AMDGPU: Eliminate BUFFER_ATOMIC_PK_ADD_F16 node

This is redundant with the other no return buffer atomic node, and we
don't really need a separate type profile for it.

diff --git a/llvm/lib/Target/AMDGPU/AMDGPUISelLowering.cpp b/llvm/lib/Target/AMDGPU/AMDGPUISelLowering.cpp
index 095018a..ce6c8fa 100644 (file)
--- a/llvm/lib/Target/AMDGPU/AMDGPUISelLowering.cpp
+++ b/llvm/lib/Target/AMDGPU/AMDGPUISelLowering.cpp
@@ -4371,7 +4371,6 @@ const char* AMDGPUTargetLowering::getTargetNodeName(unsigned Opcode) const {
   NODE_NAME_CASE(BUFFER_ATOMIC_CMPSWAP)
   NODE_NAME_CASE(BUFFER_ATOMIC_CSUB)
   NODE_NAME_CASE(BUFFER_ATOMIC_FADD)
-  NODE_NAME_CASE(BUFFER_ATOMIC_PK_FADD)
   NODE_NAME_CASE(ATOMIC_PK_FADD)
 
   case AMDGPUISD::LAST_AMDGPU_ISD_NUMBER: break;

diff --git a/llvm/lib/Target/AMDGPU/AMDGPUISelLowering.h b/llvm/lib/Target/AMDGPU/AMDGPUISelLowering.h
index e995fdc..932c442 100644 (file)
--- a/llvm/lib/Target/AMDGPU/AMDGPUISelLowering.h
+++ b/llvm/lib/Target/AMDGPU/AMDGPUISelLowering.h
@@ -536,7 +536,6 @@ enum NodeType : unsigned {
   BUFFER_ATOMIC_CMPSWAP,
   BUFFER_ATOMIC_CSUB,
   BUFFER_ATOMIC_FADD,
-  BUFFER_ATOMIC_PK_FADD,
   ATOMIC_PK_FADD,
 
   LAST_AMDGPU_ISD_NUMBER

diff --git a/llvm/lib/Target/AMDGPU/BUFInstructions.td b/llvm/lib/Target/AMDGPU/BUFInstructions.td
index bb04ae0..bf0177b 100644 (file)
--- a/llvm/lib/Target/AMDGPU/BUFInstructions.td
+++ b/llvm/lib/Target/AMDGPU/BUFInstructions.td
@@ -1432,7 +1432,7 @@ multiclass BufferAtomicPatterns_NO_RTN<SDPatternOperator name, ValueType vt,
 
 let SubtargetPredicate = HasAtomicFaddInsts in {
 defm : BufferAtomicPatterns_NO_RTN<SIbuffer_atomic_fadd, f32, "BUFFER_ATOMIC_ADD_F32">;
-defm : BufferAtomicPatterns_NO_RTN<SIbuffer_atomic_pk_fadd, v2f16, "BUFFER_ATOMIC_PK_ADD_F16">;
+defm : BufferAtomicPatterns_NO_RTN<SIbuffer_atomic_fadd, v2f16, "BUFFER_ATOMIC_PK_ADD_F16">;
 }
 
 def : GCNPat<

diff --git a/llvm/lib/Target/AMDGPU/SIISelLowering.cpp b/llvm/lib/Target/AMDGPU/SIISelLowering.cpp
index 74c3527..877ece0 100644 (file)
--- a/llvm/lib/Target/AMDGPU/SIISelLowering.cpp
+++ b/llvm/lib/Target/AMDGPU/SIISelLowering.cpp
@@ -7531,10 +7531,9 @@ SDValue SITargetLowering::LowerINTRINSIC_VOID(SDValue Op,
 
     auto *M = cast<MemSDNode>(Op);
     M->getMemOperand()->setOffset(Offset);
-    unsigned Opcode = VT.isVector() ? AMDGPUISD::BUFFER_ATOMIC_PK_FADD
-                                    : AMDGPUISD::BUFFER_ATOMIC_FADD;
 
-    return DAG.getMemIntrinsicNode(Opcode, DL, Op->getVTList(), Ops, VT,
+    return DAG.getMemIntrinsicNode(AMDGPUISD::BUFFER_ATOMIC_FADD, DL,
+                                   Op->getVTList(), Ops, VT,
                                    M->getMemOperand());
   }
 

diff --git a/llvm/lib/Target/AMDGPU/SIInstrInfo.td b/llvm/lib/Target/AMDGPU/SIInstrInfo.td
index d146114..b882508 100644 (file)
--- a/llvm/lib/Target/AMDGPU/SIInstrInfo.td
+++ b/llvm/lib/Target/AMDGPU/SIInstrInfo.td
@@ -173,10 +173,9 @@ class SDBufferAtomic<string opcode> : SDNode <opcode,
   [SDNPMemOperand, SDNPHasChain, SDNPMayLoad, SDNPMayStore]
 >;
 
-class SDBufferAtomicNoRtn<string opcode, ValueType ty> : SDNode <opcode,
+class SDBufferAtomicNoRtn<string opcode> : SDNode <opcode,
   SDTypeProfile<0, 8,
-      [SDTCisVT<0, ty>,    // vdata
-       SDTCisVT<1, v4i32>, // rsrc
+      [SDTCisVT<1, v4i32>, // rsrc
        SDTCisVT<2, i32>,   // vindex(VGPR)
        SDTCisVT<3, i32>,   // voffset(VGPR)
        SDTCisVT<4, i32>,   // soffset(SGPR)
@@ -199,8 +198,7 @@ def SIbuffer_atomic_xor : SDBufferAtomic <"AMDGPUISD::BUFFER_ATOMIC_XOR">;
 def SIbuffer_atomic_inc : SDBufferAtomic <"AMDGPUISD::BUFFER_ATOMIC_INC">;
 def SIbuffer_atomic_dec : SDBufferAtomic <"AMDGPUISD::BUFFER_ATOMIC_DEC">;
 def SIbuffer_atomic_csub : SDBufferAtomic <"AMDGPUISD::BUFFER_ATOMIC_CSUB">;
-def SIbuffer_atomic_fadd : SDBufferAtomicNoRtn <"AMDGPUISD::BUFFER_ATOMIC_FADD", f32>;
-def SIbuffer_atomic_pk_fadd : SDBufferAtomicNoRtn <"AMDGPUISD::BUFFER_ATOMIC_PK_FADD", v2f16>;
+def SIbuffer_atomic_fadd : SDBufferAtomicNoRtn <"AMDGPUISD::BUFFER_ATOMIC_FADD">;
 
 def SIbuffer_atomic_cmpswap : SDNode <"AMDGPUISD::BUFFER_ATOMIC_CMPSWAP",
   SDTypeProfile<1, 9,