i965: Delete vestiges of resource streamer code.

We never actually used the resource streamer in any shipping build
of Mesa.  We have no plans to do so in the future.  We looked into
using it in Vulkan, and concluded that it was unusable.  We're not
the only ones to arrive at the conclusion that it's not worth using.

So, drop the last vestiges of resource streamer support and move on.

Signed-off-by: Kenneth Graunke <kenneth@whitecape.org>
Reviewed-by: Jason Ekstrand <jason@jlekstrand.net>
Reviewed-by: Emil Velikov <emil.velikov@collabora.com>

i965: Drop duplicate #defines now that we've bumped libdrm requirements.

We've updated our libdrm requirement, and it will already provide these.

Signed-off-by: Kenneth Graunke <kenneth@whitecape.org>
Reviewed-by: Jason Ekstrand <jason@jlekstrand.net>
Reviewed-by: Emil Velikov <emil.velikov@collabora.com>

getteximage: fix _mesa_GetTextureSubImage()

Oops.

Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=100088
Fixes: 5ae54c0cf7 ("getteximage: avoid to lookup textures with id 0")
Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Marek Olšák <marek.olsak@amd.com>

ralloc: don't leave out the alignment factor

Experimentation shows that without alignment factor gcc and clang choose
a factor of 16 even on IA-32, which doesn't match what malloc() uses (8).
The problem is it makes gcc assume the pointer is 16 byte aligned, so
with -O3 it starts using aligned SSE instructions that later fault,
so always specify a suitable alignment factor.

Cc: Jonas Pfeil <pfeiljonas@gmx.de>
Fixes: cd2b55e5 "ralloc: Make sure ralloc() allocations match malloc()'s alignment."
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=100049
Signed-off-by: Grazvydas Ignotas <notasas@gmail.com>
Tested by: Mike Lothian <mike@fireburn.co.uk>
Tested by: Jonas Pfeil <pfeiljonas@gmx.de>

i965: don't require 64bit cmpxchg

There are still some distributions trying to support unfortunate people
with old or exotic CPUs that don't have 64bit atomic operations. The
only thing preventing compile of the Intel driver for them seems to be
initialization of a debug variable.

v2: use call_once() instead of unsafe code, as suggested by Matt Turner

Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=93089
Signed-off-by: Grazvydas Ignotas <notasas@gmail.com>

radv: Emit pending flushes before executing a secondary command buffer

If we have any pending flushes on the primary command buffer, these
must be performed before executing the secondary buffer.

This fixes potential corruption when the contents of a subpass which
clears any of its render targets are given in a secondary buffer: the
flushes after a fast clear would not have been performed until the
vkCmdEndRenderPass call.

Signed-off-by: Alex Smith <asmith@feralinteractive.com>
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Cc: 13.0 17.0 <mesa-stable@lists.freedesktop.org>

mesa/main: remove useless check in _mesa_IsSampler()

_mesa_lookup_samplerobj() returns NULL if sampler is 0.

v2: use _mesa_lookup...(...) != NULL

Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Marek Olšák <marek.olsak@amd.com>

getteximage: avoid to lookup textures with id 0

This fixes the following assertion when the key is 0.

main/hash.c:181: _mesa_HashLookup_unlocked: Assertion `key' failed.

Fixes: 633c959fae ("getteximage: Return correct error value when texure object is not found")
Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Marek Olšák <marek.olsak@amd.com>

docs/relnotes/17.1.0: document the new LLVM requirement

gallium/radeon: don't monitor SDMA busyness on EG/Cayman/SI

It's always busy.

Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=99955

Reviewed-by: Dave Airlie <airlied@redhat.com>

radeonsi: drop support for LLVM 3.6 & 3.7

They are too old.

Reviewed-by: Dave Airlie <airlied@redhat.com>

radeonsi: set the convergent attribute where needed

Reviewed-by: Dave Airlie <airlied@redhat.com>

gallivm,ac: add LP_FUNC_ATTR_CONVERGENT

Reviewed-by: Dave Airlie <airlied@redhat.com>

radeonsi: fix LLVM 3.9 - don't use non-matching attributes on declarations

Call site attributes are used since LLVM 4.0.

This also reverts commit b19caecbd6f310c1663b0cfe483d113ae3bd5fe2
"radeon/ac: fix intrinsic version check", because this is the correct fix.

Reviewed-by: Dave Airlie <airlied@redhat.com>

st/omx: Set end-of-frame flag on bitstream output buffers

Since all output buffers are whole frames, this should always be set.

Technically, setting this flag is is optional (see OpenMAX IL section
3.1.2.7.1), but some clients assume that it will be used and
therefore buffer indefinitely thinking that all output buffers are
fragments of the first frame when it is not set.

Signed-off-by: Mark Thompson <sw@jkqxz.net>
Reviewed-by: Christian König <christian.koenig@amd.com>

st/omx: Fix port format enumeration

From OpenMAX IL section 4.3.5:
"The value of nIndex is the range 0 to N-1, where N is the number of
formats supported by the port.  There is no need for the port to
report N, as the caller can determine N by enumerating all the
formats supported by the port.  Each port shall support at least one
format.  If there are no more formats, OMX_GetParameter returns
OMX_ErrorNoMore (i.e., nIndex is supplied where the value is N or
greater)."

Only one format is supported, so N = 1 and OMX_ErrorNoMore should be
returned if nIndex >= 1.  The previous code here would return the
same format for all values of nIndex, resulting in an infinite loop
when a client attempts to enumerate all formats.

Signed-off-by: Mark Thompson <sw@jkqxz.net>
Reviewed-by: Christian König <christian.koenig@amd.com>

st/va: Fix forward/backward referencing for deinterlacing

The VAAPI documentation is not very clear here, but the intent
appears to be that a forward reference is forward from a frame in the
past, not forward to a frame in the future (that is, forward as in
forward prediction, not as in a forward reference in source code).
This interpretation is derived from other implementations, in
particular the i965 driver and the gstreamer client.

In order to match those other implementations, this patch swaps the
meaning of forward and backward references as they currently appear
for motion-adaptive deinterlacing.

Signed-off-by: Mark Thompson <sw@jkqxz.net>
Reviewed-by: Christian König <christian.koenig@amd.com>

st/va: Support fractional framerate in misc parameter

Signed-off-by: Mark Thompson <sw@jkqxz.net>
Acked-by: Christian König <christian.koenig@amd.com>

st/va encode handle ntsc framerate rate control

Tested with ffmpeg and gst-vaapi. Without this bits per
frame is set way too low for fractional framerates.

v2: Mark Thompson: simplify calculation.
    Use float.

Signed-off-by: Andy Furniss <adf.lists@gmail.com>
Acked-by: Christian König <christian.koenig@amd.com>

radv: Use the new L2 writeback flag.

Signed-off-by: Bas Nieuwenhuizen <basni@google.com>
Reviewed-by: Dave Airlie <airlied@redhat.com>

radv: Add L2 writeback.

Signed-off-by: Bas Nieuwenhuizen <basni@google.com>
Reviewed-by: Dave Airlie <airlied@redhat.com>

util/disk_cache: fix make check

Fixes make check after 11f0efec2e615f5233d which caused disk cache
to create an additional directory.

radv/ac: use bitfield extract new intrinsics.

Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Signed-off-by: Dave Airlie <airlied@redhat.com>

radv/ac: move to new kill build.

Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Signed-off-by: Dave Airlie <airlied@redhat.com>

radv/ac: move to using new export intrinsics.

This uses the new code in build to do exports.

Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Signed-off-by: Dave Airlie <airlied@redhat.com>

radv/ac: switch to new intrinsics for pkrtz and clamp.

Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Signed-off-by: Dave Airlie <airlied@redhat.com>

radv: drop Z24 support.

This isn't exposed in -pro, the hw docs say it is deprecated,
so let's not bother with it.

Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Cc: "13.0 17.0" <mesa-stable@lists.freedesktop.org>
Signed-off-by: Dave Airlie <airlied@redhat.com>

radv: use VK_NULL_HANDLE for handles

Avoids warnings on 32bit.

Signed-off-by: Grazvydas Ignotas <notasas@gmail.com>
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>

radv: check for upload alloc failure

Mainly to avoid gcc's complains about uninitialized ptr and offset use
later in that code.

Signed-off-by: Grazvydas Ignotas <notasas@gmail.com>
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>

radv: don't use uninitialized value on failure

Mainly to avoid a warning.

Signed-off-by: Grazvydas Ignotas <notasas@gmail.com>
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>

radv: avoid casting warnings on 32bit

Use the same helpers as for other handle<->pointer conversions.

Signed-off-by: Grazvydas Ignotas <notasas@gmail.com>
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>

radv/amdgpu: Add some debug flags.

Signed-off-by: Bas Nieuwenhuizen <basni@google.com>
Reviewed-by: Dave Airlie <airlied@redhat.com>

radv: Cache command buffers in command pool.

So that we don't keep allocating BOs for the IBs and upload buffers.

We run some risk of memory increase with e.g. a bimodal size
distribution of command buffers, but I haven't noticed a significant
increase with dota2 and talos.

Signed-off-by: Bas Nieuwenhuizen <basni@google.com>
Reviewed-by: Dave Airlie <airlied@redhat.com>

Revert "glsl: Switch to disable-by-default for the GLSL shader cache"

This reverts commit 0f60c6616e93cba72bff4fbfedb72a753ef78e05.

Piglit and all games tested so far seem to be working without
issue. This change will allow wide user testing and we can decided
before the next release if we need to turn it off again.
Reviewed-by: Marek Olšák <marek.olsak@amd.com>

docs: update envvars.html to reflect having a cache per arch

util/disk_cache: support caches for multiple architectures

Previously we were deleting the entire cache if a user switched
between 32 and 64 bit applications.

V2: make the check more generic, it should now work with any
platform we are likely to support.

V3: Use suggestion from Emil to make even more generic/fix issue
with __ILP32__ not being declared on gcc for regular 32-bit builds.

Tested-by: Grazvydas Ignotas <notasas@gmail.com>
Tested-by: Dieter Nützel <Dieter@nuetzel-hh.de>

util/disk_cache: mark read-only arguments const

No functional changes.

Signed-off-by: Grazvydas Ignotas <notasas@gmail.com>
Reviewed-by: Timothy Arceri <tarceri@itsqueeze.com>

radeon/ac: fix intrinsic version check

Reported-by: 375gnu@gmail.com
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=100068

Signed-off-by: Dave Airlie <airlied@redhat.com>

radv: Merge fast clear flushes.

Don't flush multiple times if we clear multiple attachments. Also allows
doing the depth clear in parallel with the fast color clears.

Signed-off-by: Bas Nieuwenhuizen <basni@google.com>
Reviewed-by: Dave Airlie <airlied@redhat.com>

relnotes: [swr] note addition of gs, increased llvm requirement

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

docs: update features.txt for swr geometry shaders

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

swr: [rasterizer core] fix primID provoking vertex for GS

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

swr: implement geometry shaders

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

configure.ac: increase required swr llvm to 3.9.0

GS implementation uses the masked.{gather,store} intrinsics,
introduced in llvm-3.9.0.  swr llvm version requirement in
automake and scons now match (scons already needed >= 3.9).

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

i965: Clamp texture buffer size to GL_MAX_TEXTURE_BUFFER_SIZE.

The OpenGL 4.5 specification's description of TexBuffer says:

"The number of texels in the texture image is then clamped to an
 implementation-dependent limit, the value of MAX_TEXTURE_BUFFER_SIZE."

We set GL_MAX_TEXTURE_BUFFER_SIZE to 2^27.  For buffers with a byte
element size, this is the maximum possible size we can encode in
SURFACE_STATE.  If you bind a buffer object larger than this as a
texture buffer object, we'll exceed that limit and hit an isl assert:

   assert(num_elements <= (1ull << 27));

To fix this, clamp the size in bytes to MaxTextureSize / texel_size.

Signed-off-by: Kenneth Graunke <kenneth@whitecape.org>
Reviewed-by: Anuj Phogat <anuj.phogat@gmail.com>

automake: move wayland-drm prior to Vulkan

Earlier commit was picked from a larger series, but did not consider
that it removed the vulkan <> wayland-drm interdependency.

Rather than reverting everything, temporarily move wayland-drm further
up to resolve the issue. Since it [wayland-drm] does not have any
in-mesa dependencies that's perfectly safe.

Cc: Vedran Miletić <vedran@miletic.net>
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=100060
Fixes: e135ce6f088 ("vulkan: Build common Vulkan code earlier")
Signed-off-by: Emil Velikov <emil.l.velikov@gmail.com>
Tested-by: Javier Jardón <jjardon@gnome.org>

android: fix libz dynamic library dependencies

Fixes a series of libz related building errors:

target SharedLib: gallium_dri_32
(out/target/prod...SHARED_LIBRARIES/gallium_dri_intermediates/LINKED/gallium_dri.so)
external/elfutils/libelf/elf_compress.c:117: error: undefined reference to 'deflateInit_'
...
external/elfutils/libelf/elf_compress.c:244: error: undefined reference to 'inflateEnd'
clang++: error: linker command failed with exit code 1 (use -v to see
invocation)

Fixes: 85a9b1b "util/disk_cache: compress individual cache entries"

svga: pass NULL to ureg_get_tokens()

The number of tokens in never used and the pointer is NULL checked
so just pass NULL.

Reviewed-by: Brian Paul <brianp@vmware.com>

nvc0: take extra pushbuf space into account for pushbuf_space calls

See detailed explanation of why this is needed in commit eb60a89bc3a.
This spot was missed/overlooked. Basically as a result of the fact
that BEGIN_* ends up calling PUSH_SPACE, which in turn adds an extra 8
to the requested amount, we have to be mindful of that when doing bare
nouveau_pushbuf_space calls.

Reportedly this fixes some crashes when replaying a hitman trace taken
on radeonsi.

Fixes: eb60a89bc3a ("nouveau: take extra push space into account for pushbuf_space calls")
Cc: "13.0 17.0" <mesa-stable@lists.freedesktop.org>
Reported-by: Karol Herbst <nouveau@karolherbst.de>
Signed-off-by: Ilia Mirkin <imirkin@alum.mit.edu>
Reviewed-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>

nvc0: increase alignment to 256 for texture buffers on fermi

When binding as textures, the alignment can be 16. However when binding
as an image, the address has to be aligned to 256. (Also when binding as
an RT, but that can't happen with GL or current gallium APIs.)

Reported-by: Roy Spliet <nouveau@spliet.org>
Signed-off-by: Ilia Mirkin <imirkin@alum.mit.edu>
Acked-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>

android: fix outdir for gen_enum_to_str files

when files are being generated the value of $intermediates var content can be
completely random, this makes sure that outdir is the wanted one.

Fixes: 3f2cb699 ("android: vulkan: add support for libmesa_vulkan_util")
Signed-off-by: Tapani Pälli <tapani.palli@intel.com>
Reviewed-by: Eric Engestrom <eric.engestrom@imgtec.com>
Acked-by: Emil Velikov <emil.velikov@collabora.com>

EGL/Android: Add EGL_EXT_buffer_age extension

This patch implements the EGL_EXT_buffer_age extension for Android.
https://www.khronos.org/registry/EGL/extensions/EXT/EGL_EXT_buffer_age.txt

Reviewed-by: Tapani Pälli <tapani.palli@intel.com>
Acked-by: Emil Velikov <emil.velikov@collabora.com>

docs: add news item and link release notes for 17.0.1

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

docs: add sha256 checksums for 17.0.1

Signed-off-by: Emil Velikov <emil.velikov@collabora.com>
(cherry picked from commit 5c9273152c59777771fa6c7b546316caf3f091d8)

docs: add release notes for 17.0.1

Signed-off-by: Emil Velikov <emil.velikov@collabora.com>
(cherry picked from commit 8fee1d348cc3d91a88319c0d72689acabaa2bf47)

gallium/targets: don't leave an empty target directory(ies)

Some drivers do not support certain targets - for example nouveau
doesn't do VAAPI, while freedreno doesn't do of the video backends.

As such if we enter vdpau when building freedreno/ilo/etc, a vdpau/
folder will be created, empty library will be build and almost
immediately removed. Thus keeping an empty vdpau/ folder around.

There are two ways to fix this.

 * add substantial tracking in configure/makefiles so that we never end
up in targets/vdpau
 Downsides:
Error prone, as the configure checks and the 'include
gallium/drivers/foo/Automake.inc' can easily get out of sync.

 * remove the folder, if empty, alongside the empty library.
 Downsides:
In the latter case vdpau/ might be empty before the mesa build has
started, yet we'll remove it either way.

This patch implements the latter option, as the downside isn't that
significant, plus the patch is way shorter ;-)

v2: use has_drivers to track since TARGET_DRIVERS can contain space,
hence neither string comparison nor -n/-z works correctly.

Gentoo Bugzilla: https://bugs.gentoo.org/545230
Signed-off-by: Emil Velikov <emil.l.velikov@gmail.com>
Reviewed-by: Matt Turner <mattst88@gmail.com>

radv: use enum_to_str util functions.

Port of e9dcb17962f7e58a81c93bae7bd33885675b1043
vulkan/util: Add generator for enum_to_str functions

Cc: Bas Nieuwenhuizen <basni@google.com>
Signed-off-by: Emil Velikov <emil.l.velikov@gmail.com>
Reviewed-by: Eric Engestrom <eric.engestrom@imgtec.com>

vulkan: Build common Vulkan code earlier

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

anv: Advertise shaderInt64 on Broadwell and above

Reviewed-by: Samuel Iglesias Gonsálvez <siglesias@igalia.com>

nir/int64: Properly handle imod/irem

The previous implementation was fine for GLSL which doesn't really have
a signed modulus/remainder.  They just leave the behavior undefined
whenever either source is negative.  However, in SPIR-V, there is a
defined behavior for negative arguments.  This commit beefs up the pass
so that it handles both correctly.  Tested using a hacked up version of
the Vulkan CTS test to get 64-bit support.

Reviewed-by: Matt Turner <mattst88@gmail.com>

nir/builder: Add an int64 immediate helper

Reviewed-by: Matt Turner <mattst88@gmail.com>

genxml: Fill out Gen4 and G45 XML.

This is a work in progress - some things may still need fixing.
But it should be in pretty decent shape.

Signed-off-by: Kenneth Graunke <kenneth@whitecape.org>
Signed-off-by: Jason Ekstrand <jason@jlekstrand.net>

ac: normalize build helper names

s/emit/build/

Reviewed-by: Dave Airlie <airlied@redhat.com>

ac: replace SI.vs.load.input with amdgcn.buffer.load.format

Reviewed-by: Dave Airlie <airlied@redhat.com>

radeonsi: move SI.vs.load.input building into amd/common

Reviewed-by: Dave Airlie <airlied@redhat.com>

radeonsi: detect and mark loads/stores from read-only/write-only memory

ac: replace llvm.SI.tbuffer.store with llvm.amdgcn.buffer.store if ADD_TID=0

ADD_TID doesn't work. Needs more investigation.

v2: remove leftover dead code

Reviewed-by: Dave Airlie <airlied@redhat.com> (v1)

radeonsi: use the writeonly LLVM attribute

ac: remove offen parameter from ac_build_buffer_store_dword

Reviewed-by: Dave Airlie <airlied@redhat.com>

radeonsi: enable TC L2 for tessellation offchip stores

Vulkan does the same thing.

radeonsi: merge and simplify tbuffer_store functions

Reviewed-by: Dave Airlie <airlied@redhat.com>

radeonsi: set noalias on input shader pointers

radeonsi: replace AMDGPU.bfe.* with amdgcn.*bfe

Reviewed-by: Dave Airlie <airlied@redhat.com>

radeonsi: move kill intrinsic building into amd/common

just a cleanup

Reviewed-by: Dave Airlie <airlied@redhat.com>

radeonsi: set readnone on reads from read-only memory

radeonsi: replace SI.buffer.load.dword with amdgcn.buffer.load

radeonsi: replace SI.packf16 with amdgcn.cvt.pkrtz

ac: replace old image intrinsics with new ones

Reviewed-by: Dave Airlie <airlied@redhat.com>

radeonsi: remove last use of llvm.SI.resinfo

and move one function up to reuse the code.

radeonsi: move image intrinsic building to amd/common

Reviewed-by: Dave Airlie <airlied@redhat.com>

ac: replace SI.export with amdgcn.exp.*

Reviewed-by: Dave Airlie <airlied@redhat.com>

radeonsi: move llvm.SI.export building to amd/common

Reviewed-by: Dave Airlie <airlied@redhat.com>

ac: unify build_type_name_for_intr functions

Reviewed-by: Dave Airlie <airlied@redhat.com>

radeonsi: set unorm=1 for TGSI_TEXTURE_SHADOWRECT as well

It was harmless, because we also set unorm in the sampler state.

Reviewed-by: Dave Airlie <airlied@redhat.com>

gallivm, ac: add writeonly and inaccessiblememonly attributes

Reviewed-by: Dave Airlie <airlied@redhat.com>

tgsi/scan: record load/store/atomic image usage

Reviewed-by: Dave Airlie <airlied@redhat.com>

glapi: Fix a comment typo

Reviewed-by: Nicolai Hähnle <nicolai.haehnle@amd.com>
Reviewed-by: Timothy Arceri <tarceri@itsqueeze.com>

mesa/main: *TextureSubImage* generates INVALID_OPERATION on wrong target

Equivalent *TexSubImage* methods generates INVALID_ENUM.

From OpenGL 4.5 spec, section 8.6 Alternate Texture Image
Specification Commands:

   "An INVALID_ENUM error is generated by *TexSubImage* if target does
    not match the command, as shown in table 8.15."

And:

   "An INVALID_OPERATION error is generated by *TextureSubImage* if
    the effective target of texture does not match the command, as
    shown in table 8.15."

Fixes:
GL45-CTS.direct_state_access.textures_copy_errors

v2: slightly change commit summary (Samuel)

Reviewed-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>

i965: Add Kaby Lake brandstrings

While here, use the spacing defined in Ark.
https://ark.intel.com/products/codename/82879/Kaby-Lake

Reviewed-by: Jason Ekstrand <jason@jlekstrand.net>
Signed-off-by: Ben Widawsky <ben@bwidawsk.net>

tgsi/ureg: return correct token count in ureg_get_tokens

Valgrind reports that the shader cache writes uninitialized data to disk.
Turns out ureg_get_tokens() is returning the count of allocated tokens
instead of how many are actually used, so the cache writes out unused
space at the end. Use the real count instead.

This change should not cause regressions elsewhere because the only
ureg_get_tokens() user that cares about token count is the shader cache.

Signed-off-by: Grazvydas Ignotas <notasas@gmail.com>
Reviewed-by: Timothy Arceri <tarceri@itsqueeze.com>

radeonsi: add support for an on-disk shader cache

V2:
- when loading from disk cache also binary insert into memory cache.
- check that the binary loaded from disk is the correct size. If not
  delete the cache item and skip loading from cache.

V3:
- remove unrequired variable

Reviewed-by: Grigori Goronzy <greg@chown.ath.cx>
Reviewed-by: Marek Olšák <marek.olsak@amd.com>

util/disk_cache: compress individual cache entries

This reduces the cache size for Deus Ex from ~160M to ~30M for
radeonsi (these numbers differ from Grigori's results below
probably due to different graphics quality settings).

I'm also seeing the following improvements in minimum fps in the
Shadow of Mordor benchmark on an i5-6400 CPU@2.70GHz, with a HDD:

no-cache:                    ~10fps
with-cache-no-compression:   ~15fps
with-cache-and-compression:  ~20fps

Note: The with cache results are from the second run after closing
and opening the game to avoid the in-memory cache.

Since we mainly care about decompression I went with
Z_BEST_COMPRESSION as suggested on irc by Steinar H. Gunderson
who has benchmarked decompression speeds.

Grigori Goronzy provided the following stats for Deus Ex: Mankind
Divided start-up times on a Athlon X4 860k with a SSD:

No Cache                                 215 sec

Cold Cache zlib BEST_COMPRESSION         285 sec
Warm Cache zlib BEST_COMPRESSION         33 sec

Cold Cache zlib BEST_SPEED               264 sec
Warm Cache zlib BEST_SPEED               33 sec

Cold Cache no compression                266 sec
Warm Cache no compression                34 sec

The total cache size for that game is 48 MiB with BEST_COMPRESSION,
56 MiB with BEST_SPEED and 170 MiB with no compression.

These numbers suggest that it may be ok to go with Z_BEST_SPEED
but we should gather some actual decompression times before doing
so. Other options might be to do the compression in a separate
thread, this might allow us to use a higher compression algorithim
such as LZMA.

Reviewed-by: Grigori Goronzy <greg@chown.ath.cx>
Acked-by: Marek Olšák <marek.olsak@amd.com>

util/disk_cache: add support for detecting corrupt cache entries

V2: fix pointer increments for writing/reading crc

Acked-by: Marek Olšák <marek.olsak@amd.com>
Reviewed-by: Grigori Goronzy <greg@chown.ath.cx>

glsl: fix subroutine mismatch between declarations/definitions

Previously, when q.subroutine was set to 1, a new subroutine
declaration was added to the AST, while 0 meant a subroutine
definition has been detected by the parser.

Thus, setting the q.subroutine flag in both situations is
obviously wrong because a new type identifier is added instead
of trying to match the declaration. To fix it up, introduce
ast_type_qualifier::is_subroutine_decl() to differentiate
declarations and definitions easily.

This fixes a regression with:
arb_shader_subroutine/compiler/direct-call.vert

Cc: Mark Janes <mark.a.janes@intel.com>
Fixes: be8aa76afd ("glsl: remove unecessary flags.q.subroutine_def")
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=100026
Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Timothy Arceri <tarceri@itsqueeze.com>

genxml: Depend on Makefile.am for generated sources.

Depending on the generated Makefile means that all generated sources are
recreated after ./configure.

Reviewed-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>

clover: Work around build failure with AltiVec.

Bugzilla: https://bugs.gentoo.org/show_bug.cgi?id=587210
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=68504
Acked-by: Francisco Jerez <currojerez@riseup.net>

anv/image: Allow HiZ on input attachment-capable depth/stencil images

While an input attachment may only take on one of those two layouts,
other depth/stencil attachments that use the same image may have
HiZ-enabled layouts. Improves the average frame rate on a release
candidate of a proprietary Vulkan benchmark by 9.94% over 3 runs on my
SKL GT4.

Signed-off-by: Nanley Chery <nanley.g.chery@intel.com>
Reviewed-by: Jason Ekstrand <jason@jlekstrand.net>

anv/cmd_buffer: Centralize automatic layout transitions

Signed-off-by: Nanley Chery <nanley.g.chery@intel.com>
Reviewed-by: Jason Ekstrand <jason@jlekstrand.net>

anv/cmd_buffer: Add attachment transitioning functions

This is needed to transition input attachments.

Signed-off-by: Nanley Chery <nanley.g.chery@intel.com>
Reviewed-by: Jason Ekstrand <jason@jlekstrand.net>

anv/blorp: Encapsulate subpass id querying

Signed-off-by: Nanley Chery <nanley.g.chery@intel.com>
Reviewed-by: Jason Ekstrand <jason@jlekstrand.net>