[herin049]

Description

Reduces per-span overhead in the SDK hot path. Most performance gains where achieved by reordering branch statements, inlining a few function calls and switching from keyword to positional calling where possible.

Note: There are some very minor/subtle breaking changes with adding __slots__ on some classes (i.e. __dict__ goes away on Span objects). Likewise, the bounded attributes change does switch the default from None to sys.maxsize, but this one is extremely unlikely to break any existing code. If we're concerned about the __slots__ changes, we can remove this from the PR.

Type of change

Please delete options that are not relevant.

• Bug fix (non-breaking change which fixes an issue)
• New feature (non-breaking change which adds functionality)
• Breaking change (fix or feature that would cause existing functionality to not work as expected)
• This change requires a documentation update

Benchmarks (MacBook M5, CPython 3.14)

opentelemetry-sdk/benchmarks/trace/test_benchmark_trace.py

Benchmark	Before	After	Δ abs	Δ rel
start_span_with_attributes[0]	4.10 µs	3.35 µs	−0.75 µs	−18.3%
start_span_with_attributes[1]	5.04 µs	3.67 µs	−1.37 µs	−27.3%
start_span_with_attributes[10]	6.85 µs	4.60 µs	−2.25 µs	−32.8%
start_span_with_attributes[50]	15.50 µs	8.85 µs	−6.65 µs	−42.9%
start_span_with_attributes[128]	31.29 µs	16.87 µs	−14.42 µs	−46.1%
simple_start_span	5.92 µs	4.44 µs	−1.48 µs	−25.0%
simple_start_as_current_span	7.35 µs	5.85 µs	−1.50 µs	−20.4%
set_attribute[10]	7.96 µs	6.04 µs	−1.92 µs	−24.1%
read_events[10]	19.12 µs	13.29 µs	−5.83 µs	−30.5%
set_attribute_types[str]	251.45 ms	181.52 ms	−69.93 ms	−27.8%
set_attribute_types[seq_int]	710.96 ms	749.61 ms	+38.66 ms	+5.4%

Per span, memory drops by 56 B: the span object goes from 344 B to 224 B, and SpanContext from 112 B to 96 B.

Checklist:

See contributing.md for styleguide, changelog guidelines, and more.

• I am a human and not a bot
• I have run tox -e precommit
• I have run tox -e typecheck
• I have run the relevant tests and linting tox environments
• This requires the following PR in contrib:

[opentelemetry-pr-dashboard]

Pull request dashboard status

Waiting on the author · refreshed 2026-11-09 18:13 UTC

Resolve merge conflicts.

Two things need attention:

• Required checks are failing — investigate the failures.
• 3 review items — respond to each (e.g. link a commit, explain why not, ask a follow-up):
  • Inline threads: 1, 2, 3

Status above doesn't look right?

• Just replied or pushed? Anything around or after the refresh time above may not be picked up yet — give it a few minutes.
• Should this be with reviewers? Comment /dashboard route:reviewers to route it to them.
• Anything wrong — including the routing? Report it with what you expected; it helps us improve the dashboard.

[xrmx]

Can we quantify how the named arguments contribute to this? Not a fan on making things less readable.

[herin049]

IIRC it does result in a >10% speedup for Span creation. I'll add benchmarks breaking it down. I agree though, switching to positional args makes it hard to read, if you'd prefer we remove it that's OK with me.

[herin049]

Contribution of positional arguments

Apple M5 Pro, CPython 3.14.5

Benchmark	Baseline (existing)	Optimized, keyword calls	Optimized, positional calls	Positional saving	Time reduction
start_span_with_attributes[0]	4.10 µs	3.89 µs	3.13 µs	+0.76 µs	+19.6%
start_span_with_attributes[1]	5.04 µs	4.14 µs	3.38 µs	+0.75 µs	+18.2%
start_span_with_attributes[10]	6.85 µs	5.11 µs	4.38 µs	+0.73 µs	+14.4%
start_span_with_attributes[50]	15.50 µs	9.27 µs	8.55 µs	+0.72 µs	+7.8%
start_span_with_attributes[128]	31.29 µs	16.97 µs	16.25 µs	+0.72 µs	+4.2%
simple_start_span	5.92 µs	4.99 µs	4.27 µs	+0.72 µs	+14.4%
simple_start_as_current_span	7.35 µs	6.48 µs	5.67 µs	+0.80 µs	+12.4%
set_attribute[10]	7.96 µs	6.58 µs	5.81 µs	+0.77 µs	+11.7%
read_events[10]	19.12 µs	13.69 µs	13.05 µs	+0.64 µs	+4.7%
set_attribute_types[str]	251.45 ms	176.58 ms	172.61 ms	+3.97 ms	+2.2%
set_attribute_types[seq_int]	710.96 ms	709.95 ms	706.70 ms	+3.26 ms	+0.5%

@xrmx

[opentelemetry-pr-dashboard]

Hi @herin049 — just a friendly reminder that this pull request is waiting on you. The dashboard status comment has the open items and is kept current.

• Replying is enough to hand it off — answer, explain why no change is needed, or ask a follow-up. The dashboard routes it onward once nothing on the list is waiting on you.
• To hand it back for any other reason, including the dashboard getting this wrong, comment /dashboard route:reviewers.

[rads-1996]

Could we keep input iteration and cleanup outside this lock? For a span with existing attributes, span.set_attributes(span.attributes) tries to acquire the same lock twice and hangs.

[rads-1996]

Updating an existing attribute below capacity no longer moves it to the end, so a recently updated attribute can now be dropped instead of an older one. Could we preserve the previous eviction behavior?