MFC r265925:

Upgrade our copy of llvm/clang to 3.4.1 release.  This release contains
mostly fixes, for the following upstream bugs:

http://llvm.org/PR16365 http://llvm.org/PR17473 http://llvm.org/PR18000
http://llvm.org/PR18068 http://llvm.org/PR18102 http://llvm.org/PR18165
http://llvm.org/PR18260 http://llvm.org/PR18290 http://llvm.org/PR18316
http://llvm.org/PR18460 http://llvm.org/PR18473 http://llvm.org/PR18515
http://llvm.org/PR18526 http://llvm.org/PR18600 http://llvm.org/PR18762
http://llvm.org/PR18773 http://llvm.org/PR18860 http://llvm.org/PR18994
http://llvm.org/PR19007 http://llvm.org/PR19010 http://llvm.org/PR19033
http://llvm.org/PR19059 http://llvm.org/PR19144 http://llvm.org/PR19326

1 files changed, 30 insertions, 12 deletions

diff --git a/contrib/llvm/lib/Target/R600/SILowerControlFlow.cpp b/contrib/llvm/lib/Target/R600/SILowerControlFlow.cpp
index 958763d..ef867d3 100644
--- a/contrib/llvm/lib/Target/R600/SILowerControlFlow.cpp
+++ b/contrib/llvm/lib/Target/R600/SILowerControlFlow.cpp
@@ -109,6 +109,23 @@ FunctionPass *llvm::createSILowerControlFlowPass(TargetMachine &tm) {
   return new SILowerControlFlowPass(tm);
 }
 
+static bool isDS(unsigned Opcode) {
+  switch(Opcode) {
+  default: return false;
+  case AMDGPU::DS_ADD_U32_RTN:
+  case AMDGPU::DS_SUB_U32_RTN:
+  case AMDGPU::DS_WRITE_B32:
+  case AMDGPU::DS_WRITE_B8:
+  case AMDGPU::DS_WRITE_B16:
+  case AMDGPU::DS_READ_B32:
+  case AMDGPU::DS_READ_I8:
+  case AMDGPU::DS_READ_U8:
+  case AMDGPU::DS_READ_I16:
+  case AMDGPU::DS_READ_U16:
+    return true;
+  }
+}
+
 bool SILowerControlFlowPass::shouldSkip(MachineBasicBlock *From,
                                         MachineBasicBlock *To) {
 
@@ -145,7 +162,9 @@ void SILowerControlFlowPass::SkipIfDead(MachineInstr &MI) {
   MachineBasicBlock &MBB = *MI.getParent();
   DebugLoc DL = MI.getDebugLoc();
 
-  if (!shouldSkip(&MBB, &MBB.getParent()->back()))
+  if (MBB.getParent()->getInfo<SIMachineFunctionInfo>()->ShaderType !=
+      ShaderType::PIXEL ||
+      !shouldSkip(&MBB, &MBB.getParent()->back()))
     return;
 
   MachineBasicBlock::iterator Insert = &MI;
@@ -296,9 +315,11 @@ void SILowerControlFlowPass::Kill(MachineInstr &MI) {
   MachineBasicBlock &MBB = *MI.getParent();
   DebugLoc DL = MI.getDebugLoc();
 
-  // Kill is only allowed in pixel shaders
+  // Kill is only allowed in pixel / geometry shaders
   assert(MBB.getParent()->getInfo<SIMachineFunctionInfo>()->ShaderType ==
-         ShaderType::PIXEL);
+         ShaderType::PIXEL ||
+         MBB.getParent()->getInfo<SIMachineFunctionInfo>()->ShaderType ==
+         ShaderType::GEOMETRY);
 
   // Clear this pixel from the exec mask if the operand is negative
   BuildMI(MBB, &MI, DL, TII->get(AMDGPU::V_CMPX_LE_F32_e32), AMDGPU::VCC)
@@ -431,6 +452,11 @@ bool SILowerControlFlowPass::runOnMachineFunction(MachineFunction &MF) {
 
       Next = llvm::next(I);
       MachineInstr &MI = *I;
+      if (isDS(MI.getOpcode())) {
+        NeedM0 = true;
+        NeedWQM = true;
+      }
+
       switch (MI.getOpcode()) {
         default: break;
         case AMDGPU::SI_IF:
@@ -491,14 +517,6 @@ bool SILowerControlFlowPass::runOnMachineFunction(MachineFunction &MF) {
           IndirectDst(MI);
           break;
 
-        case AMDGPU::DS_READ_B32:
-          NeedWQM = true;
-          // Fall through
-        case AMDGPU::DS_WRITE_B32:
-        case AMDGPU::DS_ADD_U32_RTN:
-          NeedM0 = true;
-          break;
-
         case AMDGPU::V_INTERP_P1_F32:
         case AMDGPU::V_INTERP_P2_F32:
         case AMDGPU::V_INTERP_MOV_F32:
@@ -517,7 +535,7 @@ bool SILowerControlFlowPass::runOnMachineFunction(MachineFunction &MF) {
             AMDGPU::M0).addImm(0xffffffff);
   }
 
-  if (NeedWQM && MFI->ShaderType != ShaderType::COMPUTE) {
+  if (NeedWQM && MFI->ShaderType == ShaderType::PIXEL) {
     MachineBasicBlock &MBB = MF.front();
     BuildMI(MBB, MBB.getFirstNonPHI(), DebugLoc(), TII->get(AMDGPU::S_WQM_B64),
             AMDGPU::EXEC).addReg(AMDGPU::EXEC);