Device property concurrentKernels is added to hipDeviceProp_t struct.

For HCC path concurrentKernels is set to true since all ROCR hardware supports this feature.
For NVCC path concurrentKernels is obtained from CUDA's device property cudaDeviceProp::concurrentKernels.


[ROCm/clr commit: 4d4ca3ef3f]

projects/clr/hipamd/include/hip_runtime_api.h

@@ -93,6 +93,7 @@ typedef struct hipDeviceProp_t {
 	int clockInstructionRate ;   ///< Frequency in khz of the timer used by the device-side "clock*" instructions.  New for HIP.
 
     hipDeviceArch_t arch;  ///< Architectural feature flags.  New for HIP.
+    int concurrentKernels; ///< Device can possibly execute multiple kernels concurrently
  } hipDeviceProp_t;
 
 

projects/clr/hipamd/include/nvcc_detail/hip_runtime_api.h

@@ -207,6 +207,7 @@ inline static hipError_t hipDeviceGetProperties(hipDeviceProp_t *p_prop, int dev
     p_prop->arch.has3dGrid                   =  (ccVers >= 200);
     p_prop->arch.hasDynamicParallelism       =  (ccVers >= 350);
 
+    p_prop->concurrentKernels = cdprop.concurrentKernels;
 
     return hipCUDAErrorTohipError(cerror);
 }

projects/clr/hipamd/src/hip_hcc.cpp

@@ -383,6 +383,7 @@ hipError_t ihipDevice_t::getProperties(hipDeviceProp_t* prop)
     prop->arch.has3dGrid                   = 1;
     prop->arch.hasDynamicParallelism       = 0;
 
+    prop->concurrentKernels = 1; // All ROCR hardware supports executing multiple kernels concurrently
     return e;
 }