GPU Programming basics - PhD course

Registro lezioni 2017

---

24 Gennaio 2017 - 3 ore:

Introduction to GPU h/w and CUDA programming

• GPU architecture
• comparison w.r.t CPU architecture
• SIMT
• What is CUDA
• Hello World in CUDA
• CUDA Memory organization: introduction
• Threads organization and setup
• Exercises:
• summing vectors
• matrix multiplication

• Grids, blocks and threads
• thread scheduling
• Threads and divergence

Slides:

• GPU CUDA 1
• GPU CUDA 2 

Examplesi:

• CUDA Hello World (CLion)
• CUDA Hello World (NSight)

---

25 Gennaio 2017 - 3 ore:

Good practices in blocks and threads organization

• Good practices in blocks and threads organization
• 2D CUDA kernels
• 3D CUDA kernels
• CUDA memory hierarchy
• shared, device and constant memory

Efficient CUDA programming

• tiling: introduction
• tiling: how to
• synchronization barriers
• phase-based processing
• border checking

Lucidi:

• GPU CUDA 2
• GPU CUDA 3

Esercizi:

• Modificare il codice di esempio di moltiplicazione tra matrici implementando il tiling

Possibile soluzione:

• -

---

26 Gennaio 2017 - 3 ore:

Performance considerations

• DRAM burst
• memory coalescing

GPU programming techniques

• atomic instructions
• privatization
• example: histogram computation
• Parallel patterns: convolution
• 1D convolution
• how to use constant memory
• tiling
• 2D convolution
• 2D convolution tiling
• Tiling performance evaluation

Lucidi:

• GPU CUDA 4
• GPU CUDA 5 

Esercizi:

• -

Possibile soluzione:

• -

---

27 Gennaio 2017 - 4 ore: lab.

Esercitazione di laboratorio: convoluzione 2D di immagini

• -

Lucidi:

• Description of the exercise

Esercizi:

• Convoluzione 2D di immagini

Possibile soluzione:

• -

---