[Mesa-dev] [PATCH 3/8] i965: Faking the ETC2 compression on Gen < 8 GPUs using two miptrees.

[Eleni Maria Stea]

On 1/22/19 9:25 PM, Nanley Chery wrote:
[...]
> 
> The performance difference should be negligible if the function is
> declared static inline in the intel_mipmap_tree.h header. The compiler
> should include the body of function (which should be small) and avoid
> the overhead of a function call.

[...]

> 
> Firstly, it's not information that's generally useful for most
> intel_mipmap_tree objects. Having too much of such state makes debugging
> and reading the struct definition more difficult.
> 
> Secondly, it adds to the amount of state-dependent variables I have to
> keep in mind when looking at the code. I have to start asking, when is
> needs_fake_etc initialized? Is needs_fake_etc ever modified later? I'm
> already familiar with the other variables needs_fake_etc can be computed
> by: the gen, the miptree format, and the shadow_mt. I hope that helps.
> 
> -Nanley
> 

Ok, I understand, I am going to change the code to use an inline
function then.

Thank you very much,
Eleni