mesa: simplify handling the return value of update_program

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

mesa: simplify a loop in _mesa_update_texture_state

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

mesa: replace VP/FP/ATIfs _Enabled flags with helper functions

These are only used in the GL compatibility profile.

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

mesa: don't update draw buffer bounds in _mesa_update_state

st/mesa doesn't need the draw bounds for draw calls. I've added the call
where it's necessary in core Mesa and drivers, but I suspect that most
drivers can just move the call to the right places.

The core Mesa places aren't hot paths, so the call overhead doesn't matter
there.

For now, only st/mesa is made such that this function is invoked very
rarely.

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

mesa: remove update_framebuffer_size

For the default framebuffer, _mesa_resize_framebuffer updates it.
For FBOs, _mesa_test_framebuffer_completeness updates it.

This code is redundant.

Reviewed-by: Nicolai Hähnle <nicolai.haehnle@amd.com>
Reviewed-by: Brian Paul <brianp@vmware.com>

mesa: replace ctx->Polygon._FrontBit with a helper function

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

mesa: replace ctx->VertexProgram._TwoSideEnabled with a helper function

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

mesa: stop using _NEW_STENCIL with st/mesa, use DriverFlags.NewStencil instead

This bypasses _mesa_update_state_locked.

Before:
   DrawElements ( 1 VBOs, 4 UBOs,  8 Tex) w/ stencil enable change:    3.99 million
   DrawArrays   ( 1 VBOs, 4 UBOs,  8 Tex) w/ stencil enable change:    4.56 million

After:
   DrawElements ( 1 VBOs, 4 UBOs,  8 Tex) w/ stencil enable change:    4.93 million
   DrawArrays   ( 1 VBOs, 4 UBOs,  8 Tex) w/ stencil enable change:    5.84 million

It's quite a difference in the draw call rate when ctx->NewState stays
equal to 0 the whole time.

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

mesa: replace _mesa_update_stencil() with helper functions

The idea is to remove the dependency on _mesa_update_state_locked,
so that st/mesa can skip it for stencil state updates, and then stop
setting _NEW_STENCIL in mesa/main if the driver is st/mesa.

The main motivation is to stop invoking _mesa_update_state_locked for
certain state groups.

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

meta: do the full FBO completeness check in decompress_texture_image

_mesa_update_state will no longer recompute Width/Height if the framebuffer
is complete. We now rely on the FBO completeness check to do it.

The only code that needs to be fixed seems to be this one.

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

i965/gen6: Use isl-based miptree also for stencil rbs

Fixes dEQP-EGL.functional.image.render_multiple_contexts.
gles2_renderbuffer_stencil_stencil_buffer

Signed-off-by: Topi Pohjolainen <topi.pohjolainen@intel.com>
Reviewed-by: Jason Ekstrand <jason@jlekstrand.net>

i965: Remove spurious mutex frobbing around call to intel_miptree_blit

These locks were added in 2f28a0dc, but I don't see anything in the
intel_miptree_blit path that should make this necessary.

When asked, Kristian says:

    I doubt it's needed now with the new blorp. If I remember correctly,
    I had to drop the lock there since intel_miptree_blit() could hit
    the XY blit path that requires a fast clear resolve. The fast
    resolve being meta, would then try to lock the texture again.

Signed-off-by: Ian Romanick <ian.d.romanick@intel.com>
Reviewed-by: Kristian Høgsberg <krh@bitplanet.net>

egl: turn one more boolean `int` into a `bool`

Same as the previous commit, but this one was split out because it's
a bit more complicated: this field is given as a pointer to a function,
so the function had to be changed as well, and the function was use in
a bunch of places, which needed updating as well.

Signed-off-by: Eric Engestrom <eric@engestrom.ch>
Reviewed-by: Emil Velikov <emil.velikov@collabora.com>

egl: turn boolean `int`s into `bool`s

Signed-off-by: Eric Engestrom <eric@engestrom.ch>
Reviewed-by: Emil Velikov <emil.velikov@collabora.com>

i965/miptree: Move isl_surf_get_(hiz|mcs)_surf out of the assert

Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=101535
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=101538
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=101539
Reviewed-by: Topi Pohjolainen <topi.pohjolainen@intel.com>

intel/genxml: Use the same naming convention for Floating Point Mode.

In newer gens, this field has a prefix and the non-IEEEE-745 mode is called
"Alternate", instead of simply "Alt".

Signed-off-by: Rafael Antognolli <rafael.antognolli@intel.com>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>

intel/genxml: Normalize URB Data field in WM_STATE.

On gen6+, this is called "Dispatch GRF Start Register For Constant/Setup Data
0", while on gen5 and lower it's called only "Dispatch GRF Start Register For
URB Data", but it's essentially the same thing (URB data), so rename it to
match newer gens and simplify the C code that handles it.

Signed-off-by: Rafael Antognolli <rafael.antognolli@intel.com>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>

intel/genxml: Rename field on WM_STATE to match gen6+.

"Pixel Shader Kill Pixel" -> "Pixel Shader Kills Pixel", which is how it's
called on newer gens.

Signed-off-by: Rafael Antognolli <rafael.antognolli@intel.com>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>

intel/genxml: Normalize fields on WM_STATE.

On gen4, WM_STATE only has one Kernel Start Pointer and one GRF Register
Count, but we can make the code that handles this on multiple gens simpler if
we add an index 0 to it too.

Signed-off-by: Rafael Antognolli <rafael.antognolli@intel.com>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>

intel/genxml: Add missing field to CLIP_STATE.

Just because it's not set doesn't mean that it doesn't exist. And since the
field is there on newer gens, having it on gen5 simplifies the code when
porting gen5 and lower.

Also add missing value to API Mode on CLIP_STATE on gen4.

Signed-off-by: Rafael Antognolli <rafael.antognolli@intel.com>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>

intel/genxml: Fix type of UserClipFlags ClipTest Enable Bitmask.

This is a bitmask, so it can't be a boolean. Also rename it so it matches
gen6+.

Signed-off-by: Rafael Antognolli <rafael.antognolli@intel.com>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>

intel/genxml: Add missing fields to CLIP_STATE on gen4-5.

These fields are set by brw_clip_unit, so we need them when converting to
genxml.

Signed-off-by: Rafael Antognolli <rafael.antognolli@intel.com>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>

intel/genxml: Normalize GS_STATE.

Rename "Rendering Enable" to "Rendering Enabled", so it matches gen6+.

Signed-off-by: Rafael Antognolli <rafael.antognolli@intel.com>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>

i915: Always emit W on gen3

Unlike the older gen2 hardware, gen3 performs perspective
correct interpolation even for the primary/secondary colors.
To do that it naturally needs us to emit W for the vertices.

Currently we emit W only when at least one texture coordinate
set gets emitted. This means the interpolation of color will
change depending on whether texcoords/varyings are used or not.
That's probably not what anyone would expect, so let's just
always emit W to get consistent behaviour. Trying to avoid
emitting W seems like more hassle than it's worth, especially
as bspec seems to suggest that the hardware will perform the
perspective division anyway.

This used to be broken until it was accidentally fixed it in
commit c349031c27b7 ("i915: Fix texcoord vs. varying collision
in fragment programs") by introducing a bug that made the driver
always emit W. After fixing that bug in commit c1eedb43f32f
("i915: Fix wpos_tex vs. -1 comparison") we went back to the
old behaviour and caused an apparent regression.

Fixes: c1eedb43f32f ("i915: Fix wpos_tex vs. -1 comparison")
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=101451
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Reviewed-by: Ian Romanick <ian.d.romanick@intel.com>

mesa: add KHR_no_error support for glStencilOp()

Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Timothy Arceri <tarceri@itsqueeze.com>

mesa: add stencil_op() helper

Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Timothy Arceri <tarceri@itsqueeze.com>

mesa: add KHR_no_error support for glStencilFunc()

Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Timothy Arceri <tarceri@itsqueeze.com>

mesa: add stencil_func() helper

Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Timothy Arceri <tarceri@itsqueeze.com>

mesa: add KHR_no_error support for glStencilOpSeparate()

Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Timothy Arceri <tarceri@itsqueeze.com>

mesa: add stencil_op_separate() helper

Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Timothy Arceri <tarceri@itsqueeze.com>

mesa: add KHR_no_error support for glStencilMaskSeparate()

Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Timothy Arceri <tarceri@itsqueeze.com>

mesa: add stencil_mask_separate() helper

Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Timothy Arceri <tarceri@itsqueeze.com>

mesa: add KHR_no_error support for glStencilFuncSeparate()

Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Timothy Arceri <tarceri@itsqueeze.com>

mesa: add stencil_func_separate() helper

Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Timothy Arceri <tarceri@itsqueeze.com>

etnaviv: fix blend color for RB swapped rendertargets

Same as with the colormasks, the blend color needs to be swizzled according
to the rendertarget format.

Signed-off-by: Lucas Stach <dev@lynxeye.de>
Reviewed-by: Wladimir J. van der Laan <laanwj@gmail.com>
Reviewed-by: Christian Gmeiner <christian.gmeiner@gmail.com>

spirv: Work around the Doom shader bug

Doom shipped with a broken version of GLSLang which handles samplers as
function arguments in a way that isn't spec-compliant.  In particular,
it creates a temporary local sampler variable and copies the sampler
into it.  While Dave has had a hack patch out for a while that gets it
working, we've never landed it because we've been hoping that a game
update would come out with fixed shaders.  Unfortunately, no game update
appears on to be on the horizon and I've found this issue in yet another
application so I think we're stuck working around it.  Hopefully, we can
delete this code one day.

Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=99467
Cc: "17.1" <mesa-stable@lists.freedesktop.org>
Tested-by: Grazvydas Ignotas <notasas@gmail.com>
Reviewed-by: Dave Airlie <airlied@redhat.com>

glsl: Update build instructions for int64.glsl

Trivial

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

glsl: Fix indent in dump code

Signed-off-by: Elie Tournier <elie.tournier@collabora.com>
Reviewed-by: Ian Romanick <ian.d.romanick@intel.com>

st/xvmc: deal with drivers wanting different texture formats

Previously, texture formats were being used unconditionally without
checking. However nv30 supports neither RGBX8 nor R4A4/A4R4 formats. Add
sufficient fallbacks so that the nv30 driver can have working OSD.

Tested on a NV44A/PCI.

Signed-off-by: Ilia Mirkin <imirkin@alum.mit.edu>

nvc0: fix transfer of larger rectangles with DmaCopy on gk104 and up

By treating the rectangles as 1cpp, we can run up against some internal
copy engine limits and trigger a MEM2MEM_RECT_OUT_OF_BOUNDS error check
at launch time.

This commit enables the REMAP hardware, which allows us to specify both
the component size and number of components for a transfer.  We're then
able to pass in the real width/nblocksx values and not hit the limits.

There's a couple of "supported" CPPs in the list that we can't actually
hit, but are there simply because they're possible.

Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
Acked-by: Ilia Mirkin <imirkin@alum.mit.edu>

nvc0: copy engine surface params are only relevant for tiled surfaces

Aside from reducing pushbuf usage in some situations, this commit should
have no other effect, and is just to make it somewhat obvious that those
methods have zero effect on linear surfaces.

Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
Acked-by: Ilia Mirkin <imirkin@alum.mit.edu>

st/mesa: fix assert to be simpler

I just noticed a warning with a non-debug build, but really
this could all be one line, and I'm not even 100% the assert
makes sense here.

Reviewed-by: Timothy Arceri <tarceri@itsqueeze.com>
Reviewed-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Marek Olšák <marek.olsak@amd.com>

intel: compiler/i965: fix is_broxton checks

In 5f2fe9302c is_geminilake was introduced for the differenciate
broxton from geminilake. Unfortunately I failed as verifying that
is_broxton is throughout the code base to mean Gen9lp.

Fixes: 5f2fe9302c ("intel: common: add flag to identify platforms by name")
Signed-off-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>

mesa/main: Move NULL pointer check.

In blit_framebuffer we're already doing a NULL
pointer check for readFb and drawFb so it makes
sense to do it before we actually use the pointers.

CID: 1412569
Signed-off-by: Plamena Manolova <plamena.manolova@intel.com>
Reviewed-by: Rafael Antognolli <rafael.antognolli@intel.com>

swr: Include definition of missing function

Inline function SWR_MULTISAMPLE_POS::PrecalcSampleData() was missing
definition.  Include definition in core/state_funcs.h.

Fixes windows build.

Reviewed-by: Tim Rowley <timothy.o.rowley@intel.com>

i965/cnl: Add l3 configuration for Cannonlake

V2 (Anuj):
Squash the changes in one patch rebase on master.
Address the review comments made by Francisco Jerez.
Do the URB allocation per slice (not per bank).

V3 (Anuj):
Update the comment.
Format the table as other l3 config tables.

Signed-off-by: Ben Widawsky <ben@bwidawsk.net>
Signed-off-by: Anuj Phogat <anuj.phogat@gmail.com>
Reviewed-by: Francisco Jerez <currojerez@riseup.net>
---
V1 was sent out with the heading:
"i965/cnl: Properly handle l3 configuration"

i965: Add a variable for way size per bank in get_l3_way_size()

Adding this variable better explains the computation of L3 way
size in the function.

V2: Use const variable for way_size_per_bank.

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

i965: Fix broxton 2x6 l3 config

The new table added in this patch matches with the table
in gfxspecs. We were programming the wrong values earlier.

V2: Update the comment.

Cc: "17.1" <mesa-stable@lists.freedesktop.org>
Signed-off-by: Anuj Phogat <anuj.phogat@gmail.com>
Reviewed-by: Francisco Jerez <currojerez@riseup.net>

i965: Fall back to normal blorp clear instead of meta clear

When intel_miptree_alloc_non_msrt_mcs fails, fall back to normal blorp
color clear instead of falling back to meta.  With this change,
brw_blorp_clear_color can never fail.

v2: Combine two if-statements to remove a level of indentation.
Suggested by Jason.

Signed-off-by: Ian Romanick <ian.d.romanick@intel.com>
Reviewed-by: Jason Ekstrand <jason@jlekstrand.net>

intel/blorp: Apply source offset in the TEX case

Previously the offset was only applied in the TXF case.

Signed-off-by: Ian Romanick <ian.d.romanick@intel.com>
Suggested-by: Jason Ekstrand <jason.ekstrand@intel.com>
Reviewed-by: Jason Ekstrand <jason@jlekstrand.net>

intel/blorp: Apply Gen4 coord. normalization after cubemap sizes are adjusted

Otherwise the values used for coordinate normalization use the wrong
sizes.

Signed-off-by: Ian Romanick <ian.d.romanick@intel.com>
Suggested-by: Jason Ekstrand <jason.ekstrand@intel.com>
Reviewed-by: Jason Ekstrand <jason@jlekstrand.net>

intel/blorp: Set needs_(dst|src)_offset for Gen4 cubemaps

We call convert_to_single_slice so they may end up with a non-trivial
offset that needs to be taken into account.

v2 (idr): Also set needs_src_offset.  Suggested by Jason.

Fixes ES2-CTS.functional.texture.specification.basic_copyteximage2d.cube_rgba
and ES2-CTS.functional.texture.specification.basic_copytexsubimage2d.cube_rgba
on G45.

Signed-off-by: Ian Romanick <ian.d.romanick@intel.com>
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=101284
Reviewed-by: Jason Ekstrand <jason@jlekstrand.net>

meta/blit: Silence unused parameter warning

drivers/common/meta_blit.c: In function ‘setup_glsl_msaa_blit_scaled_shader’:
drivers/common/meta_blit.c:62:58: warning: unused parameter ‘filter’ [-Wunused-parameter]
                                    GLenum target, GLenum filter)
                                                          ^~~~~~

Signed-off-by: Ian Romanick <ian.d.romanick@intel.com>
Reviewed-by: Jason Ekstrand <jason@jlekstrand.net>

meta: Silence unused parameter warning

drivers/common/meta.c:2694:71: warning: unused parameter ‘dims’ [-Wunused-parameter]
 copytexsubimage_using_blit_framebuffer(struct gl_context *ctx, GLuint dims,
                                                                       ^~~~

Signed-off-by: Ian Romanick <ian.d.romanick@intel.com>
Reviewed-by: Jason Ekstrand <jason@jlekstrand.net>

i965: Fix incorrect comment

There is no intel_miptree_slice_has_hiz function, but there is a
intel_miptree_level_has_hiz function.  I assume that's the correct one
to use.

Signed-off-by: Ian Romanick <ian.d.romanick@intel.com>
Reviewed-by: Jason Ekstrand <jason@jlekstrand.net>

mesa: simplify _mesa_IsVertexArray()

_mesa_lookup_vao() already returns NULL if id is zero.

v2: - change the conditional (Ian)

Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Alejandro Piñeiro <apinheiro@igalia.com> (v1)

mesa/format_info: use designated initialiser list

Also, make that table const, since no-one is supposed to modify it anyway.

Signed-off-by: Eric Engestrom <eric.engestrom@imgtec.com>
Reviewed-by: Emil Velikov <emil.velikov@collabora.com>
Reviewed-by: Ian Romanick <ian.d.romanick@intel.com>

vc4: Clean up release build warnings using MAYBE_UNUSED.

These variables are all used in an assert(), so release builds see no
usages.

vc4: Allow VBOs to be mapped during execution.

There's no reason we can't -- the mappings we expose are basically
equivalent to persistent/coherent, already.

Improves mesa-demos drawoverhead (no state change) performance by
5.21362% +/- 1.25078% (n=11).

gallium/vbuf: avoid segfault when we get invalid glDrawRangeElements()

A common user error is to call glDrawRangeElements() with the 'end'
argument being one too large.  If we use the vbuf module to translate
some vertex attributes this error can cause us to read past the end of
the mapped hardware buffer, resulting in a crash.

This patch adjusts the vertex count to avoid that issue.  Typically,
the vertex_count gets decremented by one.

This fixes crashes with the Unigine Tropics and Sanctuary demos with older
VMware hardware versions.  The issue isn't hit with VGPU10 because we
don't hit this fallback.

No piglit changes.

CC: mesa-stable@lists.freedesktop.org
Reviewed-by: Marek Olšák <marek.olsak@amd.com>

gallium/vbuf: add some const qualifiers

Helps understandability a bit.

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

translate: whitespace fixes in translate_generic.c

softpipe: remove unused softpipe_context::line_stipple_counter

Trivial.

radeonsi: set correct usage flag according to image access type

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

winsys/amdgpu: fix a deadlock when waiting for submission_in_progress

First this happens:

1) amdgpu_cs_flush (lock bo_fence_lock)
   -> amdgpu_add_fence_dependency
   -> os_wait_until_zero (wait for submission_in_progress) - WAITING

2) amdgpu_bo_create
   -> pb_cache_reclaim_buffer (lock pb_cache::mutex)
   -> pb_cache_is_buffer_compat
   -> amdgpu_bo_wait (lock bo_fence_lock) - WAITING

So both bo_fence_lock and pb_cache::mutex are held. amdgpu_bo_create can't
continue. amdgpu_cs_flush is waiting for the CS ioctl to finish the job,
but the CS ioctl is trying to release a buffer:

3) amdgpu_cs_submit_ib (CS thread - job entrypoint)
   -> amdgpu_cs_context_cleanup
   -> pb_reference
   -> pb_destroy
   -> amdgpu_bo_destroy_or_cache
   -> pb_cache_add_buffer (lock pb_cache::mutex) - DEADLOCK

The simple solution is not to wait for submission_in_progress, which we
need in order to create the list of dependencies for the CS ioctl. Instead
of building the list of dependencies as a direct input to the CS ioctl,
build the list of dependencies as a list of fences, and make the final list
of dependencies in the CS thread itself.

Therefore, amdgpu_cs_flush doesn't have to wait and can continue.
Then, amdgpu_bo_create can continue and return. And then amdgpu_cs_submit_ib
can continue.

Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=101294

Cc: 17.1 <mesa-stable@lists.freedesktop.org>
Reviewed-by: Nicolai Hähnle <nicolai.haehnle@amd.com>

radeonsi: update all resident texture descriptors when needed

To avoid useless DCC fetches when DCC is disabled, descriptors
have to be updated in order to reflect this change. This is
quite similar to how we update descriptors of bound textures.

As a side effect, this should also prevent VM faults when
bindless textures are invalidated, because the VA in the
descriptor has to be updated accordingly as well.

I don't see any performance improvements with DOW3.

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

radeonsi: keep track of the sampler state for texture handles

Needed for updating all resident texture descriptors when
dirty_tex_counter changes.

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

i965: perf: use gen_device_info rather then brw_context

Signed-off-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>

intel: common: add number of thread per eu

This will be used by to normalize OA counters.

Signed-off-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>

intel: common: express timestamps units in frequency

Rather than storing the period as a double that looses some precision.

Also fixes the Gen9LP timestamp frequency which is no 19200123 but
19200000 as pointed by Ville :

https://lists.freedesktop.org/archives/intel-gfx/2017-April/125126.html

Finally add the Cannonlake timestamp frequency.

Signed-off-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>

i965: convert MI_REPORT_PERF_COUNT to genxml

Also make it available from gen7 only to gen7+.

Signed-off-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>

i965: perf: fix codegen with single operand equation

We did support single value operand equations, but not single variable
operand ones. In particular we were failing on "$Sampler0Bottleneck".

Signed-off-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>

intel: common: add flag to identify platforms by name

The perf infrastructure needs to identify specific platforms, not just
generations.

Signed-off-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>

i965/wm: Use stored hiz surface instead of creating copy

Now the last user of intel_miptree_get_aux_isl_surf() is gone.

Reviewed-by: Jason Ekstrand <jason@jlekstrand.net>
Signed-off-by: Topi Pohjolainen <topi.pohjolainen@intel.com>

i965/blorp: Use hiz surface instead of creating copy

Reviewed-by: Jason Ekstrand <jason@jlekstrand.net>
Signed-off-by: Topi Pohjolainen <topi.pohjolainen@intel.com>

i965/miptree/gen7+: Use isl for hiz layouts

v2: Use better assert by checking isl_surf_get_hiz_surf()

Reviewed-by: Jason Ekstrand <jason@jlekstrand.net>
Signed-off-by: Topi Pohjolainen <topi.pohjolainen@intel.com>

i965/miptree: Drop BO_ALLOC_FOR_RENDER in intel_miptree_alloc_mcs()

because buffers get unconditionally initialised by cpu writing.

Reviewed-by: Jason Ekstrand <jason@jlekstrand.net>
Signed-off-by: Topi Pohjolainen <topi.pohjolainen@intel.com>

i965/miptree: Use isl for mcs layouts

and pass the ccs isl surface to blorp instead of creating a
copy.

v2 (Jason): Explain ccs change and use better assert checking
            isl_surf_get_mcs_surf()

Reviewed-by: Jason Ekstrand <jason@jlekstrand.net>
Signed-off-by: Topi Pohjolainen <topi.pohjolainen@intel.com>

i965/miptree: Refactor aux surface allocation

v2 (Jason): Drop unused argument in intel_alloc_aux_buffer() and
            move assignment of "buf->surf" in intel_alloc_aux_buffer()
            into this patch.

Reviewed-by: Jason Ekstrand <jason@jlekstrand.net>
Signed-off-by: Topi Pohjolainen <topi.pohjolainen@intel.com>

i965/gen6: Use isl for hiz

Reviewed-by: Jason Ekstrand <jason@jlekstrand.net>
Signed-off-by: Topi Pohjolainen <topi.pohjolainen@intel.com>

i965/miptree: Refactor isl aux usage resolver

Reviewed-by: Jason Ekstrand <jason@jlekstrand.net>
Signed-off-by: Topi Pohjolainen <topi.pohjolainen@intel.com>

i965/gen6: Use isl for stencil surfaces

Reviewed-by: Jason Ekstrand <jason@jlekstrand.net>
Signed-off-by: Topi Pohjolainen <topi.pohjolainen@intel.com>

i965/miptree: Prepare range getter for isl based

Reviewed-by: Jason Ekstrand <jason@jlekstrand.net>
Signed-off-by: Topi Pohjolainen <topi.pohjolainen@intel.com>

i965/miptree: Prepare stencil mapping for isl based

Reviewed-by: Jason Ekstrand <jason@jlekstrand.net>
Signed-off-by: Topi Pohjolainen <topi.pohjolainen@intel.com>

i965/blorp: Prepare for isl based miptrees

Reviewed-by: Jason Ekstrand <jason@jlekstrand.net>
Signed-off-by: Topi Pohjolainen <topi.pohjolainen@intel.com>

i965: Add isl based miptree creator

v2: Use new brw_bo_alloc_tiled() interface

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

i965/miptree: Add option to resolve offsets using isl_surf

v2 (Nanley): Add comment telling why "level -= mt->first_level"

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

i965: Prepare slice copy for isl based miptrees

v2 (Jason): Fix a helper variable only used for assert -
            open code instead.

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

i965/tex: Prepare image update for isl based miptrees

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

i965: Prepare framebuffer validator for isl based miptrees

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

i965: Prepare slice validator for isl based miptrees

v2 (Nanley): Minify depth in case of 3D surface. Also moved to
             .c file to get minify() without additional
             header inclusions

Reviewed-by: Jason Ekstrand <jason@jlekstrand.net>
Signed-off-by: Topi Pohjolainen <topi.pohjolainen@intel.com>

i965: Prepare image validation for isl based miptrees

Reviewed-by: Jason Ekstrand <jason@jlekstrand.net>
Signed-off-by: Topi Pohjolainen <topi.pohjolainen@intel.com>

i965: Prepare up/downsampling for isl based miptrees

Reviewed-by: Nanley Chery <nanley.g.chery@intel.com>
Signed-off-by: Topi Pohjolainen <topi.pohjolainen@intel.com>

i965/miptree: Add isl surface

Reviewed-by: Nanley Chery <nanley.g.chery@intel.com>
Signed-off-by: Topi Pohjolainen <topi.pohjolainen@intel.com>

i965: Add helper for converting isl tiling to bufmgr tiling

Reviewed-by: Nanley Chery <nanley.g.chery@intel.com>
Signed-off-by: Topi Pohjolainen <topi.pohjolainen@intel.com>

i965/miptree: Refactor mapping table alloc

v2 (Nanley): Use minify() instead of direct shift

Reviewed-by: Jason Ekstrand <jason@jlekstrand.net>
Signed-off-by: Topi Pohjolainen <topi.pohjolainen@intel.com>

i965/gen6: Declare minify(depth, level) layers for 3D stencil

Keeps following patch refactoring the table allocation
non-functional.

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

i965/gen4: Add support for single layer in alignment workaround

On gen < 6 one doesn't have level or layer specifiers available
for render and depth targets. In order to support rendering to
specific level/layer, driver needs to manually offset the surface
to the desired slice.
There are, however, alignment restrictions to respect as well and
in come cases the only option is to use temporary single slice
surface which driver copies after rendering to the full miptree.

Current alignment workaround introduces new texture images which
are added to the parent texture object. Texture validation later
on copies the additional levels back to the surface that contains
the full mipmap.
This only works for non-arrayed surfaces and driver currently
creates new arrayed images in vain - individual layers within the
newly created are still unaligned the same as before.

This patch drops this mechanism and instead attaches single
temporary slice into the render buffer. This gets immediately
copied back to the mipmapped and/or arrayed surface just after
the render is done.

Sitting on top of earlier series cleaning up the depth buffer
state, this patch additionally fixes the following piglit tests:

    arb_framebuffer_object.fbo-generatemipmap-cubemap.g965m64
    arb_texture_cube_map.copyteximage cube.g965m64
    arb_texture_cube_map.copyteximage cube.ilkm64
    arb_pixel_buffer_object.texsubimage array pbo.g965m64
    ext_framebuffer_object.fbo-cubemap.g965m64
    ext_texture_array.copyteximage 1d_array.g45m64
    ext_texture_array.copyteximage 1d_array.g965m64
    ext_texture_array.copyteximage 1d_array.ilkm64
    ext_texture_array.copyteximage 2d_array.g45m64
    ext_texture_array.copyteximage 2d_array.g965m64
    ext_texture_array.copyteximage 2d_array.ilkm64
    ext_texture_array.fbo-array.g965m64
    ext_texture_array.fbo-generatemipmap-array.g965m64
    ext_texture_array.gen-mipmap.g965m64

Reviewed-by: Jason Ekstrand <jason@jlekstrand.net>
Signed-off-by: Topi Pohjolainen <topi.pohjolainen@intel.com>

i965/miptree: Separate src and dst slice specifiers in slice copy

Reviewed-by: Jason Ekstrand <jason@jlekstrand.net>
Signed-off-by: Topi Pohjolainen <topi.pohjolainen@intel.com>

i965/miptree: Clarify face/level/layer in slice copy

Reviewed-by: Jason Ekstrand <jason@jlekstrand.net>
Signed-off-by: Topi Pohjolainen <topi.pohjolainen@intel.com>