Two different linearZ inside Falcor

[zheng95z]

I found that there are two different implementation of linearZ inside Falcor:

Falcor/Source/RenderPasses/GBuffer/GBuffer/GBufferRaster.3d.slang

Lines 128 to 133 in 2e087af

	// Linear z and its derivative
	if (is_valid(gLinearZAndDeriv))
	{
	const float linearZ = vsOut.posH.z * vsOut.posH.w;
	gLinearZAndDeriv[ipos] = float2(linearZ, max(abs(ddx(linearZ)), abs(ddy(linearZ))));
	}

and

Falcor/Source/RenderPasses/GBuffer/GBuffer/GBufferRT.slang

Lines 278 to 292 in 900a50b

	if (is_valid(gLinearZ))
	{
	float4 curPosH = mul(gScene.camera.data.viewProjMatNoJitter, float4(sd.posW, 1.f));
	float curLinearZ = curPosH.w;
	// TODO: Improve computation of derivatives:
	float3 ddxPosW = computeDdxPosW(sd.posW, sd.faceN, invFrameDim);
	float3 ddyPosW = computeDdyPosW(sd.posW, sd.faceN, invFrameDim);
	float4 curPosH_dx = mul(gScene.camera.data.viewProjMatNoJitter, float4(sd.posW + ddxPosW, 1.f));
	float4 curPosH_dy = mul(gScene.camera.data.viewProjMatNoJitter, float4(sd.posW + ddxPosW, 1.f));
	float ddxLinearZ = abs(curPosH_dx.w - curLinearZ);
	float ddyLinearZ = abs(curPosH_dy.w - curLinearZ);
	float dLinearZ = max(ddxLinearZ, ddyLinearZ);
	gLinearZ[pixel] = float2(curLinearZ, dLinearZ);
	}

Although one of them is generated by rasterization and another is by rt, I believe they both should be posH.w since its view space depth.