Pixel Local Storage cannot reduce bandwidth in my app

I use PLS to reduce the reading and writing of a rendertarget(rgba8).But streamline shows that there is no reduction in bandwidth.

UE4 uses two passes to achieve distortion. 'accumulate pass' and 'merge pass'. The first pass draws uv to a texture(rgba8) and the second pass reads this texture.I use PLS to combine two passes into one.But streamline shows that there is no reduction in bandwidth.

In my expectation, this will reduce the bandwidth consumption caused by the reading and writing of a texture.

我用PLS来减少一张rgba8的RT，但是arm的streamline的各项指标都显示没有任何的带宽下降。

在实现扭曲效果时，使用一个累积pass和一个合并pass，累积pass输出uv到一张rgba8纹理上，合并pass读取这张纹理，我期望PLS减少这种纹理的带宽