[PATCH] R600: Custom lower f32 = uint_to_fp i64

[Tom Stellard]

From: Tom Stellard <thomas.stellard at amd.com>

---
 lib/Target/R600/AMDGPUISelLowering.cpp | 22 ++++++++++++++++++++++
 lib/Target/R600/AMDGPUISelLowering.h   |  1 +
 test/CodeGen/R600/uint_to_fp.ll        | 23 +++++++++++++++++++----
 3 files changed, 42 insertions(+), 4 deletions(-)

diff --git a/lib/Target/R600/AMDGPUISelLowering.cpp b/lib/Target/R600/AMDGPUISelLowering.cpp
index 009c94e..337dd18 100644
--- a/lib/Target/R600/AMDGPUISelLowering.cpp
+++ b/lib/Target/R600/AMDGPUISelLowering.cpp
@@ -136,6 +136,8 @@ AMDGPUTargetLowering::AMDGPUTargetLowering(TargetMachine &TM) :
   setOperationAction(ISD::FNEG, MVT::v2f32, Expand);
   setOperationAction(ISD::FNEG, MVT::v4f32, Expand);
 
+  setOperationAction(ISD::UINT_TO_FP, MVT::i64, Custom);
+
   setOperationAction(ISD::MUL, MVT::i64, Expand);
 
   setOperationAction(ISD::UDIV, MVT::i32, Expand);
@@ -254,6 +256,7 @@ SDValue AMDGPUTargetLowering::LowerOperation(SDValue Op, SelectionDAG &DAG)
   case ISD::FrameIndex: return LowerFrameIndex(Op, DAG);
   case ISD::INTRINSIC_WO_CHAIN: return LowerINTRINSIC_WO_CHAIN(Op, DAG);
   case ISD::UDIVREM: return LowerUDIVREM(Op, DAG);
+  case ISD::UINT_TO_FP: return LowerUINT_TO_FP(Op, DAG);
   }
   return Op;
 }
@@ -691,6 +694,25 @@ SDValue AMDGPUTargetLowering::LowerUDIVREM(SDValue Op,
   return DAG.getMergeValues(Ops, 2, DL);
 }
 
+SDValue AMDGPUTargetLowering::LowerUINT_TO_FP(SDValue Op,
+                                               SelectionDAG &DAG) const {
+  SDValue S0 = Op.getOperand(0);
+  SDLoc DL(Op);
+  if (Op.getValueType() != MVT::f32 || S0.getValueType() != MVT::i64)
+    return SDValue();
+
+  // f32 uint_to_fp i64
+  SDValue Lo = DAG.getNode(ISD::EXTRACT_ELEMENT, DL, MVT::i32, S0,
+                           DAG.getConstant(0, MVT::i32));
+  SDValue FloatLo = DAG.getNode(ISD::UINT_TO_FP, DL, MVT::f32, Lo);
+  SDValue Hi = DAG.getNode(ISD::EXTRACT_ELEMENT, DL, MVT::i32, S0,
+                           DAG.getConstant(1, MVT::i32));
+  SDValue FloatHi = DAG.getNode(ISD::UINT_TO_FP, DL, MVT::f32, Hi);
+  FloatHi = DAG.getNode(ISD::FMUL, DL, MVT::f32, FloatHi,
+                        DAG.getConstantFP(4294967296.0f, MVT::f32)); // 2^32
+  return DAG.getNode(ISD::FADD, DL, MVT::f32, FloatLo, FloatHi);
+
+}
 
 //===----------------------------------------------------------------------===//
 // Helper functions
diff --git a/lib/Target/R600/AMDGPUISelLowering.h b/lib/Target/R600/AMDGPUISelLowering.h
index e167f12..dacb086 100644
--- a/lib/Target/R600/AMDGPUISelLowering.h
+++ b/lib/Target/R600/AMDGPUISelLowering.h
@@ -38,6 +38,7 @@ private:
   /// \brief Split a vector store into multiple scalar stores.
   /// \returns The resulting chain. 
   SDValue LowerUDIVREM(SDValue Op, SelectionDAG &DAG) const;
+  SDValue LowerUINT_TO_FP(SDValue Op, SelectionDAG &DAG) const;
 
 protected:
 
diff --git a/test/CodeGen/R600/uint_to_fp.ll b/test/CodeGen/R600/uint_to_fp.ll
index 3f6e4c8..a5ac355 100644
--- a/test/CodeGen/R600/uint_to_fp.ll
+++ b/test/CodeGen/R600/uint_to_fp.ll
@@ -1,10 +1,10 @@
 ; RUN: llc < %s -march=r600 -mcpu=redwood | FileCheck %s --check-prefix=R600-CHECK
 ; RUN: llc < %s -march=r600 -mcpu=SI -verify-machineinstrs | FileCheck %s --check-prefix=SI-CHECK
 
-; R600-CHECK: @uint_to_fp_v2i32
+; R600-CHECK-LABEL: @uint_to_fp_v2i32
 ; R600-CHECK-DAG: UINT_TO_FLT * T{{[0-9]+\.[XYZW]}}, KC0[2].W
 ; R600-CHECK-DAG: UINT_TO_FLT * T{{[0-9]+\.[XYZW]}}, KC0[3].X
-; SI-CHECK: @uint_to_fp_v2i32
+; SI-CHECK-LABEL: @uint_to_fp_v2i32
 ; SI-CHECK: V_CVT_F32_U32_e32
 ; SI-CHECK: V_CVT_F32_U32_e32
 define void @uint_to_fp_v2i32(<2 x float> addrspace(1)* %out, <2 x i32> %in) {
@@ -13,12 +13,12 @@ define void @uint_to_fp_v2i32(<2 x float> addrspace(1)* %out, <2 x i32> %in) {
   ret void
 }
 
-; R600-CHECK: @uint_to_fp_v4i32
+; R600-CHECK-LABEL: @uint_to_fp_v4i32
 ; R600-CHECK: UINT_TO_FLT * T{{[0-9]+\.[XYZW], T[0-9]+\.[XYZW]}}
 ; R600-CHECK: UINT_TO_FLT * T{{[0-9]+\.[XYZW], T[0-9]+\.[XYZW]}}
 ; R600-CHECK: UINT_TO_FLT * T{{[0-9]+\.[XYZW], T[0-9]+\.[XYZW]}}
 ; R600-CHECK: UINT_TO_FLT * T{{[0-9]+\.[XYZW], T[0-9]+\.[XYZW]}}
-; SI-CHECK: @uint_to_fp_v4i32
+; SI-CHECK-LABEL: @uint_to_fp_v4i32
 ; SI-CHECK: V_CVT_F32_U32_e32
 ; SI-CHECK: V_CVT_F32_U32_e32
 ; SI-CHECK: V_CVT_F32_U32_e32
@@ -29,3 +29,18 @@ define void @uint_to_fp_v4i32(<4 x float> addrspace(1)* %out, <4 x i32> addrspac
   store <4 x float> %result, <4 x float> addrspace(1)* %out
   ret void
 }
+
+; R600-CHECK-LABEL: @uint_to_fp_i64_f32
+; R600-CHECK: UINT_TO_FLT
+; R600-CHECK: UINT_TO_FLT
+; R600-CHECK: MULADD_IEEE
+; SI-CHECK-LABEL: @uint_to_fp_i64_f32
+; SI-CHECK: V_CVT_F32_U32_e32
+; SI-CHECK: V_CVT_F32_U32_e32
+; SI-CHECK: V_MAD_F32
+define void @uint_to_fp_i64_f32(float addrspace(1)* %out, i64 %in) {
+entry:
+  %0 = uitofp i64 %in to float
+  store float %0, float addrspace(1)* %out
+  ret void
+}
-- 
1.7.11.4