[llvm] [LV] Convert scatter w/uniform addr and mask being header mask to scalar store. (PR #172799)

[Elvis Wang via llvm-commits]

================
@@ -14,8 +14,8 @@ define void @truncate_to_minimal_bitwidths_widen_cast_recipe(ptr %src) {
 ; CHECK-NEXT:    br label %[[VECTOR_BODY:.*]]
 ; CHECK:       [[VECTOR_BODY]]:
 ; CHECK-NEXT:    [[AVL:%.*]] = phi i64 [ 9, %[[VECTOR_PH]] ], [ [[AVL_NEXT:%.*]], %[[VECTOR_BODY]] ]
-; CHECK-NEXT:    [[TMP7:%.*]] = call i32 @llvm.experimental.get.vector.length.i64(i64 [[AVL]], i32 8, i1 true)
-; CHECK-NEXT:    call void @llvm.vp.scatter.nxv8i8.nxv8p0(<vscale x 8 x i8> zeroinitializer, <vscale x 8 x ptr> align 1 splat (ptr null), <vscale x 8 x i1> splat (i1 true), i32 [[TMP7]])
+; CHECK-NEXT:    [[TMP7:%.*]] = call i32 @llvm.experimental.get.vector.length.i64(i64 [[AVL]], i32 2, i1 true)
+; CHECK-NEXT:    call void @llvm.vp.scatter.nxv2i8.nxv2p0(<vscale x 2 x i8> zeroinitializer, <vscale x 2 x ptr> align 1 splat (ptr null), <vscale x 2 x i1> splat (i1 true), i32 [[TMP7]])
----------------
ElvisWang123 wrote:

The larger VF (`vscale x 8|16`) now won't generate any vector instruction.  And then consider to the TC (`9`), choosing `vscale x 2`.

https://github.com/llvm/llvm-project/pull/172799