[Mesa-dev] [PATCH 06/12] glsl: Add lowering pass for ir_triop_vector_insert

Ian Romanick idr at freedesktop.org
Tue May 7 10:25:02 PDT 2013


On 05/07/2013 09:53 AM, Eric Anholt wrote:
> Ian Romanick <idr at freedesktop.org> writes:
>
>> From: Ian Romanick <ian.d.romanick at intel.com>
>>
>> This will eventually replace do_vec_index_to_cond_assign.  This lowering
>> pass is called in all the places where do_vec_index_to_cond_assign or
>> do_vec_index_to_swizzle is called.
>>
>> v2: Use WRITEMASK_* instead of integer literals.  Use a more concise
>> method of generating broadcast_index.  Both suggested by Eric.
>
> I still object to using a vec4 cmp_result.  It's awful for code
> generation.  Just compare in the if statements.

So you're telling me that it's unacceptable to continue generating the 
same code that we currently generate?  ...for a case that happens is how 
many real shaders?

> IFs vs cond assigns I don't care about -- that's something the driver
> should be deciding.




More information about the mesa-dev mailing list