AMDGPU: Detect kernarg segment pointer

This is necessary to pass the kernarg segment pointer
to callee functions. Also don't unconditionally enable
for kernels.

llvm-svn: 307978

2 files changed, 9 insertions, 2 deletions

diff --git a/llvm/lib/Target/AMDGPU/AMDGPUAnnotateKernelFeatures.cpp b/llvm/lib/Target/AMDGPU/AMDGPUAnnotateKernelFeatures.cpp
index c837180334c..d1580ff4e19 100644
--- a/llvm/lib/Target/AMDGPU/AMDGPUAnnotateKernelFeatures.cpp
+++ b/llvm/lib/Target/AMDGPU/AMDGPUAnnotateKernelFeatures.cpp
@@ -150,6 +150,9 @@ static StringRef intrinsicToAttrName(Intrinsic::ID ID, bool &IsQueuePtr) {
     return "amdgpu-dispatch-ptr";
   case Intrinsic::amdgcn_dispatch_id:
     return "amdgpu-dispatch-id";
+  case Intrinsic::amdgcn_kernarg_segment_ptr:
+  case Intrinsic::amdgcn_implicitarg_ptr:
+    return "amdgpu-kernarg-segment-ptr";
   case Intrinsic::amdgcn_queue_ptr:
   case Intrinsic::trap:
   case Intrinsic::debugtrap:
@@ -181,7 +184,8 @@ static void copyFeaturesToFunction(Function &Parent, const Function &Callee,
     { "amdgpu-work-group-id-y" },
     { "amdgpu-work-group-id-z" },
     { "amdgpu-dispatch-ptr" },
-    { "amdgpu-dispatch-id" }
+    { "amdgpu-dispatch-id" },
+    { "amdgpu-kernarg-segment-ptr" }
   };
 
   if (handleAttr(Parent, Callee, "amdgpu-queue-ptr"))
diff --git a/llvm/lib/Target/AMDGPU/SIMachineFunctionInfo.cpp b/llvm/lib/Target/AMDGPU/SIMachineFunctionInfo.cpp
index 3203c38dae3..7fe671a533a 100644
--- a/llvm/lib/Target/AMDGPU/SIMachineFunctionInfo.cpp
+++ b/llvm/lib/Target/AMDGPU/SIMachineFunctionInfo.cpp
@@ -92,7 +92,7 @@ SIMachineFunctionInfo::SIMachineFunctionInfo(const MachineFunction &MF)
 
   CallingConv::ID CC = F->getCallingConv();
   if (CC == CallingConv::AMDGPU_KERNEL || CC == CallingConv::SPIR_KERNEL) {
-    KernargSegmentPtr = true;
+    KernargSegmentPtr = !F->arg_empty();
     WorkGroupIDX = true;
     WorkItemIDX = true;
   } else if (CC == CallingConv::AMDGPU_PS) {
@@ -154,6 +154,9 @@ SIMachineFunctionInfo::SIMachineFunctionInfo(const MachineFunction &MF)
       ImplicitBufferPtr = true;
   }
 
+  if (F->hasFnAttribute("amdgpu-kernarg-segment-ptr"))
+    KernargSegmentPtr = true;
+
   // We don't need to worry about accessing spills with flat instructions.
   // TODO: On VI where we must use flat for global, we should be able to omit
   // this if it is never used for generic access.