Using clmagma and clAmdBlas with NVIDIA OpenCL library Problems encountered using NVIDIA OpenCL libr

Accelerated Computing CUDA CUDA Programming and Performance
1

Hi,
I want in my project to use clmagma with clAmdBlas. But I have NVIDIA boards. So, as OpenCL should be platform agnostic it should compile and run without problems.
I have installed both, NVIDIA and AMD SKD.
When I compile with AMD OpenCL library all is ok. Compile gives no errors but segmentation fault is got on execution.
If I link in compile time to NVIDIA OpenCL library I get error like ‘undefined reference to OpenCL_function_name@OPENCL_1.0’

So, I have read ELF definition section of NVIDIA OpenCL library but there is such a section. It seems no provide OPENCL_1.0.
If I do same for AMD OpenCL library I get this:

Version definition section ‘.gnu.version_d’ contains 4 entries:
Addr: 0x0000000000001a38 Offset: 0x001a38 Link: 3 (.dynstr)
000000: Rev: 1 Flags: BASE Index: 1 Cnt: 1 Name: libOpenCL.so.1
0x001c: Rev: 1 Flags: none Index: 2 Cnt: 1 Name: OPENCL_1.0
0x0038: Rev: 1 Flags: none Index: 3 Cnt: 2 Name: OPENCL_1.1
0x0054: Parent 1: OPENCL_1.0
0x005c: Rev: 1 Flags: none Index: 4 Cnt: 2 Name: OPENCL_1.2
0x0078: Parent 1: OPENCL_1.1

Thanks in advance.
PD: The reason of using clAmdBlas, clMagma and NVIDIA cards is a requirerment of the project.

2

I have found the solution.
It seems that AMD and NVIDIA use different methods to version their libraries.
It is possible to compile against OpenCL AMD library but in clmagma files it is necessary to modify clGetPlatformIDs if you have installed more than one vendor implementation. If not, segmentation fault is got at runtime.