feat(otel_thread_ctx): thread-level ctx publication

[yannham]

What does this PR do?

This PR implements a first version of publishing side of the OTel proposal on thread-level context sharing.

The changes are two-folds:

1. Add a C shim for accessing the correct thread-local variable that holds the thread context. Indeed the spec mandates the use of the TLSDESC dialect, but this isn't possible in Rust, which uses the classical TLS dialect (gnu instead of gnu2) by default. This is not possible to configure on stable Rust. See additional notes below for more details.
2. Add a new otel_thread_ctx module, similar to the otel_process_ctx one, which provides an abstraction over the thread-level context and handle attach/detach/modify.

The part of the spec around the interned string table, hooking it up in the process-level ctx and tracer metadata is left for future work.

Motivation

See the corresponding OTEP linked above for more details on the motivation.

Additional Notes

TLSDESC is chosen in the spec for performance reasons. I initially found a bit sad that we have to call to a C function from libdatadog to retrieve the TLS address, which incurs an additional cost. I researched potential alternatives:

• force Rust to use the TLSDESC dialect. This is just not possible/supported.
• cache the address (which is stable per thread). But this must be put in...drumroll... another thread-local variable, since it's well, thread-local, so back to square one. One possibility would be to use a cached_addr thread-local and force the access to use the initial-exec model, which is very fast (the price to pay is that libdatadog coulnd't be dlopen at runtime, but I'm not sure there's any usage for that). We would trade a function call for an offset and loads, which I expect to be faster. But this requires nightly Rust, unfortunately, and would apply to the entirety of libdatadog.
• inline-asm: the access sequence for TLSDESC is really simple (it's a few LLVM IR instructions, calling a function obtained from the global offset table). I looked into inline LLVM assembly for Rust, but it's not supported and is anon-goal (LLVM is almost considered as an implementation detail, as rust-gcc could one day become a viable alternative, for example). Native ASM is just too much hassle to pay off.

All in all, I think there's no simple, reasonable, portable and maintainable alternative to the C shim for now. It is worth noting that calling to the C shim is likely to be still faster than the default TLS model chosen by Rust in a dynamic library (the latter requires a function call to __get_tls_addr anyway, but that function is more involved). I wonder if some aggressive cross-language late LTO could inline the C shim.

How to test the change?

There are a bunch of tests in this PR. Ideally we will later test this against other thread-level ctx readers implementations once we hook this work in the FFI; this is left for a follow-up.

[github-actions]

📚 Documentation Check Results

⚠️ 798 documentation warning(s) found

📦 libdd-library-config - 150 warning(s)

📦 libdd-profiling - 648 warning(s)

---

Updated: 2026-03-27 16:19:14 UTC | Commit: ea242af | missing-docs job results

[github-actions]

Clippy Allow Annotation Report

Comparing clippy allow annotations between branches:

• Base Branch: origin/main
• PR Branch: origin/yannham/thread-ctx-tls-shim

Summary by Rule

Rule	Base Branch	PR Branch	Change

Annotation Counts by File

File	Base Branch	PR Branch	Change

Annotation Stats by Crate

Crate	Base Branch	PR Branch	Change
clippy-annotation-reporter	5	5	No change (0%)
datadog-ffe-ffi	1	1	No change (0%)
datadog-ipc	20	20	No change (0%)
datadog-live-debugger	6	6	No change (0%)
datadog-live-debugger-ffi	10	10	No change (0%)
datadog-profiling-replayer	4	4	No change (0%)
datadog-remote-config	3	3	No change (0%)
datadog-sidecar	55	55	No change (0%)
libdd-common	10	10	No change (0%)
libdd-common-ffi	12	12	No change (0%)
libdd-data-pipeline	5	5	No change (0%)
libdd-ddsketch	2	2	No change (0%)
libdd-dogstatsd-client	1	1	No change (0%)
libdd-profiling	13	13	No change (0%)
libdd-telemetry	19	19	No change (0%)
libdd-tinybytes	4	4	No change (0%)
libdd-trace-normalization	2	2	No change (0%)
libdd-trace-obfuscation	8	8	No change (0%)
libdd-trace-utils	15	15	No change (0%)
Total	195	195	No change (0%)

About This Report

This report tracks Clippy allow annotations for specific rules, showing how they've changed in this PR. Decreasing the number of these annotations generally improves code quality.

[github-actions]

🔒 Cargo Deny Results

✅ No issues found!

📦 libdd-library-config - ✅ No issues

📦 libdd-profiling - ✅ No issues

---

Updated: 2026-03-27 16:22:39 UTC | Commit: ea242af | dependency-check job results

[datadog-datadog-prod-us1]

✅ Tests

🎉 All green!

❄️ No new flaky tests detected
🧪 All tests passed

🎯 Code Coverage (details)
• Patch Coverage: 98.80%
• Overall Coverage: 71.28% (+0.11%)

_{This comment will be updated automatically if new data arrives.
🔗 Commit SHA: ce8e143 | Docs | Datadog PR Page | Was this helpful? React with 👍/👎 or give us feedback!}

[codecov-commenter]

Codecov Report

❌ Patch coverage is 98.80478% with 3 lines in your changes missing coverage. Please review.
✅ Project coverage is 71.28%. Comparing base (a29b90b) to head (ce8e143).

Additional details and impacted files

@@            Coverage Diff             @@
##             main    #1791      +/-   ##
==========================================
+ Coverage   71.16%   71.28%   +0.11%     
==========================================
  Files         414      415       +1     
  Lines       65739    65990     +251     
==========================================
+ Hits        46786    47038     +252     
+ Misses      18953    18952       -1     

Components	Coverage Δ
libdd-crashtracker	65.33% <ø> (+0.06%)	⬆️
libdd-crashtracker-ffi	35.24% <ø> (ø)
libdd-alloc	98.77% <ø> (ø)
libdd-data-pipeline	87.55% <ø> (ø)
libdd-data-pipeline-ffi	75.77% <ø> (ø)
libdd-common	79.79% <ø> (ø)
libdd-common-ffi	73.87% <ø> (ø)
libdd-telemetry	62.48% <ø> (ø)
libdd-telemetry-ffi	16.75% <ø> (ø)
libdd-dogstatsd-client	82.64% <ø> (ø)
datadog-ipc	72.56% <ø> (ø)
libdd-profiling	82.06% <98.80%> (+0.43%)	⬆️
libdd-profiling-ffi	64.94% <ø> (ø)
datadog-sidecar	30.79% <ø> (ø)
datdog-sidecar-ffi	9.37% <ø> (ø)
spawn-worker	54.69% <ø> (ø)
libdd-tinybytes	93.16% <ø> (ø)
libdd-trace-normalization	81.71% <ø> (ø)
libdd-trace-obfuscation	87.37% <ø> (ø)
libdd-trace-protobuf	68.25% <ø> (ø)
libdd-trace-utils	88.64% <ø> (ø)
datadog-tracer-flare	86.88% <ø> (ø)
libdd-log	74.69% <ø> (ø)

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.
• 📦 JS Bundle Analysis: Save yourself from yourself by tracking and limiting bundle sizes in JS merges.

[pr-commenter]

Benchmarks

Comparison

Benchmark execution time: 2026-03-27 16:36:02

Comparing candidate commit ce8e143 in PR branch yannham/thread-ctx-tls-shim with baseline commit a29b90b in branch main.

Found 0 performance improvements and 2 performance regressions! Performance is the same for 60 metrics, 0 unstable metrics.

Explanation

This is an A/B test comparing a candidate commit's performance against that of a baseline commit. Performance changes are noted in the tables below as:

• 🟩 = significantly better candidate vs. baseline
• 🟥 = significantly worse candidate vs. baseline

We compute a confidence interval (CI) over the relative difference of means between metrics from the candidate and baseline commits, considering the baseline as the reference.

If the CI is entirely outside the configured SIGNIFICANT_IMPACT_THRESHOLD (or the deprecated UNCONFIDENCE_THRESHOLD), the change is considered significant.

Feel free to reach out to #apm-benchmarking-platform on Slack if you have any questions.

More details about the CI and significant changes

You can imagine this CI as a range of values that is likely to contain the true difference of means between the candidate and baseline commits.

CIs of the difference of means are often centered around 0%, because often changes are not that big:

---------------------------------(------|---^--------)-------------------------------->
                              -0.6%    0%  0.3%     +1.2%
                                 |          |        |
         lower bound of the CI --'          |        |
sample mean (center of the CI) -------------'        |
         upper bound of the CI ----------------------'

As described above, a change is considered significant if the CI is entirely outside the configured SIGNIFICANT_IMPACT_THRESHOLD (or the deprecated UNCONFIDENCE_THRESHOLD).

For instance, for an execution time metric, this confidence interval indicates a significantly worse performance:

----------------------------------------|---------|---(---------^---------)---------->
                                       0%        1%  1.3%      2.2%      3.1%
                                                  |   |         |         |
       significant impact threshold --------------'   |         |         |
                      lower bound of CI --------------'         |         |
       sample mean (center of the CI) --------------------------'         |
                      upper bound of CI ----------------------------------'

scenario:receiver_entry_point/report/2598

• 🟥 execution_time [+260.175µs; +271.632µs] or [+7.523%; +7.854%]

scenario:tags/replace_trace_tags

• 🟥 execution_time [+96.550ns; +103.368ns] or [+4.040%; +4.325%]

Candidate

Candidate benchmark details

Group 1

cpu_model	git_commit_sha	git_commit_date	git_branch
Intel(R) Xeon(R) Platinum 8259CL CPU @ 2.50GHz	ce8e143	1774628300	yannham/thread-ctx-tls-shim

scenario	metric	min	mean ± sd	median ± mad	p75	p95	p99	max	peak_to_median_ratio	skewness	kurtosis	cv	sem	runs	sample_size
tags/replace_trace_tags	execution_time	2.457µs	2.490µs ± 0.018µs	2.486µs ± 0.010µs	2.499µs	2.532µs	2.535µs	2.536µs	2.02%	0.900	0.202	0.73%	0.001µs	1	200

scenario	metric	95% CI mean	Shapiro-Wilk pvalue	Ljung-Box pvalue (lag=1)	Dip test pvalue
tags/replace_trace_tags	execution_time	[2.487µs; 2.492µs] or [-0.102%; +0.102%]	None	None	None

Group 2

cpu_model	git_commit_sha	git_commit_date	git_branch
Intel(R) Xeon(R) Platinum 8259CL CPU @ 2.50GHz	ce8e143	1774628300	yannham/thread-ctx-tls-shim

scenario	metric	min	mean ± sd	median ± mad	p75	p95	p99	max	peak_to_median_ratio	skewness	kurtosis	cv	sem	runs	sample_size
two way interface	execution_time	13.892µs	14.324µs ± 0.389µs	14.078µs ± 0.090µs	14.651µs	15.035µs	15.539µs	15.662µs	11.25%	1.064	0.423	2.71%	0.028µs	1	200

scenario	metric	95% CI mean	Shapiro-Wilk pvalue	Ljung-Box pvalue (lag=1)	Dip test pvalue
two way interface	execution_time	[14.270µs; 14.378µs] or [-0.376%; +0.376%]	None	None	None

Group 3

cpu_model	git_commit_sha	git_commit_date	git_branch
Intel(R) Xeon(R) Platinum 8259CL CPU @ 2.50GHz	ce8e143	1774628300	yannham/thread-ctx-tls-shim

scenario	metric	min	mean ± sd	median ± mad	p75	p95	p99	max	peak_to_median_ratio	skewness	kurtosis	cv	sem	runs	sample_size
profile_add_sample_frames_x1000	execution_time	4.170ms	4.174ms ± 0.003ms	4.173ms ± 0.001ms	4.174ms	4.178ms	4.181ms	4.198ms	0.60%	3.963	28.178	0.07%	0.000ms	1	200

scenario	metric	95% CI mean	Shapiro-Wilk pvalue	Ljung-Box pvalue (lag=1)	Dip test pvalue
profile_add_sample_frames_x1000	execution_time	[4.173ms; 4.174ms] or [-0.009%; +0.009%]	None	None	None

Group 4

cpu_model	git_commit_sha	git_commit_date	git_branch
Intel(R) Xeon(R) Platinum 8259CL CPU @ 2.50GHz	ce8e143	1774628300	yannham/thread-ctx-tls-shim

scenario	metric	min	mean ± sd	median ± mad	p75	p95	p99	max	peak_to_median_ratio	skewness	kurtosis	cv	sem	runs	sample_size
sdk_test_data/rules-based	execution_time	144.724µs	146.500µs ± 1.855µs	146.223µs ± 0.479µs	146.693µs	148.158µs	153.849µs	165.115µs	12.92%	6.468	55.230	1.26%	0.131µs	1	200

scenario	metric	95% CI mean	Shapiro-Wilk pvalue	Ljung-Box pvalue (lag=1)	Dip test pvalue
sdk_test_data/rules-based	execution_time	[146.243µs; 146.757µs] or [-0.175%; +0.175%]	None	None	None

Group 5

cpu_model	git_commit_sha	git_commit_date	git_branch
Intel(R) Xeon(R) Platinum 8259CL CPU @ 2.50GHz	ce8e143	1774628300	yannham/thread-ctx-tls-shim

scenario	metric	min	mean ± sd	median ± mad	p75	p95	p99	max	peak_to_median_ratio	skewness	kurtosis	cv	sem	runs	sample_size
normalization/normalize_name/normalize_name/Too-Long-.Too-Long-.Too-Long-.Too-Long-.Too-Long-.Too-Lo...	execution_time	185.634µs	186.180µs ± 0.332µs	186.138µs ± 0.178µs	186.332µs	186.612µs	187.777µs	187.959µs	0.98%	2.065	8.146	0.18%	0.023µs	1	200
normalization/normalize_name/normalize_name/Too-Long-.Too-Long-.Too-Long-.Too-Long-.Too-Long-.Too-Lo...	throughput	5320310.632op/s	5371175.791op/s ± 9544.837op/s	5372347.320op/s ± 5149.375op/s	5377178.672op/s	5382725.256op/s	5385552.208op/s	5386950.006op/s	0.27%	-2.034	7.955	0.18%	674.922op/s	1	200
normalization/normalize_name/normalize_name/bad-name	execution_time	17.826µs	17.905µs ± 0.028µs	17.904µs ± 0.018µs	17.923µs	17.955µs	17.967µs	17.983µs	0.44%	0.135	0.073	0.16%	0.002µs	1	200
normalization/normalize_name/normalize_name/bad-name	throughput	55608230.205op/s	55850150.642op/s ± 88674.839op/s	55854056.655op/s ± 54803.845op/s	55905696.956op/s	55990145.735op/s	56070701.630op/s	56097517.284op/s	0.44%	-0.126	0.073	0.16%	6270.258op/s	1	200
normalization/normalize_name/normalize_name/good	execution_time	10.541µs	10.587µs ± 0.026µs	10.582µs ± 0.015µs	10.603µs	10.626µs	10.643µs	10.784µs	1.91%	2.408	14.597	0.25%	0.002µs	1	200
normalization/normalize_name/normalize_name/good	throughput	92732771.551op/s	94453964.274op/s ± 231031.474op/s	94499790.067op/s ± 134660.123op/s	94617212.501op/s	94722360.694op/s	94759155.430op/s	94867591.434op/s	0.39%	-2.330	13.834	0.24%	16336.392op/s	1	200

scenario	metric	95% CI mean	Shapiro-Wilk pvalue	Ljung-Box pvalue (lag=1)	Dip test pvalue
normalization/normalize_name/normalize_name/Too-Long-.Too-Long-.Too-Long-.Too-Long-.Too-Long-.Too-Lo...	execution_time	[186.134µs; 186.226µs] or [-0.025%; +0.025%]	None	None	None
normalization/normalize_name/normalize_name/Too-Long-.Too-Long-.Too-Long-.Too-Long-.Too-Long-.Too-Lo...	throughput	[5369852.969op/s; 5372498.614op/s] or [-0.025%; +0.025%]	None	None	None
normalization/normalize_name/normalize_name/bad-name	execution_time	[17.901µs; 17.909µs] or [-0.022%; +0.022%]	None	None	None
normalization/normalize_name/normalize_name/bad-name	throughput	[55837861.162op/s; 55862440.122op/s] or [-0.022%; +0.022%]	None	None	None
normalization/normalize_name/normalize_name/good	execution_time	[10.584µs; 10.591µs] or [-0.034%; +0.034%]	None	None	None
normalization/normalize_name/normalize_name/good	throughput	[94421945.534op/s; 94485983.015op/s] or [-0.034%; +0.034%]	None	None	None

Group 6

cpu_model	git_commit_sha	git_commit_date	git_branch
Intel(R) Xeon(R) Platinum 8259CL CPU @ 2.50GHz	ce8e143	1774628300	yannham/thread-ctx-tls-shim

scenario	metric	min	mean ± sd	median ± mad	p75	p95	p99	max	peak_to_median_ratio	skewness	kurtosis	cv	sem	runs	sample_size
sql/obfuscate_sql_string	execution_time	288.948µs	289.514µs ± 0.479µs	289.458µs ± 0.157µs	289.594µs	289.829µs	291.467µs	293.769µs	1.49%	5.857	44.277	0.17%	0.034µs	1	200

scenario	metric	95% CI mean	Shapiro-Wilk pvalue	Ljung-Box pvalue (lag=1)	Dip test pvalue
sql/obfuscate_sql_string	execution_time	[289.448µs; 289.581µs] or [-0.023%; +0.023%]	None	None	None

Group 7

cpu_model	git_commit_sha	git_commit_date	git_branch
Intel(R) Xeon(R) Platinum 8259CL CPU @ 2.50GHz	ce8e143	1774628300	yannham/thread-ctx-tls-shim

scenario	metric	min	mean ± sd	median ± mad	p75	p95	p99	max	peak_to_median_ratio	skewness	kurtosis	cv	sem	runs	sample_size
benching deserializing traces from msgpack to their internal representation	execution_time	49.805ms	50.142ms ± 0.901ms	50.002ms ± 0.076ms	50.109ms	50.233ms	55.375ms	59.155ms	18.30%	7.901	66.243	1.79%	0.064ms	1	200

scenario	metric	95% CI mean	Shapiro-Wilk pvalue	Ljung-Box pvalue (lag=1)	Dip test pvalue
benching deserializing traces from msgpack to their internal representation	execution_time	[50.017ms; 50.267ms] or [-0.249%; +0.249%]	None	None	None

Group 8

cpu_model	git_commit_sha	git_commit_date	git_branch
Intel(R) Xeon(R) Platinum 8259CL CPU @ 2.50GHz	ce8e143	1774628300	yannham/thread-ctx-tls-shim

scenario	metric	min	mean ± sd	median ± mad	p75	p95	p99	max	peak_to_median_ratio	skewness	kurtosis	cv	sem	runs	sample_size
profile_serialize_compressed_pprof_timestamped_x1000	execution_time	911.572µs	914.303µs ± 1.310µs	914.276µs ± 0.917µs	915.154µs	916.601µs	917.439µs	918.241µs	0.43%	0.365	-0.088	0.14%	0.093µs	1	200

scenario	metric	95% CI mean	Shapiro-Wilk pvalue	Ljung-Box pvalue (lag=1)	Dip test pvalue
profile_serialize_compressed_pprof_timestamped_x1000	execution_time	[914.121µs; 914.484µs] or [-0.020%; +0.020%]	None	None	None

Group 9

cpu_model	git_commit_sha	git_commit_date	git_branch
Intel(R) Xeon(R) Platinum 8259CL CPU @ 2.50GHz	ce8e143	1774628300	yannham/thread-ctx-tls-shim

scenario	metric	min	mean ± sd	median ± mad	p75	p95	p99	max	peak_to_median_ratio	skewness	kurtosis	cv	sem	runs	sample_size
profile_add_sample2_frames_x1000	execution_time	742.216µs	743.110µs ± 0.463µs	743.099µs ± 0.283µs	743.343µs	744.053µs	744.266µs	745.268µs	0.29%	0.923	2.165	0.06%	0.033µs	1	200

scenario	metric	95% CI mean	Shapiro-Wilk pvalue	Ljung-Box pvalue (lag=1)	Dip test pvalue
profile_add_sample2_frames_x1000	execution_time	[743.046µs; 743.174µs] or [-0.009%; +0.009%]	None	None	None

Group 10

cpu_model	git_commit_sha	git_commit_date	git_branch
Intel(R) Xeon(R) Platinum 8259CL CPU @ 2.50GHz	ce8e143	1774628300	yannham/thread-ctx-tls-shim

scenario	metric	min	mean ± sd	median ± mad	p75	p95	p99	max	peak_to_median_ratio	skewness	kurtosis	cv	sem	runs	sample_size
ip_address/quantize_peer_ip_address_benchmark	execution_time	5.049µs	5.105µs ± 0.047µs	5.082µs ± 0.017µs	5.149µs	5.192µs	5.194µs	5.198µs	2.28%	0.738	-1.069	0.92%	0.003µs	1	200

scenario	metric	95% CI mean	Shapiro-Wilk pvalue	Ljung-Box pvalue (lag=1)	Dip test pvalue
ip_address/quantize_peer_ip_address_benchmark	execution_time	[5.099µs; 5.112µs] or [-0.128%; +0.128%]	None	None	None

Group 11

cpu_model	git_commit_sha	git_commit_date	git_branch
Intel(R) Xeon(R) Platinum 8259CL CPU @ 2.50GHz	ce8e143	1774628300	yannham/thread-ctx-tls-shim

scenario	metric	min	mean ± sd	median ± mad	p75	p95	p99	max	peak_to_median_ratio	skewness	kurtosis	cv	sem	runs	sample_size
normalization/normalize_trace/test_trace	execution_time	242.329ns	254.525ns ± 14.115ns	248.872ns ± 4.346ns	257.673ns	286.725ns	295.341ns	296.056ns	18.96%	1.644	1.499	5.53%	0.998ns	1	200

scenario	metric	95% CI mean	Shapiro-Wilk pvalue	Ljung-Box pvalue (lag=1)	Dip test pvalue
normalization/normalize_trace/test_trace	execution_time	[252.569ns; 256.481ns] or [-0.769%; +0.769%]	None	None	None

Group 12

cpu_model	git_commit_sha	git_commit_date	git_branch
Intel(R) Xeon(R) Platinum 8259CL CPU @ 2.50GHz	ce8e143	1774628300	yannham/thread-ctx-tls-shim

scenario	metric	min	mean ± sd	median ± mad	p75	p95	p99	max	peak_to_median_ratio	skewness	kurtosis	cv	sem	runs	sample_size
single_flag_killswitch/rules-based	execution_time	190.341ns	192.504ns ± 1.677ns	192.448ns ± 1.325ns	193.413ns	195.366ns	197.761ns	198.081ns	2.93%	0.805	0.578	0.87%	0.119ns	1	200

scenario	metric	95% CI mean	Shapiro-Wilk pvalue	Ljung-Box pvalue (lag=1)	Dip test pvalue
single_flag_killswitch/rules-based	execution_time	[192.271ns; 192.736ns] or [-0.121%; +0.121%]	None	None	None

Group 13

cpu_model	git_commit_sha	git_commit_date	git_branch
Intel(R) Xeon(R) Platinum 8259CL CPU @ 2.50GHz	ce8e143	1774628300	yannham/thread-ctx-tls-shim

scenario	metric	min	mean ± sd	median ± mad	p75	p95	p99	max	peak_to_median_ratio	skewness	kurtosis	cv	sem	runs	sample_size
redis/obfuscate_redis_string	execution_time	33.657µs	34.338µs ± 1.242µs	33.776µs ± 0.054µs	33.841µs	37.022µs	37.055µs	37.496µs	11.01%	1.698	0.913	3.61%	0.088µs	1	200

scenario	metric	95% CI mean	Shapiro-Wilk pvalue	Ljung-Box pvalue (lag=1)	Dip test pvalue
redis/obfuscate_redis_string	execution_time	[34.166µs; 34.511µs] or [-0.501%; +0.501%]	None	None	None

Group 14

cpu_model	git_commit_sha	git_commit_date	git_branch
Intel(R) Xeon(R) Platinum 8259CL CPU @ 2.50GHz	ce8e143	1774628300	yannham/thread-ctx-tls-shim

scenario	metric	min	mean ± sd	median ± mad	p75	p95	p99	max	peak_to_median_ratio	skewness	kurtosis	cv	sem	runs	sample_size
write only interface	execution_time	5.397µs	5.461µs ± 0.028µs	5.467µs ± 0.020µs	5.482µs	5.498µs	5.511µs	5.520µs	0.98%	-0.362	-0.777	0.50%	0.002µs	1	200

scenario	metric	95% CI mean	Shapiro-Wilk pvalue	Ljung-Box pvalue (lag=1)	Dip test pvalue
write only interface	execution_time	[5.457µs; 5.464µs] or [-0.070%; +0.070%]	None	None	None

Group 15

cpu_model	git_commit_sha	git_commit_date	git_branch
Intel(R) Xeon(R) Platinum 8259CL CPU @ 2.50GHz	ce8e143	1774628300	yannham/thread-ctx-tls-shim

scenario	metric	min	mean ± sd	median ± mad	p75	p95	p99	max	peak_to_median_ratio	skewness	kurtosis	cv	sem	runs	sample_size
benching serializing traces from their internal representation to msgpack	execution_time	13.767ms	13.814ms ± 0.029ms	13.809ms ± 0.015ms	13.827ms	13.866ms	13.914ms	13.956ms	1.06%	1.640	4.474	0.21%	0.002ms	1	200

scenario	metric	95% CI mean	Shapiro-Wilk pvalue	Ljung-Box pvalue (lag=1)	Dip test pvalue
benching serializing traces from their internal representation to msgpack	execution_time	[13.810ms; 13.818ms] or [-0.029%; +0.029%]	None	None	None

Group 16

cpu_model	git_commit_sha	git_commit_date	git_branch
Intel(R) Xeon(R) Platinum 8259CL CPU @ 2.50GHz	ce8e143	1774628300	yannham/thread-ctx-tls-shim

scenario	metric	min	mean ± sd	median ± mad	p75	p95	p99	max	peak_to_median_ratio	skewness	kurtosis	cv	sem	runs	sample_size
profile_add_sample_timestamped_x1000	execution_time	4.183ms	4.187ms ± 0.008ms	4.186ms ± 0.001ms	4.187ms	4.190ms	4.196ms	4.296ms	2.62%	11.959	155.352	0.19%	0.001ms	1	200

scenario	metric	95% CI mean	Shapiro-Wilk pvalue	Ljung-Box pvalue (lag=1)	Dip test pvalue
profile_add_sample_timestamped_x1000	execution_time	[4.186ms; 4.188ms] or [-0.027%; +0.027%]	None	None	None

Group 17

cpu_model	git_commit_sha	git_commit_date	git_branch
Intel(R) Xeon(R) Platinum 8259CL CPU @ 2.50GHz	ce8e143	1774628300	yannham/thread-ctx-tls-shim

scenario	metric	min	mean ± sd	median ± mad	p75	p95	p99	max	peak_to_median_ratio	skewness	kurtosis	cv	sem	runs	sample_size
benching string interning on wordpress profile	execution_time	160.478µs	161.521µs ± 0.241µs	161.485µs ± 0.115µs	161.608µs	161.950µs	162.293µs	162.396µs	0.56%	0.544	2.734	0.15%	0.017µs	1	200

scenario	metric	95% CI mean	Shapiro-Wilk pvalue	Ljung-Box pvalue (lag=1)	Dip test pvalue
benching string interning on wordpress profile	execution_time	[161.488µs; 161.555µs] or [-0.021%; +0.021%]	None	None	None

Group 18

cpu_model	git_commit_sha	git_commit_date	git_branch
Intel(R) Xeon(R) Platinum 8259CL CPU @ 2.50GHz	ce8e143	1774628300	yannham/thread-ctx-tls-shim

scenario	metric	min	mean ± sd	median ± mad	p75	p95	p99	max	peak_to_median_ratio	skewness	kurtosis	cv	sem	runs	sample_size
normalization/normalize_service/normalize_service/A0000000000000000000000000000000000000000000000000...	execution_time	495.452µs	496.384µs ± 0.645µs	496.262µs ± 0.349µs	496.638µs	497.446µs	498.879µs	499.641µs	0.68%	1.957	5.513	0.13%	0.046µs	1	200
normalization/normalize_service/normalize_service/A0000000000000000000000000000000000000000000000000...	throughput	2001435.697op/s	2014570.884op/s ± 2609.172op/s	2015064.962op/s ± 1417.204op/s	2016374.134op/s	2017442.822op/s	2018019.714op/s	2018358.541op/s	0.16%	-1.943	5.437	0.13%	184.496op/s	1	200
normalization/normalize_service/normalize_service/Data🐨dog🐶 繋がっ⛰てて	execution_time	370.786µs	371.580µs ± 0.378µs	371.539µs ± 0.250µs	371.818µs	372.219µs	372.557µs	373.478µs	0.52%	0.837	2.232	0.10%	0.027µs	1	200
normalization/normalize_service/normalize_service/Data🐨dog🐶 繋がっ⛰てて	throughput	2677533.884op/s	2691212.747op/s ± 2737.222op/s	2691507.644op/s ± 1808.505op/s	2693193.057op/s	2695157.734op/s	2695790.913op/s	2696973.456op/s	0.20%	-0.827	2.182	0.10%	193.551op/s	1	200
normalization/normalize_service/normalize_service/Test Conversion 0f Weird !@#$%^&**() Characters	execution_time	167.859µs	168.250µs ± 0.283µs	168.178µs ± 0.094µs	168.299µs	168.978µs	169.387µs	169.478µs	0.77%	2.484	6.506	0.17%	0.020µs	1	200
normalization/normalize_service/normalize_service/Test Conversion 0f Weird !@#$%^&**() Characters	throughput	5900470.048op/s	5943546.994op/s ± 9967.730op/s	5946076.827op/s ± 3324.289op/s	5948829.689op/s	5952750.364op/s	5954290.203op/s	5957392.452op/s	0.19%	-2.472	6.450	0.17%	704.825op/s	1	200
normalization/normalize_service/normalize_service/[empty string]	execution_time	36.821µs	37.024µs ± 0.108µs	37.025µs ± 0.090µs	37.110µs	37.196µs	37.238µs	37.241µs	0.59%	-0.016	-0.984	0.29%	0.008µs	1	200
normalization/normalize_service/normalize_service/[empty string]	throughput	26851831.753op/s	27009674.026op/s ± 78671.729op/s	27009107.923op/s ± 65760.423op/s	27076324.439op/s	27139319.893op/s	27154629.707op/s	27158777.205op/s	0.55%	0.025	-0.984	0.29%	5562.931op/s	1	200
normalization/normalize_service/normalize_service/test_ASCII	execution_time	46.268µs	46.374µs ± 0.048µs	46.368µs ± 0.027µs	46.399µs	46.467µs	46.498µs	46.557µs	0.41%	0.650	0.777	0.10%	0.003µs	1	200
normalization/normalize_service/normalize_service/test_ASCII	throughput	21479186.091op/s	21563989.791op/s ± 22468.519op/s	21566690.178op/s ± 12773.367op/s	21577776.283op/s	21595327.414op/s	21608055.460op/s	21613295.384op/s	0.22%	-0.643	0.763	0.10%	1588.764op/s	1	200

scenario	metric	95% CI mean	Shapiro-Wilk pvalue	Ljung-Box pvalue (lag=1)	Dip test pvalue
normalization/normalize_service/normalize_service/A0000000000000000000000000000000000000000000000000...	execution_time	[496.295µs; 496.474µs] or [-0.018%; +0.018%]	None	None	None
normalization/normalize_service/normalize_service/A0000000000000000000000000000000000000000000000000...	throughput	[2014209.278op/s; 2014932.490op/s] or [-0.018%; +0.018%]	None	None	None
normalization/normalize_service/normalize_service/Data🐨dog🐶 繋がっ⛰てて	execution_time	[371.528µs; 371.632µs] or [-0.014%; +0.014%]	None	None	None
normalization/normalize_service/normalize_service/Data🐨dog🐶 繋がっ⛰てて	throughput	[2690833.394op/s; 2691592.099op/s] or [-0.014%; +0.014%]	None	None	None
normalization/normalize_service/normalize_service/Test Conversion 0f Weird !@#$%^&**() Characters	execution_time	[168.211µs; 168.289µs] or [-0.023%; +0.023%]	None	None	None
normalization/normalize_service/normalize_service/Test Conversion 0f Weird !@#$%^&**() Characters	throughput	[5942165.563op/s; 5944928.426op/s] or [-0.023%; +0.023%]	None	None	None
normalization/normalize_service/normalize_service/[empty string]	execution_time	[37.009µs; 37.039µs] or [-0.040%; +0.040%]	None	None	None
normalization/normalize_service/normalize_service/[empty string]	throughput	[26998770.881op/s; 27020577.171op/s] or [-0.040%; +0.040%]	None	None	None
normalization/normalize_service/normalize_service/test_ASCII	execution_time	[46.367µs; 46.380µs] or [-0.014%; +0.014%]	None	None	None
normalization/normalize_service/normalize_service/test_ASCII	throughput	[21560875.870op/s; 21567103.711op/s] or [-0.014%; +0.014%]	None	None	None

Group 19

cpu_model	git_commit_sha	git_commit_date	git_branch
Intel(R) Xeon(R) Platinum 8259CL CPU @ 2.50GHz	ce8e143	1774628300	yannham/thread-ctx-tls-shim

scenario	metric	min	mean ± sd	median ± mad	p75	p95	p99	max	peak_to_median_ratio	skewness	kurtosis	cv	sem	runs	sample_size
concentrator/add_spans_to_concentrator	execution_time	14.793ms	14.824ms ± 0.016ms	14.821ms ± 0.009ms	14.832ms	14.852ms	14.870ms	14.906ms	0.58%	1.577	4.743	0.11%	0.001ms	1	200

scenario	metric	95% CI mean	Shapiro-Wilk pvalue	Ljung-Box pvalue (lag=1)	Dip test pvalue
concentrator/add_spans_to_concentrator	execution_time	[14.822ms; 14.827ms] or [-0.015%; +0.015%]	None	None	None

Group 20

cpu_model	git_commit_sha	git_commit_date	git_branch
Intel(R) Xeon(R) Platinum 8259CL CPU @ 2.50GHz	ce8e143	1774628300	yannham/thread-ctx-tls-shim

scenario	metric	min	mean ± sd	median ± mad	p75	p95	p99	max	peak_to_median_ratio	skewness	kurtosis	cv	sem	runs	sample_size
receiver_entry_point/report/2598	execution_time	3.682ms	3.724ms ± 0.033ms	3.715ms ± 0.011ms	3.727ms	3.799ms	3.855ms	3.865ms	4.05%	2.234	5.228	0.89%	0.002ms	1	200

scenario	metric	95% CI mean	Shapiro-Wilk pvalue	Ljung-Box pvalue (lag=1)	Dip test pvalue
receiver_entry_point/report/2598	execution_time	[3.720ms; 3.729ms] or [-0.124%; +0.124%]	None	None	None

Group 21

cpu_model	git_commit_sha	git_commit_date	git_branch
Intel(R) Xeon(R) Platinum 8259CL CPU @ 2.50GHz	ce8e143	1774628300	yannham/thread-ctx-tls-shim

scenario	metric	min	mean ± sd	median ± mad	p75	p95	p99	max	peak_to_median_ratio	skewness	kurtosis	cv	sem	runs	sample_size
credit_card/is_card_number/	execution_time	3.894µs	3.912µs ± 0.003µs	3.913µs ± 0.001µs	3.914µs	3.917µs	3.918µs	3.921µs	0.20%	-1.563	12.781	0.07%	0.000µs	1	200
credit_card/is_card_number/	throughput	255068094.303op/s	255594876.365op/s ± 169814.663op/s	255589546.979op/s ± 95600.394op/s	255691721.981op/s	255815248.608op/s	255862585.104op/s	256831213.078op/s	0.49%	1.587	12.976	0.07%	12007.710op/s	1	200
credit_card/is_card_number/ 3782-8224-6310-005	execution_time	78.904µs	79.609µs ± 0.356µs	79.585µs ± 0.223µs	79.789µs	80.257µs	80.590µs	80.883µs	1.63%	0.771	0.860	0.45%	0.025µs	1	200
credit_card/is_card_number/ 3782-8224-6310-005	throughput	12363602.966op/s	12561670.694op/s ± 56020.255op/s	12565217.608op/s ± 35369.580op/s	12603223.907op/s	12641114.994op/s	12661156.208op/s	12673650.384op/s	0.86%	-0.741	0.784	0.44%	3961.230op/s	1	200
credit_card/is_card_number/ 378282246310005	execution_time	72.346µs	73.014µs ± 0.354µs	72.988µs ± 0.236µs	73.217µs	73.700µs	73.833µs	73.969µs	1.35%	0.428	-0.327	0.48%	0.025µs	1	200
credit_card/is_card_number/ 378282246310005	throughput	13519102.245op/s	13696355.755op/s ± 66271.355op/s	13700952.911op/s ± 44504.692op/s	13745422.436op/s	13792730.124op/s	13813393.211op/s	13822493.556op/s	0.89%	-0.406	-0.352	0.48%	4686.092op/s	1	200
credit_card/is_card_number/37828224631	execution_time	3.894µs	3.912µs ± 0.003µs	3.912µs ± 0.001µs	3.914µs	3.917µs	3.918µs	3.921µs	0.23%	-1.104	11.247	0.07%	0.000µs	1	200
credit_card/is_card_number/37828224631	throughput	255023341.221op/s	255596902.368op/s ± 174851.908op/s	255616576.739op/s ± 96680.444op/s	255696186.526op/s	255794768.532op/s	255857332.352op/s	256825545.822op/s	0.47%	1.129	11.424	0.07%	12363.897op/s	1	200
credit_card/is_card_number/378282246310005	execution_time	69.122µs	69.610µs ± 0.259µs	69.543µs ± 0.154µs	69.757µs	70.100µs	70.244µs	70.770µs	1.76%	1.107	1.612	0.37%	0.018µs	1	200
credit_card/is_card_number/378282246310005	throughput	14130261.781op/s	14365949.547op/s ± 53160.545op/s	14379543.661op/s ± 31999.280op/s	14403283.949op/s	14431382.849op/s	14440103.659op/s	14467204.505op/s	0.61%	-1.081	1.505	0.37%	3759.018op/s	1	200
credit_card/is_card_number/37828224631000521389798	execution_time	52.234µs	52.312µs ± 0.033µs	52.310µs ± 0.023µs	52.331µs	52.377µs	52.402µs	52.457µs	0.28%	0.781	1.457	0.06%	0.002µs	1	200
credit_card/is_card_number/37828224631000521389798	throughput	19063194.888op/s	19116163.316op/s ± 12070.403op/s	19116722.778op/s ± 8306.153op/s	19125247.059op/s	19133181.859op/s	19137483.533op/s	19144488.641op/s	0.15%	-0.776	1.440	0.06%	853.506op/s	1	200
credit_card/is_card_number/x371413321323331	execution_time	5.709µs	5.858µs ± 0.064µs	5.864µs ± 0.042µs	5.902µs	5.963µs	6.003µs	6.038µs	2.98%	0.080	-0.522	1.10%	0.005µs	1	200
credit_card/is_card_number/x371413321323331	throughput	165611525.634op/s	170727061.167op/s ± 1875869.043op/s	170542747.567op/s ± 1217098.997op/s	172283412.143op/s	173689764.678op/s	174314252.645op/s	175148574.661op/s	2.70%	-0.032	-0.547	1.10%	132643.972op/s	1	200
credit_card/is_card_number_no_luhn/	execution_time	3.892µs	3.913µs ± 0.003µs	3.913µs ± 0.002µs	3.915µs	3.918µs	3.919µs	3.922µs	0.22%	-1.638	13.576	0.07%	0.000µs	1	200
credit_card/is_card_number_no_luhn/	throughput	254974597.271op/s	255548670.867op/s ± 189114.832op/s	255537005.722op/s ± 123315.796op/s	255681584.535op/s	255774008.211op/s	255827966.940op/s	256949029.231op/s	0.55%	1.666	13.812	0.07%	13372.438op/s	1	200
credit_card/is_card_number_no_luhn/ 3782-8224-6310-005	execution_time	63.849µs	64.355µs ± 0.181µs	64.351µs ± 0.119µs	64.476µs	64.640µs	64.772µs	64.952µs	0.93%	0.044	0.118	0.28%	0.013µs	1	200
credit_card/is_card_number_no_luhn/ 3782-8224-6310-005	throughput	15395899.638op/s	15538935.316op/s ± 43765.667op/s	15539820.494op/s ± 28615.164op/s	15566661.272op/s	15615568.566op/s	15629192.722op/s	15662060.916op/s	0.79%	-0.026	0.107	0.28%	3094.700op/s	1	200
credit_card/is_card_number_no_luhn/ 378282246310005	execution_time	57.821µs	58.087µs ± 0.186µs	58.031µs ± 0.109µs	58.190µs	58.433µs	58.679µs	58.766µs	1.27%	1.169	1.178	0.32%	0.013µs	1	200
credit_card/is_card_number_no_luhn/ 378282246310005	throughput	17016769.485op/s	17215819.719op/s ± 55000.148op/s	17232167.298op/s ± 32291.941op/s	17255082.002op/s	17277262.879op/s	17293564.719op/s	17294847.865op/s	0.36%	-1.152	1.121	0.32%	3889.098op/s	1	200
credit_card/is_card_number_no_luhn/37828224631	execution_time	3.892µs	3.912µs ± 0.003µs	3.911µs ± 0.001µs	3.913µs	3.917µs	3.920µs	3.924µs	0.33%	-0.805	14.288	0.07%	0.000µs	1	200
credit_card/is_card_number_no_luhn/37828224631	throughput	254832296.032op/s	255636331.748op/s ± 184496.028op/s	255668309.913op/s ± 82226.393op/s	255741647.530op/s	255817436.841op/s	255858509.976op/s	256969480.628op/s	0.51%	0.839	14.494	0.07%	13045.839op/s	1	200
credit_card/is_card_number_no_luhn/378282246310005	execution_time	54.594µs	55.031µs ± 0.314µs	54.970µs ± 0.211µs	55.217µs	55.632µs	55.983µs	56.180µs	2.20%	1.021	0.928	0.57%	0.022µs	1	200
credit_card/is_card_number_no_luhn/378282246310005	throughput	17799966.890op/s	18172200.658op/s ± 103074.143op/s	18191696.664op/s ± 70019.476op/s	18252408.671op/s	18295727.967op/s	18310347.803op/s	18317056.843op/s	0.69%	-0.989	0.829	0.57%	7288.443op/s	1	200
credit_card/is_card_number_no_luhn/37828224631000521389798	execution_time	52.216µs	52.320µs ± 0.037µs	52.322µs ± 0.025µs	52.346µs	52.372µs	52.393µs	52.441µs	0.23%	-0.102	-0.134	0.07%	0.003µs	1	200
credit_card/is_card_number_no_luhn/37828224631000521389798	throughput	19069164.555op/s	19113011.804op/s ± 13371.691op/s	19112293.684op/s ± 9116.912op/s	19121945.652op/s	19135496.349op/s	19141896.764op/s	19151393.763op/s	0.20%	0.106	-0.136	0.07%	945.521op/s	1	200
credit_card/is_card_number_no_luhn/x371413321323331	execution_time	5.704µs	5.848µs ± 0.063µs	5.852µs ± 0.048µs	5.896µs	5.940µs	5.981µs	6.008µs	2.67%	-0.003	-0.641	1.08%	0.004µs	1	200
credit_card/is_card_number_no_luhn/x371413321323331	throughput	166431075.746op/s	171016170.666op/s ± 1850209.584op/s	170868191.812op/s ± 1390944.919op/s	172380978.288op/s	174022598.218op/s	174422478.667op/s	175302186.236op/s	2.59%	0.046	-0.654	1.08%	130829.574op/s	1	200

scenario	metric	95% CI mean	Shapiro-Wilk pvalue	Ljung-Box pvalue (lag=1)	Dip test pvalue
credit_card/is_card_number/	execution_time	[3.912µs; 3.913µs] or [-0.009%; +0.009%]	None	None	None
credit_card/is_card_number/	throughput	[255571341.686op/s; 255618411.044op/s] or [-0.009%; +0.009%]	None	None	None
credit_card/is_card_number/ 3782-8224-6310-005	execution_time	[79.559µs; 79.658µs] or [-0.062%; +0.062%]	None	None	None
credit_card/is_card_number/ 3782-8224-6310-005	throughput	[12553906.826op/s; 12569434.563op/s] or [-0.062%; +0.062%]	None	None	None
credit_card/is_card_number/ 378282246310005	execution_time	[72.965µs; 73.063µs] or [-0.067%; +0.067%]	None	None	None
credit_card/is_card_number/ 378282246310005	throughput	[13687171.182op/s; 13705540.327op/s] or [-0.067%; +0.067%]	None	None	None
credit_card/is_card_number/37828224631	execution_time	[3.912µs; 3.913µs] or [-0.009%; +0.009%]	None	None	None
credit_card/is_card_number/37828224631	throughput	[255572669.575op/s; 255621135.161op/s] or [-0.009%; +0.009%]	None	None	None
credit_card/is_card_number/378282246310005	execution_time	[69.574µs; 69.646µs] or [-0.051%; +0.051%]	None	None	None
credit_card/is_card_number/378282246310005	throughput	[14358582.006op/s; 14373317.087op/s] or [-0.051%; +0.051%]	None	None	None
credit_card/is_card_number/37828224631000521389798	execution_time	[52.307µs; 52.316µs] or [-0.009%; +0.009%]	None	None	None
credit_card/is_card_number/37828224631000521389798	throughput	[19114490.474op/s; 19117836.158op/s] or [-0.009%; +0.009%]	None	None	None
credit_card/is_card_number/x371413321323331	execution_time	[5.849µs; 5.867µs] or [-0.152%; +0.152%]	None	None	None
credit_card/is_card_number/x371413321323331	throughput	[170467083.759op/s; 170987038.575op/s] or [-0.152%; +0.152%]	None	None	None
credit_card/is_card_number_no_luhn/	execution_time	[3.913µs; 3.914µs] or [-0.010%; +0.010%]	None	None	None
credit_card/is_card_number_no_luhn/	throughput	[255522461.370op/s; 255574880.364op/s] or [-0.010%; +0.010%]	None	None	None
credit_card/is_card_number_no_luhn/ 3782-8224-6310-005	execution_time	[64.330µs; 64.380µs] or [-0.039%; +0.039%]	None	None	None
credit_card/is_card_number_no_luhn/ 3782-8224-6310-005	throughput	[15532869.815op/s; 15545000.816op/s] or [-0.039%; +0.039%]	None	None	None
credit_card/is_card_number_no_luhn/ 378282246310005	execution_time	[58.061µs; 58.113µs] or [-0.044%; +0.044%]	None	None	None
credit_card/is_card_number_no_luhn/ 378282246310005	throughput	[17208197.227op/s; 17223442.210op/s] or [-0.044%; +0.044%]	None	None	None
credit_card/is_card_number_no_luhn/37828224631	execution_time	[3.911µs; 3.912µs] or [-0.010%; +0.010%]	None	None	None
credit_card/is_card_number_no_luhn/37828224631	throughput	[255610762.373op/s; 255661901.123op/s] or [-0.010%; +0.010%]	None	None	None
credit_card/is_card_number_no_luhn/378282246310005	execution_time	[54.987µs; 55.074µs] or [-0.079%; +0.079%]	None	None	None
credit_card/is_card_number_no_luhn/378282246310005	throughput	[18157915.573op/s; 18186485.743op/s] or [-0.079%; +0.079%]	None	None	None
credit_card/is_card_number_no_luhn/37828224631000521389798	execution_time	[52.315µs; 52.325µs] or [-0.010%; +0.010%]	None	None	None
credit_card/is_card_number_no_luhn/37828224631000521389798	throughput	[19111158.616op/s; 19114864.991op/s] or [-0.010%; +0.010%]	None	None	None
credit_card/is_card_number_no_luhn/x371413321323331	execution_time	[5.839µs; 5.857µs] or [-0.150%; +0.150%]	None	None	None
credit_card/is_card_number_no_luhn/x371413321323331	throughput	[170759749.412op/s; 171272591.920op/s] or [-0.150%; +0.150%]	None	None	None

Baseline

Omitted due to size.

[dd-octo-sts]

Artifact Size Benchmark Report

aarch64-alpine-linux-musl

Artifact	Baseline	Commit	Change
/aarch64-alpine-linux-musl/lib/libdatadog_profiling.so	8.76 MB	8.76 MB	0% (0 B) 👌
/aarch64-alpine-linux-musl/lib/libdatadog_profiling.a	101.90 MB	101.90 MB	+0% (+286 B) 👌

aarch64-unknown-linux-gnu

Artifact	Baseline	Commit	Change
/aarch64-unknown-linux-gnu/lib/libdatadog_profiling.a	118.84 MB	118.85 MB	+0% (+4.72 KB) 👌
/aarch64-unknown-linux-gnu/lib/libdatadog_profiling.so	11.36 MB	11.36 MB	0% (0 B) 👌

libdatadog-x64-windows

Artifact	Baseline	Commit	Change
/libdatadog-x64-windows/debug/dynamic/datadog_profiling_ffi.dll	27.39 MB	27.39 MB	0% (0 B) 👌
/libdatadog-x64-windows/debug/dynamic/datadog_profiling_ffi.lib	80.69 KB	80.69 KB	0% (0 B) 👌
/libdatadog-x64-windows/debug/dynamic/datadog_profiling_ffi.pdb	187.21 MB	187.19 MB	-0% (-16.00 KB) 👌
/libdatadog-x64-windows/debug/static/datadog_profiling_ffi.lib	924.91 MB	925.25 MB	+.03% (+349.11 KB) 🔍
/libdatadog-x64-windows/release/dynamic/datadog_profiling_ffi.dll	9.06 MB	9.06 MB	0% (0 B) 👌
/libdatadog-x64-windows/release/dynamic/datadog_profiling_ffi.lib	80.69 KB	80.69 KB	0% (0 B) 👌
/libdatadog-x64-windows/release/dynamic/datadog_profiling_ffi.pdb	26.98 MB	26.98 MB	0% (0 B) 👌
/libdatadog-x64-windows/release/static/datadog_profiling_ffi.lib	61.28 MB	61.28 MB	+0% (+174 B) 👌

libdatadog-x86-windows

Artifact	Baseline	Commit	Change
/libdatadog-x86-windows/debug/dynamic/datadog_profiling_ffi.dll	23.20 MB	23.20 MB	0% (0 B) 👌
/libdatadog-x86-windows/debug/dynamic/datadog_profiling_ffi.lib	81.94 KB	81.94 KB	0% (0 B) 👌
/libdatadog-x86-windows/debug/dynamic/datadog_profiling_ffi.pdb	191.48 MB	191.48 MB	0% (0 B) 👌
/libdatadog-x86-windows/debug/static/datadog_profiling_ffi.lib	908.20 MB	907.97 MB	--.02% (-236.40 KB) 💪
/libdatadog-x86-windows/release/dynamic/datadog_profiling_ffi.dll	6.90 MB	6.90 MB	-0% (-512 B) 👌
/libdatadog-x86-windows/release/dynamic/datadog_profiling_ffi.lib	81.94 KB	81.94 KB	0% (0 B) 👌
/libdatadog-x86-windows/release/dynamic/datadog_profiling_ffi.pdb	29.11 MB	29.11 MB	0% (0 B) 👌
/libdatadog-x86-windows/release/static/datadog_profiling_ffi.lib	57.68 MB	57.68 MB	-0% (-486 B) 👌

x86_64-alpine-linux-musl

Artifact	Baseline	Commit	Change
/x86_64-alpine-linux-musl/lib/libdatadog_profiling.a	88.72 MB	88.72 MB	+0% (+2.41 KB) 👌
/x86_64-alpine-linux-musl/lib/libdatadog_profiling.so	10.32 MB	10.32 MB	0% (0 B) 👌

x86_64-unknown-linux-gnu

Artifact	Baseline	Commit	Change
/x86_64-unknown-linux-gnu/lib/libdatadog_profiling.a	111.54 MB	111.54 MB	+0% (+5.18 KB) 👌
/x86_64-unknown-linux-gnu/lib/libdatadog_profiling.so	12.07 MB	12.07 MB	0% (0 B) 👌

[nsavoire]

Concerning cross language LTO, I did an experiment there that inlines a small C function into rust:
https://github.com/nsavoire/cross_language_lto_demo

[ivoanjo]

👍 Left some notes but really excited to see this starting to land!

Note: My rust-fu is kinda sucky so may be worth getting a second opinion on the rusty bits, I may have missed something obvious there ;)

[ivoanjo]

Concerning cross language LTO, I did an experiment there that inlines a small C function into rust:

That's very cool! Although I soft-wonder if it'll still respect the TLSDESC... 👀

[yannham]

That's very cool! Although I soft-wonder if it'll still respect the TLSDESC... 👀

I'm rather confident on that front. What I want and that @nsavoire managed to do is just to inline the C function that has the right instructions for TLSDESC access as LLVM bytecode or native code. The TLSDESC access pattern should be a simple read the GOT and call into something, which is an inlineable snippet - in fact we're really using C as a portable ASM macro here (plus the fact that it sets the proper relocation table). In any case, it seems simple enough that I'll definitely try to do it (it's also interesting that in high-perf Rust code you could use this technique to get TLSDESC perf for selected thread-locals).

[ivoanjo]

I'm rather confident on that front.

Coolio! Let's use it then! This is all experimental anyway so we can always go back if we find any problems with the approach.

[scottgerring]

Nice!

[scottgerring]

Do we ever strip, and if we do, is this sufficient to ensure the TLS symbol remains? In the demo model I used a linker script but tbh I don't know if that was necessary

[yannham]

Good question. Let me research to see if the TLS symbols could be stripped; the script looks simple enough that it's ok to include it if needed.

[yannham]

Thanks for pointing this out. I've done some experimentation by adding a dummy wrapper in the -ffi crate and indeed by default the symbol is stripped (or made local, to be precise). It turns out a working solution is to use a linker version script like:

{
    global: custom_labels_current_set_v2;
}

on the FFI crate. The reason is that rustc generates itself a version script that looks like local: *, hiding all symbols not explicitly exported/whitelisted by default, and this happens after all the C business. Thankfully LLD support multiple version scripts (kinda merge them), and we can make sure the TLS variable makes it to the final shared library by providing our own additional one.

Though all of this is required, I believe it doesn't really affect this PR, because this will be needed on the -ffi crate producing the shared library, not the pure Rust part.

[scottgerring]

@ivoanjo are we going to launch from 'v2' for this

[ivoanjo]

Good point -- let's discuss in the channel.

[nsavoire]

That's very cool! Although I soft-wonder if it'll still respect the TLSDESC... 👀

I'm rather confident on that front. What I want and that @nsavoire managed to do is just to inline the C function that has the right instructions for TLSDESC access as LLVM bytecode or native code. The TLSDESC access pattern should be a simple read the GOT and call into something, which is an inlineable snippet - in fact we're really using C as a portable ASM macro here (plus the fact that it sets the proper relocation table). In any case, it seems simple enough that I'll definitely try to do it (it's also interesting that in high-perf Rust code you could use this technique to get TLSDESC perf for selected thread-locals).

I made some experiments and turns out @ivoanjo is right: TLSDESC is lost in the C -> llvm IR translation.
Adding RUSTFLAGS= "-C link-arg=-Wl,-plugin-opt=--enable-tlsdesc" (in addition to -C linker-plugin-lto, required for cross-language LTO) forces use of TLSDESC.
But then we might as well use RUSTFLAGS="-C llvm-args=--enable-tlsdesc" with pure Rust TLS variable.

[yannham]

I made some experiments and turns out @ivoanjo is right: TLSDESC is lost in the C -> llvm IR translation.
Adding RUSTFLAGS= "-C link-arg=-Wl,-plugin-opt=--enable-tlsdesc" (in addition to -C linker-plugin-lto, required for cross-language LTO) forces use of TLSDESC.

A good demonstration that being confident doesn't equal to being right 😅 though it's beyond me how LTO could change the TLS model that is used, since the emitted instructions are different, so somehow the TLS access pattern is "already compiled and decided" at the point where the linker does its magic (well, it's not entirely true for the relocations, but at least for the instructions emitted).

But then we might as well use RUSTFLAGS="-C llvm-args=--enable-tlsdesc" with pure Rust TLS variable.

Won't that permeate through at least the whole crate, if not all of libdatadog? Do you know how much could we scope this option?

[yannham]

So indeed linkers might actually change the TLS access model at link-time: Link-time TLS relaxation. Though I don't see how this could affect a dynamic library. I can see how it could happen in an executable, though.

In any case, I think cross-language inlining is premature for this PR, so I'll ignore this for now and experiment later when we have a proper FFI in "real" conditions, if that sounds good to everyone.