[libomptarget][amdgpu] Address compiler warnings, drive by fixes

[JonChesterfield]

Summary

[libomptarget][amdgpu] Address compiler warnings, drive by fixes

Initialize some variables, remove unused ones.
Changes the debug printing condition to align with the aomp test suite.

[pdhaliwal]

Looks good, thanks.

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

 /// Use a single entity to encode a kernel and a set of flags
 struct KernelTy {
   // execution mode of kernel
   // 0 - SPMD mode (without master warp)
   // 1 - Generic mode (with master warp)
   int8_t ExecutionMode;
   int16_t ConstWGSize;
   int32_t device_id;
-  void *CallStackAddr;
+  void *CallStackAddr = nullptr;
   const char *Name;
 
   KernelTy(int8_t _ExecutionMode, int16_t _ConstWGSize, int32_t _device_id,
            void *_CallStackAddr, const char *_Name,
            uint32_t _kernarg_segment_size)
       : ExecutionMode(_ExecutionMode), ConstWGSize(_ConstWGSize),
         device_id(_device_id), CallStackAddr(_CallStackAddr), Name(_Name) {
     DP("Construct kernelinfo: ExecMode %d\n", ExecutionMode);
void operator()(void *p) {
       atmi_free(p); // ignore failure to free
     }
   };
 
   // device_State shared across loaded binaries, error if inconsistent size
   std::vector<std::pair<std::unique_ptr<void, atmiFreePtrDeletor>, uint64_t>>
       deviceStateStore;
 
-  static const int HardTeamLimit = 1 << 20; // 1 Meg
+  static const unsigned HardTeamLimit =
+      (1 << 16) - 1; // 64K needed to fit in uint16
   static const int DefaultNumTeams = 128;
   static const int Max_Teams =
       llvm::omp::AMDGPUGpuGridValues[llvm::omp::GVIDX::GV_Max_Teams];
   static const int Warp_Size =
       llvm::omp::AMDGPUGpuGridValues[llvm::omp::GVIDX::GV_Warp_Size];
   static const int Max_WG_Size =
       llvm::omp::AMDGPUGpuGridValues[llvm::omp::GVIDX::GV_Max_WG_Size];
   static const int Default_WG_Size =
err = hsa_agent_get_info(
       &compute_units);
   if (err != HSA_STATUS_SUCCESS) {
     DeviceInfo.ComputeUnits[device_id] = 1;
     DP("Error getting compute units : settiing to 1\n");
   } else {
     DeviceInfo.ComputeUnits[device_id] = compute_units;
     DP("Using %d compute unis per grid\n", DeviceInfo.ComputeUnits[device_id]);
   }
-  if (print_kernel_trace > 1)
+  if (print_kernel_trace == 4)
     fprintf(stderr, "Device#%-2d CU's: %2d\n", device_id,
             DeviceInfo.ComputeUnits[device_id]);
 
   // Query attributes to determine number of threads/block and blocks/grid.
   uint16_t workgroup_max_dim[3];
   err = hsa_agent_get_info(agent, HSA_AGENT_INFO_WORKGROUP_MAX_DIM,
                            &workgroup_max_dim);
   if (err != HSA_STATUS_SUCCESS) {
__tgt_target_table *__tgt_rtl_load_binary(int32_t device_id,
   DeviceInfo.load_run_lock.lock();
   __tgt_target_table *res = __tgt_rtl_load_binary_locked(device_id, image);
   DeviceInfo.load_run_lock.unlock();
   return res;
 }
 
 __tgt_target_table *__tgt_rtl_load_binary_locked(int32_t device_id,
                                                  __tgt_device_image *image) {
+  // This function loads the device image onto gpu[device_id] and does other
+  // per-image initialization work. Specifically:
+  //
+  // - Initialize an omptarget_device_environmentTy instance embedded in the
+  //   image at the symbol "omptarget_device_environment"
+  //   Fields debug_level, device_num, num_devices. Used by the deviceRTL.
+  //
+  // - Allocate a large array per-gpu (could be moved to init_device)
+  //   - Read a uint64_t at symbol omptarget_nvptx_device_State_size
+  //   - Allocate at least that many bytes of gpu memory
+  //   - Zero initialize it
+  //   - Write the pointer to the symbol omptarget_nvptx_device_State
+  //
+  // - Pulls some per-kernel information together from various sources and
+  //   records it in the KernelsList for quicker access later
+  //
+  // The initialization can be done before or after loading the image onto the
+  // gpu. This function presently does a mixture. Using the hsa api to get/set
+  // the information is simpler to implement, in exchange for more complicated
+  // runtime behaviour. E.g. launching a kernel or using dma to get eight bytes
+  // back from the gpu vs a hashtable lookup on the host.
 
   const size_t img_size = (char *)image->ImageEnd - (char *)image->ImageStart;
 
   DeviceInfo.clearOffloadEntriesTable(device_id);
 
   // We do not need to set the ELF version because the caller of this function
   // had to do that to decide the right runtime to use
 
if (rc != 0) {
       DP("Finding global device environment '%s' - symbol missing.\n",
          device_env_Name);
       // no need to return FAIL, consider this is a not a device debug build.
       return ATMI_STATUS_SUCCESS;
     }
     if (si.size != sizeof(host_device_env)) {
       return ATMI_STATUS_ERROR;
     }
-    DP("Setting global device environment %lu bytes\n", si.size);
+    DP("Setting global device environment %u bytes\n", si.size);
     uint64_t offset = (char *)si.addr - (char *)image->ImageStart;
     void *pos = (char *)data + offset;
     memcpy(pos, &host_device_env, sizeof(host_device_env));
     return ATMI_STATUS_SUCCESS;
   };
 
   atmi_status_t err;
   {
for (__tgt_offload_entry *e = HostBegin; e != HostEnd; ++e) {
 
     // get Kernel Descriptor if present.
     // Keep struct in sync wih getTgtAttributeStructQTy in CGOpenMPRuntime.cpp
     struct KernDescValType {
       uint16_t Version;
       uint16_t TSize;
       uint16_t WG_Size;
       uint8_t Mode;
-      uint8_t HostServices;
     };
     struct KernDescValType KernDescVal;
     std::string KernDescNameStr(e->name);
     KernDescNameStr += "_kern_desc";
     const char *KernDescName = KernDescNameStr.c_str();
 
     void *KernDescPtr;
     uint32_t KernDescSize;
-    void *CallStackAddr;
+    void *CallStackAddr = nullptr;
     err = interop_get_symbol_info((char *)image->ImageStart, img_size,
                                   KernDescName, &KernDescPtr, &KernDescSize);
 
     if (err == ATMI_STATUS_SUCCESS) {
       if ((size_t)KernDescSize != sizeof(KernDescVal))
         DP("Loading global computation properties '%s' - size mismatch (%u != "
            "%lu)\n",
            KernDescName, KernDescSize, sizeof(KernDescVal));
 
       memcpy(&KernDescVal, KernDescPtr, (size_t)KernDescSize);
 
       // Check structure size against recorded size.
       if ((size_t)KernDescSize != KernDescVal.TSize)
         DP("KernDescVal size %lu does not match advertized size %d for '%s'\n",
            sizeof(KernDescVal), KernDescVal.TSize, KernDescName);
 
       DP("After loading global for %s KernDesc \n", KernDescName);
       DP("KernDesc: Version: %d\n", KernDescVal.Version);
       DP("KernDesc: TSize: %d\n", KernDescVal.TSize);
       DP("KernDesc: WG_Size: %d\n", KernDescVal.WG_Size);
       DP("KernDesc: Mode: %d\n", KernDescVal.Mode);
-      DP("KernDesc: HostServices: %x\n", KernDescVal.HostServices);
 
       // Get ExecMode
       ExecModeVal = KernDescVal.Mode;
       DP("ExecModeVal %d\n", ExecModeVal);
       if (KernDescVal.WG_Size == 0) {
         KernDescVal.WG_Size = RTLDeviceInfoTy::Default_WG_Size;
         DP("Setting KernDescVal.WG_Size to default %d\n", KernDescVal.WG_Size);
       }
void getLaunchVals(int &threadsPerGroup, int &num_groups, int ConstWGSize,
                    int num_teams, int thread_limit, uint64_t loop_tripcount) {
 
   int Max_Teams = DeviceInfo.EnvMaxTeamsDefault > 0
                       ? DeviceInfo.EnvMaxTeamsDefault
                       : DeviceInfo.Max_Teams;
   if (Max_Teams > DeviceInfo.HardTeamLimit)
     Max_Teams = DeviceInfo.HardTeamLimit;
 
-  if (print_kernel_trace > 1) {
+  if (print_kernel_trace == 4) {
     fprintf(stderr, "RTLDeviceInfoTy::Max_Teams: %d\n",
             RTLDeviceInfoTy::Max_Teams);
     fprintf(stderr, "Max_Teams: %d\n", Max_Teams);
     fprintf(stderr, "RTLDeviceInfoTy::Warp_Size: %d\n",
             RTLDeviceInfoTy::Warp_Size);
     fprintf(stderr, "RTLDeviceInfoTy::Max_WG_Size: %d\n",
             RTLDeviceInfoTy::Max_WG_Size);
     fprintf(stderr, "RTLDeviceInfoTy::Default_WG_Size: %d\n",
if (thread_limit > 0) {
     }
   }
   // check flat_max_work_group_size attr here
   if (threadsPerGroup > ConstWGSize) {
     threadsPerGroup = ConstWGSize;
     DP("Reduced threadsPerGroup to flat-attr-group-size limit %d\n",
        threadsPerGroup);
   }
-  if (print_kernel_trace > 1)
+  if (print_kernel_trace == 4)
     fprintf(stderr, "threadsPerGroup: %d\n", threadsPerGroup);
   DP("Preparing %d threads\n", threadsPerGroup);
 
   // Set default num_groups (teams)
   if (DeviceInfo.EnvTeamLimit > 0)
     num_groups = (Max_Teams < DeviceInfo.EnvTeamLimit)
                      ? Max_Teams
                      : DeviceInfo.EnvTeamLimit;
   else
     num_groups = Max_Teams;
   DP("Set default num of groups %d\n", num_groups);
 
-  if (print_kernel_trace > 1) {
+  if (print_kernel_trace == 4) {
     fprintf(stderr, "num_groups: %d\n", num_groups);
     fprintf(stderr, "num_teams: %d\n", num_teams);
   }
 
   // Reduce num_groups if threadsPerGroup exceeds RTLDeviceInfoTy::Max_WG_Size
   // This reduction is typical for default case (no thread_limit clause).
   // or when user goes crazy with num_teams clause.
   // FIXME: We cant distinguish between a constant or variable thread limit.
   // So we only handle constant thread_limits.
   if (threadsPerGroup >
       RTLDeviceInfoTy::Default_WG_Size) //  256 < threadsPerGroup <= 1024
     // Should we round threadsPerGroup up to nearest RTLDeviceInfoTy::Warp_Size
     // here?
     num_groups = (Max_Teams * RTLDeviceInfoTy::Max_WG_Size) / threadsPerGroup;
 
   // check for num_teams() clause
   if (num_teams > 0) {
     num_groups = (num_teams < num_groups) ? num_teams : num_groups;
   }
-  if (print_kernel_trace > 1) {
+  if (print_kernel_trace == 4) {
     fprintf(stderr, "num_groups: %d\n", num_groups);
     fprintf(stderr, "DeviceInfo.EnvNumTeams %d\n", DeviceInfo.EnvNumTeams);
     fprintf(stderr, "DeviceInfo.EnvTeamLimit %d\n", DeviceInfo.EnvTeamLimit);
   }
 
   if (DeviceInfo.EnvNumTeams > 0) {
     num_groups = (DeviceInfo.EnvNumTeams < num_groups) ? DeviceInfo.EnvNumTeams
                                                        : num_groups;
if (num_teams <= 0) {
            "threads per block %d\n",
            num_groups, loop_tripcount, threadsPerGroup);
       }
     } else {
       num_groups = num_teams;
     }
     if (num_groups > Max_Teams) {
       num_groups = Max_Teams;
-      if (print_kernel_trace > 1)
+      if (print_kernel_trace == 4)
         fprintf(stderr, "Limiting num_groups %d to Max_Teams %d \n", num_groups,
                 Max_Teams);
     }
     if (num_groups > num_teams && num_teams > 0) {
       num_groups = num_teams;
-      if (print_kernel_trace > 1)
+      if (print_kernel_trace == 4)
         fprintf(stderr, "Limiting num_groups %d to clause num_teams %d \n",
                 num_groups, num_teams);
     }
   }
 
   // num_teams clause always honored, no matter what, unless DEFAULT is active.
   if (num_teams > 0) {
     num_groups = num_teams;
     // Cap num_groups to EnvMaxTeamsDefault if set.
     if (DeviceInfo.EnvMaxTeamsDefault > 0 &&
         num_groups > DeviceInfo.EnvMaxTeamsDefault)
       num_groups = DeviceInfo.EnvMaxTeamsDefault;
   }
-  if (print_kernel_trace > 1) {
+  if (print_kernel_trace == 4) {
     fprintf(stderr, "threadsPerGroup: %d\n", threadsPerGroup);
     fprintf(stderr, "num_groups: %d\n", num_groups);
     fprintf(stderr, "loop_tripcount: %ld\n", loop_tripcount);
   }
   DP("Final %d num_groups and %d threadsPerGroup\n", num_groups,
      threadsPerGroup);
 }
 
int32_t __tgt_rtl_run_target_team_region_locked(
   getLaunchVals(threadsPerGroup, num_groups, KernelInfo->ConstWGSize,
                 KernelInfo->ExecutionMode, DeviceInfo.EnvTeamLimit,
                 DeviceInfo.EnvNumTeams,
                 num_teams,     // From run_region arg
                 thread_limit,  // From run_region arg
                 loop_tripcount // From run_region arg
   );
 
-  if (print_kernel_trace > 0)
+  if (print_kernel_trace == 4)
     // enum modes are SPMD, GENERIC, NONE 0,1,2
     fprintf(stderr,
             "DEVID:%2d SGN:%1d ConstWGSize:%-4d args:%2d teamsXthrds:(%4dX%4d) "
             "reqd:(%4dX%4d) n:%s\n",
             device_id, KernelInfo->ExecutionMode, KernelInfo->ConstWGSize,
             arg_num, num_groups, threadsPerGroup, num_teams, thread_limit,
             KernelInfo->Name);