[llvm] [SystemZ] SLP reductions: cost functions of reductions and scalarization (PR #112491)

[Jonas Paulsson via llvm-commits]

================
@@ -104,13 +108,20 @@ declare double @llvm.fmuladd.f64(double, double, double)
 ; which is recognized in SystemZTTImpl::getScalarizationOverhead().
 define void @fun2(ptr %0, ptr %Dst) {
 ; CHECK-LABEL: define void @fun2(
-; CHECK: insertelement
-; CHECK: store <2 x i64>
+; CHECK-SAME: ptr [[TMP0:%.*]], ptr [[DST:%.*]]) #[[ATTR0]] {
+; CHECK-NEXT:    [[TMP2:%.*]] = load i64, ptr [[TMP0]], align 8
+; CHECK-NEXT:    [[TMP3:%.*]] = icmp eq i64 [[TMP2]], 0
+; CHECK-NEXT:    br i1 [[TMP3]], label %[[BB4:.*]], label %[[BB5:.*]]
+; CHECK:       [[BB4]]:
+; CHECK-NEXT:    ret void
+; CHECK:       [[BB5]]:
+; CHECK-NEXT:    [[TMP6:%.*]] = getelementptr i8, ptr [[DST]], i64 24
+; CHECK-NEXT:    store i64 [[TMP2]], ptr [[TMP6]], align 8
+; CHECK-NEXT:    [[TMP7:%.*]] = getelementptr i8, ptr [[DST]], i64 16
+; CHECK-NEXT:    store i64 0, ptr [[TMP7]], align 8
+; CHECK-NEXT:    br label %[[BB4]]
 ;
-; REMARK-LABEL: Function: fun2
-; REMARK: Args:
-; REMARK-NEXT: - String:          'Stores SLP vectorized with cost '
-; REMARK-NEXT: - Cost:            '-1'
+; REMARK-NOT: Function: fun2
----------------
JonPsson1 wrote:

I would, but the problem is that SLPVectorizer prints nothing for fun2 - I guess because it doesn't see any opportunity. This seems to be the case also in other tests, like AArch64/gather-cost.ll.

https://github.com/llvm/llvm-project/pull/112491