ISCTE SINTRA
GALLERY
SIMULATION
Details
Metrics
Settings
Visibility
Appearance
BENCHMARK
PROFILING
REPORTING
SETTINGS
Simulation
Global
ABOUT
A GPU-Accelerated Finite Element Analysis Platform
~
- Governed by
- The Laplace Equation
- ∇²u = 0
- Solved using
- Conjugate Gradient Method
- with Jacobi Preconditioning
- Discretized by
- Quad-8 Isoparametric Elements
- 8 nodes • 3×3 Gauss-Legendre
~
The Development Team
ISCTE-IUL • Sintra Campus
Lead Architect
António Cruz
Chief Quadrilateral Officer
GPU Whisperer
Bruno Santos
CUDA Kernel Choreographer
Mesh Wrangler
Pedro Miranda
Geometry Acquisition Specialist
Convergence Guru
Ricardo Kayseller
Residual Reduction Consultant
~
Starring
The Solver Implementations
The Reliable One
CPU Baseline
NumPy / SciPy
The Speed Demon
GPU CuPy
20-30× speedup
The Compiler
Numba JIT
Production without GPU
The Kernel Master
Numba CUDA
Fine-grained control
The Multitasker
CPU Threaded
ThreadPoolExecutor
The Core Collector
CPU Multiprocess
Pool initializer pattern
~
Visual Effects
Three.js
Reality Renderer
WebGL
Pixel Pusher
Socket.IO
Progress Streamer
ECharts
Chart Conjurer
~
Mesh Catering
Serving Fine Geometries Since 2025
- Backward-Facing Step
- Elbow 90°
- S-Bend
- T-Junction
- Venturi
- Y-Shaped
~
Backend Services
FastAPI
Request Handler
Uvicorn
ASGI Conductor
Docker
Container Warden
NVIDIA CUDA
Parallel Universe
~
Soundtrack
Songs to Iterate By
- "Conjugate My Heart"
- The Gradient Descenders
- "Still Iterating"
- 15,000 Iterations & The Residuals
- "Laplace Your Bets"
- Elliptic PDE Anthem
- "Sparse but Not Forgotten"
- A Ballad for Large Matrices
~
Special Thanks
SciPy
Sparse matrix therapist
CuPy
Making GPUs useful since 2015
NumPy
The foundation of everything
Numba
JIT happens
~
No nodes were displaced beyond their elastic limits during this simulation.
All boundary conditions were applied humanely and in accordance
with the Dirichlet-Robin Convention of 1850.
The Conjugate Gradient method converged on the first try.
Just kidding. It took 12,847 iterations.
~