Skip to content

Commit 5eeb3e2

Browse files
authored
Sets the value for rpc: in LIBOMPTARGET_KERNEL_TRACE (llvm#813)
2 parents 0f7a60b + fc4c76c commit 5eeb3e2

File tree

1 file changed

+7
-1
lines changed
  • offload/plugins-nextgen/amdgpu/src

1 file changed

+7
-1
lines changed

offload/plugins-nextgen/amdgpu/src/rtl.cpp

Lines changed: 7 additions & 1 deletion
Original file line numberDiff line numberDiff line change
@@ -810,6 +810,8 @@ struct AMDGPUKernelTy : public GenericKernelTy {
810810
INFO(OMP_INFOTYPE_PLUGIN_KERNEL, Device.getDeviceId(),
811811
"Could not read extra information for kernel %s.", getName());
812812

813+
HasRPC = AMDImage.hasDeviceSymbol(Device, "__llvm_rpc_client");
814+
813815
return Plugin::success();
814816
}
815817

@@ -865,6 +867,10 @@ struct AMDGPUKernelTy : public GenericKernelTy {
865867
uint32_t PrivateSize;
866868
bool DynamicStack;
867869

870+
/// Device init sets this to true if image has symbol indicating that RPC
871+
/// service threads are used in this image. Only used for trace display.
872+
bool HasRPC;
873+
868874
/// The size of implicit kernel arguments.
869875
uint32_t ImplicitArgsSize;
870876

@@ -5119,7 +5125,7 @@ void AMDGPUKernelTy::printAMDOneLineKernelTrace(GenericDeviceTy &GenericDevice,
51195125
GenericDevice.getDeviceId(), getExecutionModeFlags(), ConstWGSize,
51205126
KernelArgs.NumArgs, NumBlocks[0], NumThreads[0], 0, 0, GroupSegmentSize,
51215127
SGPRCount, VGPRCount, AGPRCount, SGPRSpillCount, VGPRSpillCount,
5122-
KernelArgs.Tripcount, /*FIXME*/ 0, isMultiDeviceKernel(), MultiDeviceLB,
5128+
KernelArgs.Tripcount, HasRPC, isMultiDeviceKernel(), MultiDeviceLB,
51235129
MultiDeviceUB, MaxOccupancy, AchievedOccupancy, getName());
51245130
}
51255131

0 commit comments

Comments
 (0)