docs: remove stray paragraph-close

This tag tries to close a non-existent paragraph. Let's get rid of it!

Signed-off-by: Erik Faye-Lund <erik.faye-lund@collabora.com>
Reviewed-by: Eric Engestrom <eric.engestrom@intel.com>

docs: properly escape ampersand

Even in preformatted blocks, ampersands should be escaped. Let's correct
this, in case of strict parsers.

Signed-off-by: Erik Faye-Lund <erik.faye-lund@collabora.com>
Reviewed-by: Eric Engestrom <eric.engestrom@intel.com>

docs: properly escape '>'

The '>'-symbol should usually be escaped to avoid confusing strict
parsers. While it's very unlikely to cause issues as-is, let's quite it
for good measure.

Signed-off-by: Erik Faye-Lund <erik.faye-lund@collabora.com>
Reviewed-by: Eric Engestrom <eric.engestrom@intel.com>

radv: fix set_output_usage_mask() with composite and 64-bit types

It previously used var->type instead of deref_instr->type and didn't
handle 64-bit outputs.

This fixes lots of transform feedback CTS tests involving transform
feedback and geometry shaders (mostly
dEQP-VK.transform_feedback.fuzz.random_geometry.*)

v2: fix writemask widening when comp != 0
v3: fix 64-bit variables when comp != 0, again

Signed-off-by: Rhys Perry <pendingchaos02@gmail.com>
Cc: 19.0 19.1 <mesa-stable@lists.freedesktop.org>
Reviewed-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>

docs: do not hard-code header-height

It's generally nicer to do this in terms of em units, as that scales
better with text-sizes, if we ever decide to change them.

The result is slightly larger than before, but only by a couple of
pixels.

Signed-off-by: Erik Faye-Lund <erik.faye-lund@collabora.com>
Reviewed-by: Eric Engestrom <eric.engestrom@intel.com>

docs: simplify css-centering

With "display: flex;" we can make this a bit more automatic, not
requiring a bunch of values to be of specific values to get the right
centering.

Signed-off-by: Erik Faye-Lund <erik.faye-lund@collabora.com>
Reviewed-by: Eric Engestrom <eric.engestrom@intel.com>

docs: use multiple background-images for header

This is a bit tidier than to set a background on the h1-text, requiring
it to be full height and all.

Signed-off-by: Erik Faye-Lund <erik.faye-lund@collabora.com>
Reviewed-by: Eric Engestrom <eric.engestrom@intel.com>

docs: remove spurious newline

Signed-off-by: Erik Faye-Lund <erik.faye-lund@collabora.com>
Reviewed-by: Eric Engestrom <eric.engestrom@intel.com>

docs: avoid repeating the color

The color attribute is inherited in CSS, so there's no point in repeating
this.

Signed-off-by: Erik Faye-Lund <erik.faye-lund@collabora.com>
Reviewed-by: Eric Engestrom <eric.engestrom@intel.com>

docs: avoid repeating the font

The font attribute is inherited in CSS, so there's no point in repeating
this.

Signed-off-by: Erik Faye-Lund <erik.faye-lund@collabora.com>
Reviewed-by: Eric Engestrom <eric.engestrom@intel.com>

docs: add missing semicolon

While it's legal to omit the last semicolon in a CSS block, it's
generally not considered good style, as it makes it harder to add new
lines.

Signed-off-by: Erik Faye-Lund <erik.faye-lund@collabora.com>
Reviewed-by: Eric Engestrom <eric.engestrom@intel.com>

docs: remove long commented out css

These attributes has been commented out since 2005; I don't think
there's a big chance of them making a return as-is.

Signed-off-by: Erik Faye-Lund <erik.faye-lund@collabora.com>
Reviewed-by: Eric Engestrom <eric.engestrom@intel.com>

docs: remove non-existent css attribute

There's no CSS-attribute named "link", so let's remove it.

Signed-off-by: Erik Faye-Lund <erik.faye-lund@collabora.com>
Reviewed-by: Eric Engestrom <eric.engestrom@intel.com>

docs: normaize css-indent style

Tabs has been around as the indention style of this file since it was
created. Some newer CSS has added double-spaces, but let's keep it
consistent.

Signed-off-by: Erik Faye-Lund <erik.faye-lund@collabora.com>
Reviewed-by: Eric Engestrom <eric.engestrom@intel.com>

winsys/svga: Don't abort on EBUSY errors from execbuffer

This error code typically indicated that a buffer object that was referenced
by the command stream was being used for CPU access by another client.
The correct action here is to retry after a while. Use usleep() until we
have proper kernel support for this wait.

Signed-off-by: Thomas Hellstrom <thellstrom@vmware.com>
Reviewed-by: Brian Paul <brianp@vmware.com>

winsys/svga: Update the drm interface file

The file vmwgfx_drm.h was a bit outdated. Update to a recent version,
including defines supporting coherent memory.

Signed-off-by: Thomas Hellstrom <thellstrom@vmware.com>
Reviewed-by: Brian Paul <brianp@vmware.com>

svga: Avoid bouncing buffer data in malloced buffers

Some constant- and texture upload buffer data may bounce in malloced
buffers before being transferred to hardware buffers. In the case of
texture upload buffers this seems to be an oversight. In the case of
constant buffers, code comments indicate that we want to avoid mapping
hardware buffers for reading when copying out of buffers that need
modification before being passed to hardware. In this case we avoid
data bouncing for upload manager buffers but make sure buffers that
we read out from stay in malloced memory.

Signed-off-by: Thomas Hellstrom <thellstrom@vmware.com>
Reviewed-by: Brian Paul <brianp@vmware.com>

winsys/svga: Enable the transfer_from_buffer GPU command for vgpu10

We didn't have the path using this command enabled as
typically we take an alternate path using DMA uploads.
Emable it so that we can exercise that code-path by turning off
the DMA path.

Signed-off-by: Thomas Hellstrom <thellstrom@vmware.com>
Reviewed-by: Brian Paul <brianp@vmware.com>

winsys/svga: Add an environment variable to force host-backed operation

The vmwgfx kernel module has a compatibility mode for user-space that is
not guest-backed resource aware. Add an environment variable to facilitate
testing of this mode on guest-backed aware kernels: if the environment
variable SVGA_FORCE_HOST_BACKED is defined, the driver will use host-backed
operation.

Signed-off-by: Thomas Hellstrom <thellstrom@vmware.com>
Reviewed-by: Deepak Rawat <drawat@vmware.com>
Reviewed-by: Brian Paul <brianp@vmware.com>

ac: tidy up ac_build_llvm8_tbuffer_{load,store}

For consistency with ac_build_llvm8_buffer_{load,store}_common
helpers and that will help a bit for removing the vec3 restriction.

Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>

radv: implement a workaround for VK_EXT_conditional_rendering

Per the Vulkan spec 1.1.107, the predicate is a 32-bit value. Though
the AMD hardware treats it as a 64-bit value which means it might
fail to discard.

I don't know why this extension has been drafted like that but this
definitely not fit with AMD. The hardware doesn't seem to support
a 32-bit value for the predicate, so we need to implement a workaround.

This fixes an issue when DXVK enables conditional rendering with RADV,
this also fixes the Sasha conditionalrender demo.

Fixes: e45ba51ea45 ("radv: add support for VK_EXT_conditional_rendering")
Reported-by: Philip Rebohle <philip.rebohle@tu-dortmund.de>
Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>

radv: fix color conversions for normalized uint/sint formats

The hardware actually rounds before conversion. This now matches
what values are used when performing fast clears vs slow clears.

This fixes a rendering issue with Far Cry 3&4. This also fixes
a bunch of CTS tests that use a 8-bit UNORM format (only when
the 512*512 image size hint is manually disabled).

Cc: "19.0" "19.1" <mesa-stable@lists.freedesktop.org>
Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>

radv: do not need to force emit the TCS regs on Vega20

This chip doesn't need the fixup. This fixes a bunch of
dEQP-VK.tessellation tests and avoid random GPU hangs.

Cc: "19.0" "19.1" <mesa-stable@lists.freedesktop.org>
Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>

util/bitset: Return an actual bool from test macros

I want to be able to do BITSET_TEST() != BITSET_TEST() and this isn't
currently possible because BITSET_TEST() returns a random bit.  Compare
to zero to get an actual Boolean.

Reviewed-by: Eric Engestrom <eric.engestrom@intel.com>
Reviewed-by: Ian Romanick <ian.d.romanick@intel.com>

glsl: work around MinGW 7.x compiler bug

I'm not sure what triggered this, but building with
scons platform=windows toolchain=crossmingw machine=x86 build=profile
with MinGW g++ 7.3 or 7.4 causes an internal compiler error.

We can work around it by forcing -O1 optimization.

Reviewed-by: Jose Fonseca <jfonseca@vmware.com>
Reviewed-by: Neha Bhende <bhenden@vmware.com>

llvmpipe: init some vars to NULL to silence MinGW compiler warnings

Reviewed-by: Neha Bhende <bhenden@vmware.com>

radeonsi: set sampler state and view functions for compute-only contexts

radeonsi: use new atomic LLVM helpers

This depends on "ac,ac/nir: use a better sync scope for shared atomics"

st/mesa: don't flush the front buffer if it's a pbuffer

This is the best guess I can make here.

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

mesa: fix pbuffers because internally they are front buffers

This fixes the egl_ext_device_base piglit test, which uses EGL pbuffers.

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

mesa: rework error handling in glDrawBuffers

It's needed by the next pbuffer fix, which changes the behavior of
draw_buffer_enum_to_bitmask, so it can't be used to help with error
checking.

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

radv: Restrict YUVY formats to 1 layer.

Fixes: 8bb3cec7c9b "radv: Expose VK_EXT_ycbcr_image_arrays."
Reviewed-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>

radv: Set is_array in lowered ycbcr tex instructions.

Fixes array tests.

Fixes: 91702374d5d "radv: Add ycbcr lowering pass."
Reviewed-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>

radv: Fix hang width YCBCR array textures.

Forgot to apply the width/height divisor for CB writes resulting in
the CB using larger than expected slice sizes.

Fixes: 42d159f2766 "radv: Add multiple planes to images."
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=110530
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=110526
Reviewed-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>

lima/gpir: add limit of max 512 instructions

It has been noted that the lima GP has a limit of 512 instructions,
after which the shaders don't work and fail silently.
This commit adds a check to make the shader compilation abort when the
shader exceeds this limit, so that we get a clear reason for why the
program will not work.

Signed-off-by: Erico Nunes <nunes.erico@gmail.com>
Reviewed-by: Qiang Yu <yuq825@gmail.com>

panfrost: Fix blend shader upload

Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>

panfrost/decode: Hit MRT blend shader enable bits

Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>

panfrost: Remove shader dump

Redundant via the midgard shader dump.

Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>

virgl: Re-use and extend queue transfers for intersecting buffer subdatas.

Small buffer subdatas which are essentially doing a memcpy were getting
bogged down by all the overhead of creating new transfers.

Signed-off-by: David Riley <davidriley@chromium.org>
Reviewed-by: Gurchetan Singh <gurchetansingh@chromium.org>

virgl: Allow transfer queue entries to be found and extended.

Intersecting transfer queue entries allow for the possibility of
extending an existing transfer instead of creating a new one (and all
the associated mappign/unmapping).

Signed-off-by: David Riley <davidriley@chromium.org>
Reviewed-by: Gurchetan Singh <gurchetansingh@chromium.org>

virgl: Store mapped hw resource with transfer object.

Signed-off-by: David Riley <davidriley@chromium.org>
Reviewed-by: Gurchetan Singh <gurchetansingh@chromium.org>

iris: Fix imageBuffer and PBO download.

Recently we added checks to try and deny multisampled shader images.
Unfortunately, this messed up imageBuffers, which have sample_count = 0,
which are also used in PBO download, causing us hit CPU map fallbacks.

Fixes: b15f5cfd20c iris: Do not advertise multisampled image load/store.
Reviewed-by: Rafael Antognolli <rafael.antognolli@intel.com>

r600: reset tex array override even when no view bound

If no view is bound we still should reset the override to 0
and array mode.

This should fix misrendering in firefox WebRender since
the pbo sampler was removed.

Fixes: 1250383e36 (st/mesa: remove sampler associated with buffer texture in pbo logic)

nir: Saturating integer arithmetic is not associative

In 8-bits,

    iadd_sat(iadd_sat(0x7f, 0x7f), -1) =
    iadd_sat(0x7f, -1) =
    0x7e

but,

    iadd_sat(0x7f, iadd_sat(0x7f, -1)) =
    iadd_sat(0x7f, 0x7e) =
    0x7f

Fixes: 272e927d0e9 ("nir/spirv: initial handling of OpenCL.std extension opcodes")
Reviewed-by: Karol Herbst <kherbst@redhat.com>
Reviewed-by: Jason Ekstrand <jason@jlekstrand.net>

util: move #include out of #if linux

This #include is needed for `NULL`, which is used on all OSes, not just Linux.

Reported-by: Juan A. Suarez Romero <jasuarez@igalia.com>
Fixes: 316964709e21286c2af5 "util: add os_read_file() helper"
Signed-off-by: Eric Engestrom <eric.engestrom@intel.com>
Reviewed-by: Juan A. Suarez <jasuarez@igalia.com>

swr/rast: Add general SWTag statistics

Update Archrast parser to use stats, used with an internal tool

Reviewed-by: Bruce Cherniak <bruce.cherniak@intel.com>

swr/rast: Add string handling to AR event framework

For use by an internal tool

Reviewed-by: Bruce Cherniak <bruce.cherniak@intel.com>

swr/rast: Add initial SWTag proto definitions

Update gen_archrast.py to properly generate event IDs

Reviewed-by: Bruce Cherniak <bruce.cherniak@intel.com>

swr/rast: Cleanup and generalize gen_archrast

- Update meson.build
  - Includes current_build_dir() fix
    meson/swr: replace hard-coded path with current_build_dir()

Fixes: 93cd9905c8fbb98985ae "swr/rast: Cleanup and generalize gen_archrast"
Signed-off-by: Eric Engestrom <eric.engestrom@intel.com>
Reviewed-by: Alok Hota <alok.hota@intel.com>
Reviewed-by: Dylan Baker <dylan@pnwbakers.com>
  - Clean up meson.build (remove foreach loop, replace with single call)
- Update SConscript
  - use `$SOURCES` to call `CodeGenerate` with multiple source files

Reviewed-by: Bruce Cherniak <bruce.cherniak@intel.com>

gitlab-ci: build vulkan drivers in clang build

Signed-off-by: Eric Engestrom <eric.engestrom@intel.com>
Reviewed-by: Michel Dänzer <michel.daenzer@amd.com>

softpipe: setup pixel_offset for all primitive types

If we don't update this for all primitive-types, we end up rendering
slightly offset points and lines up until the point where the first
triangle gets drawn. This is obviously not correct, and violates
OpenGL's repeatability rule.

Signed-off-by: Erik Faye-Lund <erik.faye-lund@collabora.com>
Fixes: ca9c413647b ("softpipe: Respect gl_rasterization_rules in
       primitive setup.")
Reviewed-By: Gert Wollny <gert.wollny@collabora.com>

nir: improve convert_yuv_to_rgb

Use a different arrangement of constants to allow more ffma.

A vec4 backend will now use 3 fma for yuv_to_rgb. On freedreno/ir3, it is
down from 10 to 7 alu (4 fma, 3 mul, 3 add to 7 fma). Other backends
shouldn't be hurt.

Signed-off-by: Jonathan Marek <jonathan@marek.ca>
Reviewed-by: Eric Anholt <eric@anholt.net>
Tested-by: Ian Romanick <ian.d.romanick@intel.com>

doc: Update feature matrix

Since softpipe doesn't truely support multisample, I've not added softpipe
to the "Enhanced per-sample shading" even though with the advertised GLSL
level ARB_gpu_shader5 is advertised.

Signed-off-by: Gert Wollny <gert.wollny@collabora.com>
Reviewed-by: Roland Scheidegger <sroland@vmware.com>

softpipe: Increase the GLSL feature level

This will enable calls to the interpolateAt* functions, but also a bunch
of other features.

Signed-off-by: Gert Wollny <gert.wollny@collabora.com>
Reviewed-by: Roland Scheidegger <sroland@vmware.com>

softpipe: Add support for TGSI_OPCODE_INTERP_CENTROID

Like with interpolatAtSample this is also not really implementing the
according sampling and will only work correctly for pixels that are fully
covered, but since softpipe only supports one sample this is good enough
for now.

v2: Correct spelling (Roland Scheidegger)

Signed-off-by: Gert Wollny <gert.wollny@collabora.com>
Reviewed-by: Roland Scheidegger <sroland@vmware.com>

softpipe: Add support for TGSI_OPCODE_INTERP_OFFSET

Since for this opcode the offsets are given manually the function
should actually also work for non-zero offsets, but the related piglits
only ever test with offset 0. Accordingly the patch satisfies
"fs-interpolateatoffset-*".

Signed-off-by: Gert Wollny <gert.wollny@collabora.com>
Reviewed-by: Roland Scheidegger <sroland@vmware.com>

softpipe: Add (fake) support for TGSI_OPCODE_INTERP_SAMPLE

Softpipe doesn't support more than one sample, so this function
implements the interpolation at sample 0 and adds a stub to make it
possible to interpolate at other samples.

As it is this makes the piglits "fs-interpolateatsample-*" pass, but
they only ever test sample 0 anyway.

Signed-off-by: Gert Wollny <gert.wollny@collabora.com>
Reviewed-by: Roland Scheidegger <sroland@vmware.com>

softpipe: Add an per-input array for interpolator correctors to machine

This adds entry points for correcting the interpolation values if the
interpolation is done by using one of the interpolateAt* functions.

Signed-off-by: Gert Wollny <gert.wollny@collabora.com>
Reviewed-by: Roland Scheidegger <sroland@vmware.com>

softpipe: Factor out evaluation of the source indices

We will need these for per sample interpolation as well

Signed-off-by: Gert Wollny <gert.wollny@collabora.com>
Reviewed-by: Roland Scheidegger <sroland@vmware.com>

softpipe: evaluate cube the faces on a per sample bases

Now that the LOD is evaluated up front the cube faces can also be
evauate on a per sample basis instead of using the quad.

This fixes a large number of deqp gles 3 and 31 cube texture tests.

Signed-off-by: Gert Wollny <gert.wollny@collabora.com>
Reviewed-by: Roland Scheidegger <sroland@vmware.com>

softpipe: keep input lod for explicite derivatives

This only affects anisotropic interpolation.

Signed-off-by: Gert Wollny <gert.wollny@collabora.com>
Reviewed-by: Roland Scheidegger <sroland@vmware.com>

softpipe: tie in new code path for lod evaluation

This enables the use of explicit gradients.

Also remove an unused parameter when changing the interfaces.

Signed-off-by: Gert Wollny <gert.wollny@collabora.com>
Reviewed-by: Roland Scheidegger <sroland@vmware.com>

softpipe: Move selection of shadow values up and clean parameter list

The shadow evaluation compare parameter is stored in different locations,
depending on the texture type. Move the values to a common location free
the lod storage and to be able to reduce the number of parameters.

Signed-off-by: Gert Wollny <gert.wollny@collabora.com>
Reviewed-by: Roland Scheidegger <sroland@vmware.com>

softpipe: Pipe gather_comp through from st_tgsi_get_samples

The value is stored in the lod components and this will be overwritten
when swithcing to the new code path.

Signed-off-by: Gert Wollny <gert.wollny@collabora.com>
Reviewed-by: Roland Scheidegger <sroland@vmware.com>

softpipe: Prepare handling explicit gradients

This only adds corde that is not yet enabled.

Signed-off-by: Gert Wollny <gert.wollny@collabora.com>
Reviewed-by: Roland Scheidegger <sroland@vmware.com>

softpipe: Factor gradient evaluation out of the lambda evaluation

this is useful when we want to use explicit gradients.

Signed-off-by: Gert Wollny <gert.wollny@collabora.com>
Reviewed-by: Roland Scheidegger <sroland@vmware.com>

egl: return correct error code for a case req ver < 3 with forward-compatible

The EGL_KHR_create_context spec says:
   "If an OpenGL context is requested and the values for attributes
    EGL_CONTEXT_MAJOR_VERSION_KHR and EGL_CONTEXT_MINOR_VERSION_KHR,
    when considered together with the value for attribute
    EGL_CONTEXT_OPENGL_FORWARD_COMPATIBLE_BIT_KHR, specify an OpenGL
    version and feature set that are not defined, than an
    EGL_BAD_MATCH error is generated."

This case is already correctly handled a bit below in
the same source file.
The correct handling was added by commit: 63beb3df

Reported-by: Ian Romanick <idr@freedesktop.org>
Here: https://bugzilla.freedesktop.org/show_bug.cgi?id=92552#c9
Fixes: 11cabc45b712 "egl: rework handling EGL_CONTEXT_FLAGS"
Reviewed-by: Ian Romanick <ian.d.romanick@intel.com>
Signed-off-by: Andrii Simiklit <andrii.simiklit@globallogic.com>

radeonsi/nir: call radeonsi nir opts before the scan pass

Some of the opts are not called in the general optimastion loop
in the state trackers glsl -> nir conversion. We need to call
the radeonsi specific optimisation once before scanning over
the nir otherwise we can end up gathering info on code that
is later removed.

Fixes an assert in the piglit test:

./bin/varying-struct-centroid_gles3

Reviewed-by: Marek Olšák <marek.olsak@amd.com>

radeonsi/nir: create si_nir_opts() helper

We will make use of this in the following commit.

Reviewed-by: Marek Olšák <marek.olsak@amd.com>

swr/rast: early exit on empty triangle mask

Reviewed-by: Bruce Cherniak <bruce.cherniak@intel.com>

swr/rast: add guards for cpuid on Linux

Reviewed-by: Bruce Cherniak <bruce.cherniak@intel.com>

swr/rast: add flat shading

Reviewed-by: Bruce Cherniak <bruce.cherniak@intel.com>

swr/rast: add SWR_STATIC_ASSERT() macro

Reviewed-by: Bruce Cherniak <bruce.cherniak@intel.com>

swr/rast: update guardband rects at draw setup

It's dependent on other state fields

Reviewed-by: Bruce Cherniak <bruce.cherniak@intel.com>

swr/rast: add more llvm intrinsics

Reviewed-by: Bruce Cherniak <bruce.cherniak@intel.com>

st/va: properly set stride and offset in vlVaDeriveImage

Using the new resource_get_info function.

Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=110443
Signed-off-by: Julien Isorce <jisorce@oblong.com>
Reviewed-by: Leo Liu <leo.liu@amd.com>

radeonsi: implement resource_get_info

Re-use existing si_texture_get_offset.

Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=110443
Signed-off-by: Julien Isorce <jisorce@oblong.com>
Reviewed-by: Marek Olšák <marek.olsak@amd.com>

gallium: add resource_get_info to pipe_screen

Generic plumbing.

Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=110443
Signed-off-by: Julien Isorce <jisorce@oblong.com>
Reviewed-by: Marek Olšák <marek.olsak@amd.com>

freedreno/ir3: fixes for half reg in/out

Needs to update max_half_reg, or be remapped to full reg and update
max_reg accordingly, depending on generation..

Signed-off-by: Rob Clark <robdclark@chromium.org>

st/nine: Check discard_delayed_release is set before allocating more

When discard_delayed_release is set (default), we allocate more buffers
and use a different buffer wait path.
Check if it is set, and use the old paths if not
(the alternative buffer wait path could still be used, but there is no
advantage to using it in this case).

Signed-off-by: Axel Davy <davyaxel0@gmail.com>

st/nine: Throttle rendering similarly for thread_submit

thread_submit's throttling depending on the number of internal
back buffers, and wasn't affected by the driver requested
throttling value.
Now it is.

Signed-off-by: Axel Davy <davyaxel0@gmail.com>

st/nine: Optimize a bit writeonly buffers

Optimize writeonly by passing PIPE_TRANSFER_WRITE
for these buffers instead of the safer
PIPE_TRANSFER_READ_WRITE.

This seems to improve the performance of d3d8 games
using d3d8to9.

Signed-off-by: Axel Davy <davyaxel0@gmail.com>

st/nine: Use TGSI_SEMANTIC_GENERIC for fog

We used TGSI_SEMANTIC_FOG for fog,
however on vs/ps 3, fog is allowed to have
4 components (even on the ff pipeline according
to a wine test).
Since gallium's TGSI_SEMANTIC_FOG has only one
component, use TGSI_SEMANTIC_GENERIC instead.

Fixes:
https://github.com/iXit/Mesa-3D/issues/346

Signed-off-by: Axel Davy <davyaxel0@gmail.com>

st/nine: Enable computing const_ranges

All the pieces for constant compact are ready,
thus enable the path.

Signed-off-by: Axel Davy <davyaxel0@gmail.com>

st/nine: Handle const_ranges in nine_state

Handle slot mapping if there is one.

Signed-off-by: Axel Davy <davyaxel0@gmail.com>

st/nine: Cache constant buffer size

The shader constant buffer size with the
constant compaction code can vary depending
on the shader variant compiled (for example if
fog constants are required, etc).
Thus instead of using fixed size for the shader,
add in the variant cache the size required, pass it
to the context, and use this value.

Signed-off-by: Axel Davy <davyaxel0@gmail.com>

st/nine: Propagate const_range to context

As with the constant compaction we map the constant
slots to new slots, we need to pass that information
to the context which is in charge of uploading
the constants.

Signed-off-by: Axel Davy <davyaxel0@gmail.com>

st/nine: Prepare constant compaction in nine_shader

When indirect addressing is not used, we know exactly
which constants are accessed, and thus can
have them located in consecutive slots.
We thus parse again the shader with a slot map
for compaction.

The path contains the work inside nine_shader.c for this
path, but it needs some other commits to work, and thus
is not enabled yet by this commit.

Signed-off-by: Axel Davy <davyaxel0@gmail.com>

st/nine: Refactor counting of constants

Track the number of slots used

Signed-off-by: Axel Davy <davyaxel0@gmail.com>

st/nine: Track constant slots used

This tracking will be useful for constant compaction

Signed-off-by: Axel Davy <davyaxel0@gmail.com>

st/nine: Refactor ct_ctor

The refactoring will make it easier to parse the shader
twice for the constant compaction path.

Signed-off-by: Axel Davy <davyaxel0@gmail.com>

st/nine: Make swvp_on imply IS_VS

swvp cannot happen with ps, thus it makes
sense to force it to false with ps.

Signed-off-by: Axel Davy <davyaxel0@gmail.com>

st/nine: Refactor shader constants ureg_src computation

Put the shader constant code in one place to better change
that code in future commits.

Signed-off-by: Axel Davy <davyaxel0@gmail.com>

st/nine: Manually upload vs and ps constants

In future commits we will introduce more fine-grained
uploads

Signed-off-by: Axel Davy <davyaxel0@gmail.com>

st/nine: use helper ureg_DECL_sampler everywhere

Signed-off-by: Axel Davy <davyaxel0@gmail.com>

st/nine: Compact pixel shader key

Compact the shader key to make room for new
elements.

Signed-off-by: Axel Davy <davyaxel0@gmail.com>

st/nine: Compact nine_ff_get_projected_key

Only the first four sampler slots can be used
by ff ps < 0x14, thus the size of the key
can be reduced.

Signed-off-by: Axel Davy <davyaxel0@gmail.com>

st/nine: Refactor param->rel

Refactor param->rel to enable different paths
for constants and inputs relative addressing.

Signed-off-by: Axel Davy <davyaxel0@gmail.com>

st/nine: Regroup param->rel tests

Regroup all the param->rel assertions into one assertion for better clarity
and better covering.
param->rel on an input can only happen with float constants for vs,
or with inputs on vs/ps 3.0.

Signed-off-by: Axel Davy <davyaxel0@gmail.com>

st/nine: Control shader constant inlining with drirc

Until we use async shader compilation for constant inlining,
don't enable it unless user asks for it.

Signed-off-by: Axel Davy <davyaxel0@gmail.com>