v0.4.0

Tachyon: Much Faster Plonky3

See below for more details:

• medium

What's Changed

• feat: enable eigen matrix inversion by @chokobole in #472
• perf: optimize poseidon sbox by @chokobole in #474
• feat: enable pedersen gpu by @chokobole in #473
• build: update bazel version v6.3.0 by @GideokKim in #477
• chore: add .dockerignore by @chokobole in #475
• feat: implement snark verifier poseidon by @chokobole in #476
• refac: prepare shuffle argument by @chokobole in #479
• feat(math): implement packed field extension by @chokobole in #478
• chore: use short version of GitHub url by @chokobole in #481
• docs: update README for better visibility by @TomTaehoonKim in #482
• fix: enable halo2 proof generation by @chokobole in #480
• build: remove halo2 option by @chokobole in #483
• test: optimize benchmark test by @chokobole in #484
• refac(zk): use Evaluator by @chokobole in #485
• docs: fix circuit testing doc by @ashjeong in #487
• build(math): include a different header if gcc version ≤ 10 by @chokobole in #489
• feat(zk): add shuffle argument by @chokobole in #488
• docs(benchmark): update benchmark images by @chokobole in #494
• perf(circom): optimize zkey wtns reading by @chokobole in #490
• docs(circom): add missing change from #490 by @chokobole in #495
• refac: prepare circle stark by @chokobole in #493
• perf/optimize-prime-field-inversion-bernstein-yang-method by @AlekseiVambol in #491
• ci: fix warnings and improve runtime by @batzor in #498
• feat: use pmr vector for custom allocation by @batzor in #486
• feat: impl repetitive execution msm gpu by @GideokKim in #496
• build: bump up sp1 prover to v1.0.1 by @dongchangYoo in #497
• ci: reduce cache storage usage by @batzor in #500
• build: fix cuda build by @GideokKim in #502
• build: support building tachyon based on python 3.10 docker image by @chokobole in #501
• test: optimize openmp test by @chokobole in #503
• feat: support pse halo2 (Part 1) by @chokobole in #499
• refac: revert change to pmr vector by @batzor in #504
• feat: support pse halo2 part2 by @chokobole in #505
• feat: impl two adic fri pcs by @ashjeong in #492
• perf: optimize proof generation by @chokobole in #506
• feat: support icicle api for all the fields by @chokobole in #508
• feat: integrate perfetto for profiling by @batzor in #507
• feat(math): add TwoInv constant by @ashjeong in #510
• perf: optimize proof generation by @chokobole in #509
• build: add openmp flags to goldilocks by @batzor in #513
• chore: fix naming and error messages by @ashjeong in #514
• perf: optimize msm gpu memory usage by @GideokKim in #511
• feat: add BabyBear4 C api by @chokobole in #512
• build: fix cuda build by @batzor in #516
• perf: optimize fft batch by @chokobole in #517
• test(benchmark): add baby bear poseidon2 benchmark by @ashjeong in #519
• perf(math): unset require initialization by @chokobole in #520
• feat: add rust binding for baby bear row major matrix by @chokobole in #518
• feat: add FFTBatch benchmark by @batzor in #515
• perf(math): optimize radix2 domain by @chokobole in #521
• refac(benchmark): refactor benchmark by @chokobole in #522
• perf: optimize packed prime field operations by @chokobole in #523
• perf: optimize fft batch by @batzor in #524
• feat: add rust binding for two adic fri pcs by @chokobole in #525
• fix(base): enable building with clang++-15 by @chokobole in #527
• refac: refactor two adic fri pcs by @chokobole in #526
• test: add mmcs benchmark by @batzor in #528
• feat: enable rust binding for TwoAdicFriPcs::Open() by @chokobole in #529
• perf: optimize poseidon2 pt1 by @ashjeong in #531
• feat: enable rust binding for TwoAdicFriPcs::Verify() by @chokobole in #532
• perf: compute twiddle cache just once per degree by @chokobole in #534
• perf: optimize poseidon2 pt2 by @ashjeong in #536
• test: add fri opening benchmark by @batzor in #533
• test: use appropriate values for shift and log_blowup by @batzor in #537
• Fix spelling issues by @nnsW3 in #538
• fix(benchmark): fix faulty fri benchmark by @ashjeong in #540
• feat: enable using rust allocated ldes by @chokobole in #539
• Update README by @Ocheretovich in #543
• feat: make TwoAdicFriPcs stateless by @chokobole in #541
• docs fix typos by @cratiu222 in #545
• feat(circom): introduce witness generation tool by @batzor in #546
• feat: enable sp1 proof generation by @chokobole in #547
• perf: integrate icicle mmcs by @GideokKim in #544
• perf: reduce redundant zero initializations by @batzor in #548
• perf: add optimization flags by @batzor in #549
• refac: clean up fri protocol & benchmarks by @ashjeong in #551
• chore: update benchmarks by @ashjeong in #550
• feat: add Dockerfile for circom by @batzor in #552
• refac: refactor TwoAdicFri verify by @chokobole in #553
• perf(crypto): optimize TwoAdicFri by @chokobole in #555
• build: release tachyon v0.4.0 by @chokobole in #556

New Contributors

• @AlekseiVambol made their first contribution in #491
• @nnsW3 made their first contribution in #538
• @Ocheretovich made their first contribution in #543
• @cratiu222 made their first contribution in #545

Full Changelog: v0.3.0...v0.4.0

v0.3.1

Groth16 proof becomes very fast

Highlights

Optimizations

• Poseidon hash becomes 2x faster.
• Zkey parsing becomes 10x faster.

What's Changed

• feat: enable eigen matrix inversion by @chokobole in #472
• perf: optimize poseidon sbox by @chokobole in #474
• feat: enable pedersen gpu by @chokobole in #473
• build: update bazel version v6.3.0 by @GideokKim in #477
• chore: add .dockerignore by @chokobole in #475
• feat: implement snark verifier poseidon by @chokobole in #476
• refac: prepare shuffle argument by @chokobole in #479
• feat(math): implement packed field extension by @chokobole in #478
• chore: use short version of GitHub url by @chokobole in #481
• docs: update README for better visibility by @TomTaehoonKim in #482
• fix: enable halo2 proof generation by @chokobole in #480
• build: remove halo2 option by @chokobole in #483
• test: optimize benchmark test by @chokobole in #484
• refac(zk): use Evaluator by @chokobole in #485
• docs: fix circuit testing doc by @ashjeong in #487
• build(math): include a different header if gcc version ≤ 10 by @chokobole in #489
• feat(zk): add shuffle argument by @chokobole in #488
• docs(benchmark): update benchmark images by @chokobole in #494
• perf(circom): optimize zkey wtns reading by @chokobole in #490
• docs(circom): add missing change from #490 by @chokobole in #495

Full Changelog: v0.3.0...v0.3.1

v0.3.0

ICICLE Integration to Tachyon

Highlights

Optimizations

• MSM and FFT is optimized with icicle. See MSM Gpu benchmark and FFT Gpu benchmark.
• Circom proof generation becomes 4x faster using icicle.
• Parsing zkey and wtns files becomes 3x faster.

New Features

• Poseidon2
• BinaryField
• FieldMerkleTreeMMCS
• ChaCha20Rng

What's Changed

• feat: implement poseidon2 by @chokobole in #413
• build: support building with clang on ubuntu by @chokobole in #414
• refac: generalize example circuit tests by @ashjeong in #415
• test: implement fibonacci circuits by @TomTaehoonKim in #279
• feat(circom): enable circom proof generation without compilation by @chokobole in #419
• refac(math): refactor prime fields by @chokobole in #422
• perf: prepare to skip zkey parsing by @chokobole in #425
• docs: add zip to Ubuntu prerequisites by @Insun35 in #427
• refac: fix zero case for inverse and div by @ashjeong in #424
• feat(math): implement binary field initially by @chokobole in #416
• test: add no_zk and verify flag for circom benchmarking by @TomTaehoonKim in #423
• refac(math): move binary fields by @chokobole in #429
• feat: add initial template two adic fri pcs by @chokobole in #428
• refac: improve #424 by @chokobole in #431
• feat(math): add simd int by @chokobole in #432
• feat(math): impl remove zeros for sparse univariate poly by @batzor in #430
• feat: support openmp on mac by @batzor in #434
• feat: implement LogUp scheme by @Insun35 in #420
• feat: implement hash wrappers by @chokobole in #433
• perf(zk): optimize log derivative halo2 lookup by @chokobole in #437
• feat(math): enable poseidon to hash packed prime fields by @chokobole in #435
• refac: apply c type traits to point types by @chokobole in #438
• feat!(math): improve prime field (de)serialization by @chokobole in #440
• feat: add FieldMerkleTree by @chokobole in #436
• docs: update readme by @chokobole in #439
• refac(c/zk): apply c type traits by @chokobole in #445
• feat(crypto): add FieldMerkleTreeMMCS by @chokobole in #442
• feat(math): implement field extension for stark fields by @chokobole in #443
• docs: update circom README.md by @chokobole in #449
• feat(circom): introduce -n flag for specifying number of proof generation by @whoisgautxm in #452
• refac(zk): modularize expressions by @batzor in #446
• feat: impl Radix2DitParallel by @ashjeong in #441
• feat: add extension field merkle tree mmcs by @chokobole in #444
• feat(zk): implement plonky3 challenges by @chokobole in #448
• perf!: use icicle msm by @chokobole in #426
• perf: optimize fft by @GideokKim in #456
• feat: add rust binding for DuplexChallenger by @chokobole in #451
• feat(c/zk): generalize halo2 prover and verifier c API by @chokobole in #458
• feat: impl two adic multiplicative coset by @ashjeong in #457
• docs(halo2): update Halo2 migration guide by @chokobole in #461
• build: prepare circom proof generation with cuda by @chokobole in #462
• feat(zk): implement air constraint system by @batzor in #453
• perf: use msm gpu when generating groth16 proof by @chokobole in #463
• feat: support scroll halo2 1.1 by @chokobole in #464
• fix(halo2)!: remove constraint system serialization from vk.write() by @chokobole in #467
• perf(halo2): apply batch commitment in create_proof() by @chokobole in #465
• perf: apply icicle ntt by @chokobole in #466
• feat: support ChaCha20RNG by @chokobole in #468
• perf: optimize circom proof generation by @chokobole in #469
• perf(circom): optimize zkey parsing by @chokobole in #471
• bulid: release tachyon v0.3.0 by @chokobole in #470

New Contributors

• @batzor made their first contribution in #430
• @whoisgautxm made their first contribution in #452
• @GideokKim made their first contribution in #456

Full Changelog: v0.2.0...v0.3.0

v0.2.0

Tachyon in Kroma Mainnet

See medium for more details.

Highlights

• Becomes 1.5x Faster than Halo2 by 1) reducing the data copy and 2) optimizing FFT. see benchmark.

What's Changed

• docs: run markdown linter by @chokobole in #367
• perf(math): avoid copying in prime field operations by @chokobole in #368
• fix: fix broken build with cuda by @chokobole in #363
• build: fix bazel build configurations for halo2 to be on linux by @TomTaehoonKim in #369
• perf(math): avoid copying twice on bigint operations by @chokobole in #366
• test: enable benchmarking with rapidsnark by @chokobole in #357
• perf(crypto): avoid copying twice when returning a polynomial by @chokobole in #371
• perf: avoid in place operator where ineffective by @chokobole in #372
• test(zk): enable SelectorReplacerTest.Selector test on optimization mode by @chokobole in #374
• refac: organize Dockerfiles into docker directory by @TomTaehoonKim in #373
• ci: add opt option to ci test by @Insun35 in #375
• fix: change compile option maxvx512f to mavx2 for goldilocks by @chokobole in #376
• ci: use setup-python v5 by @Insun35 in #381
• fix: fix gcc-13 compile errors by @ashjeong in #377
• docs: add telegram contact to README.md by @chokobole in #378
• build(math): generate m31 field as non fft friendly field by @chokobole in #385
• docs: update how to build by @chokobole in #380
• refac(zk): refactor circuit polynomial builder by @chokobole in #384
• test: run (I)FFT benchmark under uniform conditions by @chokobole in #382
• perf: avoid copying on semigroup operations by @chokobole in #379
• refac: allow string conversions to accept malformed input by @chokobole in #388
• feat(math): add range check to string conversion methods by @chokobole in #389
• perf: avoid copying on group operations by @chokobole in #383
• perf(math): streamline operations by @ashjeong in #390
• fix(math): add missing changes from #388 by @chokobole in #392
• refac: rename group internal operations by @chokobole in #391
• chore(zk): clean up zk by @Insun35 in #393
• docs: update tachyon vs halo2 benchmarking results by @ashjeong in #395
• build(math): update build settings for goldilocks by @chokobole in #397
• feat(math): add circle point initially by @chokobole in #387
• docs(zk): add item to the list of the Halo2 optimizations by @chokobole in #398
• refac(math): rename filenames by @chokobole in #399
• chore: fix typos by @simlecode in #403
• feat: add packed mersenne31 by @chokobole in #396
• perf(math): add fast reduction for a special case by @chokobole in #404
• build: add build option to disable assembly optimized field by @chokobole in #406
• perf(math): parallelize SwapElements() by @chokobole in #408
• docs: update tachyon vs halo2 benchmarking results by @dongchangYoo in #410
• feat(circom): enable loading witness from json by @chokobole in #407
• build: add intel openmp support by @chokobole in #409
• refac!: abstract lookup scheme by @Insun35 in #400
• feat(math): add baby and koala bear fields by @chokobole in #405
• feat(math): prepare poseidon2 by @chokobole in #411
• build: update tachyon version for releasing v0.2.0 by @dongchangYoo in #417

New Contributors

• @simlecode made their first contribution in #403

Full Changelog: v0.1.0...v0.2.0

Release v0.1.0

Tachyon: A Much Faster and Lighter Halo2

See below for more details:

• medium
• tachyon/zk/plonk/halo2/optimizaitons.md
• vendors/halo2/README.md