Design an AVX-512 Vectorization Proposal for a Hot Simulation Kernel
Overview
What this challenge is about.
Analyze a hot simulation kernel's dependencies, then implement an AVX-512 vectorization proposal using LLVM IR or C++ intrinsics. Earn a verifiable certificate.
The scenario
The company builds finite-element structural-simulation software that industrial customers rely on to certify load-bearing designs, so its numerical kernels run for hours on customer-owned Intel Xeon clusters. A single inner loop now dominates that runtime, and customers have asked for faster turnaround without any change to the certified numerical results.
The Brief
What you'll do, and what you'll demonstrate.
Vectorize a 38-percent-hot inner numerical loop for AVX-512 to achieve at least a 3x kernel speedup while proving the rewrite is legal and produces bit-for-bit identical results to the scalar baseline.
Earning criteria — what you'll demonstrate
- Perform dataflow and dependency analysis to prove a loop can be legally vectorized
- Implement AVX-512 vectorization either as an LLVM intermediate-representation pass or via hand-generated intrinsics
- Design a benchmarking methodology that produces speedup figures with confidence intervals
- Verify bit-exact numerical equivalence between scalar and vectorized code paths
- Write a technical design proposal that defends one engineering approach against a credible alternative
Program Fit
Where this fits in your program.
Sharpens the same skills your degree expects you to demonstrate.
Aligned coursework coming soon.
Skills
Skills you'll demonstrate.
Each one shows up on your verified credential.
- Vectorization
Apply vectorization to solve real industry problems and demonstrate production-level capability.
- Llvm
Apply llvm to solve real industry problems and demonstrate production-level capability.
- Loop Optimization
Apply loop optimization to solve real industry problems and demonstrate production-level capability.
- Simd
Apply simd to solve real industry problems and demonstrate production-level capability.
- Benchmarking
Apply benchmarking to solve real industry problems and demonstrate production-level capability.
- Dataflow Analysis
Apply dataflow analysis to solve real industry problems and demonstrate production-level capability.
Careers
Career paths this challenge builds toward
Completing this challenge demonstrates skills that transfer directly to these roles:
Compiler / Performance Engineer
Building and proving an AVX-512 vectorization mirrors the daily work of compiler and performance engineers who must accelerate hot code paths without altering program behavior, defending each transformation with rigorous legality and measurement evidence.
This challenge sharpens
- vectorization
- llvm
- loop-optimization
High-Performance Computing Engineer
HPC engineers optimize numerical kernels on SIMD-capable hardware for scientific and engineering simulation, exactly the skill set exercised by vectorizing this structural-simulation loop and benchmarking it to a measured speedup target.
This challenge sharpens
- simd
- benchmarking
- loop-optimization
Systems Software Engineer
Systems engineers reason about data dependencies and memory safety when rewriting low-level loops; the dependency analysis and bit-exact verification here translate directly into shipping correct, fast systems code under strict correctness constraints.
This challenge sharpens
- dataflow-analysis
- vectorization
- benchmarking