[NEON] Fix combining of vldx_dup intrinsics with updating of base addresses

Resolves:
Unsupported ARM Neon intrinsics in Target-specific DAG combine
function for VLDDUP
https://bugs.llvm.org/show_bug.cgi?id=38031

Related diff: D48439

Differential Revision: https://reviews.llvm.org/D48920

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@336325 91177308-0d34-0410-b5e6-96231b3b80d8

diff --git a/lib/Target/ARM/ARMISelLowering.cpp b/lib/Target/ARM/ARMISelLowering.cpp
index 6e5942f..9b143dc 100644 (file)
--- a/lib/Target/ARM/ARMISelLowering.cpp
+++ b/lib/Target/ARM/ARMISelLowering.cpp
@@ -11488,6 +11488,12 @@ static SDValue CombineBaseUpdate(SDNode *N,
         NumVecs = 3; break;
       case Intrinsic::arm_neon_vld4:     NewOpc = ARMISD::VLD4_UPD;
         NumVecs = 4; break;
+      case Intrinsic::arm_neon_vld2dup:
+      case Intrinsic::arm_neon_vld3dup:
+      case Intrinsic::arm_neon_vld4dup:
+        // TODO: Support updating VLDxDUP nodes. For now, we just skip
+        // combining base updates for such intrinsics.
+        continue;
       case Intrinsic::arm_neon_vld2lane: NewOpc = ARMISD::VLD2LN_UPD;
         NumVecs = 2; isLaneOp = true; break;
       case Intrinsic::arm_neon_vld3lane: NewOpc = ARMISD::VLD3LN_UPD;

diff --git a/test/CodeGen/ARM/arm-vlddup-update.ll b/test/CodeGen/ARM/arm-vlddup-update.ll
new file mode 100644 (file)
index 0000000..c39dfc7
--- /dev/null
+++ b/test/CodeGen/ARM/arm-vlddup-update.ll
@@ -0,0 +1,43 @@
+; RUN: llc < %s -mtriple=armv8-linux-gnueabi -verify-machineinstrs \
+; RUN:     -asm-verbose=false | FileCheck %s
+
+%struct.uint32x2x2_t = type { <2 x i32>, <2 x i32> }
+%struct.uint32x2x3_t = type { <2 x i32>, <2 x i32>, <2 x i32> }
+%struct.uint32x2x4_t = type { <2 x i32>, <2 x i32>, <2 x i32>, <2 x i32> }
+
+declare %struct.uint32x2x2_t @llvm.arm.neon.vld2dup.v2i32.p0i8(i8*, i32)
+declare %struct.uint32x2x3_t @llvm.arm.neon.vld3dup.v2i32.p0i8(i8*, i32)
+declare %struct.uint32x2x4_t @llvm.arm.neon.vld4dup.v2i32.p0i8(i8*, i32)
+
+; CHECK-LABEL: test_vld2_dup_update
+; CHECK: vld2.32 {d16[], d17[]}, {{\[}}[[SRC_R:r[0-9]+]]]
+; CHECK: add {{r[0-9]+|lr}}, [[SRC_R]], #4
+define i8* @test_vld2_dup_update(%struct.uint32x2x2_t* %dest, i8* %src) {
+entry:
+  %tmp = tail call %struct.uint32x2x2_t @llvm.arm.neon.vld2dup.v2i32.p0i8(i8* %src, i32 4)
+  store %struct.uint32x2x2_t %tmp, %struct.uint32x2x2_t* %dest, align 8
+  %updated_src = getelementptr inbounds i8, i8* %src, i32 4
+  ret i8* %updated_src
+}
+
+; CHECK-LABEL: test_vld3_dup_update
+; CHECK: vld3.32 {d16[], d17[], d18[]}, {{\[}}[[SRC_R:r[0-9]+]]]
+; CHECK: add {{r[0-9]+|lr}}, [[SRC_R]], #4
+define i8* @test_vld3_dup_update(%struct.uint32x2x3_t* %dest, i8* %src) {
+entry:
+  %tmp = tail call %struct.uint32x2x3_t @llvm.arm.neon.vld3dup.v2i32.p0i8(i8* %src, i32 4)
+  store %struct.uint32x2x3_t %tmp, %struct.uint32x2x3_t* %dest, align 8
+  %updated_src = getelementptr inbounds i8, i8* %src, i32 4
+  ret i8* %updated_src
+}
+
+; CHECK-LABEL: test_vld4_dup_update
+; CHECK: vld4.32 {d16[], d17[], d18[], d19[]}, {{\[}}[[SRC_R:r[0-9]+]]]
+; CHECK: add {{r[0-9]+|lr}}, [[SRC_R]], #4
+define i8* @test_vld4_dup_update(%struct.uint32x2x4_t* %dest, i8* %src) {
+entry:
+  %tmp = tail call %struct.uint32x2x4_t @llvm.arm.neon.vld4dup.v2i32.p0i8(i8* %src, i32 4)
+  store %struct.uint32x2x4_t %tmp, %struct.uint32x2x4_t* %dest, align 8
+  %updated_src = getelementptr inbounds i8, i8* %src, i32 4
+  ret i8* %updated_src
+}