Direct3D 11.3 Functional Specification (2015) | Hacker News Reader

Direct3D 11.3 Functional Specification (2015)

microsoft.github.io

59 pointsatesti21 comments

MS has had a tendency to do things differently for the sake of doing things differently (eg. for instance clip-space range/w from 0..1 instead of -1..1) but I love the candor in this document:

"D3D9 and prior had a terrible Pixel Coordinate System where the origin was the center of the top left pixel on the RenderTarget"

There are precision issues with -1...1. 0...1 is actually the right choice IMO.

I'm currently updating myself with Vulkan (which is mostly based on modern OpenGL). The Vulkan clip range is 0.0 to 1.0.

ARB_clip_control extension allows to use 1..0 range, improving precision: https://developer.nvidia.com/content/depth-precision-visuali...

> MS has had a tendency to do things differently for the sake of doing things differently

It's probably for the sake of "making sure it's hard to port".

E.g., .net in 2008 called "culture" a thing that has been referred to as a "locale" for at least 30 years at that point, with naming conventions and all. I find it extremely hard to believe that they were unaware of it -- yet, the vast majority of .net programmers I've known had no idea what "locale" was, and believed microsoft invented "culture".

It's not necessarily candor. Could just be people in the current team criticizing the decisions of previous team members.

Might be, but this is an official document. You don't usually find these type of statements in such documents, that's why I like it.

This document may have been written in 2015 but it was not published publicly until this week. I think the (2015) label is misleading.

There is nothing really comparable for OpenGL and it shows, really. Awesome work.

sorry... what ? https://www.khronos.org/registry/OpenGL/specs/gl/glspec45.co...

From someone who wrote graphics drivers, D3D's spec is quite a bit better: https://twitter.com/rygorous/status/1115716179413176320

I'm (too) familiar with the document but it isn't nearly as accurate and unambiguous as the DX one.

GL specs are generally quite accurate, now that there are real conformance tests for implementations. They can feel imprecise though, as they tend to leave a bunch more as undefined, whereas d3d tends to be more strict, leaving fewer choices to implementations.

Of course, the D3D11 spec didn't used to be public at all.

The Fahrenheit project should've been a thing. Just declare Direct3D the new standard.

DX has poor documented features. Ask DXVK developers about it.

My biggest issue with it is that what is written in the specs often doesn't hold true when using lousy drivers ;) .

Maybe it's just me, but PDF is harder to navigate/code/copy-paste than a plain HTML page.

I would be curious to know which tools peoples preferring PDF as documentation uses.

Khronos publish GLSL spec and Vulkan Spec in several format, including HTML[0][1], so they definitly don't have anything against HTML/Plain text.

I don't know why OpenGL specs are still only available has PDFs, maybe because they first wrote it using a format that not easily convertible to HTML.

[0] https://www.khronos.org/registry/vulkan/specs/1.0/html/vkspe...

[1] https://www.khronos.org/registry/OpenGL/specs/gl/GLSLangSpec...

There doesn't seem to be anything comparable for Direct3D 12 either, the relevant page [1] in the same site has only scattered documents about it instead of a single coherent one.

[1] https://microsoft.github.io/DirectX-Specs/

This page is incompatible with mobile devices. I think it is causing relayouts on chrome mobile that make it unusable or it is just too long.

It looks to be an incredibly simple, static page. No javascript even. Don't we see a "you shouldn't be using a fancy framework/javascript" article every week or so?

Reply on news.ycombinator.com