Task List for VK_KHR_fragment_shading_rate

[Tobski]

The Vulkan 1.2.158 spec release includes the VK_KHR_fragment_shading_rate extension. This extension adds the ability to change the rate at which fragments are shaded. Rather than the usual single fragment invocation for each pixel covered by a primitive, multiple pixels can be shaded by a single fragment shader invocation.

The task list for the VK_KHR_fragment_shading_rate extension release is:

• Vulkan specification VK_KHR_fragment_shading_rate
• SPIRV specification SPV_KHR_fragment_shading_rate
• SPIRV headers (Added SPV_KHR_fragment_shading_rate SPIRV-Headers#172)
• SPIRV tools (Support SPV_KHR_fragment_shading_rate SPIRV-Tools#3943)
• glslang (Add GL_EXT_fragment_shading_rate glslang#2428)
• GLSL extension (Add GLSL_EXT_fragment_shading_rate GLSL#137)
• Validation layer (Add validation for VK_KHR_fragment_shading_rate Vulkan-ValidationLayers#2266)
• Conformance tests
• DirectX Shader Compiler ([spirv] Add VK_KHR_variable_rate_fragment_shading SPIR-V support microsoft/DirectXShaderCompiler#3664)

As each component is made public, the task will be checked off. When all tasks have been completed this issue will be closed and the extension will be fully released.

[pdaniell-nv]

Support for NVIDIA Turing and Ampere products can be found in our 457.00 (Windows) and 455.26.02 (Linux) drivers here: developer.nvidia.com/vulkan-driver.

[SaschaWillems]

Already posted this at the accompanying validation layer issue:

Is there a reason for VkPhysicalDeviceFragmentShadingRateKHR to have an sType member? That structure type is only used in a get function, but with it having an sType member needing initialization (to silence validation), it makes up from some odd code.

Take this for example:

std::vector<VkPhysicalDeviceFragmentShadingRateKHR> fragment_shading_rates{};
uint32_t fragment_shading_rate_count = 0;
vkGetPhysicalDeviceFragmentShadingRatesKHR(device, &fragment_shading_rate_count, nullptr);
if (fragment_shading_rate_count > 0)
{
	fragment_shading_rates.resize(fragment_shading_rate_count);
	vkGetPhysicalDeviceFragmentShadingRatesKHR(device, &fragment_shading_rate_count, fragment_shading_rates.data());
}

This will trigger validation layer message for all entries in the fragment_shading_rates vector. So to silence these and adhere to the valid usage, one would have to set the sType for each entry in that vector. This looks a bit odd and counter-intuitive.

[Tobski]

@SaschaWillems I've moved your question over to a dedicated issue, and I'll post some comments there too.

[marty-johnson59]

Doing some year-end house-cleaning on open public issues... @Tobski I see CTS isn't checked here. Is that still needed - and if so, is there an issue/owner for that task? Thanks,

Marty

[Tobski]

@marty-johnson59 CTS is on track to be updated as the tests are merged internally, but they've not made it to a public release yet. Once they are on GitHub we'll check this off.

[Tobski]

With microsoft/DirectXShaderCompiler#3664 merged finally, this can now be closed.