Analysis of performance of the Cado-NFS implementation (M/F)

Job Description

Missions

**Software Comparison:** Compare the Cado-NFS software base with the good ideas contained in competing projects (msieve, lasieve, bwxl, yafu) in order to get the best out of them.

**Scaling Analysis:** Analyze how the code can be used to tackle larger problem sizes, taking the RSA-1024 size as a major milestone.

**Integration and Sharing:** Integrate all achievements into the public Cado-NFS codebase on the GitLab platform.

**Refinement of Computational Estimations:** Refine the computational cost estimation for RSA-1024 so that it can be carried out exclusively with Cado-NFS.

Activity

**Analysis and Optimization of the Sieving Step:** Analyze, reimplement, and determine how to best optimize the undocumented algorithms present in the `lasieve` codebase for the sieving stage.

**Establishment of Performance Regression Tests:** Establish performance regression tests for Cado-NFS and discuss how to present the results regarding performance non-regression.

**Exploration of GPU Potential:** Contribute to ongoing exploratory work regarding the potential benefits of GPUs for all stages of Cado-NFS (notably the sieving step, polynomial selection, and sparse linear algebra).

**HPC Optimization (Optional):** Depending on personal interests, further adapt the linear algebra step to an HPC environment to maximize its performance.

Your Profil

Skills

**Mathematics and Algorithmics:** Possess solid mathematical and algorithmic knowledge to be comfortable with the NFS algorithm.

**C++ Development:** Essential ability to understand and write C++20 code.

**Assembly Reading:** Capability to understand routines written in x86-64 assembly when necessary, with the potential use of dedicated tools such as Ghidra.

**Versioning Tools:** Be perfectly comfortable with the Git tool and the GitLab platform.

**Complementary Skills:** Good knowledge of HPC environments and GPUs would be a plus, depending on the explored areas.

Your Work Environment

**Team and Location:** The work will be carried out within the Caramba team at LORIA in Vandœuvre-lès-Nancy.

**Supervision:** The project is supervised by Emmanuel Thomé.

**Duration and Timeline:** The contract has a duration of 11 months, starting in October 2026 (10/2026).

**Software Environment:** The project centers on Cado-NFS, a substantial academic software written mainly in C++, consisting of approximately 300,000 lines of code.

**Hardware Resources:** Access to computing resources of various scales, notably including the Grid5000 platform to provide a testing framework for HPC or specific GPU generations.

Constraints and risks

**On-site Presence:** The work is expected to be done primarily in person, though telecommuting is possible within the standard frameworks of the CNRS.

**Complexity and Readability of Legacy Code:** Examining the `lasieve` codebase is made difficult by the nature of the code itself and the fact that it contains algorithms described nowhere else.

**Risk of Performance Side Effects:** Cado-NFS is a complex software structure where a seemingly harmless or locally positive modification can have a detrimental or even disastrous side effect in other contexts.

**Technical Difficulty:** Porting to GPU presents varying difficulties depending on the stage, with the sieving step likely being the most difficult to port to this architecture.

Compensation and benefits

Compensation

Roughly 2,000 EUR/month

Annual leave and RTT

44 jours

Remote Working practice and compensation

Pratique et indemnisation du TT

Transport

Prise en charge à 75% du coût et forfait mobilité durable jusqu’à 300€

About the offer

About the CNRS

The CNRS is a major player in fundamental research on a global scale. The CNRS is the only French organization active in all scientific fields. Its unique position as a multi-specialist allows it to bring together different disciplines to address the most important challenges of the contemporary world, in connection with the actors of change.

CNRS

The research professions