Project Log
Semester 1: Performance benchmarking of professional COTS (commercial off the shelf) 3D CAD applications such as AutoCAD and media creation applications such as Premier Pro using Nvidia CUDA in standard x86 Compucon systems as well as with the ARM development kit from Nvidia. The ARM kit will need to run under Linux at this stage.
Semester 2: Repeat performance benchmarking of scientific, research applications on ARM, Quadro, Tesla and CUDA. Fine tune the conversion of applications from running under x86 to ARM computing environment.
Fortnightly tasks in Semester 1
- S1F1 Learn about CUDA and how Nvidia has implemented it commercially and academically
- S1F2 Learn about AutoCAD eco-system and how various software applications relate to each other
- S1F3 Start performance benchmarking AutoCAD with Z68, X79 and Q2000
- S1F4 Benchmark CATIA with X79, Q2000, and C2075
- S1F5 Review open space position with respect to what we have achieved
- S1F6 Benchmark a university research application (this requires Mano to facilitate)
- S1F7 Analyze hardware performance for computing, visualisation, rendering and simulation
Quadro 2000
Tesla C2075
Fortnightly tasks in Semester 2
- S2F1 Learn about Semi-global matching algorithm
- S2F2 Learn about CUDA implementation and implementing SGM on CPU first
- S2F3 Continuous the implementation and finalises the structure of SGM
- S2F4 Semi-global matching on CUDA
- S2F5 Optimization of SGM on CUDA
- S2F6 Test the result and collect data
- S2F7 Final seminar & Final report preparation