AMDGPU: Don't pass MachineFunction if only the IR Function is used

diff --git a/llvm/lib/Target/AMDGPU/AMDGPUAsmPrinter.cpp b/llvm/lib/Target/AMDGPU/AMDGPUAsmPrinter.cpp
index fe0462a..42bf515 100644 (file)
--- a/llvm/lib/Target/AMDGPU/AMDGPUAsmPrinter.cpp
+++ b/llvm/lib/Target/AMDGPU/AMDGPUAsmPrinter.cpp
@@ -1165,7 +1165,7 @@ void AMDGPUAsmPrinter::getSIProgramInfo(SIProgramInfo &ProgInfo,
       S_00B84C_LDS_SIZE(STM.isAmdHsaOS() ? 0 : ProgInfo.LDSBlocks) |
       S_00B84C_EXCP_EN(0);
 
-  ProgInfo.Occupancy = STM.computeOccupancy(MF, ProgInfo.LDSSize,
+  ProgInfo.Occupancy = STM.computeOccupancy(MF.getFunction(), ProgInfo.LDSSize,
                                             ProgInfo.NumSGPRsForWavesPerEU,
                                             ProgInfo.NumVGPRsForWavesPerEU);
 }

diff --git a/llvm/lib/Target/AMDGPU/AMDGPUSubtarget.cpp b/llvm/lib/Target/AMDGPU/AMDGPUSubtarget.cpp
index 64e1e8f..a30e981 100644 (file)
--- a/llvm/lib/Target/AMDGPU/AMDGPUSubtarget.cpp
+++ b/llvm/lib/Target/AMDGPU/AMDGPUSubtarget.cpp
@@ -635,13 +635,12 @@ unsigned GCNSubtarget::getReservedNumSGPRs(const MachineFunction &MF) const {
   return 2; // VCC.
 }
 
-unsigned GCNSubtarget::computeOccupancy(const MachineFunction &MF,
-                                        unsigned LDSSize,
+unsigned GCNSubtarget::computeOccupancy(const Function &F, unsigned LDSSize,
                                         unsigned NumSGPRs,
                                         unsigned NumVGPRs) const {
   unsigned Occupancy =
     std::min(getMaxWavesPerEU(),
-             getOccupancyWithLocalMemSize(LDSSize, MF.getFunction()));
+             getOccupancyWithLocalMemSize(LDSSize, F));
   if (NumSGPRs)
     Occupancy = std::min(Occupancy, getOccupancyWithNumSGPRs(NumSGPRs));
   if (NumVGPRs)

diff --git a/llvm/lib/Target/AMDGPU/AMDGPUSubtarget.h b/llvm/lib/Target/AMDGPU/AMDGPUSubtarget.h
index 34a7efd..c833bfb 100644 (file)
--- a/llvm/lib/Target/AMDGPU/AMDGPUSubtarget.h
+++ b/llvm/lib/Target/AMDGPU/AMDGPUSubtarget.h
@@ -1100,7 +1100,7 @@ public:
   /// registers if provided.
   /// Note, occupancy can be affected by the scratch allocation as well, but
   /// we do not have enough information to compute it.
-  unsigned computeOccupancy(const MachineFunction &MF, unsigned LDSSize = 0,
+  unsigned computeOccupancy(const Function &F, unsigned LDSSize = 0,
                             unsigned NumSGPRs = 0, unsigned NumVGPRs = 0) const;
 
   /// \returns true if the flat_scratch register should be initialized with the

diff --git a/llvm/lib/Target/AMDGPU/SIMachineFunctionInfo.cpp b/llvm/lib/Target/AMDGPU/SIMachineFunctionInfo.cpp
index 5515e15..d33193e 100644 (file)
--- a/llvm/lib/Target/AMDGPU/SIMachineFunctionInfo.cpp
+++ b/llvm/lib/Target/AMDGPU/SIMachineFunctionInfo.cpp
@@ -53,7 +53,7 @@ SIMachineFunctionInfo::SIMachineFunctionInfo(const MachineFunction &MF)
   FlatWorkGroupSizes = ST.getFlatWorkGroupSizes(F);
   WavesPerEU = ST.getWavesPerEU(F);
 
-  Occupancy = ST.computeOccupancy(MF, getLDSSize());
+  Occupancy = ST.computeOccupancy(F, getLDSSize());
   CallingConv::ID CC = F.getCallingConv();
 
   // FIXME: Should have analysis or something rather than attribute to detect