Add initial hipProfileStart/Stop

And modify sample to show how to use.
Still needs some work to understand interaction with CXL.

Change-Id: I2579824d2dd7863ea23874d34f0dabb3cb305d3e

include/hip/hcc_detail/hip_runtime_api.h

@@ -1737,11 +1737,24 @@ hipError_t hipModuleLaunchKernel(hipFunction_t f,
  *
  *  @warning The cudaProfilerInitialize API format for "configFile" is not supported.
  *
- *  On AMD platforms, hipProfilerStart and hipProfilerStop require installation of AMD's GPU
- *  perf counter API and defining GPU_PERF
  */
 
 
+// TODO - expand descriptions:
+/**
+ * @brief Start recording of profiling information
+ * @warning : hipProfilerStart API is under development.
+ */
+hipError_t hipProfilerStart();
+
+
+/**
+ * @brief Stop recording of profiling information.
+ * @warning : hipProfilerStop API is under development.
+ */
+hipError_t hipProfilerStop();
+
+
 /**
  * @}
  */

include/hip/nvcc_detail/hip_runtime_api.h

@@ -643,6 +643,18 @@ inline static hipError_t hipMemcpyPeerAsync ( void* dst, int  dstDevice, const v
     return hipCUDAErrorTohipError(cudaMemcpyPeerAsync(dst, dstDevice, src, srcDevice, count, stream));
 }
 
+// Profile APIs:
+inline hipError_t hipProfilerStart() 
+{ 
+    return hipCUDAErrorTohipError(cudaProfileStart());
+}
+
+inline hipError_t hipProfilerStop() 
+{ 
+    return hipCUDAErrorTohipError(cudaProfileStop());
+}
+
+
 inline static hipError_t hipSetDeviceFlags (unsigned int flags)
 {
     return hipCUDAErrorTohipError(cudaSetDeviceFlags(flags));
@@ -802,6 +814,8 @@ inline static hipError_t hipModuleLaunchKernel(hipFunction_t f,
                     sharedMemBytes, stream, kernelParams, extra));
 }
 
+
+
 #ifdef __cplusplus
 }
 #endif