[llvm] [LV] Avoid crashing for vector calls with scalar byte types (PR #197417)

[Graham Hunter via llvm-commits]

================
@@ -0,0 +1,210 @@
+; NOTE: Assertions have been autogenerated by utils/update_test_checks.py UTC_ARGS: --check-globals none --version 6
+; RUN: opt < %s -passes=loop-vectorize -force-vector-width=2 -force-vector-interleave=1 -S | FileCheck %s
+
+;; Safe to vectorize normally.
+define void @byte_type_vector(ptr noalias %dst, ptr readonly %src, i64 %n) {
+; CHECK-LABEL: define void @byte_type_vector(
+; CHECK-SAME: ptr noalias [[DST:%.*]], ptr readonly [[SRC:%.*]], i64 [[N:%.*]]) {
+; CHECK-NEXT:  [[ENTRY:.*]]:
+; CHECK-NEXT:    [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[N]], 2
+; CHECK-NEXT:    br i1 [[MIN_ITERS_CHECK]], label %[[SCALAR_PH:.*]], label %[[VECTOR_PH:.*]]
+; CHECK:       [[VECTOR_PH]]:
+; CHECK-NEXT:    [[N_MOD_VF:%.*]] = urem i64 [[N]], 2
+; CHECK-NEXT:    [[N_VEC:%.*]] = sub i64 [[N]], [[N_MOD_VF]]
+; CHECK-NEXT:    br label %[[VECTOR_BODY:.*]]
+; CHECK:       [[VECTOR_BODY]]:
+; CHECK-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
+; CHECK-NEXT:    [[TMP0:%.*]] = getelementptr b64, ptr [[SRC]], i64 [[INDEX]]
+; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <2 x b64>, ptr [[TMP0]], align 8
+; CHECK-NEXT:    [[TMP1:%.*]] = call <2 x b64> @foo_vector(<2 x b64> [[WIDE_LOAD]])
+; CHECK-NEXT:    [[TMP2:%.*]] = getelementptr inbounds b64, ptr [[DST]], i64 [[INDEX]]
+; CHECK-NEXT:    store <2 x b64> [[TMP1]], ptr [[TMP2]], align 4
+; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 2
+; CHECK-NEXT:    [[TMP3:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
+; CHECK-NEXT:    br i1 [[TMP3]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]]
+; CHECK:       [[MIDDLE_BLOCK]]:
+; CHECK-NEXT:    [[CMP_N:%.*]] = icmp eq i64 [[N]], [[N_VEC]]
+; CHECK-NEXT:    br i1 [[CMP_N]], label %[[EXIT:.*]], label %[[SCALAR_PH]]
+; CHECK:       [[SCALAR_PH]]:
+; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], %[[MIDDLE_BLOCK]] ], [ 0, %[[ENTRY]] ]
+; CHECK-NEXT:    br label %[[FOR_BODY:.*]]
+; CHECK:       [[FOR_BODY]]:
+; CHECK-NEXT:    [[IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], %[[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], %[[FOR_BODY]] ]
+; CHECK-NEXT:    [[GEP_SRC:%.*]] = getelementptr b64, ptr [[SRC]], i64 [[IV]]
+; CHECK-NEXT:    [[DATA:%.*]] = load b64, ptr [[GEP_SRC]], align 8
+; CHECK-NEXT:    [[CALL:%.*]] = call b64 @foo(b64 [[DATA]]) #[[ATTR0:[0-9]+]]
+; CHECK-NEXT:    [[GEP_DST:%.*]] = getelementptr inbounds b64, ptr [[DST]], i64 [[IV]]
+; CHECK-NEXT:    store b64 [[CALL]], ptr [[GEP_DST]], align 4
+; CHECK-NEXT:    [[IV_NEXT]] = add nuw nsw i64 [[IV]], 1
+; CHECK-NEXT:    [[EXIT_COND:%.*]] = icmp eq i64 [[IV_NEXT]], [[N]]
+; CHECK-NEXT:    br i1 [[EXIT_COND]], label %[[EXIT]], label %[[FOR_BODY]], !llvm.loop [[LOOP3:![0-9]+]]
+; CHECK:       [[EXIT]]:
+; CHECK-NEXT:    ret void
+;
+entry:
+  br label %for.body
+
+for.body:
+  %iv = phi i64 [ 0, %entry ], [ %iv.next, %for.body ]
+  %gep.src = getelementptr b64, ptr %src, i64 %iv
+  %data = load b64, ptr %gep.src, align 8
+  %call = call b64 @foo(b64 %data) #0
+  %gep.dst = getelementptr inbounds b64, ptr %dst, i64 %iv
+  store b64 %call, ptr %gep.dst
+  %iv.next = add nuw nsw i64 %iv, 1
+  %exit.cond = icmp eq i64 %iv.next, %n
+  br i1 %exit.cond, label %exit, label %for.body
+
+exit:
+  ret void
+}
+
+;; Must be scalarized, since we can't form a SCEV from a byte type.
----------------
huntergr-arm wrote:

done

https://github.com/llvm/llvm-project/pull/197417