = (walls[idx] == UP) && (up == DOWN); } // Returns true if there is a collision return (colUp || colDown || colLeft || colRight); } void generateBoard(space *board, wall *walls) { /* Generate the board For each wall, identify the board spaces that it effects Determine the effect of each affected space's mobility */ int numSpaces = SPACE_LENGTH * SPACE_WIDTH; for (int i = 0; i < WALL_WIDTH; i++) { for (int j = 0; j < WALL_LENGTH; j++) { int idx = (i * WALL_LENGTH) + j; //printf("Maze Generated: %d - %d\n", idx, walls[idx]); // Determine the 4 adjacent spaces to this wall int TL = idx + i; int TR = TL +1; int BL = TL + SPACE_LENGTH; int BR = BL +1; if (board[TL].right) board[TL].right = (walls[idx] != UP); if (board[TL].down) board[TL].down = (walls[idx] != LEFT); if (board[TR].left) board[TR].left = board[TL].right; if (board[TR].down) board[TR].down = (walls[idx] != RIGHT); if (board[BL].right) board[BL].right = (walls[idx] != DOWN); if (board[BL].up) board[BL].up = board[TL].down; if (board[BR].left) board[BR].left = board[BL].right; if (board[BR].up) board[BR].up = board[TR].down; } } board[0].start = true; board[numSpaces - 1].finish = true; } void boardInit(space *board) { printf("Initialize Board\n"); // Initialize the board, blank for (int i = 0; i < SPACE_LENGTH; i++) { for (int j = 0; j < SPACE_WIDTH; j++) { int idx = (i * SPACE_WIDTH) + j; //board[idx] = blankSpace; /* if (i == 0) board[idx].up = false; if (j == 0) board[idx].left = false; if (i == (SPACE_WIDTH - 1)) board[idx].down = false; if (j == (SPACE_LENGTH - 1)) board[idx].right = false; */ // Better to avoid divergence board[idx].up = (i != 0); board[idx].left = (j != 0); board[idx].down = (i != (SPACE_WIDTH - 1)); board[idx].right = (j != (SPACE_LENGTH - 1)); board[idx].start = false; board[idx].finish = false; board[idx].parent = -1; board[idx].distance = 0; board[idx].state = UNEXPLORED; } } } void generateWalls(wall *walls) { /* Randomly generate the walls for the board */ srand(1024); for (int i = 0; i < WALL_WIDTH; i++) { for (int j = 0; j < WALL_LENGTH; j++) { int idx = (i * WALL_LENGTH) + j; // == walls[i][j]; walls[idx] = (wall)(rand() % 4); printf("IDX %d - %d\n", idx, walls[idx]); } } // Check for any wall collisions and re-randomize if necessary for (int i = 0; i < WALL_LENGTH; i++) { for (int j = 0; j < WALL_WIDTH; j++) { int idx = (i * WALL_WIDTH) + j; while (checkWallCollisions(walls, idx)) { printf("IDX No Overlap: %d - %d\n", idx, walls[idx]); walls[idx] = (wall)(rand() % 4); } } } } //output board configuration by numbers void outputBoard(space *in) { for (int i = 0; i < SPACE_WIDTH; i++) { for (int j = 0; j < SPACE_LENGTH; j++) { int idx = (i * SPACE_WIDTH) + j; // == board[i][j]; printf("outputBoard: Space #: %d, UP: %d, DOWN: %d, LEFT: %d, RIGHT: %d \n", idx, in[idx].up, in[idx].down, in[idx].left, in[idx].right); } } } //display board configuration visually void displayBoard(space *in) { int bWidth = 1+2*SPACE_WIDTH; int bLength = 1+2*SPACE_LENGTH; char boardArray[bLength][bWidth]; printf("\n"); printf("Display Board:\n"); printf("(s = starting position; e = end goal)\n"); //initialize boardArray to blank spaces for (int i=0; i<bWidth; i++) { for (int j=0;j<bLength;j++) { boardArray[i][j] = ' '; } } //populate boardArray for (int i=0; i<SPACE_WIDTH; i++) { for (int j=0; j< SPACE_LENGTH; j++) { int idx = (i*SPACE_WIDTH)+j; if (in[idx].up==1) { boardArray[2*i][2*j+1] = '*'; boardArray[2*i][2*j+2] = '*'; } else boardArray[2*i][2*j+1] = ' '; if (in[idx].right == 1) { boardArray[2*i+1][2*j+2] = '*'; boardArray[2*i+2][2*j+2] = '*'; boardArray[2*i][2*j+2] = '*'; } else boardArray[2*i+1][2*j+2] = ' '; } } //set board vertical borders for (int i = 0; i<bWidth; i++) { boardArray[i][0] = '*'; boardArray[i][bWidth-1] = '*'; } //set board horizontal borders for (int i = 0; i< bWidth; i++) { for (int j=0; j<bLength; j++) { if (i == 0) //if it's the first row boardArray[i][j] = '*'; //set border if (i == bWidth-1) //if it's the last row boardArray[i][j] = '*'; //set border } } //set start and end points boardArray[1][1] = 's'; boardArray[bLength-2][bWidth-2] = 'e'; //print boardArray for (int i=0; i<bWidth; i++) { for (int j = 0; j<bLength; j++) { printf("%c", boardArray[i][j]); } printf("\n"); } printf("\n"); } void outputResults(nextMove *results, int numResults) { /* Output the best valued move for each space */ printf("----- RESULTS -----\n"); for (int i = 0; i < numResults; i++) { printf("Best Move for Space %d\n", results[i].space); printf("Move wall %d to direction %d\n", results[i].wallIdx, results[i].newDir); printf("Player Score: %d, Opponent Score: %d\n\n", results[i].playerScore, results[i].oppScore); } } /* Used strictly for debugging. This should not be used in the final version */ void printAdjList(int adjList[][POSSIBLE_DIRECTIONS]) { int i = 0; int numSpaces = SPACE_LENGTH * SPACE_WIDTH; for (i = 0; i < numSpaces; ++i) { printf("Space #%d's neighbors: UP: %d, DOWN: %d, LEFT: %d, RIGHT: %d \n", i, adjList[i][0], adjList[i][1], adjList[i][2], adjList[i][3]); } } /* Set all neighbors to -1, then cycle through and add neighbors for each space All spaces marked with -1 afterwards means the neighbor is invalid and can be ignored */ void initializeAdjList(int adjList[][POSSIBLE_DIRECTIONS]) { int i = 0; int numSpaces = SPACE_LENGTH * SPACE_WIDTH; for (i = 0; i < numSpaces; ++i) { int j; for (j = 0; j < POSSIBLE_DIRECTIONS; ++j) { adjList[i][j] = -1; } } for (i = 0; i < numSpaces; ++i) { // Add up neighbor to list if (i >= SPACE_WIDTH) adjList[i][0] = i - SPACE_LENGTH; // Add down neighbor to list if (i < (numSpaces - SPACE_WIDTH)) adjList[i][1] = i + SPACE_LENGTH; // Add left neighbor to list if (i % SPACE_WIDTH != 0) adjList[i][2] = i - 1; // Add right neighbor to list if (i % SPACE_WIDTH != (SPACE_WIDTH - 1)) adjList[i][3] = i + 1; } // printAdjList(adjList); } nextSpace findMinimum(space *in, int adjList[][POSSIBLE_DIRECTIONS], int idx) { int min = 9999; int min_idx = -1; int j; const int WALL_COST = 3; nextSpace next; // Find the best next step based on our index's neighbors. for (j = 0; j < POSSIBLE_DIRECTIONS; ++j) { if (adjList[idx][j] == -1 || in[adjList[idx][j]].state == VISITED) continue; if (j == 0) { if (in[idx].up && min > WALL_COST) { min = WALL_COST; min_idx = adjList[idx][j]; } else if (!in[idx].up && min > 1) { min = 1; min_idx = adjList[idx][j]; } } if (j == 1) { if (in[idx].down && WALL_COST <= min) { min = WALL_COST; min_idx = adjList[idx][j]; } else if (!in[idx].down && min > 1) { min = 1; min_idx = adjList[idx][j]; } } if (j == 2) { if (in[idx].left && min > WALL_COST) { min = WALL_COST; min_idx = adjList[idx][j]; } else if (!in[idx].left && min > 1) { min = 1; min_idx = adjList[idx][j]; } } if (j == 3) { if (in[idx].right && min > WALL_COST) { min = WALL_COST; min_idx = adjList[idx][j]; } else if (!in[idx].right && min > 1) { min = 1; min_idx = adjList[idx][j]; } } } next.index = min_idx; next.distance = min; return next; } void resetSpaces(space *in) { int i; int numSpaces = SPACE_LENGTH * SPACE_WIDTH; for (i = 0; i < numSpaces; ++i) { in[i].parent = -1; in[i].state = UNEXPLORED; } return; } int shortestPath(space *in, int idxIn = 0) { int adjList[SPACE_LENGTH*SPACE_WIDTH][POSSIBLE_DIRECTIONS]; initializeAdjList(adjList); int i = idxIn; nextSpace next; int distance; // If shortestPath is used multiple times then we need to reset the parent & state. resetSpaces(in); // Iterate through the board until we reach the finish node. while (!in[i].finish) { // Run greedy shortest path on all of the current space's neighbors. in[i].state = VISITED; int tmp = i; next = findMinimum(in, adjList, i); i = next.index; if (i == -1) { i = in[tmp].parent; } else { in[i].parent = tmp; in[i].distance = in[in[i].parent].distance + next.distance; } } distance = in[i].distance; printf("Total distance: %d\n", distance); while (!in[i].start) { printf("Space #%d\n", i); i = in[i].parent; } printf("Space #%d\n", i); return distance; } /* */ void moveWall(wall *in, int wallIdx, wall newDir, nextMove *move, int moveIdx, int oppPos) { wall oldDir = in[wallIdx]; // Temp store the old wall direction bool sameDir = in[wallIdx] == newDir; // Set the walls[idx] to the new direction in[wallIdx] = newDir; bool collision = checkWallCollisions(in, wallIdx); // If same direction or a collision // Reset to old direction and return -1 if (sameDir || collision) { in[wallIdx] = oldDir; return; } printf("No collision - Space: %d, WallID: %d\n", move[moveIdx].space, wallIdx); space *board = (space *)malloc(sizeof(space) * NUM_SPACES); boardInit(board); generateBoard(board, in); int playerScore = 0; playerScore = shortestPath(board, move[moveIdx].space); int oppScore = 0; oppScore = shortestPath(board, oppPos); printf("Space: %d, WallID: %d, Wall New Dir: %d, Player Score: %d, Opp Score: %d\n", move[moveIdx].space, wallIdx, newDir, playerScore, oppScore); if (playerScore < move[moveIdx].playerScore || oppScore > move[moveIdx].oppScore) { move[moveIdx].playerScore = playerScore; move[moveIdx].oppScore = oppScore; move[moveIdx].wallIdx = wallIdx; move[moveIdx].newDir = newDir; } // Reset to the old direction in[wallIdx] = oldDir; free(board); } void moveAllWalls(wall *walls, int oppIdx, nextMove *results, int resultsIdx) { /* For all walls, orient them each possible direction Check for collisions or if same direction If not, see if it changes the player or opponent's shortest path */ for (int i = 0; i < (WALL_LENGTH * WALL_WIDTH); i++) { for (int j = 0; j < 3; j++) { // Check player (walls, wallID, direction, playerSpace, oppSpace) moveWall(walls, i, (wall) j, results, resultsIdx, oppIdx); } } } /* *********************************************************** *** Inputs: space *in (our board) * *** idx (the space # we're finding neighbors for) * *********************************************************** *** Output: 12-element int array with indexes to all * *** potential neighbors. -1 is entered for any * *** elements that are unnecessary. * *********************************************************** *** Purpose: Find all possible neighbors a space can * *** visit on the current turn. If a wall is * *** encountered after a step is taken then the * *** adjacent neighbors are added to the array. * *********************************************************** */ int* findNeighbors(space *in, int idx) { int const NUM_NEIGHBORS = 12; int *neighbors = (int *)malloc(NUM_NEIGHBORS*sizeof(int)); int const MAX_DISTANCE = 3; int numSpaces = SPACE_LENGTH * SPACE_WIDTH; int count = 0; int i = idx; int neighborIdx = 0; int const WALL_COST = 3; // Grab neighbors going up while (count < MAX_DISTANCE) { // Don't gather upward neighbors if index is in the top row. if (idx < SPACE_WIDTH) break; i -= SPACE_WIDTH; if (i >= numSpaces || i < 0) { break; } else if (in[i].down && count > 0) { // Get left neighbors if we're reached a wall i += SPACE_WIDTH; int tmp = i; int tmpCount = count; while (count < MAX_DISTANCE) { i--; if (in[i].right || (i % SPACE_WIDTH) == 0) { count += MAX_DISTANCE; break; } else { neighbors[neighborIdx++] = i; count++; } } // Get right neighbors if we've reached a wall i = tmp; count = tmpCount; while (count < MAX_DISTANCE) { i++; if (in[i].left || (i % SPACE_WIDTH == (SPACE_WIDTH - 1))) { count += MAX_DISTANCE; break; } else { neighbors[neighborIdx++] = i; count++; } } } else if (in[i].down && count == 0) { count += WALL_COST; neighbors[neighborIdx++] = i; } else { count++; neighbors[neighborIdx++] = i; } } // Grab neighbors going down count = 0; i = idx; while (count < MAX_DISTANCE) { // Don't grab downward neighbors if index is in the bottom row. if (i >= (numSpaces - SPACE_WIDTH)) break; i += SPACE_WIDTH; if (i >= numSpaces || i < 0) { break; } else if (in[i].up && count > 0) { // Get left neighbors if we're reached a wall i -= SPACE_WIDTH; int tmp = i; int tmpCount = count; while (count < MAX_DISTANCE) { i--; if (in[i].right || (i % SPACE_WIDTH) == 0) { count += MAX_DISTANCE; break; } else { neighbors[neighborIdx++] = i; count++; } } // Get right neighbors if we've reached a wall i = tmp; count = tmpCount; while (count < MAX_DISTANCE) { i++; if (in[i].left || (i % SPACE_WIDTH == (SPACE_WIDTH - 1))) { count += MAX_DISTANCE; break; } else { neighbors[neighborIdx++] = i; count++; } } } else if (in[i].up && count == 0) { count += WALL_COST; neighbors[neighborIdx++] = i; } else { count++; neighbors[neighborIdx++] = i; } } // Grab neighbors going left count = 0; i = idx; while (count < MAX_DISTANCE) { // Don't gather leftward neighbors if index is in the left column. if ((idx % SPACE_WIDTH == 0)) break; i--; if (i >= numSpaces || i < 0) { break; } else if (in[i].right && count > 0) { // Get up neighbors if we're reached a wall i++; int tmp = i; int tmpCount = count; while (count < MAX_DISTANCE) { i -= SPACE_WIDTH; if (i < 0 || in[i].down) { count += MAX_DISTANCE; break; } else { neighbors[neighborIdx++] = i; count++; } } // Get down neighbors if we've reached a wall i = tmp; count = tmpCount; while (count < MAX_DISTANCE) { i += SPACE_WIDTH; if (in[i].up || (i >= (numSpaces - SPACE_WIDTH))) { count += MAX_DISTANCE; break; } else { neighbors[neighborIdx++] = i; count++; } } } else if (in[i].right && count == 0) { count += WALL_COST; neighbors[neighborIdx++] = i; } else { count++; neighbors[neighborIdx++] = i; } } // Grab neighbors going right count = 0; i = idx; while (count < MAX_DISTANCE) { // Don't gather rightward neighbors if index is in the right column. if (idx % SPACE_WIDTH == (SPACE_WIDTH - 1)) break; i++; if (i >= numSpaces || i < 0) { break; } else if (in[i].left && count > 0) { // Get up neighbors if we're reached a wall i--; int tmp = i; int tmpCount = count; while (count < MAX_DISTANCE) { i -= SPACE_WIDTH; if (i < 0 || in[i].down) { count += MAX_DISTANCE; break; } else { neighbors[neighborIdx++] = i; count++; } } // Get down neighbors if we've reached a wall i = tmp; count = tmpCount; while (count < MAX_DISTANCE) { i += SPACE_WIDTH; if (in[i].up || (i >= (numSpaces - SPACE_WIDTH))) { count += MAX_DISTANCE; break; } else { neighbors[neighborIdx++] = i; count++; } } } else if (in[i].left && count == 0) { count += WALL_COST; neighbors[neighborIdx++] = i; } else { count++; neighbors[neighborIdx++] = i; } } // Set all unused spaces in our return array to -1 to indicate they don't exist for (i = neighborIdx; i < NUM_NEIGHBORS; ++i) neighbors[i] = -1; return neighbors; } nextMove *getPossibleMoves(space *board, int pos, int *numSpaces) { // Counter number of possible spaces @ pos int *neighbors = findNeighbors(board, pos); int possibleSpaces = 0; for(int i = 0; i < 12; i++) { printf("Neighbors for space %d: %d\n", pos, neighbors[i]); if (neighbors[i] != -1) possibleSpaces++; } nextMove *moves = (nextMove *) malloc( sizeof(nextMove) * possibleSpaces ); int j = 0; for(int i = 0; i < 12 && j < possibleSpaces; i++) { if (neighbors[i] == -1 ) break; moves[j].space = neighbors[i]; moves[j].playerScore = 100; // Intentionally high preset moves[j].oppScore = -1; moves[j].wallIdx = -1; moves[j].newDir = (wall) 0; j++; } free(neighbors); //numSpaces = possibleSpaces; return moves; } nextMove pickBestMove(nextMove *moves, int possibleSpaces) { nextMove best; if (possibleSpaces > 0) { best.space = moves[0].space; best.playerScore = moves[0].playerScore; best.oppScore = moves[0].oppScore; best.wallIdx = moves[0].wallIdx; best.newDir = moves[0].newDir; } int diff = best.oppScore - best.playerScore; int i; for (i = 0; i < possibleSpaces; ++i) { int tmpdiff = moves[i].oppScore - moves[i].playerScore; if (tmpdiff > diff) { best.space = moves[i].space; best.playerScore = moves[i].playerScore; best.oppScore = moves[i].oppScore; best.wallIdx = moves[i].wallIdx; best.newDir = moves[i].newDir; diff = tmpdiff; } } return best; } int main(int argc, char const *argv[]) { int playerPos = 0; int oppPos = 0; int numSpaces = SPACE_LENGTH * SPACE_WIDTH; int spaceSize = sizeof(space) * numSpaces; int numWalls = WALL_LENGTH * WALL_WIDTH; int wallSize = sizeof(wall) * numWalls; // Malloc the array of wall / board wall *walls = (wall *)malloc(wallSize); space *board = (space *)malloc(spaceSize); // Initialize, zero out the board boardInit(board); // Generate walls generateWalls(walls); generateBoard(board, walls); // Start the timer gettimeofday(&startTime, &Idunno); //display board outputBoard(board); //display by numbers displayBoard(board); //display visually // Testing //shortestPath(board, 0); //shortestPath(board, 0); // board, walls, playerIdx to be moved to, current opponent idx, results //moveAllWalls(board, walls, 0, 0, results); // Report the running time //report_running_time(); // Get neighbors of a space int *neighbors = findNeighbors(board, 7); int i; for (i = 0; i < 12; ++i) { printf("Neighbors for space #7: %d\n", neighbors[i]); } neighbors = findNeighbors(board, 17); for (i = 0; i < 12; ++i) { printf("Neighbors for space #17: %d\n", neighbors[i]); } neighbors = findNeighbors(board, playerPos); for (i = 0; i < 12; ++i) { printf("Neighbors for space #%d: %d\n", playerPos, neighbors[i]); } // Count the number of possible spaces = # of blocks int possibleSpaces = 0; for (int i = 0; i < 12; i++) { if (neighbors[i] != -1) { possibleSpaces++; } } // Malloc an array nextMove[ # of neighbors ] nextMove *moves = (nextMove *)malloc( sizeof(nextMove) * possibleSpaces ); printf("DEBUG: successful malloc of results\n"); // Zero-out the results array and set each move.space ot the neighbor space int j = 0; for (int i = 0; i < 12 && j < possibleSpaces; i++) { if (neighbors[i] != -1) { printf("Init results array. Moves[%d], Space: %d\n", j, neighbors[i]); moves[j].space = neighbors[i]; moves[j].playerScore = 100; // Intentionally high preset moves[j].oppScore = -1; moves[j].wallIdx = -1; moves[j].newDir = (wall) 0; j++; } } // TODO: DEBUG this function //int *possibleSpaces = getPossibleMoves(board, playerPos, possibleSpaces); /* start counting time */ gettimeofday(&startTime, &Idunno); // Set the nextSpace in the array to each part in array /* For each possible space --> Move all 16 walls all possible direction Determine the shortest path */ for (int i = 0; i < possibleSpaces; i++) { moveAllWalls(walls, oppPos, moves, i); } /* check the total running time */ report_running_time(); outputResults(moves, possibleSpaces); // PICK THE BEST MOVE nextMove bestMove = pickBestMove(moves, possibleSpaces); printf("Best Move: %d\n", bestMove.space); printf("Memory Sizes - walls: %d, board: %d, nextMove (maximum): %d, neighbors: %d\n", wallSize, spaceSize, sizeof(nextMove)*12, sizeof(neighbors)); free(walls); free(board); //free(possibleSpaces); free(neighbors); free(moves); return 0; }
19,774
#include "includes.h" #define DOUBLE #ifdef DOUBLE #define Complex cufftDoubleComplex #define Real double #define Transform CUFFT_Z2Z #define TransformExec cufftExecZ2Z #else #define Complex cufftComplex #define Real float #define Transform CUFFT_C2C #define TransformExec cufftExecC2C #endif #define TILE_DIM 8 // synchronize blocks __global__ void copy_0(int nx, int ny, int nz, Real *in, Real *out) { int tj = threadIdx.x; int td = blockDim.x; int jj = (blockIdx.y*nx*ny/4 + blockIdx.x*nx/2); int jj1 = ((blockIdx.y+nz/2)*nx*ny + (blockIdx.x+ny/2)*nx); out[jj1+tj+td] = in[jj+tj]; }
19,775
#ifdef _WIN32 # define EXPORT __declspec(dllexport) #else # define EXPORT #endif void __global__ file1_kernel(int x, int& r) { r = -x; } EXPORT int file1_launch_kernel(int x) { int r = 0; file1_kernel<<<1, 1>>>(x, r); return r; }
19,776
#include <cuda_runtime_api.h> #include <cuda.h> #include <stdio.h> #include <stdlib.h> #include <iostream> using namespace std; cudaStream_t createStreamWithFlags(){ cudaStream_t stream; cudaStreamCreateWithFlags(&stream, cudaStreamNonBlocking); return stream; }
19,777
#include<stdio.h> #include<stdlib.h> #include<math.h> #include "answer.cuh" // Compute vector sum C = A+B //CUDA kernel. Each thread performes one pair-wise addition __global__ void vector_add(float *a, float *b, float *c, int N) { /* insert code to calculate the index properly using blockIdx.x, blockDim.x, threadIdx.x */ int index = blockIdx.x * blockDim.x + threadIdx.x; int stride = blockDim.x * gridDim.x; for (int i = index; i < N; i += stride) c[i] = a[i] + b[i]; } /* experiment with N */ /* how large can it be? */ #define THREADS_PER_BLOCK 1000 void add(float *a, float *b, float *c, int N) { float *d_a, *d_b, *d_c; int size = N * sizeof( float ); /* allocate space for device copies of a, b, c */ cudaMalloc( (void **) &d_a, size ); cudaMalloc( (void **) &d_b, size ); cudaMalloc( (void **) &d_c, size ); /* copy inputs to device */ /* fix the parameters needed to copy data to the device */ cudaMemcpy( d_a, a, size, cudaMemcpyHostToDevice ); cudaMemcpy( d_b, b, size, cudaMemcpyHostToDevice ); /* launch the kernel on the GPU */ /* insert the launch parameters to launch the kernel properly using blocks and threads */ // vector_add<<< (N+THREADS_PER_BLOCK-1)/THREADS_PER_BLOCK, THREADS_PER_BLOCK >>>( d_a, d_b, d_c, N); vector_add<<<1, 1 >>>( d_a, d_b, d_c, N); //Synchronize threads cudaThreadSynchronize(); /* copy result back to host */ /* fix the parameters needed to copy data back to the host */ cudaMemcpy( c, d_c, size, cudaMemcpyDeviceToHost ); /* clean up */ cudaFree( d_a ); cudaFree( d_b ); cudaFree( d_c ); return; }
19,778
#include<stdio.h> #include<stdlib.h> #include<ctype.h> #include<math.h> #include<time.h> __global__ void euler_method(float *y, float *sum, float delta_t, int N) { int y0 = 4; int tId = threadIdx.x + blockIdx.x*blockDim.x; if(tId < N){ y[tId] = y0 + delta_t * sum[tId]; } } float edo_resuelta(float t); float edo_original(float t); void sumatoria(float *sum, float delta_t, int N); int main(){ //FILE *fp; //int i; FILE *fp_time; int j, N, counter = 0; float delta_t[6] = {powf(10, -1), powf(10, -2), powf(10, -3), powf(10, -4), powf(10, -5), powf(10, -6)}; float *y, *y_dev, *sum, *sum_dev; int block_size, grid_size; //fp = fopen("../1_c", "w"); fp_time = fopen("1_c_time", "w"); cudaEvent_t ct1, ct2; float dt; cudaEventCreate(&ct1); cudaEventCreate(&ct2); for(j = 0 ; j < 6 ; j++) { block_size = 256; N = 10 / delta_t[j]; grid_size = (int)ceil((float) N / block_size); cudaMalloc(&y_dev, sizeof(float) * N); cudaMalloc(&sum_dev, sizeof(float) * N); sum = (float*) malloc(sizeof(float) * N); y = (float*) malloc(sizeof(float) * N); sumatoria(sum, delta_t[j], N); cudaEventRecord(ct1); cudaMemcpy(sum_dev, sum, N *sizeof(float), cudaMemcpyHostToDevice); euler_method<<<grid_size,block_size>>>(y_dev, sum_dev, delta_t[j], N); cudaEventRecord(ct2); cudaMemcpy(y, y_dev, N *sizeof(float), cudaMemcpyDeviceToHost); cudaEventSynchronize(ct2); cudaEventElapsedTime(&dt, ct1, ct2); /* fprintf(fp, "*********************************\n"); fprintf(fp, "Con delta = %f\n", delta_t[j]); fprintf(fp, "*********************************\n"); for(i = 0 ; i < N; i++) { fprintf(fp, "t = %f\n", i+1 * delta_t[j]); fprintf(fp, "y[%i] = %f , %f\n", i + 1, *(y + i), edo_resuelta((i+1) * delta_t[j])); } */ counter++; printf("Tiempo que demora en HYBRID = %f [ms] para delta numero %d\n", dt, counter); fprintf(fp_time, "%f %f ",delta_t[j], dt); free(y); free(sum); cudaFree(y_dev); } fclose(fp_time); return 0; } void sumatoria(float *sum, float delta_t, int N){ sum[0] = edo_original(0); for(int i = 1; i < N; i++){ sum[i] = sum[i-1] + edo_original(i*delta_t); } } float edo_original(float t) { return 9 * (powf(t, 2)) - 4 * t + 5; } float edo_resuelta(float t) { return 3 * (powf(t, 3)) - 2 * (powf(t, 2)) + 5 * t + 4; }
19,779
/********************************************************************************** This code performs a calculation of pi using the monte carlo method using cuda GPU parallelisation. Created by: George Tall Email: george.tall@seh.ox.ac.uk /*********************************************************************************/ #include <math.h> #include <stdio.h> #include <stdlib.h> #include <cuda.h> #include <curand.h> __global__ void point_test(int *N, float *d_x, float *d_y, float *d_R, float *d_A){ long int index = blockDim.x*blockIdx.x + threadIdx.x; //printf("%d\n", index); // Now each value of d_R is computed d_R[index] = d_x[index]*d_x[index] + d_y[index]*d_y[index]; //printf("Thread %d d_R is %f \n", index, d_R[index]); // sync threads at this point to prevent deadlock __syncthreads(); if(d_R[index] < 1.0f) atomicAdd(&d_A[blockIdx.x], 1); //printf("\nPoints in block %d = %d", blockIdx.x, d_A[blockIdx.x]); } __global__ void area_reduction(float *d_A){ // allocate shared memory extern __shared__ float shared_array[]; // copy passed array into shared array int tid = threadIdx.x; //long int index = blockIdx.x*blockDim.x + threadIdx.x; shared_array[tid] = d_A[tid]; __syncthreads(); for(long int d = blockDim.x/2; d > 0; d /= 2){ if(tid < d){ atomicAdd(&shared_array[tid], shared_array[tid+d]); } __syncthreads(); } __syncthreads(); // if you're the first thread get the value from shared array if(tid == 0){ d_A[0] = shared_array[0]; } } int main() { // N is the number of random points. // area stores the number of random points that fall into // the area of the quadrant of a circle of radius 1 //size_t N = 2^10; int N = 6536; float area=0; //initalize the GPU int nBlocks = N/256; int nThreads = 256; int deviceid = 0; // using GPU with id 0 int devCount; // gets number of GPU available cudaGetDeviceCount(&devCount); // check if we have enough GPUs if(deviceid<devCount) { // tell CUDA that we want to use GPU 0 cudaSetDevice(deviceid); } else return(1); //random variable gen curandGenerator_t gen; //pointers to host memory and device memory we have a pointer for a radius in the device to calculate that before conditional is operated //we are also going to have an area count per block to prevent confusion in the kernal if statement later float *h_x, *h_y, *h_A; float *d_x, *d_y, *d_R, *d_A; //allocate host memory h_x = (float*)malloc(N*sizeof(float)); h_y = (float*)malloc(N*sizeof(float)); h_A = (float*)malloc(nBlocks*sizeof(float)); //allocate device memory cudaMalloc((void**)&d_x, N*sizeof(float)); cudaMalloc((void**)&d_y, N*sizeof(float)); cudaMalloc((void**)&d_R, N*sizeof(float)); cudaMalloc((void**)&d_A, nBlocks*sizeof(float)); // Create a pseudo-random number generator curandCreateGenerator(&gen, CURAND_RNG_PSEUDO_DEFAULT); // Set a seed curandSetPseudoRandomGeneratorSeed(gen, 1234ULL); // Generate N pseudo random numbers on device for x curandGenerateUniform(gen, d_x, N); curandGenerateUniform(gen, d_y, N); // Kernal for testing if points lie in area or not point_test<<<nBlocks, nThreads>>>(&N, d_x, d_y, d_R, d_A); // Syncronise the device here in order for all the blocks to finish calculating their area data points cudaDeviceSynchronize(); // Kernal for reducing the sum of the areas of the blocks area_reduction<<<nBlocks/nThreads, nThreads, nBlocks*sizeof(float)>>>(d_A); //Copy the generated numbers back to host // I've bought the other data back aside from h_A because it appears my reduction doesn't work cudaMemcpy(h_x, d_x, N*sizeof(float), cudaMemcpyDeviceToHost); cudaMemcpy(h_y, d_y, N*sizeof(float), cudaMemcpyDeviceToHost); cudaMemcpy(h_A, d_A, nBlocks*sizeof(float), cudaMemcpyDeviceToHost); for(int i = 0; i < nBlocks; i++){ printf("%f \n",h_A[i]); } area = h_A[0]; printf("\nPi from reduction:\t%f\n", (4.0*area)/(float)N); // reset area to zero so that we can do a monte carlo method on the host, this now limits the number of points we can use area = 0; for(int i = 0; i < N; i++){ if(h_x[i]*h_x[i] + h_y[i]*h_y[i] < 1.0f) area++; } printf("\nPi from host:\t%f\n", (4.0*area)/(float)N); /* area = 0; for(int i=0; i<N; i++) { double x = ((double)rand())/RAND_MAX; double y = ((double)rand())/RAND_MAX; if(x*x + y*y <= 1.0) area++; } printf("\nPi:\t%f\n", (4.0*area)/(double)N); */ // Free memory on host and device cudaFree(d_x); cudaFree(d_y); cudaFree(d_R); cudaFree(d_A); free(h_x); free(h_y); free(h_A); return(0); }
19,780
#include <stdio.h> #include <vector> #include <algorithm> #include <functional> #include <cuda_runtime.h> #include <cstdlib> #include <string> #include <map> #include <vector> #include <math.h> #include <cuda.h> #include <float.h> #include <thrust/host_vector.h> #include <thrust/device_vector.h> #include <thrust/transform.h> #include <thrust/sequence.h> #include <thrust/copy.h> #include <thrust/fill.h> #include <thrust/replace.h> #include <thrust/functional.h> #include <thrust/sort.h> #include <thrust/gather.h> #include <thrust/execution_policy.h> #include <iostream> #include <thrust/binary_search.h> #include <thrust/random.h> #include <curand.h> #include <time.h> #include <stdlib.h> #define THREADS_PER_BLOCK 256 #define gpuErrchk(ans) { gpuAssert((ans), __FILE__, __LINE__); } inline void gpuAssert(cudaError_t code, const char *file, int line, bool abort=true) { if (code != cudaSuccess) { fprintf(stderr,"GPUassert: %s %s %d\n", cudaGetErrorString(code), file, line); if (abort) exit(code); } } struct GenUnifRands { __device__ float operator () (int idx) { thrust::default_random_engine randEng; thrust::uniform_real_distribution<float> uniDist; randEng.discard(idx); return uniDist(randEng); } }; __global__ void segmentation_kernel(float* device_xyz,float* device_rgb,int* device_offset, int* neighbor_id,float* device_pdens,int yz_idx, int z_idx,int* parents,float* distances){ int block_i = blockIdx.x; int block_j = blockIdx.y; int block_k = blockIdx.z; int blockId = yz_idx * block_i + z_idx* block_j + block_k ; int threadId = threadIdx.x; int num_threads = THREADS_PER_BLOCK; int my_num_pts = device_offset[blockId + 1] - device_offset[blockId]; float3 nbr_xyz; float3 my_xyz; int my_idx; int nbr_idx; // Iterate over all points in voxel, num_threads points at a time. for(int i = 0 ; i < my_num_pts ; i+= num_threads){ if(threadId + i > my_num_pts){ return; } my_idx = device_offset[blockId] + i + threadId; my_xyz = *(float3*) &device_xyz[3*my_idx]; // Iterate over the neighbouring blocks including yourself. float min_distance = 10000000.0; float current_parent = my_idx; for(int j = 0 ; j < 7 ; j++){ if(neighbor_id[7*blockId+j] == -1){ continue; } int nbr_num_pts = device_offset[neighbor_id[7*blockId +j] + 1] - device_offset[neighbor_id[7*blockId + j]]; // Iterate over all the points in the neighbouring block. for(int k = 0; k < nbr_num_pts; k++){ nbr_idx = device_offset[neighbor_id[blockId*7 + j]] + k; nbr_xyz = *(float3*) &device_xyz[3*nbr_idx]; if(nbr_idx == my_idx){ continue; } float xyz_dist = pow(my_xyz.x - nbr_xyz.x,2.0) + pow((my_xyz.y - nbr_xyz.y),2.0) + pow((my_xyz.z - nbr_xyz.z),2.0); if(device_pdens[nbr_idx] > device_pdens[my_idx] && xyz_dist < min_distance){ min_distance = xyz_dist; current_parent = nbr_idx; } } } distances[my_idx] = min_distance; parents[my_idx] = current_parent; } } __global__ void compute_weights_kernel(float *device_xyz, float *device_rgb, int *device_offset, int *neighbor_id, int yz_idx, int z_idx, float *imp_wt, float* pdensity) { int block_i = blockIdx.x; int block_j = blockIdx.y; int block_k = blockIdx.z; int blockId = yz_idx * block_i + z_idx* block_j + block_k ; int threadId = threadIdx.x; int num_threads = THREADS_PER_BLOCK; float pdensity_sum = 0.0; float nbr_feature[6]; float norm_sum = 0.0; int my_num_pts = device_offset[blockId + 1] - device_offset[blockId]; float3 nbr_rgb; float3 nbr_xyz; float3 my_rgb; float3 my_xyz; float sigma_sq = 0.00005; float Aij_ew; float Aij_ew2; float lambda = 0.5; // Iterate over all points in voxel, num_threads points at a time. for(int i = 0 ; i < my_num_pts ; i+= num_threads){ pdensity_sum = 0.0; norm_sum = 0.0; nbr_feature[0] = 0.0; nbr_feature[1] = 0.0; nbr_feature[2] = 0.0; nbr_feature[3] = 0.0; nbr_feature[4] = 0.0; nbr_feature[5] = 0.0; if(threadId + i > my_num_pts){ return; } my_xyz = *(float3*) &device_xyz[3*(device_offset[blockId] + i + threadId)]; my_rgb = *(float3*) &device_rgb[3*(device_offset[blockId] + i + threadId)]; // Iterate over the neighbouring blocks including yourself. for(int j = 0 ; j < 7 ; j++){ if(neighbor_id[7*blockId+j] == -1){ continue; } int nbr_num_pts = device_offset[neighbor_id[7*blockId +j] + 1] - device_offset[neighbor_id[7*blockId + j]]; // Iterate over all the points in the neighbouring block. for(int k = 0; k < nbr_num_pts; k++){ nbr_xyz = *(float3*) &device_xyz[3*(device_offset[neighbor_id[blockId*7 + j]] + k)]; nbr_rgb = *(float3*) &device_rgb[3*(device_offset[neighbor_id[blockId*7 + j]] + k)]; float xyz_dist = pow(my_xyz.x - nbr_xyz.x,2.0) + pow((my_xyz.y - nbr_xyz.y),2.0) + pow((my_xyz.z - nbr_xyz.z),2.0); float rgb_dist = pow((my_rgb.x - nbr_rgb.x),2.0) + pow((my_rgb.y - nbr_rgb.y),2.0) + pow((my_rgb.z - nbr_rgb.z),2.0); Aij_ew = __expf(-1.0f * (xyz_dist/sigma_sq)); Aij_ew2 = __expf(-1.0f * ((lambda * xyz_dist) + ((1 - lambda) * rgb_dist))/(0.5 * sigma_sq)); pdensity_sum += Aij_ew2; nbr_feature[0] += Aij_ew * nbr_xyz.x; nbr_feature[1] += Aij_ew * nbr_xyz.y; nbr_feature[2] += Aij_ew * nbr_xyz.z; nbr_feature[3] += Aij_ew * nbr_rgb.x; nbr_feature[4] += Aij_ew * nbr_rgb.y; nbr_feature[5] += Aij_ew * nbr_rgb.z; } } norm_sum += pow(my_xyz.x - nbr_feature[0],2.0); norm_sum += pow(my_xyz.y - nbr_feature[1],2.0); norm_sum += pow(my_xyz.z - nbr_feature[2],2.0); norm_sum += pow(my_rgb.x - nbr_feature[3],2.0); norm_sum += pow(my_rgb.y - nbr_feature[4],2.0); norm_sum += pow(my_rgb.z - nbr_feature[5],2.0); pdensity[device_offset[blockId] + i + threadId] = pdensity_sum; imp_wt[device_offset[blockId] + i + threadId] = norm_sum; } } void cuda_resampling(int num_pts, int num_voxels, float *flattenXYZ, float *flattenRGB, int *voxel_offset, int *neighbor_ids, int x_idx, int y_idx, int z_idx,int num_samples,uint *sample_arr,float *pdens) { float *device_xyz, *device_rgb; int *device_offset, *device_neighbor_ids; cudaMalloc(&device_xyz, num_pts*3*sizeof(float)); cudaMalloc(&device_rgb, num_pts*3*sizeof(float)); cudaMalloc(&device_offset, (num_voxels+1)*sizeof(int)); cudaMalloc(&device_neighbor_ids , num_voxels * 7 * sizeof(int)); thrust::device_vector<float> dev_imp_wt(num_pts); thrust::device_vector<float> dev_pdensity(num_pts); thrust::host_vector<float> host_imp_wt(num_pts); thrust::host_vector<float> host_pdensity(num_pts); float* imp_wt = thrust::raw_pointer_cast(dev_imp_wt.data()); float* pdensity = thrust::raw_pointer_cast(dev_pdensity.data()); gpuErrchk(cudaMemcpy(device_offset,voxel_offset, (num_voxels+1)*sizeof(int),cudaMemcpyHostToDevice)); gpuErrchk(cudaMemcpy(device_xyz, flattenXYZ, num_pts*3*sizeof(float),cudaMemcpyHostToDevice)); gpuErrchk(cudaMemcpy(device_rgb, flattenRGB, num_pts*3*sizeof(float),cudaMemcpyHostToDevice)); gpuErrchk(cudaMemcpy(device_neighbor_ids, neighbor_ids, num_voxels*7*sizeof(int),cudaMemcpyHostToDevice)); int yz_idx = y_idx * z_idx; dim3 gridDim(x_idx,y_idx,z_idx); dim3 blockDim(THREADS_PER_BLOCK,1,1); compute_weights_kernel<<<gridDim,blockDim>>>(device_xyz,device_rgb,device_offset,device_neighbor_ids,yz_idx,z_idx,imp_wt,pdensity); printf("finished computing weights! \n"); cudaDeviceSynchronize(); thrust::copy(dev_imp_wt.begin(),dev_imp_wt.end(),host_imp_wt.begin()); thrust::copy(dev_pdensity.begin(),dev_pdensity.end(),host_pdensity.begin()); /* WEIGHTED SAMPLING */ printf("====================\n"); // step 1: normalize: (TODO: use functor to make fast ) float norm_sum = thrust::reduce(dev_imp_wt.begin(),dev_imp_wt.end()); float norm_factor = 1.0/norm_sum; thrust::device_vector<float> temp(dev_imp_wt.size()); thrust::fill(temp.begin(),temp.end(),norm_factor); thrust::transform(dev_imp_wt.begin(),dev_imp_wt.end(),temp.begin(),dev_imp_wt.begin(),thrust::multiplies<float>()); // step 2: compute prefix sum (inclusive scan) thrust::device_vector<float> wts_rs(dev_imp_wt.size()); thrust::inclusive_scan(dev_imp_wt.begin(),dev_imp_wt.end(),wts_rs.begin()); // step 3: generate uniform random numbers: srand(time(NULL)); int seed = rand(); thrust::device_vector<float> d_unifrands(num_samples); thrust::transform( thrust::make_counting_iterator(seed), thrust::make_counting_iterator(seed + num_samples), d_unifrands.begin(),GenUnifRands()); // step 4 : generate (weighted) random samples thrust::device_vector<unsigned int> samples(num_samples); thrust::lower_bound(wts_rs.begin(),wts_rs.end(),d_unifrands.begin(),d_unifrands.end(),samples.begin()); thrust::host_vector<unsigned int> h_samples(num_samples); thrust::copy(samples.begin(),samples.end(),h_samples.begin()); uint* pc_samples = thrust::raw_pointer_cast(h_samples.data()); memcpy(sample_arr, pc_samples,num_samples*sizeof(uint)); // step 5 : gather sample_arr indices of p_density thrust::device_vector<float> dev_pdensity_rs(num_samples); thrust::host_vector<float> host_pdensity_rs(num_samples); thrust::gather(thrust::device,samples.begin(),samples.end(),dev_pdensity.begin(),dev_pdensity_rs.begin()); thrust::copy(dev_pdensity_rs.begin(),dev_pdensity_rs.end(),host_pdensity_rs.begin()); float* host_pdensity_rs_ptr = thrust::raw_pointer_cast(host_pdensity_rs.data()); memcpy(pdens, host_pdensity_rs_ptr,num_samples*sizeof(float)); printf("finished weighted sampling of points\n"); return; } void cuda_segmentation(int num_pts, int num_voxels, float* pdens, float *flattenXYZ, float *flattenRGB, int *voxel_offset, int *neighbor_ids, int x_idx, int y_idx, int z_idx,int* parents_ptr) { float *device_xyz, *device_rgb, *device_pdens; int *device_offset, *device_neighbor_ids; cudaMalloc(&device_xyz, num_pts*3*sizeof(float)); cudaMalloc(&device_rgb, num_pts*3*sizeof(float)); cudaMalloc(&device_offset, (num_voxels+1)*sizeof(int)); cudaMalloc(&device_neighbor_ids , num_voxels * 7 * sizeof(int)); cudaMalloc(&device_pdens, num_pts*sizeof(float)); thrust::device_vector<int> dev_parents(num_pts); thrust::host_vector<int> host_parents(num_pts); thrust::device_vector<float> dev_distances(num_pts); thrust::host_vector<float> host_distances(num_pts); int* parents = thrust::raw_pointer_cast(dev_parents.data()); float* distances = thrust::raw_pointer_cast(dev_distances.data()); gpuErrchk(cudaMemcpy(device_offset,voxel_offset, (num_voxels+1)*sizeof(int),cudaMemcpyHostToDevice)); gpuErrchk(cudaMemcpy(device_xyz, flattenXYZ, num_pts*3*sizeof(float),cudaMemcpyHostToDevice)); gpuErrchk(cudaMemcpy(device_rgb, flattenRGB, num_pts*3*sizeof(float),cudaMemcpyHostToDevice)); gpuErrchk(cudaMemcpy(device_neighbor_ids, neighbor_ids, num_voxels*7*sizeof(int),cudaMemcpyHostToDevice)); gpuErrchk(cudaMemcpy(device_pdens, pdens, num_pts * sizeof(float), cudaMemcpyHostToDevice)); int yz_idx = y_idx * z_idx; dim3 gridDim(x_idx,y_idx,z_idx); dim3 blockDim(THREADS_PER_BLOCK,1,1); segmentation_kernel<<<gridDim,blockDim>>>(device_xyz,device_rgb,device_offset,device_neighbor_ids, device_pdens,yz_idx,z_idx,parents,distances); cudaDeviceSynchronize(); printf("finished segmentation!\n"); int num_gather = 6; thrust::device_vector<int> temp_1(num_pts); thrust::device_vector<int> temp_2(num_pts); thrust::copy(dev_parents.begin(),dev_parents.end(),temp_1.begin()); for(int i = 0 ; i < num_gather ; i++){ thrust::gather(thrust::device, dev_parents.begin(),dev_parents.end(),temp_1.begin(),temp_2.begin()); thrust::copy(temp_2.begin(),temp_2.end(),temp_1.begin()); thrust::copy(temp_2.begin(),temp_2.end(),dev_parents.begin()); } printf("finished tree cutting!\n"); thrust::copy(dev_parents.begin(),dev_parents.end(),host_parents.begin()); int* host_parents_ptr = thrust::raw_pointer_cast(host_parents.data()); memcpy(parents_ptr, host_parents_ptr,num_pts*sizeof(int)); return; }
19,781
#include <stdio.h> #include <iostream> #include <cuda.h> #include <vector> #include "demo.cuh" __global__ void add_kernel(float* A, float* B, float* C) { const unsigned int idx = blockIdx.x * blockDim.x + threadIdx.x; C[idx] = A[idx] + B[idx]; //printf("%d\n", C[idx]); } int testCUDA() { int num = 100; float a[num]; float b[num]; float c[num]; float a_test[num]; std::vector<float> a_vec; std::vector<float> b_vec; std::vector<float> c_vec; int block_num = 20; int block_size = num / block_num; cudaError_t cudaStatus; for (int i = 0; i < num; i++) { a[i] = i; b[i] = i; //a_vec.push_back(i); //b_vec.push_back(i); a_vec.emplace_back(i); b_vec.emplace_back(i); } float *a_d, *b_d, *c_d; cudaMalloc((void **)&a_d, sizeof(float) * num); cudaMalloc((void **)&b_d, sizeof(float) * num); cudaMalloc((void **)&c_d, sizeof(float) * num); //cudaMemcpy(a_d, a, num * sizeof(float), cudaMemcpyHostToDevice); //cudaMemcpy(b_d, b, num * sizeof(float), cudaMemcpyHostToDevice); cudaMemcpy(a_d, a_vec.data(), num * sizeof(int), cudaMemcpyHostToDevice); cudaMemcpy(b_d, b_vec.data(), num * sizeof(int), cudaMemcpyHostToDevice); add_kernel<<<block_num, block_size>>>(a_d, b_d, c_d); cudaStatus = cudaGetLastError(); if(cudaStatus != cudaSuccess) { std::cout << "add_kernel failed: " << cudaGetErrorString(cudaStatus) << std::endl; goto Error; } //cudaThreadSynchronize(); cudaStatus = cudaMemcpy(a_test, a_d, num * sizeof(int), cudaMemcpyDeviceToHost); if(cudaStatus != cudaSuccess) { std::cout << "cuda memcpy failed: " << cudaGetErrorString(cudaStatus) << std::endl; goto Error; } for (int i=0; i<num; i++) { std::cout << i << ": " << *(a_test + i) << std::endl; } cudaStatus = cudaMemcpy(c, c_d, num * sizeof(int), cudaMemcpyDeviceToHost); if(cudaStatus != cudaSuccess) { std::cout << "cuda memcpy failed: " << cudaGetErrorString(cudaStatus) << std::endl; goto Error; } for (int i = 0; i < num; i++) { //std::cout << "c[" << i << "]: " << c[i] << std::endl; } Error: cudaFree(a_d); cudaFree(b_d); cudaFree(c_d); return 0; }
19,782
/* autor fredy m uaem desonses@gmail.com para mas comentarios */ #include <device_functions.h> #include <stdio.h> #include <stdlib.h> #include <vector_types.h> #include <cuda.h> #include <math.h> #include "cuda_runtime.h" #include "device_launch_parameters.h" #define N 33 /* realiza la transpuesta de una matriz */ // definicio de memoria constante CUDA __constant__ float dev_A[N][N]; //GLOBAL: func desde el host y ejecutada en el kernel(DEVICE) __global__ void transpuesta(float *dev_B) { int columna = threadIdx.x; int fila = threadIdx.y; int pos = columna + N * fila; // cada hilo coloca un elemento de la matriz final dev_B[pos] = dev_A[columna][fila]; } int main(int argc, char** argv) { float *hst_A, *hst_B; float *dev_B; int size = N * N * sizeof(float); //reserva de memoria en el host hst_A = (float*)malloc(size); hst_B = (float*)malloc(size); //reserva de memoria en el device cudaMalloc((void**)&dev_B, size); //llenar la matriz for (int i = 0; i < N*N; i++) { hst_A[i] = float(i) + 1; } //copiar los datos hacia el device cudaError_t error = cudaMemcpyToSymbol(dev_A, hst_A, size); if (error != cudaSuccess) { printf("Error Memori const\n"); } //dimensiones del kernel a lanzar dim3 bloques(1); dim3 hilos(N, N); //lanzamiento del kernel transpuesta <<<bloques, hilos >>> (dev_B); //recoger los datos cudaMemcpy(hst_B, dev_B, size, cudaMemcpyDeviceToHost); //impresion de los datos printf("Matriz original:\n"); for (int i = 0; i < N; i++) { for (int j = 0; j < N; j++) { printf("%.2f\t", hst_A[j + i * N]); } printf("\n"); } printf("Matriz transpuesta:\n"); for (int i = 0; i < N; i++) { for (int j = 0; j < N; j++) { printf("%.2f\t", hst_B[j + i * N]); } printf("\n"); } // printf("\n pulsa INTRO para salir:\n"); fflush(stdin); char tecla = getchar(); return 0; }
19,783
/* NiuTrans.Tensor - an open-source tensor library * Copyright (C) 2017, Natural Language Processing Lab, Northeastern University. * All rights reserved. * * Licensed under the Apache License, Version 2.0 (the "License"); * you may not use this file except in compliance with the License. * You may obtain a copy of the License at * * http://www.apache.org/licenses/LICENSE-2.0 * * Unless required by applicable law or agreed to in writing, software * distributed under the License is distributed on an "AS IS" BASIS, * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. * See the License for the specific language governing permissions and * limitations under the License. */ /* * $Created by: Xu Chen (email: hello_master1954@163.com) 2018-09-12 */ #include "Dropout.h" #include "Dropout.cuh" #include "Loss.cuh" #include "../XDevice.h" #ifdef USE_CUDA // the CUDA stuff #include <cuda_runtime.h> #include <cublas_v2.h> #include <cuda.h> #endif namespace nts{ // namespace nts(NiuTrans.Tensor) #ifdef USE_CUDA /* dropout function (Cuda kernel) >> x - input data pointer >> y - output data pointer >> m - mask indicator to set zero >> s - the scale factor >> size - size of input/output */ __global__ void KernelDropoutCompute(DTYPE * x, DTYPE * y, DTYPE * m, DTYPE s, int size) { int i = blockDim.x * blockIdx.x + threadIdx.x; if (i < size){ y[i] = x[i] * m[i] * s; } } /* dropout function (Cuda version) >> x - input tensor >> y - output tensor >> mask - mask tensor to set 0 >> scaleFactor - the scale factor */ void _CudaDropout(const XTensor * x, XTensor * y, const XTensor * mask, DTYPE scaleFactor) { if(x->dataType == DEFAULT_DTYPE && y->dataType == DEFAULT_DTYPE){ CheckNTErrors(!x->isSparse && !y->isSparse, "the activation function (rectify) does not support sparse matrices."); CheckNTErrors(x->unitNum && y->unitNum, "we require two vectors with the same length."); int gridSize[3], blockSize[3]; GDevs.GetCudaThread(x->devID, x->unitNum, gridSize, blockSize); int devIDBackup; ProtectCudaDev(x->devID, devIDBackup); KernelDropoutCompute<<<dim3(gridSize[0]), dim3(blockSize[0])>>>((DTYPE*)x->data, (DTYPE*)y->data, (DTYPE*)mask->data, scaleFactor, x->unitNum); BacktoCudaDev(x->devID, devIDBackup); } else ShowNTErrors("TODO!"); } /* backward computation of dropout function (Cuda kernel) dE/dx = dE/dy * dy/dx >> dedy - dE/dy >> dedx - dE/dx >> m - mask indicator to set zero >> s - the scale factor >> size - size of input/output */ __global__ void KernelDropoutBackward(DTYPE * dedy, DTYPE * dedx, DTYPE * m, DTYPE s, int size) { int i = blockDim.x * blockIdx.x + threadIdx.x; if (i < size){ dedx[i] = dedy[i] * m[i] * s; } } /* backward computation of dropout function (Cuda version) dE/dx = dE/dy * dy/dx >> y - output of the dropout function >> x - input of the dropout function >> dedy - dE/dy >> dedx - dE/dx >> mask - mask tensor to set 0 >> scaleFactor - the scale factor */ void _CudaDropoutBackward(const XTensor * y, const XTensor * x, const XTensor * dedy, XTensor * dedx, const XTensor * mask, DTYPE scaleFactor) { int gridSize[3], blockSize[3]; if(x->dataType == DEFAULT_DTYPE && y->dataType == DEFAULT_DTYPE){ GDevs.GetCudaThread(x->devID, x->unitNum, gridSize, blockSize); int devIDBackup; ProtectCudaDev(x->devID, devIDBackup); /* dE/ds = dE/dy * dy/ds */ KernelDropoutBackward<<<dim3(gridSize[0]),dim3(blockSize[0])>>> ((DTYPE*)dedy->data, (DTYPE*)dedx->data, (DTYPE*)mask->data, scaleFactor, x->unitNum); BacktoCudaDev(x->devID, devIDBackup); } else ShowNTErrors("TODO!"); } #endif } // namespace nts(NiuTrans.Tensor)
19,784
#include <cuda.h> #include <assert.h> #include <stdio.h> template <int input_per_thread, int filter_per_thread, int input_per_block, int filter_per_block> __global__ static void _cwc_kern_convolutional_forward_propagate(const int strides, const int border, const int batch, float* input, const int rows, const int cols, const int channels_per_partition, const int partition, float* out, const int out_rows, const int out_cols, float* filter, const int filter_rows, const int filter_cols, const int count, float* const biases) { assert(gridDim.x * partition * filter_per_block * input_per_block == out_cols * batch * count); assert(gridDim.y == out_rows); assert(gridDim.z == partition); extern __shared__ float shared[]; float* shared_block = &shared[0]; float* shared_weights = &shared[input_per_block]; float* shared_bias = &shared[input_per_block + filter_per_block]; float prod[filter_per_thread][input_per_thread]; assert(input_per_block == input_per_thread * blockDim.x); assert(filter_per_block == filter_per_thread * blockDim.y); const int thidx = threadIdx.x + threadIdx.y * blockDim.x; const int thcnt = blockDim.x * blockDim.y; int c, i, j, x, y; #pragma unroll for (i = 0; i < filter_per_thread; i++) #pragma unroll for (j = 0; j < input_per_thread; j++) prod[i][j] = 0; const int origin_x = blockIdx.x % out_cols; const int origin_y = blockIdx.y; const int input_group_count = batch / input_per_block; const int input_group_idx = (blockIdx.x % (out_cols * input_group_count)) / out_cols; const int filter_group_idx = blockIdx.z * count / (filter_per_block * partition) + blockIdx.x / (out_cols * input_group_count); // for the partitioned filter group input += (blockIdx.z * channels_per_partition * rows * cols + origin_y * strides * cols + origin_x * strides) * batch + input_group_idx * input_per_block; assert(thcnt >= input_per_block); assert(thcnt >= filter_per_block); if (thidx < filter_per_block) shared_bias[thidx] = biases[filter_group_idx * filter_per_block + thidx]; const int start_x = max(origin_x * strides - border, 0) - (origin_x * strides - border); const int end_x = min(origin_x * strides - border + filter_cols, cols) - (origin_x * strides - border); const int start_y = max(origin_y * strides - border, 0) - (origin_y * strides - border); const int end_y = min(origin_y * strides - border + filter_rows, rows) - (origin_y * strides - border); filter += filter_group_idx * filter_per_block; for (c = 0; c < channels_per_partition; c++) { for (y = start_y; y < end_y; y++) for (x = start_x; x < end_x; x++) { if (thidx < input_per_block) shared_block[thidx] = input[((y - border) * cols + x - border) * batch + thidx]; if (thidx < filter_per_block) shared_weights[thidx] = filter[(y * filter_cols + x) * count + thidx]; __syncthreads(); #pragma unroll for (i = 0; i < filter_per_thread; i++) #pragma unroll for (j = 0; j < input_per_thread; j++) prod[i][j] += shared_block[j + threadIdx.x * input_per_thread] * shared_weights[i + threadIdx.y * filter_per_thread]; __syncthreads(); } input += rows * cols * batch; filter += filter_rows * filter_cols * count; } const int outcnt = out_rows * out_cols * batch; out += (filter_group_idx * filter_per_block + threadIdx.y * filter_per_thread) * outcnt + (origin_y * out_cols + origin_x) * batch + input_group_idx * input_per_block + threadIdx.x * input_per_thread; #pragma unroll for (i = 0; i < filter_per_thread; i++) { const float bias = shared_bias[i + threadIdx.y * filter_per_thread]; #pragma unroll for (j = 0; j < input_per_thread; j++) out[j] = max(0.0, prod[i][j] + bias); out += outcnt; } } int main(int argc, char** argv) { float* in = 0; float* out = 0; cudaMalloc(&in, sizeof(float) * (55 * 55 * 96 * 256)); cudaMalloc(&out, sizeof(float) * (27 * 27 * 256 * 256)); float* in_host = 0; float* out_host = 0; int i, j, c, k; cudaMallocHost(&in_host, sizeof(float) * 55 * 55 * 96 * 128); for (i = 0; i < 55; i++) for (j = 0; j < 55; j++) for (c = 0; c < 96; c++) for (k = 0; k < 128; k++) in_host[i * 55 * 96 * 128 + j * 96 * 128 + c * 128 + k] = c * k; cudaMemcpy(in, in_host, sizeof(float) * 55 * 55 * 96 * 128, cudaMemcpyHostToDevice); cudaMallocHost(&out_host, sizeof(float) * 27 * 27 * 256 * 128); for (i = 0; i < 27; i++) for (j = 0; j < 27; j++) for (c = 0; c < 256; c++) for (k = 0; k < 128; k++) out_host[i * 27 * 256 * 128 + j * 256 * 128 + c * 128 + k] = c * k; cudaMemcpy(out, out_host, sizeof(float) * 27 * 27 * 256 * 128, cudaMemcpyHostToDevice); float* w = 0; cudaMalloc(&w, sizeof(float) * (256 * 96 / 2) * 5 * 5); float* biases = 0; cudaMalloc(&biases, sizeof(float) * 256); dim3 thread_per_block(64 / 4, 32 / 8); dim3 num_blocks(27 * 2 * 256 / (32 * 2), 27, 2); int shared_memory_size = sizeof(float) * (64 + 32 * 2); cudaFuncSetCacheConfig(_cwc_kern_convolutional_forward_propagate<4, 8, 64, 32>, cudaFuncCachePreferL1); _cwc_kern_convolutional_forward_propagate <4, 8, 64, 32> <<<num_blocks, thread_per_block, shared_memory_size>>> (2, 1, 128, in, 55, 55, 96 / 2, 2, out, 27, 27, w, 5, 5, 256, biases); cudaFree(biases); cudaFree(w); cudaFree(out); cudaFree(in); cudaFreeHost(out_host); cudaFreeHost(in_host); return 0; }
19,785
__global__ void KNNSearch( float * result, const int * args, const float * pc1, const float * pc2) { int cudaNumBlocks = args[0]; int cudaNumThreads = args[1]; int pc1NumPts = args[2]; int pc2NumPts = args[3]; int pc2Idx = blockIdx.x * cudaNumThreads + threadIdx.x; float currPtX = pc2[pc2NumPts * 0 + pc2Idx]; float currPtY = pc2[pc2NumPts * 1 + pc2Idx]; float currPtZ = pc2[pc2NumPts * 2 + pc2Idx]; if (pc2Idx < pc2NumPts) { int nnIdx = 0; float nnDist = 100000.0f; for (int i = 0; i < pc1NumPts; i++) { float otherPtX = pc1[pc1NumPts * 0 + i]; float otherPtY = pc1[pc1NumPts * 1 + i]; float otherPtZ = pc1[pc1NumPts * 2 + i]; float checkDist = (currPtX - otherPtX) * (currPtX - otherPtX) + (currPtY - otherPtY) * (currPtY - otherPtY) + (currPtZ - otherPtZ) * (currPtZ - otherPtZ); if (checkDist < nnDist) { nnDist = checkDist; nnIdx = i; } } result[pc2NumPts * 0 + pc2Idx] = nnIdx + 1; result[pc2NumPts * 1 + pc2Idx] = nnDist; } }
19,786
#include "includes.h" // Author: Jose F. Martinez Rivera // Course: ICOM4036 - 040 // Professor: Wilson Rivera Gallego // Assignment 2 - CUDA Implementation #define V 8 #define E 11 #define MAX_WEIGHT 1000000 #define TRUE 1 #define FALSE 0 typedef int boolean; // //Represents an edge or path between Vertices typedef struct { int u; int v; } Edge; //Represents a Vertex typedef struct { int title; boolean visited; } Vertex; //Finds the weight of the path from vertex u to vertex v __device__ __host__ int findEdge(Vertex u, Vertex v, Edge *edges, int *weights) { int i; for(i = 0; i < E; i++) { if(edges[i].u == u.title && edges[i].v == v.title) { return weights[i]; } } return MAX_WEIGHT; } __global__ void Find_Vertex(Vertex *vertices, Edge *edges, int *weights, int *length, int *updateLength) { int u = threadIdx.x; if(vertices[u].visited == FALSE) { vertices[u].visited = TRUE; int v; for(v = 0; v < V; v++) { //Find the weight of the edge int weight = findEdge(vertices[u], vertices[v], edges, weights); //Checks if the weight is a candidate if(weight < MAX_WEIGHT) { //If the weight is shorter than the current weight, replace it if(updateLength[v] > length[u] + weight) { updateLength[v] = length[u] + weight; } } } } }
19,787
#include<stdio.h> #include<stdlib.h> #include<stdbool.h> #include<string.h> #include<cuda.h> #include<time.h> #include<cuda_runtime_api.h> #include<device_launch_parameters.h> #include <device_functions.h> #define MASK_WIDTH 3 //Here MASK_WIDTH = MASK_HEIGHT = 2*N + 1 where N is half-width of the chosen square mask #define BLOCK_WIDTH 32 __global__ void convolution_kernel(unsigned char *input_img, unsigned char *output_img, int height, int width) { int x,y,row,col,chan,i,j; unsigned char pixval; int freq[256]; for(chan=0;chan<3;chan++) //3 Channel Image { int tx = threadIdx.x; int ty = threadIdx.y; int row = blockIdx.x*BLOCK_WIDTH + tx; int col = blockIdx.y*BLOCK_WIDTH + ty; if(row < height && col < width) { for(i=0;i<256;i++)freq[i]=0; for(x=row-MASK_WIDTH/2;x<=row+MASK_WIDTH/2;x++) { for(y=col-MASK_WIDTH/2;y<=col+MASK_WIDTH/2;y++) { if((x >= 0) && (x < height) && (y>=0) && (y < width)) { pixval = input_img[(x*width + y)*3 + chan]; } else { if(x<0 && y<0) { pixval = input_img[chan]; } else if(x<0 && y<width) { pixval = input_img[3*y + chan]; } else if(x<0) { pixval = input_img[3*(width-1) + chan]; } else if(x<height && y<0) { pixval = input_img[x*width*3 + chan]; } else if(x<height && y>width) { pixval = input_img[(x*width +width-1)*3 + chan]; } else if(x>height && y<0) { pixval = input_img[width*(height-1)*3 + chan]; } else if(x>height && y<width) { pixval = input_img[(width*(height-1)+y)*3 + chan]; } else { pixval = input_img[(width*(height-1) + (width-1))*3 + chan]; } } freq[pixval]++; } } j=0; for(i=0;i<256;i++) { j=j+freq[i]; if(j>((MASK_WIDTH*MASK_WIDTH)/2))break; } output_img[(row*width + col)*3 + chan] = i; } } } unsigned char* readImg(char *filename,int *height_out, int *width_out, int *maxval_out) { FILE *fptr; char buf[16]; int i,j,k; int height,width,maxval; unsigned char *input_img; int index = 0; fptr = fopen(filename,"rb"); if(!fptr) { printf("Unable to open file '%s'\n",filename); return NULL; } if (!fgets(buf, sizeof(buf), fptr)) { printf("Error reading format\n"); return NULL; } int c = getc(fptr); while(c == '#'){ while(getc(fptr)!='\n'); c = getc(fptr); } ungetc(c, fptr); if(fscanf(fptr,"%d %d",&height,&width) !=2){ printf("ERROR Reading Dimension\n"); return NULL; } if(fscanf(fptr,"%d",&maxval)!=1){ printf("ERROR Reading MAXDEPTH\n"); return NULL; } while(fgetc(fptr) != '\n'); printf("%d\t%d\t%d\n",height,width,maxval); int pix = width*height; input_img = (unsigned char*)(malloc((3*pix)*sizeof(unsigned char))); if (fread(input_img,sizeof(unsigned char),3*pix, fptr) != 3*pix) { printf("Error loading image '%s'\n", filename); return NULL; } *height_out = height; *width_out = width; *maxval_out = maxval; fclose(fptr); printf("Image read successfully\n"); return input_img; } int writeImg(int width, int height, int maxval, unsigned char *output_img) { FILE *fptr; fptr = fopen("output.ppm","wb"); if(!fptr) { printf("Error opening file\n"); return 0; } fprintf(fptr,"P6\n"); fprintf(fptr, "%d %d\n",height,width); // rgb component depth fprintf(fptr, "%d\n",maxval); int pix = 3*width*height; // pixel data int j; if((j=fwrite(output_img,sizeof(unsigned char),pix, fptr))!=pix) { printf("ERROR WRITING %d\n",j); } fclose(fptr); return 1; } int main(int argc, char **argv) { int height,width,maxval; unsigned char *input_img,*output_img; unsigned char *dev_input_img,*dev_output_img; input_img = readImg(argv[1], &height, &width, &maxval); //printf("%d %d %d\n",height,width,maxval); int pix = width*height; cudaMalloc((void**)&dev_input_img,3*pix*sizeof(unsigned char)); cudaMalloc((void**)&dev_output_img,3*pix*sizeof(unsigned char)); //if(input_img == NULL)printf("WHY\n"); cudaMemcpy(dev_input_img,input_img,3*pix*sizeof(unsigned char),cudaMemcpyHostToDevice); int BlockX = ceil(((float)height)/BLOCK_WIDTH); int BlockY = ceil(((float)width)/BLOCK_WIDTH); dim3 dimGrid(BlockX, BlockY,1); dim3 dimBlock(BLOCK_WIDTH,BLOCK_WIDTH,1); cudaEvent_t start, stop; cudaEventCreate(&start); cudaEventCreate(&stop); cudaEventRecord(start); convolution_kernel<<<dimGrid,dimBlock>>>(dev_input_img, dev_output_img, height, width); cudaEventRecord(stop); cudaEventSynchronize(stop); float parallel_time = 0; cudaEventElapsedTime(&parallel_time, start, stop); cudaDeviceSynchronize(); output_img = (unsigned char*)(malloc(3*pix*sizeof(unsigned char))); cudaMemcpy(output_img,dev_output_img,3*pix*sizeof(unsigned char),cudaMemcpyDeviceToHost); int i = writeImg(width,height,maxval,output_img); if(i==0) return 1; cudaError_t error = cudaGetLastError(); if(error != cudaSuccess) { // print the CUDA error message and exit printf("CUDA error: %s\n", cudaGetErrorString(error)); exit(-1); } printf("%f\n",parallel_time); }
19,788
/* 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,int var_2,float var_3,float var_4,float var_5,float var_6,int 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,float var_34,float var_35,float var_36,float var_37,float var_38,float var_39,float var_40,float var_41,float var_42) { for (int i=0; i < var_1; ++i) { for (int i=0; i < var_2; ++i) { if (comp < -1.3991E35f + var_3 * +1.4162E9f / var_4) { if (comp <= -1.1995E-9f - -1.7017E-37f / var_5 / (var_6 + +1.8188E-35f)) { float tmp_1 = (var_8 / var_9 - +0.0f); comp += tmp_1 / (+1.9210E35f / var_10 * -1.7824E35f + (-1.4496E-37f * expf((var_11 - fmodf(+1.2371E-44f - (-1.4016E-36f + var_12), (+1.4381E36f / +1.3855E34f * +1.6777E-27f)))))); for (int i=0; i < var_7; ++i) { var_13[i] = +0.0f * powf((-1.0946E5f * (+1.9144E-43f + var_14 - var_15 - var_16 / var_17)), var_18 - (-1.0446E-35f + var_19 * (+1.2310E-35f / (var_20 / var_21)))); float tmp_2 = +1.2608E-22f; comp = tmp_2 * var_13[i] + (+1.0021E35f - floorf((+1.6976E36f / var_22 - (var_23 - tanhf(var_24 / (var_25 * (var_26 * -0.0f))))))); } if (comp <= (+0.0f * (var_27 - +1.3131E34f))) { comp += var_28 - +1.9225E36f; } if (comp > (var_29 - (-1.7616E-41f * (var_30 - var_31)))) { comp = var_32 + -1.3167E36f - (-1.4523E1f * var_33 / (var_34 / var_35)); float tmp_3 = +1.6180E-44f / var_36 - (var_37 + var_38 - (-1.8663E-37f + -0.0f)); comp = tmp_3 / (-1.2382E35f * powf((+1.7628E-42f + var_39 / ldexpf(var_40 / coshf(-1.5175E35f / var_41), 2)), var_42 - +0.0f)); } } } } } 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]); int tmp_3 = atoi(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]); int tmp_8 = atoi(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 = initPointer( 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 = atof(argv[23]); float tmp_24 = 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]); float tmp_35 = atof(argv[35]); float tmp_36 = atof(argv[36]); float tmp_37 = atof(argv[37]); float tmp_38 = atof(argv[38]); float tmp_39 = atof(argv[39]); float tmp_40 = atof(argv[40]); float tmp_41 = atof(argv[41]); float tmp_42 = atof(argv[42]); float tmp_43 = atof(argv[43]); 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,tmp_35,tmp_36,tmp_37,tmp_38,tmp_39,tmp_40,tmp_41,tmp_42,tmp_43); cudaDeviceSynchronize(); return 0; }
19,789
#include "includes.h" __global__ void matmul_double(double* A, double* B , double* C, int M, int N, int K) { int bx = blockIdx.x ; int by = blockIdx.y ; int tx = threadIdx.x ; int ty = threadIdx.y ; int row = by * TILE_WIDTH + ty ; int col = bx * TILE_WIDTH + tx ; __shared__ double SA[TILE_WIDTH][TILE_WIDTH+1] ; __shared__ double SB[TILE_WIDTH][TILE_WIDTH+1] ; double Csub = 0; for (int i = 0; i < (K-1)/TILE_WIDTH +1 ; ++i) { /* code */ if ( (row < M) && (i * TILE_WIDTH + tx < K ) ){ SA[ty][tx] = A[row*K + i * TILE_WIDTH + tx] ; } else{ SA[ty][tx] = 0; } if ( (col < N ) && ( i * TILE_WIDTH + ty < K)){ SB[tx][ty] = B[ col * K + i*TILE_WIDTH + ty] ; } else{ SB[tx][ty] = 0; } __syncthreads() ; for (int k = 0; k < TILE_WIDTH; ++k){ Csub += SA[ty][k]*SB[tx][k] ; } __syncthreads() ; } //C[row*n + col] = Csub ; if ( (row < M ) && ( col < N )){ C[ row * N + col] = Csub ; } }
19,790
#include "includes.h" __global__ void matrixAddKernel(float* A, float* B, float* C, int n) { int Row = blockIdx.y * blockDim.y + threadIdx.y; int Col = blockIdx.x * blockDim.x + threadIdx.x; if((Row < n) && (Col < n)) C[Row * n + Col] = A[Row * n + Col] + B[Row * n + Col]; }
19,791
#include"stdio.h" #include<cuda_runtime.h> #include <sys/time.h> #define N 1024 // Kernel definition __global__ void VecAdd(float* A, float* B, float* C) { int i = threadIdx.x; for(int j=0;j<1000;j++) C[i] = (A[i] * B[i]); } long getCurrentTime() { struct timeval tv; gettimeofday(&tv,NULL); return tv.tv_sec * 1000000 + tv.tv_usec; } void cpu_VecAdd(int i,float* A, float* B, float* C) { for(int j=0;j<1000;j++) C[i] = (A[i] * B[i]); } int main() { // Kernel invocation with N threads printf("Hello,World\n"); float *A=new float[N],*B=new float[N],*C=new float[N]; for(int i=0;i<N;i++) { A[i]=i; B[i]=2*i; } size_t size = N * sizeof(float); float* d_A; cudaMalloc(&d_A, size); float* d_B; cudaMalloc(&d_B, size); float* d_C; cudaMalloc(&d_C, size); float *e=new float[N]; long st=getCurrentTime(); cudaMemcpy(d_A, A, size, cudaMemcpyHostToDevice); cudaMemcpy(d_B, B, size, cudaMemcpyHostToDevice); VecAdd<<<1, N>>>(d_A, d_B, d_C); cudaMemcpy(C, d_C, size, cudaMemcpyDeviceToHost); long ed=getCurrentTime(); printf("gpu running time:%ld\n",ed-st); st=getCurrentTime(); for(int i=0;i<N;i++) cpu_VecAdd(i,A,B,e); ed=getCurrentTime(); printf("cpu running time:%ld\n",ed-st); cudaFree(d_A); cudaFree(d_B); cudaFree(d_C); for(int i=0;i<N;i++) { //printf("%f ",C[i]); } printf("\n"); }
19,792
#include "cuda_runtime.h" #include "device_launch_parameters.h" #include <stdio.h> // 1 block of threads --> 8 values, grid = 1 __global__ void unique_idx_calc_threadIdx(int * input) { int tid = threadIdx.x; printf("threadIdx : %d, value : %d \n", tid, input[tid]); } // 4 blocks, each block - 4 threads. __global__ void unique_gid_calculation(int * input) { int tid = threadIdx.x; int offset = blockIdx.x * blockDim.x; int gid = tid + offset; printf("blockIdx.x : %d, threadIdx.x : %d, blockDim.x : %d, gridDim.x: %d, value : %d \n", blockIdx.x, tid, blockDim.x , gridDim.x, input[gid]); } int main() { int array_size = 16; int array_bite_size = sizeof(int) * array_size; int h_data[] = {23, 9, 4, 53, 64, 12, 1, 33, 22, 11, 9, 12, 13, 89, 90, 77}; for (int i=0; i < array_size; i++) { printf("%d ", h_data[i]); } printf ("\n \n"); int * d_data; cudaMalloc((void **)&d_data, array_bite_size); cudaMemcpy(d_data, h_data, array_bite_size, cudaMemcpyHostToDevice); dim3 block(4); dim3 grid(4); //unique_idx_calc_threadIdx<<<grid, block>>>(d_data); unique_gid_calculation<<<grid, block>>>(d_data); cudaDeviceSynchronize(); cudaDeviceReset(); }
19,793
// // Created by hina on 2021-03-13. // #include "activation.cuh" __device__ float activation::func_relu(float input_num) { return input_num > 0 ? input_num : 0.0f; } __device__ float activation::deriv_relu(float input_num) { return input_num > 0 ? 1.0f : 0.0f; }
19,794
#include "includes.h" __global__ void bias_grad(float *pre_grad, float *output, int rows, int cols) { int i = blockIdx.x * blockDim.x + threadIdx.x; if (i >= rows) return; output[i] = 0; for (int k = 0; k < cols; k++) { output[i] += pre_grad[i * cols + k]; } }
19,795
#include <stdio.h> #include <stdlib.h> #include "cuda.h" #include <curand.h> #include <curand_kernel.h> #include <math.h> #include <time.h> #include <sys/time.h> #include <unistd.h> #define HASH_STEP 720 #define WARP_SIZE 32 #define gpuErrchk(ans) { gpuAssert((ans), __FILE__, __LINE__); } inline void gpuAssert(cudaError_t code, char *file, int line, bool abort=true) { if (code != cudaSuccess) { fprintf(stderr,"GPUassert: %s %s %d\n", cudaGetErrorString(code), file, line); if (abort) exit(code); } } __global__ void generate_randoms(float *d_random_r, int numDim, unsigned long seed) { int idx = threadIdx.x + blockIdx.x*blockDim.x; curandState localState; curand_init (seed, idx, 0, &localState); for(int i=0;i<numDim;i++) { d_random_r[idx*(numDim+1)+i] = curand_uniform(&localState); // ri [0,1] } d_random_r[idx*(numDim+1)+numDim] = curand_uniform(&localState)*HASH_STEP; // bi [0,720] } __device__ float matrix_multiplication(float *d_data, float *d_random_r, int numDim) { float sum = 0; for(int i=0;i<numDim;i++) { sum += d_data[i]*d_random_r[i]; } return(sum); } __global__ void hash_f(float *d_data, int *d_hash_codes, float *d_random_r, int numDim, unsigned long seed) { int idx = threadIdx.x + blockIdx.x*blockDim.x; d_hash_codes[idx] = floorf((matrix_multiplication(&d_data[(idx/numDim)*numDim],&d_random_r[idx*(numDim+1)],numDim) + d_random_r[idx*(numDim+1)+numDim])/HASH_STEP); } __global__ void initial_count(int *d_hash_codes, int *d_offsets, int digit, int numDim, int NTHREADS) { int idx = threadIdx.x + blockIdx.x*blockDim.x; int s_zeros = 0,s_ones = 0; for(int i=0;i<64;i++) { if(d_hash_codes[idx*numDim*64+i*numDim+numDim-digit] == 0) { s_zeros++; } else if(d_hash_codes[idx*numDim*64+i*numDim+numDim-digit] == 1){ s_ones++; } } d_offsets[idx*2+2] = s_zeros; // First 2 slots are 0 in the table d_offsets[idx*2+3] = s_ones; // Same } __global__ void lsd_count(int *d_hash_codes, int *d_offsets, int *d_ids, int digit, int numDim, int NTHREADS) { int idx = threadIdx.x + blockIdx.x*blockDim.x; int s_zeros = 0,s_ones = 0; for(int i=0;i<64;i++) { if(d_hash_codes[d_ids[idx*64+i]*numDim+numDim-digit] == 0) { s_zeros++; } else if(d_hash_codes[d_ids[idx*64+i]*numDim+numDim-digit] == 1){ s_ones++; } } d_offsets[idx*2+2] = s_zeros; // First 2 slots are 0 in the table d_offsets[idx*2+3] = s_ones; // Same } __global__ void initialize_sorting(int *d_hash_codes, int *d_offsets, int *d_ids_1, int digit, int numDim, int NTHREADS) { int idx = threadIdx.x + blockIdx.x*blockDim.x; int offset_zero = d_offsets[idx*2]; int offset_one = d_offsets[idx*2+1]+d_offsets[2*NTHREADS]; // d_offsets[2*NTHREADS] holds the total number of zeros for(int i=0;i<64;i++) { if(d_hash_codes[idx*numDim*64+i*numDim+numDim-digit] == 0) { d_ids_1[offset_zero] = idx*64+i; offset_zero++; } else if(d_hash_codes[idx*numDim*64+i*numDim+numDim-digit] == 1){ d_ids_1[offset_one] = idx*64+i; offset_one++; } } // At this point offset_zero + offset_one = 64 } __global__ void radix_sort(int *d_hash_codes, int *d_offsets, int *d_ids_1, int *d_ids_2, int digit, int numDim, int NTHREADS) { int idx = threadIdx.x + blockIdx.x*blockDim.x; int offset_zero = d_offsets[idx*2]; int offset_one = d_offsets[idx*2+1]+d_offsets[2*NTHREADS]; // d_offsets[2*NTHREADS] holds the total number of zeros for(int i=0;i<64;i++) { if(d_hash_codes[d_ids_1[idx*64+i]*numDim+numDim-digit] == 0) { d_ids_2[offset_zero] = d_ids_1[idx*64+i]; offset_zero++; } else if(d_hash_codes[d_ids_1[idx*64+i]*numDim+numDim-digit] == 1){ d_ids_2[offset_one] = d_ids_1[idx*64+i]; offset_one++; } } // At this point offset_zero + offset_one = 64 } __global__ void write_sorted_hash_codes(int *d_hash_codes, int *d_sorted_hash_codes, int *d_ids, int numDim) { int idx = threadIdx.x + blockIdx.x*blockDim.x; for(int i=0;i<64;i++) { for(int j=0;j<numDim;j++) { d_sorted_hash_codes[idx*64*numDim+i*numDim+j] = d_hash_codes[d_ids[idx*64+i]*numDim+j]; } } } void lsh(int numObjects, int numDim, float *h_data) { /* Generate hash codes */ int NTHREADS = ceil((float)numObjects*numDim/(float)(64*WARP_SIZE)); //Concurrent threads per loop int BLOCK_SIZE = 256; int GRID_SIZE = ceil((float)NTHREADS/(float)BLOCK_SIZE); dim3 grid(GRID_SIZE,1); dim3 threads(BLOCK_SIZE,1); printf("NTHREADS:%d GRID_SIZE:%d BLOCK_SIZE:%d\n", NTHREADS,GRID_SIZE,BLOCK_SIZE); float *d_random_r; cudaMalloc((void**)&d_random_r,NTHREADS*(numDim+1)*sizeof(float)); // numDim+1 because ri has numDim values and bi has 1 float *d_data; cudaMalloc((void**)&d_data,numObjects*numDim*sizeof(float)); cudaMemcpy(d_data,h_data,numObjects*numDim*sizeof(float),cudaMemcpyHostToDevice); int *d_hash_codes; cudaMalloc((void**)&d_hash_codes, numObjects*numDim*sizeof(int)); for(int i=0;i<64*WARP_SIZE;i++) { generate_randoms<<<grid,threads>>>(d_random_r,numDim,time(NULL)); hash_f<<<grid,threads>>>(&d_data[i*NTHREADS],&d_hash_codes[i*NTHREADS],d_random_r,numDim,time(NULL)); } cudaFree(d_random_r); cudaFree(d_data); int *h_hash_codes; h_hash_codes = (int *)malloc(numObjects*numDim*sizeof(int)); cudaMemcpy(h_hash_codes,d_hash_codes,numObjects*numDim*sizeof(int),cudaMemcpyDeviceToHost); /* Sort hash codes */ NTHREADS = ceil((float)numObjects/(float)64); //Concurrent threads per loop BLOCK_SIZE = 256; GRID_SIZE = ceil((float)NTHREADS/(float)BLOCK_SIZE); dim3 grid_2(GRID_SIZE,1); dim3 threads_2(BLOCK_SIZE,1); printf("NTHREADS:%d GRID_SIZE:%d BLOCK_SIZE:%d\n", NTHREADS,GRID_SIZE,BLOCK_SIZE); int *d_ids_1; cudaMalloc((void**)&d_ids_1,numObjects*sizeof(int)); int *d_ids_2; cudaMalloc((void**)&d_ids_2,numObjects*sizeof(int)); int *d_offsets; cudaMalloc((void**)&d_offsets,(2*NTHREADS+2)*sizeof(int)); int *d_sorted_hash_codes; cudaMalloc((void**)&d_sorted_hash_codes, numObjects*numDim*sizeof(int)); initial_count<<<grid_2,threads_2>>>(d_hash_codes,d_offsets,1,numDim,NTHREADS); cudaDeviceSynchronize(); int *h_offsets; h_offsets = (int *)malloc((2*NTHREADS+2)*sizeof(int)); cudaMemcpy(h_offsets,d_offsets,(2*NTHREADS+2)*sizeof(int),cudaMemcpyDeviceToHost); //Calculate offsets serially h_offsets[0] = 0; h_offsets[1] = 0; for(int j=2;j<(2*NTHREADS+2);j++) { h_offsets[j] += h_offsets[j-2]; } cudaMemcpy(d_offsets,h_offsets,(2*NTHREADS+2)*sizeof(int),cudaMemcpyHostToDevice); initialize_sorting<<<grid_2,threads_2>>>(d_hash_codes,d_offsets,d_ids_1,1,numDim,NTHREADS); for(int i=1;i<numDim;i++) { if (i%2!=0) { lsd_count<<<grid_2,threads_2>>>(d_hash_codes,d_offsets,d_ids_1,(i+1),numDim,NTHREADS); cudaMemcpy(h_offsets,d_offsets,(2*NTHREADS+2)*sizeof(int),cudaMemcpyDeviceToHost); //Calculate offsets serially h_offsets[0] = 0; h_offsets[1] = 0; for(int j=2;j<(2*NTHREADS+2);j++) { h_offsets[j] += h_offsets[j-2]; } cudaMemcpy(d_offsets,h_offsets,(2*NTHREADS+2)*sizeof(int),cudaMemcpyHostToDevice); radix_sort<<<grid_2, threads_2>>>(d_hash_codes,d_offsets,d_ids_1,d_ids_2,(i+1),numDim,NTHREADS); } else { lsd_count<<<grid_2,threads_2>>>(d_hash_codes,d_offsets,d_ids_2,(i+1),numDim,NTHREADS); cudaMemcpy(h_offsets,d_offsets,(2*NTHREADS+2)*sizeof(int),cudaMemcpyDeviceToHost); //Calculate offsets serially h_offsets[0] = 0; h_offsets[1] = 0; for(int j=2;j<(2*NTHREADS+2);j++) { h_offsets[j] += h_offsets[j-2]; } cudaMemcpy(d_offsets,h_offsets,(2*NTHREADS+2)*sizeof(int),cudaMemcpyHostToDevice); radix_sort<<<grid_2, threads_2>>>(d_hash_codes,d_offsets,d_ids_2,d_ids_1,(i+1),numDim,NTHREADS); } } write_sorted_hash_codes<<<grid_2, threads_2>>>(d_hash_codes,d_sorted_hash_codes,d_ids_2,numDim); int *h_sorted_hash_codes; h_sorted_hash_codes = (int *)malloc(numObjects*numDim*sizeof(int)); cudaMemcpy(h_sorted_hash_codes,d_sorted_hash_codes,numObjects*numDim*sizeof(int),cudaMemcpyDeviceToHost); } int main(int argc, char** argv) { int numObjects = atoi(argv[1]); int numDim = atoi(argv[2]); FILE *dataset; float *h_data; h_data = (float *)malloc(numObjects*numDim*sizeof(float)); dataset = fopen("/export/home/dhmtasos/Ergasia4/data.bin", "rb"); if(dataset == NULL) { printf("Error opening data.bin\n"); } size_t a = fread(h_data, sizeof(float), numObjects*numDim, dataset); if(a!=numObjects*numDim) { printf("Error reading data from data.bin\n"); } fclose(dataset); struct timeval first, second, lapsed; struct timezone tzp; gettimeofday(&first, &tzp); lsh(numObjects,numDim,h_data); gettimeofday(&second, &tzp); if(first.tv_usec>second.tv_usec){ second.tv_usec += 1000000; second.tv_sec--; } lapsed.tv_usec = second.tv_usec - first.tv_usec; lapsed.tv_sec = second.tv_sec - first.tv_sec; printf("Time elapsed: %d, %d s\n", lapsed.tv_sec, lapsed.tv_usec); }
19,796
__host__ __device__ int get_raw(int idx, int K_k) { return idx / K_k; } __host__ __device__ int get_col(int idx, int K_k) { return idx % K_k; } __host__ __device__ void get_mul(double* res, double* a, double* b, int idx, int K_m, int K_n, int K_k ) { int m = get_raw(idx, K_k); int k = get_col(idx, K_k); for (int n = 0; n < K_n; n++) res[m * K_k + k] += a[idx] * b[n * K_k + k]; } __global__ void mmul(double* a, double* b, double* res, int K_m, int K_n, int K_k) { int id = blockDim.x * blockIdx.x + threadIdx.x; if (id < K_m * K_k) get_mul(res, a, b, id, K_m, K_n, K_k); }
19,797
#include <stdio.h> #include <cuda.h> #define n 10 __global__ void add(int*a, int*max) { int i = blockIdx.x * blockDim.x + threadIdx.x; if(i < n) { for(i=0;i<n;i++) { if(a[i]>*max) *max=a[i]; } } } int main() { int a[n]; int i; int max; int* dev_a; int* dev_max; cudaMalloc((void**)&dev_max, sizeof(int)); cudaMalloc((void**)&dev_a, n * sizeof(int)); printf("\narray elements:\n"); for(i=0;i<n;i++) { scanf("%d",&a[i]); } max = a[0]; cudaMemcpy(dev_a, a, n * sizeof(int), cudaMemcpyHostToDevice); cudaMemcpy(dev_max, &max, sizeof(int), cudaMemcpyHostToDevice); add<<<1,1>>>(dev_a, dev_max); cudaMemcpy(&max, dev_max, sizeof(int), cudaMemcpyDeviceToHost); printf("\nMax is %d\n",max); cudaFree(dev_max); cudaFree(dev_a); cudaDeviceReset(); return 0; }
19,798
#include<bits/stdc++.h> int main (void) { printf("Hello World!\n"); return 0; }
19,799
#include "includes.h" __global__ void bcnn_op_cuda_tanh_grad_kernel(int n, float *x, float *dx) { int i = (blockIdx.x + blockIdx.y * gridDim.x) * blockDim.x + threadIdx.x; if (i < n) { dx[i] *= (1 - x[i] * x[i]); } return; }
19,800
// Copyright (c) 2012-2017 VideoStitch SAS // Copyright (c) 2018 stitchEm #include <cstdio> int main() { int devices; cudaError_t err = cudaGetDeviceCount(&devices); if (err != cudaSuccess) { fprintf(stderr, "cudaGetDeviceCount failed: %s\n", cudaGetErrorString(err)); return 1; } if (devices == 0) { fprintf(stderr, "No CUDA device found!\n"); return 1; } cudaDeviceProp props_1; cudaDeviceProp props_2; for (int i = 0; i < devices; ++i) { props_2 = props_1; err = cudaGetDeviceProperties(&props_1, i); if (err != cudaSuccess) { fprintf(stderr, "cudaGetDeviceProperties failed: %s\n", cudaGetErrorString(err)); return 1; } if (i > 0 && (props_1.major != props_2.major || props_1.minor != props_2.minor)) { fprintf(stderr, "Multiple CUDA arch not supported at the moment\n"); return 1; } } printf("CUDA compute capability: %d%d", props_1.major, props_1.minor); // Easiest cross-platform way to pass the number seems to be the exit code // nvcc --run on windows seems to print the filename through cl.exe (that can't be silenced) return props_1.major * 10 + props_1.minor; }