Question why sampling in low level cube lod for SSGI fallback ?

if (RAYMARCHINGFALLBACKHIERARCHY_REFLECTION_PROBES & _RayMarchingFallbackHierarchy)
TraceReflectionProbes(posInput, normalData.normalWS, sampleDir, weight, color);


Using reflection probes for SSGI fallback would be very common.
In fact, Unreal 5's lumen's lumen card is almost the same.

However, the value we want to replace now is the indirect diffuse value, not the specular value.

if (s_envLightIdx >= v_envLightIdx)
{
v_envLightListOffset++;
if (IsEnvIndexCubemap(envLightData.envIndex) && totalWeight < 1.0)
{
float3 R = rayDirection;
float weight = 1.0f;
float intersectionDistance = EvaluateLight_EnvIntersection(posInput.positionWS, normalWS, envLightData, envLightData.influenceShapeType, R, weight);

int index = abs(envLightData.envIndex) - 1;
float3 probeResult = SAMPLE_TEXTURECUBE_ARRAY_LOD_ABSTRACT(_EnvCubemapTextures, s_trilinear_clamp_sampler, R, _EnvSliceSize * index, 0).rgb * envLightData.rangeCompressionFactorCompensation;
probeResult = ClampToFloat16Max(probeResult);

UpdateLightingHierarchyWeights(totalWeight, weight);
result += weight * probeResult * envLightData.multiplier;
}
}

SAMPLE_TEXTURECUBE_ARRAY_LOD_ABSTRACT

I don't understand why you set 0 to LOD.

Shouldn't it be the highest level?

 

It uses LOD 0 because LOD 0 is the highest quality. If you go up in LOD level, the resolution will be lower, details will be lost and the result will be inaccurate.