AMDGPU: Only use legal inline immediates with kill pseudo

Only if the value is negative or positive is what matters,
so use a constant that doesn't require an instruction to
materialize.

These should really just emit the write exec directly,
but for stick with the kill pseudo-terminator.

llvm-svn: 275988

5 files changed, 15 insertions, 3 deletions

diff --git a/llvm/lib/Target/AMDGPU/AMDGPUISelLowering.cpp b/llvm/lib/Target/AMDGPU/AMDGPUISelLowering.cpp
index e24f54e4b8f..a69e6693aec 100644
--- a/llvm/lib/Target/AMDGPU/AMDGPUISelLowering.cpp
+++ b/llvm/lib/Target/AMDGPU/AMDGPUISelLowering.cpp
@@ -2834,6 +2834,7 @@ const char* AMDGPUTargetLowering::getTargetNodeName(unsigned Opcode) const {
   NODE_NAME_CASE(BUILD_VERTICAL_VECTOR)
   NODE_NAME_CASE(CONST_DATA_PTR)
   NODE_NAME_CASE(PC_ADD_REL_OFFSET)
+  NODE_NAME_CASE(KILL)
   case AMDGPUISD::FIRST_MEM_OPCODE_NUMBER: break;
   NODE_NAME_CASE(SENDMSG)
   NODE_NAME_CASE(INTERP_MOV)
diff --git a/llvm/lib/Target/AMDGPU/AMDGPUISelLowering.h b/llvm/lib/Target/AMDGPU/AMDGPUISelLowering.h
index f9376945aa6..071390beb93 100644
--- a/llvm/lib/Target/AMDGPU/AMDGPUISelLowering.h
+++ b/llvm/lib/Target/AMDGPU/AMDGPUISelLowering.h
@@ -303,6 +303,7 @@ enum NodeType : unsigned {
   INTERP_P1,
   INTERP_P2,
   PC_ADD_REL_OFFSET,
+  KILL,
   FIRST_MEM_OPCODE_NUMBER = ISD::FIRST_TARGET_MEMORY_OPCODE,
   STORE_MSKOR,
   LOAD_CONSTANT,
diff --git a/llvm/lib/Target/AMDGPU/AMDGPUInstrInfo.td b/llvm/lib/Target/AMDGPU/AMDGPUInstrInfo.td
index cf087b7ebc3..85675c800cf 100644
--- a/llvm/lib/Target/AMDGPU/AMDGPUInstrInfo.td
+++ b/llvm/lib/Target/AMDGPU/AMDGPUInstrInfo.td
@@ -40,6 +40,8 @@ def AMDGPUFmasOp : SDTypeProfile<1, 4,
   [SDTCisFP<0>, SDTCisSameAs<0, 1>, SDTCisSameAs<0, 2>, SDTCisSameAs<0, 3>, SDTCisInt<4>]
 >;
 
+def AMDGPUKillSDT : SDTypeProfile<0, 1, [SDTCisInt<0>]>;
+
 //===----------------------------------------------------------------------===//
 // AMDGPU DAG Nodes
 //
@@ -246,6 +248,9 @@ def AMDGPUinterp_p2 : SDNode<"AMDGPUISD::INTERP_P2",
                       SDTypeProfile<1, 4, [SDTCisFP<0>]>,
                       [SDNPInGlue]>;
 
+def AMDGPUkill : SDNode<"AMDGPUISD::KILL", AMDGPUKillSDT,
+  [SDNPHasChain, SDNPSideEffect]>;
+
 //===----------------------------------------------------------------------===//
 // Flow Control Profile Types
 //===----------------------------------------------------------------------===//
diff --git a/llvm/lib/Target/AMDGPU/SIISelLowering.cpp b/llvm/lib/Target/AMDGPU/SIISelLowering.cpp
index 7da2e081b76..a2c4e6a4056 100644
--- a/llvm/lib/Target/AMDGPU/SIISelLowering.cpp
+++ b/llvm/lib/Target/AMDGPU/SIISelLowering.cpp
@@ -2290,12 +2290,17 @@ SDValue SITargetLowering::LowerINTRINSIC_VOID(SDValue Op,
                                    Op->getVTList(), Ops, VT, MMO);
   }
   case AMDGPUIntrinsic::AMDGPU_kill: {
-    if (const ConstantFPSDNode *K = dyn_cast<ConstantFPSDNode>(Op.getOperand(2))) {
+    SDValue Src = Op.getOperand(2);
+    if (const ConstantFPSDNode *K = dyn_cast<ConstantFPSDNode>(Src)) {
       if (!K->isNegative())
         return Chain;
+
+      SDValue NegOne = DAG.getTargetConstant(FloatToBits(-1.0f), DL, MVT::i32);
+      return DAG.getNode(AMDGPUISD::KILL, DL, MVT::Other, Chain, NegOne);
     }
 
-    return Op;
+    SDValue Cast = DAG.getNode(ISD::BITCAST, DL, MVT::i32, Src);
+    return DAG.getNode(AMDGPUISD::KILL, DL, MVT::Other, Chain, Cast);
   }
   default:
     return SDValue();
diff --git a/llvm/lib/Target/AMDGPU/SIInstructions.td b/llvm/lib/Target/AMDGPU/SIInstructions.td
index a6485069567..f1e8c232737 100644
--- a/llvm/lib/Target/AMDGPU/SIInstructions.td
+++ b/llvm/lib/Target/AMDGPU/SIInstructions.td
@@ -1996,7 +1996,7 @@ def SI_END_CF : PseudoInstSI <
 let Uses = [EXEC], Defs = [EXEC,VCC] in {
 def SI_KILL : PseudoInstSI <
   (outs), (ins VSrc_32:$src),
-  [(int_AMDGPU_kill f32:$src)]> {
+  [(AMDGPUkill i32:$src)]> {
   let isConvergent = 1;
   let usesCustomInserter = 1;
 }