feat(bb): Introduce chunks for univariate computation for the AVM #12707

jeanmon · 2025-03-13T09:27:04Z

This PR introduces a more even vertically distributed trace chunks processing among threads in the univariate computation as part of sumcheck. This leads to a substantial speed up of sumcheck.

Let t be the number of threads. The processing of the rows of a circuit used to be

thread_1 (round_size/t rows)
...
thread_t (round_size/t rows) [end of circuit]

while this PR introduces the possibility of having the processing be interleaved, and therefore the load is more uniformly balanced across threads:

thread_1 (chunk_thread_portion_size rows)
...
thread_t (chunk_thread_portion_size rows)
thread_1 (chunk_thread_portion_size rows)
...
thread_t (chunk_thread_portion_size rows)
...
[end of circuit]

This PR improves #12703 measurement from 8.5 seconds to 2.4 seconds.

fcarreiro · 2025-03-13T09:35:09Z

The gains look amazing! but... I think we should wait until we have the capacity to do a full trace in VM2 for this. Also I think think

the current approach couples the sumcheck_round with the concept of avm flavor, and requires ifdefs, which kind of pollutes the code a lot. I think we can do better.
we should have a more detailed PR explanation with tracy graphs (before/after) on the threads, explaining why this approach works both for short and fuller traces

jeanmon · 2025-03-13T10:18:04Z

@fcarreiro I addressed 1) as it was indeed ugly.

fcarreiro · 2025-03-13T11:15:04Z

The gains look amazing! but... I think we should wait until we have the capacity to do a full trace in VM2 for this. Also I think think

the current approach couples the sumcheck_round with the concept of avm flavor, and requires ifdefs, which kind of pollutes the code a lot. I think we can do better.

we should have a more detailed PR explanation with tracy graphs (before/after) on the threads, explaining why this approach works both for short and fuller traces

Pushed some changes to fix 1

barretenberg/cpp/src/barretenberg/sumcheck/sumcheck_round.hpp

fcarreiro · 2025-03-13T13:10:57Z

Having fixed (1) I'm not against merging this as long as crypto is ok with it. We should however revisit it later.

barretenberg/cpp/src/barretenberg/vm2/generated/flavor.hpp

fcarreiro · 2025-03-13T17:06:52Z

Actually now it looks good to merge!

lucasxia01

Looks fine, and the results are great! I'm not sure how well tested this code is - some of the logic is a little tricky so I would hope for some better testing. Also feel like it needs to be more readable.

barretenberg/cpp/src/barretenberg/vm2/generated/flavor.hpp

lucasxia01 · 2025-03-14T04:43:26Z

barretenberg/cpp/src/barretenberg/sumcheck/sumcheck_round.hpp

+
+            // When the trace is shrunk to a point where the chunk portion size per thread is lower than 2,
+            // we fall back to a single chunk, i.e., we keep the "non-AVM" values.
+            if (thread_portion_size_candidate >= 2) {


why is there an if here? Seems unnecessary?

I added more explanations and actually could simplify a bit the logic. See new version.

lucasxia01 · 2025-03-14T04:43:42Z

barretenberg/cpp/src/barretenberg/sumcheck/sumcheck_round.hpp

+            static_assert(Flavor::MAX_CHUNK_THREAD_PORTION_SIZE >= 2);
+            static_assert((Flavor::MAX_CHUNK_THREAD_PORTION_SIZE & (Flavor::MAX_CHUNK_THREAD_PORTION_SIZE - 1)) == 0);
+
+            const auto thread_portion_size_candidate =


naming is not great.. can't think of proper naming at the moment but at least requires a comment to what this is

yeah, not so easy to have great terminology. I did no change it but added plenty of explanations.

lucasxia01 · 2025-03-14T04:44:25Z

barretenberg/cpp/src/barretenberg/sumcheck/sumcheck_round.hpp


+        // In the AVM, the trace is more dense at the top and therefore it is worth to split the work over the threads


this section just has a lot of logic thats hard to follow. All of the divisions and unclear names make it hard to parse.

lucasxia01 · 2025-03-14T04:45:31Z

barretenberg/cpp/src/barretenberg/sumcheck/sumcheck_round.hpp


+        // In the AVM, the trace is more dense at the top and therefore it is worth to split the work over the threads
+        // a bit more evenly on the vertical axis. To achieve this, we split the trace into chunks and each thread


this comment could have more detail in terms what you mean by "more evenly on the vertical axis"

jeanmon · 2025-03-14T09:16:30Z

@lucasxia01 I added a significant number of explanations and explained the required properties to be satisfied and added a little proof of why the code satisfies this. I hope this gives enough confidence that the code is correct. it is a bit hard to unit test this. In any case, I think there is enough safeguard that the new code does not affect non-AVM parts in any way.

ledwards2225 · 2025-03-14T21:34:42Z

@jeanmon This is more or less exactly what I had in mind with this issue. @lucasxia01 do you see any reason why this same mechanism isn't applicable for us in the PG context?

fcarreiro · 2025-03-14T22:28:15Z

Personally I think this approach is noticeably better if you don't have uniform density (like in our case, we get many times improvement), and even if you do, it should work at least as good (assuming poor cache locality, which in any case you could probably improve with a bigger thread chunk; but that needs hard data).

lucasxia01

Thanks for the fantastic comments!

lucasxia01 · 2025-03-16T16:06:12Z

@ledwards2225 It's not clear exactly what we do in PG, but I could see it applying similarly. It might not be as effective or easy to implement because of how we structure the trace so the nonzero blocks are all over the place.

🤖 I have created a new Aztec Packages release --- ## [0.82.0](v0.81.0...v0.82.0) (2025-03-21) ### ⚠ BREAKING CHANGES * `AztecNode.findLeavesIndexes` returning block info ([#12890](#12890)) * make `ResolverError::UnnecessaryPub` a hard error (noir-lang/noir#7664) ### Features * `AztecNode.findLeavesIndexes` returning block info ([#12890](#12890)) ([9770e15](9770e15)) * add `EmbeddedCurvePoint::generator()` to return generator point (noir-lang/noir#7754) ([ce84b2d](ce84b2d)) * add AMM support to transaction bot ([#12897](#12897)) ([b2c35cc](b2c35cc)) * add minter role to TestERC20 ([#12889](#12889)) ([716ab4f](716ab4f)), closes [#12887](#12887) [#12882](#12882) * allow `fn` returning `()` without having to write `-> ()` (noir-lang/noir#7717) ([e9526cf](e9526cf)) * avm merkle gadget in vm2 ([#12726](#12726)) ([898d50e](898d50e)) * **avm:** instruction fetching parsing error ([#12804](#12804)) ([09a09d5](09a09d5)) * **bb:** Introduce chunks for univariate computation for the AVM ([#12707](#12707)) ([c912bd6](c912bd6)) * can associate error regex with entry in flake file ([#12785](#12785)) ([da48e38](da48e38)) * capture app stacks for bb usage ([#12383](#12383)) ([293eb9d](293eb9d)) * **docs:** Docs on shared mutable, contract classes, and injecting data ([#12043](#12043)) ([94beffe](94beffe)) * Fallback to blobscan API on blob miss ([#12857](#12857)) ([f2f0d72](f2f0d72)), closes [#12856](#12856) * Fast node_modules caching in CI. ([#12371](#12371)) ([09f50d7](09f50d7)) * generate subrelation-label comment in generated relation hpp ([#12914](#12914)) ([fa2bf95](fa2bf95)) * lets try grinding ([4342491](4342491)) * Montgomery optimisation (partial) ([#12822](#12822)) ([c524339](c524339)) * **noir sync:** Calculate noir hash based on just `noir-repo-ref` and `noir-repo.patch` ([#12861](#12861)) ([fa5991f](fa5991f)) * observe L1 and init proposal counter ([#12681](#12681)) ([1e3d71e](1e3d71e)) * precomputed ClientIVC VKs ([#12126](#12126)) ([65bd276](65bd276)) * reapplying reverted circuits recorder with a fix ([#12919](#12919)) ([bf9a034](bf9a034)) * recording circuit inputs + oracles ([#12148](#12148)) ([5436627](5436627)) * rotate over instances on request fail ([#12270](#12270)) ([7ee2e24](7ee2e24)) * Separate Aztec Node Admin API ([#12790](#12790)) ([ccac9ff](ccac9ff)), closes [#12789](#12789) * slim wallet ([#12803](#12803)) ([3f9f30a](3f9f30a)) * **sol:** setup epoch - sampling without replacement ([#12753](#12753)) ([096f739](096f739)) * **ssa:** Basic control dependent LICM (noir-lang/noir#7660) ([ce84b2d](ce84b2d)) * **ssa:** Dominance frontiers (noir-lang/noir#7692) ([e9526cf](e9526cf)) * **ssa:** Post dominator tree (noir-lang/noir#7595) ([e9526cf](e9526cf)) ### Bug Fixes * add missing inputs to `BlackBoxFuncCall::get_inputs_vec` for EcAdd (noir-lang/noir#7752) ([ce84b2d](ce84b2d)) * add random deployment salt to sandbox ([#12869](#12869)) ([49acd71](49acd71)) * agent reporting jobs completion correctly ([#12835](#12835)) ([18d68f0](18d68f0)) * allow method call after block, if and match (noir-lang/noir#7655) ([e9526cf](e9526cf)) * allow omitting ';' after last block statement if it's an assignment (noir-lang/noir#7718) ([e9526cf](e9526cf)) * allow referring to comptime locals at runtime (noir-lang/noir#7681) ([e9526cf](e9526cf)) * allow renaming a trait when importing it (noir-lang/noir#7688) ([e9526cf](e9526cf)) * avm merkle cpp broke darwin ([#12917](#12917)) ([75c9098](75c9098)) * **bb.js:** remove size metadata from UH proof ([#12775](#12775)) ([5b064bc](5b064bc)) * bench upload logic ([#12800](#12800)) ([52575d8](52575d8)) * bring back matrix on grind. try several things to resolve txe port in use. ([79af7a3](79af7a3)) * CI Tweaks ([#12908](#12908)) ([326767c](326767c)) * ci.sh gh-bench logic ([#12776](#12776)) ([8080f47](8080f47)) * **ci:** Exclude inliner specific reference count tests from Brillig trace report (noir-lang/noir#7761) ([ce84b2d](ce84b2d)) * **ci:** Fail the CI job on a Brillig report failure (noir-lang/noir#7762) ([ce84b2d](ce84b2d)) * clientivc capture benchmarks include authwits ([#12873](#12873)) ([5df2aea](5df2aea)) * config ordering ([#12893](#12893)) ([44af76e](44af76e)) * consensus URL as CLI config ([#12796](#12796)) ([66e8028](66e8028)) * correctly format let followed by comment before unsafe (noir-lang/noir#7659) ([e9526cf](e9526cf)) * cpp ivc bench ([#12815](#12815)) ([95a0ec1](95a0ec1)) * disable gpg signing for noir bootstrap ([#12840](#12840)) ([2a76c8c](2a76c8c)) * Disallow registration of contract classes with no public bytecode ([#12910](#12910)) ([41bf13e](41bf13e)) * doc comments on functions warn unexpectedly (noir-lang/noir#7721) ([e9526cf](e9526cf)) * don't gpg sign the noir patch ([#12912](#12912)) ([43191e0](43191e0)) * Don't log config ([#12876](#12876)) ([545b4e0](545b4e0)) * Experiment with `DecrementRc` (noir-lang/noir#7629) ([66a62d0](66a62d0)) * Fix prover node publisher for multi-proofs ([#12924](#12924)) ([42733a6](42733a6)) * Fix stdout testing when running test programs (noir-lang/noir#7741) ([66a62d0](66a62d0)) * Fixes regressed LSP and adds a test to sanity check. ([#12915](#12915)) ([30f9087](30f9087)) * generate ivc benchmarks ([#12811](#12811)) ([3ee32a9](3ee32a9)) * handle predicate value reduction in array get also for the databus (noir-lang/noir#7730) ([66a62d0](66a62d0)) * hotfix git user config ([d58efbb](d58efbb)) * kind on sepolia + balance consolidation + sepolia values service indices ([#12473](#12473)) ([000d515](000d515)) * misleading test ([#12877](#12877)) ([dc8ab31](dc8ab31)) * nightly versioning ([#12872](#12872)) ([7586192](7586192)) * oracles handlers ([#12864](#12864)) ([4bae397](4bae397)) * pass bot salt ([#12923](#12923)) ([7aa0b87](7aa0b87)) * post-checkout hook installation to include `$@` ([#12781](#12781)) ([323c958](323c958)) * redact sensitive fields in logs ([#12913](#12913)) ([31feb0b](31feb0b)) * redo "fix: make vk metadata actual witnesses" ([#12535](#12535)) ([392abc2](392abc2)) * **redo:** "fix: Switch to `noir-repo` context for cache content hashing" ([#12825](#12825)) ([5a98d20](5a98d20)) * Remove hack to register contract class directly on node ([#12795](#12795)) ([eff2501](eff2501)), closes [#10007](#10007) * Removed logged config object in L1 Tx Utils ([#12901](#12901)) ([5d871f8](5d871f8)) * Restart proving job if epoch content changes ([#12655](#12655)) ([613994f](613994f)) * retries starting txe ([3fcc601](3fcc601)) * revert "Switch to `noir-repo` context for cache content hashing" ([#12824](#12824)) ([37ccc38](37ccc38)) * Run aztec-up tests as ubuntu not root. ([#12875](#12875)) ([41cf5a4](41cf5a4)) * run nightly as bot ([#12833](#12833)) ([3799330](3799330)) * Some basic re-org handling ([#12812](#12812)) ([558e315](558e315)) * source_refname ([#12827](#12827)) ([b4a0b71](b4a0b71)) * **ssa:** don't check Brillig calls for coverage if they don't return anything (e.g. println) (noir-lang/noir#7644) ([e9526cf](e9526cf)) * Stop blockstream on world-state sync error ([#12854](#12854)) ([810ee2d](810ee2d)) * Switch to `noir-repo` context for cache content hashing ([#12784](#12784)) ([6214c8c](6214c8c)) * update join split test hash ([#12798](#12798)) ([2fd47f6](2fd47f6)) * Use schnorr lib with reduced sig.e ([#12844](#12844)) ([d03d61c](d03d61c)) * validators verify tx proofs ([#12939](#12939)) ([c02132d](c02132d)) * wrong printing of line comment in quoted (noir-lang/noir#7694) ([e9526cf](e9526cf)) * yolo annotated tag on nightlies. only run kind smoke test. ([8cdfb9a](8cdfb9a)) * yolo debug info ([6681c8f](6681c8f)) * yolo fix ([aec93df](aec93df)) * yolo mark flakey discv5 timeout test ([b45a994](b45a994)) * yolo trying to fix txe port ([ea553d2](ea553d2)) * yolo wait for docker to be up on aztec-up tests. ([8dcc1b5](8dcc1b5)) ### Miscellaneous * add `mapi`/`for_eachi` functions (noir-lang/noir#7705) ([66a62d0](66a62d0)) * add `shared` module within `noirc_frontend` (noir-lang/noir#7746) ([ce84b2d](ce84b2d)) * add acir_test for bb.js classes ([#12892](#12892)) ([ffe1f4f](ffe1f4f)) * add cargo deny advisory (noir-lang/noir#7691) ([e9526cf](e9526cf)) * Add comment on verifyHistoricBlock ([#12933](#12933)) ([cb978b5](cb978b5)) * add extra docs lint (noir-lang/noir#7742) ([66a62d0](66a62d0)) * Add GITHUB_TOKEN for downloading prost_prebuilt to acvm.js build (noir-lang/noir#7745) ([ce84b2d](ce84b2d)) * add kind diagnostics during startup ([#12805](#12805)) ([450f9bb](450f9bb)) * add lambda calculus test (noir-lang/noir#7646) ([e9526cf](e9526cf)) * add more test suites to CI (noir-lang/noir#7757) ([ce84b2d](ce84b2d)) * add regression tests for PR [#7570](#7570) from lambda interpreter test (noir-lang/noir#7638) ([e9526cf](e9526cf)) * add support for caching `node_modules` within a nested repository ([#12862](#12862)) ([16232c8](16232c8)) * add tests for trait renaming in imports (noir-lang/noir#7631) ([e9526cf](e9526cf)) * add timeouts to CI (noir-lang/noir#7725) ([e9526cf](e9526cf)) * add trailing slash to link on docs homepage (noir-lang/noir#7682) ([e9526cf](e9526cf)) * add workflow to publish rustdoc to github pages (noir-lang/noir#7687) ([e9526cf](e9526cf)) * address recurring typo in docs (noir-lang/noir#7656) ([e9526cf](e9526cf)) * allow individual service data map size configuration ([#12853](#12853)) ([b9e6a19](b9e6a19)), closes [#12831](#12831) * **artifact_cli:** Print circuit output to stdout (noir-lang/noir#7696) ([e9526cf](e9526cf)) * **avm:** Constrain pc_size_in_bits column and rename ([#12899](#12899)) ([d69901f](d69901f)) * **avm:** make contract db returns optional ([#12867](#12867)) ([cf2beb2](cf2beb2)) * **avm:** vm2 lazy bytecode loading ([#12847](#12847)) ([233ca3e](233ca3e)) * **avm:** vm2 recursive execution ([#12842](#12842)) ([04981e2](04981e2)) * avoid syncing recursive proofs between repositories ([#12769](#12769)) ([7259b92](7259b92)) * begin splitting out note discovery ([#12819](#12819)) ([75d545a](75d545a)) * begin the introduction of log type id and standard log layouts ([#12823](#12823)) ([e86a258](e86a258)) * bump bb (noir-lang/noir#7726) ([66a62d0](66a62d0)) * bump external pinned commits (noir-lang/noir#7667) ([e9526cf](e9526cf)) * bump external pinned commits (noir-lang/noir#7728) ([e9526cf](e9526cf)) * bump JS dependencies (noir-lang/noir#7669) ([e9526cf](e9526cf)) * bump node to v22.18.3 (noir-lang/noir#7668) ([e9526cf](e9526cf)) * bump wasm-pack to 0.13.1 (noir-lang/noir#7675) ([e9526cf](e9526cf)) * check test program execution success output (noir-lang/noir#7713) ([e9526cf](e9526cf)) * **ci:** enforce rustdoc lints (noir-lang/noir#7738) ([66a62d0](66a62d0)) * **ci:** Exclude enum tests from Brillig reports (noir-lang/noir#7661) ([e9526cf](e9526cf)) * **ci:** private-kernel-inner timeout bump (noir-lang/noir#7732) ([66a62d0](66a62d0)) * Cleanup and re-specify sequencer config in RC1 ([#12898](#12898)) ([13aa4f5](13aa4f5)) * comment out kind tests ([#12793](#12793)) ([39986e5](39986e5)) * delete honk programs from `test_programs` (noir-lang/noir#7727) ([66a62d0](66a62d0)) * **docs:** Avoid colliding filenames (noir-lang/noir#7771) ([ce84b2d](ce84b2d)) * **docs:** Brillig opcodes (noir-lang/noir#7722) ([e9526cf](e9526cf)) * **docs:** Document BlackBoxFuncCall enum (noir-lang/noir#7702) ([e9526cf](e9526cf)) * **docs:** Extend stable documentation versions to build to cover multiple `beta.n` releases (noir-lang/noir#7685) ([e9526cf](e9526cf)) * **docs:** Landing page jargonless intro (noir-lang/noir#7649) ([66a62d0](66a62d0)) * **docs:** Minor fixes on local documentation development workflows (noir-lang/noir#7684) ([e9526cf](e9526cf)) * **docs:** More acir docs (noir-lang/noir#7731) ([66a62d0](66a62d0)) * **docs:** update bb commands to match the new version (noir-lang/noir#7677) ([e9526cf](e9526cf)) * **docs:** update profiler usage docs ([#12782](#12782)) ([3b2fccb](3b2fccb)) * don't switch out bb.js install in noir repo ([#12771](#12771)) ([79c3c62](79c3c62)) * easier way to test monormophization errors (noir-lang/noir#7679) ([e9526cf](e9526cf)) * enable wtr debug logging ([#12848](#12848)) ([0fc5528](0fc5528)) * encapsulate `Index` within `LocalModuleId` (noir-lang/noir#7719) ([e9526cf](e9526cf)) * Faster tx production on masternet ([#12878](#12878)) ([17d0301](17d0301)) * fix archiver.test.ts ([#12907](#12907)) ([fd7adb1](fd7adb1)) * fix gemini ([#12602](#12602)) ([3eb9e9a](3eb9e9a)) * fix rustdoc issues (noir-lang/noir#7712) ([e9526cf](e9526cf)) * Fix rustdocs error (noir-lang/noir#7750) ([ce84b2d](ce84b2d)) * fixing timeouts (noir-lang/noir#7666) ([e9526cf](e9526cf)) * **frontend:** Regression test for creating a mutable reference to an array element (noir-lang/noir#7699) ([e9526cf](e9526cf)) * hide Ident fields (noir-lang/noir#7709) ([e9526cf](e9526cf)) * Increase bracket_depth to 2048 ([#12801](#12801)) ([2e9a7a3](2e9a7a3)) * make `ResolverError::UnnecessaryPub` a hard error (noir-lang/noir#7664) ([e9526cf](e9526cf)) * migrate to use new flat eslint config file (noir-lang/noir#7676) ([e9526cf](e9526cf)) * minor `CommitmentsDB` cleanup ([#12817](#12817)) ([ff8a199](ff8a199)) * more descriptive SSA tests (noir-lang/noir#7697) ([e9526cf](e9526cf)) * more logging in kind setup ([#12852](#12852)) ([9824336](9824336)) * note getter internals ([#12809](#12809)) ([3a89a24](3a89a24)) * nuking MemoryArchiveStore ([#12826](#12826)) ([c5405d8](c5405d8)) * once again skip the uniswap tests ([#12859](#12859)) ([4f37dec](4f37dec)) * **p2p:** add tx queue to prevent ddos attacks ([#12603](#12603)) ([a96a908](a96a908)) * pr linking job for merge queue ([#12546](#12546)) ([a0cee9f](a0cee9f)) * pull most logic from `get_all_contracts` up out of the `CrateDefMap` (noir-lang/noir#7715) ([e9526cf](e9526cf)) * pull out pure functions from interpreter (noir-lang/noir#7755) ([ce84b2d](ce84b2d)) * push users towards nargo in tutorial (noir-lang/noir#7736) ([ce84b2d](ce84b2d)) * re-enable nightly tests ([#12673](#12673)) ([d2f8f18](d2f8f18)), closes [#12107](#12107) * Refactor entrypoints ([#12868](#12868)) ([5f48055](5f48055)) * **refactor:** Move resolved error structures out of acir crate (noir-lang/noir#7734) ([66a62d0](66a62d0)) * remove `examples` from testing of noir ([#12768](#12768)) ([cd6a3af](cd6a3af)) * remove `noir-lang/ec` dependency ([#12507](#12507)) ([181d6e0](181d6e0)) * remove bun from docs in favour of yarn (noir-lang/noir#7756) ([ce84b2d](ce84b2d)) * Remove magic number from AVM bytecode ([#12900](#12900)) ([4d96fc0](4d96fc0)) * Remove seemingly unused compose files ([#12794](#12794)) ([1a75797](1a75797)) * remove some unnecessary mod.nr files ([#12797](#12797)) ([99fc705](99fc705)) * remove some unused HIR code (noir-lang/noir#7643) ([e9526cf](e9526cf)) * remove ultraplonk tests (noir-lang/noir#7680) ([e9526cf](e9526cf)) * remove unconditional sleep ([#12814](#12814)) ([e61c376](e61c376)) * remove whyle, use native while ([#12820](#12820)) ([889bc48](889bc48)) * rename note discovery to message discovery ([#12755](#12755)) ([ae4f935](ae4f935)) * replace relative paths to noir-protocol-circuits ([070ad0a](070ad0a)) * replace relative paths to noir-protocol-circuits ([f62614d](f62614d)) * replace relative paths to noir-protocol-circuits ([742162c](742162c)) * replace relative paths to noir-protocol-circuits ([801244c](801244c)) * replace relative paths to noir-protocol-circuits ([c39d88b](c39d88b)) * replace relative paths to noir-protocol-circuits ([2cb1a37](2cb1a37)) * replace relative paths to noir-protocol-circuits ([a42309b](a42309b)) * Resolve various rustdoc warnings (noir-lang/noir#7724) ([e9526cf](e9526cf)) * run `noir_wasm` over `test_programs` (noir-lang/noir#7765) ([ce84b2d](ce84b2d)) * separated multi map and fixed indexeddb map ([#12896](#12896)) ([05af647](05af647)) * Set max txs per block ([#12837](#12837)) ([74476ff](74476ff)) * simplify note getter oracle ([#12807](#12807)) ([2fc01a3](2fc01a3)) * **ssa:** Do not print entire functions in underconstrained values check trace (noir-lang/noir#7665) ([e9526cf](e9526cf)) * update docusaurus config to correct trailing slash issue (noir-lang/noir#7720) ([e9526cf](e9526cf)) * update examples to use UltraHonk (noir-lang/noir#7653) ([e9526cf](e9526cf)) * Update README.md to add trailing docs `/` (noir-lang/noir#7689) ([e9526cf](e9526cf)) * Update references to GH issues to reflect recent changes ([#12722](#12722)) ([100d31f](100d31f)) * update yarn version to 4.5.2 (noir-lang/noir#7678) ([e9526cf](e9526cf)) * use rollup.ts more consistenty ([#12863](#12863)) ([bd9f3ce](bd9f3ce)) * vars for --network ignition-testnet ([#12886](#12886)) ([6efc8ce](6efc8ce)) ### Documentation * Add sponsoredFPC ([#12485](#12485)) ([87bb6da](87bb6da)) * additions ([#12551](#12551)) ([c9ea1cd](c9ea1cd)) --- This PR was generated with [Release Please](https://github.com/googleapis/release-please). See [documentation](https://github.com/googleapis/release-please#release-please).

jeanmon marked this pull request as ready for review March 13, 2025 09:27

jeanmon requested review from IlyasRidhuan, fcarreiro and Maddiaa0 as code owners March 13, 2025 09:27

jeanmon requested review from lucasxia01 and ledwards2225 and removed request for IlyasRidhuan and Maddiaa0 March 13, 2025 09:27

jeanmon force-pushed the jm/sumcheck-chunks branch from 871a40e to f9f88dd Compare March 13, 2025 10:15

jeanmon marked this pull request as draft March 13, 2025 10:44

fcarreiro reviewed Mar 13, 2025

View reviewed changes

barretenberg/cpp/src/barretenberg/sumcheck/sumcheck_round.hpp Show resolved Hide resolved

jeanmon force-pushed the jm/sumcheck-chunks branch from 8aba26d to f006b50 Compare March 13, 2025 16:38

fcarreiro reviewed Mar 13, 2025

View reviewed changes

barretenberg/cpp/src/barretenberg/vm2/generated/flavor.hpp Outdated Show resolved Hide resolved

barretenberg/cpp/src/barretenberg/vm2/generated/flavor.hpp Outdated Show resolved Hide resolved

jeanmon marked this pull request as ready for review March 13, 2025 17:15

fcarreiro changed the title ~~chore: Introduce chunks for univariate computation for the AVM~~ feat(bb): Introduce chunks for univariate computation for the AVM Mar 13, 2025

lucasxia01 reviewed Mar 14, 2025

View reviewed changes

jeanmon and others added 8 commits March 14, 2025 07:30

Introduce chunks for univariate computation for the AVM

bde40e1

Removing dependency of sumcheck to avm

13cd912

rollback last change

febe4bc

use concepts

f46b226

fix concept

eea7f1c

Addressing review feedback

b3f12ac

Further adressing of review feedback

c8073a2

Addressing review feedback

56df29c

jeanmon force-pushed the jm/sumcheck-chunks branch from 4dc4d59 to 56df29c Compare March 14, 2025 09:11

jeanmon requested a review from lucasxia01 March 14, 2025 09:16

lucasxia01 approved these changes Mar 16, 2025

View reviewed changes

jeanmon merged commit c912bd6 into master Mar 16, 2025
8 checks passed

jeanmon deleted the jm/sumcheck-chunks branch March 16, 2025 16:43

AztecBot mentioned this pull request Mar 16, 2025

chore(master): release 0.82.0 #12770

Merged

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

feat(bb): Introduce chunks for univariate computation for the AVM #12707

feat(bb): Introduce chunks for univariate computation for the AVM #12707

jeanmon commented Mar 13, 2025 •

edited by fcarreiro

Loading

fcarreiro commented Mar 13, 2025

jeanmon commented Mar 13, 2025

fcarreiro commented Mar 13, 2025

fcarreiro commented Mar 13, 2025

fcarreiro commented Mar 13, 2025

lucasxia01 left a comment

lucasxia01 Mar 14, 2025

jeanmon Mar 14, 2025

lucasxia01 Mar 14, 2025

jeanmon Mar 14, 2025

lucasxia01 Mar 14, 2025

lucasxia01 Mar 14, 2025

jeanmon commented Mar 14, 2025

ledwards2225 commented Mar 14, 2025

fcarreiro commented Mar 14, 2025

lucasxia01 left a comment

lucasxia01 commented Mar 16, 2025


		// In the AVM, the trace is more dense at the top and therefore it is worth to split the work over the threads


		// In the AVM, the trace is more dense at the top and therefore it is worth to split the work over the threads
		// a bit more evenly on the vertical axis. To achieve this, we split the trace into chunks and each thread

feat(bb): Introduce chunks for univariate computation for the AVM #12707

feat(bb): Introduce chunks for univariate computation for the AVM #12707

Conversation

jeanmon commented Mar 13, 2025 • edited by fcarreiro Loading

fcarreiro commented Mar 13, 2025

jeanmon commented Mar 13, 2025

fcarreiro commented Mar 13, 2025

fcarreiro commented Mar 13, 2025

fcarreiro commented Mar 13, 2025

lucasxia01 left a comment

Choose a reason for hiding this comment

lucasxia01 Mar 14, 2025

Choose a reason for hiding this comment

jeanmon Mar 14, 2025

Choose a reason for hiding this comment

lucasxia01 Mar 14, 2025

Choose a reason for hiding this comment

jeanmon Mar 14, 2025

Choose a reason for hiding this comment

lucasxia01 Mar 14, 2025

Choose a reason for hiding this comment

lucasxia01 Mar 14, 2025

Choose a reason for hiding this comment

jeanmon commented Mar 14, 2025

ledwards2225 commented Mar 14, 2025

fcarreiro commented Mar 14, 2025

lucasxia01 left a comment

Choose a reason for hiding this comment

lucasxia01 commented Mar 16, 2025

jeanmon commented Mar 13, 2025 •

edited by fcarreiro

Loading