Special fp value get wrong result for fmod

Description

fmod with special input get wrong result.

Steps to Reproduce

dxc -Tvs_6_0

RWBuffer<float4> results;

[shader("vertex")]
void main() {
      // Explicit float
      // 0.0, -0.0, NaN, -NaN
      results[0] = float4(fmod(0.0f, 1.0f),
                                        fmod(-0.0f, 1.0f),
                         fmod(5.5f, 0.0f),
                         fmod(-5.5f, 0.0f));
}

Actual Behavior

0.0, 0.0, 0.0, 0.0 is stored into result[0].

Environment

• DXC version : 1.8
• Host Operating System : Windows

Reproduce the shader with dxc -Tvs_6_0 and inspect the handling of fmod for the listed special inputs. The issue names no source file or test path, so first locate the compiler entry point and relevant floating-point tests; done means the cases no longer produce the reported all-zero result.