[v3 resend 3/6] d3dx9: implement fxlc constants (expressions) in effect.

[Paul Gofman]

On 03/15/2016 05:54 AM, Matteo Bruni wrote:
>
> +static float regstore_get_float(struct d3dx_regstore *rs, unsigned int table, unsigned int offset)
> +{
> +    BYTE *p;
> +
> +    p = (BYTE *)rs->tables[table] + table_info[table].value_size * offset;
> +    switch (table_info[table].type)
> +    {
> +        case PRES_VT_FLOAT : return *(float *)p;
> +        case PRES_VT_DOUBLE: return *(double *)p;
> +        case PRES_VT_INT   : return *(int *)p;
> +        case PRES_VT_BOOL  : return *(BOOL *)p ? 1.0f : -0.0f;
> I guess that -0.0f comes from some test?
No, I just saw a lot of constructs in preshader disassembly like "cmp
r1.x, c0.x, 1.0, -0.0", somehow they like to put -0.0 for immediate
constants for such sort of cmp, and was thinking of this when I was
writing it. There is currently no usecase I know for preshader inputting
booleans or ints, so I can't test this. So I think this just need to be
changed to normal 0.0.
>
> +HRESULT d3dx_param_eval_get_shader_parameters(struct d3dx_param_eval *peval, struct d3dx_parameter ***param,
> +        unsigned int *nparam, D3DXCONSTANT_DESC **cdesc)
> +{
> +    *nparam = peval->shader_inputs.ninputs;
> +    *param = peval->shader_inputs.inputs_param;
> +    *cdesc = peval->shader_inputs.inputs;
> +    return D3D_OK;
> +}
> You don't use this function in this patch either. Also it doesn't seem
> useful, it isn't an improvement to just assigning those three pointers
> directly in the caller IMO.
>
I would like to remove this function but then I will need to put the
param_eval structure into shared header file (now there is only a
forward declaration). Is it OK?