[OpenMP][CUDA] Remove hard thread limit in the plugin

[tianshilei1992]

Summary

OpenMP specification doesn't limit the max number of threads a team can
use. Like we did in D119313 to remove the hard limit of number of teams, we also
do it for number of threads.

[jdoerfert]

OpenMP doesn't say we need to use as many threads as the user wants either. Do we ask the hardware what it can do?

[tianshilei1992]

In D121255#3368474, @jdoerfert wrote:

OpenMP doesn't say we need to use as many threads as the user wants either. Do we ask the hardware what it can do?

Yes. That's what ThreadsPerBlock is for.

[tianshilei1992]

Move to D121262 for more cleanup.

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

class DeviceRTLTy {
   uint64_t DynamicMemorySize;
 
   /// Number of initial streams for each device.
   int NumInitialStreams = 32;
 
   /// Number of initial events for each device.
   int NumInitialEvents = 8;
 
-  static constexpr const int32_t HardThreadLimit = 1024;
   static constexpr const int32_t DefaultNumTeams = 128;
   static constexpr const int32_t DefaultNumThreads = 128;
 
   using StreamPoolTy = ResourcePoolTy<StreamAllocatorTy>;
   std::vector<std::unique_ptr<StreamPoolTy>> StreamPool;
 
   using EventPoolTy = ResourcePoolTy<EventAllocatorTy>;
   std::vector<std::unique_ptr<EventPoolTy>> EventPool;
if (Err != CUDA_SUCCESS) {
 
       if (EnvTeamThreadLimit > 0 &&
           DeviceData[DeviceId].ThreadsPerBlock > EnvTeamThreadLimit) {
         DP("Max CUDA threads per block %d exceeds the thread limit %d set by "
            "OMP_TEAMS_THREAD_LIMIT, capping at the limit\n",
            DeviceData[DeviceId].ThreadsPerBlock, EnvTeamThreadLimit);
         DeviceData[DeviceId].ThreadsPerBlock = EnvTeamThreadLimit;
       }
-      if (DeviceData[DeviceId].ThreadsPerBlock > DeviceRTLTy::HardThreadLimit) {
-        DP("Max CUDA threads per block %d exceeds the hard thread limit %d, "
-           "capping at the hard limit\n",
-           DeviceData[DeviceId].ThreadsPerBlock, DeviceRTLTy::HardThreadLimit);
-        DeviceData[DeviceId].ThreadsPerBlock = DeviceRTLTy::HardThreadLimit;
-      }
     }
 
     // Get and set warp size
     int WarpSize;
     Err =
         cuDeviceGetAttribute(&WarpSize, CU_DEVICE_ATTRIBUTE_WARP_SIZE, Device);
     if (Err != CUDA_SUCCESS) {
       DP("Error getting warp size, assume default value 32\n");