nir: Use !block_ends_in_jump() in a few places rather than open-coding.

Connor introduced this helper recently; we should use it here too.

I had to move the function earlier in the file for it to be available.

Signed-off-by: Kenneth Graunke <kenneth@whitecape.org>
Reviewed-by: Jason Ekstrand <jason.ekstrand@intel.com>

nir/cf: reimplement nir_cf_node_remove() using the new API

This gives us some testing of it. Also, the old nir_cf_node_remove()
wasn't handling phi nodes correctly and was calling cleanup_cf_node()
too late.

Signed-off-by: Connor Abbott <connor.w.abbott@intel.com>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>

nir/cf: add new control modification API's

These will help us do a number of things, including:

- Early return elimination.
- Dead control flow elimination.
- Various optimizations, such as replacing:

if (foo) {
    ...
}
if (!foo) {
    ...
}

with:

if (foo) {
    ...
} else {
    ...
}

Signed-off-by: Connor Abbott <connor.w.abbott@intel.com>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>

nir/cf: use a cursor for inserting control flow

Signed-off-by: Connor Abbott <connor.w.abbott@intel.com>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>

nir/cf: add split_block_cursor()

This is a helper that will be shared between the new control flow
insertion and modification code.

Signed-off-by: Connor Abbott <connor.w.abbott@intel.com>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>

nir/cf: add split_block_before_instr()

Signed-off-by: Connor Abbott <connor.w.abbott@intel.com>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>

nir/cf: add a cursor structure

For now, it allows us to refactor the control flow insertion API's so
that there's a single entrypoint (with some wrappers). More importantly,
it will allow us to reduce the combinatorial explosion in the extract
function. There, we need to specify two points to extract, which may be
at the beginning of a block, the end of a block, or in the middle of a
block. And then there are various wrappers based off of that (before a
control flow node, before a control flow list, etc.). Rather than having
9 different functions, we can have one function and push the actual
logic of determining which variant to use down to the split function,
which will be shared with nir_cf_node_insert().

In the future, we may want to make the instruction insertion API's as
well as the builder use this, but that's a future cleanup.

Signed-off-by: Connor Abbott <connor.w.abbott@intel.com>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>

nir/cf: fix link_blocks() when there are no successors

When we insert a single basic block A into another basic block B, we
will split B into C and D, insert A in the middle, and then splice
together C, A, and D. When we splice together C and A, we need to move
the successors of A into C -- except A has no successors, since it
hasn't been inserted yet. So in move_successors(), we need to handle the
case where the block whose successors are to be moved doesn't have any
successors. Fixing link_blocks() here prevents a segfault and makes it
work correctly.

Signed-off-by: Connor Abbott <connor.w.abbott@intel.com>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>

nir/cf: clean up jumps when cleaning up CF nodes

We may delete a control flow node which contains structured jumps to
other parts of the program. We need to remove the jump as a predecessor,
as well as remove any phi node sources which reference it. Right now,
the same problem exists for blocks that don't end in a jump instruction,
but with the new API it shouldn't be an issue, since blocks that don't
end in a jump must either point to another block in the same extracted
CF list or not point to anything at all.

Signed-off-by: Connor Abbott <connor.w.abbott@intel.com>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>

nir/cf: remove uses of SSA definitions that are being deleted

Unlike calling nir_instr_remove(), calling nir_cf_node_remove() (and
later in the series, the nir_cf_list_delete()) implies that you're
removing instructions that may still have uses, except those
instructions are never executed so any uses will be undefined. When
cleaning up a CF node for deletion, we must clean up any uses of the
deleted instructions by making them point to undef instructions instead.

Signed-off-by: Connor Abbott <connor.w.abbott@intel.com>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>

nir/cf: handle jumps better in stitch_blocks()

In particular, handle the case where the earlier block ends in a jump
and the later block is empty. In that case, we want to preserve the jump
and remove any traces of the later block. Before, we would only hit this
case when removing a control flow node after a jump, which wasn't a
common occurance, but we'll need it to handle inserting a control flow
list which ends in a jump, which should be more common/useful.

Signed-off-by: Connor Abbott <connor.w.abbott@intel.com>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>

nir/cf: handle jumps in split_block_end()

Before, we would only split a block with a jump at the end if we were
inserting something after a block with a jump, which never happened in
practice. But now, we want to use this to extract control flow lists
which may end in a jump, in which case we really need to do the correct
patching up. As a side effect, when removing jumps we now correctly
insert undef phi sources in some corner cases, which can't hurt.

Signed-off-by: Connor Abbott <connor.w.abbott@intel.com>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>

nir/cf: add block_ends_in_jump()

Signed-off-by: Connor Abbott <connor.w.abbott@intel.com>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>

nir/cf: handle phi nodes better in split_block_beginning()

Signed-off-by: Connor Abbott <connor.w.abbott@intel.com>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>

nir/cf: split up and improve nir_handle_remove_jumps()

Before, the process of removing a jump and wiring up the remaining block
correctly was atomic, but with the new control flow modification it's
split into two parts: first, we extract the jump, which creates a new
block with re-wired successors as well as a free-floating jump, and then
we delete the control flow containing the jump, which removes the entry
in the predecessors and any phi node sources. Split up
nir_handle_remove_jumps() to accomodate this, and add the missing
support for removing phi node sources.

Signed-off-by: Connor Abbott <connor.w.abbott@intel.com>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>

nir/cf: add remove_phi_src() helper

Signed-off-by: Connor Abbott <connor.w.abbott@intel.com>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>

nir: add nir_foreach_phi_src_safe()

Signed-off-by: Connor Abbott <connor.w.abbott@intel.com>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>

nir/cf: add insert_phi_undef() helper

Signed-off-by: Connor Abbott <connor.w.abbott@intel.com>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>

nir: move control flow modification to its own file

We want to start reworking and expanding this code, but it'll be a lot
easier to do once we disentangle it from the rest of the stuff in nir.c.
Unfortunately, there are a few unavoidable dependencies in nir.c on
methods we'd rather not expose publicly, since if not used in very
specific situations they can cause Bad Things (tm) to happen. Namely, we
need to do some magical control flow munging when adding/removing jumps.
In the future, we may disallow adding/removing jumps in
nir_instr_insert_*() and nir_instr_remove(), and use separate functions
that are part of the control flow modification code, but for now we
expose them and put them in a separate, private header.

Signed-off-by: Connor Abbott <connor.w.abbott@intel.com>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>

nir: make cleanup_cf_node() not use remove_defs_uses()

cleanup_cf_node() is part of the control flow modification code, which
we're going to split into its own file, but remove_defs_uses() is an
internal function used by nir_instr_remove(). Break the dependency by
making cleanup_cf_node() use nir_instr_remove() instead, which simply
calls remove_defs_uses() and then removes the instruction from the list.
nir_instr_remove() does do extra things for jumps, though, so we avoid
calling it on jumps which matches the previous behavior (this will be
fixed later in the series).

Signed-off-by: Connor Abbott <connor.w.abbott@intel.com>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>

nir: inline block_add_pred() a few places

It was being used to initialize function impls and loops, even though
it's really a control flow modification helper. It's pretty trivial, so
just inline it to avoid the dependency.

Signed-off-by: Connor Abbott <connor.w.abbott@intel.com>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>

nir/validate: check successors/predecessors more carefully

We should be checking almost everything now.

Signed-off-by: Connor Abbott <connor.w.abbott@intel.com>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>

nir: Delete the nir_function_impl::start_block field.

It's simply the first nir_cf_node in the nir_function_impl::body list,
which is easy enough to access - we don't to store a pointer to it
explicitly.  Removing it means we don't need to maintain the pointer
when, say, splitting the start block when modifying control flow.

Thanks to Connor Abbott for suggesting this.

Signed-off-by: Kenneth Graunke <kenneth@whitecape.org>
Reviewed-by: Connor Abbott <cwabbott0@gmail.com>

mesa/formats: only do type and component lookup for uncompressed formats

Only uncompressed formats have a non-void type and actual
components per pixel. Rename _mesa_format_to_type_and_comps
to _mesa_uncompressed_format_to_type_and_comps and require
callers to check if the format is not compressed.

v2. include compressed format cases to avoid gcc warnings (Chad).

Reviewed-by: Chad Versace <chad.versace@intel.com>
Signed-off-by: Nanley Chery <nanley.g.chery@intel.com>

freedreno/a4xx: formats update

Fixes glamor, which wants to use R8 integer textures.

Signed-off-by: Rob Clark <robclark@freedesktop.org>

freedreno: update generated headers

Signed-off-by: Rob Clark <robclark@freedesktop.org>

i965: Always re-emit the pipeline select during invariant state emission

On the older platforms where we don't have logical contexts preserving
state across batches, we emit the invariant state setup on every batch
using the brw_invariant_state atom. This includes the pipeline selection
which is cached with the introduction of

commit 0e0e23ef537c9add672ff322f34e129a07edc55e
Author: Jordan Justen <jordan.l.justen@intel.com>
Date:   Wed Apr 22 11:43:50 2015 -0700

    i965/state: Emit pipeline select when changing pipelines

However, we do not reset the cache between batches on context-less
platforms resulting in us not setting the pipeline selection and can
cause GPU hangs if a media pipelined was loaded in the meantime (e.g.
mixing mplayer/gstreamer using libva and gnome-shell). A simple solution
is to just forcibly re-emit the pipeline select along with the invariant
state and reset the cache at that point.

Reported-and-tested-by: Tomasz C. <tomaszc@o2.pl>
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=91254
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Jordan Justen <jordan.l.justen@intel.com>
Cc: Kenneth Graunke <kenneth@whitecape.org>
Reviewed-by: Jordan Justen <jordan.l.justen@intel.com>
Cc: "10.6 11.0" <mesa-stable@lists.freedesktop.org>

Revert "radeon/winsys: increase the IB size for VM"

This reverts commit 567394112d904096abff1d994ab952f475dfb444.

It regressed performance. It looks like smaller IBs are better, because
the GPU goes idle quicker and there is less waiting for buffers and fences.

Cc: 11.0 <mesa-stable@lists.freedesktop.org>

nv50: fix 2d engine blits for 64- and 128-bit formats

This fixes bin/ext_framebuffer_multisample-formats all_samples

Signed-off-by: Ilia Mirkin <imirkin@alum.mit.edu>
Cc: "11.0" <mesa-stable@lists.freedesktop.org>

nv50: account for the int RT0 rule for alpha-to-one/cov

Same as commit 1af0641db but for nvc0. If an integer texture is
bound to RT0, don't do alpha-to-one or alpha-to-coverage.

Signed-off-by: Ilia Mirkin <imirkin@alum.mit.edu>
Cc: "11.0" <mesa-stable@lists.freedesktop.org>

mesa/arb_gpu_shader_fp64: add support for glGetUniformdv

This was missed when I did fp64, I've sent a piglit test to cover
the case as well.

Reviewed-by: Timothy Arceri <t_arceri@yahoo.com.au>
Cc: "11.0" <mesa-stable@lists.freedesktop.org>
Signed-off-by: Dave Airlie <airlied@redhat.com>

nv50,nvc0: disable depth bounds test on blit

Signed-off-by: Ilia Mirkin <imirkin@alum.mit.edu>
Cc: "11.0" <mesa-stable@lists.freedesktop.org>

i965/bdw: Fix 3DSTATE_VF_INSTANCING when the edge flag is used

When the edge flag element is enabled then the elements are slightly
reordered so that the edge flag is always the last one. This was
confusing the code to upload the 3DSTATE_VF_INSTANCING state because
that is uploaded with a separate loop which has an instruction for
each element. The indices used in these instructions weren't taking
into account the reordering so the state would be incorrect.

v2: Use nr_elements instead of brw->vb.nr_enabled so that it will cope
    when gl_VertexID is used.

Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=91292
Cc: <mesa-stable@lists.freedesktop.org>
Reviewed-by: Ben Widawsky <ben@bwidawsk.net>
Signed-off-by: Ben Widawsky <ben@bwidawsk.net>
Tested-by: Mark Janes <mark.a.janes@intel.com>

i965: Swap the order of the vertex ID and edge flag attributes

The edge flag data on Gen6+ is passed through the fixed function hardware as
an extra attribute. According to the PRM it must be the last valid
VERTEX_ELEMENT structure. However if the vertex ID is also used then another
extra element is added to source the VID. This made it so the vertex ID is in
the wrong register in the vertex shader and the edge attribute is no longer in
the last element.

v2: Also implement for BDW+

v3 [by Ben]: Remove 10.5 tag. Too late.

Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=84677
Cc: <mesa-stable@lists.freedesktop.org>
Reviewed-by: Kristian Høgsberg <krh@bitplanet.net>
Signed-off-by: Ben Widawsky <ben@bwidawsk.net>
Tested-by: Ben Widawsky <ben@bwidawsk.net>
Tested-by: Mark Janes <mark.a.janes@intel.com>

r600g: Fix assert in tgsi_cmp

Fixes https://bugs.freedesktop.org/show_bug.cgi?id=91726

Signed-off-by: Glenn Kennard <glenn.kennard@gmail.com>
Cc: "11.0" <mesa-stable@lists.freedesktop.org>
Signed-off-by: Dave Airlie <airlied@gmail.com>

egl: scons: fix the haiku build, do not build the dri2 backend

Reviewed-by: Emil Velikov <emil.l.velikov@gmail.com>

docs: add 11.1.0-devel release notes template, bump version

Signed-off-by: Emil Velikov <emil.l.velikov@gmail.com>

egl/wayland: define set_cloexec_or_close only when mkostemp is not present

Fixes a compiler warning of defined but not used function when
HAVE_MKOSTEMP is defined.

Fixes: eb3e2562a4b(configure.ac: check for mkostemp())
Signed-off-by: Boyan Ding <boyan.j.ding@gmail.com>
Reviewed-by: Emil Velikov <emil.l.velikov@gmail.com>
Reviewed-by: Matt Turner <mattst88@gmail.com>
Reviewed-by: Thomas Helland <thomashelland90@gmail.com>

mapi: ship ARB_tessellation_shader.xml

Fixes: e2b59a39cbb(mapi: add ARB_tessellation_shader)
Signed-off-by: Emil Velikov <emil.l.velikov@gmail.com>

nouveau: add codegen/unordered_set.h to the tarball

Signed-off-by: Emil Velikov <emil.l.velikov@gmail.com>

winsys/sw/kms-dri: don't attempt to bundle the sconscript

The build/file was removed with an earlier commit while the EXTRA_DIST
was forgotten.

Fixes: 66d77cd71c6 (scons: don't build the kms-dri winsys)
Signed-off-by: Emil Velikov <emil.l.velikov@gmail.com>

winsys/amdgpu: automake: remove missing headers

The files are not referenced in any other place in whole of
mesa. They are likely remnants of the early development stage.

Signed-off-by: Emil Velikov <emil.l.velikov@gmail.com>

automake: build all drivers but vc4 during distcheck

vc4 conflicts with ilo, when build on x86 as it's build for emulation
purposes. In that mode a i965-like symbol is exported by vc4, which
conflicts with the ilo one in the gallium-dri megadriver.

Signed-off-by: Emil Velikov <emil.l.velikov@gmail.com>

android: enable amdgpu winsys in radeonsi driver

Reviewed-by: Edward O'Callaghan <eocallaghan@alterapraxis.com>
Reviewed-by: Emil Velikov <emil.l.velikov@gmail.com>

android: fix cflags and includes for amdgpu winsys

Reviewed-by: Edward O'Callaghan <eocallaghan@alterapraxis.com>
Reviewed-by: Emil Velikov <emil.l.velikov@gmail.com>

docs: add news item and link release notes for 10.6.5

Signed-off-by: Emil Velikov <emil.l.velikov@gmail.com>

docs: add sha256 checksums for 10.6.5

Signed-off-by: Emil Velikov <emil.l.velikov@gmail.com>
(cherry picked from commit fa34225167396008e75e93f23696666caba8a7bf)

docs: add release notes for 10.6.5

Signed-off-by: Emil Velikov <emil.l.velikov@gmail.com>
(cherry picked from commit a43b3dd99bd4c114d0f3e90f4fd4792164fe7539)

i965: Move control flush into pipelined conditional render

The nv_conditional_render piglits were sporadically failing. Moving
the control flush from the write and placing it just before the read
was sufficient to make the piglits pass a 1000/1000 times. The bspec
says that the flush enable bit "waits until all previous writes of
immediate data from post sync circles are complete before executing the
next command" - the operative word being previous!

Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=90691
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Neil Roberts <neil@linux.intel.com>
Cc: Kenneth Graunke <kenneth@whitecape.org>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>

vc4: Actually allow math results to allocate into r4.

I switched us to tracking whether the results *could* go to r4, but then
didn't make a separate register class for the class bits that included r4.
Switch the "any" class to actually be "any", and name the "any but r4"
class more appropriately.

total instructions in shared programs: 96798 -> 94680 (-2.19%)
instructions in affected programs:     62736 -> 60618 (-3.38%)

vc4: Fold the 16-bit integer pack into the instructions generating it.

total instructions in shared programs: 97580 -> 96798 (-0.80%)
instructions in affected programs:     52826 -> 52044 (-1.48%)

vc4: Reuse QPU dumping for packing bits in QIR.

vc4: Make _dest variants of qir ALU helpers to provide an explicit dest.

vc4: Use the SSA defs list for figuring out eligible MOVs for copy prop.

I thought I'd converted this over previously.  It was copy propagating
MOVs badly with the new destination packing flags.

st/nine: Always use user constant buffers

We had several reports of users hitting bugs
with the other path to upload constants,
and switching to the user constant buffer
path solves the bugs.

User constant buffers are expected to be slower
for Nvidia cards, so ideally this patch should be
reverted when the path is fixed.

Reviewed-by: Axel Davy <axel.davy@ens.fr>
Signed-off-by: Krzysztof Sobiecki <sobkas@gmail.com>

st/nine: Silent warning in nine_ff

release build was complaining

Signed-off-by: Axel Davy <axel.davy@ens.fr>

st/nine: Silent warning in sm1_declusage_to_tgsi

release build was complaining

Signed-off-by: Axel Davy <axel.davy@ens.fr>

st/nine: Silent warning in NineCubeTexture9_ctor

The compiler was complaining the value may be uninitialised
when it is used (which is wrong). Initialize to NULL to silent
the warning.

st/nine: Silent warning in update_vertex_buffer

There was an unused variable

st/nine: Catch setting the same shader

This is quite rare that an app does set again
the same shaders, but it isn't an expensive check
either.

Signed-off-by: Axel Davy <axel.davy@ens.fr>

st/nine: Avoid Constant upload when there is no change

It is very common for d3d9 apps to set again the constants
they need before every draw call, even if nothing changed.

Since we are mostly gpu bound, it is better to check
for change, and upload constants again (and thus use
gpu bandwith) only if the constants changed.

Signed-off-by: Axel Davy <axel.davy@ens.fr>

st/nine: Fix the number of texture stages

The number of texture stages is 8.

'tex_stage' array was too big, and thus
the checks with 'Elements(state->ff.tex_stage)' were passing,
causing some invalid API calls to pass, and crash because of
out of bounds write since bumpmap_vars was just the correct size.

Signed-off-by: Axel Davy <axel.davy@ens.fr>

st/nine: Use CSO cache for sampler views

The CSO cache unbinds views that are not needed anymore,
which we don't do.
It checks for change before committing the views.

Signed-off-by: Axel Davy <axel.davy@ens.fr>

st/nine: Calculate dummy sampler state only once

Signed-off-by: Axel Davy <axel.davy@ens.fr>

st/nine: Better check shader constant limits

Signed-off-by: Axel Davy <axel.davy@ens.fr>

st/nine: Remove NINED3DRS_ZBIASSCALE

It wasn't giving the expected result.

This fixes some object being transparents
in games like FEAR.

Signed-off-by: Axel Davy <axel.davy@ens.fr>

st/nine: Implement special DOTPRODUCT3 behaviour

Taken from wine tests

Signed-off-by: Axel Davy <axel.davy@ens.fr>

st/nine: Implement ff vertex data passthrough

Fixes Wine tests

Signed-off-by: Axel Davy <axel.davy@ens.fr>

st/nine: Change nine_state_update order

nine_update_state called every draw call.

This patch attemps to change the order
of the checks to have better control flow

Signed-off-by: Axel Davy <axel.davy@ens.fr>

st/nine: Programmable ps D3DTTSS_PROJECTED support

The implementation used Wine tests for conformance

Signed-off-by: Axel Davy <axel.davy@ens.fr>

st/nine: Complete ff texture transform implementation

Wine tests were used to get it right.

Signed-off-by: Axel Davy <axel.davy@ens.fr>

st/nine: Change a few advertised caps

There were flags all sm3 cards do advertise,
and we weren't.
Some games can trigger buggy rendering path
if the caps are not what they expect.

Signed-off-by: Axel Davy <axel.davy@ens.fr>

st/nine: Advertise Fog flags

Signed-off-by: Axel Davy <axel.davy@ens.fr>

st/nine: Revert to userbuf path when needed

Automatically switch to userbuf path when
we would need to upload fog or bumpmat
constants

Signed-off-by: Axel Davy <axel.davy@ens.fr>

st/nine: Finish Fog implementation

Signed-off-by: Axel Davy <axel.davy@ens.fr>

st/nine: Rework shader states

Separate state setting and commit
Changes how the shader key is computed

Signed-off-by: Axel Davy <axel.davy@ens.fr>

st/nine: Remove some useless variables

Signed-off-by: Axel Davy <axel.davy@ens.fr>
Reviewed-by: David Heidelberg <david@ixit.cz>

st/nine: Fix nine_ff_ps_key padding

Signed-off-by: Axel Davy <axel.davy@ens.fr>

st/nine: Begin programmable shader fog support

Signed-off-by: Axel Davy <axel.davy@ens.fr>

st/nine: Fix fixed function fog support

Previous code had only a subcase of fog working right.

Signed-off-by: Axel Davy <axel.davy@ens.fr>

st/nine: Rework ff constant buffers

Always use a user constant buffer for ff.
It means we have to:
. commit the user constant buffer for ff when we use it
. commit back the non-ff constant buffer when we stop using it

Signed-off-by: Axel Davy <axel.davy@ens.fr>

st/nine: Rework constant buffer state handling

We have two paths:
. One that uses a fixed constant buffer, and updates it when needed
. One that uses a user constant buffer, and uploads it when needed.

This patch separates the preparation of the constant buffer
and the commit.

It also removes NineDevice9_RestoreNonCSOState, which was
used to restore all states. Instead the commit of the constant
buffer is moved to nine_state, and the other field settings
moved to other functions where more appropriate.

Signed-off-by: Axel Davy <axel.davy@ens.fr>

st/nine: Rework blend states

Separate state preparation and state commit

Signed-off-by: Axel Davy <axel.davy@ens.fr>

st/nine: Improve fallback when driver doesn't support user buffers.

For now the path updated is only used by Amd drivers, but a later
patch will make it used by all drivers. Some drivers like llvmpipe
doesn't support the uploading of constants from user buffers, so improve
the path to work for all drivers

Inspired from the gl state tracker.

Signed-off-by: Axel Davy <axel.davy@ens.fr>

st/nine: Avoid useless updates in SetSamplerState

Check for redundant settings

Signed-off-by: Axel Davy <axel.davy@ens.fr>
Reviewed-by: David Heidelberg <david@ixit.cz>

st/nine: Rework rasterizer states

Separate state preparation and state commit

Signed-off-by: Axel Davy <axel.davy@ens.fr>

st/nine: Reorder DSA state settings

Separate state preparation and state commit

Signed-off-by: Axel Davy <axel.davy@ens.fr>

st/nine: Reorder nine_state.

Instead of mixing state preparation (filling pipe_****)
and state commit (pipe->set_*****),
begin doing so in two separate functions.

This will allow to implement efficient Stateblocks,
and eventually lead to optimisation where the complete
pipe_*** structure is only partially updated.

Signed-off-by: Axel Davy <axel.davy@ens.fr>

st/nine: Remove group_mask argument from nine_update_state

It was only used to discriminate update framebuffer vs update
everything. Instead use two functions.

Signed-off-by: Axel Davy <axel.davy@ens.fr>

st/nine: Implement TEXBEM,TEXBEML and BEM

Signed-off-by: Tiziano Bacocco <tizbac2@gmail.com>

st/nine: Fix use of uninitialized values

Set all values to 0 after allocation. Found using valgrind.

Reviewed-by: Axel Davy <axel.davy@ens.fr>
Signed-off-by: Patrick Rudolph <siro@das-labor.org>

st/nine: Prevent possible crash

In case NineBaseTexture9_ctor returns an error
This->surfaces[l] might be NULL.

Reviewed-by: Axel Davy <axel.davy@ens.fr>
Signed-off-by: Patrick Rudolph <siro@das-labor.org>

st/nine: Return correct error codes in NineDevice9_Reset

Allow more than two errors, and return D3DERR_INVALIDCALL
for failed display resolution changes.

Reviewed-by: Axel Davy <axel.davy@ens.fr>
Signed-off-by: Patrick Rudolph <siro@das-labor.org>

st/nine: Fail on D3DUSAGE_DYNAMIC for D3DPOOL_SCRATCH textures

Texture with pool D3DPOOL_SCRATCH and D3DPOOL_MANAGED
cannot be used with flag D3DUSAGE_DYNAMIC.

Signed-off-by: Patrick Rudolph <siro@das-labor.org>

st/nine: Fix Lock Checks for Compressed textures

Signed-off-by: Axel Davy <axel.davy@ens.fr>

st/nine: Impose restrictions on DXTN texture sizes

This is the expected behaviour.

Fixes wine tests.

Signed-off-by: Axel Davy <axel.davy@ens.fr>
Reviewed-by: David Heidelberg <david@ixit.cz>

st/nine: Return NULL pointer in lock error cases

Tests showed, that in case of errors, the pBits pointer is set to NULL.
The pBits field isn't set to NULL in case of an already locked object.

Reviewed-by: Axel Davy <axel.davy@ens.fr>
Signed-off-by: Patrick Rudolph <siro@das-labor.org>

st/nine: Fix resource SetPriority/GetPriority

Return 0 for non MANAGED textures and surfaces.
Fixes failing wine d3d9 tests device.c test_resource_priority.

Reviewed-by: Axel Davy <axel.davy@ens.fr>
Signed-off-by: Patrick Rudolph <siro@das-labor.org>

st/nine: Clean GetPrivateData

Move the assert to return error codes in the correct order.
Always set the pSizeOfData to the required buffer size.
Fixes failing wine test device.c test_private_data()

Reviewed-by: Axel Davy <axel.davy@ens.fr>
Signed-off-by: Patrick Rudolph <siro@das-labor.org>

st/nine: Allow lock coordinates outside range

This fixes wine test device.c test_lockrect_invalid()

Mimic WindowsXp behaviour and allow negative values in the rectangle passed.
Add comment to point out behaviour used.

Reviewed-by: Axel Davy <axel.davy@ens.fr>
Signed-off-by: Patrick Rudolph <siro@das-labor.org>