Название: The History of the GPU - Eras and Environment
Автор: Jon Peddie
Издательство: Springer
Год: 2022
Страниц: 355
Язык: английский
Формат: pdf (true)
Размер: 13.4 MB

This is the second book in a three-part series that traces the development of the GPU, which is defined as a single chip with an integrated transform and lighting (T&L) capability. This feature previously was found in workstations as a stand-alone chip that only performed geometry functions. Enabled by Moore’s law, the first era of GPUs began in the late 1990s.

Silicon Graphics (SGI) introduced T&L first in 1996 with the Nintendo 64 chipset with integrated T&L but didn’t follow through. ArtX developed a chipset with integrated T&L but didn’t bring it to market until November 1999.

The need to integrate the transform and lighting functions in the graphics controller was well understood and strongly desired by dozens of companies. Nvidia was the first to produce a PC consumer level single chip with T&L in October 1999. All in all, fifteen companies came close, they had designs and experience, but one thing or another got in their way to prevent them succeeding.

All the forces and technology were converging; the GPU was ready to emerge. Several of the companies involved did produce an integrated GPU, but not until early 2000.

This is the account of those companies, the GPU and the environment needed to support it. The GPU has become ubiquitous and can be found in every platform that involves a computer and a user interface.

An application programming interface (API) is the software link between a processor such as a GPU and the computer’s programs, the applications, and the operating system. It is a form of a software interface. It can also be considered a translator, where high-level code (software) is converted to low-level machine code. An API hides the internal details of how a GPU works, exposing only those parts a programmer would find helpful or essential.

GPUs, like several other peripherals, require a software program called a driver. An API is a translator that reads instructions from an application and translates them into instructions native to the GPU. The reverse path is similar; GPU instructions get translated into equivalents that an application can understand. APIs helped the industry progress and even defined graphics in some ways. The driver is a program that runs on the CPU. It contains a compiler that turns the shader code of the program into the native code of the GPU and offers an application program interface (API). The instruction set for a CPU in a PC or workstation is the ×86 standard. Programs written for Android and Apple phones are compiled for the ARM instruction set.

In contrast, there is no standard instruction set for GPUs—that is one of the few unique things about a GPU. There were only standards for a minimal set of unaccelerated 2D functionalities: VGA and VESA BIOS Extensions (VBEs). Those standards apply to the boot text and logo that appears before the drivers of the GPU were loaded; they were used for limited graphics such as Windows in safe mode and ancient DOS games, but the CPU did all the calculations involved in drawing anything. In today’s GPUs, the most used API standards are Direct3D, Metal, and Vulkan.

Contents:
1. Introduction
2. The GPUs’ Functions
3. The Major GPU Eras
4. The First Era of GPUs
5. The GPU Environment—Hardware
6. Application Program Interface (API)
7. The GPU Environment—Software Extensions and Custom Features