[OpenMP] Remove macro guards for device debugging

[jhuber6]

Summary

The plugin currently uses a macro to check if this is a debug built
before assigning the debug kind variable to the device environment
struct. This is being deprecated because the new device runtime does not
maintain separate debug builds and should always be availible.

[tianshilei1992]

Why not. :-)

[JonChesterfield]

Seems inconsistent. Could we leave the debug output available in release builds across the host plugins? Measured in the noise for performance on rocm (branches are easily predicted) and it's useful to toggle debug output on without recompiling.

[jhuber6]

In D112083#3073140, @JonChesterfield wrote:

Seems inconsistent. Could we leave the debug output available in release builds across the host plugins? Measured in the noise for performance on rocm (branches are easily predicted) and it's useful to toggle debug output on without recompiling.

This is specifically for toggling debugging using the new device runtime, which will be the default very soon. The new device runtime has no concept of a debug version, and is instead configured at compile-time by the user by specifying -fopenmp-target-debug=<N>. At runtime the user then specifies LIBOMPTARGET_DEVICE_RTL_DEBUG=<N> to activate those features. Having the actual use of this being guarded by a separate debug build defeats the purpose at least on the device side.

[JonChesterfield]

I really like the scheme of choosing debug level at application compile time, instead of at compiler runtime build time. I'm suggesting we run with that idea for the host runtime, effectively building the print expansion of DP(...) into the release build instead of eliding it.

[jdoerfert]

We want to get rid of OMPTARGET_DEBUG as much as possible, which probably means completely. This should have never been guarded. LGTM

openmp/libomptarget/plugins/amdgpu/src/rtl.cpp

struct device_environment {
   device_environment(int device_id, int number_devices,
                      __tgt_device_image *image, const size_t img_size)
       : image(image), img_size(img_size) {
 
     host_device_env.NumDevices = number_devices;
     host_device_env.DeviceNum = device_id;
     host_device_env.DebugKind = 0;
     host_device_env.DynamicMemSize = 0;
-#ifdef OMPTARGET_DEBUG
     if (char *envStr = getenv("LIBOMPTARGET_DEVICE_RTL_DEBUG")) {
       host_device_env.DebugKind = std::stoi(envStr);
     }
-#endif
 
     int rc = get_symbol_info_without_loading((char *)image->ImageStart,
                                              img_size, sym(), &si);
     if (rc != 0) {
       DP("Finding global device environment '%s' - symbol missing.\n", sym());
       return;
     }
 

openmp/libomptarget/plugins/cuda/src/rtl.cpp

__tgt_target_table *loadBinary(const int DeviceId,
 
     // send device environment data to the device
     {
       // TODO: The device ID used here is not the real device ID used by OpenMP.
       DeviceEnvironmentTy DeviceEnv{0, static_cast<uint32_t>(NumberOfDevices),
                                     static_cast<uint32_t>(DeviceId),
                                     static_cast<uint32_t>(DynamicMemorySize)};
 
-#ifdef OMPTARGET_DEBUG
       if (const char *EnvStr = getenv("LIBOMPTARGET_DEVICE_RTL_DEBUG"))
         DeviceEnv.DebugKind = std::stoi(EnvStr);
-#endif
 
       const char *DeviceEnvName = "omptarget_device_environment";
       CUdeviceptr DeviceEnvPtr;
       size_t CUSize;
 
       Err = cuModuleGetGlobal(&DeviceEnvPtr, &CUSize, Module, DeviceEnvName);
       if (Err == CUDA_SUCCESS) {
         if (CUSize != sizeof(DeviceEnv)) {