Architecture guides

Start with the fundamentals and work toward modern vector and processor architectures.

Whether you are a student, systems programmer or performance engineer, these guides are designed to explain both what a mechanism does and why the architecture needs it.

Start here

• What is instruction-level parallelism?
• What is a superscalar processor?
• What is SIMD?
• What is a vector processor?
• SIMD vs vector processing
• CPU vs GPU execution models

Modern vector architectures

• AVX2 explained
• AVX-512 explained
• ARM NEON explained
• ARM SVE explained
• RISC-V Vector explained
• Vector-length agnostic programming

CPU microarchitecture

• Out-of-order execution
• Register renaming
• Branch prediction
• Reorder buffers
• Execution ports
• Cache hierarchies
• Memory-level parallelism

Performance engineering

• Why compilers fail to vectorize loops
• Understanding compiler vectorization reports
• Loop-carried dependencies
• Memory aliasing
• Alignment and SIMD
• Gather and scatter
• Vector reductions
• When SIMD makes code slower