radv: optimize CmdClear{Color,DepthStencil}Image() for layered textures

If all layers are bound we can perform a fast color or depth clear
instead of iterating over all layers. This has the advantage
to avoid trashing the framebuffer for nothing if you we end up by
doing a fast clear when calling radv_clear_image_layer(), and
clearing all layers in one shot is obviously faster.

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

radv: refactor the fast clear path for better re-use

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

radv: simplify a check in emit_fast_color_clear()

Currently only true if RADV_PERFTEST=dccmsaa is set.

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

radv: add radv_can_fast_clear_{color,depth}() helpers

For further optimisations.

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

radv: add radv_image_view_can_fast_clear() helper

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

radv: add radv_image_can_fast_clear() helper

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

radv: remove useless check in emit_fast_color_clear()

The driver doesn't support DCC/CMASK for mipmapped textures.

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

freedreno: Fix autotools build.

Fix build error.

  CXXLD    pipe_msm.la
../../../../src/gallium/drivers/freedreno/.libs/libfreedreno.a(freedreno_batch.o): In function `batch_init':
src/gallium/drivers/freedreno/freedreno_batch.c:54: undefined reference to `fd_device_version'
src/gallium/drivers/freedreno/freedreno_batch.c:59: undefined reference to `fd_submit_new'
src/gallium/drivers/freedreno/freedreno_batch.c:61: undefined reference to `fd_submit_new_ringbuffer'
src/gallium/drivers/freedreno/freedreno_batch.c:64: undefined reference to `fd_submit_new_ringbuffer'
src/gallium/drivers/freedreno/freedreno_batch.c:66: undefined reference to `fd_submit_new_ringbuffer'
src/gallium/drivers/freedreno/freedreno_batch.c:70: undefined reference to `fd_submit_new_ringbuffer'

Fixes: b4476138d5ad ("freedreno: move drm to common location")
Fixes: aa0fed10d357 ("freedreno: move ir3 to common location")
Signed-off-by: Vinson Lee <vlee@freedesktop.org>
Reviewed-by: Rob Clark <robdclark@gmail.com>

radeonsi: add memory management stress tests for GDS

Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>

winsys/amdgpu: add support for allocating GDS and OA resources

Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>

radeonsi: allow si_cp_dma_clear_buffer to clear GDS from any IB

Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>

winsys/amdgpu: use optimal VM alignment for CPU allocations

Acked-by: Christian König <christian.koenig@amd.com>

winsys/amdgpu: use optimal VM alignment for imported buffers

Window system buffers didn't use the optimal alignment.

Acked-by: Christian König <christian.koenig@amd.com>

winsys/amdgpu,radeon: pass vm_alignment to buffer_from_handle

Acked-by: Christian König <christian.koenig@amd.com>

winsys/amdgpu: overallocate buffers for faster address translation on Gfx9

Sadly, the 3 games I tested (DeusEx:MD, DiRT Rally, DOTA 2) are unaffected
by the overallocation, because I guess their buffers don't fall into
the small range below a power-of-two size.

Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>

winsys/amdgpu: increase the VM alignment to the MSB of the size for Gfx9

Reviewed-by: Christian König <christian.koenig@amd.com>
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>

winsys/amdgpu: use >= instead of > for VM address alignment

Reviewed-by: Christian König <christian.koenig@amd.com>
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>

winsys/amdgpu: clean up code around BO VM alignment

Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>

winsys/amdgpu: optimize slab allocation for 2 MB amdgpu page tables

- the slab buffer size increased from 128 KB to 2 MB (PTE fragment size)
- the max suballocated buffer size increased from 64 KB to 256 KB,
  this increases memory usage because it wastes memory
- the number of suballocators increased from 1 to 3 and they are layered
  on top of each other to minimize unused space in slabs

The final increase in memory usage is:
  DeusEx:MD:  1.8%
  DOTA 2:     1.75%
  DiRT Rally: 0.2%

The kernel driver will also receive fewer buffers.

radeonsi: generalize the slab allocator code to allow layered slab allocators

There is no change in behavior. It just makes it easier to change the number
of slab allocators.

winsys/amdgpu: always reclaim/release slabs if there is not enough memory

radeonsi: fix is_oneway_access_only for bindless images

radeonsi/nir: parse more information about bindless usage

fill more tgsi_shader_info fields.

tgsi/scan: add more information about bindless usage

radeonsi will use this.

radeonsi: small cleanup for memory opcodes

radeonsi: fix is_oneway_access_only for image stores

We need to look at the Dst for image stores.

radeonsi: use structured buffer intrinsics for image views

to stop using the workaround in si_make_buffer_descriptor.

radeonsi: clean up primitive binning enablement

no change in behavior.

Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>

virgl: fix undefined shift to use unsigned.

Ported from virglrenderer.

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

r600: make suballocator 256-bytes align

Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=108311
Cc: <mesa-stable@lists.freedesktop.org>

intel/compiler: Use nir's info when checking uses_streams.

Vulkan and Gallium don't use Mesa's gl_program data structure, so they
can't poke at 'prog'.  But we can simply use the copy of the shader info
stored with the NIR shader, which is guaranteed to exist.

Reviewed-by: Jason Ekstrand <jason@jlekstrand.net>
Reviewed-by: Iago Toral Quiroga <itoral@igalia.com>

nir/derefs: Add a nir_derefs_do_not_alias enum value

This makes some of the code more clear.

Reviewed-by: Thomas Helland <thomashelland90@gmail.com>

egl: add missing #include <stddef.h> in egldevice.h

Otherwise, I get this error:

main/egldevice.h:54:13: error: ‘NULL’ undeclared (first use in this function)
       dev = NULL;
             ^~~~
with this config:

./autogen.sh --enable-gles1 --enable-gles2 --with-platforms='surfaceless' --disable-glx
             --with-dri-drivers="i965" --with-gallium-drivers="" --enable-gbm

v3: Use stddef.h (Matt)
v4: Modify commit message (Eric)

Reviewed-by: Matt Turner <mattst88@gmail.com>
Reviewed-by: Eric Engestrom <eric.engestrom@intel.com>

gallivm: Use nextafterf(0.5, 0.0) as rounding constant

The common truncf(x + 0.5) fails for the floating-point value just less
than 0.5 (nextafterf(0.5, 0.0)). nextafterf(0.5, 0.0) + 0.5, after
rounding is 1.0, thus truncf does not produce the desired value.

The solution is to add nextafterf(0.5, 0.0) instead of 0.5 before
truncating. This works for all values.

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

docs: update calendar, add news item and link release notes for 18.2.6

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

docs: add sha256 checksums for 18.2.6

Signed-off-by: Juan A. Suarez Romero <jasuarez@igalia.com>
(cherry picked from commit cfd1f8b92cae9dde3e5bed42109b5142f50a2ee5)

docs: add release notes for 18.2.6

Signed-off-by: Juan A. Suarez Romero <jasuarez@igalia.com>
(cherry picked from commit 3e741344d79e3ae67b1ad645e7d56fe6c0fb2ae2)

egl/wayland: rather obvious build fix

Fixes: ce74a7bb8de7 ("egl/wayland: plug memory leak in drm_handle_device()")
Fixes: c59d3aa4b9bc ("egl/wayland: bail out when drmGetMagic fails")

winsys/amdgpu: explicitly declare whether buffer_map is permanent or not

Introduce a new driver-private transfer flag RADEON_TRANSFER_TEMPORARY
that specifies whether the caller will use buffer_unmap or not. The
default behavior is set to permanent maps, because that's what drivers
do for Gallium buffer maps.

This should eliminate the need for hacks in libdrm. Assertions are added
to catch when the buffer_unmap calls don't match the (temporary)
buffer_map calls.

I did my best to update r600 for consistency (r300 needs no changes
because it never calls buffer_unmap), even though the radeon winsys
ignores the new flag.

As an added bonus, this should actually improve the performance of
the normal fast path, because we no longer call into libdrm at all
after the first map, and there's one less atomic in the winsys itself
(there are now no atomics left in the UNSYNCHRONIZED fast path).

Cc: Leo Liu <leo.liu@amd.com>
v2:
- remove comment about visible VRAM (Marek)
- don't rely on amdgpu_bo_cpu_map doing an atomic write
Reviewed-by: Marek Olšák <marek.olsak@amd.com>

winsys/amdgpu: add amdgpu_winsys_bo::lock

We'll use it in the upcoming mapping change. Sparse buffers have always
had one.

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

vulkan/wsi: fix s/,/;/ typo

Fixes: 59e58c348e6af16a5f2dd "vulkan/wsi: Only wait on semaphores on the first swapchain"
Signed-off-by: Eric Engestrom <eric.engestrom@intel.com>
Reviewed-by: Jason Ekstrand <jason@jlekstrand.net>

egl/wayland: plug memory leak in drm_handle_device()

As we fail to open the node, we leak the node/device name.

v2: Log and then free() (Eric)

Cc: mesa-stable@lists.freedesktop.org
Signed-off-by: Emil Velikov <emil.velikov@collabora.com>
Reviewed-by: Eric Engestrom <eric.engestrom@intel.com>

egl/wayland: bail out when drmGetMagic fails

Currently as the function fails, we pass uninitialized data to the
authentication function. Stop doing that and print an warning when
the function fails.

v2: Plug memory leak in error path (Eric)

Cc: mesa-stable@lists.freedesktop.org
Signed-off-by: Emil Velikov <emil.velikov@collabora.com>
Reviewed-by: Tapani Pälli <tapani.palli@intel.com> (v1)
Reviewed-by: Eric Engestrom <eric.engestrom@intel.com>

wsi/display: fix mem leak when freeing swapchains

Fixes: da997ebec92942193955 "vulkan: Add KHR_display extension using DRM [v10]"
Signed-off-by: Eric Engestrom <eric.engestrom@intel.com>
Reviewed-by: Keith Packard <keithp@keithp.com>

i965: Set the FBO error state INCOMPLETE_ATTACHMENT only for SRGB_R8

Originally the driver reported GL_FRAMEBUFFER_UNSUPPORTED in all cases,
adding more specific error messages was not correct and broke many tests.
Mostly revert this and only report GL_FRAMEBUFFER_INCOMPLETE_ATTACHMENT
for MESA_FORMAT_R_SRGB8.

Fixes: ebcde3454552adc6d3fea8af2207aafaba857796
  i965: be more specific about FBO completeness errors

Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=108805

Signed-off-by: Gert Wollny <gert.wollny@collabora.com>
Reviewed-by: Tapani Pälli <tapani.palli@intel.com>

i965: Explicitely handle swizzles for MESA_FORMAT_R_SRGB8

The format is emulated by using ISL_FORMAT_L8_SRGB, therefore we need to
force swizzles for the GBA channels. However, doing this only based on the
data type GL_RED breaks other formats, therefore, test specifically for the
format.

Fixes: c5363869d4971780401b21bb75083ef2518c12be
  i965: Force zero swizzles for unused components in GL_RED and GL_RG
Signed-off-by: Gert Wollny <gert.wollny@collabora.com>
Reviewed-by: Tapani Pälli <tapani.palli@intel.com>

virgl: Don't try handling server fences when they are not supported

vtest doesn't implement the according API and would segfault:

Program received signal SIGSEGV, Segmentation fault.
  #0  0x0000000000000000 in ?? ()
  #1  in virgl_fence_server_sync  at
       src/gallium/drivers/virgl/virgl_context.c:1049
  #2  in st_server_wait_sync  at
       src/mesa/state_tracker/st_cb_syncobj.c:155

so just don't do the call when the function pointers are not set.

Fixes dEQP:
  dEQP-GLES3.functional.fence_sync.wait_sync_smalldraw
  dEQP-GLES3.functional.fence_sync.wait_sync_largedraw

Fixes: d1a1c21e7621b5177febf191fcd3d3b8ef69dc96
  virgl: native fence fd support

Signed-off-by: Gert Wollny <gert.wollny@collabora.com>
Reviewed-by: Dave Airlie <airlied@redhat.com>
Reviewed-by: Robert Foss <robert.foss@collabora.com>

virgl,vtest: Initialize return value

Avoids:
Conditional jump or move depends on uninitialised value(s)
  at 0x9E2B39F: virgl_vtest_winsys_resource_cache_create (virgl_vtest_winsys.c:379)
  by 0x9E2725F: virgl_buffer_create (virgl_buffer.c:169)
  by 0x9E246D5: virgl_resource_create (virgl_resource.c:60)
  by 0xA0C1B9F: bufferobj_data (st_cb_bufferobjects.c:344)
  by 0xA0C1B9F: st_bufferobj_data (st_cb_bufferobjects.c:390)
  by 0x9F4ACE3: vbo_use_buffer_objects (vbo_exec_api.c:1136)
  by 0xA0C68C3: st_create_context_priv (st_context.c:416)
  by 0xA0C707A: st_create_context (st_context.c:598)
  by 0x9F81C6B: st_api_create_context (st_manager.c:918)
  by 0x9BBE591: dri_create_context (dri_context.c:161)
  by 0x9BB6931: driCreateContextAttribs (dri_util.c:473)
  by 0x4E97A44: drisw_create_context_attribs (drisw_glx.c:630)
  by 0x4E7C591: glXCreateContextAttribsARB (create_context.c:78)
Uninitialised value was created by a stack allocation
  at 0x9E2B249: virgl_vtest_winsys_resource_cache_create (virgl_vtest_winsys.c:342)

Signed-off-by: Gert Wollny <gert.wollny@collabora.com>
Reviewed-by: Dave Airlie <airlied@redhat.com>
Reviewed-by: Robert Foss <robert.foss@collabora.com>

intel/compiler: fix register allocation in opt_peephole_sel

This wasn't handling 64-bit cases properly. Found by inspection.

Reviewed-by: Ian Romanick <ian.d.romanick@intel.com>

glsl: Remove unused member variable

Reviewed-by: Tapani Pälli <tapani.palli@intel.com>

nir: Call fflush() at the end of nir_print_shader()

We normally call with stderr which is unbuffered, so this won't affect
that, but it does let me call nir_print_shader(nir, fopen("log", "w+"))
from gdb and actually get the whole shader in my file.

Reviewed-by: Tapani Pälli <tapani.palli@intel.com>

v3d: Add renderonly support.

I've been using this with the kmsro series to test v3d on VKMS without my
old KMS hack in the v3d kernel driver.  KMSRO still needs some cleanup,
but v3d RO support seems reasonable.

gallium: Remove unused variable in u_tests.

Fixes: 0d17b685b1ff ("gallium/u_tests: add a compute shader test that clears an image")
Reviewed-by: Marek Olšák <marek.olsak@amd.com>

radv: Align large buffers to the fragment size.

Improves performance in Talos by about 15% (and significant improvements
in RotR and possibly other but did not bench with final patch) on
kernel 4.19 and earlier.

On 4.20+ a similar effect comes from

433ca054949a "drm/amdgpu: try allocating VRAM as power of two"

v2: Do not impact the alignment of the physical memory.

Reviewed-by: Dave Airlie <airlied@redhat.com>
Reviewed-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
CC: <mesa-stable@lists.freedesktop.org>

freedreno: implements get_sample_position

Since 1285f71d3e landed, it needs to provide apps with proper sample
position for MSAA.

Currently no way to query this to hw, these are taken from blob driver.

Fixes: dEQP-GLES31.functional.texture.multisample.samples_#.sample_position
Signed-off-by: Rob Clark <robdclark@gmail.com>

freedreno/a3xx: also set FSSUPERTHREADENABLE

We set equiv bit in SP_FS_CTRL_REG0.  Somehow the hw doesn't hang with
this mismatched config, but does run slower.  It is faster with either
neither bit set, or both bits set, but both is the fastest of the three
configurations.  Worth a bit over 10% gain in glmark2.

Spotted-by: Jonathan Marek <jonathan@marek.ca>
Signed-off-by: Rob Clark <robdclark@gmail.com>

freedreno: use MSM_BO_SCANOUT with scanout buffers

Signed-off-by: Jonathan Marek <jonathan@marek.ca>

freedreno: use GENERIC instead of TEXCOORD for blit program

blip_fp uses GENERIC as input, so blit_vp should match for linking

Signed-off-by: Jonathan Marek <jonathan@marek.ca>
Signed-off-by: Rob Clark <robdclark@gmail.com>

freedreno: a2xx texture update

Adds all missing texture related logic. For everything to work it also
needs changes to ir2/fd2_program, which are part of the ir2 update patch.

Note: it needs rnndb update

Signed-off-by: Jonathan Marek <jonathan@marek.ca>
[remove stray patch]
Signed-off-by: Rob Clark <robdclark@gmail.com>

freedreno/a2xx: Compute depth base in gmem correctly

Note: it needs rnndb update

Signed-off-by: Marek Vasut <marex@denx.de>
Signed-off-by: Jonathan Marek <jonathan@marek.ca>
Signed-off-by: Rob Clark <robdclark@gmail.com>

freedreno/a2xx: set VIZ_QUERY_ID on a20x

Signed-off-by: Jonathan Marek <jonathan@marek.ca>
Signed-off-by: Rob Clark <robdclark@gmail.com>

freedreno: add missing a20x ids

200: 256KiB GMEM A200 (imx53)
201: 128KiB GMEM A200 (imx51)

Signed-off-by: Jonathan Marek <jonathan@marek.ca>
Signed-off-by: Rob Clark <robdclark@gmail.com>

freedreno/a2xx: fix POINT_MINMAX_MAX overflow

As it stands, it overflows to zero.

Signed-off-by: Jonathan Marek <jonathan@marek.ca>
Signed-off-by: Rob Clark <robdclark@gmail.com>

freedreno: a2xx: fd2_draw update

Signed-off-by: Jonathan Marek <jonathan@marek.ca>
Signed-off-by: Rob Clark <robdclark@gmail.com>

nir: add fceil lowering

lowers ceil(x) as -floor(-x)

Signed-off-by: Jonathan Marek <jonathan@marek.ca>
Reviewed-by: Jason Ekstrand <jason@jlekstrand.net>
Reviewed-by: Christian Gmeiner <christian.gmeiner@gmail.com>
Signed-off-by: Rob Clark <robdclark@gmail.com>

freedreno: update generated headers

Signed-off-by: Rob Clark <robdclark@gmail.com>

freedreno/a6xx: set guardband clip

On older gens, the CLIP_ADJ bitfields were actually 3.6 fixed point.
Which might make more sense.  Although this formula comes up with values
pretty close to what blob does for various viewport sizes (for at least
a5xx and a6xx), and seems to work.

Signed-off-by: Rob Clark <robdclark@gmail.com>

freedreno/a6xx: disable LRZ for z32

f6131d4ec7a had the side effect of enabling LRZ w/ 32b depth buffers.
But there are some bugs with this, which aren't fully understood yet,
so for now just skip LRZ w/ z32..

Fixes: f6131d4ec7a freedreno/a6xx: Clear z32 and separate stencil with blitter
Signed-off-by: Rob Clark <robdclark@gmail.com>

freedreno/a6xx: Clear gmem buffers at flush time

We generate an IB to clear the gmem at flush time and jump to it
before rendering each tile. This lets us get rid of the command stream
patching for gmem offsets.

Signed-off-by: Kristian H. Kristensen <hoegsberg@chromium.org>
Signed-off-by: Rob Clark <robdclark@gmail.com>

freedreno/a6xx: Move resolve blits to an IB

Signed-off-by: Kristian H. Kristensen <hoegsberg@chromium.org>
Signed-off-by: Rob Clark <robdclark@gmail.com>

freedreno/a6xx: Move restore blits to IB

Signed-off-by: Kristian H. Kristensen <hoegsberg@chromium.org>
Signed-off-by: Rob Clark <robdclark@gmail.com>

mesa/st: better colormask check for clear fallback

For RGB surfaces (for example) we don't really care that the colormask
is 0x7 instead of 0xf.  This should not trigger clear_with_quad()
slowpath.

Signed-off-by: Rob Clark <robdclark@gmail.com>
Reviewed-by: Marek Olšák <marek.olsak@amd.com>

mesa/st: swap order of clear() and clear_with_quad()

If we can't clear all the buffers with pctx->clear() (say, for example,
because of ColorMask), push the buffers we *can* clear with pctx->clear()
first.  Tilers want to see clears coming before draws to enable fast-
paths, and clearing one of the attachments with a quad-draw first
confuses that logic.

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

freedreno: move ir3 to common location

Move (most of) the ir3 compiler to src/freedreno/ir3 so that it can be
re-used by some future vulkan driver.  The parts that are gallium
specific have been refactored out and remain in the gallium driver.

Getting the move done now so that it can happen before further
refactoring to support a6xx specific instructions.

NOTE also removes ir3_cmdline compiler tool from autotools build since
that was easier than fixing it and I normally use meson build.  Waiting
patiently for the day that we can remove *everything* from the autotools
build.

Signed-off-by: Rob Clark <robdclark@gmail.com>

freedreno/ir3: remove u_inlines usage

Signed-off-by: Rob Clark <robdclark@gmail.com>

freedreno/ir3: split up ir3_shader

Split the parts that are gallium specific into ir3_gallium so the rest
can move to a common location outside of gallium.

Signed-off-by: Rob Clark <robdclark@gmail.com>

freedreno/ir3: remove pipe_stream_output_info dependency

A bit annoying to have to copy into our own struct.  But this is
something the compiler really needs to know, at least on earlier
generations where streamout is implemented in shader.

Signed-off-by: Rob Clark <robdclark@gmail.com>

freedreno/ir3: some header file cleanup

Clean up some of the low-hanging-fruit usages of freedreno_util.h

Signed-off-by: Rob Clark <robdclark@gmail.com>

freedreno/ir3: use env_var_as_unsigned()

Signed-off-by: Rob Clark <robdclark@gmail.com>

util: env_var_as_unsigned() helper

So I can drop env2u() helper from freedreno_util.h and get rid of one
small ir3 dependency on gallium/freedreno

Signed-off-by: Rob Clark <robdclark@gmail.com>

freedreno/ir3: move disasm and optmsgs debug flags

Move them to IR3_SHADER_DEBUG so we can remove ir3's dependency on
fd_mesa_debug.

Signed-off-by: Rob Clark <robdclark@gmail.com>

freedreno: FD_SHADER_DEBUG -> IR3_SHADER_DEBUG

Only used by ir3, so move it into ir3 to be more self contained.

Signed-off-by: Rob Clark <robdclark@gmail.com>

freedreno: remove shader_stage_name()

Signed-off-by: Rob Clark <robdclark@gmail.com>

freedreno: shader_t -> gl_shader_stage

Just massive search/replace for the most part.

Step towards removing ir3 dependency on disasm.h which is shared by
a2xx.  One step closer to being able to move ir3 out of gallium.

Signed-off-by: Rob Clark <robdclark@gmail.com>

freedreno/ir3: standalone compiler updates

Signed-off-by: Rob Clark <robdclark@gmail.com>

freedreno: move drm to common location

So that we can re-use at least parts of it for vulkan driver, and so
that we can move ir3 to a common location (which uses fd_bo to allocate
storage for shaders)

Signed-off-by: Rob Clark <robdclark@gmail.com>

freedreno/drm: remove dependency on gallium driver

Prep work to move drm to a common location.

Slightly hacky, but the softpin debug flag is only temporary.

Signed-off-by: Rob Clark <robdclark@gmail.com>

util: promote u_memory to src/util

as well as os_memory*
Reviewed-by: Rob Clark <robdclark@gmail.com>

gallium: Fix uninitialized variable warning in compute test.

The compiler doesn't know that ny != 0, so x might be uninitialized for
the printf at the end.

Reviewed-by: Elie Tournier <elie.tournier@collabora.com>

radv: Clamp gfx9 image view extents to the allocated image extents.

Mirrors AMDVLK. Looks like if we go over the alignment of height
we actually start to change the addressing. Seems like the extra
miplevels actually work with this.

Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=108245
Fixes: f6cc15dccd5 "radv/gfx9: fix block compression texture views. (v2)"
Reviewed-by: Dave Airlie <airlied@redhat.com>
Reviewed-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>

intel/compiler: fix indentation style in opt_algebraic()

anv/icl: Set use full ways in L3CNTLREG

L3 allocation table in h/w specification recommends using 4 KB
granularity for programming allocation fields in L3CNTLREG.

Signed-off-by: Anuj Phogat <anuj.phogat@gmail.com>
Reviewed-by: Francisco Jerez <currojerez@riseup.net>

intel/icl: Set way_size_per_bank to 4

Signed-off-by: Anuj Phogat <anuj.phogat@gmail.com>
Reviewed-by: Francisco Jerez <currojerez@riseup.net>

i965/icl: Set use full ways in L3CNTLREG

L3 allocation table in h/w specification recommends using 4 KB
granularity for programming allocation fields in L3CNTLREG.

Signed-off-by: Anuj Phogat <anuj.phogat@gmail.com>
Reviewed-by: Francisco Jerez <currojerez@riseup.net>

i965/icl: Fix L3 configurations

Use L3 configuration specified in h/w specification.

V2: Drop configs which do under allocation of l3 cache.
    Bump up the comment above table.

Signed-off-by: Anuj Phogat <anuj.phogat@gmail.com>
Reviewed-by: Francisco Jerez <currojerez@riseup.net>

build: stop defining unused VERSION

Scons and autotools don't define it, and as of last commit nothing
uses it.

`VERSION` is also a generic enough name that something somewhere will
eventually clash, and we don't want to repeat the LLVM `DEBUG` fiasco.

Signed-off-by: Eric Engestrom <eric.engestrom@intel.com>
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Reviewed-by: Emil Velikov <emil.velikov@collabora.com>
Reviewed-by: Dylan Baker <dylan@pnwbakers.com>

vulkan/utils: s/VERSION/PACKAGE_VERSION/

Everything else uses PACKAGE_VERSION, so let's be consistent, and
VERSION and PACKAGE_VERSION are currently defined to be the same in
meson and android, while VERSION is undefined in autotools and scons.

Signed-off-by: Eric Engestrom <eric.engestrom@intel.com>
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Reviewed-by: Emil Velikov <emil.velikov@collabora.com>
Reviewed-by: Dylan Baker <dylan@pnwbakers.com>

anv: correctly use vulkan 1.0 by default

Per chapter 3.2 "Instances":
> Providing a NULL VkInstanceCreateInfo::pApplicationInfo or providing
> an apiVersion of 0 is equivalent to providing an apiVersion of
> VK_MAKE_VERSION(1,0,0).

Reported-by: Niklas Haas <git@haasn.xyz>
Fixes: 8c048af5890d43578ca4 "anv: Copy the appliation info into the instance"
Signed-off-by: Eric Engestrom <eric.engestrom@intel.com>
Reviewed-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>

mesa/main: fixup requirements for GL_PRIMITIVES_GENERATED

This enum is also allowed by EXT_tessellation_shader, which is supported
on older i965 HW (as opposed to OES_geometry_shader). This was missed
when narrowing this code-path, leading to dEQP regressions.

Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=108868
Fixes: f09d94fbd11 "mesa/main: fix validation of transform-feedback queries"
Signed-off-by: Erik Faye-Lund <erik.faye-lund@collabora.com>
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Tested-by: Mark Janes <mark.a.janes@intel.com>

mesa/main: fix incorrect depth-error

If glGetTexImage or glGetnTexImage is called with a level that doesn't
exist, we get an error message on this form:

Mesa: User error: GL_INVALID_VALUE in glGetTexImage(depth = 0)

This is clearly nonsensical, because these APIs don't even have a
depth-parameter. The reason is that get_texture_image_dims() return
all-zero dimensions for non-existent texture-images, and we go on to
validate these dimensions as if they were user-input, because
glGetTextureSubImage requires checking.

So let's split this logic in two, so glGetTextureSubImage can have
stricter input-validation. All arguments that are no longer validated
are generated internally by mesa, so there's no use in validating them.

Fixes: 42891dbaa12 "gettextsubimage: verify zoffset and depth are correct"
Signed-off-by: Erik Faye-Lund <erik.faye-lund@collabora.com>
Reviewed-by: Juan A. Suarez <jasuarez@igalia.com>