Feat/ggh lut 2601 #33
Conversation
* add cuda * compile worked * GpuDCRTPoly tests worked * update build.rs * Add gpu_dcrt_poly matrix * add multi pipelines optimizations * fix compile errors * Revert "fix compile errors" This reverts commit d3bc55a. * Revert "add multi pipelines optimizations" This reverts commit 2ea09ec. * add tests * gpu matrix tests passed * Add gpu samplers * test_gpu_dcrtpoly~ and test_gpu_matrix~ tests passed * all gpu tests passed * use rayon more * cuda optimizations * Add test_arithmetic_circuit_operations_gpu_ggh15 * minor optimizations * optimize matrix mul and add benchmark codes * remove unused codes * fix clone of GpuDCRTPoly * optimize mul kernel * Add available_gpu_ids * CI: disable OpenMP in CMake configuration; remove memory logging for optimization * fix compile error * Feat/gpu without pipeline (#31) * Revert "fix compile errors" This reverts commit d3bc55a. * Revert "add multi pipelines optimizations" This reverts commit 2ea09ec. * add tests * gpu matrix tests passed * Add gpu samplers * test_gpu_dcrtpoly~ and test_gpu_matrix~ tests passed * all gpu tests passed * use rayon more * cuda optimizations * Add test_arithmetic_circuit_operations_gpu_ggh15 * minor optimizations * optimize matrix mul and add benchmark codes * remove unused codes * fix clone of GpuDCRTPoly * optimize mul kernel * Add available_gpu_ids * CI: disable OpenMP in CMake configuration; remove memory logging for optimization * fix compile error * make all tests sequential * fix compile errors
![chatgpt-codex-connector[bot]](https://avatars.githubusercontent.com/in/1144995?s=60&v=4){kind=small}
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: f47dc8fd9d
ℹ️ About Codex in GitHub
Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you
- Open a pull request for review
- Mark a draft as ready
- Comment "@codex review".
If Codex has suggestions, it will comment; otherwise it will react with 👍.
Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".
| // Input is assumed to be a constant polynomial; use its constant coefficient as the key. | ||
| let const_coeff = input | ||
| .coeffs() | ||
| .first() | ||
| .cloned() |
There was a problem hiding this comment.
The new plaintext LUT evaluator derives the lookup key from only the first coefficient and always returns a constant polynomial. If any circuit applies public_lookup to a non-constant input (e.g., packed plaintext with multiple coefficients), the remaining coefficients are silently ignored and the lookup result becomes incorrect. This is a behavioral regression from the earlier “whole polynomial” lookup and will corrupt results unless all LUT inputs are guaranteed constant; consider validating constness or applying the LUT per coefficient instead.
Useful? React with 👍 / 👎.
There was a problem hiding this comment.
We assume all LUT inputs are constant.
Can you add debug_assert lines to ensure that all inputs to the LUT is constant?
…r improved logging consistency
…chunk size, and modify the Clone impl of DCRTPoly
No description provided.