docs: Expand ARB_gpu_shader5 to describe status of individual features

This extension is a huge grab-bag of "stuff that's in DX11". Break it
apart to make it clear what still needs to be done.

Signed-off-by: Chris Forbes <chrisf@ijw.co.nz>
Reviewed-by: Matt Turner <mattst88@gmail.com>

docs: Mark off ARB_texture_view and add to release notes for 10.2.

V4: Don't claim Gen8 yet.

Signed-off-by: Chris Forbes <chrisf@ijw.co.nz>
Reviewed-by: Ian Romanick <ian.d.romanick@intel.com>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Acked-by: Eric Anholt <eric@anholt.net>

i965: Enable ARB_texture_view on Gen7

V4: Don't enable this for Gen8 yet -- that still needs wired up.

Signed-off-by: Chris Forbes <chrisf@ijw.co.nz>
Reviewed-by: Ian Romanick <ian.d.romanick@intel.com>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Acked-by: Eric Anholt <eric@anholt.net>

i965: Account for view parameters in blit CTSI path

Signed-off-by: Chris Forbes <chrisf@ijw.co.nz>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Acked-by: Eric Anholt <eric@anholt.net>

i965: Account for MinLayer/MinLevel in blorp CTSI path

Signed-off-by: Chris Forbes <chrisf@ijw.co.nz>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Acked-by: Eric Anholt <eric@anholt.net>

i965: Account for view parameters in fast depth clears

V2: - No need for layer_multiplier; multisampled depth surfaces are IMS.
    - Remove unused num_layers.

Signed-off-by: Chris Forbes <chrisf@ijw.co.nz>
Reviewed-by: Ian Romanick <ian.d.romanick@intel.com>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Acked-by: Eric Anholt <eric@anholt.net>

i965/blorp: Account for nonzero MinLayer in layered clears.

Signed-off-by: Chris Forbes <chrisf@ijw.co.nz>
Reviewed-by: Ian Romanick <ian.d.romanick@intel.com>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Acked-by: Eric Anholt <eric@anholt.net>

i965/blorp: Use irb->layer_count in clear

Signed-off-by: Chris Forbes <chrisf@ijw.co.nz>
Reviewed-by: Ian Romanick <ian.d.romanick@intel.com>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Acked-by: Eric Anholt <eric@anholt.net>

i965: Add layer_count to intel_renderbuffer

This is the effective layer count, for clears etc. This differs from the
depth of the miptree level when views are involved.

Signed-off-by: Chris Forbes <chrisf@ijw.co.nz>
Reviewed-by: Ian Romanick <ian.d.romanick@intel.com>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Acked-by: Eric Anholt <eric@anholt.net>

i965: Pull out layer_multiplier in intel_update_renderbuffer_wrapper

We're about to need this in another place.

Signed-off-by: Chris Forbes <chrisf@ijw.co.nz>
Reviewed-by: Ian Romanick <ian.d.romanick@intel.com>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Acked-by: Eric Anholt <eric@anholt.net>

i965: Add `layered` parameter to intel_update_renderbuffer_wrapper

We're about to need this so we can determine the layer count of the
wrapper.

Signed-off-by: Chris Forbes <chrisf@ijw.co.nz>
Reviewed-by: Ian Romanick <ian.d.romanick@intel.com>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Acked-by: Eric Anholt <eric@anholt.net>

i965: Adjust renderbuffer wrapper to account for MinLevel/MinLayer

Signed-off-by: Chris Forbes <chrisf@ijw.co.nz>
Reviewed-by: Ian Romanick <ian.d.romanick@intel.com>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Acked-by: Eric Anholt <eric@anholt.net>

i965: Enable texture upload fast path with MinLevel

We'll still avoid MinLayer here since the fast path doesn't understand
arrays at all, but it's straightforward to do levels.

Signed-off-by: Chris Forbes <chrisf@ijw.co.nz>
Reviewed-by: Ian Romanick <ian.d.romanick@intel.com>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Acked-by: Eric Anholt <eric@anholt.net>

i965: Account for MinLevel in texture upload fast path

Signed-off-by: Chris Forbes <chrisf@ijw.co.nz>
Reviewed-by: Ian Romanick <ian.d.romanick@intel.com>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Acked-by: Eric Anholt <eric@anholt.net>

i965: Adjust map/unmap code for MinLevel/MinLayer

This allows core mesa's TexSubImage paths etc to work correctly
with views which have nonzero MinLevel or MinLayer.

Signed-off-by: Chris Forbes <chrisf@ijw.co.nz>
Reviewed-by: Ian Romanick <ian.d.romanick@intel.com>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Acked-by: Eric Anholt <eric@anholt.net>

i965: Don't try to use fast upload path for nontrivial views

This will eventually be relaxed, but we'll get the fallback path
working first.

Signed-off-by: Chris Forbes <chrisf@ijw.co.nz>
Reviewed-by: Ian Romanick <ian.d.romanick@intel.com>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Acked-by: Eric Anholt <eric@anholt.net>

i965: Adjust surface_state emission to account for view parameters

V4: Comment style, remove magic shift.

Signed-off-by: Chris Forbes <chrisf@ijw.co.nz>
Reviewed-by: Ian Romanick <ian.d.romanick@intel.com>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Acked-by: Eric Anholt <eric@anholt.net>

i965: Add _Format to intel_texobj.

This is the actual mesa_format to use. In non-view cases this is always
the same as the mt's format.

V4: Comment style

Signed-off-by: Chris Forbes <chrisf@ijw.co.nz>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Acked-by: Eric Anholt <eric@anholt.net>

i965: Add driver hook for TextureView

We need to wire the original texture's mt into the view. All the hard
work of setting up an appropriate tree of gl_texture_image structures
has already been done by core mesa.

Signed-off-by: Chris Forbes <chrisf@ijw.co.nz>
Reviewed-by: Ian Romanick <ian.d.romanick@intel.com>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Acked-by: Eric Anholt <eric@anholt.net>

i965: Ensure that texture validation is skipped for immutable textures.

If we were to relayout the miptree, we'd break any views that are
sharing it.

(Simplified based on suggestions from Eric)

Signed-off-by: Chris Forbes <chrisf@ijw.co.nz>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Acked-by: Eric Anholt <eric@anholt.net>

i965: refactor format selection for unsupported ETC* formats

We will need to call this to munge view formats.

Signed-off-by: Chris Forbes <chrisf@ijw.co.nz>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Acked-by: Eric Anholt <eric@anholt.net>

i965: refactor format munging for separate stencil

We will need this for munging the view's format.

Signed-off-by: Chris Forbes <chrisf@ijw.co.nz>
Reviewed-by: Ian Romanick <ian.d.romanick@intel.com>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Acked-by: Eric Anholt <eric@anholt.net>

i965: Include #slices in miptree debug

Signed-off-by: Chris Forbes <chrisf@ijw.co.nz>
Reviewed-by: Ian Romanick <ian.d.romanick@intel.com>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Acked-by: Eric Anholt <eric@anholt.net>

mesa: Adjust _MaxLevel computation to account for views

Signed-off-by: Chris Forbes <chrisf@ijw.co.nz>
Reviewed-by: Ian Romanick <ian.d.romanick@intel.com>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Acked-by: Eric Anholt <eric@anholt.net>

mesa: Prefer non-swizzled formats for most sized internalformats

These formats can be cast to others (with different component types or
sizes) via ARB_texture_view or ARB_shader_image_load_store. We want
them to be laid out consistently so that we can just reinterpret the
memory with a different format.

In V1, this was done conditionally on a 'prefer_no_swizzle' flag which
was set in TexStorage/TextureView paths, but we need the same behavior
for ARB_shader_image_load_store (which also works with images created
via TexImage, so we don't want it to be conditional.

Signed-off-by: Chris Forbes <chrisf@ijw.co.nz>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Acked-by: Eric Anholt <eric@anholt.net>

i965: Render R8G8B8X8 as R8G8B8A8

The sampler can handle R8G8B8X8 (and substitute 1.0 for the fourth
component) but we can't use it as a render target.

Signed-off-by: Chris Forbes <chrisf@ijw.co.nz>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Acked-by: Eric Anholt <eric@anholt.net>

i965: Pretend we don't support BRW_SURFACEFORMAT_R16G16B16_FLOAT for textures.

None of the other 3-component 16bpc formats are directly supported, so
they get promoted to XRGB equivalents. *Not* promoting RGB16F the same
way makes texture views much more fiddly -- we don't want to have to do
crazy copying behind the scenes.

(with my other master + my experimental ARB_texture_view support) fixes
the piglit test: `spec/ARB_texture_view/view compare 48bit formats`

No regressions in gpu.tests on Haswell.

V4: Don't alter the formats table -- just don't match it to a mesa_format. [Kenneth]

Signed-off-by: Chris Forbes <chrisf@ijw.co.nz>
Reviewed-by: Ian Romanick <ian.d.romanick@intel.com>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Acked-by: Eric Anholt <eric@anholt.net>

i965: Enable R10G10B10A2_UNORM format

This is supported by all generations, and is required for memory layout
consistency for texture_view.

Signed-off-by: Chris Forbes <chrisf@ijw.co.nz>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Acked-by: Eric Anholt <eric@anholt.net>

i965: Enable R8G8B8A8_UNORM_SRGB format

Now this is the preferred format for GL_SRGB8_ALPHA8.

Signed-off-by: Chris Forbes <chrisf@ijw.co.nz>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Acked-by: Eric Anholt <eric@anholt.net>

swrast: Add support for fetching from MESA_FORMAT_R10G10B10A2_UNORM

V4: Fix rebase conflicts with Brian's renaming of the texfetch
functions.

Signed-off-by: Chris Forbes <chrisf@ijw.co.nz>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Reviewed-by: Brian Paul <brianp@vmware.com>
Acked-by: Eric Anholt <eric@anholt.net>

mesa: fix packing of float texels to GL_SHORT/GL_BYTE

Previously, we would unpack the texels to floats using *_TO_FLOAT_TEX,
and then pack them into the desired format using FLOAT_TO_*. Unfortunately,
this isn't quite the inverse operation, and so some texel values would
end up off-by-one.

This fixes the GL_RGB8_SNORM and GL_RGB16_SNORM subcases in piglit's
arb_texture_view-format-consistency-get test on i965. The similar 1-, 2-
and 4-component cases already worked because they took the memcpy path
rather than repacking.

Signed-off-by: Chris Forbes <chrisf@ijw.co.nz>
Reviewed-by: Brian Paul <brianp@vmware.com>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Acked-by: Eric Anholt <eric@anholt.net>

r600g: Don't leak bytecode on shader compile failure

Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=74868

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

build: force .so extension for the gallium dri modules

While linux uses .so as a default extension for shared libraries that is
not the case for other platforms. The loader in libGL (and others) assumes
that the dri module will always have a .so extension, thus it will fail
to load on the affected platforms.

Spotted-by: Jon TURNEY <jon.turney@dronecode.org.uk>
Signed-off-by: Emil Velikov <emil.l.velikov@gmail.com>

Partially revert bba9c28 "configure: use LIB_EXT rather than hardcoded .so"

Filenames passed to dlopen() don't need to use the platform's default extension
for shared libraries.

Using the '.so' extension when dlopen()ing DRI drivers is hardcoded into mesa
and the X server, so it should be hardcoded here in the Makefile as well.

A similar fix is probably also needed for gallium DRI drivers.

(Consider that if we were starting from scratch, perhaps we would use a custom
extension like .dri instead)

Cc: Emil Velikov <emil.l.velikov@gmail.com>
Signed-off-by: Jon TURNEY <jon.turney@dronecode.org.uk>
Reviewed-by: Emil Velikov <emil.l.velikov@gmail.com>

Partially revert "st/xa: Fix advertized version number and try to avoid future discrepancies"

This reverts commit 61bedc3d6b08943f015f9d590c07a6af36c2a92c.

As the header is the one defining the API/ABI and is distributed
during installation, we should be using it rather than re-defining
the XA version in configure.ac.

Bump the version in the header to 2.2.0, to reflect what was the
original intent of commit 42158926c6d7d3ddbe61b9a04d60544ff1b50a96.

Signed-off-by: Emil Velikov <emil.l.velikov@gmail.com>
Reviewed-by: Thomas Hellstrom <thellstrom@vmware.com>

glx: drop obsolete _XUnlock_Mutex in __glXInitialize error path

With commit 1f1928db001(glx: Drop _Xglobal_lock while we create and
initialize glx display) we've split the big _Xglobal_lock handling in
a more fine grained manner.

Unfortunatelly we forgot to drop the unlock_mutex on the error paths,
leading to undefined behaviour as the mutex is already unlocked.

Cc: Kristian Høgsberg <krh@bitplanet.net>
Cc: "9.2 10.0 10.1" <mesa-stable@lists.freedesktop.org>
Signed-off-by: Emil Velikov <emil.l.velikov@gmail.com>
Reviewed-by: Ian Romanick <ian.d.romanick@intel.com>

freedreno/a3xx: assert() -> debug_assert()

We hit this assert with some piglit tests.  Which appears to be a bug
outside of freedreno.  Previously we were relying on assert() being
redefined to debug_assert() so that we didn't crash in release builds.
Somehow that stopped working.  So just use debug_assert() directly.

Signed-off-by: Rob Clark <robclark@freedesktop.org>

svga: move LIST_INITHEAD(dirty_buffers) earlier in svga_context_create()

Fixes a crash in svga_context_flush_buffers() if we use the 'draw' module
for AA lines (when the device doesn't support that feature).  We need to
initialize this list before we setup the swtnl pieces.

Found/fixed by Charmaine Lee.

Cc: "10.0" <mesa-stable@lists.freedesktop.org>
Reviewed-by: Thomas Hellstrom <thellstrom@vmware.com>
Reviewed-by: Jakob Bornecrantz <jakob@vmware.com>

i965: Stop advertising GL_MESA_ycbcr_texture.

The "new" fragment shader backend has never supported the necessary
color conversion code for this to work.  We began using the new backend
in Mesa 7.10 for GLSL (commit a81d423d93f22a948f3aa4bf73, October 2010),
and for ARB_fragment_program in Mesa 9.1 (commit 97615b2d8c7c3cea6fd3a4,
August 2012).

I haven't heard any complaints, so I don't think anyone will miss this
feature.  I believe mplayer used it at one point, but these days
defaults to other paths anyway.

Cc: mesa-stable@lists.freedesktop.org
Signed-off-by: Kenneth Graunke <kenneth@whitecape.org>
Reviewed-by: Ian Romanick <idr@freedesktop.org>
Reviewed-by: Eric Anholt <eric@anholt.net>

freedreno/a3xx/compiler: add CEIL

fixes piglit glsl-fs-ceil

Signed-off-by: Rob Clark <robclark@freedesktop.org>

freedreno/a3xx/compiler: fix neg mov's

create_mov() was fixed up to handle neg/abs properly for interal mov's,
using absneg.f, but forgot to fix it for TGSI MOV's.  The problem with
using add.f to handle negated mov's is that we can only take a single
const reg src.  So:

  MOV TEMP[n], -CONST[m]

would turn into:

  add.f Rdst, (neg)CONST[m], 0.0

which would not work.  Anyways, just remove the extra code and always
use create_mov() which DTRT.

This fixes piglit vs-op-neg-int test.

Signed-off-by: Rob Clark <robclark@freedesktop.org>

radeonsi: allow fast color clear and Hyper-Z with 1D-tiled surfaces on CIK

This depends on my kernel fix. Hyper-Z is still disabled by default.

r600g,radeonsi: add a bunch of useful queries for the HUD

r600g,radeonsi: set correct initial domain for shared resources

gallium/radeon: fix warnings

tnl: Merge _tnl_vbo_draw_prims() into _tnl_draw_prims().

This should help prevent situations where we render without proper index
bounds. For example: https://bugs.freedesktop.org/show_bug.cgi?id=59455

Reviewed-by: Ian Romanick <ian.d.romanick@intel.com>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Reviewed-by: Eric Anholt <eric@anholt.net>

i965: Remove unused sampler key fields

Signed-off-by: Topi Pohjolainen <topi.pohjolainen@intel.com>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>

mesa: move declaration before code in etc2_unpack_rgb8()

To fix MSVC build since cb4ad1368551b.

i965: Delete "fast color clear unsupported" performance warning.

Applications frequently clear to colors other than 0.0 or 1.0, which
prevents us from doing fast color clears.  In that case, we issue this
performance warning on basically every glClear call, resulting in so
much spam that it's nearly impossible to see any other messages.

Plus, I don't think it's useful.  We aren't suggesting a better way to
do what the application developers want---we're just telling them it
would be faster to do something they don't want.

Driver developers have no control over the clear color, so this message
is totally useless to them.

A better alternative to get this sort of information is to use
INTEL_DEBUG=blorp, which tells you whether color clears were fast,
simd16 repdata, or slow.

v2: Rebase on has_color_component changes.

Signed-off-by: Kenneth Graunke <kenneth@whitecape.org>
Reviewed-by: Ian Romanick <ian.d.romanick@intel.com>

freedreno/a3xx: deal with optimized tex instructions

Keep track of whether we actually have any sam instructions in the
resulting shader, rather than using TGSI SAMP declarations.  If the sam
instruction is optimized out, because the result is not used, we don't
want to emit texture state, etc.  In fact emitting sampler state and/or
setting PIXLODENABLE bit when there are no texture fetches seems to
cause lockup.

In theory this should never happen for a "normal" shader, unless the
state tracker is wonky.  But it is a very real possibility for binning
pass shaders.

Signed-off-by: Rob Clark <robclark@freedesktop.org>

mesa: add bounds checking to eliminate buffer overrun

Decompressing ETC2 textures was causing intermitent segfault
by copying resulting 4x4 texel block to the destination texture
regardless of the size of the destination texture. Issue found
via application crash in GLBenchmark 3.0's Manhattan test.

v2: add more detail comment. Compute limit outside inner loops.
v3: add bugzilla reference
v4: Correct cc syntax in commit log
v5: really grab the right patch

Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=74988
Cc: "9.2 10.0 10.1" <mesa-stable@lists.freedesktop.org>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Reviewed-by: Ian Romanick <ian.d.romanick@intel.com> [v1, suggested v2-3]

st/omx/enc: cleanup omx/vid_enc.c

cleanup by moving each step into a separate function

Signed-off-by: Leo Liu <leo.liu@amd.com>
Reviewed-by: Christian König <christian.koenig@amd.com>

st/omx/enc: allocate input buffer private on demand

v2: move allocation to a function as first step
    to clean vid_enc_EncodeFrame

Signed-off-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Leo Liu <leo.liu@amd.com>

svga: replace sampler assertion with conditional

For TEX instructions, the set of samplers and sampler views should
be consistent.  The XA state tracker sometimes passes an inconsistent
set of samplers and sampler views.  Rather than assert and die, issue
a warning.

v2: add debugging code to detect inconsistent state.
v3: also check for null sampler in svga_state_tss.c

Cc: "10.0" "10.1" <mesa-stable@lists.freedesktop.org>
Reviewed-by: Thomas Hellstrom <thellstrom@vmware.com>

i965/vec4: fix record clearing in copy propagation

Given

  mov vgrf7, vgrf9.xyxz
  add vgrf9.xyz, vgrf4.xyzw, vgrf5.xyzw
  add vgrf10.x, vgrf6.xyzw, vgrf7.wwww

the last instruction would be wrongly changed to

  add vgrf10.x, vgrf6.xyzw, vgrf9.zzzz

during copy propagation.

The issue is that when deciding if a record should be cleared, the old code
checked for

  inst->dst.writemask & (1 << ch)

instead of

  inst->dst.writemask & (1 << BRW_GET_SWZ(src->swizzle, ch))

Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=76749
Signed-off-by: Chia-I Wu <olv@lunarg.com>
Cc: Jordan Justen <jljusten@gmail.com>
Cc: Matt Turner <mattst88@gmail.com>
Reviewed-by: Ian Romainck <ian.d.romanick@intel.com>
Reviewed-by: Eric Anholt <eric@anholt.net>
Cc: "10.1" <mesa-stable@freedesktop.org>

i965/vec4: Add a test for copy propagation behavior.

I thought I was seeing a bug in the code while reviewing, but it's not
there.

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

i965/fs: Track whether we're doing dual source in a more obvious way.

I'm going to be turning dual_src_output into an array in a moment.

Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>

i965/fs: Add a couple more global special regs to special[]

Nothing bad came of this because they weren't used after visitor running,
but leaving them in a bad state seems like a recipe for pain later.

Suggested-by: Kenneth Graunke <kenneth@whitecape.org>

i965/fs: Handle arrays of special regs more cleanly.

Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>

i965/fs: Fix dump_instructions() on uniforms.

All of a vec4 uniform was being printed as "u0"

Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>

i965/fs: Fix vgrf0 live interval when no interpolation was done.

When you've got a simple solid-color shader that doesn't generate
pixel_x/y interpolation, we were deciding that the first vgrf was both the
undefined pixel_x and pixel_y, and extending its live interval to avoid
the stride problem.  That tricked other optimization that tries to see if
a particular instruction is the last use of a variable.

Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>

i965: Drop pointless check for variable declarations in splitting.

We're walking the whole instruction stream, so we know the declaration
will be found.

Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>

i965: Remove stale comment.

We stopped doing variable index lowering for uniforms in
a64c1eb9b110f29b8abf803a8256306702629bdc, 5 months after the comment was
added.

Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>

glsl: Move tree grafting's debug output to stderr.

The rest of our compiler dumps are there, now.

Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>

glsl: Skip making a temporary for assignments when we don't need one.

While we wish our optimization passes could identify all the cases where
we can coalesce our variables, we miss out on a lot of opportunities.

total instructions in shared programs: 1673849 -> 1673166 (-0.04%)
instructions in affected programs:     299521 -> 298838 (-0.23%)
GAINED:                                7
LOST:                                  0

Note that many programs are "hurt".  The notable ones are where we produce
unrolling in cases we didn't before (presumably just because of the lower
instruction count).  But there are also some cases where pushing things
right into the variables prevents copy propagation and tree grafting,
since we don't split our variable usage webs apart.

Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>

i915: Fix build error.

is_power_of_two() is now provided by mesa so its definition must be removed
from the i915 driver code.

Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>

glsl: Pass ctx->Const.NativeIntegers to do_algebraic.

The next patch will introduce an optimization that only works when
integers are not represented as floating point values.

Signed-off-by: Kenneth Graunke <kenneth@whitecape.org>
Reviewed-by: Ian Romanick <ian.d.romanick@intel.com>
Reviewed-by: Matt Turner <mattst88@gmail.com>

glsl: Pass ctx->Const.NativeIntegers to do_common_optimization().

The next few patches will introduce an optimization that only works when
integers are not represented as floating point values.

v2: Re-word-wrap a line, as requested by Ian Romanick.

Signed-off-by: Kenneth Graunke <kenneth@whitecape.org>
Reviewed-by: Ian Romanick <ian.d.romanick@intel.com>
Reviewed-by: Matt Turner <mattst88@gmail.com>

glsl: Validate that base types match for a number of binops.

The IR is not supposed to support implicit type conversions; we just
failed to validate it.

Signed-off-by: Kenneth Graunke <kenneth@whitecape.org>
Reviewed-by: Ian Romanick <ian.d.romanick@intel.com>
Reviewed-by: Matt Turner <mattst88@gmail.com>

glsl: Fix lack of i2u in lower_ubo_reference.

ir_binop_ubo_load takes unsigned integer operands.  However, the array
index used to compute these offsets may be a signed integer.  (For
example, see Piglit's spec/glsl-1.40/uniform_buffer/fs-bvec-array).

For some reason, we were missing an ir_binop_i2u cast, and ir_validator
was failing to catch that.

Without this change, ir_builder's type inference code broke for me when
writing a new optimization pass.

Cc: mesa-stable@lists.freedesktop.org
Signed-off-by: Kenneth Graunke <kenneth@whitecape.org>
Reviewed-by: Ian Romanick <ian.d.romanick@intel.com>
Reviewed-by: Matt Turner <mattst88@gmail.com>

i965/fs: Skip emitting MACH/MOV for small integers.

The vector backend already implemented this optimization, but
surprisingly, we never bothered to implement it in the scalar backend.

In addition to saving two instructions, this eliminates a use of the
accumulator as an explicit source, which is unsupported in SIMD16 mode
on Gen7+, which could help us gain SIMD16 programs.

Cuts 19.23% of the instructions in dolphin/efb2ram.shader_test.

v2: Rebase on is_16bit_integer_constant -> is_uint16_constant rename.

Signed-off-by: Kenneth Graunke <kenneth@whitecape.org>
Reviewed-by: Ian Romanick <ian.d.romanick@intel.com>
Reviewed-by: Matt Turner <mattst88@gmail.com>

glsl: Make is_16bit_constant from i965 an ir_constant method.

The i965 MUL instruction doesn't natively support 32-bit by 32-bit
integer multiplication; additional instructions (MACH/MOV) are required.
However, we can avoid those if we know one of the operands can be
represented in 16 bits or less.  The vector backend's is_16bit_constant
static helper function checks for this.

We want to be able to use it in the scalar backend as well, which means
moving the function to a more generally-usable location.  Since it isn't
i965 specific, I decided to make it an ir_constant method, in case it
ends up being useful to other people as well.

v2: Rename from is_16bit_integer_constant to is_uint16_constant, as
    suggested by Ilia Mirkin.  Update comments to clarify that it does
    apply to both int and uint types, as long as the value is
    non-negative and fits in 16-bits.

Signed-off-by: Kenneth Graunke <kenneth@whitecape.org>
Reviewed-by: Ian Romanick <ian.d.romanick@intel.com>
Reviewed-by: Matt Turner <mattst88@gmail.com>

mesa: Move is_power_of_two() function from brw_context.h to macros.h.

This makes the function available from core Mesa code, including the
GLSL compiler.

Signed-off-by: Kenneth Graunke <kenneth@whitecape.org>
Reviewed-by: Ian Romanick <ian.d.romanick@intel.com>
Reviewed-by: Matt Turner <mattst88@gmail.com>

i965: Fix "SIMD16 unsupported" messages via KHR_debug.

Performance warnings are logged via KHR_debug in addition to when the
INTEL_DEBUG=perf environment variable is set.  Without this, messages in
debug contexts would have "(null)" for the reason.

Signed-off-by: Kenneth Graunke <kenneth@whitecape.org>
Reviewed-by: Ian Romanick <ian.d.romanick@intel.com>
Reviewed-by: Matt Turner <mattst88@gmail.com>

i965: Fix missing dirty bits in the gen8_sbe_state atom.

These are clearly needed---the comments in the function are even present
for each one of them.   I originally had two separate state atoms for
3DSTATE_SBE and 3DSTATE_SBE_SWIZ.  When I combined the functions, I must
have forgotten to add the atoms for 3DSTATE_SBE_SWIZ.

Signed-off-by: Kenneth Graunke <kenneth@whitecape.org>
Reviewed-by: Eric Anholt <eric@anholt.net>

i965: Drop BRW_NEW_RASTERIZER_DISCARD flag from Broadwell SOL atom.

Nothing actually uses this---we handle rasterizer discard in the
clipper in order for statistics counters to work.

Signed-off-by: Kenneth Graunke <kenneth@whitecape.org>
Reviewed-by: Eric Anholt <eric@anholt.net>

i965: Use the correct program when uploading Broadwell SOL state.

This is the equivalent of commit 43e77215b13b2f86e461cd8a62b542f.

Signed-off-by: Kenneth Graunke <kenneth@whitecape.org>
Reviewed-by: Eric Anholt <eric@anholt.net>

st/xa: Make sure unused samplers are set to NULL

renderer_copy_prepare was setting the first sampler but never telling
the cso code how many samplers were actually used. Fix this.

Cc: "10.1" <mesa-stable@freedesktop.org>
Signed-off-by: Thomas Hellstrom <thellstrom@vmware.com>
Reviewed-by: Brian Paul <brianp@vmware.com>

st/xa: Bind destination before setting new state

Binding a new destination may cause the svga driver to emit draw calls
while propagating the surface. Make sure this doesn't happen in the middle
of sampler state setup where state may be incosistent.

In practice, surface propagation should never happen here and even if it did,
it wouldn't be a valid reason for the svga driver to emit partially set up
state, but to avoid future uncertainties, make sure this doesn't happen
anyway.

Found while auditing the state tracker for inconsistent sampler state /
sampler view setup.

Cc: "10.1" <mesa-stable@freedesktop.org>
Signed-off-by: Thomas Hellstrom <thellstrom@vmware.com>
Reviewed-by: Jakob Bornecrantz <jakob@vmware.com>

glapi: Fix libglapi build.

This line appears to have been accidentally dropped from the last commit,
and the resulting libglapi was missing symbols.

glapi/build: Add headers to distribution.

Acked-by: Emil Velikov <emil.l.velikov@gmail.com>

glapi/gen: Ship more Python files

Acked-by: Emil Velikov <emil.l.velikov@gmail.com>

glapi/gen: Ship XML and Python files

Acked-by: Emil Velikov <emil.l.velikov@gmail.com>

glapi/gen: Add missing XML files to API_XML

Also (re)move XML files from COMMON to API_XML.

Acked-by: Emil Velikov <emil.l.velikov@gmail.com>

src/build: Add getopt to distribution.

Acked-by: Emil Velikov <emil.l.velikov@gmail.com>

gbm/build: Add headers to distribution.

Acked-by: Emil Velikov <emil.l.velikov@gmail.com>

egl/build: Sort egl sources alphabetically.

Acked-by: Emil Velikov <emil.l.velikov@gmail.com>

egl/build: Remove unused -DXF86VIDMODE.

Acked-by: Emil Velikov <emil.l.velikov@gmail.com>

egl/build: Include headers and XML in distribution.

Acked-by: Emil Velikov <emil.l.velikov@gmail.com>

egl/build: Drop two unnecessary Makefiles.

Acked-by: Emil Velikov <emil.l.velikov@gmail.com>

i965/fs: Remove left-over 'removed' variable.

I think this was used for coalescing out partly dead large virtual
registers, but the patch that enabled that caused regressions and didn't
make it upstream.

Reviewed-by: Eric Anholt <eric@anholt.net>

i965/fs: Check for interference after finding all channels.

It's more likely that we won't find writes to all channels than one will
interfere, and calculating interference is more expensive. This change
will also help prepare for coalescing load_payload instructions'
operands.

Also update the live intervals for all channels, and not just the last
that we saw.

Reviewed-by: Eric Anholt <eric@anholt.net>

i965: initialize more device info fields for Cherryview

The intent in 9b6b084eb7b10d006b44e3cd22585fc3e39e0c00 was
for urb .size and .min_vs_entries fields to use the values
from the GEN8_FEATURES macro.

Signed-off-by: Jordan Justen <jordan.l.justen@intel.com>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>

swrast: reindent s_texfetch_temp.h, remove trailing whitespace

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

swrast: remove out of date comments in s_texfetch_tmp.h

The comments were out of date and redundant (the functions are
pretty much self-explanatory).

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

swrast: rename texture fetch functions (pt. 7)

sed commands:
s/f_z24_s8/S8_UINT_Z24_UNORM/g
s/f_s8_z24/Z24_UNORM_S8_UINT/g
s/f_z16/Z_UNORM16/g
s/f_z32/Z_UNORM32/g
s/z32f_x24s8/Z32_FLOAT_S8X24_UINT/g
s/f_ycbcr_rev/YCBCR_REV/g
s/f_ycbcr/YCBCR/g
s/dudv8/DUDV8/g

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

swrast: rename texture fetch functions (pt. 6)

sed commands:
s/rgb9_e5/R9G9B9E5_FLOAT/g
s/r11_g11_b10f/R11G11B10_FLOAT/g
s/f_alpha_f16/A_FLOAT16/g
s/f_alpha_f32/A_FLOAT32/g
s/f_luminance_f16/L_FLOAT16/g
s/f_luminance_f32/L_FLOAT32/g
s/f_luminance_alpha_f16/LA_FLOAT16/g
s/f_luminance_alpha_f32/LA_FLOAT32/g
s/f_intensity_f16/I_FLOAT16/g
s/f_intensity_f32/I_FLOAT32/g
s/f_r_f16/R_FLOAT16/g
s/f_r_f32/R_FLOAT32/g
s/f_rg_f16/RG_FLOAT16/g
s/f_rg_f32/RG_FLOAT32/g
s/f_rgb_f16/RGB_FLOAT16/g
s/f_rgb_f32/RGB_FLOAT32/g
s/f_rgba_f16/RGBA_FLOAT16/g
s/f_rgba_f32/RGBA_FLOAT32/g
s/xbgr16161616_float/RGBX_FLOAT16/g
s/xbgr32323232_float/RGBX_FLOAT32/g

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

swrast: rename texture fetch functions (pt. 5)

sed commands:
s/srgba8/A8B8G8R8_SRGB/g
s/sargb8/B8G8R8A8_SRGB/g
s/sabgr8/R8G8B8A8_SRGB/g
s/sxbgr8/R8G8B8X8_SRGB/g
s/sla8/L8A8_SRGB/g
s/sl8/L_SRGB8/g
s/srgb8/BGR_SRGB8/g

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

swrast: rename texture fetch functions (pt. 4)

sed commands:
s/signed_rg1616/R16G16_SNORM/g
s/signed_rg88_rev/R8G8_SNORM/g
s/signed_al88/L8A8_SNORM/g
s/signed_a8/A_SNORM8/g
s/signed_a16/A_SNORM16/g
s/signed_l8/L_SNORM8/g
s/signed_l16/L_SNORM16/g
s/signed_i8/I_SNORM8/g
s/signed_i16/I_SNORM16/g
s/signed_r8/R_SNORM8/g
s/signed_r16/R_SNORM16/g
s/signed_al1616/LA_SNORM16/g
s/signed_rgb_16/RGB_SNORM16/g
s/signed_rgba_16/RGBA_SNORM16/g

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

swrast: rename texture fetch functions (pt. 3)

Rename functions to match format names.

sed commands:
s/f_rg1616_rev/G16R16_UNORM/g
s/f_rg1616/R16G16_UNORM/g
s/f_argb2101010/B10G10R10A2_UNORM/g
s/f_a8/A_UNORM8/g
s/f_a16/A_UNORM16/g
s/f_i8/I_UNORM8/g
s/f_i16/I_UNORM16/g
s/f_r8/R_UNORM8/g
s/f_r16/R_UNORM16/g
s/f_rgb888/BGR_UNORM8/g
s/f_bgr888/RGB_UNORM8/g
s/f_l8/L_UNORM8/g
s/f_l16/L_UNORM16/g
s/xbgr16161616_unorm/RGBX_UNORM16/g

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