[Mesa-dev] [PATCH 1/7] softpipe: Move the faces array from view to filter_args

[Emil Velikov]

On 11 September 2015 at 19:07, Krzesimir Nowak <krzesimir at kinvolk.io> wrote:
> With that, sp_sampler_view instances are not abused anymore as a local
> storage, so we can later make them constant.
> ---
>  src/gallium/drivers/softpipe/sp_tex_sample.c | 36 +++++++++++++++++-----------
>  src/gallium/drivers/softpipe/sp_tex_sample.h |  4 +---
>  2 files changed, 23 insertions(+), 17 deletions(-)
>
> diff --git a/src/gallium/drivers/softpipe/sp_tex_sample.c b/src/gallium/drivers/softpipe/sp_tex_sample.c
> index 489cae7..d5a7ed6 100644
> --- a/src/gallium/drivers/softpipe/sp_tex_sample.c
> +++ b/src/gallium/drivers/softpipe/sp_tex_sample.c

> @@ -3594,11 +3596,16 @@ sp_tgsi_get_samples(struct tgsi_sampler *tgsi_sampler,
>        float cs[TGSI_QUAD_SIZE];
>        float ct[TGSI_QUAD_SIZE];
>        float cp[TGSI_QUAD_SIZE];
> +      float faces[TGSI_QUAD_SIZE];
>
> -      convert_cube(sp_sview, sp_samp, s, t, p, c0, cs, ct, cp);
> +      convert_cube(sp_sview, sp_samp, s, t, p, c0, cs, ct, cp, faces);
>
> +      filt_args.faces = faces;
If I remember it correctly the contents of faces will become invalid
and as we exit the function, thus any attempt to use them (via
filt_args.faces) and things will go crazy.

>        sample_mip(sp_sview, sp_samp, cs, ct, cp, c0, lod, &filt_args, rgba);
>     } else {
> +      static const float zero_faces[TGSI_QUAD_SIZE] = {0.0f, 0.0f, 0.0f, 0.0f};
> +
> +      filt_args.faces = zero_faces;
Here we should be safe due to the static qualifier.

>        sample_mip(sp_sview, sp_samp, s, t, p, c0, lod, &filt_args, rgba);
>     }
>  }

> diff --git a/src/gallium/drivers/softpipe/sp_tex_sample.h b/src/gallium/drivers/softpipe/sp_tex_sample.h
> index 72b4a1a..6743b7e 100644
> --- a/src/gallium/drivers/softpipe/sp_tex_sample.h
> +++ b/src/gallium/drivers/softpipe/sp_tex_sample.h
> @@ -72,6 +72,7 @@ typedef void (*img_filter_func)(struct sp_sampler_view *sp_sview,
>  struct filter_args {
>     enum tgsi_sampler_control control;
>     const int8_t *offset;
> +   const float *faces;
Afaict during calculation of face (in convert_cube) uint type is used.
Won't this cause unnecessary int <> float conversions ?

Cheers,
Emil