About¶
Build steps¶
Revision history¶
Version 0.8, not released yet.
Version 0.7, Released October 4, 2026.
sw.rst: refine English. hw.rst: add section ‘Host & Cluster Interconnects’
Version 0.6, Released September 27, 2026.
opengl.rst: refine section ‘3D Rendering’. hw.rst: add section ‘Framebuffer hierarchy’. hw.rst: move section ‘System Features – Buffers’ to opengl.rst ‘OpenGL Buffers’ and refine it. sw.rst: refine section ‘Unified IR Conversion Flows’.
Version 0.5, Released September 20, 2026.
opengl.rst: add section ‘Mesh Construction’. refine section ‘3D Rendering’. hw.rst: refine section ‘Geometry Units’. Refine English.
Version 0.4, Released September 7, 2026.
opengl.rst: Add section ‘Ray Tracing Pipeline’. opengl.rst: Refine the meshlet section. hw.rst: Add dgpu.gv and igpu.gv. hw-sw-stack.rst: Refine the VSync section. Correct typos.
Version 0.3, Released March 14, 2026.
geo-math.rst: refine ‘section Projection’ and ‘section Mapping data in GPU’. sw.rst: add ‘section Vector Processor’.
Version 0.2, Released February 28, 2026.
Reorganize sections.
Version 0.1, Released February 22, 2026.
Initial version.
Licensing¶
Chen Chung-Shu
Motivation¶
When I began my career as a graphics GPU compiler engineer, I could not find a book that clearly explained the material from a compiler engineer’s perspective. As a result, I decided to write this book and have been adding to it over time as I continue my studies.
Preface¶
Basically, a CPU is a SISD (Single Instruction Single Data) architecture in each core. The multimedia instructions in CPUs are smaller-scale forms of SIMD (Single Instruction Multiple Data), while GPUs are large-scale SIMD processors, capable of coloring millions of image pixels in just a few milliseconds.
Since 2D and 3D graphic processing offers great potential for parallel data processing, GPU hardware typically includes tens of thousands of functional units per chip, as seen in products by NVIDIA and other manufacturers.
This book provides an overview of how 3D animation is created and executed on a CPU–GPU system. It then introduces GPU compilers and hardware features relevant to graphics applications. Finally, it explains how GPUs have taken on computational tasks traditionally handled by CPUs through the GPGPU (general-purpose computing on graphics processing units) concept and related standards.
Website: Basic Theory of 3D Graphics with OpenGL [1].