Vulkan allows for shuffle_xor, but DX12 doesn't. Why is that?
Why does DX12 not have `shuffle_xor` SIMD intrinsic, unlike Vulkan?
3,468
February 09, 2020 10:59 AM
It's almost certainly because only a limited set of hardware is capable of support the operation. The list of supported Wave Intrinsics introduced with SM6.0 are either all supported or none are supported. Requiring shuffle_xor would likely have meant a significantly larger proportion of hardware would not have been able to claim support for them all.
Breaking wave ops up further into a series of ‘tiers’ may have been possible, but you'd still be left with a significant portion of the hardware install base without support anyway.
Adam Miles - Principal Software Development Engineer - Microsoft Xbox Advanced Technology Group
February 10, 2020 06:52 PM
How should one go about doing matrix transpositions efficiently then, using DX12?
13,459
February 11, 2020 05:59 AM
Direct3D has need of instancing, but we do not. We have plenty of glVertexAttrib calls.
3,468
February 11, 2020 04:20 PM
bzm3r said:
How should one go about doing matrix transpositions efficiently then, using DX12?
It depends how large a matrix we're talking about?
Adam Miles - Principal Software Development Engineer - Microsoft Xbox Advanced Technology Group
This topic is closed to new replies.
Advertisement
Popular Topics
Advertisement
