Throughput of Resteasy Reactive's significantly decreased in 2.12.0.Final

[yazinnnn]

Describe the bug

when i use 2.11.3.Final, the endpoint's throughput would be 410k rps

wrk -t10 -c500 -d60s http://localhost:8080/bench
Running 1m test @ http://localhost:8080/bench
  10 threads and 500 connections
  Thread Stats   Avg      Stdev     Max   +/- Stdev
    Latency     2.90ms    4.52ms  94.42ms   87.86%
    Req/Sec    41.19k     5.32k   93.90k    72.64%
  24591994 requests in 1.00m, 1.86GB read
Requests/sec: 409195.71
Transfer/sec:     31.61MB

when upgrade to 2.12.0.Final, throughput would decrease to 46k rps

wrk -t10 -c500 -d60s http://localhost:8080/bench
Running 1m test @ http://localhost:8080/bench
  10 threads and 500 connections
  Thread Stats   Avg      Stdev     Max   +/- Stdev
    Latency    11.41ms    6.35ms  88.30ms   71.20%
    Req/Sec     4.55k   349.58     7.88k    71.64%
  2714402 requests in 1.00m, 209.68MB read
Requests/sec:  45169.21
Transfer/sec:      3.49MB

what make this difference?

dependencies

    implementation("io.quarkus:quarkus-resteasy-reactive-jackson")
    implementation("io.quarkus:quarkus-kotlin")
    implementation("org.jetbrains.kotlin:kotlin-stdlib-jdk8")
    implementation("io.quarkus:quarkus-arc")
    implementation("io.quarkus:quarkus-resteasy-reactive")

test endpoint

fun uni(): Uni<String> = Uni.createFrom().item("uni")

cpu

intel 10400 6c12t

Expected behavior

No response

Actual behavior

No response

How to Reproduce?

No response

Output of uname -a or ver

Linux yazi-m820 5.15.57-2-MANJARO #1 SMP PREEMPT Mon Jul 25 01:32:51 UTC 2022 x86_64 GNU/Linux

Output of java -version

openjdk version "17.0.2" 2022-01-18 LTS

GraalVM version (if different from Java)

OpenJDK Runtime Environment GraalVM 22.0.0.2 (build 17.0.2+9-LTS)

Quarkus version or git rev

[2.12.0.Final]

Build tool (ie. output of mvnw --version or gradlew --version)

gradle-7.4.2

Additional information

No response

[quarkus-bot]

/cc @FroMage, @evanchooly, @geoand, @stuartwdouglas

[geoand]

Thanks for reporting.

That's a pretty extrordinary regression, I'll look into it ASAP

[geoand]

I can confirm this, we have a massive performance drop.

[geoand]

This was actually picked up by us as well yesterday and fixed in #27736. The fix will be part of 2.12.1.Final

[yazinnnn]

This was actually picked up by us as well yesterday and fixed in #27736. The fix will be part of 2.12.1.Final

Thanks a lot！

[geoand]

💪🏼