[LV] Add test where interleave group start pointer is incorrect.

Test case from #106431.
llvm · Oct 16, 2024 · cc5b5ca · cc5b5ca
1 parent 282ab2f
commit cc5b5ca
Showing 1 changed file with 88 additions and 0 deletions.
diff --git a/llvm/test/Transforms/LoopVectorize/interleaved-accesses-different-insert-position.ll b/llvm/test/Transforms/LoopVectorize/interleaved-accesses-different-insert-position.ll
@@ -154,11 +154,99 @@ loop.latch:
 exit:
   ret void
 }
+
+; FIXME: Currently the start address of the interleav group is computed
+; incorrectly.
+define i64 @interleave_group_load_pointer_type(ptr %start, ptr %end) {
+; CHECK-LABEL: define i64 @interleave_group_load_pointer_type(
+; CHECK-SAME: ptr [[START:%.*]], ptr [[END:%.*]]) {
+; CHECK-NEXT:  [[ENTRY:.*]]:
+; CHECK-NEXT:    [[START2:%.*]] = ptrtoint ptr [[START]] to i64
+; CHECK-NEXT:    [[END1:%.*]] = ptrtoint ptr [[END]] to i64
+; CHECK-NEXT:    [[TMP0:%.*]] = sub i64 [[END1]], [[START2]]
+; CHECK-NEXT:    [[TMP1:%.*]] = udiv i64 [[TMP0]], 24
+; CHECK-NEXT:    [[TMP2:%.*]] = add nuw nsw i64 [[TMP1]], 1
+; CHECK-NEXT:    [[MIN_ITERS_CHECK:%.*]] = icmp ule i64 [[TMP2]], 4
+; CHECK-NEXT:    br i1 [[MIN_ITERS_CHECK]], label %[[SCALAR_PH:.*]], label %[[VECTOR_PH:.*]]
+; CHECK:       [[VECTOR_PH]]:
+; CHECK-NEXT:    [[N_MOD_VF:%.*]] = urem i64 [[TMP2]], 4
+; CHECK-NEXT:    [[TMP3:%.*]] = icmp eq i64 [[N_MOD_VF]], 0
+; CHECK-NEXT:    [[TMP4:%.*]] = select i1 [[TMP3]], i64 4, i64 [[N_MOD_VF]]
+; CHECK-NEXT:    [[N_VEC:%.*]] = sub i64 [[TMP2]], [[TMP4]]
+; CHECK-NEXT:    [[TMP5:%.*]] = mul i64 [[N_VEC]], 24
+; CHECK-NEXT:    [[IND_END:%.*]] = getelementptr i8, ptr [[START]], i64 [[TMP5]]
+; CHECK-NEXT:    br label %[[VECTOR_BODY:.*]]
+; CHECK:       [[VECTOR_BODY]]:
+; CHECK-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
+; CHECK-NEXT:    [[VEC_PHI:%.*]] = phi <4 x i64> [ zeroinitializer, %[[VECTOR_PH]] ], [ [[TMP12:%.*]], %[[VECTOR_BODY]] ]
+; CHECK-NEXT:    [[OFFSET_IDX:%.*]] = mul i64 [[INDEX]], 24
+; CHECK-NEXT:    [[TMP6:%.*]] = add i64 [[OFFSET_IDX]], 0
+; CHECK-NEXT:    [[NEXT_GEP:%.*]] = getelementptr i8, ptr [[START]], i64 [[TMP6]]
+; CHECK-NEXT:    [[TMP7:%.*]] = getelementptr i8, ptr [[NEXT_GEP]], i64 16
+; CHECK-NEXT:    [[TMP8:%.*]] = getelementptr i8, ptr [[TMP7]], i32 0
+; CHECK-NEXT:    [[WIDE_VEC:%.*]] = load <12 x ptr>, ptr [[TMP8]], align 8
+; CHECK-NEXT:    [[STRIDED_VEC:%.*]] = shufflevector <12 x ptr> [[WIDE_VEC]], <12 x ptr> poison, <4 x i32> <i32 0, i32 3, i32 6, i32 9>
+; CHECK-NEXT:    [[STRIDED_VEC3:%.*]] = shufflevector <12 x ptr> [[WIDE_VEC]], <12 x ptr> poison, <4 x i32> <i32 1, i32 4, i32 7, i32 10>
+; CHECK-NEXT:    [[TMP9:%.*]] = ptrtoint <4 x ptr> [[STRIDED_VEC3]] to <4 x i64>
+; CHECK-NEXT:    [[TMP10:%.*]] = ptrtoint <4 x ptr> [[STRIDED_VEC]] to <4 x i64>
+; CHECK-NEXT:    [[TMP11:%.*]] = or <4 x i64> [[TMP9]], [[TMP10]]
+; CHECK-NEXT:    [[TMP12]] = or <4 x i64> [[TMP11]], [[VEC_PHI]]
+; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
+; CHECK-NEXT:    [[TMP13:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
+; CHECK-NEXT:    br i1 [[TMP13]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP6:![0-9]+]]
+; CHECK:       [[MIDDLE_BLOCK]]:
+; CHECK-NEXT:    [[TMP14:%.*]] = call i64 @llvm.vector.reduce.or.v4i64(<4 x i64> [[TMP12]])
+; CHECK-NEXT:    br label %[[SCALAR_PH]]
+; CHECK:       [[SCALAR_PH]]:
+; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi ptr [ [[IND_END]], %[[MIDDLE_BLOCK]] ], [ [[START]], %[[ENTRY]] ]
+; CHECK-NEXT:    [[BC_MERGE_RDX:%.*]] = phi i64 [ [[TMP14]], %[[MIDDLE_BLOCK]] ], [ 0, %[[ENTRY]] ]
+; CHECK-NEXT:    br label %[[LOOP:.*]]
+; CHECK:       [[LOOP]]:
+; CHECK-NEXT:    [[PTR_IV:%.*]] = phi ptr [ [[BC_RESUME_VAL]], %[[SCALAR_PH]] ], [ [[PTR_IV_NEXT:%.*]], %[[LOOP]] ]
+; CHECK-NEXT:    [[RED:%.*]] = phi i64 [ [[BC_MERGE_RDX]], %[[SCALAR_PH]] ], [ [[RED_NEXT:%.*]], %[[LOOP]] ]
+; CHECK-NEXT:    [[GEP_16:%.*]] = getelementptr i8, ptr [[PTR_IV]], i64 16
+; CHECK-NEXT:    [[L_16:%.*]] = load ptr, ptr [[GEP_16]], align 8
+; CHECK-NEXT:    [[P_16:%.*]] = ptrtoint ptr [[L_16]] to i64
+; CHECK-NEXT:    [[GEP_8:%.*]] = getelementptr i8, ptr [[PTR_IV]], i64 8
+; CHECK-NEXT:    [[L_8:%.*]] = load ptr, ptr [[GEP_8]], align 8
+; CHECK-NEXT:    [[P_8:%.*]] = ptrtoint ptr [[L_8]] to i64
+; CHECK-NEXT:    [[OR_1:%.*]] = or i64 [[P_16]], [[P_8]]
+; CHECK-NEXT:    [[RED_NEXT]] = or i64 [[OR_1]], [[RED]]
+; CHECK-NEXT:    [[PTR_IV_NEXT]] = getelementptr nusw i8, ptr [[PTR_IV]], i64 24
+; CHECK-NEXT:    [[EC:%.*]] = icmp eq ptr [[PTR_IV]], [[END]]
+; CHECK-NEXT:    br i1 [[EC]], label %[[EXIT:.*]], label %[[LOOP]], !llvm.loop [[LOOP7:![0-9]+]]
+; CHECK:       [[EXIT]]:
+; CHECK-NEXT:    [[RED_NEXT_LCSSA:%.*]] = phi i64 [ [[RED_NEXT]], %[[LOOP]] ]
+; CHECK-NEXT:    ret i64 [[RED_NEXT_LCSSA]]
+;
+entry:
+  br label %loop
+
+loop:
+  %ptr.iv = phi ptr [ %start, %entry ], [ %ptr.iv.next, %loop ]
+  %red = phi i64 [ 0, %entry ], [ %red.next, %loop ]
+  %gep.16 = getelementptr i8, ptr %ptr.iv, i64 16
+  %l.16 = load ptr, ptr %gep.16, align 8
+  %p.16 = ptrtoint ptr %l.16 to i64
+  %gep.8 = getelementptr i8, ptr %ptr.iv, i64 8
+  %l.8 = load ptr, ptr %gep.8, align 8
+  %p.8 = ptrtoint ptr %l.8 to i64
+  %or.1 = or i64 %p.16, %p.8
+  %red.next = or i64 %or.1, %red
+  %ptr.iv.next = getelementptr nusw i8, ptr %ptr.iv, i64 24
+  %ec = icmp eq ptr %ptr.iv, %end
+  br i1 %ec, label %exit, label %loop
+
+exit:
+  ret i64 %red.next
+}
 ;.
 ; CHECK: [[LOOP0]] = distinct !{[[LOOP0]], [[META1:![0-9]+]], [[META2:![0-9]+]]}
 ; CHECK: [[META1]] = !{!"llvm.loop.isvectorized", i32 1}
 ; CHECK: [[META2]] = !{!"llvm.loop.unroll.runtime.disable"}
 ; CHECK: [[LOOP3]] = distinct !{[[LOOP3]], [[META2]], [[META1]]}
 ; CHECK: [[LOOP4]] = distinct !{[[LOOP4]], [[META1]], [[META2]]}
 ; CHECK: [[LOOP5]] = distinct !{[[LOOP5]], [[META2]], [[META1]]}
+; CHECK: [[LOOP6]] = distinct !{[[LOOP6]], [[META1]], [[META2]]}
+; CHECK: [[LOOP7]] = distinct !{[[LOOP7]], [[META2]], [[META1]]}
 ;.