Hi, all:
I am working on the OpenACC compiler for the AMD GPUs. It is extremely difficult to debug the OpenCL kernel in AMD platform without the printf.
This is the simple kernel I am using.
#pragma OPENCL EXTENSION cl_amd_printf : enable
kernel void helloworld()
{
printf("hello world!\n");
}
However, it doesn't printf anything.
I use the version AMDAPPSDK-3.0-0-Beta.
This is the output that I received from my machine:
[0] Platform Name: AMD Accelerated Parallel Processing
Choose Platform [0].
32692 CPU device available.
1 GPU device available.
Choose GPU Device:1
Device Name = Hawaii
Device Vendor = Advanced Micro Devices, Inc.
The sample code is attached. Any suggestion is appreciated.
Thank you very much,
Daniel
Solved! Go to Solution.
Why are you using 3.0 Beta ? The actual release in available, in which they fixed the printf bug.
Please upgrade to the full release.
Please, someone help me solve it. This is very important for me to debug my compiler CL kernel generation phase.
I dont think so you can use printf in the kernel code. The extension which you are specifying , make sure that your hardware/compilier supports that. It would be helpful to debug if you can give the clinfo of your system.
I also tried to pull off the printf in kernel. You can refer to my program at:
https://github.com/pradyotsn/opencl_printf
Thank you
Why are you using 3.0 Beta ? The actual release in available, in which they fixed the printf bug.
Please upgrade to the full release.