[Mesa-dev] Initial radeonsi NIR tess support

[Timothy Arceri]

On 12/12/17 21:12, Nicolai Hähnle wrote:
> On 11.12.2017 03:43, Timothy Arceri wrote:
>> There are is still more work to be done but this gets us passing
>> a bunch of piglit tests without any lock ups (although a few
>> test do hang for a while).
> 
> Good stuff! I've sent a bunch of comments, although many are either easy 
> to fix or (in the case of the comments on refactored/moved code) should 
> probably be in separate commits.
> 
> Patches 1, 2, 4-12, 14 (with a trivial change), 15-19:
> 
> Reviewed-by: Nicolai Hähnle <nicolai.haehnle at amd.com>
> 
> ... with the caveat that some of the patches will probably be 
> (trivially) affected by the following:
> 
> Considering patches 20 & 21: on second thought, I believe the type 
> argument in the ac_shader_abi callbacks should just be removed. It's 
> true that radeonsi uses it, but it might as well just use ctx->ac.i32 in 
> the relevant places.
> 
> Furthermore, *if* a type argument is required after all, it should 
> probably be
> 
>    get_def_type(ctx, &instr->dest.ssa)
> 
> instead of a nir2llvm of the variable type. But again, it's probably 
> better to just remove the type argument and treat the loads as typeless.

I tried using get_def_type(ctx, &instr->dest.ssa) but I was getting some 
really strange behavior as if it was causing some kind of buffer 
overflow somewhere, so gave up and switched back. I'll try using 
ctx->ac.i32 and see if I have better luck. Thanks for the review :)

> 
> Cheers,
> Nicolai
> 
> 
> 
>>
>> _______________________________________________
>> mesa-dev mailing list
>> mesa-dev at lists.freedesktop.org
>> https://lists.freedesktop.org/mailman/listinfo/mesa-dev
>>
> 
>