[AMDGPU] Add FP_CLASS to the add/setcc combine

This is one of the nodes which also compile as v_cmp_*.

Differential Revision: https://reviews.llvm.org/D34485

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@305970 91177308-0d34-0410-b5e6-96231b3b80d8

diff --git a/lib/Target/AMDGPU/SIISelLowering.cpp b/lib/Target/AMDGPU/SIISelLowering.cpp
index ad9c5e49320d9925bc0722fba0976068666c93f4..3563eb365b96ec83cc494d969e8a2f9596ff1cff 100644 (file)
--- a/lib/Target/AMDGPU/SIISelLowering.cpp
+++ b/lib/Target/AMDGPU/SIISelLowering.cpp
@@ -4872,7 +4872,9 @@ SDValue SITargetLowering::performAddCombine(SDNode *N,
   case ISD::SIGN_EXTEND:
   case ISD::ANY_EXTEND: {
     auto Cond = RHS.getOperand(0);
-    if (Cond.getOpcode() != ISD::SETCC) break;
+    if (Cond.getOpcode() != ISD::SETCC &&
+        Cond.getOpcode() != AMDGPUISD::FP_CLASS)
+      break;
     SDVTList VTList = DAG.getVTList(MVT::i32, MVT::i1);
     SDValue Args[] = { LHS, DAG.getConstant(0, SL, MVT::i32), Cond };
     Opc = (Opc == ISD::SIGN_EXTEND) ? ISD::SUBCARRY : ISD::ADDCARRY;

diff --git a/test/CodeGen/AMDGPU/combine-cond-add-sub.ll b/test/CodeGen/AMDGPU/combine-cond-add-sub.ll
index 6026a047d88e74b99c7f71dae3540d5d24220266..187fb24dfb66560942991ef7f7e857aad4613c63 100644 (file)
--- a/test/CodeGen/AMDGPU/combine-cond-add-sub.ll
+++ b/test/CodeGen/AMDGPU/combine-cond-add-sub.ll
@@ -116,6 +116,42 @@ bb:
   ret void
 }
 
+; GCN-LABEL: {{^}}zext_flclass:
+; GCN: v_cmp_class_f32_e{{32|64}} [[CC:[^,]+]],
+; GCN: v_addc_u32_e{{32|64}} v{{[0-9]+}}, {{[^,]+}}, 0, v{{[0-9]+}}, [[CC]]
+; GCN-NOT: v_cndmask
+
+define amdgpu_kernel void @zext_flclass(i32 addrspace(1)* nocapture %arg, float %x) {
+bb:
+  %id = tail call i32 @llvm.amdgcn.workitem.id.x()
+  %gep = getelementptr inbounds i32, i32 addrspace(1)* %arg, i32 %id
+  %v = load i32, i32 addrspace(1)* %gep, align 4
+  %cmp = tail call zeroext i1 @llvm.amdgcn.class.f32(float %x, i32 608)
+  %ext = zext i1 %cmp to i32
+  %add = add i32 %v, %ext
+  store i32 %add, i32 addrspace(1)* %gep, align 4
+  ret void
+}
+
+; GCN-LABEL: {{^}}sext_flclass:
+; GCN: v_cmp_class_f32_e{{32|64}} [[CC:[^,]+]],
+; GCN: v_subb_u32_e{{32|64}} v{{[0-9]+}}, {{[^,]+}}, v{{[0-9]+}}, 0, [[CC]]
+; GCN-NOT: v_cndmask
+
+define amdgpu_kernel void @sext_flclass(i32 addrspace(1)* nocapture %arg, float %x) {
+bb:
+  %id = tail call i32 @llvm.amdgcn.workitem.id.x()
+  %gep = getelementptr inbounds i32, i32 addrspace(1)* %arg, i32 %id
+  %v = load i32, i32 addrspace(1)* %gep, align 4
+  %cmp = tail call zeroext i1 @llvm.amdgcn.class.f32(float %x, i32 608)
+  %ext = sext i1 %cmp to i32
+  %add = add i32 %v, %ext
+  store i32 %add, i32 addrspace(1)* %gep, align 4
+  ret void
+}
+
+declare i1 @llvm.amdgcn.class.f32(float, i32) #0
+
 declare i32 @llvm.amdgcn.workitem.id.x() #0
 
 declare i32 @llvm.amdgcn.workitem.id.y() #0