Add testcase.

Author: NexMing

llvm/test/CodeGen/RISCV/rvv/sink-splat-operands.ll

@@ -5890,3 +5890,70 @@ vector.body:                                      ; preds = %vector.body, %entry
 for.cond.cleanup:                                 ; preds = %vector.body
   ret void
 }
+
+define void @sink_vp_splat(ptr nocapture %out, ptr nocapture %in) {
+; CHECK-LABEL: sink_vp_splat:
+; CHECK:       # %bb.0: # %entry
+; CHECK-NEXT:    li a2, 0
+; CHECK-NEXT:    li a3, 1024
+; CHECK-NEXT:    lui a4, 1
+; CHECK-NEXT:  .LBB129_1: # %vector.body
+; CHECK-NEXT:    # =>This Loop Header: Depth=1
+; CHECK-NEXT:    # Child Loop BB129_2 Depth 2
+; CHECK-NEXT:    vsetvli a5, a3, e32, m4, ta, ma
+; CHECK-NEXT:    slli a6, a2, 2
+; CHECK-NEXT:    vmv.v.i v8, 0
+; CHECK-NEXT:    vmv.v.i v12, 3
+; CHECK-NEXT:    add a7, a1, a6
+; CHECK-NEXT:    li t0, 1024
+; CHECK-NEXT:  .LBB129_2: # %for.body424
+; CHECK-NEXT:    # Parent Loop BB129_1 Depth=1
+; CHECK-NEXT:    # => This Inner Loop Header: Depth=2
+; CHECK-NEXT:    vle32.v v16, (a7)
+; CHECK-NEXT:    addi t0, t0, -1
+; CHECK-NEXT:    vmacc.vv v8, v16, v12
+; CHECK-NEXT:    add a7, a7, a4
+; CHECK-NEXT:    bnez t0, .LBB129_2
+; CHECK-NEXT:  # %bb.3: # %vector.latch
+; CHECK-NEXT:    # in Loop: Header=BB129_1 Depth=1
+; CHECK-NEXT:    add a6, a0, a6
+; CHECK-NEXT:    sub a3, a3, a5
+; CHECK-NEXT:    vse32.v v8, (a6)
+; CHECK-NEXT:    add a2, a2, a5
+; CHECK-NEXT:    bnez a3, .LBB129_1
+; CHECK-NEXT:  # %bb.4: # %for.cond.cleanup
+; CHECK-NEXT:    ret
+entry:
+  br label %vector.body
+
+vector.body:                                      ; preds = %vector.latch, %entry
+  %scalar.ind = phi i64 [ 0, %entry ], [ %next.ind, %vector.latch ]
+  %trip.count = phi i64 [ 1024, %entry ], [ %remaining.trip.count, %vector.latch ]
+  %evl = tail call i32 @llvm.experimental.get.vector.length.i64(i64 %trip.count, i32 8, i1 true)
+  %vp.splat1 = tail call <vscale x 8 x i32> @llvm.experimental.vp.splat.nxv8i32(i32 0, <vscale x 8 x i1> splat(i1 true), i32 %evl)
+  %vp.splat2 = tail call <vscale x 8 x i32> @llvm.experimental.vp.splat.nxv8i32(i32 3, <vscale x 8 x i1> splat(i1 true), i32 %evl)
+  %evl.cast = zext i32 %evl to i64
+  br label %for.body424
+
+for.body424:                                      ; preds = %for.body424, %vector.body
+  %scalar.phi = phi i64 [ 0, %vector.body ], [ %indvars.iv.next27, %for.body424 ]
+  %vector.phi = phi <vscale x 8 x i32> [ %vp.splat1, %vector.body ], [ %vp.binary26, %for.body424 ]
+  %arrayidx625 = getelementptr inbounds [1024 x i32], ptr %in, i64 %scalar.phi, i64 %scalar.ind
+  %widen.load = tail call <vscale x 8 x i32> @llvm.vp.load.nxv8i32.p0(ptr %arrayidx625, <vscale x 8 x i1> splat (i1 true), i32 %evl)
+  %vp.binary = tail call <vscale x 8 x i32> @llvm.vp.mul.nxv8i32(<vscale x 8 x i32> %widen.load, <vscale x 8 x i32> %vp.splat2, <vscale x 8 x i1> splat (i1 true), i32 %evl)
+  %vp.binary26 = tail call <vscale x 8 x i32> @llvm.vp.add.nxv8i32(<vscale x 8 x i32> %vector.phi, <vscale x 8 x i32> %vp.binary, <vscale x 8 x i1> splat (i1 true), i32 %evl)
+  %indvars.iv.next27 = add nuw nsw i64 %scalar.phi, 1
+  %exitcond.not28 = icmp eq i64 %indvars.iv.next27, 1024
+  br i1 %exitcond.not28, label %vector.latch, label %for.body424
+
+vector.latch:                                     ; preds = %for.body424
+  %arrayidx830 = getelementptr inbounds i32, ptr %out, i64 %scalar.ind
+  tail call void @llvm.vp.store.nxv8i32.p0(<vscale x 8 x i32> %vp.binary26, ptr %arrayidx830, <vscale x 8 x i1> splat (i1 true), i32 %evl)
+  %remaining.trip.count = sub nuw i64 %trip.count, %evl.cast
+  %next.ind = add i64 %scalar.ind, %evl.cast
+  %6 = icmp eq i64 %remaining.trip.count, 0
+  br i1 %6, label %for.cond.cleanup, label %vector.body
+
+for.cond.cleanup:                                 ; preds = %vector.latch
+  ret void
+}