Merge "Fix SF blur to be as blurry as Gaussian blur" into sc-dev am: e78acd5b

    // Kawase is an approximation of Gaussian, but it behaves differently from it.

    // A radius transformation is required for approximating them, and also to introduce

    // non-integer steps, necessary to smoothly interpolate large radii.

    float tmpRadius = (float)blurRadius / 6.0f;

    float tmpRadius = (float)blurRadius / 2.0f;

    float numberOfPasses = std::min(kMaxPasses, (uint32_t)ceil(tmpRadius));

    float radiusByPasses = tmpRadius / (float)numberOfPasses;