feat: add blst bindings #18

spiral-ladder · 2025-09-30T16:40:22Z

This PR implements the blst bun bindings as in ChainSafe/blst-z#49

What this PR does:

moved tests, backing TS code, and eth_c_abi.zig within that PR here as blst.zig
cleaned up some commented out / unused TS code in that PR

What this PR does NOT do:

include benchmark code. Left out test/perf for a separate PR

spiral-ladder · 2025-09-30T17:02:19Z

test/blst/aggregateWithRandomness.test.ts

+				aggregateWithRandomness(
+					sets.concat({
+						pk: sets[0].pk,
+						//TODO: (@matthewkeil) this throws error "Public key is infinity" not signature because there is only one blst error


what does this TODO mean, is this still relevant? @matthewkeil

yes its still relevant. Check out the error type https://github.com/ChainSafe/blst-z/pull/49/files#diff-31ebad738598504e01f7313f21e76c7eb1c128f9cc636dca0fa8957b7ec15748R22 which wraps c.BLST_PK_IS_INFINITY, but no corresponding c.BLST_SIG_IS_INFINITY exists. The way the error is used is that it is returned whenever either p1 or p2 is infinity.

So it shouldn't be a TODO, but a note?

wemeetagain · 2025-10-16T22:48:31Z

I just added a few benchmarks to sanity check, somehow looks quite bad?

  blst
    ✔ blst - keygen                                                       372023.8 ops/s    2.688000 us/op        -     510549 runs   1.62 s
    ✔ other - keygen                                                      376081.2 ops/s    2.659000 us/op        -     635747 runs   2.02 s
    ✔ blst - sign                                                         1688.576 ops/s    592.2150 us/op        -        509 runs  0.814 s
    ✔ other - sign                                                        1942.408 ops/s    514.8250 us/op        -        584 runs  0.808 s
    ✔ blst - verify                                                       463.1062 ops/s    2.159332 ms/op        -         71 runs  0.890 s
    ✔ other - verify                                                      579.2484 ops/s    1.726375 ms/op        -        213 runs   1.23 s
    ✔ blst - aggregateVerify 16                                           83.31737 ops/s    12.00230 ms/op        -         37 runs   2.05 s
    ✔ other - aggregateVerify 16                                          308.3152 ops/s    3.243434 ms/op        -        106 runs   4.02

twoeths · 2025-10-17T02:11:22Z

looks like we did not ultilize multi-threads? also need to make sure a Release build of zig, refer to ChainSafe/blst-bun#14 (comment)

aggregateVerify
      ✔ aggregateVerify - 1 sets                                            968.8627 ops/s    1.032138 ms/op   x1.038        197 runs  0.706 s
      ✔ aggregateVerify - 8 sets                                            468.5928 ops/s    2.134049 ms/op   x0.647        142 runs  0.812 s
      ✔ aggregateVerify - 32 sets                                           337.2375 ops/s    2.965269 ms/op   x0.259        474 runs   1.94 s
      ✔ aggregateVerify - 128 sets                                          128.2682 ops/s    7.796166 ms/op   x0.177         27 runs  0.737 s

spiral-ladder · 2025-10-17T09:30:46Z

@twoeths is right, the port did not contain the mt version. I've since ported it over in ChainSafe/blst-z#56, and is using it in a branch that uses this as base: #21

The aggregation numbers look better but is still underperforming even under zig build -Doptimize=ReleaseFast:

before:

bench/blst.bench.ts
    ✔ blst - aggregateVerify 16                                           144.0573 ops/s    6.941682 ms/op        -         29 runs   1.53 s
    ✔ other - aggregateVerify 16                                          846.1060 ops/s    1.181885 ms/op        -         55 runs   3.70 s

after:

bench/blst.bench.ts
    ✔ blst - aggregateVerify 16                                           498.5373 ops/s    2.005868 ms/op        -        116 runs   3.71 s
    ✔ other - aggregateVerify 16                                          841.3663 ops/s    1.188543 ms/op        -         81 runs   3.90 s

verify is also 2x slower on mac and that probably needs a separate look.

This missing c flag as well as the macro were causing some regressions in performance within lodestar-bun versus production blst-ts in ChainSafe/lodestar-bun#18.

* use new aggregateverify * update blst * remove unnecessary bytes.bench.ts from bench cmd * Update blst aggregateVerify * (change me) temporary changes to bench blst aggregateVerify only * bist: fix init

spiral-ladder · 2025-10-31T08:37:13Z

latest numbers with proper thread pool/memory pool impl. On M2 Mac it looks bad still:

  blst
    ✔ blst - aggregateVerify 1                                            1158.990 ops/s    862.8200 us/op        -        160 runs  0.942 s
    ✔ other - aggregateVerify 1                                           1532.520 ops/s    652.5200 us/op        -        288 runs   1.11 s
    ✔ blst - aggregateVerify 16                                           764.2134 ops/s    1.308535 ms/op        -         93 runs   3.72 s
    ✔ other - aggregateVerify 16                                          833.3125 ops/s    1.200030 ms/op        -         70 runs   3.85 s
    ✔ blst - aggregateVerify 32                                           510.0693 ops/s    1.960518 ms/op        -         29 runs   4.27 s
    ✔ other - aggregateVerify 32                                          580.7518 ops/s    1.721906 ms/op        -         29 runs   4.78 s
    ✔ blst - aggregateVerify 128                                          176.6967 ops/s    5.659416 ms/op        -          7 runs   5.52 s
    ✔ other - aggregateVerify 128                                         182.8872 ops/s    5.467850 ms/op        -          5 runs   5.65 s

On Linux it's more reasonable, and outperforms at smaller sizes:

bench/blst.bench.ts
  blst
    ✔ blst - aggregateVerify 1                                            911.1152 ops/s    1.097556 ms/op        -        193 runs   1.03 s
    ✔ other - aggregateVerify 1                                           868.3586 ops/s    1.151598 ms/op        -       1092 runs   2.57 s
    ✔ blst - aggregateVerify 16                                           344.3300 ops/s    2.904191 ms/op        -         87 runs   2.90 s
    ✔ other - aggregateVerify 16                                          326.3093 ops/s    3.064577 ms/op        -         86 runs   2.83 s
    ✔ blst - aggregateVerify 32                                           220.8602 ops/s    4.527752 ms/op        -         28 runs   2.77 s
    ✔ other - aggregateVerify 32                                          198.3364 ops/s    5.041940 ms/op        -         61 runs   3.48 s
    ✔ blst - aggregateVerify 128                                          67.22979 ops/s    14.87436 ms/op        -         15 runs   4.19 s
    ✔ other - aggregateVerify 128                                         67.98385 ops/s    14.70938 ms/op        -         29 runs   5.21 s

wemeetagain · 2025-10-31T15:06:44Z

bench/blst.bench.ts

+  //    },
+  //  });
+
+  bench({


can this code be deduped a bit?

even just a for loop with the various number of signatures/messages

For sure, this was supposed to be temporary so I quickly did this. The version before was pretty concise, I'll switch to that

wemeetagain · 2025-10-31T15:08:03Z

src/binding.ts

+  "init": {
+    "args": [],
+    "returns": "u32"
+  },
+  "deinit": {
+    "args": [],
+    "returns": "void"
+  },


these function names probably need to be qualified.

spiral-ladder self-assigned this Sep 30, 2025

spiral-ladder force-pushed the bing/blst-bun branch 4 times, most recently from 726a8c4 to 062fc28 Compare September 30, 2025 17:09

spiral-ladder marked this pull request as ready for review September 30, 2025 17:14

spiral-ladder requested a review from a team as a code owner September 30, 2025 17:14

spiral-ladder commented Sep 30, 2025

View reviewed changes

spiral-ladder mentioned this pull request Oct 20, 2025

fix: add -O2 flag and sha optimizations for arm ChainSafe/blst.zig#1

Merged

spiral-ladder and others added 17 commits October 31, 2025 13:57

add blst

aa6617b

fix bindings

b898003

feat: move blst tests

44a2415

clean up blst dep

159a5e7

rm blst/.benchrc.yaml

4e2fb84

remove redundant code

03096d3

remove misisng file from export

71b55a9

centralize blst exports at src/blst.ts

39bbcb5

update build.zig.zon

21b021c

remove seemingly redundant commented out code

5cbeb80

remove commented out code in aggregatedWithRandomness.test.ts

6141690

remove afterAll closeBinding in tests

229168b

use toEqual for deep equals

4762165

remove benchmark TODO in verifyMultipleAggregateSignatures

000b4d3

remove more comments

f3c671a

update blst

b26d201

chore: move blst dependency in alpha order

8b1e5f8

wemeetagain added 3 commits October 31, 2025 13:58

chore: simplify hex utils

9b53bb7

chore: small import cleanup

7926626

feat: add some blst benchmarks

361b6a2

spiral-ladder force-pushed the bing/blst-bun branch from 0effe9e to 361b6a2 Compare October 31, 2025 05:59

spiral-ladder mentioned this pull request Oct 31, 2025

use new aggregateverify #21

Merged

use new aggregateverify (#21)

69bfae9

* use new aggregateverify * update blst * remove unnecessary bytes.bench.ts from bench cmd * Update blst aggregateVerify * (change me) temporary changes to bench blst aggregateVerify only * bist: fix init

spiral-ladder added 3 commits October 31, 2025 19:30

fix blst

d85fc71

regen bindings

b62e36c

fix bun-ffi-z dep

81f6198

spiral-ladder force-pushed the bing/blst-bun branch from 022f861 to 81f6198 Compare October 31, 2025 11:35

spiral-ladder added 2 commits October 31, 2025 22:12

update blst

6e849dc

blst(bun): init memory pool before tests

1a1a0d6

wemeetagain reviewed Oct 31, 2025

View reviewed changes

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Uh oh!

feat: add blst bindings #18

feat: add blst bindings #18

Uh oh!

spiral-ladder commented Sep 30, 2025 •

edited

Loading

Uh oh!

spiral-ladder Sep 30, 2025

Uh oh!

wemeetagain Sep 30, 2025

Uh oh!

spiral-ladder Oct 2, 2025

Uh oh!

wemeetagain Oct 16, 2025

Uh oh!

wemeetagain commented Oct 16, 2025

Uh oh!

twoeths commented Oct 17, 2025

Uh oh!

spiral-ladder commented Oct 17, 2025 •

edited

Loading

Uh oh!

spiral-ladder commented Oct 31, 2025 •

edited

Loading

Uh oh!

wemeetagain Oct 31, 2025

Uh oh!

spiral-ladder Oct 31, 2025

Uh oh!

wemeetagain Oct 31, 2025

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

4 participants

+                //    },
+                //  });
+                bench({

feat: add blst bindings #18

Are you sure you want to change the base?

feat: add blst bindings #18

Uh oh!

Conversation

spiral-ladder commented Sep 30, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

spiral-ladder Sep 30, 2025

Choose a reason for hiding this comment

Uh oh!

wemeetagain Sep 30, 2025

Choose a reason for hiding this comment

Uh oh!

spiral-ladder Oct 2, 2025

Choose a reason for hiding this comment

Uh oh!

wemeetagain Oct 16, 2025

Choose a reason for hiding this comment

Uh oh!

wemeetagain commented Oct 16, 2025

Uh oh!

twoeths commented Oct 17, 2025

Uh oh!

spiral-ladder commented Oct 17, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

spiral-ladder commented Oct 31, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

wemeetagain Oct 31, 2025

Choose a reason for hiding this comment

Uh oh!

spiral-ladder Oct 31, 2025

Choose a reason for hiding this comment

Uh oh!

wemeetagain Oct 31, 2025

Choose a reason for hiding this comment

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

4 participants

spiral-ladder commented Sep 30, 2025 •

edited

Loading

spiral-ladder commented Oct 17, 2025 •

edited

Loading

spiral-ladder commented Oct 31, 2025 •

edited

Loading