We currently use CL_INVOCATION_COUNT for the GL_PRIMITIVES_GENERATED
query, which involves passing all primitives to the clipper. When
rasterizer discard is enabled, we program the clipper in REJECT_ALL
mode, rather than using the SOL stage's "Rendering Disable" feature.
See commit f09b91f782 for an explanation
of why we implement GL_PRIMITIVES_GENERATED this way.
Apparently the SOL stage's "Rendering Disable" feature is a lot faster
than having the clipper reject all primitives. It's safe to use when
no GL_PRIMITIVES_GENERATED query is active, as we don't care about
CL_INVOCATION_COUNT incrementing.
This patch makes us use SO_RENDERING_DISABLE when no query is active,
but continues falling back to the clipper in REJECT_ALL mode when the
queries are enabled. It brings back the perf_debug for the clipper
case (which I removed in commit 1f9445ff57, thinking it wasn't useful).
Improves performance in Gl32GSCloth by 84.8303% +/- 2.07132% (n = 10)
on my Broadwell GT2 laptop.
Cc: mesa-stable@lists.freedesktop.org
Signed-off-by: Kenneth Graunke <kenneth@whitecape.org>
Reviewed-by: Jason Ekstrand <jason@jlekstrand.net>
(cherry picked from commit b0629e6894)
They're basically the same. Let's avoid the code duplication.
v2: Fix SO_BUFFER_ENABLE stuff to only happen on Gen < 8 (caught
by Jason Ekstrand).
Cc: mesa-stable@lists.freedesktop.org
Signed-off-by: Kenneth Graunke <kenneth@whitecape.org>
Reviewed-by: Jason Ekstrand <jason@jlekstrand.net>
(cherry picked from commit 4db98f8beb)
Constant propagation on arrays doesn't make a lot of sense. If the
array is only accessed with constant indexes, then opt_array_splitting
would split it up. Otherwise, we have variable indexing. If there's
multiple accesses, then constant propagation would end up replicating
the data.
The lower_const_arrays_to_uniforms pass creates uniforms for each
ir_constant with array type that it encounters. This means that it
creates redundant uniforms for each copy of the constant, which means
uploading too much data. It can even mean exceeding the maximum number
of uniform components, causing link failures.
We could try and teach the pass to de-duplicate the data by hashing
constants, but it makes more sense to avoid duplicating it in the first
place. We should promote constant arrays to uniforms, then propagate
the uniform access.
Fixes the TressFX shaders from Tomb Raider, which exceeded the maximum
number of uniform components by a huge margin and failed to link.
On Broadwell:
total instructions in shared programs: 9067702 -> 9068202 (0.01%)
instructions in affected programs: 10335 -> 10835 (4.84%)
helped: 10 (Hoard, Shadow of Mordor, Amnesia: The Dark Descent)
HURT: 20 (Natural Selection 2)
loops in affected programs: 4 -> 0
The hurt programs appear to no longer have a constarray uniform, as
all constants were successfully propagated. Apparently before this
patch, we successfully unrolled a loop containing array access, but
only after promoting constant arrays to uniforms. With this patch,
we unroll it first, so all array access is direct, and the array
is split up, and individual constants are propagated. This seems
better.
Cc: mesa-stable@lists.freedesktop.org
Reported-by: Karol Herbst <nouveau@karolherbst.de>
Signed-off-by: Kenneth Graunke <kenneth@whitecape.org>
Reviewed-by: Timothy Arceri <timothy.arceri@collabora.com>
(cherry picked from commit fb857b5eea)
There's really no point in looking at ir_dereference_array of a
constant. It also misses cases like:
(assign () (var_ref tmp) (constant (array ...) ...))
No changes in shader-db, but keeps it working after the next commit.
Cc: mesa-stable@lists.freedesktop.org
Signed-off-by: Kenneth Graunke <kenneth@whitecape.org>
Reviewed-by: Timothy Arceri <timothy.arceri@collabora.com>
(cherry picked from commit ef78df8d3b)
The scalar backend currently doesn't support variable indexing on
temporary arrays, but it does support it on uniform arrays, and
some stages support it for input arrays. Make sure these are
propagated through before exploding indirects into piles of
if-ladders unnecessarily.
On Broadwell, no instruction count change in shader-db.
total cycles in shared programs: 80675652 -> 80674928 (-0.00%)
cycles in affected programs: 649972 -> 649248 (-0.11%)
helped: 386
HURT: 165
This will help avoid code quality regressions in a future commit.
Cc: mesa-stable@lists.freedesktop.org
Signed-off-by: Kenneth Graunke <kenneth@whitecape.org>
Reviewed-by: Timothy Arceri <timothy.arceri@collabora.com>
(cherry picked from commit f7741c5211)
The new uniform may need precise as well.
Fixes copy propagation of constant array uniforms in Tomb Raider shaders.
Cc: mesa-stable@lists.freedesktop.org
Signed-off-by: Kenneth Graunke <kenneth@whitecape.org>
Reviewed-by: Timothy Arceri <timothy.arceri@collabora.com>
(cherry picked from commit 586f4a42e7)
Previously, we failed to split constant arrays. Code such as
int[2] numbers = int[](1, 2);
would generates a whole-array assignment:
(assign () (var_ref numbers)
(constant (array int 4) (constant int 1) (constant int 2)))
opt_array_splitting generally tried to visit ir_dereference_array nodes,
and avoid recursing into the inner ir_dereference_variable. So if it
ever saw a ir_dereference_variable, it assumed this was a whole-array
read and bailed. However, in the above case, there's no array deref,
and we can totally handle it - we just have to "unroll" the assignment,
creating assignments for each element.
This was mitigated by the fact that we constant propagate whole arrays,
so a dereference of a single component would usually get the desired
single value anyway. However, I plan to stop doing that shortly;
early experiments with disabling constant propagation of arrays
revealed this shortcoming.
This patch causes some arrays in Gl32GSCloth's geometry shaders to be
split, which allows other optimizations to eliminate unused GS inputs.
The VS then doesn't have to write them, which eliminates the entire VS
(5 -> 2 instructions). It still renders correctly.
No other change in shader-db.
v2: Drop !AOA check and improve a comment (feedback from Tim Arceri).
Cc: mesa-stable@lists.freedesktop.org
Signed-off-by: Kenneth Graunke <kenneth@whitecape.org>
Reviewed-by: Timothy Arceri <timothy.arceri@collabora.com>
(cherry picked from commit c264fdbc07)
opt_constant_propagation.cpp contains constant folding code which can
actually do constant propagation in some cases. It was happily
propagating constants into the left-hand-side of assignments.
For example,
(assign () (var_ref temp) (constant ...))
would brilliantly be turned into:
(assign () (constant ...) (constant ....))
This is a bigger hammer than necessary - it prevents propagation
into the left-hand-side altogether. We could certainly do better
someday. Notably, the constant propagation pass itself already
takes this approach - it's just the constant propagation pass's
built-in constant folding code (which actually propagates, too)
that was broken.
No change in shader-db, but prevents regressions after future commits.
It seems plausible that this could be hit today, but I haven't seen it
happen.
Cc: mesa-stable@lists.freedesktop.org
Signed-off-by: Kenneth Graunke <kenneth@whitecape.org>
Reviewed-by: Timothy Arceri <timothy.arceri@collabora.com>
(cherry picked from commit acf5444044)
Earlier MSVC 2013 releases have troubles compiling some of our C99 code,
so make sure we have Update 4 to avoid confusion.
Cc: mesa-stable@lists.freedesktop.org
Reviewed-by: Brian Paul <brianp@vmware.com>
(cherry picked from commit 805dbdf06d)
This solves a race condition where we can end up having different stages
stomp on each other because they're all trying to scratch in the same BO
but they have different views of its layout.
Signed-off-by: Jason Ekstrand <jason@jlekstrand.net>
Cc: "12.0" <mesa-stable@lists.freedesktop.org>
(cherry picked from commit c2f2c8e407)
While we're here, we also fixup MEDIA_VFE_STATE and rename the field in
3DSTATE_VS on gen6-7.5 to be consistent with the others.
Signed-off-by: Jason Ekstrand <jason@jlekstrand.net>
Cc: "12.0" <mesa-stable@lists.freedesktop.org>
(cherry picked from commit 45c0f60999)
The pack header generation scripts can't handle the case where you have
two addresses in the same dword; they just take whatever is the last one.
This meant that the MCS address wasn't properly getting handled. Since we
don't care about append counters, we can just re-arrange the XML for now.
Reviewed-by: Chad Versace <chad.versace@intel.com>
Cc: "12.0" <mesa-stable@lists.freedesktop.org>
(cherry picked from commit 89ded099f8)
ISL was being a bit too clever for its own good and lowering the format for
us. This is all well and good *if* we always want to lower it. However,
the GL driver selectively lowers the format depending on whether the
surface is write-only or not.
Reviewed-by: Chad Versace <chad.versace@intel.com>
Cc: "12.0" <mesa-stable@lists.freedesktop.org>
(cherry picked from commit d82322eb18)
Ivy Bridge and above can handle up to 2^31 elements for RAW buffer
surfaces.
Reviewed-by: Chad Versace <chad.versace@intel.com>
Cc: "12.0" <mesa-stable@lists.freedesktop.org>
(cherry picked from commit 97f12773b8)
This field is ignored by the hardware in this case and, on very large 1-D
textures, it can end up being larger than the maximum allowed value.
Reviewed-by: Chad Versace <chad.versace@intel.com>
Cc: "12.0" <mesa-stable@lists.freedesktop.org>
(cherry picked from commit ce24097abe)
This matches better what happens on gen8 where the "Tiled Surface" and
"Tile Walke" bits are combined into a single two-bit value. This is also
more consistent with what the GL driver does.
Reviewed-by: Chad Versace <chad.versace@intel.com>
Cc: "12.0" <mesa-stable@lists.freedesktop.org>
(cherry picked from commit f47e23a8b6)
This hasn't ever been a problem in the past but it is recommended by the
hardware docs.
Reviewed-by: Chad Versace <chad.versace@intel.com>
Cc: "12.0" <mesa-stable@lists.freedesktop.org>
(cherry picked from commit 96706bad5f)
It seems safe to set it all the time, but this reduces the diff between
the way i965 does it and what ISL does.
Reviewed-by: Chad Versace <chad.versace@intel.com>
Cc: "12.0" <mesa-stable@lists.freedesktop.org>
(cherry picked from commit 14d7c16e50)
For depth/stencil 1-D textures on SKL, we want them layed out in the old
format that has been used since gen4. In order for the surface state
fill-out code to handle, this it needs to distinguish based on layout
rather than just dimensionality.
Reviewed-by: Chad Versace <chad.versace@intel.com>
Cc: "12.0" <mesa-stable@lists.freedesktop.org>
(cherry picked from commit 5d24e9cfa1)
This fixes 688 Vulkan CTS tests on Haswell.
Reviewed-by: Chad Versace <chad.versace@intel.com>
Cc: "12.0" <mesa-stable@lists.freedesktop.org>
(cherry picked from commit 6a43204afa)
The docs specify that this only matters for render targets and surfaces
used with typed dataport messages. On some platforms (gen4-6) the Depth
field has more bits than RenderTargetViewExtent so we can have textures
with more levels than we can render to.
Reviewed-by: Chad Versace <chad.versace@intel.com>
Cc: "12.0" <mesa-stable@lists.freedesktop.org>
(cherry picked from commit 215282c9f4)
According to the PRM, you can't set SurfaceArray for 3D or buffer textures.
There doesn't seem to be a good reason not to set it when we can. On the
other hand, if we don't set it we can end up getting strange results for
1-layer array textures such as textureSize() returning the wrong results.
Reviewed-by: Chad Versace <chad.versace@intel.com>
Cc: "12.0" <mesa-stable@lists.freedesktop.org>
(cherry picked from commit bb326f7b01)
We already set the bit in the few cases where it's required by the docs so
there's no need to set it all the time. This has no noticable perf impact
for Dota 2 on Vulkan with the time demo I have.
Reviewed-by: Chad Versace <chad.versace@intel.com>
Cc: "12.0" <mesa-stable@lists.freedesktop.org>
(cherry picked from commit d050ffbce9)
This commit switches clear colors to use #if's instead of a C if. This
lets us properly handle SNB where the clear color field doesn't exist.
Reviewed-by: Chad Versace <chad.versace@intel.com>
Cc: "12.0" <mesa-stable@lists.freedesktop.org>
(cherry picked from commit 87f0ffa646)
This moves the #if's around so that halign and valign have different sets
of #if conditions. This also prepares us for SNB because isl_to_gen_halign
is not defined at all on gen6.
Reviewed-by: Chad Versace <chad.versace@intel.com>
Cc: "12.0" <mesa-stable@lists.freedesktop.org>
(cherry picked from commit 62a5e6e031)
This is purely cosmetic, but it makes things look a bit more readable.
Reviewed-by: Chad Versace <chad.versace@intel.com>
Cc: "12.0" <mesa-stable@lists.freedesktop.org>
(cherry picked from commit a60ae9e10a)
This is purely cosmetic, but it makes things look a bit more readable.
Reviewed-by: Chad Versace <chad.versace@intel.com>
Cc: "12.0" <mesa-stable@lists.freedesktop.org>
(cherry picked from commit 70c8afc0c8)
This is purely cosmetic, but it makes things look a bit more readable.
Reviewed-by: Chad Versace <chad.versace@intel.com>
Cc: "12.0" <mesa-stable@lists.freedesktop.org>
(cherry picked from commit e66e70ef47)
They're already zero-initialized and we have no plans of doing anything
more interesting with them.
Reviewed-by: Chad Versace <chad.versace@intel.com>
Cc: "12.0" <mesa-stable@lists.freedesktop.org>
(cherry picked from commit 39baea551f)
While designated initializers are nice, they also force us to put some
things in the initializer and some things later. Surface state setup is
complicated enough that this really hurts readability in the long run.
Reviewed-by: Chad Versace <chad.versace@intel.com>
Cc: "12.0" <mesa-stable@lists.freedesktop.org>
(cherry picked from commit caf2af4181)
This is what gen7 does and it's nice to have a prefix
Reviewed-by: Chad Versace <chad.versace@intel.com>
Cc: "12.0" <mesa-stable@lists.freedesktop.org>
(cherry picked from commit de1d194856)
The PRM states that the values put in Width, Height, and Depth should be
various bits from the value size - 1. We seem to have done this wrong
more-or-less from the start.
Reviewed-by: Chad Versace <chad.versace@intel.com>
Cc: "11.1 11.2 12.0" <mesa-stable@lists.freedesktop.org>
(cherry picked from commit 2a1cc94d27)
Previously, we were incrementing length but not actually putting anything
in the Y coordinate. This meant that 1-D TXF operations had a garbage
array index. If the surface is emitted as 1-D non-array, the coordinate
gets discarded and it works fine. If it happens to be bound as an array
surface, it may count as an out-of-bounds array access and you get zero.
Reviewed-by: Ian Romanick <ian.d.romanick@intel.com>
Cc: "11.1 11.2 12.0" <mesa-stable@lists.freedesktop.org>
(cherry picked from commit 0195299c86)
We were adding in the base which is wrong because the values given in the
miptree are relative to zero and not the base layer/level.
Reviewed-by: Chad Versace <chad.versace@intel.com>
Reviewed-by: Topi Pohjolainen <topi.pohjolainen@intel.com>
Cc: "11.1 11.2 12.0" <mesa-stable@lists.freedesktop.org>
(cherry picked from commit 620f81d2ed)
The RenderTargetViewExtent field of RENDER_SURFACE_STATE is supposed to be
set to the depth of a 3-D texture when rendering. Unfortunatley, that
field is only 9 bits on Sandy Bridge and prior so we can't actually bind
a 3-D texturing for rendering if it has depth > 512. On Ivy Bridge, this
field was bumpped to 11 bits so we can go all the way up to 2048. On Iron
Lake and prior, we don't support layered rendering and we use OffsetX/Y
hacks to render to particular layers so 2048 is ok there too.
Signed-off-by: Jason Ekstrand <jason@jlekstrand.net>
Reviewed-by: Ian Romanick <ian.d.romanick@intel.com>
Cc: "11.1 11.2 12.0" <mesa-stable@lists.freedesktop.org>
(cherry picked from commit 6ba88bce64)
This is basically a direct translation of what we do for gen7.
Signed-off-by: Jason Ekstrand <jason@jlekstrand.net>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=83036
Cc: "11.1 11.2 12.0" <mesa-stable@lists.freedesktop.org>
(cherry picked from commit 0f9cd74aab)
This makes texture views sort-of work. It doesn't add full texture view
support for gen4-5 but it is enough to fix the GL_ARB_copy_image formats
piglit test on Iron Lake.
Signed-off-by: Jason Ekstrand <jason@jlekstrand.net>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=83036
Cc: "11.1 11.2 12.0" <mesa-stable@lists.freedesktop.org>
(cherry picked from commit ee39d3ba91)
There's special logic around finding gl_FragData. It latches onto any
array with FRAG_RESULT_DATA0. However gl_SecondaryFragDataEXT[], added
by GL_EXT_blend_func_extended, fits those parameters as well. The real
frag data array should have index 0 though, so we can use that to
distinguish them.
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=96617
Signed-off-by: Ilia Mirkin <imirkin@alum.mit.edu>
Cc: "11.1 11.2 12.0" <mesa-stable@lists.freedesktop.org>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
(cherry picked from commit 36ed1b695e)
The start instance is applied as an offset into the buffer directly,
ignoring the divisor, not as an instance id offset that respects the
divisor.
Signed-off-by: Ilia Mirkin <imirkin@alum.mit.edu>
Cc: "11.2 12.0" <mesa-stable@lists.freedesktop.org>
Reviewed-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
(cherry picked from commit 1f4bca798d)
The generic version gets this right already, but this was using an
incorrect formula in SSE.
Signed-off-by: Ilia Mirkin <imirkin@alum.mit.edu>
Cc: "11.2 12.0" <mesa-stable@lists.freedesktop.org>
Reviewed-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
(cherry picked from commit 5b0d64886d)
Ever since c2581a9375, the binding table layout has depended on the
pipeline. This means that whenever we change pipelines we also need to
re-emit binding tables for the new layout.
Signed-off-by: Jason Ekstrand <jason@jlekstrand.net>
Cc: "12.0" <mesa-stable@lists.freedesktop.org>
(cherry picked from commit 35b53c8d47)
It's tiny and fully generic so there's really no reason for it to be in a
gen7-specific file.
Signed-off-by: Jason Ekstrand <jason@jlekstrand.net>
Cc: "12.0" <mesa-stable@lists.freedesktop.org>
(cherry picked from commit 2bfe0c3374)
