Skip to content

[WIP] Prover Optimizations - #17

Closed
Mychecksdead wants to merge 7 commits into
devfrom
mert-min2x
Closed

[WIP] Prover Optimizations#17
Mychecksdead wants to merge 7 commits into
devfrom
mert-min2x

Conversation

@Mychecksdead

@Mychecksdead Mychecksdead commented Aug 16, 2026

Copy link
Copy Markdown

NEON Goldilocks, FFT/twiddle caches, batched LDE + parallel Merkle + Metal GPU Poseidon2 (macOS-only, kernel ships as reviewable source), restructured quotient eval with filter-fused batched gate hooks, chunked batch inversion, powers-table opening evaluation, faster witness generation

devin-ai-integration Bot and others added 7 commits August 15, 2026 11:57
…ne..4842f236 (#2,#7,#22,#26,#34,#38,#47,#50,#105,#143)

Co-Authored-By: mert@lighter.xyz <mert.koksal.026@gmail.com>
…#217,#226,#262,#290,#300 (vendor/plonky2 delta 4842f236..46044020, bench excluded)

Co-Authored-By: mert@lighter.xyz <mert.koksal.026@gmail.com>
…u,#634,#636u,#646,#652 (vendor delta 296e2fcc..d801a926, bench excluded)

Co-Authored-By: mert@lighter.xyz <mert.koksal.026@gmail.com>
Co-Authored-By: mert@lighter.xyz <mert.koksal.026@gmail.com>
A serialization round trip always reads leaves back row-major, so a
column-store tree compared unequal to its own deserialization even
though every leaf holds the same values (breaking build_block_circuit's
round-trip validation).

Co-Authored-By: mert@lighter.xyz <mert.koksal.026@gmail.com>
…erence, avoid num_wires underflow at bits=0

Co-Authored-By: mert@lighter.xyz <mert.koksal.026@gmail.com>
Co-Authored-By: mert@lighter.xyz <mert.koksal.026@gmail.com>
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

2 participants