[Piglit] [PATCH 2/3] arb_shader_precision: add framework for calculating tolerances for complex functions

Ilia Mirkin imirkin at alum.mit.edu
Thu Feb 26 14:22:16 PST 2015


On Thu, Feb 26, 2015 at 5:09 PM, Micah Fedke
<micah.fedke at collabora.co.uk> wrote:
>>> +def _mod_ref(args):
>>
>>
>> What is 'args' precisely here? Let's say I have
>>
>> vec4 a, b;
>> mod(a, b)
>>
>> Is args an array of 2 things, which in turn might themselves be arrays
>> or not? If so, I'd much rather _mod_ref take 2 args explicitly. (It'd
>> also mean that your _to_bigfloat_list thing won't work properly.)
>>
>>> +    bfargs = _to_bigfloat_list(args)
>>
>>
>> Can you not fix this up in the caller? Or is the caller that
>> _synthesize_bla in the other file? [In which case this is fine... you
>> could be fancy and use a decorator, but you don't have to.]
>
>
> The matrix multiplication reference functions perform some logic that is
> specific to themselves that relies on the original data types.  This would
> look out of place in the caller.

Bottom line is... I have no clue what the arguments to this (or any
other) function are. Until I do, I can't review this change. Perhaps
I'm just being slow here and it's obvious to others, dunno.

  -ilia


More information about the Piglit mailing list