bench: add and use undo-data for BlockToJsonVerbosity3 bench #36254

0xB10C commented at 7:56 PM on September 14, 2026: contributor

blockToJSON only includes prevouts and fees when we have undo-data. The benchmarks did not have any undo-data, so TxVerbosity::SHOW_DETAILS and TxVerbosity::SHOW_DETAILS_AND_PREVOUT did exactly the same work.

This adds the 136 KB of undo-data of block 413567 and uses it in the benchmarks.

$ ./build/bin/bench_bitcoin --filter=BlockToJson.* -min-time=5000

Before:

ns/op	op/s	err%	total	benchmark
10,273,470.25	97.34	0.2%	5.51	`BlockToJsonVerboseWrite`
104,973.68	9,526.20	0.1%	5.51	`BlockToJsonVerbosity1`
17,071,753.65	58.58	0.1%	5.52	`BlockToJsonVerbosity2`
17,256,754.10	57.95	0.1%	5.51	`BlockToJsonVerbosity3`

Note that the BlockToJsonVerbosity2 and BlockToJsonVerbosity3 times were the same.

After:

ns/op	op/s	err%	total	benchmark
16,207,066.75	61.70	0.7%	5.45	`BlockToJsonVerboseWrite`
109,119.99	9,164.22	0.4%	5.59	`BlockToJsonVerbosity1`
17,907,169.45	55.84	0.1%	5.49	`BlockToJsonVerbosity2`
34,274,126.50	29.18	0.1%	5.30	`BlockToJsonVerbosity3`

BlockToJsonVerboseWrite: slower as it now also writes the prevouts too
BlockToJsonVerbosity1: unchanged
BlockToJsonVerbosity2: slightly slower as we now read the undo-data and calculate the fee field
BlockToJsonVerbosity3: slower as it's using undo-data now

DrahtBot added the label Tests on Sep 14, 2026

DrahtBot commented at 7:57 PM on September 14, 2026: contributor

The following sections might be updated with supplementary metadata relevant to reviewers and maintainers.

Code Coverage & Benchmarks

For details see: https://corecheck.dev/bitcoin/bitcoin/pulls/36254.

Reviews

See the guideline and AI policy for information on the review process.

Type	Reviewers
ACK	ismaelsadeeq

If your review is incorrectly listed, please copy-paste <code></code> into the comment that the bot should ignore.

DrahtBot added the label CI failed on Sep 14, 2026

DrahtBot commented at 8:58 PM on September 14, 2026: contributor

🚧 At least one of the CI tasks failed. <sub>Task iwyu: https://github.com/bitcoin/bitcoin/actions/runs/34889958781/job/104129847496</sub> <sub>LLM reason (✨ experimental): CI failed because IWYU detected include issues (“^^^ ⚠️ Failure generated from IWYU”) and exited with a non-zero status.</sub>

<details><summary>Hints</summary>

Try to run the tests locally, according to the documentation. However, a CI failure may still happen due to a number of reasons, for example:

Possibly due to a silent merge conflict (the changes in this pull request being incompatible with the current code in the target branch). If so, make sure to rebase on the latest commit of the target branch.
A sanitizer issue, which can only be found by compiling with the sanitizer and running the affected test.
An intermittent issue.

Leave a comment here, if you need help tracking down a confusing failure.

</details>

bench: add undo-data for BlockToJsonVerbosity3

blockToJSON only includes prevouts and fees when we have undo-data.
The benchmarks did not have any undo-data, so TxVerbosity::SHOW_DETAILS
and TxVerbosity::SHOW_DETAILS_AND_PREVOUT did exactly the same work.

This adds the 139KB of undo-data of block 413567 and uses it in the
benchmarks.

```
$ ./build/bin/bench_bitcoin --filter=BlockToJson.* -min-time=5000
```

Before:

|               ns/op |                op/s |    err% |     total | benchmark
|--------------------:|--------------------:|--------:|----------:|:----------
|       10,273,470.25 |               97.34 |    0.2% |      5.51 | `BlockToJsonVerboseWrite`
|          104,973.68 |            9,526.20 |    0.1% |      5.51 | `BlockToJsonVerbosity1`
|       17,071,753.65 |               58.58 |    0.1% |      5.52 | `BlockToJsonVerbosity2`
|       17,256,754.10 |               57.95 |    0.1% |      5.51 | `BlockToJsonVerbosity3`  (same as Verbosity2 as undo-data is missing)

After:

|               ns/op |                op/s |    err% |     total | benchmark
|--------------------:|--------------------:|--------:|----------:|:----------
|       16,207,066.75 |               61.70 |    0.7% |      5.45 | `BlockToJsonVerboseWrite`
|          109,119.99 |            9,164.22 |    0.4% |      5.59 | `BlockToJsonVerbosity1`
|       17,907,169.45 |               55.84 |    0.1% |      5.49 | `BlockToJsonVerbosity2`
|       34,274,126.50 |               29.18 |    0.1% |      5.30 | `BlockToJsonVerbosity3`

- `BlockToJsonVerboseWrite`: slower as it now also writes the prevouts too
- `BlockToJsonVerbosity1`: unchanged
- `BlockToJsonVerbosity2`: slightly slower as we now read the undo-data and calculate the fee field
- `BlockToJsonVerbosity3`: slower as it's using undo-data now

0771dd89d6

0xB10C force-pushed on Sep 14, 2026

DrahtBot removed the label CI failed on Sep 14, 2026

ismaelsadeeq approved

ismaelsadeeq commented at 9:13 AM on September 15, 2026: member

Tested ACK 0771dd89d6eec31ea39f25b283aacbf2092c7c15

I reproduced the benchmark and verified the undo-data blob.

Metric	Verbosity2 ins/op	Verbosity3 ins/op
before	326,698,108	326,688,650
after	341,707,431	621,328,726

0xB10C commented at 7:23 PM on September 15, 2026: contributor

As a side note:

There's also the question how reliable a block from 10 years ago is for benchmarking BlockToJsonVerbosity3 today. The output types changed a lot and since we spend a lot of time on per-script work (address encoding, descriptor inference, checksums), the script type mix decides what the benchmark measures. With the old block, probably don't have good benchmark-coverage.

We could add more recent blocks here, but not sure if that's worth the effort and repository size bloat. Maybe a crafted benchmark-block including P2PKH, P2SH, SegWit, Taproot, ... is an alternative. Or use @vostrnad's transaction for something like this.

Still worth it to fix this benchmark now. Everything else can be a followup, if fast-ish verbose getblock is important to people.

sedited commented at 9:54 PM on September 15, 2026: contributor

We could add more recent blocks here, but not sure if that's worth the effort and repository size bloat. Maybe a crafted benchmark-block including P2PKH, P2SH, SegWit, Taproot, ... is an alternative.

You might be interested in reviewing #32554 then.