bench: add BIP-352 pubkey extraction benchmarks (standard + P2PKH worst-case) #36475

pull theStack wants to merge 2 commits into bitcoin:master from theStack:bip352-bench-pubkey_extraction changing 2 files +112 −0

theStack commented at 2:30 PM on October 8, 2026: contributor

As there are ongoing changes to the Silent Payments public key extraction function GetPubKeyFromInput (see recently merged #36338, upcoming #36417), it seems reasonable to add benchmarks for it. The first commit covers the standard shape of each eligible input type (P2PKH, P2SH-P2WPKH, P2WPKH and P2TR script-path spend; see BIP-352). The second adds a worst-case P2PKH input with a maximally malleated scriptSig, where every push has to be hashed during the scan.

Results on my arm64 machine (as per $ ./build/bin/bench_bitcoin -filter='BIP352_.*'):

ns/input	input/s	err%	total	benchmark
3,034.12	329,584.35	0.2%	0.01	`BIP352_GetPubKey_P2PKH`
69,868.20	14,312.66	0.1%	0.01	`BIP352_GetPubKey_P2PKH_WorstCase`
2,841.35	351,945.44	0.1%	0.01	`BIP352_GetPubKey_P2SH_P2WPKH`
2,768.00	361,272.01	0.1%	0.01	`BIP352_GetPubKey_P2TR`
2,757.57	362,637.66	0.0%	0.01	`BIP352_GetPubKey_P2WPKH`

DrahtBot added the label Tests on Oct 8, 2026
DrahtBot commented at 2:30 PM on October 8, 2026: contributor


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



External sites
- Code coverage & benchmarks: https://corecheck.dev/bitcoin/bitcoin/pulls/36475
- Ralph review <img src=https://review.fish.foo/assets/bitcoin/bitcoin/36475.png alt="Ralph review status"> https://git.fish.foo/bitcoin/bitcoin/pulls/36475#author-ralph


Reviews

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

Type Reviewers

ACK Yudis-bit, fametrano

Concept ACK w0xlt

If your review is incorrectly listed, please copy-paste <code></code> into the comment that the bot should ignore.
theStack force-pushed on Oct 8, 2026
DrahtBot added the label CI failed on Oct 8, 2026
w0xlt commented at 5:52 PM on October 8, 2026: contributor

Concept ACK
Yudis-bit commented at 2:53 AM on October 9, 2026: none

Concept and approach ACK.

Numbers make sense given that IsFullyValid() point validation dominates the standard paths. In P2PKH_WorstCase, hoisting pubkey_candidate outside the while loop in GetPubKeyFromInput would save ~290 heap reallocations across the dummy pushes.

It might also be worth adding P2TR_ScriptPath and P2TR_ScriptPath_NUMS to benchmark the control-block validation and the early rejection path.

Type	Reviewers
ACK	Yudis-bit, fametrano
Concept ACK	w0xlt

bench: add benchmarks for BIP-352 pubkey extraction (standard inputs)

Benchmark the public key extraction (`GetPubKeyFromInput`) for the
standard shape of each Silent Payments eligible input type, i.e.
P2PKH, P2SH-P2WPKH, P2WPKH and P2TR. For P2TR, a script-path spend is
used, as it runs all of the key-path extraction logic plus the control
block checks and the NUMS-H internal key comparison.

d71bc7127f

theStack force-pushed on Oct 10, 2026
theStack force-pushed on Oct 10, 2026

bench: add worst-case P2PKH benchmark for BIP-352 pubkey extraction

Add a P2PKH input with a malleated scriptSig that prepends as many
33-byte pushes as the consensus scriptSig size limit (MAX_SCRIPT_SIZE)
allows, each of which has to be hashed during the scan, followed by the
actual signature and public key. The input respects all other consensus
limits (data pushes don't count towards MAX_OPS_PER_SCRIPT, the stack
stays below MAX_STACK_SIZE), but is non-standard as it exceeds
MAX_STANDARD_SCRIPTSIG_SIZE.

Results on my arm64 machine:
```
$ ./build/bin/bench_bitcoin -filter='BIP352_.*'
[...]

|            ns/input |             input/s |    err% |     total | benchmark
|--------------------:|--------------------:|--------:|----------:|:----------
|            3,021.72 |          330,937.16 |    0.1% |      0.01 | `BIP352_GetPubKey_P2PKH`
|           69,754.07 |           14,336.08 |    0.1% |      0.01 | `BIP352_GetPubKey_P2PKH_WorstCase`
|            2,839.92 |          352,122.64 |    0.2% |      0.01 | `BIP352_GetPubKey_P2SH_P2WPKH`
|            2,756.93 |          362,722.60 |    0.1% |      0.01 | `BIP352_GetPubKey_P2TR`
|            2,752.45 |          363,312.34 |    0.0% |      0.01 | `BIP352_GetPubKey_P2WPKH`

```

56c0f168d1

theStack force-pushed on Oct 10, 2026
theStack commented at 5:07 PM on October 10, 2026: contributor

It might also be worth adding P2TR_ScriptPath and P2TR_ScriptPath_NUMS to benchmark the control-block validation and the early rejection path.

Due to the strong similarity I think it's fine to have only a single representative P2TR benchmark, but I switched the one from key-path to script-path spending as the pubkey extraction contains (slightly) more logic. Also fixed the IWYU CI error. Thanks for the review!
DrahtBot removed the label CI failed on Oct 10, 2026

Yudis-bit commented at 11:51 AM on October 17, 2026: none

re-tACK 56c0f168d1

Makes sense to stick to one P2TR case. Windows numbers:

ns/input	input/s	err%	total	benchmark
4,660.44	214,571.81	0.2%	0.01	`BIP352_GetPubKey_P2PKH`
81,266.67	12,305.17	0.2%	0.01	`BIP352_GetPubKey_P2PKH_WorstCase`
4,624.56	216,236.72	0.4%	0.01	`BIP352_GetPubKey_P2SH_P2WPKH`
4,330.00	230,946.88	0.6%	0.01	`BIP352_GetPubKey_P2TR`
4,335.53	230,652.50	0.2%	0.01	`BIP352_GetPubKey_P2WPKH`

fametrano commented at 7:37 AM on October 21, 2026: contributor

tACK 56c0f168d1

The five benchmarks build and run on an Apple M5 (Release build, bench_bitcoin -filter='BIP352_.*' -min-time=2000, three runs). The four standard inputs take 2.2-2.6 µs each. The worst case, a 10,000-byte P2PKH scriptSig with 291 dummy pushes, takes 39-43 µs. I also rebuilt the five input shapes independently. A separate BIP352 implementation (btclib-wallet) extracts the expected key from each.

Contributors

Labels

Linked (view graph)

#36338 BIP-352: fix P2PKH pubkey extraction from malleated scriptSig #36417 bip352: verify redeem script and pubkey hash in GetPubKeyFromInput

bench: add BIP-352 pubkey extraction benchmarks (standard + P2PKH worst-case) #36475

External sites

Reviews