Add data packing builtin functions

[dneto0]

• Add placeholder for unpacking builtin functions
• Expand the description of floating point conversions now that
  we have to handle edge cases for f32 -> f16.
  I put in the behaviour of an NVIDIA GPU for the concerning edge
  case, but added an Issue to review this for portability.

[dneto0]

Partly addresses #988

[dneto0]

HTML preview is at https://gistpreview.github.io/?55aec7912b75ea316e79597eef8a5053

[dneto0]

I think I just saw SwiftShader convert a finite f32 value larger than the largest f16 finite value into a f16 NaN. That's not permitted in the Vulkan spec, but let me verify my findings first.

[dneto0]

Rebased and addressed @dj2's comment.

The main issue is the addition of general language about converting on floating point type value to another floating point type value. The interesting case if f32 to f16 as it introduces cases where finite source numbers should map to either infinite f16 or max normal f16 values. It's not clear what the dividing line is. I've added an "Issue" to verify behaviour and revisit in the future.

[kvark]

Do we have an investigation somewhere showing how these map to shading languages?

[grorg]

Discussed at 2020-11-03 meeting.

[grorg]

Resolved: Accept PR (although @kvark mentioned a small error)

[dneto0]

UGH. I made a mistake claiming these were already present 1-1 in all the backends. I had checked Metal but skimped on HLSL. (!)

Do we have an investigation somewhere showing how these map to shading languages?

I'll add details here.

[dneto0]

• WGSL pack2x16float(v : vec2<f32>) -> u32

  • SPIR-V GLSL.std.450 extended instruction PackHalf2x16
  • MSL: as_type<uint>(packed_half2(v.x, v.y))
  • HLSL: f32tof16(f.x) | (f32tof16(f.y) << 16)

• WGSL: pack4x8snorm(e: vec4<f32>)

  • SPIR-V GLSL.std.450 extended instruction PackSnorm4x8
  • MSL: pack_float_to_snorm4x8(v)
  • HLSL: uint4 temp = round(clamp(v, -1.0f, 1.0f)*127.0); return (temp.x & 0xff) | ((temp.y&0xff) <<8) | ((temp.z&0xff) <<16) | ((temp.w & 0xff) <<24);

• WGSL: pack4x8unorm(e: vec4<f32>)

  • SPIR-V GLSL.std.450 extended instruction PackUnorm4x8
  • MSL: pack_float_to_unorm4x8(v)
  • HLSL: uint4 temp = round(clamp(v, 0.0f, 1.0f)*255.0); return (temp.x & 0xff) | ((temp.y&0xff) <<8) | ((temp.z&0xff) <<16) | ((temp.w & 0xff) <<24);

• WGSL: pack2x16snorm(e: vec4<f32>)

  • SPIR-V GLSL.std.450 extended instruction PackSnorm2x16
  • MSL: pack_float_to_snorm2x16(v)
  • HLSL: uint4 temp = round(clamp(v, -1.0f, 1.0f)*32767.0); return (temp.x & 0xffff) | ((temp.y&0xffff) <<16);

• WGSL: pack2x16unorm(e: vec4<f32>)

  • SPIR-V GLSL.std.450 extended instruction PackUnorm2x16
  • MSL: pack_float_to_unorm2x16(v)
  • HLSL: uint4 temp = round(clamp(v, 0.0f, 1.0f)*65535.0); return (temp.x & 0xffff) | ((temp.y&0xffff) <<16);

• WGSL: pack2x16unorm(e: vec4<f32>)

  • SPIR-V GLSL.std.450 extended instruction PackUnorm2x16
  • MSL: pack_float_to_unorm2x16(v)
  • HLSL: uint4 temp = round(clamp(v, 0.0f, 1.0f)*65535.0); return (temp.x & 0xffff) | ((temp.y&0xffff) <<16);

[dneto0]

I filed #1201 to track the edge case behaviour of conversion from f32 to f16

[grorg]

Discussed at the 2020-11-10 meeting.

Resolved: Accept with @dneto0 's changes.

[dneto0]

Resolved: Accept with @dneto0 's changes.

The "changes" are really implementor's notes. The behaviour is already fully specified in the WGSL spect text in the PR.