Massive mali non-fragment gpu clock cost when game rendered on opengl platform using mali-G76

Hi there,

Our game is developed using UE 4.24 and the shaders are compiled using dxc. We found that when the game is on opengl platform and running on Huawei P30(Mali-G76), the mali non-fragment gpu clock and gpu clock active counters have a much larger value than that of our game running on vulkan platform. All the scene/rendering settings are the same besides the shader platform. We tried to simplify the vertex shaders so that they only calculate the positions, but it doesn't help. And we are pretty sure it is caused by the tiling stage since the non-fragment gpu clock decrease to zero if we hide all the objects in the scene. We didn't see such problems on Huawei P40(another subtype of G76 i believe) or other qualcomm equipped phones.

I wondered if anyone could help us figure out the cause.

Thank you so much!