search

microsoft / hlsl-specs

HLSL Specifications
MIT License
110 stars 29 forks source link

[0003] {#IND} HLSL 5.1 Deprecation without alternative for HLSL 6.x. #210

Open soufianekhiat opened 3 months ago

soufianekhiat commented 3 months ago

Description On HLSL 5.1. we can have:

float nan_f32()     { return  1.#IND; }

Without HLSL 6.x. alternative: error: invalid suffix '#IND' on floating constant

Do we have plan for: https://microsoft.github.io/hlsl-specs/proposals/0003-numeric-constants.html hlsl::numeric_limits<float>::quiet_NaN() An unelegant alternative would be: float nan_f32() { return 1.0f/0.0f; } Which produce: 0x7FF0000000000000 cf. https://shader-playground.timjones.io/293af62284312601053e69ca30eeda87

Steps to Reproduce // Not working: error: invalid suffix '#IND' on floating constant HLSL 6.0. https://shader-playground.timjones.io/4d91695886406f7a2c0cd8e21027f104 HLSL 6.6. https://shader-playground.timjones.io/3c52ea5b0687d9a72cabd99ff8cc20dd // Working: HLSL 5.1. https://shader-playground.timjones.io/606ded7e9933d6f62dd16e4f1d1464ea

Actual Behavior error: invalid suffix '#IND' on floating constant

Environment DXC 6.6. Windows

damyanp commented 2 months ago

We should verify that proposal 0003 covers this scenario.