Unexpected Float usage in trivial … | Apple Developer Forums

Unexpected Float usage in trivial fragment shader

Is there an explanation why "Capture GPU Frame" reports this fragment shader uses Float instead of Half instructions.

Code Block Metalfragment half4 texturedQuadFragment/*ConstantColorNoParameters*/()
{
    METAL_CONST half4 color={0.5h,0.25h,0.75h,1.0h};
    return color;
}

Code Block Metal
fragment half4 texturedQuadFragment/ConstantColorNoParameters/()
{
METAL_CONST half4 color={0.5h,0.25h,0.75h,1.0h};
return color;
}

The Pipeline Statistics, Remarks pane reports:

High Float-to-Half Ratio
A relatively high amount of floating-point operations use 32-bit float data types
Instead, use 16-bit half data types to increase performance.
Ratio is 100.0% - TextureQuadShader.metal:texturedQuadFragmentMultiline

Total Number of Instructions 4

Performance, Counters, Fragment Shader reports:

FS ALU Instructions, 4,216,440
FS ALU Float Instructions, 98.5%
FS ALU Half Instructions, 1.51%

Assembly listing doesn't indicate Float usage:

Code Block define <4 x half> @texturedQuadFragment() local_unnamed_addr #0 {
  ret <4 x half> <half 0xH3800, half 0xH3400, half 0xH3A00, half 0xH3C00>
}

Code Block
define <4 x half> @texturedQuadFragment() local_unnamed_addr #0 {

ret <4 x half> <half 0xH3800, half 0xH3400, half 0xH3A00, half 0xH3C00>

}

Fragment shader is rendering to CAMetalLayer Display Drawable, 1920x1080 BGRA8Unorm

Apple TV 4K
tvOS 13.4.6 (17L570)
Xcode 11.5 (11E608c)
MacOS 10.15.5 (19F101)