Archives Discussions

[Skysnake]

Hi,

I have a very big problem with the OpenCL c++ sampel. I have solved several problems, but now i stuck and can´t find any solution for the Problem.

First my System:

Core2Duo E8400, HIS 5870, Cat 10.8 (8.14.10.0775), SDK2.2, VS2010 Prof, win 7 64

I start the project as a Visual C++ Win32 console appllication. After that I add to the linker $ATISTREAMSDKROOT)/lib/x86

Also i include the C:\...\ATI Sream\unclude\CL and C:\...\ATI Stream\include as Includedirectorys

As bilbiothekdirectorys i include C:\...\ATI Stream\lib

The Code is this:

So i can get it so far, that i can compile the programm without an arrow, the win32 console open, and i get the following results:

Platform number is: 1

Platform is by Advanced Micro Devices, Inc.

ERROR: Kernel() (-46)

Press any button...

I try very much, and use the search function for hours, but can´t find a solution for my problem

I hope somebody can help me.

PS: Excuse my english, i am no nativ speaker and schooltime is far away So when you don´t understand something, say it please.

EDIT: here the VS2010 project

http://s7.directupload.net/images/100914/9cvmyc3v.jpg

#################### lesson.cpp #################### // OpenCLAMD.cpp : Definiert den Einstiegspunkt für die Konsolenanwendung. // #include "stdafx.h" #include <utility> #define __NO_STD_VECTOR //Use cl::vektor instead of STL version //#define __NO_STD_STRING #include <malloc.h> #define alloca _alloca #include <CL\cl.hpp> const std::string hw("Hello World\n"); inline void checkErr(cl_int err, const char* name)//Fehlerbehandlung da eine Anwendung geöffnet wird { if (err != CL_SUCCESS){ std::cerr<<"ERROR: "<<name<<" ("<<err<<")"<<std::endl; exit(EXIT_FAILURE); //Wenn OpenCL nicht richtig initialisiert wird, wird ein Fehler ausgegeben und das Programm beendet //Wenn kein Fehler auftaucht passiert nichts } } //int _tmain(int argc, _TCHAR* argv[]) int main() { cl_int err; cl::vector< cl::Platform > platformList; cl::Platform::get(&platformList); checkErr(platformList.size()!=0 ? CL_SUCCESS : -1, "cl::Platform::get"); std::cerr<<"Platform number is: "<< platformList.size()<<std::endl; std::string platformVendor; platformList[0].getInfo((cl_platform_info)CL_PLATFORM_VENDOR, &platformVendor); std::cerr<<"Platform is by: "<<platformVendor<<"\n"; cl_context_properties cprops[3]={CL_CONTEXT_PLATFORM, (cl_context_properties)(platformList[0])(),0}; cl::Context context( CL_DEVICE_TYPE_CPU, cprops, NULL, NULL, &err); checkErr(err, "Context::Context()"); char* outH=new char[hw.length()+1]; cl::Buffer outCL( context, CL_MEM_WRITE_ONLY | CL_MEM_USE_HOST_PTR, hw.length()+1, outH, &err); checkErr(err, "Buffer::Buffer()"); cl::vector<cl::Device> devices; devices=context.getInfo<CL_CONTEXT_DEVICES>(); checkErr( devices.size() > 0 ? CL_SUCCESS : -1, "devices.size() > 0"); std::ifstream file("lesson1_kernels.cl"); checkErr(file.is_open() ? CL_SUCCESS:-1, "lesson1_kernel.cl"); std::string prog( std::istreambuf_iterator<char>(file), (std::istreambuf_iterator<char>())); cl::Program::Sources source( 1, std::make_pair(prog.c_str(),prog.length()+1)); cl::Program program(context, source); err = program.build(devices,""); checkErr(file.is_open() ? CL_SUCCESS : -1, "Program::build()"); cl::Kernel kernel(program, "hello", &err); checkErr(err, "Kernel::Kernel()"); err = kernel.setArg(0, outCL); checkErr(err, "Kernel::setArg()"); cl::CommandQueue queue(context, devices[0], 0, &err); checkErr(err, "CommandQueue::CommandQueue()"); cl::Event event; err = queue.enqueueNDRangeKernel( kernel, cl::NullRange, cl::NDRange(hw.length()+1), cl::NDRange(1,1), NULL, &event); checkErr(err, "CommandQueue::endqueueNDRangeKernel()"); event.wait(); err = queue.enqueueReadBuffer( outCL, CL_TRUE, 0, hw.length()+1, outH); checkErr(err, "CommandQueue::endqueueReadBuffer()"); std::cout<<outH; return EXIT_SUCCESS; } /* //ab hier kein OpenCL mehr printf("Hallo Welt"); std::cin.get(); return 0; } */ ############ lesson1_kernels.cl ############ #pragma OPENCL EXTENSION cl_khr_byte_addressable_store : enable __constant char hw[] = "Hello World\n"; __kernel void hello (__global char * out) { size_t tid = get_global_id(0); out [tid]= hw[tid]; }

[nou]

-46 is CL_INVALID_KERNEL_NAME so double chceck if you load kernel source properly

[Skysnake]

Ok, now i know whot the Problem is, i also thought that it is a Problem with the Cernel, the debugger stops in the line, where the cernel load.

But my problem now is, that i don´t understand how this helps me to fix the bug.

EDIT: I have only two lines, with the cernel:

cl::vector devices;
    devices=context.getInfo();
    checkErr( devices.size() > 0 ? CL_SUCCESS : -1, "devices.size() > 0");

    std::ifstream file("lesson1_kernels.cl");
    checkErr(file.is_open() ? CL_SUCCESS:-1, "lesson1_kernels.cl");

    std::string prog(
        std::istreambuf_iterator(file),
        (std::istreambuf_iterator()));

The underlined s wasn´t there, but it doesnt matter if it is there or not. Allways the same failure

EDIT: perhaps it helps, that i can´t do this step:

On Windows, with a Visual Studio command window, an example is:
cl /Fehello_world.exe /Ipath-OpenCL-include lesson.cpp path-OpenCL-libdir/OpenCL.lib

I just compile it with "F7" or "STRG+ALT+F7" and than start it without debugger

[Skysnake]

Nobody out there who knows whot the mistake is?

With another OpenCL program that didn´t use a .cl kernel file i have no problems.

[ryta1203]

So nou's suggestion was incorrect? Why do you send clBuildProgram an empty string?

[Skysnake]

I just follow the instructions under http://developer.amd.com/gpu/ATIStreamSDK/pages/TutorialOpenCL.aspx

And there also just stand:

cl::Program program(context, source);
 err = program.build(devices,"");
 checkErr(file.is_open() ? CL_SUCCESS : -1, "Program::build()");

I learn atm OpenCL and i think this exampel ist very confusing. I understand some basics, but not everything. So i can´t say you why there is a emty sting.

It realy looks strange.

[ryta1203]

That looks like a horribly put together tutorial, I wouldn't follow that if I were you.

[Skysnake]

Yeah, but the Problem is, that i want to bring a matlab programm from CPU to GPU with OpenCL. Atm it takse 6 hour for a picture. The goal ist to solve the problem with a consumer card in a few minutes or less. So performance is very important. And so i think there is now way without c++ bindings, and i know no ohter exampel, than this.

And also it must work -.-

[himanshu_gautam]

hi skysnake,

Have you tried running hellocl sample that came with the SDK.

[Skysnake]

Yes i try some sampels from the SDK, and i also have no problems with my own OpenCL programs.

Only when the source is in a etxra file i have problems.

[LeeHowes]

Try putting the full path instead of the cl file name alone in case the search path is the problem. Does that help at all? Remember if you're doing it with \ in the quotes to double them up (\\) so that they are correctly escaped in the string.

I try to avoid loading cl code from files for this reason. Generally I setup the build system with a little embedding script (or, at a push, the C preprocess) that loads the cl file and wraps it in quotes within a header file that's then used directly.

[Skysnake]

I will try your tip in the next days.