Replace JCTools queues with VarHandle-based implementations for Java 9+

[amarziali]

What Does This Do

This PR introduces a set of queue implementations in order to replace the JCTools-based queues, eliminating direct usage of sun.misc.Unsafe and providing full compatibility with Java 9+ runtimes through the VarHandle API.

The goal is to achieve similar high-performance concurrent queue behavior as JCTools while using supported, standard Java mechanisms.

A new Queues factory class is introduced to dynamically select the optimal queue implementation based on the Java runtime environment:

• On Java 9 and newer, the factory instantiates the new VarHandle-based queues
• On Java 1.8, it falls back to the existing JCTools-based queues to maintain backward compatibility and performance consistency.

Introduced Classes Summary

Class	Pattern	Description
SpscArrayQueueVarHandle	Single-Producer / Single-Consumer	Lock-free SPSC queue using VarHandles with acquire/release semantics. When possible, eliminates redundant volatile reads via cached head and tail.
SpmcArrayQueueVarHandle	Single-Producer / Multiple-Consumer	Lock-free SPMC queue supporting concurrent consumers using atomic head updates (CAS). Uses consumerLimit caching to reduce volatile contention.
MpscArrayQueueVarHandle<E>	Multiple-Producer / Single-Consumer	Lock-free MPSC queue where producers claim slots via CAS on TAIL_HANDLE. Maintains a producerLimit to minimize volatile head reads.
MpscBlockingConsumerArrayQueueVarHandle<E>	Multiple-Producer / Single-Consumer (Blocking)	Extends MPSC with blocking consumer support. Uses CONSUMER_THREAD_HANDLE to park/unpark the waiting consumer efficiently.

Memory Padding

All queue state fields (head, tail, cached limits, etc.) are cache-line padded to prevent false sharing between producers and consumers.
This ensures that frequently accessed hot fields do not reside on the same cache line across threads, minimizing cache invalidations and improving throughput under contention.

Memory Fence Semantics

Memory fences were explicitly chosen for each access type to minimize volatile overhead while maintaining correct visibility guarantees:

• setRelease / getAcquire for publishing and consuming elements — provides correct inter-thread ordering without full barriers.
• setOpaque / getOpaque for relaxed head/tail updates — avoids unnecessary synchronization costs where ordering is not required.
• getVolatile only used when full memory fences are really required (e.g. refreshing limits to ensure visibility when the queue might be full or empty).

Queue Benchmark Results (ops/us)

Note: SPSC benchmark shows contentions on slow path (i.e. queue is full/queue is empty). This should less frequently happen in our case. Increasing the queue size (hence reducing the probability that's full) shows good performances.

MPSCBlockingConsumer Queue Benchmark (ops/us)

Implementation	Capacity	Total	Consume	Produce
JCTools	1024	41,149	30,661	10,488
VarHandle	1024	258,074	246,683	11,391
JCTools	65536	32,413	24,680	7,733
VarHandle	65536	224,982	217,498	7,485

MPSC Queue Benchmark (ops/us)

Implementation	Capacity	Total	Consume	Produce
JCTools	1024	41,784	31,070	10,715
VarHandle	1024	238,609	222,383	16,226
JCTools	65536	39,589	32,370	7,219
VarHandle	65536	262,729	250,627	12,102

SPSC Queue Benchmark (ops/us)

Implementation	Capacity	Total	Consume	Produce
JCTools	1024	259,418	129,694	129,724
VarHandle	1024	101,007	72,542	28,465
JCTools	65536	537,111	268,577	268,534
VarHandle	65536	353,161	191,188	161,973

Takeaways:

• MPSC queues can be replaced with a VarHandle equivalent. In some cases, the new implementation shows even better performances.
• jctools still outperforms in SPSC queues. Perhaps the VarHandle implementation might be optimised further.

Room for future improvements

In high-throughput scenarios where multiple producers compete for queue space, contention on the CAS operation can become a bottleneck.

Idea to mitigate this, when the queue is likely not full, a getAndAdd operation can be used instead of a CAS to claim slots since it will never fail. This optimization allows multiple producers to advance the tail index with reduced atomic contention. However, when the queue is nearly full, the getAndAdd cannot be reliably done hence the classic CAS loop (slow path) can be used instead.

Motivation

Additional Notes

Contributor Checklist

• Format the title according the contribution guidelines
• Assign the type: and (comp: or inst:) labels in addition to any useful labels
• Don't use close, fix or any linking keywords when referencing an issue.
  Use solves instead, and assign the PR milestone to the issue
• Update the CODEOWNERS file on source file addition, move, or deletion
• Update the public documentation in case of new configuration flag or behavior

Jira ticket: [PROJ-IDENT]

[pr-commenter]

Debugger benchmarks

Parameters

	Baseline	Candidate
baseline_or_candidate	baseline	candidate
ci_job_date	1762350659	1762351004
end_time	2025-11-05T13:52:20	2025-11-05T13:58:05
git_branch	master	andrea.marziali/remove-jctools-queues
git_commit_sha	8db72c0	a880c67
start_time	2025-11-05T13:51:00	2025-11-05T13:56:45

See matching parameters

	Baseline	Candidate
ci_job_id	1217030429	1217030429
ci_pipeline_id	81317491	81317491
cpu_model	Intel(R) Xeon(R) Platinum 8259CL CPU @ 2.50GHz	Intel(R) Xeon(R) Platinum 8259CL CPU @ 2.50GHz
git_commit_date	1762349967	1762349967

Summary

Found 0 performance improvements and 0 performance regressions! Performance is the same for 10 metrics, 5 unstable metrics.

See unchanged results

scenario	Δ mean agg_http_req_duration_min	Δ mean agg_http_req_duration_p50	Δ mean agg_http_req_duration_p75	Δ mean agg_http_req_duration_p99	Δ mean throughput
scenario:noprobe	unstable [-16.952µs; +19.877µs] or [-6.074%; +7.122%]	unstable [-27.156µs; +31.309µs] or [-8.553%; +9.861%]	unstable [-36.996µs; +42.158µs] or [-11.187%; +12.748%]	unstable [-142.099µs; +61.104µs] or [-13.747%; +5.911%]	same
scenario:basic	same	same	same	unstable [-226.500µs; -8.890µs] or [-25.732%; -1.010%]	same
scenario:loop	unsure [+0.257µs; +4.484µs] or [+0.003%; +0.051%]	unsure [-7.436µs; -0.629µs] or [-0.083%; -0.007%]	unsure [-8.505µs; -0.543µs] or [-0.094%; -0.006%]	same	same

Request duration reports for reports

gantt
    title reports - request duration [CI 0.99] : candidate=None, baseline=None
    dateFormat X
    axisFormat %s
section baseline
noprobe (317.508 µs) : 291, 344
.   : milestone, 318,
basic (294.053 µs) : 287, 301
.   : milestone, 294,
loop (8.959 ms) : 8956, 8963
.   : milestone, 8959,
section candidate
noprobe (319.585 µs) : 290, 349
.   : milestone, 320,
basic (293.196 µs) : 286, 300
.   : milestone, 293,
loop (8.955 ms) : 8952, 8958
.   : milestone, 8955,

Loading

• baseline results

Scenario	Request median duration [CI 0.99]
noprobe	317.508 µs [291.339 µs, 343.677 µs]
basic	294.053 µs [286.681 µs, 301.425 µs]
loop	8.959 ms [8.956 ms, 8.963 ms]

• candidate results

Scenario	Request median duration [CI 0.99]
noprobe	319.585 µs [290.19 µs, 348.98 µs]
basic	293.196 µs [286.34 µs, 300.053 µs]
loop	8.955 ms [8.952 ms, 8.958 ms]

[pr-commenter]

Benchmarks

Startup

Parameters

	Baseline	Candidate
baseline_or_candidate	baseline	candidate
git_branch	master	andrea.marziali/remove-jctools-queues
git_commit_date	1764344838	1764345098
git_commit_sha	17c7fcf	9a902be
release_version	1.57.0-SNAPSHOT~17c7fcf3e9	1.57.0-SNAPSHOT~9a902be485

See matching parameters

	Baseline	Candidate
application	insecure-bank	insecure-bank
ci_job_date	1764346917	1764346917
ci_job_id	1262017328	1262017328
ci_pipeline_id	84364713	84364713
cpu_model	Intel(R) Xeon(R) Platinum 8259CL CPU @ 2.50GHz	Intel(R) Xeon(R) Platinum 8259CL CPU @ 2.50GHz
kernel_version	Linux runner-zfyrx7zua-project-304-concurrent-1-mxxtzcsw 6.8.0-1031-aws #33~22.04.1-Ubuntu SMP Thu Jun 26 14:22:30 UTC 2025 x86_64 x86_64 x86_64 GNU/Linux	Linux runner-zfyrx7zua-project-304-concurrent-1-mxxtzcsw 6.8.0-1031-aws #33~22.04.1-Ubuntu SMP Thu Jun 26 14:22:30 UTC 2025 x86_64 x86_64 x86_64 GNU/Linux
module	Agent	Agent
parent	None	None

Summary

Found 3 performance improvements and 0 performance regressions! Performance is the same for 52 metrics, 10 unstable metrics.

scenario	Δ mean execution_time	candidate mean execution_time	baseline mean execution_time
scenario:startup:insecure-bank:iast:Flare Poller	better [-7.114ms; -6.569ms] or [-65.496%; -60.474%]	4.021ms	10.863ms
scenario:startup:petclinic:iast:Flare Poller	better [-6.966ms; -6.462ms] or [-64.882%; -60.181%]	4.023ms	10.737ms
scenario:startup:petclinic:profiling:Debugger	better [-456.469µs; -182.922µs] or [-6.685%; -2.679%]	6.508ms	6.828ms

Startup time reports for insecure-bank

gantt
    title insecure-bank - global startup overhead: candidate=1.57.0-SNAPSHOT~9a902be485, baseline=1.57.0-SNAPSHOT~17c7fcf3e9

    dateFormat X
    axisFormat %s
section tracing
Agent [baseline] (1.051 s) : 0, 1050647
Total [baseline] (8.702 s) : 0, 8702333
Agent [candidate] (1.054 s) : 0, 1053780
Total [candidate] (8.695 s) : 0, 8694955
section iast
Agent [baseline] (1.204 s) : 0, 1203661
Total [baseline] (9.395 s) : 0, 9394979
Agent [candidate] (1.196 s) : 0, 1195813
Total [candidate] (9.354 s) : 0, 9354072

Loading

• baseline results

Module	Variant	Duration	Δ tracing
Agent	tracing	1.051 s	-
Agent	iast	1.204 s	153.014 ms (14.6%)
Total	tracing	8.702 s	-
Total	iast	9.395 s	692.646 ms (8.0%)

• candidate results

Module	Variant	Duration	Δ tracing
Agent	tracing	1.054 s	-
Agent	iast	1.196 s	142.033 ms (13.5%)
Total	tracing	8.695 s	-
Total	iast	9.354 s	659.117 ms (7.6%)

gantt
    title insecure-bank - break down per module: candidate=1.57.0-SNAPSHOT~9a902be485, baseline=1.57.0-SNAPSHOT~17c7fcf3e9

    dateFormat X
    axisFormat %s
section tracing
crashtracking [baseline] (1.501 ms) : 0, 1501
crashtracking [candidate] (1.49 ms) : 0, 1490
BytebuddyAgent [baseline] (705.353 ms) : 0, 705353
BytebuddyAgent [candidate] (711.982 ms) : 0, 711982
GlobalTracer [baseline] (249.125 ms) : 0, 249125
GlobalTracer [candidate] (245.517 ms) : 0, 245517
AppSec [baseline] (31.986 ms) : 0, 31986
AppSec [candidate] (32.292 ms) : 0, 32292
Debugger [baseline] (6.425 ms) : 0, 6425
Debugger [candidate] (6.398 ms) : 0, 6398
Remote Config [baseline] (676.635 µs) : 0, 677
Remote Config [candidate] (691.263 µs) : 0, 691
Telemetry [baseline] (16.688 ms) : 0, 16688
Telemetry [candidate] (11.081 ms) : 0, 11081
Flare Poller [baseline] (4.111 ms) : 0, 4111
Flare Poller [candidate] (9.399 ms) : 0, 9399
section iast
crashtracking [baseline] (1.484 ms) : 0, 1484
crashtracking [candidate] (1.491 ms) : 0, 1491
BytebuddyAgent [baseline] (839.472 ms) : 0, 839472
BytebuddyAgent [candidate] (844.407 ms) : 0, 844407
GlobalTracer [baseline] (239.787 ms) : 0, 239787
GlobalTracer [candidate] (233.853 ms) : 0, 233853
AppSec [baseline] (30.888 ms) : 0, 30888
AppSec [candidate] (29.319 ms) : 0, 29319
Debugger [baseline] (6.173 ms) : 0, 6173
Debugger [candidate] (6.002 ms) : 0, 6002
Remote Config [baseline] (635.14 µs) : 0, 635
Remote Config [candidate] (624.26 µs) : 0, 624
Telemetry [baseline] (8.074 ms) : 0, 8074
Telemetry [candidate] (7.935 ms) : 0, 7935
Flare Poller [baseline] (10.863 ms) : 0, 10863
Flare Poller [candidate] (4.021 ms) : 0, 4021
IAST [baseline] (31.36 ms) : 0, 31360
IAST [candidate] (32.955 ms) : 0, 32955

Loading

Startup time reports for petclinic

gantt
    title petclinic - global startup overhead: candidate=1.57.0-SNAPSHOT~9a902be485, baseline=1.57.0-SNAPSHOT~17c7fcf3e9

    dateFormat X
    axisFormat %s
section tracing
Agent [baseline] (1.057 s) : 0, 1057036
Total [baseline] (10.788 s) : 0, 10788279
Agent [candidate] (1.049 s) : 0, 1048628
Total [candidate] (10.773 s) : 0, 10773229
section appsec
Agent [baseline] (1.223 s) : 0, 1222705
Total [baseline] (10.889 s) : 0, 10889247
Agent [candidate] (1.221 s) : 0, 1221015
Total [candidate] (10.874 s) : 0, 10874446
section iast
Agent [baseline] (1.195 s) : 0, 1195366
Total [baseline] (11.167 s) : 0, 11166550
Agent [candidate] (1.184 s) : 0, 1184296
Total [candidate] (11.173 s) : 0, 11173407
section profiling
Agent [baseline] (1.196 s) : 0, 1196480
Total [baseline] (10.915 s) : 0, 10914884
Agent [candidate] (1.193 s) : 0, 1193101
Total [candidate] (10.902 s) : 0, 10902174

Loading

• baseline results

Module	Variant	Duration	Δ tracing
Agent	tracing	1.057 s	-
Agent	appsec	1.223 s	165.669 ms (15.7%)
Agent	iast	1.195 s	138.329 ms (13.1%)
Agent	profiling	1.196 s	139.444 ms (13.2%)
Total	tracing	10.788 s	-
Total	appsec	10.889 s	100.968 ms (0.9%)
Total	iast	11.167 s	378.272 ms (3.5%)
Total	profiling	10.915 s	126.606 ms (1.2%)

• candidate results

Module	Variant	Duration	Δ tracing
Agent	tracing	1.049 s	-
Agent	appsec	1.221 s	172.387 ms (16.4%)
Agent	iast	1.184 s	135.669 ms (12.9%)
Agent	profiling	1.193 s	144.473 ms (13.8%)
Total	tracing	10.773 s	-
Total	appsec	10.874 s	101.217 ms (0.9%)
Total	iast	11.173 s	400.178 ms (3.7%)
Total	profiling	10.902 s	128.945 ms (1.2%)

gantt
    title petclinic - break down per module: candidate=1.57.0-SNAPSHOT~9a902be485, baseline=1.57.0-SNAPSHOT~17c7fcf3e9

    dateFormat X
    axisFormat %s
section tracing
crashtracking [baseline] (1.495 ms) : 0, 1495
crashtracking [candidate] (1.471 ms) : 0, 1471
BytebuddyAgent [baseline] (711.52 ms) : 0, 711520
BytebuddyAgent [candidate] (707.738 ms) : 0, 707738
GlobalTracer [baseline] (249.308 ms) : 0, 249308
GlobalTracer [candidate] (245.056 ms) : 0, 245056
AppSec [baseline] (32.159 ms) : 0, 32159
AppSec [candidate] (32.124 ms) : 0, 32124
Debugger [baseline] (6.428 ms) : 0, 6428
Debugger [candidate] (6.382 ms) : 0, 6382
Remote Config [baseline] (671.387 µs) : 0, 671
Remote Config [candidate] (671.318 µs) : 0, 671
Telemetry [baseline] (14.913 ms) : 0, 14913
Telemetry [candidate] (10.281 ms) : 0, 10281
Flare Poller [baseline] (5.61 ms) : 0, 5610
Flare Poller [candidate] (10.178 ms) : 0, 10178
section appsec
crashtracking [baseline] (1.473 ms) : 0, 1473
crashtracking [candidate] (1.47 ms) : 0, 1470
BytebuddyAgent [baseline] (728.101 ms) : 0, 728101
BytebuddyAgent [candidate] (731.141 ms) : 0, 731141
GlobalTracer [baseline] (239.975 ms) : 0, 239975
GlobalTracer [candidate] (235.947 ms) : 0, 235947
IAST [baseline] (24.746 ms) : 0, 24746
IAST [candidate] (24.635 ms) : 0, 24635
AppSec [baseline] (174.278 ms) : 0, 174278
AppSec [candidate] (174.001 ms) : 0, 174001
Debugger [baseline] (6.307 ms) : 0, 6307
Debugger [candidate] (6.118 ms) : 0, 6118
Remote Config [baseline] (683.246 µs) : 0, 683
Remote Config [candidate] (685.189 µs) : 0, 685
Telemetry [baseline] (8.218 ms) : 0, 8218
Telemetry [candidate] (8.183 ms) : 0, 8183
Flare Poller [baseline] (3.977 ms) : 0, 3977
Flare Poller [candidate] (4.044 ms) : 0, 4044
section iast
crashtracking [baseline] (1.483 ms) : 0, 1483
crashtracking [candidate] (1.491 ms) : 0, 1491
BytebuddyAgent [baseline] (833.751 ms) : 0, 833751
BytebuddyAgent [candidate] (834.347 ms) : 0, 834347
GlobalTracer [baseline] (238.528 ms) : 0, 238528
GlobalTracer [candidate] (233.959 ms) : 0, 233959
IAST [baseline] (32.75 ms) : 0, 32750
IAST [candidate] (33.388 ms) : 0, 33388
AppSec [baseline] (28.637 ms) : 0, 28637
AppSec [candidate] (27.083 ms) : 0, 27083
Debugger [baseline] (6.038 ms) : 0, 6038
Debugger [candidate] (6.001 ms) : 0, 6001
Remote Config [baseline] (611.783 µs) : 0, 612
Remote Config [candidate] (593.025 µs) : 0, 593
Telemetry [baseline] (7.979 ms) : 0, 7979
Telemetry [candidate] (7.891 ms) : 0, 7891
Flare Poller [baseline] (10.737 ms) : 0, 10737
Flare Poller [candidate] (4.023 ms) : 0, 4023
section profiling
crashtracking [baseline] (1.427 ms) : 0, 1427
crashtracking [candidate] (1.451 ms) : 0, 1451
BytebuddyAgent [baseline] (732.71 ms) : 0, 732710
BytebuddyAgent [candidate] (734.4 ms) : 0, 734400
GlobalTracer [baseline] (222.019 ms) : 0, 222019
GlobalTracer [candidate] (217.783 ms) : 0, 217783
AppSec [baseline] (32.14 ms) : 0, 32140
AppSec [candidate] (32.331 ms) : 0, 32331
Debugger [baseline] (6.828 ms) : 0, 6828
Debugger [candidate] (6.508 ms) : 0, 6508
Remote Config [baseline] (700.953 µs) : 0, 701
Remote Config [candidate] (716.138 µs) : 0, 716
Telemetry [baseline] (16.456 ms) : 0, 16456
Telemetry [candidate] (16.472 ms) : 0, 16472
Flare Poller [baseline] (4.129 ms) : 0, 4129
Flare Poller [candidate] (4.134 ms) : 0, 4134
ProfilingAgent [baseline] (110.996 ms) : 0, 110996
ProfilingAgent [candidate] (109.961 ms) : 0, 109961
Profiling [baseline] (111.631 ms) : 0, 111631
Profiling [candidate] (110.623 ms) : 0, 110623

Loading

Load

Parameters

	Baseline	Candidate
baseline_or_candidate	baseline	candidate
git_branch	master	andrea.marziali/remove-jctools-queues
git_commit_date	1764344838	1764345098
git_commit_sha	17c7fcf	9a902be
release_version	1.57.0-SNAPSHOT~17c7fcf3e9	1.57.0-SNAPSHOT~9a902be485

See matching parameters

	Baseline	Candidate
application	insecure-bank	insecure-bank
ci_job_date	1764347418	1764347418
ci_job_id	1262017330	1262017330
ci_pipeline_id	84364713	84364713
cpu_model	Intel(R) Xeon(R) Platinum 8259CL CPU @ 2.50GHz	Intel(R) Xeon(R) Platinum 8259CL CPU @ 2.50GHz
kernel_version	Linux runner-zfyrx7zua-project-304-concurrent-1-vpizpu5p 6.8.0-1031-aws #33~22.04.1-Ubuntu SMP Thu Jun 26 14:22:30 UTC 2025 x86_64 x86_64 x86_64 GNU/Linux	Linux runner-zfyrx7zua-project-304-concurrent-1-vpizpu5p 6.8.0-1031-aws #33~22.04.1-Ubuntu SMP Thu Jun 26 14:22:30 UTC 2025 x86_64 x86_64 x86_64 GNU/Linux

Summary

Found 4 performance improvements and 4 performance regressions! Performance is the same for 11 metrics, 17 unstable metrics.

scenario	Δ mean agg_http_req_duration_p50	Δ mean agg_http_req_duration_p95	Δ mean throughput	candidate mean agg_http_req_duration_p50	candidate mean agg_http_req_duration_p95	candidate mean throughput	baseline mean agg_http_req_duration_p50	baseline mean agg_http_req_duration_p95	baseline mean throughput
scenario:load:insecure-bank:iast:high_load	worse [+168.703µs; +253.184µs] or [+6.713%; +10.075%]	worse [+370.853µs; +695.788µs] or [+5.070%; +9.512%]	unstable [-243.368op/s; +44.243op/s] or [-17.144%; +3.117%]	2.724ms	7.848ms	1320.000op/s	2.513ms	7.315ms	1419.562op/s
scenario:load:insecure-bank:iast_FULL:high_load	better [-316.023µs; -106.488µs] or [-5.999%; -2.022%]	same [-615.548µs; +32.558µs] or [-4.961%; +0.262%]	unstable [-59.279op/s; +105.404op/s] or [-7.559%; +13.441%]	5.056ms	12.117ms	807.281op/s	5.268ms	12.408ms	784.219op/s
scenario:load:petclinic:profiling:high_load	worse [+0.869ms; +1.854ms] or [+4.777%; +10.196%]	worse [+1.045ms; +2.649ms] or [+3.535%; +8.958%]	unstable [-39.782op/s; +9.657op/s] or [-15.818%; +3.840%]	19.544ms	31.423ms	236.438op/s	18.183ms	29.575ms	251.500op/s
scenario:load:petclinic:iast:high_load	better [-1308.411µs; -551.847µs] or [-7.080%; -2.986%]	unsure [-1.916ms; -0.525ms] or [-6.351%; -1.741%]	unstable [-10.721op/s; +39.658op/s] or [-4.353%; +16.103%]	17.551ms	28.948ms	260.750op/s	18.481ms	30.169ms	246.281op/s
scenario:load:petclinic:code_origins:high_load	better [-2.048ms; -1.279ms] or [-10.864%; -6.783%]	better [-2.547ms; -1.285ms] or [-8.405%; -4.240%]	unstable [-4.108op/s; +47.921op/s] or [-1.682%; +19.620%]	17.191ms	28.391ms	266.156op/s	18.854ms	30.307ms	244.250op/s

Request duration reports for insecure-bank

gantt
    title insecure-bank - request duration [CI 0.99] : candidate=1.57.0-SNAPSHOT~9a902be485, baseline=1.57.0-SNAPSHOT~17c7fcf3e9
    dateFormat X
    axisFormat %s
section baseline
no_agent (1.2 ms) : 1188, 1211
.   : milestone, 1200,
iast (3.231 ms) : 3185, 3277
.   : milestone, 3231,
iast_FULL (5.897 ms) : 5837, 5956
.   : milestone, 5897,
iast_GLOBAL (3.72 ms) : 3660, 3781
.   : milestone, 3720,
profiling (2.004 ms) : 1986, 2022
.   : milestone, 2004,
tracing (1.829 ms) : 1813, 1844
.   : milestone, 1829,
section candidate
no_agent (1.201 ms) : 1189, 1212
.   : milestone, 1201,
iast (3.472 ms) : 3423, 3520
.   : milestone, 3472,
iast_FULL (5.723 ms) : 5666, 5780
.   : milestone, 5723,
iast_GLOBAL (3.582 ms) : 3530, 3634
.   : milestone, 3582,
profiling (2.134 ms) : 2116, 2153
.   : milestone, 2134,
tracing (1.816 ms) : 1800, 1831
.   : milestone, 1816,

Loading

• baseline results

Variant	Request duration [CI 0.99]	Δ no_agent
no_agent	1.2 ms [1.188 ms, 1.211 ms]	-
iast	3.231 ms [3.185 ms, 3.277 ms]	2.032 ms (169.4%)
iast_FULL	5.897 ms [5.837 ms, 5.956 ms]	4.697 ms (391.6%)
iast_GLOBAL	3.72 ms [3.66 ms, 3.781 ms]	2.52 ms (210.1%)
profiling	2.004 ms [1.986 ms, 2.022 ms]	804.678 µs (67.1%)
tracing	1.829 ms [1.813 ms, 1.844 ms]	629.1 µs (52.4%)

• candidate results

Variant	Request duration [CI 0.99]	Δ no_agent
no_agent	1.201 ms [1.189 ms, 1.212 ms]	-
iast	3.472 ms [3.423 ms, 3.52 ms]	2.271 ms (189.1%)
iast_FULL	5.723 ms [5.666 ms, 5.78 ms]	4.522 ms (376.6%)
iast_GLOBAL	3.582 ms [3.53 ms, 3.634 ms]	2.382 ms (198.3%)
profiling	2.134 ms [2.116 ms, 2.153 ms]	933.663 µs (77.8%)
tracing	1.816 ms [1.8 ms, 1.831 ms]	614.867 µs (51.2%)

Request duration reports for petclinic

gantt
    title petclinic - request duration [CI 0.99] : candidate=1.57.0-SNAPSHOT~9a902be485, baseline=1.57.0-SNAPSHOT~17c7fcf3e9
    dateFormat X
    axisFormat %s
section baseline
no_agent (17.824 ms) : 17638, 18011
.   : milestone, 17824,
appsec (18.674 ms) : 18481, 18867
.   : milestone, 18674,
code_origins (19.115 ms) : 18925, 19305
.   : milestone, 19115,
iast (18.952 ms) : 18760, 19145
.   : milestone, 18952,
profiling (18.56 ms) : 18376, 18744
.   : milestone, 18560,
tracing (18.807 ms) : 18619, 18995
.   : milestone, 18807,
section candidate
no_agent (19.045 ms) : 18851, 19240
.   : milestone, 19045,
appsec (18.367 ms) : 18181, 18554
.   : milestone, 18367,
code_origins (17.53 ms) : 17358, 17703
.   : milestone, 17530,
iast (17.898 ms) : 17718, 18078
.   : milestone, 17898,
profiling (19.752 ms) : 19547, 19956
.   : milestone, 19752,
tracing (18.84 ms) : 18646, 19034
.   : milestone, 18840,

Loading

• baseline results

Variant	Request duration [CI 0.99]	Δ no_agent
no_agent	17.824 ms [17.638 ms, 18.011 ms]	-
appsec	18.674 ms [18.481 ms, 18.867 ms]	849.698 µs (4.8%)
code_origins	19.115 ms [18.925 ms, 19.305 ms]	1.291 ms (7.2%)
iast	18.952 ms [18.76 ms, 19.145 ms]	1.128 ms (6.3%)
profiling	18.56 ms [18.376 ms, 18.744 ms]	735.371 µs (4.1%)
tracing	18.807 ms [18.619 ms, 18.995 ms]	982.494 µs (5.5%)

• candidate results

Variant	Request duration [CI 0.99]	Δ no_agent
no_agent	19.045 ms [18.851 ms, 19.24 ms]	-
appsec	18.367 ms [18.181 ms, 18.554 ms]	-678.018 µs (-3.6%)
code_origins	17.53 ms [17.358 ms, 17.703 ms]	-1.515 ms (-8.0%)
iast	17.898 ms [17.718 ms, 18.078 ms]	-1.148 ms (-6.0%)
profiling	19.752 ms [19.547 ms, 19.956 ms]	706.306 µs (3.7%)
tracing	18.84 ms [18.646 ms, 19.034 ms]	-205.506 µs (-1.1%)

Dacapo

Parameters

	Baseline	Candidate
baseline_or_candidate	baseline	candidate
git_branch	master	andrea.marziali/remove-jctools-queues
git_commit_date	1764344838	1764345098
git_commit_sha	17c7fcf	9a902be
release_version	1.57.0-SNAPSHOT~17c7fcf3e9	1.57.0-SNAPSHOT~9a902be485

See matching parameters

	Baseline	Candidate
application	biojava	biojava
ci_job_date	1764347140	1764347140
ci_job_id	1262017332	1262017332
ci_pipeline_id	84364713	84364713
cpu_model	Intel(R) Xeon(R) Platinum 8259CL CPU @ 2.50GHz	Intel(R) Xeon(R) Platinum 8259CL CPU @ 2.50GHz
kernel_version	Linux runner-zfyrx7zua-project-304-concurrent-2-yuiicez9 6.8.0-1031-aws #33~22.04.1-Ubuntu SMP Thu Jun 26 14:22:30 UTC 2025 x86_64 x86_64 x86_64 GNU/Linux	Linux runner-zfyrx7zua-project-304-concurrent-2-yuiicez9 6.8.0-1031-aws #33~22.04.1-Ubuntu SMP Thu Jun 26 14:22:30 UTC 2025 x86_64 x86_64 x86_64 GNU/Linux

Summary

Found 0 performance improvements and 0 performance regressions! Performance is the same for 11 metrics, 1 unstable metrics.

Execution time for biojava

gantt
    title biojava - execution time [CI 0.99] : candidate=1.57.0-SNAPSHOT~9a902be485, baseline=1.57.0-SNAPSHOT~17c7fcf3e9
    dateFormat X
    axisFormat %s
section baseline
no_agent (15.404 s) : 15404000, 15404000
.   : milestone, 15404000,
appsec (14.613 s) : 14613000, 14613000
.   : milestone, 14613000,
iast (18.56 s) : 18560000, 18560000
.   : milestone, 18560000,
iast_GLOBAL (17.961 s) : 17961000, 17961000
.   : milestone, 17961000,
profiling (14.806 s) : 14806000, 14806000
.   : milestone, 14806000,
tracing (14.554 s) : 14554000, 14554000
.   : milestone, 14554000,
section candidate
no_agent (15.362 s) : 15362000, 15362000
.   : milestone, 15362000,
appsec (14.667 s) : 14667000, 14667000
.   : milestone, 14667000,
iast (18.461 s) : 18461000, 18461000
.   : milestone, 18461000,
iast_GLOBAL (18.006 s) : 18006000, 18006000
.   : milestone, 18006000,
profiling (14.889 s) : 14889000, 14889000
.   : milestone, 14889000,
tracing (14.759 s) : 14759000, 14759000
.   : milestone, 14759000,

Loading

• baseline results

Variant	Execution Time [CI 0.99]	Δ no_agent
no_agent	15.404 s [15.404 s, 15.404 s]	-
appsec	14.613 s [14.613 s, 14.613 s]	-791.0 ms (-5.1%)
iast	18.56 s [18.56 s, 18.56 s]	3.156 s (20.5%)
iast_GLOBAL	17.961 s [17.961 s, 17.961 s]	2.557 s (16.6%)
profiling	14.806 s [14.806 s, 14.806 s]	-598.0 ms (-3.9%)
tracing	14.554 s [14.554 s, 14.554 s]	-850.0 ms (-5.5%)

• candidate results

Variant	Execution Time [CI 0.99]	Δ no_agent
no_agent	15.362 s [15.362 s, 15.362 s]	-
appsec	14.667 s [14.667 s, 14.667 s]	-695.0 ms (-4.5%)
iast	18.461 s [18.461 s, 18.461 s]	3.099 s (20.2%)
iast_GLOBAL	18.006 s [18.006 s, 18.006 s]	2.644 s (17.2%)
profiling	14.889 s [14.889 s, 14.889 s]	-473.0 ms (-3.1%)
tracing	14.759 s [14.759 s, 14.759 s]	-603.0 ms (-3.9%)

Execution time for tomcat

gantt
    title tomcat - execution time [CI 0.99] : candidate=1.57.0-SNAPSHOT~9a902be485, baseline=1.57.0-SNAPSHOT~17c7fcf3e9
    dateFormat X
    axisFormat %s
section baseline
no_agent (1.472 ms) : 1460, 1483
.   : milestone, 1472,
appsec (3.646 ms) : 3433, 3860
.   : milestone, 3646,
iast (2.205 ms) : 2140, 2269
.   : milestone, 2205,
iast_GLOBAL (2.252 ms) : 2187, 2317
.   : milestone, 2252,
profiling (2.076 ms) : 2023, 2130
.   : milestone, 2076,
tracing (2.038 ms) : 1987, 2088
.   : milestone, 2038,
section candidate
no_agent (1.476 ms) : 1464, 1487
.   : milestone, 1476,
appsec (3.703 ms) : 3485, 3922
.   : milestone, 3703,
iast (2.202 ms) : 2138, 2266
.   : milestone, 2202,
iast_GLOBAL (2.252 ms) : 2187, 2317
.   : milestone, 2252,
profiling (2.067 ms) : 2015, 2119
.   : milestone, 2067,
tracing (2.045 ms) : 1995, 2096
.   : milestone, 2045,

Loading

• baseline results

Variant	Execution Time [CI 0.99]	Δ no_agent
no_agent	1.472 ms [1.46 ms, 1.483 ms]	-
appsec	3.646 ms [3.433 ms, 3.86 ms]	2.175 ms (147.8%)
iast	2.205 ms [2.14 ms, 2.269 ms]	732.779 µs (49.8%)
iast_GLOBAL	2.252 ms [2.187 ms, 2.317 ms]	780.218 µs (53.0%)
profiling	2.076 ms [2.023 ms, 2.13 ms]	604.676 µs (41.1%)
tracing	2.038 ms [1.987 ms, 2.088 ms]	565.89 µs (38.4%)

• candidate results

Variant	Execution Time [CI 0.99]	Δ no_agent
no_agent	1.476 ms [1.464 ms, 1.487 ms]	-
appsec	3.703 ms [3.485 ms, 3.922 ms]	2.228 ms (151.0%)
iast	2.202 ms [2.138 ms, 2.266 ms]	726.576 µs (49.2%)
iast_GLOBAL	2.252 ms [2.187 ms, 2.317 ms]	776.393 µs (52.6%)
profiling	2.067 ms [2.015 ms, 2.119 ms]	591.001 µs (40.1%)
tracing	2.045 ms [1.995 ms, 2.096 ms]	569.823 µs (38.6%)

[github-actions]

Hi! 👋 Thanks for your pull request! 🎉

To help us review it, please make sure to:

• Add at least one type, and one component or instrumentation label to the pull request

If you need help, please check our contributing guidelines.

[franz1981]

Hi @amarziali I am one of the developers of JCTools and we are super happy if we could bring a var handle generation variant in our lib as well.
I can see the value and faster feedback/different ownership of having a stripped version of our dependency (which can still be obtain via shading actually...), but I believe would be a great community value if we could join efforts...plus, we love contributions ☺️

Note: JCTools is at the very core of other frameworks which will soon hit the "no unsafe world" JVM barrier, including Netty.
I'm recently working hard to improve it re this aspect, and JCtools is one of the key but missing pieces there too.
Which means that contributing to JCTools would bring an enormous value to Netty and to many others very impactful projects as well
🙏

[bric3]

First batch of comments, questions, open thoughts...

[bric3]

suggestion:

Avoids spotless meaningless reformat

Suggested change

	// -1 in two's complement = 0xFFFF_FFFF (all bits set to 1). Unsigned right-shifting by n: (-1
	// >>> n) produces a mask of (32 - n) one-bits.
	// -1 in two's complement = 0xFFFF_FFFF (all bits set to 1).
	// Unsigned right-shifting by n: (-1 // >>> n) produces a mask
	// of (32 - n) one-bits.

[bric3]

thought: Do we need javadoc for a submodule ?

[amarziali]

Good question. To be fair I copied paste from another in the same utils folder. I can perhaps get rid of

[bric3]

thought: It might worth it to have benchmarks on aarch64 and x86_64 ?

As there might be differences how VarHandle translates to machine code.

[amarziali]

As far as I know, fences are cheaper on x86_64, so running the JMH benchmark on the worst-case architecture should already give a meaningful picture. Just keep in mind that benchmarks of this type tend to be very sensitive to machine conditions because the operations are so small. My goal was only to get a rough estimate of the expected behaviour, not to push the microbenchmarking further. In realistic scenarios, I’m more interested in understanding the impact on the host application, and that’s already covered by the macro benchmarks we have.

[bric3]

suggestion: Removed tabs + alignment

Suggested change

	/*
	Benchmark (capacity) Mode Cnt Score Error Units
	MPSCQueueBenchmark.queueTest 1024 thrpt 238,609 ops/us
	MPSCQueueBenchmark.queueTest:consume 1024 thrpt 222,383 ops/us
	MPSCQueueBenchmark.queueTest:produce 1024 thrpt 16,226 ops/us
	MPSCQueueBenchmark.queueTest 65536 thrpt 262,729 ops/us
	MPSCQueueBenchmark.queueTest:consume 65536 thrpt 250,627 ops/us
	MPSCQueueBenchmark.queueTest:produce 65536 thrpt 12,102 ops/us
	*/
	/*
	Benchmark (capacity) Mode Cnt Score Error Units
	MPSCQueueBenchmark.queueTest 1024 thrpt 238,609 ops/us
	MPSCQueueBenchmark.queueTest:consume 1024 thrpt 222,383 ops/us
	MPSCQueueBenchmark.queueTest:produce 1024 thrpt 16,226 ops/us
	MPSCQueueBenchmark.queueTest 65536 thrpt 262,729 ops/us
	MPSCQueueBenchmark.queueTest:consume 65536 thrpt 250,627 ops/us
	MPSCQueueBenchmark.queueTest:produce 65536 thrpt 12,102 ops/us
	*/

[bric3]

suggestion:

Suggested change

	final Object[] localBuffer = this.buffer;
	@SuppressWarnings("UnnecessaryLocalVariable")
	final Object[] localBuffer = this.buffer;

[bric3]

note: It doesn't seem to be always used as "sequence".

[bric3]

nitpick:

Suggested change

	public BaseQueue(int requestedCapacity) {
	public BaseQueue(int requestedCapacity) {
	if (requestedCapacity < 1) { throw new IllArgumentException("Size needs to be at least 1"); }

[bric3]

Suggested change

;

[bric3]

thought: I wonder if it makes to communicate via the type, the properties of this queue here, i.e. multiple producers single consumer?
So reader of this code don't have to wonder the connection between the customers and producers.

[amarziali]

It would require creating too many additional types, which I deliberately want to avoid. Functionally, it wouldn’t change anything. I think it’s sufficient for the reader to simply check which queue variant is requested at creation time.

[bric3]

Might be worth some larger comment here ?