False positive in shader float control validation

Environment:

• OS: Arch Linux (rolling)
• GPU: Radeon RX6800 on RADV
• SDK or header version if building from repo: Built VVL d1bb20e47fe9f22fd0aadec44e899ab58f3f2458
• Options enabled (synchronization, best practices, etc.): Just validation.

Describe the Issue

When declaring a module with only FP16 denorm mode or FP64 denorm mode, a false positive is triggered.

Expected behavior

For 32_BIT_ONLY independence, it is expected that validation is only triggered if FP16 and FP64 modes are explicitly declared differently. If 64-bit floats are not used in a module, it should not trigger false positive.

(If denorm mode is not declared for FP64, it should inherit the explicit mode from FP16 I think, but that might not be required in the specification.)

Valid Usage ID
06289

Additional context

463ac:err:vkd3d_debug_messenger_callback: "Validation Error: [ VUID-RuntimeSpirv-denormBehaviorIndependence-06289 ] Object 0: handle = 0x8f1c31000000006d, name = 60c9660a21aafad8, type = VK_OBJECT_TYPE_SHADER_MODULE; | MessageID = 0xb4fec51a | Shader uses different denormals execution modes for 16 bit floats but roundingModeIndependence is VK_SHADER_FLOAT_CONTROLS_INDEPENDENCE_32_BIT_ONLY The Vulkan spec states: If denormBehaviorIndependence is VK_SHADER_FLOAT_CONTROLS_INDEPENDENCE_32_BIT_ONLY, then the entry point must use the same denormal"...