[Mesa-dev] [PATCH 03/11] i965/gen6: Fix blending state when no color buffer is bound.

[Kenneth Graunke]

On 05/12/2011 09:05 PM, Eric Anholt wrote:
> This is part of fixing fbo-alphatest-nocolor -- a regression in
> 35e8fe5c99b285f348cb8a1bba2931f120f7c0a1 after the initial regression,
> that had us using a garbage BLEND_STATE[0] (in particular, the alpha
> test enable) if no color buffer was bound.
> ---
>   src/mesa/drivers/dri/i965/gen6_cc.c |   12 +++++++++++-
>   1 files changed, 11 insertions(+), 1 deletions(-)
>
> diff --git a/src/mesa/drivers/dri/i965/gen6_cc.c b/src/mesa/drivers/dri/i965/gen6_cc.c
> index 66357f0..2b16d6c 100644
> --- a/src/mesa/drivers/dri/i965/gen6_cc.c
> +++ b/src/mesa/drivers/dri/i965/gen6_cc.c
> @@ -39,8 +39,18 @@ prepare_blend_state(struct brw_context *brw)
>      struct gen6_blend_state *blend;
>      int b;
>      int nr_draw_buffers = ctx->DrawBuffer->_NumColorDrawBuffers;
> -   int size = sizeof(*blend) * nr_draw_buffers;
> -
> +   int size;
> +
> +   /* We need at least one BLEND_STATE written, because we might do
> +    * thread dispatch even if _NumColorDrawBuffers is 0 (for example
> +    * for computed depth or alpha test), which will do an FB write

The above comment mentions depth or alpha...

> +    * with render target 0, which will reference BLEND_STATE[0] for
> +    * alpha test enable.
> +    */
> +   if (nr_draw_buffers == 0&&  ctx->Color.AlphaEnabled)
> +      nr_draw_buffers = 1;

...but this doesn't do anything with depth?  Does it need to?

> +   size = sizeof(*blend) * nr_draw_buffers;
>      blend = brw_state_batch(brw, size, 64,&brw->cc.blend_state_offset);
>
>      memset(blend, 0, size);

I'm not familiar enough with this code to offer a Reviewed-by, sorry.

--Kenneth