[llvm] [LV] Convert uniform-address unmasked scatters to scalar store. (PR #166114)

[Florian Hahn via llvm-commits]

================
@@ -0,0 +1,51 @@
+; NOTE: Assertions have been autogenerated by utils/update_test_checks.py UTC_ARGS: --version 5
+; RUN: opt -passes=loop-vectorize -force-vector-width=2 -mtriple=riscv64 -mattr=+v -S %s | FileCheck %s
+define void @truncate_i16_to_i8_cse(ptr noalias %src, ptr noalias %dst) {
+; CHECK-LABEL: define void @truncate_i16_to_i8_cse(
+; CHECK-SAME: ptr noalias [[SRC:%.*]], ptr noalias [[DST:%.*]]) #[[ATTR0:[0-9]+]] {
+; CHECK-NEXT:  [[ENTRY:.*:]]
+; CHECK-NEXT:    br label %[[VECTOR_PH:.*]]
+; CHECK:       [[VECTOR_PH]]:
+; CHECK-NEXT:    br label %[[VECTOR_BODY:.*]]
+; CHECK:       [[VECTOR_BODY]]:
+; CHECK-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
+; CHECK-NEXT:    [[TMP0:%.*]] = load i16, ptr [[SRC]], align 2
+; CHECK-NEXT:    [[BROADCAST_SPLATINSERT:%.*]] = insertelement <2 x i16> poison, i16 [[TMP0]], i64 0
+; CHECK-NEXT:    [[BROADCAST_SPLAT:%.*]] = shufflevector <2 x i16> [[BROADCAST_SPLATINSERT]], <2 x i16> poison, <2 x i32> zeroinitializer
+; CHECK-NEXT:    [[TMP1:%.*]] = trunc <2 x i16> [[BROADCAST_SPLAT]] to <2 x i8>
+; CHECK-NEXT:    [[TMP2:%.*]] = extractelement <2 x i8> [[TMP1]], i32 1
+; CHECK-NEXT:    store i8 [[TMP2]], ptr null, align 1
+; CHECK-NEXT:    store i8 [[TMP2]], ptr [[DST]], align 1
+; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 2
+; CHECK-NEXT:    [[TMP3:%.*]] = icmp eq i64 [[INDEX_NEXT]], 4294967296
+; CHECK-NEXT:    br i1 [[TMP3]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]]
+; CHECK:       [[MIDDLE_BLOCK]]:
+; CHECK-NEXT:    br label %[[EXIT:.*]]
+; CHECK:       [[EXIT]]:
+; CHECK-NEXT:    ret void
+;
+entry:
+  br label %loop
+
+loop:
+  %iv = phi i64 [ 0, %entry ], [ %iv.next, %loop ]
+  %count = phi i32 [ 0, %entry ], [ %count.next, %loop ]
+  %val = load i16, ptr %src, align 2
+  %val.zext = zext i16 %val to i64
+  %val.trunc.zext = trunc i64 %val.zext to i8
+  store i8 %val.trunc.zext, ptr null, align 1
----------------
fhahn wrote:

please avoid immediate UB, perhaps pass another pointer.

https://github.com/llvm/llvm-project/pull/166114