CUDA By Examples 6 - 使用常量内存 Constant Memory_综合

1. 使用常量内存的好处

访问常量内存时, GPU会针对访问同一个地址的half-warp(16个threads)只读取一次此地址.
被访问的常量内存被cache, 之后的对此地址的访问可以更加快捷.
如果half-warp内的threads需要访问不同的地址, 那么这些访问就会串行进行, 速度会比使用global memory要慢. 因为访问global memory是可以并行的.

2. 光线追踪问题

可以简化为: 从某一个像素射出的垂直于bitmap平面的射线,和当前球体相交,求出最近的一个交点.
可以设置不同大小, 位置和颜色的球体.
这里写图片描述

#include "cuda.h"
#include "../common/book.h"
#include "../common/cpu_bitmap.h"#define INF 2e10f
#define rnd( x ) (x * rand() / RAND_MAX)
#define SPHERES 20 //球体数量
#define DIM 1024 // bitmap图大小struct Sphere{float r,b,g;float radius;float x,y,z; //球心//从某一个像素射出的垂直于bitmap平面的射线,和当前球体相交,求出最近的一个交点//没有交点就返回负无穷__device__ float hit(float ox, float oy, float *n){float dx = ox - x;float dy = oy - y;if(dx*dx + dy*dy < radius*radius){float dz = sqrtf(radius*radius - dx*dx - dy*dy);*n = dz / sqrtf( radius * radius);return dz + z;}return -INF;}
};// kernel函数应该加入第二个参数Sphere* s
// 因为main函数里的s是host变量,不能由kernel使用
__global__ void kernel( unsigned char *ptr, Sphere *s)
{int x = threadIdx.x + blockIdx.x * blockDim.x;int y = threadIdx.y + blockIdx.y * blockDim.y;int offset = x + y * blockDim.x * gridDim.x;float ox = (x - DIM/2);float oy = (y - DIM/2);float r=0, g=0, b=0;float maxz = -INF;for (int i=0; i<SPHERES; i++){float n;float t = s[i].hit(ox, oy, &n);if (t > maxz){float fscale = n;r = s[i].r * fscale;g = s[i].g * fscale;b = s[i].b * fscale;}}//四个通道赋值ptr[offset*4 + 0] = (int)(r*255);ptr[offset*4 + 1] = (int)(g*255);ptr[offset*4 + 2] = (int)(b*255);ptr[offset*4 + 3] = 255;
}int main(void)
{cudaEvent_t start, stop;HANDLE_ERROR( cudaEventCreate( &start ) );HANDLE_ERROR( cudaEventCreate( &stop ) );HANDLE_ERROR( cudaEventCreate( &start, 0 ) );CPUBitmap bitmap(DIM, DIM);unsigned char *dev_bitmap;Sphere *s;HANDLE_ERROR( cudaMalloc( (void**)&dev_bitmap,bitmap.image_size() ) );HANDLE_ERROR( cudaMalloc( (void**)&s,sizeof(Sphere) * SPHERES ) );//球体参数在CPU上初始化Sphere *temp_s = (Sphere*)malloc( sizeof(Sphere) * SPHERES );for (int i=0; i<SPHERES; i++){temp_s[i].r = rnd( 1.0f );temp_s[i].g = rnd( 1.0f );temp_s[i].b = rnd( 1.0f );temp_s[i].x = rnd( 1000.0f ) - 500;temp_s[i].y = rnd( 1000.0f ) - 500;temp_s[i].z = rnd( 1000.0f ) - 500;temp_s[i].radius = rnd( 100.0f ) + 20;}//把Sphere从主机拷贝到GPU上HANDLE_ERROR( cudaMemcpy( s, temp_s,sizeof(Sphere) * SPHERES,cudaMemcpyHostToDevice ) );free( temp_s );// GPU上做处理dim3 grids(DIM/16, DIM/16);dim3 threads(16, 16);kernel<<<grids,threads>>>(dev_bitmap, s);//从GPU拷贝到Host上HANDLE_ERROR( cudaMemcpy( bitmap.get_ptr(),dev_bitmap,bitmap.image_size(),cudaMemcpyDeviceToHost ) );bitmap.display_and_exit();//free memorycudaFree( dev_bitmap );cudaFree( s );}

输出:
这里写图片描述

3. 在光线追踪问题中使用Constant Memory

使用常量内存要注意:
1. 分配常量内存不使用动态分配, 直接在编译时分配. 而且此声明不能放在函数体内部.

__constant__ Sphere s[SPHERES];

将数据从Host拷贝到GPU的constant memory上需要使用:

cudaMemcpyToSymbol( s, temp_s, sizeof(Sphere) * SPHERES );

完整代码:

#include "cuda.h"
#include "../common/book.h"
#include "../common/cpu_bitmap.h"#define INF 2e10f
#define rnd( x ) (x * rand() / RAND_MAX)
#define SPHERES 20 //球体数量
#define DIM 1024 // bitmap图大小struct Sphere{float r,b,g;float radius;float x,y,z; //球心//从某一个像素射出的垂直于bitmap平面的射线,和当前球体相交,求出最近的一个交点//没有交点就返回负无穷__device__ float hit(float ox, float oy, float *n){float dx = ox - x;float dy = oy - y;if(dx*dx + dy*dy < radius*radius){float dz = sqrtf(radius*radius - dx*dx - dy*dy);*n = dz / sqrtf( radius * radius);return dz + z;}return -INF;}
};// 不是指针
// 常量内存的声明不能放在函数体内部.
__constant__ Sphere s[SPHERES];
__global__ void kernel( unsigned char *ptr)
{int x = threadIdx.x + blockIdx.x * blockDim.x;int y = threadIdx.y + blockIdx.y * blockDim.y;int offset = x + y * blockDim.x * gridDim.x;float ox = (x - DIM/2);float oy = (y - DIM/2);float r=0, g=0, b=0;float maxz = -INF;for (int i=0; i<SPHERES; i++){float n;float t = s[i].hit(ox, oy, &n);if (t > maxz){float fscale = n;r = s[i].r * fscale;g = s[i].g * fscale;b = s[i].b * fscale;}}//四个通道赋值ptr[offset*4 + 0] = (int)(r*255);ptr[offset*4 + 1] = (int)(g*255);ptr[offset*4 + 2] = (int)(b*255);ptr[offset*4 + 3] = 255;
}int main(void)
{cudaEvent_t start, stop;HANDLE_ERROR( cudaEventCreate( &start ) );HANDLE_ERROR( cudaEventCreate( &stop ) );HANDLE_ERROR( cudaEventCreate( &start, 0 ) );CPUBitmap bitmap(DIM, DIM);unsigned char *dev_bitmap;HANDLE_ERROR( cudaMalloc( (void**)&dev_bitmap,bitmap.image_size() ) );//常量内存不需要为之开辟GPU存储空间.//HANDLE_ERROR( cudaMalloc( (void**)&s,// sizeof(Sphere) * SPHERES ) );//球体参数在CPU上初始化Sphere *temp_s = (Sphere*)malloc( sizeof(Sphere) * SPHERES );for (int i=0; i<SPHERES; i++){temp_s[i].r = rnd( 1.0f );temp_s[i].g = rnd( 1.0f );temp_s[i].b = rnd( 1.0f );temp_s[i].x = rnd( 1000.0f ) - 500;temp_s[i].y = rnd( 1000.0f ) - 500;temp_s[i].z = rnd( 1000.0f ) - 500;temp_s[i].radius = rnd( 100.0f ) + 20;}//把Sphere从主机拷贝到GPU上HANDLE_ERROR( cudaMemcpyToSymbol( s, temp_s,sizeof(Sphere) * SPHERES ) );free( temp_s );// GPU上做处理dim3 grids(DIM/16, DIM/16);dim3 threads(16, 16);kernel<<<grids,threads>>>(dev_bitmap);//从GPU拷贝到Host上HANDLE_ERROR( cudaMemcpy( bitmap.get_ptr(),dev_bitmap,bitmap.image_size(),cudaMemcpyDeviceToHost ) );bitmap.display_and_exit();//free memorycudaFree( dev_bitmap );//cudaFree( s );}