[llvm] [ValueTracking] Conservative NoSync check prevents vectorization (PR #199002)

[Kshitij Paranjape via llvm-commits]

================
@@ -1456,4 +1456,364 @@ exit:
   ret void
 }
 
+define i64 @VectorizingWithoutNoSyncAttributeTest(ptr noundef nonnull readonly align 8 captures(none) dereferenceable(24) %v, i32 noundef %n) {
+; CHECK-LABEL: define i64 @VectorizingWithoutNoSyncAttributeTest(
+; CHECK-SAME: ptr noundef nonnull readonly align 8 captures(none) dereferenceable(24) [[V:%.*]], i32 noundef [[N:%.*]]) {
+; CHECK-NEXT:  [[ENTRY:.*:]]
+; CHECK-NEXT:    [[L_V:%.*]] = load ptr, ptr [[V]], align 8
+; CHECK-NEXT:    call void @llvm.assume(i1 true) [ "align"(ptr [[L_V]], i64 4) ]
+; CHECK-NEXT:    [[PI_L_V:%.*]] = ptrtoint ptr [[L_V]] to i64
+; CHECK-NEXT:    [[GEP_V:%.*]] = getelementptr inbounds nuw i8, ptr [[V]], i64 8
+; CHECK-NEXT:    [[L_GEP_V:%.*]] = load ptr, ptr [[GEP_V]], align 8
+; CHECK-NEXT:    call void @llvm.assume(i1 true) [ "align"(ptr [[L_GEP_V]], i64 4) ]
+; CHECK-NEXT:    [[PI_L_GEP_V:%.*]] = ptrtoint ptr [[L_GEP_V]] to i64
+; CHECK-NEXT:    [[SUB:%.*]] = sub i64 [[PI_L_GEP_V]], [[PI_L_V]]
+; CHECK-NEXT:    call void @llvm.assume(i1 true) [ "dereferenceable"(ptr [[L_V]], i64 [[SUB]]) ]
+; CHECK-NEXT:    call void @llvm.assume(i1 true) [ "align"(ptr [[L_V]], i64 4) ]
+; CHECK-NEXT:    call void @llvm.assume(i1 true) [ "align"(ptr [[L_GEP_V]], i64 4) ]
+; CHECK-NEXT:    [[C_1:%.*]] = icmp eq ptr [[L_V]], [[L_GEP_V]]
+; CHECK-NEXT:    br i1 [[C_1]], [[EXIT:label %.*]], label %[[LOOP_PREHEADER:.*]]
+; CHECK:       [[LOOP_PREHEADER]]:
+; CHECK-NEXT:    [[TMP0:%.*]] = add i64 [[PI_L_GEP_V]], -4
+; CHECK-NEXT:    [[TMP1:%.*]] = sub i64 [[TMP0]], [[PI_L_V]]
+; CHECK-NEXT:    [[TMP2:%.*]] = lshr i64 [[TMP1]], 2
+; CHECK-NEXT:    [[TMP3:%.*]] = add nuw nsw i64 [[TMP2]], 1
+; CHECK-NEXT:    [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[TMP3]], 2
+; CHECK-NEXT:    br i1 [[MIN_ITERS_CHECK]], label %[[SCALAR_PH:.*]], label %[[VECTOR_PH:.*]]
+; CHECK:       [[VECTOR_PH]]:
+; CHECK-NEXT:    [[N_MOD_VF:%.*]] = urem i64 [[TMP3]], 2
+; CHECK-NEXT:    [[N_VEC:%.*]] = sub i64 [[TMP3]], [[N_MOD_VF]]
+; CHECK-NEXT:    [[TMP4:%.*]] = shl i64 [[N_VEC]], 2
+; CHECK-NEXT:    [[TMP5:%.*]] = getelementptr i8, ptr [[L_V]], i64 [[TMP4]]
+; CHECK-NEXT:    [[BROADCAST_SPLATINSERT:%.*]] = insertelement <2 x i32> poison, i32 [[N]], i64 0
+; CHECK-NEXT:    [[BROADCAST_SPLAT:%.*]] = shufflevector <2 x i32> [[BROADCAST_SPLATINSERT]], <2 x i32> poison, <2 x i32> zeroinitializer
+; CHECK-NEXT:    br label %[[LOOP_HEADER:.*]]
+; CHECK:       [[LOOP_HEADER]]:
+; CHECK-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[LOOP_LATCH:.*]] ]
+; CHECK-NEXT:    [[TMP6:%.*]] = shl i64 [[INDEX]], 2
+; CHECK-NEXT:    [[NEXT_GEP:%.*]] = getelementptr i8, ptr [[L_V]], i64 [[TMP6]]
+; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <2 x i32>, ptr [[NEXT_GEP]], align 4
+; CHECK-NEXT:    [[TMP7:%.*]] = icmp slt <2 x i32> [[WIDE_LOAD]], [[BROADCAST_SPLAT]]
+; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 2
+; CHECK-NEXT:    [[TMP8:%.*]] = freeze <2 x i1> [[TMP7]]
+; CHECK-NEXT:    [[C_2:%.*]] = call i1 @llvm.vector.reduce.or.v2i1(<2 x i1> [[TMP8]])
+; CHECK-NEXT:    [[TMP10:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
+; CHECK-NEXT:    br i1 [[C_2]], label %[[LOOP_EARLY_EXIT:.*]], label %[[LOOP_LATCH]]
+; CHECK:       [[LOOP_LATCH]]:
+; CHECK-NEXT:    br i1 [[TMP10]], label %[[LOOP_COMP:.*]], label %[[LOOP_HEADER]], !llvm.loop [[LOOP24:![0-9]+]]
+; CHECK:       [[LOOP_COMP]]:
+; CHECK-NEXT:    [[CMP_N:%.*]] = icmp eq i64 [[TMP3]], [[N_VEC]]
+; CHECK-NEXT:    br i1 [[CMP_N]], [[LOOP_COMP1:label %.*]], label %[[SCALAR_PH]]
+; CHECK:       [[LOOP_EARLY_EXIT]]:
+; CHECK-NEXT:    [[TMP11:%.*]] = call i64 @llvm.experimental.cttz.elts.i64.v2i1(<2 x i1> [[TMP7]], i1 false)
+; CHECK-NEXT:    [[TMP12:%.*]] = add i64 [[INDEX]], [[TMP11]]
+; CHECK-NEXT:    [[TMP13:%.*]] = shl i64 [[TMP12]], 2
+; CHECK-NEXT:    [[TMP14:%.*]] = getelementptr i8, ptr [[L_V]], i64 [[TMP13]]
+; CHECK-NEXT:    br [[LOOP_EARLY_EXIT1:label %.*]]
+; CHECK:       [[SCALAR_PH]]:
+;
+entry:
+  %l.v = load ptr, ptr %v, align 8
+  call void @llvm.assume(i1 true) [ "align"(ptr %l.v, i64 4) ]
+  %pi.l.v = ptrtoint ptr %l.v to i64
+  %gep.v = getelementptr inbounds nuw i8, ptr %v, i64 8
+  %l.gep.v = load ptr, ptr %gep.v, align 8
+  call void @llvm.assume(i1 true) [ "align"(ptr %l.gep.v, i64 4) ]
+  %pi.l.gep.v = ptrtoint ptr %l.gep.v to i64
+  %sub = sub i64 %pi.l.gep.v, %pi.l.v
+  call void @llvm.assume(i1 true) [ "dereferenceable"(ptr %l.v, i64 %sub) ]
----------------
kshitijvp wrote:

@fhahn Isn't `deref_assumption_in_header_constant_trip_count_nofree_via_context_but_missing_nosync` getting vectorized too?

The Check-Lines contain the Vector-Preheader and Vector-Body which means it is getting vectorized.

https://github.com/llvm/llvm-project/pull/199002