[PATCH vkd3d v7 4/6] vkd3d-shader/hlsl: Perform a copy propagation pass.

[Zebediah Figura]

On 11/19/21 10:12 AM, Matteo Bruni wrote:
> On Fri, Nov 19, 2021 at 4:55 PM Giovanni Mascellani
> <gmascellani at codeweavers.com> wrote:
>>
>> Hi,
>>
>> On 18/11/21 18:28, Matteo Bruni wrote:
>>>> +struct copy_propagation_value
>>>> +{
>>>> +    struct hlsl_ir_node *node;
>>>> +    unsigned int component;
>>>> +};
>>>> +
>>>> +struct copy_propagation_variable
>>>> +{
>>>> +    struct rb_entry entry;
>>>> +    struct hlsl_ir_var *var;
>>>> +    struct copy_propagation_value *values;
>>>> +};
>>>
>>> I still haven't gotten warm to these names. What about
>>> copy_propagation_definition (or just copy_propagation_def)?
>>
>> I don't understand why it's appropriate for this structure, but fine.
> 
> It's in the sense of "definition of the current value", more or less
> like the "def" in use-def chains.

I don't really like this, though, because both copy_propagation_variable 
and copy_propagation_value are "def"s, but we need some way of 
disambiguating what's being defined.

> 
>>>> +static struct hlsl_ir_node *copy_propagation_find_replacement(struct copy_propagation_variable *variable,
>>>> +        unsigned int offset, unsigned int count, unsigned int *swizzle)
>>>
>>> I think I prefer my original suggestion of _get_ (or something else)
>>> rather than _find_; the point being that this function isn't really
>>> finding anything.
>>
>> To me it gets something as much as it finds something. I wanted to avoid
>> the word "get" because it seems overly generic, but it's true that
>> "find" is not that much more specific. Ok.
> 
> Throwing a couple more options: generate, compute.
> 

I'm... not sure why this function isn't finding something? It's not 
really generating anything (except a swizzle, I guess); rather it's 
looking up the stored node that can serve as a replacement.

Although I like "get" too.