Strategic Objectives
• Master the architecture of high-performance SIMD engines.
• Implement the cutting-edge RISC-V 'V' vector extension.
• Optimize data-parallel workloads for maximum hardware throughput.
• Bridge the gap between hardware logic and parallel software execution.
The Core Challenge
Traditional scalar processing hits a wall when facing the massive data demands of AI, graphics, and scientific simulation.
01
The Evolution of Throughput
02
Foundations of SIMD
03
Flynn’s Taxonomy and Beyond
04
The RISC-V Ecosystem
05
Anatomy of the 'V' Extension
06
Parallelism at the Data Level
07
The Vector Register File
08
Vector Lane Architecture
09
Instruction Level Parallelism vs. Vectors
10
Memory Access Patterns
11
Chaining and Pipelines
12
Masking and Predication
13
Fixed-Point and Integer Vectors
14
Floating-Point Vector Performance
15
Compiler Vectorization
16
Roofline Modeling
17
The Role of Microcode
18
Advanced Vector Extensions
19
Graphics and GPGPU
20
Security in Vector Units
21
The Future of Parallelism
Explore Research Ecosystem
Available eBook Editions
Arabic
English
French
German
Italian
Japanese
Korean
Portuguese
Spanish
