[llvm] [VPlan] Add ExtractLane VPInst to extract across multiple parts. (PR #148817)

[David Sherwood via llvm-commits]

================
@@ -46,8 +77,27 @@ define i64 @same_exit_block_pre_inc_use1() #0 {
 ; CHECK-NEXT:    [[CMP_N:%.*]] = icmp eq i64 510, [[N_VEC]]
 ; CHECK-NEXT:    br i1 [[CMP_N]], label [[LOOP_END:%.*]], label [[SCALAR_PH]]
 ; CHECK:       vector.early.exit:
+; CHECK-NEXT:    [[TMP39:%.*]] = call i64 @llvm.vscale.i64()
+; CHECK-NEXT:    [[TMP40:%.*]] = mul nuw i64 [[TMP39]], 16
+; CHECK-NEXT:    [[TMP41:%.*]] = call i64 @llvm.experimental.cttz.elts.i64.nxv16i1(<vscale x 16 x i1> [[TMP59]], i1 true)
+; CHECK-NEXT:    [[TMP42:%.*]] = mul i64 [[TMP40]], 3
+; CHECK-NEXT:    [[TMP43:%.*]] = add i64 [[TMP42]], [[TMP41]]
+; CHECK-NEXT:    [[TMP44:%.*]] = call i64 @llvm.experimental.cttz.elts.i64.nxv16i1(<vscale x 16 x i1> [[TMP31]], i1 true)
+; CHECK-NEXT:    [[TMP45:%.*]] = mul i64 [[TMP40]], 2
+; CHECK-NEXT:    [[TMP46:%.*]] = add i64 [[TMP45]], [[TMP44]]
+; CHECK-NEXT:    [[TMP47:%.*]] = icmp ne i64 [[TMP44]], [[TMP40]]
----------------
david-arm wrote:

I just realised this IR is broken I think. Since the cttz intrinsics have been split up into parts you can have situations like this:

`[[TMP41:%.*]] = call i64 @llvm.experimental.cttz.elts.i64.nxv16i1(<vscale x 16 x i1> [[TMP59]], i1 true)` -> 3
and
`[[TMP44:%.*]] = call i64 @llvm.experimental.cttz.elts.i64.nxv16i1(<vscale x 16 x i1> [[TMP31]], i1 true)` -> 7

which is because there can be multiple triggering exit conditions, but the *first* exit condition was in part 1. Therefore, it's not valid to use the condition `[[TMP44]] != [[TMP40]]` for selecting between [[TMP46]] and [[TMP43]]. I think really this should be

```
; CHECK-NEXT:    [[TMP47:%.*]] = icmp ne i64 [[TMP41]], [[TMP40]]
; CHECK-NEXT:    [[TMP48:%.*]] = select i1 [[TMP47]], i64 [[TMP46]], i64 [[TMP43]]
```

https://github.com/llvm/llvm-project/pull/148817