[PATCH] D115497: [Inline] Disable deferred inlining

[Mircea Trofin via Phabricator via llvm-commits]

mtrofin added a comment.

In D115497#3186414 <https://reviews.llvm.org/D115497#3186414>, @mtrofin wrote:

> In D115497#3186304 <https://reviews.llvm.org/D115497#3186304>, @nikic wrote:
>
>> Regarding additional benchmarks, could I ask one of you to trigger a run of Google's internal benchmarks with this patch? Those are probably pretty representative for both PGO and non-PGO workloads.
>
> I'll take care of that over the weekend.

TL;DR; no significant performance effect

I ran 2 things:

- intenal microbenchmarks we use for compiler release, which are using `google benchmark`. I used the perf counter support in that library, and measured `INSTRUCTIONS` and `CYCLES` for a subset that I pre-verified has very low `INSTRUCTIONS` variation (i.e. around 0.2%). The geomean is flat. Per-benchmark, I saw variations sub-2% for either of the 2 counters.
- an internal benchmark for a real application - no significant change


CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D115497/new/

https://reviews.llvm.org/D115497