This has the notable effect of implementing maps on deferred contexts.
Signed-off-by: Zebediah Figura <z.figura12@gmail.com>
Signed-off-by: Henri Verbeet <hverbeet@codeweavers.com>
Signed-off-by: Alexandre Julliard <julliard@winehq.org>
Somewhat similar to commit 0170bb0e54,
setting invalid render states ends up invalidating invalid state IDs.
Signed-off-by: Henri Verbeet <hverbeet@codeweavers.com>
Signed-off-by: Alexandre Julliard <julliard@winehq.org>
Allow things like wined3d_format_copy_data() to work on them.
Signed-off-by: Zebediah Figura <z.figura12@gmail.com>
Signed-off-by: Henri Verbeet <hverbeet@codeweavers.com>
Signed-off-by: Alexandre Julliard <julliard@winehq.org>
Allocate a fresh query index that does not need to be reset.
Reset query pools at creation and when the render pass is ended.
Signed-off-by: Jan Sikorski <jsikorski@codeweavers.com>
Signed-off-by: Henri Verbeet <hverbeet@codeweavers.com>
Signed-off-by: Alexandre Julliard <julliard@winehq.org>
Not every state ID between 0 and STATE_HIGHEST is valid; for example, there
are holes in the renderstate IDs. While invalidating these works fine,
attempting to subsequently apply them in e.g. context_apply_draw_state() ends
up calling state_undefined() and printing an ERR. This is especially visible
when running the d2d1 tests.
Signed-off-by: Henri Verbeet <hverbeet@codeweavers.com>
Signed-off-by: Alexandre Julliard <julliard@winehq.org>
As in wined3d_buffer_vk_upload_ranges().
Signed-off-by: Zebediah Figura <z.figura12@gmail.com>
Signed-off-by: Henri Verbeet <hverbeet@codeweavers.com>
Signed-off-by: Alexandre Julliard <julliard@winehq.org>
This can't currently happen. However, we'd like to use
wined3d_buffer_get_memory() in wined3d_cs_exec_update_sub_resource(). This would
probably also help in implementing ID3D11DeviceContext1::DiscardResource().
Signed-off-by: Zebediah Figura <z.figura12@gmail.com>
Signed-off-by: Henri Verbeet <hverbeet@codeweavers.com>
Signed-off-by: Alexandre Julliard <julliard@winehq.org>
Layer index and count fields in struct wined3d_view_desc double as the depth range of a 3D texture view.
Signed-off-by: Jan Sikorski <jsikorski@codeweavers.com>
Signed-off-by: Henri Verbeet <hverbeet@codeweavers.com>
Signed-off-by: Alexandre Julliard <julliard@winehq.org>
Instead of the view's level count, which may be different.
Signed-off-by: Henri Verbeet <hverbeet@codeweavers.com>
Signed-off-by: Alexandre Julliard <julliard@winehq.org>
It's already part of the context.
Signed-off-by: Matteo Bruni <mbruni@codeweavers.com>
Signed-off-by: Henri Verbeet <hverbeet@codeweavers.com>
Signed-off-by: Alexandre Julliard <julliard@winehq.org>
This fixes the non-default ARB shader backend, broken since
2ddb6b66a7, although the
fundamental issue was there long before that.
Originally the STATE_VDECL handler did some bookkeeping (basically,
computing what is now the stream info data) that's a prerequisite for
running other state handlers. For this reason a somewhat complicated
dance was put in place, with the dependent handlers returning right
away until the STATE_VDECL handler could prepare everything up and the
STATE_VDECL handler in turn explicitly calling these "downstream"
handlers so that they could do their job. With the commit mentioned
above the state dirty flags are cleared after the corresponding
handlers are executed, which means that the relevant handlers are
skipped twice and some stuff is never set up properly.
Stream info is computed by context_apply_draw_state() before going
through the state handler table for a long time now, getting rid of
this obscure handler interdependency. So let's just get rid of the
skipping altogether.
Signed-off-by: Matteo Bruni <mbruni@codeweavers.com>
Signed-off-by: Henri Verbeet <hverbeet@codeweavers.com>
Signed-off-by: Alexandre Julliard <julliard@winehq.org>
Specifically, tweak and make use of the already existing
WINED3D_CS_OP_RESET_STATE.
Signed-off-by: Zebediah Figura <z.figura12@gmail.com>
Signed-off-by: Henri Verbeet <hverbeet@codeweavers.com>
Signed-off-by: Alexandre Julliard <julliard@winehq.org>
Avoid storing pointers to the CS data inside individual ops, so that we can
reallocate the queue when appending data to a deferred context, or copy the
queue contents to a command list.
Signed-off-by: Zebediah Figura <z.figura12@gmail.com>
Signed-off-by: Henri Verbeet <hverbeet@codeweavers.com>
Signed-off-by: Alexandre Julliard <julliard@winehq.org>
