On Tue, 18 Oct 2011 17:05:28 -0700, Kenneth Graunke <kenn...@whitecape.org> wrote: > We definitely want CACHE_NEW_WM_PROG, not CACHE_NEW_VS_PROG. > > NOTE: This is a candidate for the 7.11 branch. > > Signed-off-by: Kenneth Graunke <kenn...@whitecape.org>
This series is Reviewed-by: Eric Anholt <e...@anholt.net>
> ---
> src/mesa/drivers/dri/i965/gen6_wm_state.c | 4 ++--
> 1 files changed, 2 insertions(+), 2 deletions(-)
>
> diff --git a/src/mesa/drivers/dri/i965/gen6_wm_state.c
> b/src/mesa/drivers/dri/i965/gen6_wm_state.c
> index 00fadb9..370516d 100644
> --- a/src/mesa/drivers/dri/i965/gen6_wm_state.c
> +++ b/src/mesa/drivers/dri/i965/gen6_wm_state.c
> @@ -49,7 +49,7 @@ gen6_prepare_wm_push_constants(struct brw_context *brw)
> /* XXX: Should this happen somewhere before to get our state flag set? */
> _mesa_load_state_parameters(ctx, fp->program.Base.Parameters);
>
> - /* CACHE_NEW_VS_PROG */
> + /* CACHE_NEW_WM_PROG */
> if (brw->wm.prog_data->nr_params != 0) {
> float *constants;
> unsigned int i;
> @@ -85,7 +85,7 @@ const struct brw_tracked_state gen6_wm_constants = {
> .mesa = _NEW_PROGRAM_CONSTANTS,
> .brw = (BRW_NEW_BATCH |
> BRW_NEW_FRAGMENT_PROGRAM),
> - .cache = CACHE_NEW_VS_PROG,
> + .cache = CACHE_NEW_WM_PROG,
> },
> .prepare = gen6_prepare_wm_push_constants,
> };
I think this first bug could have resulted in real problems if, for
example, proj_attrib_mask but nothing else fragment program related
changed between draw calls (since we'd end up with different computation
in the shader, and thus different sets of live uniform components to be
uploaded). Texture swizzles might do it, too. I'm having trouble
coming up with a likely behavior in a non-testcase app to trigger it,
but I don't want to think about the issue any more, so cherry-picking
seems like a good idea. :)
pgp9obRkCJ5uD.pgp
Description: PGP signature
_______________________________________________ mesa-dev mailing list mesa-dev@lists.freedesktop.org http://lists.freedesktop.org/mailman/listinfo/mesa-dev
