change approach

Author: Pierre-vh

llvm/lib/CodeGen/SelectionDAG/DAGCombiner.cpp

@@ -15262,21 +15262,31 @@ SDValue DAGCombiner::visitAssertExt(SDNode *N) {
     }
   }
 
-  // If we have (AssertZext (and (AssertSext X, iX), M), iY) and Y is smaller
-  // than X, and the And doesn't change the lower iX bits, we can move the
-  // AssertZext in front of the And and drop the AssertSext.
   if (Opcode == ISD::AssertZext && N0.getOpcode() == ISD::AND &&
-      N0.hasOneUse() && N0.getOperand(0).getOpcode() == ISD::AssertSext &&
-      isa<ConstantSDNode>(N0.getOperand(1))) {
-    SDValue BigA = N0.getOperand(0);
-    EVT BigA_AssertVT = cast<VTSDNode>(BigA.getOperand(1))->getVT();
+      N0.hasOneUse() && isa<ConstantSDNode>(N0.getOperand(1))) {
     const APInt &Mask = N0.getConstantOperandAPInt(1);
-    if (AssertVT.bitsLT(BigA_AssertVT) &&
-        Mask.countr_one() >= BigA_AssertVT.getScalarSizeInBits()) {
+
+    // If we have (AssertZext (and (AssertSext X, iX), M), iY) and Y is smaller
+    // than X, and the And doesn't change the lower iX bits, we can move the
+    // AssertZext in front of the And and drop the AssertSext.
+    if (N0.getOperand(0).getOpcode() == ISD::AssertSext) {
+      SDValue BigA = N0.getOperand(0);
+      EVT BigA_AssertVT = cast<VTSDNode>(BigA.getOperand(1))->getVT();
+      if (AssertVT.bitsLT(BigA_AssertVT) &&
+          Mask.countr_one() >= BigA_AssertVT.getScalarSizeInBits()) {
+        SDLoc DL(N);
+        SDValue NewAssert =
+            DAG.getNode(Opcode, DL, N->getValueType(0), BigA.getOperand(0), N1);
+        return DAG.getNode(ISD::AND, DL, N->getValueType(0), NewAssert,
+                           N0.getOperand(1));
+      }
+    }
+
+    // Remove AssertZext entirely if the mask guarantees the assertion cannot
+    // fail.
+    if (Mask.isMask() && Mask.countr_one() <= AssertVT.getScalarSizeInBits()) {
       SDLoc DL(N);
-      SDValue NewAssert =
-          DAG.getNode(Opcode, DL, N->getValueType(0), BigA.getOperand(0), N1);
-      return DAG.getNode(ISD::AND, DL, N->getValueType(0), NewAssert,
+      return DAG.getNode(ISD::AND, DL, N0.getValueType(), N0.getOperand(0),
                          N0.getOperand(1));
     }
   }

llvm/lib/Target/AMDGPU/AMDGPUISelLowering.cpp

@@ -3991,24 +3991,6 @@ SDValue AMDGPUTargetLowering::performAssertSZExtCombine(SDNode *N,
     }
   }
 
-  // AssertZext in front of these intrinsics is not necessary, the lowering of
-  // the intrinsics into a register read will insert one if it is needed.
-  if (N->getOpcode() == ISD::AssertZext &&
-      N0.getOpcode() == ISD::INTRINSIC_WO_CHAIN) {
-    unsigned IID = N0.getConstantOperandVal(0);
-    switch (IID) {
-    case Intrinsic::amdgcn_workitem_id_x:
-    case Intrinsic::amdgcn_workitem_id_y:
-    case Intrinsic::amdgcn_workitem_id_z:
-    case Intrinsic::amdgcn_workgroup_id_x:
-    case Intrinsic::amdgcn_workgroup_id_y:
-    case Intrinsic::amdgcn_workgroup_id_z:
-      return N0;
-    default:
-      break;
-    }
-  }
-
   return SDValue();
 }
 

llvm/test/CodeGen/AMDGPU/llvm.amdgcn.workitem.id-unsupported-calling-convention.ll

@@ -9,9 +9,16 @@ declare i32 @llvm.amdgcn.workitem.id.z()
 ; FIXME: It's not worth adding AssertZext to the intrinsic calls, and
 ; we don't fold out assertzext undef ->undef
 define amdgpu_ps void @undefined_workitems(ptr addrspace(1) %p, ptr addrspace(1) %q, ptr addrspace(1) %r) {
-; CHECK-LABEL: undefined_workitems:
-; CHECK:       ; %bb.0:
-; CHECK-NEXT:    s_endpgm
+; SDAG-LABEL: undefined_workitems:
+; SDAG:       ; %bb.0:
+; SDAG-NEXT:    global_store_dword v[0:1], v0, off
+; SDAG-NEXT:    global_store_dword v[2:3], v0, off
+; SDAG-NEXT:    global_store_dword v[4:5], v0, off
+; SDAG-NEXT:    s_endpgm
+;
+; GISEL-LABEL: undefined_workitems:
+; GISEL:       ; %bb.0:
+; GISEL-NEXT:    s_endpgm
   %id.x = call i32 @llvm.amdgcn.workitem.id.x()
   store i32 %id.x, ptr addrspace(1) %p
   %id.y = call i32 @llvm.amdgcn.workitem.id.y()
@@ -21,5 +28,4 @@ define amdgpu_ps void @undefined_workitems(ptr addrspace(1) %p, ptr addrspace(1)
   ret void
 }
 ;; NOTE: These prefixes are unused and the list is autogenerated. Do not add tests below this line:
-; GISEL: {{.*}}
-; SDAG: {{.*}}
+; CHECK: {{.*}}