Opening GPU Architectures (ETH Zürich, Digital Design and Computer Architecture, Lecture 19)
GPU Architectures (ETH Zürich, Digital Design and Computer Architecture, Lecture 19)
- Length
- 1 h 52 min
- Level
- Undergraduate
- Recorded
- ETH Zürich2025
Chapter 1 of 15: Before the lecture
Mutlu derives GPU warps from SIMD and vector machines, grouping scalar SPMD threads at a shared program counter.
GPU architecture grows out of SIMD array and vector processing in Lecture 19 of Mutlu's Digital Design and Computer Architecture course at ETH Zürich. Mutlu reviews SIMD execution, then shows how memory stride and bank count decide whether accesses cause bank conflicts. Automatic loop vectorization, Amdahl's law on the serial fraction, MMX packed arithmetic, and SIMD in Cerebras and FPGA machine-learning accelerators follow. The second half separates programming models from execution models, grouping SPMD threads into warps and contrasting this SIMT scheme with vector processors. Basic pipelining, memory organization and instruction set architecture concepts are assumed.
GPU architecture grows out of SIMD array and vector processing in Lecture 19 of Mutlu's Digital Design and Computer Architecture course at ETH Zürich. Mutlu reviews SIMD execution, then shows how memory stride and bank count decide whether accesses cause bank conflicts. Automatic loop vectorization, Amdahl's law on the serial fraction, MMX packed arithmetic, and SIMD in Cerebras and FPGA machine-learning accelerators follow. The second half separates programming models from execution models, grouping SPMD threads into warps and contrasting this SIMT scheme with vector processors. Basic pipelining, memory organization and instruction set architecture concepts are assumed.

This lecture
Lecture 22 of 35 of Digital Design and Computer Architecture.
Choose a node to see how it connects to this lecture. What comes before is on the leftat the top. What comes after is on the rightat the bottom. The ring is what it teaches. People and places are above. Other lectures and books are below.
Press a chosen node again to open it.
Your comment is kept while you sign up.Sign in
NVIDIA Tesla: A Unified Graphics and Computing Architecture
MMX Technology Extension to the Intel Architecture
Very High-Speed Computing Systems
Pseudo-Randomly Interleaved Memory
Validity of the single processor approach to achieving large scale computing capabilities