Reland debug impl for fieldless enums with concatenated string without using discriminant_value by panstromek · Pull Request #162535 · rust-lang/rust

panstromek · 2026-09-09T13:17:53Z

relands part of #155452
partially reverts #162164
fixes only a part of #133945

This implementation avoids the problem in the original PR by using simple match instead of indexing by discriminant_value. This avoids the cost of many small strings, generates less MIR and reduces borrowck cost, but it doesn't avoid the match checking costs (obviously).

It uses slightly different strategy to avoid more MIR costs - it doesn't generate array of indexes. It maps the enum variant directly to the start index. Names are separated by space, so the outlined function figures out the end index based on that.

Some more measurements and reasoning behind this implementation is in this comment

This PR history contains various experiments, #162535 (comment) is the perf run for the final version, which also includes results for the enum-stress, which is a benchmark that contains a stress test for this case.

I also noticed there's not really a test testing the output of some of these derives (especially big ones), so I added a case to an existing test that covers this.

panstromek · 2026-09-09T13:18:10Z

@bors try @rust-timer queue

Reland debug impl for fieldless enums with concatenated string without using discriminant_value

rust-bors · 2026-09-09T14:17:23Z

☀️ Try build successful (CI)
Build commit: 6ff0c29 (6ff0c293700f750741ca037a07d54262de940377)
Base parent: eca445e (eca445e5ae4a6679cc27d3a09106ce245e13a5a6)

rust-timer · 2026-09-09T22:37:33Z

Finished benchmarking commit (6ff0c29): comparison URL.

Overall result: ❌ regressions - no action needed

Benchmarking means the PR may be perf-sensitive. Consider adding rollup=never if this change is not fit for rolling up.

@rustbot label: -S-waiting-on-perf -perf-regression

Instruction count

Our most reliable metric. Used to determine the overall result above. However, even this metric can be noisy.

	mean	range	count
Regressions ❌ (primary)	0.1%	[0.1%, 0.1%]	1
Regressions ❌ (secondary)	0.3%	[0.2%, 0.5%]	4
Improvements ✅ (primary)	-	-	0
Improvements ✅ (secondary)	-	-	0
All ❌✅ (primary)	0.1%	[0.1%, 0.1%]	1

Max RSS (memory usage)

Results (primary 0.8%, secondary -0.0%)

A less reliable metric. May be of interest, but not used to determine the overall result above.

	mean	range	count
Regressions ❌ (primary)	2.2%	[0.5%, 5.2%]	4
Regressions ❌ (secondary)	3.8%	[3.8%, 3.8%]	1
Improvements ✅ (primary)	-2.2%	[-2.4%, -2.0%]	2
Improvements ✅ (secondary)	-1.9%	[-2.4%, -1.5%]	2
All ❌✅ (primary)	0.8%	[-2.4%, 5.2%]	6

Cycles

Results (primary 1.0%, secondary 5.9%)

A less reliable metric. May be of interest, but not used to determine the overall result above.

	mean	range	count
Regressions ❌ (primary)	1.5%	[0.4%, 3.5%]	7
Regressions ❌ (secondary)	5.9%	[5.9%, 5.9%]	1
Improvements ✅ (primary)	-0.7%	[-0.7%, -0.6%]	2
Improvements ✅ (secondary)	-	-	0
All ❌✅ (primary)	1.0%	[-0.7%, 3.5%]	9

Binary size

Results (primary 0.1%, secondary 4.2%)

A less reliable metric. May be of interest, but not used to determine the overall result above.

	mean	range	count
Regressions ❌ (primary)	0.2%	[0.0%, 1.7%]	39
Regressions ❌ (secondary)	4.2%	[0.0%, 13.7%]	21
Improvements ✅ (primary)	-0.4%	[-0.7%, -0.0%]	16
Improvements ✅ (secondary)	-	-	0
All ❌✅ (primary)	0.1%	[-0.7%, 1.7%]	55

Bootstrap: 481.923s -> 481.101s (-0.17%)
Artifact size: 403.28 MiB -> 403.16 MiB (-0.03%)

panstromek · 2026-09-10T06:20:34Z

Replacing discriminant_value with match => variant_index sadly doesn't help much. The generated MIR looks somewhat better than the default match => &str variant (there's less code in each arms basic block), but without the O(n) reduction in basic blocks, it doesn't seem to be worth it.

panstromek · 2026-10-08T14:08:54Z

@bors try @rust-timer queue

Reland debug impl for fieldless enums with concatenated string without using discriminant_value

rust-bors · 2026-10-08T14:20:17Z

💔 Test for 499edf9 failed: CI. Failed jobs:

try - dist-aarch64-linux-quick (web logs, enhanced plaintext logs)
try - dist-x86_64-linux-quick (web logs, enhanced plaintext logs)

panstromek · 2026-10-08T14:43:37Z

@bors try

Reland debug impl for fieldless enums with concatenated string without using discriminant_value

rust-bors · 2026-10-08T15:48:12Z

☀️ Try build successful (CI)
Build commit: 3b82c9e (3b82c9e09a3707bda85dc9eb54201ae70eb921c8)
Base parent: 65efa19 (65efa194201c67ba29b7384255a534d61c2aa3cb)

rust-timer · 2026-10-08T16:37:51Z

Finished benchmarking commit (3b82c9e): comparison URL.

Overall result: ✅ improvements - no action needed

Benchmarking means the PR may be perf-sensitive. It's automatically marked not fit for rolling up. Overriding is possible but disadvised: it risks changing compiler perf.

@bors rollup=never rustc-perf
@rustbot label: -S-waiting-on-perf -perf-regression

Instruction count

Our most reliable metric. Used to determine the overall result above. However, even this metric can be noisy.

	mean	range	count
Regressions ❌ (primary)	-	-	0
Regressions ❌ (secondary)	0.6%	[0.5%, 0.7%]	2
Improvements ✅ (primary)	-0.3%	[-0.6%, -0.1%]	14
Improvements ✅ (secondary)	-0.4%	[-0.8%, -0.1%]	59
All ❌✅ (primary)	-0.3%	[-0.6%, -0.1%]	14

Max RSS (memory usage)

Results (primary 0.6%, secondary -1.5%)

A less reliable metric. May be of interest, but not used to determine the overall result above.

	mean	range	count
Regressions ❌ (primary)	1.9%	[0.7%, 4.3%]	5
Regressions ❌ (secondary)	2.1%	[2.1%, 2.1%]	1
Improvements ✅ (primary)	-2.4%	[-3.5%, -1.3%]	2
Improvements ✅ (secondary)	-2.4%	[-3.6%, -1.1%]	4
All ❌✅ (primary)	0.6%	[-3.5%, 4.3%]	7

Cycles

Results (secondary -2.5%)

A less reliable metric. May be of interest, but not used to determine the overall result above.

	mean	range	count
Regressions ❌ (primary)	-	-	0
Regressions ❌ (secondary)	-	-	0
Improvements ✅ (primary)	-	-	0
Improvements ✅ (secondary)	-2.5%	[-2.9%, -2.2%]	2
All ❌✅ (primary)	-	-	0

Binary size

Results (primary -0.1%, secondary 3.9%)

A less reliable metric. May be of interest, but not used to determine the overall result above.

	mean	range	count
Regressions ❌ (primary)	1.2%	[1.1%, 1.2%]	4
Regressions ❌ (secondary)	3.9%	[0.6%, 8.0%]	13
Improvements ✅ (primary)	-0.4%	[-0.7%, -0.0%]	18
Improvements ✅ (secondary)	-	-	0
All ❌✅ (primary)	-0.1%	[-0.7%, 1.2%]	22

Bootstrap: 485.904s -> 487.919s (0.41%)
Artifact size: 406.44 MiB -> 406.38 MiB (-0.02%)

nnethercote · 2026-10-21T19:22:37Z

@bors r+

rust-bors · 2026-10-21T19:22:40Z

📌 Commit 11200f7 has been tentatively approved by nnethercote

It will be put into the queue for this repository once PR CI succeeds.

🌲 The tree is currently closed for pull requests below priority 5. This pull request will be tested once the tree is reopened.

Reason for tree closure: github outage

This reduces compile time for large enums. Co-authored-by: Makai <m4kai410@gmail.com>

This derive is only used for large nums, so it doesn't make sense to use small enum as an example

rustbot · 2026-10-21T20:02:31Z

This PR was rebased onto a different main commit. Here's a range-diff highlighting what actually changed.

Rebasing is a normal part of keeping PRs up to date, so no action is needed—this note is just to help reviewers.

panstromek · 2026-10-21T20:03:34Z

rebased

@rustbot ready

nnethercote · 2026-10-21T20:16:16Z

@bors r+

rust-bors · 2026-10-21T20:16:20Z

📌 Commit 0076e09 has been tentatively approved by nnethercote

It will be put into the queue for this repository once PR CI succeeds.

🌲 The tree is currently closed for pull requests below priority 5. This pull request will be tested once the tree is reopened.

Reason for tree closure: github outage

Rollup of 6 pull requests Successful merges: - #162535 (Reland debug impl for fieldless enums with concatenated string without using discriminant_value) - #164060 (Add hightlight to acceptable characters in macro rules) - #164583 (Run -Cpasses during OptimizerLastEP) - #164293 (Tweak missing fields error wording) - #164563 (Removes unwanted backticks from slice::as_ascii documentation.) - #164683 (Add -Zasync-panic flag)

Rollup merge of #162535 - panstromek:reland-debug-opt, r=nnethercote Reland debug impl for fieldless enums with concatenated string without using discriminant_value relands part of #155452 partially reverts #162164 fixes only a part of #133945 This implementation avoids the problem in the original PR by using simple match instead of indexing by `discriminant_value`. This avoids the cost of many small strings, generates less MIR and reduces borrowck cost, but it doesn't avoid the match checking costs (obviously). It uses slightly different strategy to avoid more MIR costs - it doesn't generate array of indexes. It maps the enum variant directly to the start index. Names are separated by space, so the outlined function figures out the end index based on that. Some more measurements and reasoning behind this implementation is in [this comment](#162535 (comment)) This PR history contains various experiments, #162535 (comment) is the perf run for the final version, which also includes results for the `enum-stress`, which is a benchmark that contains a stress test for this case. I also noticed there's not really a test testing the output of some of these derives (especially big ones), so I added a case to an existing test that covers this.

rust-timer · 2026-10-22T06:58:51Z

Note

This PR was benchmarked as part of triage of its containing rollup: triage URL.

Finished benchmarking commit (fd1f60f): comparison URL.

Overall result: ✅ improvements - no action needed

@rustbot label: -perf-regression

Instruction count

Our most reliable metric. Used to determine the overall result above. However, even this metric can be noisy.

	mean	range	count
Regressions ❌ (primary)	-	-	0
Regressions ❌ (secondary)	0.3%	[0.3%, 0.3%]	1
Improvements ✅ (primary)	-0.3%	[-0.4%, -0.1%]	3
Improvements ✅ (secondary)	-4.7%	[-10.4%, -0.0%]	14
All ❌✅ (primary)	-0.3%	[-0.4%, -0.1%]	3

Max RSS (memory usage)

Results (secondary -60.7%)

A less reliable metric. May be of interest, but not used to determine the overall result above.

	mean	range	count
Regressions ❌ (primary)	-	-	0
Regressions ❌ (secondary)	0.4%	[0.4%, 0.4%]	1
Improvements ✅ (primary)	-	-	0
Improvements ✅ (secondary)	-68.4%	[-91.1%, -1.8%]	8
All ❌✅ (primary)	-	-	0

Cycles

Results (secondary -14.0%)

A less reliable metric. May be of interest, but not used to determine the overall result above.

	mean	range	count
Regressions ❌ (primary)	-	-	0
Regressions ❌ (secondary)	2.3%	[1.6%, 3.5%]	3
Improvements ✅ (primary)	-	-	0
Improvements ✅ (secondary)	-20.1%	[-28.5%, -1.6%]	8
All ❌✅ (primary)	-	-	0