我在 cuda、pycuda 中使用复数有困难。
我在 C 中有这个：
#include <complex>
typedef std::complex<double> cmplx;
....
cmplx j(0.,1.);
另外，在相同的代码中：
#include <boost/python.hpp>
#include <boost/array.hpp>
...
typedef std::vector< boost::array<std::complex<double>,3 > > ComplexFieldType;
typedef std::vector< boost::array<double,3> > RealFieldType;
...
__global__ void compute(RealFieldType const & Rs,ComplexFieldType const & M,..)
...
我如何将其转换为与 pycuda 一起使用？
我尝试了这样的操作（根据“cuda by an example”一书）：
struct cuComplex {
float real;
float imag;
cuComplex(float a,float b): real(a),imag(b){}
cuComplex operator *(const cuComplex& a) {
return cuComplex(real*a.real -imag*a.imag ,imag*a.real +real*a.imag);
}
cuComplex operator +(const cuComplex& a) {
return cuComplex(real+a.real ,imag+a.imag);
};
cuComplex j(0.,1.); //instead of cmplx j(0.,1.);
__global__ void compute(float *Rs,cuComplex * M,..) //instead of RealFieldType const & Rs,ComplexFieldType const & M
....
我犯的一些错误是：
不允许数据成员初始值设定项
此声明没有存储类或类型说明符
谢谢！
---------------------编辑--------------------- -------------------------
我使用 #include 执行了以下操作（相对于上面) :
pycuda::complex<float> cmplx;
cmplx j(0.,1.);
以及 typedef std::vector boost::array,3> > ComplexFieldType;
和 ComplexFieldType const & M ，在全局函数内，
我尝试了“float *M”或“cmplx *M”。
到目前为止，我收到错误：
变量“cmplx”不是类型名称
如果我使用 pycuda::complex cmplx; ，然后我得到：
标识符“cmplx”未定义
名称后跟“::”必须是类或命名空间名称
另外：
<块引用>
表达式必须具有指向对象的指针类型（但也许这是来自代码的另一部分）
I have difficulties to use complex numbers in cuda,pycuda.
I have this in C:
#include <complex>
typedef std::complex<double> cmplx;
....
cmplx j(0.,1.);
Also,in the same code:
#include <boost/python.hpp>
#include <boost/array.hpp>
...
typedef std::vector< boost::array<std::complex<double>,3 > > ComplexFieldType;
typedef std::vector< boost::array<double,3> > RealFieldType;
...
__global__ void compute(RealFieldType const & Rs,ComplexFieldType const & M,..)
...
How can i convert this to use it with pycuda?
I tried sth like this (according to the book 'cuda by an example'):
struct cuComplex {
float real;
float imag;
cuComplex(float a,float b): real(a),imag(b){}
cuComplex operator *(const cuComplex& a) {
return cuComplex(real*a.real -imag*a.imag ,imag*a.real +real*a.imag);
}
cuComplex operator +(const cuComplex& a) {
return cuComplex(real+a.real ,imag+a.imag);
};
cuComplex j(0.,1.); //instead of cmplx j(0.,1.);
__global__ void compute(float *Rs,cuComplex * M,..) //instead of RealFieldType const & Rs,ComplexFieldType const & M
....
Some of the errors i take are:
data member initializer is not allowed
this declaration has no storage class or type specifier
Thank you!
---------------------EDIT----------------------------------------------
I did the following using #include <pycuda-complex.hpp> (relative to the above) :
pycuda::complex<float> cmplx;
cmplx j(0.,1.);
and as for typedef std::vector< boost::array<std::complex<double>,3 > > ComplexFieldType;
and ComplexFieldType const & M ,inside the global function,
i tried just "float *M " or "cmplx *M".
Until now , i am getting error :
variable "cmplx" is not a type name
If i use pycuda::complex cmplx; ,then i get:
identifier "cmplx" is undefined
name followed by "::" must be a class or namespace name
Also:
expression must have pointer-to-object type (but maybe this is from another part of code)
发布评论
评论（2）
确实不清楚您实际上想要做什么（如果您确实了解自己），并且随着编辑和评论的不断进行，问题变得越来越混乱。但为了稍微扩展 Andreas 的答案，这里有一段简单的、可编译的 CUDA 代码，它正确使用 pycuda 本机复杂类型：
这为您提供了普通内核的单双实数和复杂版本，并使用 nvcc 进行编译，如下所示：
也许这在某种程度上回答了你的问题......
It really isn't clear what you are actually trying to do (if you actually know yourself), and the question is getting progressively more confused as the edits and comments roll on. But to expand Andreas's answer a little, here is a simple, compilable piece of CUDA code which uses the pycuda native complex type correctly:
This gives you single and double real and complex versions of the trivial kernel and compiles with nvcc something like this:
Perhaps this goes someway to answering your question....
使用
与
std::complex<>相同的接口，实际上源自其 STLport 版本。Use
Same interface as
std::complex<>, in fact derived from the STLport version of that.