[Intel][DirectX] Failure to write third scalar value from CBuffer

[bob80905]

When targeting DXIL IR, Intel seems to consistently fail to write the third scalar value from a CBuffer containing 3 scalars into an output buffer containing a struct of 3 scalar members:

HLSL:

cbuffer CBScalars : register(b0) {
  float16_t a1;
  uint16_t a2;
  int16_t a3;
}

struct Scalars {
  float16_t a1;
  uint16_t a2;
  int16_t a3;
};

RWStructuredBuffer<Scalars> Out : register(u1);

[numthreads(1,1,1)]
void main() {
  Out[0].a1 = a1;
  Out[0].a2 = a2;
  Out[0].a3 = a3;
}

Setup:

---
Shaders:
  - Stage: Compute
    Entry: main
    DispatchSize: [1, 1, 1]
Buffers:
  - Name: CBScalars
    Format: Hex16
    Data: [ 0x3c00, 0x0001, 0xffff ]
  - Name: Out
    Format: Hex16
    Stride: 6
    FillSize: 6
DescriptorSets:
  - Resources:
    - Name: CBScalars
      Kind: ConstantBuffer
      DirectXBinding:
        Register: 0
        Space: 0
      VulkanBinding:
        Binding: 0
    - Name: Out
      Kind: RWStructuredBuffer
      DirectXBinding:
        Register: 1
        Space: 0
      VulkanBinding:
        Binding: 1

# RUN: split-file %s %t
# RUN: %dxc_target -fvk-use-dx-layout -enable-16bit-types -T cs_6_5 -Fo %t.o %t/source.hlsl
# RUN: %offloader %t/pipeline.yaml %t.o | FileCheck %s
...

We expect Data: [ 0x3C00, 0x1, 0xFFFF ]
But the actual output is: Data: [ 0x3C00, 0x1, 0x0 ]