gallivm: add compute jit interface.

This adds the jit interface for compute shaders, it's based
on the fragment shader one.

Reviewed-by: Roland Scheidegger <sroland@vmware.com>

llvmpipe: add initial compute state structs

These mirror the fragment shader structs, this is just a framework.

Reviewed-by: Roland Scheidegger <sroland@vmware.com>

llvmpipe: introduce compute shader context

The compute shader will need it's own context like the frag shader
has, this just introduces the framework struct and allocates/frees
for it in the right places.

Reviewed-by: Roland Scheidegger <sroland@vmware.com>

gallivm: add barrier support for compute shaders.

When the code is executing an hits a barrier, it will suspend
the coroutine and return control to the coroutine dispatcher.

Reviewed-by: Roland Scheidegger <sroland@vmware.com>

llvmpipe: add compute threadpool + mutex

Reviewed-by: Roland Scheidegger <sroland@vmware.com>
In order to efficiently run a number of compute blocks, use
a threadpool that just allows for jobs with unique sequential
ids to be dispatched.

gallivm: add support for compute shared memory

Reviewed-by: Roland Scheidegger <sroland@vmware.com>

gallivm: add new compute related intrinsics

Reviewed-by: Roland Scheidegger <sroland@vmware.com>

llvmpipe: reogranise jit pointer ordering

In order to share the texture/image/sampler code with compute
shaders we need to reorg them to be at the front of context
same as draw does for vs/gs sharing.

Reviewed-by: Roland Scheidegger <sroland@vmware.com>

gallivm: add coroutine pass manager support

coroutines require a proper pass manager, so add the passes
to the correct places

Reviewed-by: Roland Scheidegger <sroland@vmware.com>

gallivm: add coroutine support files to gallivm.

These wrap the coroutine intrinsics and also add some higher
level wrappers around coroutine begin, end and suspend procedures

Reviewed-by: Roland Scheidegger <sroland@vmware.com>

gallivm/flow: add counter reset for loops

This allows the counter value to be forced to a certain value

Reviewed-by: Roland Scheidegger <sroland@vmware.com>

llvmpipe: enable fb no attach

Reviewed-by: Roland Scheidegger <sroland@vmware.com>

iris: Report correct number of planes for planar images

We were only handling the modifiers case and not counting the number of
planes in actual planar images.

Fixes Piglit's ext_image_dma_buf_import-export.

Fixes: fc12fd05f56 ("iris: Implement pipe_screen::resource_get_param")
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=111509
Reviewed-by: Jordan Justen <jordan.l.justen@intel.com>

teximage: ensure that Tex*SubImage* checks format

We were previously not doing at least some of the checks. This uses the
same logic that is used in glTexImage*.

Signed-off-by: Ilia Mirkin <imirkin@alum.mit.edu>
Reviewed-by: Dave Airlie <airlied@redhat.com>
Reviewed-by: Marek Olšák <marek.olsak@amd.com>

gallium/hud: add CPU usage support for DragonFly/NetBSD/OpenBSD

Each BSD has slightly different sysctl for retrieving per-CPU times.
FreeBSD returns long while NetBSD returns uint64_t. On OpenBSD return
type differs between summation and per-CPU times. DragonFly is
compatible with FreeBSD.

Signed-off-by: Jan Beich <jbeich@FreeBSD.org>

lima: Return fence unconditionally

Based on the vc4 implementation.
Fixes Android RenderEngine::flush() routine:
android.googlesource.com/platform/frameworks/native/+/refs/tags/android-o-mr1-iot-release-smart-clock-fcs/services/surfaceflinger/RenderEngine/RenderEngine.cpp#225

Signed-off-by: Roman Stratiienko <roman.stratiienko@globallogic.com>
Reviewed-by: Qiang Yu <yuq825@gmail.com>

lima/ppir: clone uniforms and load_coords into each successor

Try more aggressive approach with cloning uniform and coord loads.

Uniform load can be inserted into any instruction, so let's do that. ARM site
claim that penalty for cache miss is one clock, so we don't lose anything if
we merge it into instruction that uses the result. As side effect we can also
pipeline it and thus decrease reg pressure.

Do the same for varyings that hold texture coords, but for different reason:
looks like there's a special path for coords that increases precision if
varying that holds it is pipelined. If we don't pipeline it and load coords
from a register its precision is fp16 and thus only 10 bits which is not enough
to accurately sample textures of size 1024 or larger.

Since instruction can hold only one uniform load and one varying load,
node_to_instr now creates a move using helper introduced in previous commit if
slot is already taken. As side effect of this change we can also try to
pipeline texture loads and create a move if attempt fails.

Reviewed-by: Erico Nunes <nunes.erico@gmail.com>
Signed-off-by: Vasily Khoruzhick <anarsoul@gmail.com>

lima/ppir: don't assume that load coords gets value from register

It can load value from varying directly as well. Also load_regs is the
only op that has a source, so add src_num field to load node and set it
accordingly.

Reviewed-by: Erico Nunes <nunes.erico@gmail.com>
Signed-off-by: Vasily Khoruzhick <anarsoul@gmail.com>

lima/ppir: add common helper for creating movs

Introduce common helper for creating movs to avoid code duplication

Reviewed-by: Erico Nunes <nunes.erico@gmail.com>
Signed-off-by: Vasily Khoruzhick <anarsoul@gmail.com>

nir: fix memleak in error path

Fixes: 2cf59861a8128a91bfdd ("nir: Add partial redundancy elimination for compares")
Signed-off-by: Eric Engestrom <eric.engestrom@intel.com>
Reviewed-by: Timothy Arceri <tarceri@itsqueeze.com>
Reviewed-by: Ian Romanick <ian.d.romanick@intel.com>

freedreno/drm-shim: fix mem leak

Fixes: 494ecef6b42198ab6c3e ("freedreno: Add support for drm-shim.")
Signed-off-by: Eric Engestrom <eric.engestrom@intel.com>
Reviewed-by: Eric Anholt <eric@anholt.net>

anv: fix format string in error message

Fixes: 9775894f102535a79186 ("anv: Move size check from anv_bo_cache_import() to caller (v2)")
Signed-off-by: Eric Engestrom <eric.engestrom@intel.com>
Reviewed-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>

util/os_file: fix double-close()

Fixes: 955c63d3643f30d7db0c ("util/os_file: resize buffer to what was actually needed")
Signed-off-by: Eric Engestrom <eric.engestrom@intel.com>
Reviewed-by: Tapani Pälli <tapani.palli@intel.com>

egl: fix deadlock in malloc error path

Fixes: cb0980e69aa921af7086 ("egl: move alloc & init out of _eglBuiltInDriver{DRI2,Haiku}")
Signed-off-by: Eric Engestrom <eric.engestrom@intel.com>
Reviewed-by: Tapani Pälli <tapani.palli@intel.com>

ttn: fix 64-bit shift on 32-bit `1`

Fixes: 4d0b2c7aaac3cf3de5af ("ttn: Update shader->info as we generate code.")
Signed-off-by: Eric Engestrom <eric.engestrom@intel.com>
Reviewed-by: Tapani Pälli <tapani.palli@intel.com>
Reviewed-by: Eric Anholt <eric@anholt.net>
Reviewed-by: Rob Clark <robdclark@gmail.com>

freedreno/ir3: use uniform base

When lowering from ubo, use the constant base field in the load_uniform
instruction for the constant part of the offset.  Doesn't change much
for constant indexing, but this will help for indirect indexing because
constant-folding can't completely clean up the result.

Signed-off-by: Rob Clark <robdclark@chromium.org>
Reviewed-by: Eric Anholt <eric@anholt.net>

freedreno/drm: fix 64b iova shifts

Should shift before splitting 64b iova into dwords

Signed-off-by: Rob Clark <robdclark@chromium.org>
Reviewed-by: Eric Anholt <eric@anholt.net>

nir: remove unused constant_fold_state

Signed-off-by: Rob Clark <robdclark@chromium.org>
Reviewed-by: Eric Anholt <eric@anholt.net>

freedreno: Fix the type of single-component scaled vertex attrs.

This looks like clear copy-and-pasteos, and fixes:

dEQP-GLES2.functional.draw.random.40

(on A307 and A630, both tested in the new CI farm)

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

radeonsi/nir: Remove uniform variable scanning

We can get all the information we need from NIR. It's slightly less
accurate, but radeonsi doesn't use the extra information. The old code
also overcounted atomic counters, which led to problems when everything
was used at once.

Fixes KHR-GL45.compute_shader.resources-max.

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

ttn: Fill out more info fields

We'll use these in radeonsi.

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

nir: Fix num_ssbos when lowering atomic counters

Otherwise it's impossible to know the maximum SSBO index for both
internal TGSI shaders from TTN (which don't have any notion of atomic
counters and no offset) as well as shaders from GLSL.

I fixed everything I could find while grepping for num_ssbos and
num_abos, which hopefully is everything (iris was the only user I could
find that uses it in a meaningful way).

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

ac/nir: Fix gather4 integer wa with unnormalized coordinates

This adds a bit of unneccesary code on radeonsi, since whether
unnormalized coordinates are used is known at compile time with GL, but
I wasn't sure if it was worth the few instructions to plumb everything
through, especially for something so rare -- my shader-db doesn't have
any instances where this changes anything.

Fixes CTS tests I created at
https://github.com/cwabbott0/VK-GL-CTS/tree/unnorm-gather-tests

Acked-by: Marek Olšák <marek.olsak@amd.com>
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>

ac/nir: Rewrite gather4 integer workaround based on radeonsi

The workaround was originally written based on amdgpu-pro traces, but
since then radeonsi has got its own slightly different version. Use the
radeonsi version instead, to be consistent and because it'll be slightly
more convenient for handling unnormalized coordinates.

Acked-by: Marek Olšák <marek.olsak@amd.com>
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>

egl: warn user if they set an invalid EGL_PLATFORM

Technically, the user might have set EGL_DISPLAY instead of
EGL_PLATFORM, but since the former is deprecated let's just mention the
latter in the warning message.

Signed-off-by: Eric Engestrom <eric.engestrom@intel.com>
Reviewed-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>

panfrost: Remove panfrost_upload

This routine was made obsolete over a series of reworks of memory
allocation; Tomeu's changes to shader memory allocation finally made
this unused as cppcheck noted.

Signed-off-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
Reviewed-by: Tomeu Vizoso <tomeu.vizoso@collabora.com>

panfrost: Fix misc. issues flagged by cppcheck

Signed-off-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
Reviewed-by: Tomeu Vizoso <tomeu.vizoso@collabora.com>

panfrost: Mark (1 << 31) as unsigned

I was not aware this incurred undefined behaviour; thank you cppcheck.

Signed-off-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
Reviewed-by: Tomeu Vizoso <tomeu.vizoso@collabora.com>

pan/midgard: Remove mir_rewrite_index_*_tag

These helpers are unused, as flagged by cppcheck.

Signed-off-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
Reviewed-by: Tomeu Vizoso <tomeu.vizoso@collabora.com>

pan/midgard: Remove mir_print_bundle

In practice, the new post-schedule print is just as useful.

Signed-off-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
Reviewed-by: Tomeu Vizoso <tomeu.vizoso@collabora.com>

pan/midgard: Remove cppwrap.cpp

It has not been used in a long time; I forgot this file even existed.
Flagged by cppcheck.

Signed-off-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
Reviewed-by: Tomeu Vizoso <tomeu.vizoso@collabora.com>

pan/midgard: Fix cppcheck issues

Miscellaneous minor issues flagged by cppcheck.

Signed-off-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
Reviewed-by: Tomeu Vizoso <tomeu.vizoso@collabora.com>

pan/midgard: Correct issues in disassemble.c

cppcheck.

Signed-off-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
Reviewed-by: Tomeu Vizoso <tomeu.vizoso@collabora.com>

pan/decode: Add missing format specifier

Signed-off-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
Reviewed-by: Tomeu Vizoso <tomeu.vizoso@collabora.com>

pan/decode: Use portable format specifier for 64-bit

Signed-off-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
Reviewed-by: Tomeu Vizoso <tomeu.vizoso@collabora.com>

pan/decode: Use %zu instead of %d

Signed-off-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
Reviewed-by: Tomeu Vizoso <tomeu.vizoso@collabora.com>

pan/decode: Fix uninitialized variables

Signed-off-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
Reviewed-by: Tomeu Vizoso <tomeu.vizoso@collabora.com>

docs: update calendar, add news item and link release notes for 19.1.6

Signed-off-by: Juan A. Suarez Romero <jasuarez@igalia.com>

docs: add sha256 checksums for 19.1.6

Signed-off-by: Juan A. Suarez Romero <jasuarez@igalia.com>
(cherry picked from commit 4ec2325dd07a768f2b52ea788ee76085586b2469)

docs: add release notes for 19.1.6

Signed-off-by: Juan A. Suarez Romero <jasuarez@igalia.com>
(cherry picked from commit 85c8f88a49aa7c8aa866faed90a4a63330c15b8b)

vulkan/overlay: bounce image back to present layout

Once we write the overlay to an image to be presented, we must not
forget to put it back into present layout.

Signed-off-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=111401
Cc: <mesa-stable@lists.freedesktop.org>
Reviewed-by: Tapani Pälli <tapani.palli@intel.com>

broadcom/vc4: Expand width of dst surface

Four bytes of src_surf will be compressed into a 32-bits data and
stored into dst_surf, and dst_surf is read as z-order, so its width
must be aligned to multiples of 8(4x2) before divided by 2.

Signed-off-by: Zhaowei Yuan <zhaowei.yuan@samsung.com>
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=111266

Reviewed-by: Alejandro Piñeiro <apinheiro@igalia.com>

swr: Fix make_unique build error.

swr_shader.cpp: In function ‘void (* swr_compile_gs(swr_context*, swr_jit_gs_key&))(HANDLE, HANDLE, SWR_GS_CONTEXT*)’:
swr_shader.cpp:732:44: error: ‘make_unique’ was not declared in this scope
    ctx->gs->map.insert(std::make_pair(key, make_unique<VariantGS>(builder.gallivm, func)));
                                            ^~~~~~~~~~~

Signed-off-by: Vinson Lee <vlee@freedesktop.org>
Reviewed-by: Jan Zielinski <jan.zielinski@intel.com>

loader: include limits.h for PATH_MAX

This is needed to build on illumos.

The location of the PATH_MAX definition in limits.h seems to be fairly standard:
https://pubs.opengroup.org/onlinepubs/009695399/basedefs/limits.h.html

Reviewed-by: Eric Engestrom <eric@engestrom.ch>

util: only allow _BitScanReverse64 on 64-bit cpus

While the documentation for _BitScanReverse64 on MSDN says that it's
available on ARM, this isn't true. It's only available on ARM64. So
let's match reality.

Signed-off-by: Erik Faye-Lund <erik.faye-lund@collabora.com>
Acked-by: Matt Turner <mattst88@gmail.com>

mesa/x86: improve SSE-checks for MSVC

This enables some more SSE optimizations on MSVC builds.

Signed-off-by: Erik Faye-Lund <erik.faye-lund@collabora.com>
Reviewed-by: Matt Turner <mattst88@gmail.com>

util: do not assume MSVC implies SSE

This is not true for MSVC on ARM.

Signed-off-by: Erik Faye-Lund <erik.faye-lund@collabora.com>
Reviewed-by: Matt Turner <mattst88@gmail.com>

util: fix SSE-version needed for double opcodes

This code generates CVTSD2SI, which requires SSE2. So let's fix the
required SSE-version.

Signed-off-by: Erik Faye-Lund <erik.faye-lund@collabora.com>
Fixes: 5de29ae (util: try to use SSE instructions with MSVC and 32-bit gcc)
Reviewed-by: Matt Turner <mattst88@gmail.com>

mesa/main: remove unused include

This has been unused since 183db3a6455 ("glsl: move half<->float
convertion to util"), Oct 10 2015. Let's drop needlessly including it.

Signed-off-by: Erik Faye-Lund <erik.faye-lund@collabora.com>
Reviewed-by: Matt Turner <mattst88@gmail.com>

nir: do not assume that the result of fexp2(a) is always an integral

It's only correct when 'a' is an integral greater or equal to 0.

Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=111493
Fixes: 5544b2cbbd2 ("nir/algebraic: Use value range analysis to eliminate useless unary ops")
Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Ian Romanick <ian.d.romanick@intel.com>

egl: fix platform selection

Add missing "device" platform

v2: Add the missing platform (Eric)

Signed-off-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Reported-by: Jean Hertel <jean.hertel@hotmail.com>
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=111529
Fixes: d6edccee8d ("egl: add EGL_platform_device support")
Reviewed-by: Eric Engestrom <eric@engestrom.ch>

iris: Lessen texture cache hack flush for blits/copies on Icelake.

Lionel found actual documentation for this at long last.  Apparently
it actually is a sampler cache limitation that was mostly fixed on
Icelake.  Unfortunately, it seems there are still issues with ASTC
and non-ASTC sampler views.  Still, we can lessen the flush condition
from "format mismatch" to "ASTC mismatch", which eliminates most of
the flushing here.

We also update the documentation to refer to the workaround name.

util: Define strchrnul on macOS.

strchrnul is not available on macOS.

pipe_loader.c:141:14: error: implicit declaration of function 'strchrnul' is invalid in C99 [-Werror,-Wimplicit-function-declaration]
      next = strchrnul(library_paths, ':');
             ^

Signed-off-by: Vinson Lee <vlee@freedesktop.org>
Acked-by: Eric Engestrom <eric.engestrom@intel.com>

gallium/auxiliary/indices: consistently apply start only to input

The majority of these only apply the start argument to the input, but a
few of them also does for the output-array. util_primconvert, the only
user of this argument expects this pass a non-zero start-argument does
not expect this to be applied to the output; if it is, it will write
outside of allocated memory, leading to VRAM corruption.

The reason this doesn't seem to have been noticed before, is that no
driver currently use util_primconvert to convert a primitive-type to
itself, which is the cases where this was broken. But for Zink, this
will no longer be true, because we need to eliminate the use of 8-bit
index-buffers.

Signed-off-by: Erik Faye-Lund <erik.faye-lund@collabora.com>
Fixes: 28f3f8d413f ("gallium/auxiliary/indices: add start param")
Reviewed-by: Rob Clark <robdclark@chromium.org>

travis: Fail build if any command in if statement fails.

Travis is checking the exit code of the entire if statement.

Fixes: 64ffc289be89 ("travis: add MacOS Scons build")
Signed-off-by: Vinson Lee <vlee@freedesktop.org>
Acked-by: Eric Engestrom <eric@engestrom.ch>
Reviewed-by: Jose Fonseca <jfonseca@vmware.com>

swr: Fix build with llvm-9.0 again.

Commit 6f7306c029a7 ("swr/rast: Refactor memory API between rasterizer
core and swr") unintentionally removed changes for llvm-9.0.

Fixes: 6f7306c029a7 ("swr/rast: Refactor memory API between rasterizer core and swr")
Fixes: 5dd9ad157005 ("swr/rasterizer: Better implementation of scatter")
Signed-off-by: Vinson Lee <vlee@freedesktop.org>
Reviewed-by: Jan Zielinski <jan.zielinski@intel.com>

pan/midgard: Use shared psiz clamp pass

We already had a perfectly cromulent pass for this, but one landed in
common NIR code so let's switch and lighten our tree.

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

pan/midgard: Remove mir_opt_post_move_eliminate

This optimization depended on RA running before scheduling. It therefore
no longer applies and is now unused.

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

pan/midgard: Schedule before RA

This is a tradeoff.

Scheduling before RA means we don't do RA on what-will-become pipeline
registers. Importantly, it means the scheduler is able to reorder
instructions, as registers have not been decided yet.

Unfortunately, it also complicates register spilling, since the spills
themselves won't get bundled optimally and we can only spill twice per
ALU bundle (only one spill per bundle allowed here). It also prevents us
from eliminating dead moves introduced by register allocation, as they
are not dead before RA. The shader-db regressions are from poor spilling
choices introduced by the new bundling requirements. These could be
solved by the combination of a post-scheduler (to combine adjacent
spills into bundles) with a VLIW-aware spill cost calculation.
Nevertheless, the change is small enough that I feel it's worth it to
eat a tiny shader-db regression for the sake of flexibility.

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

pan/midgard: Handle fragment writeout in RA

Rather than using a pile of hacks and awkward constructs in MIR to
ensure the writeout parameter gets written into r0, let's add a
dedicated shadow register class for writeout (interfering with work
register r0) so we can express the writeout condition succintly and
directly.

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

pan/midgard: Do not propagate swizzles into writeout

There's no slot for it; you'll end up writing into the void and
clobbering stuff. Don't. do it.

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

pan/midgard: Fix misc. RA issues

When running the register allocator after scheduling, the MIR looks a
little different, so we need to extend the RA to handle a few of these
extra cases correctly.

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

pan/midgard: Print MIR by the bundle

After scheduling, we still have valid MIR, but we have additional
bundling annotations which we would like to keep debug, so print these.

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

pan/midgard: Print branches in MIR

Rather than a vague "br.??" line, annotate the branch with its target
type (useful for disambiguating discards) and whether it was inverted.

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

pan/midgard: Remove texture_index

This is deadcode.

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

pan/midgard: Cleanup fragment writeout branch

I'm not sure if this is strictly necessary but it makes debugging easier
and minimizes the diff with the experimental scheduler.

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

pan/midgard: Add scheduling barriers

Scheduling occurs on a per-block basis, strongly assuming that a given
block contains at most a single branch. This does not always map to the
source NIR control flow, particularly when discard intrinsics are
involved. The solution is to allow scheduling barriers, which will
terminate a block early in code generation and open a new block.

To facilitate this, we need to move some post-block processing to a new
pass, rather than relying hackily on the current_block pointer.

This allows us to cleanup some logic analyzing branches in other parts
of the driver us well, now that the MIR is much more well-formed.

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

pan/midgard: Track shader quadword count while scheduling

This allow multiblock blend shaders to compute constant colour offsets
correctly.

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

pan/midgard: Allow NULL argument in mir_has_arg

It's sometimes convenient to call this with no instruction specified. By
definition, a missing instruction cannot reference any argument, so
let's check for NULL and shortciruit to false.

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

pan/midgard: Improve mir_mask_of_read_components

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

pan/midgard: Extend mir_special_index to writeout

The branch has the writeout specified in its source list, making this
special even if it's not explicitly part of r0.

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

pan/midgard: csel_swizzle with mir get swizzle

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

pan/midgard: Add mir_insert_instruction*scheduled helpers

In order to run register allocation after scheduling, it is sometimes
necessary to be able to insert instructions into an already-scheduled
program. This is suboptimal, since it forces us to do a worst-case
scheduling, but it is nevertheless required for correct handling of
spills/fills. Let's add helpers to insert instructions as standalone
bundles for use in spilling code.

These helpers are minimal -- they *only* work on load/store ops or
moves. They should not be used for anything but register spilling; any
other instructions should be added prior to the schedule.

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

pan/midgard: Track csel swizzle

While it doesn't matter with an unconditional move to the conditional
register (r31), when we try to elide that move we'll need to track the
swizzle explicitly, and there is no slot for that yet since ALU ops are
normally binary.

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

pan/midgard: Ensure fragment writeout is in the final block

This ensures the block only has exactly one branch, which makes
scheduling happy.

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

pan/midgard: Document Midgard scheduling requirements

Oh boy. Midgard scheduling is crazy... These are all just the
requirements, not even the algorithm yet.

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

pan/midgard: Include condition in branch->src[0]

This will allow us to reference the condition while scheduling.

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

pan/midgard: Add post-schedule iteration helpers

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

pan/midgard: Fix corner case in RA

It doesn't really matter but... meh.

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

pan/midgard: Add OP_IS_CSEL_V helper

..to distinguish from scalar csel.

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

pan/midgard: Expose mir_get/set_swizzle

The scheduler would like to use these.

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

pan/midgard: Extract instruction sizing helper

The scheduler shouldn't need to worry about this.

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

pan/midgard: Factor out mir_is_scalar

This helper doesn't need to be in the giant loop.

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

pan/midgard: Count shader-db stats by bundled instructions

This does not affect shaders in any way. Rather, it makes the shader-db
instruction count recorded in the compiler accurate with the in-order
scheduler, matching up with what we calculate from pandecode.

Though shaders are the same, instruction counts cannot be compared
across this commit for this reason.

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

freedreno/ir3: Link directly to Sethi-Ullman paper

Allow a direct link to the PDF itself from the authors themselves,
rather than a paywall splash page.

Signed-off-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
Acked-by: Rob Clark <robdclark@chromium.org>

Revert "glx: Unset the direct_support bit for GLX_EXT_import_context"

The GLX extension strings are independent of any context, so abusing the
direct_support bit to control this extension's visibility is wrong.

This reverts commit 079d0717fc896bc8086b037d0ed22642274986c7.

Reported-by: Michel Dänzer <michel@daenzer.net>
Reviewed-by: Michel Dänzer <michel@daenzer.net>

panfrost: Add transient BOs to job batches

Memory allocated through panfrost_allocate_transient() is likely to
come from the transient pool. Let's add the BO backing the allocated
memory region to the job batch so the kernel can retain this BO while
jobs are executed.

In practice that has never been a problem because the transient pool
is never shrinked, and even if it was, we still control the lifetime of
the job, so there's no reason for this BO to be freed before the GPU is
done executing the batch. But it still make sense to add the BO for
debugging purpose.

Signed-off-by: Boris Brezillon <boris.brezillon@collabora.com>
Reviewed-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>

panfrost: protect access to shared bo cache and transient pool

Both the BO cache and the transient pool are shared across
context's. Protect access to these with mutexes.

Signed-off-by: Rohan Garg <rohan.garg@collabora.com>
Reviewed-by: Boris Brezillon <boris.brezillon@collabora.com>
Reviewed-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
Signed-off-by: Boris Brezillon <boris.brezillon@collabora.com>

panfrost: Jobs must be per context, not per screen

Jobs _must_ only be shared across the same context, having
the last_job tracked in a screen causes use-after-free issues
and memory corruptions.

Signed-off-by: Rohan Garg <rohan.garg@collabora.com>
Reviewed-by: Boris Brezillon <boris.brezillon@collabora.com>
Reviewed-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
Signed-off-by: Boris Brezillon <boris.brezillon@collabora.com>

st/mesa: Allow zero as [level|layer]_override

This fix two dEQP tests for virgl:

dEQP-EGL.functional.image.create.gles2_cubemap_positive_x_rgba_texture
dEQP-EGL.functional.image.render_multiple_contexts.gles2_cubemap_positive_x_rgba8_texture

Signed-off-by: Lepton Wu <lepton@chromium.org>
Reviewed-by: Marek Olšák <marek.olsak@amd.com>