AMDGPU: Add llvm.amdgcn.interp.mov intrinsic

Reviewers: arsenm, nhaehnle

Subscribers: kzhuravl, wdng, yaxunl, tony-tye, llvm-commits

Differential Revision: https://reviews.llvm.org/D26725

llvm-svn: 288865

diff --git a/llvm/include/llvm/IR/IntrinsicsAMDGPU.td b/llvm/include/llvm/IR/IntrinsicsAMDGPU.td
index 4ce5d87..078959c 100644 (file)
--- a/llvm/include/llvm/IR/IntrinsicsAMDGPU.td
+++ b/llvm/include/llvm/IR/IntrinsicsAMDGPU.td
@@ -475,6 +475,14 @@ def int_amdgcn_s_getreg :
   GCCBuiltin<"__builtin_amdgcn_s_getreg">,
   Intrinsic<[llvm_i32_ty], [llvm_i32_ty], [IntrReadMem]>;
 
+// __builtin_amdgcn_interp_mov <param>, <attr_chan>, <attr>, <m0>
+// param values: 0 = P10, 1 = P20, 2 = P0
+def int_amdgcn_interp_mov :
+  GCCBuiltin<"__builtin_amdgcn_interp_mov">,
+  Intrinsic<[llvm_float_ty],
+            [llvm_i32_ty, llvm_i32_ty, llvm_i32_ty, llvm_i32_ty],
+            [IntrNoMem]>;
+
 // __builtin_amdgcn_interp_p1 <i>, <attr_chan>, <attr>, <m0>
 def int_amdgcn_interp_p1 :
   GCCBuiltin<"__builtin_amdgcn_interp_p1">,

diff --git a/llvm/lib/Target/AMDGPU/SIISelLowering.cpp b/llvm/lib/Target/AMDGPU/SIISelLowering.cpp
index 64f2c0a..535db8f 100644 (file)
--- a/llvm/lib/Target/AMDGPU/SIISelLowering.cpp
+++ b/llvm/lib/Target/AMDGPU/SIISelLowering.cpp
@@ -2497,6 +2497,12 @@ SDValue SITargetLowering::LowerINTRINSIC_WO_CHAIN(SDValue Op,
     return DAG.getNode(AMDGPUISD::INTERP_P2, DL, MVT::f32, P1, J,
                              Op.getOperand(1), Op.getOperand(2), Glue);
   }
+  case Intrinsic::amdgcn_interp_mov: {
+    SDValue M0 = copyToM0(DAG, DAG.getEntryNode(), DL, Op.getOperand(4));
+    SDValue Glue = M0.getValue(1);
+    return DAG.getNode(AMDGPUISD::INTERP_MOV, DL, MVT::f32, Op.getOperand(1),
+                       Op.getOperand(2), Op.getOperand(3), Glue);
+  }
   case Intrinsic::amdgcn_interp_p1: {
     SDValue M0 = copyToM0(DAG, DAG.getEntryNode(), DL, Op.getOperand(4));
     SDValue Glue = M0.getValue(1);

diff --git a/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.interp.ll b/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.interp.ll
index 74f70e5..9613a50 100644 (file)
--- a/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.interp.ll
+++ b/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.interp.ll
@@ -14,7 +14,9 @@ main_body:
   %p1_0 = call float @llvm.amdgcn.interp.p2(float %p0_0, float %j, i32 0, i32 0, i32 %3)
   %p0_1 = call float @llvm.amdgcn.interp.p1(float %i, i32 1, i32 0, i32 %3)
   %p1_1 = call float @llvm.amdgcn.interp.p2(float %p0_1, float %j, i32 1, i32 0, i32 %3)
-  call void @llvm.SI.export(i32 15, i32 1, i32 1, i32 0, i32 1, float %p0_0, float %p0_0, float %p1_1, float %p1_1)
+  %const = call float @llvm.amdgcn.interp.mov(i32 2, i32 0, i32 0, i32 %3)
+  %w = fadd float %p1_1, %const
+  call void @llvm.SI.export(i32 15, i32 1, i32 1, i32 0, i32 1, float %p0_0, float %p0_0, float %p1_1, float %w)
   ret void
 }
 
@@ -24,6 +26,8 @@ declare float @llvm.amdgcn.interp.p1(float, i32, i32, i32) #0
 ; Function Attrs: nounwind readnone
 declare float @llvm.amdgcn.interp.p2(float, float, i32, i32, i32) #0
 
+declare float @llvm.amdgcn.interp.mov(i32, i32, i32, i32) #0
+
 declare void @llvm.SI.export(i32, i32, i32, i32, i32, float, float, float, float)
 
 attributes #0 = { nounwind readnone }