Awesome-CUDA-Books: A Curated Reading List for GPU Programmers

A GitHub repository maintained by alternbits collects what its author claims is the most comprehensive public bibliography of CUDA programming books, organized by skill level and focus area. The list spans foundational texts like Sanders and Kandrot’s CUDA by Example (2010) and Kirk and Hwu’s Programming Massively Parallel Processors, through practical guides covering multi-GPU work and library ecosystems (cuBLAS, cuFFT, Thrust), into Python-oriented resources built around Numba and CuPy.

The curator emphasizes recent releases, flagging a wave of 2024–2026 titles covering CUDA 12.x and 13, Tensor Cores, Nsight debugging, and C++20/26 interop. Several of these are self-published or specialized works that surface in search results but vary in depth. The repo includes contribution guidelines favoring books from 2018 onward or enduring classics with substantial code examples.

The practical caveat embedded in the list matters more than any single recommendation: CUDA’s API and hardware capabilities shift quickly enough that printed books should be treated as companions to NVIDIA’s official CUDA C++ Programming Guide rather than standalone references.