Zedboard zynq-7000 Opencl 浮点数从类型 'double*' 强制转换为类型 'double' 无效

Zedboard zynq-7000 Opencl floating point invalid cast from type 'double*' to type 'double'

本文关键字：类型 double 转换 无效 Opencl Zedboard zynq-7000 浮点数      更新时间：2023-10-16

我是 Xilinx 开发领域的新手，我正在准备 zedboard 的硕士学位。我正在尝试使用 Vivado HLS 在 Zedboard 上开发一个带有 Opencl 的卷积加速器。

我创建了OpenCL块。

#pragma OPENCL EXTENSION cl_khr_fp64 : enable
#include <clc.h>
__kernel void __attribute__ ((reqd_work_group_size(26,1,1)))
conv_openCL( __global double*a, __global double*c) {
int i = get_global_id(0);
c[i] = a[i] * 2 ;
}

然后我在 Vivado 中使用它来创建架构： 图片 1

将其导出到 Vivado SDK 并从此处使用和修改代码后。 OpenCL 块无法正确执行，我不知道原因。

所以我的问题是为什么当我用浮点数或双精度替换 int 类型时它会给我一个错误。

旧代码：

volatile char *control = (volatile char*)0x43C00000;
volatile int *wg_x   = (volatile int*)0x43C00010;
volatile int *wg_y   = (volatile int*)0x43C00018;
volatile int *wg_z   = (volatile int*)0x43C00020;
volatile int *o_x    = (volatile int*)0x43C00028;
volatile int *o_y    = (volatile int*)0x43C00030;
volatile int *o_z    = (volatile int*)0x43C00038;
volatile int *a_addr = (volatile int*) 0x43C00040;
volatile int *c_addr = (volatile int*)0x43C00048;
...
void main(){
...
int* a;
int*c;
a = (int*)malloc(WG_SIZE_X *sizeof(int));
c = (int*)malloc(WG_SIZE_X *sizeof(int));
*a_addr =(unsigned int)a;
*c_addr =(unsigned int)c;
...


}

我的代码：

volatile char *control = (volatile char*)0x43C00000;
volatile int *wg_x   = (volatile int*)0x43C00010;
volatile int *wg_y   = (volatile int*)0x43C00018;
volatile int *wg_z   = (volatile int*)0x43C00020;
volatile int *o_x    = (volatile int*)0x43C00028;
volatile int *o_y    = (volatile int*)0x43C00030;
volatile int *o_z    = (volatile int*)0x43C00038;
volatile double *a_addr = (volatile double*) 0x43C00040;
volatile double *c_addr = (volatile double*)0x43C00048;
...
void main(){
...
int* a;
int*c;
a = (double*)malloc(WG_SIZE_X *sizeof(double));
c = (double*)malloc(WG_SIZE_X *sizeof(double));
*a_addr =(double)a;
*c_addr =(double)c;
...
}

它给了我这个错误：

错误图像

从类型"双精度*"到类型"双精度"的转换无效

请帮助我，我想将"a"数组的内容传输到 OpenCL 块的"a_addr"数组。

您正在尝试将int*、a投射到double。这不能作为指针存储地址 - 整数来完成。

您需要修改代码以：

volatile double **a_addr = (volatile double**) 0x43C00040;
volatile double **c_addr = (volatile double**)0x43C00048;
...
void main(){
...
double* a;
double* c;
a = (double*)malloc(WG_SIZE_X *sizeof(double));
c = (double*)malloc(WG_SIZE_X *sizeof(double));
*a_addr =a;
*c_addr =c;
...
}

或者你想以与原始代码类似的方式来做到这一点：

volatile int *a_addr = (volatile int*) 0x43C00040;
volatile int *c_addr = (volatile int*)0x43C00048;
...
void main(){
...
double* a;
double* c;
a = (double*)malloc(WG_SIZE_X *sizeof(double));
c = (double*)malloc(WG_SIZE_X *sizeof(double));
*a_addr =(int)a;
*c_addr =(int)c;
...
}
相关文章：