The coordinate offsets set in the m1 header are for textureOffset;
they have nothing to do with textureGrad (TXD).

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

The same as 3e43adef95 but for Gen7.

This doesn't quite fix GL_ARB_depth_texture/fbo-clear-formats; there's
still a 1 pixel wide black line on the right edge of the smaller squares.

The results were entirely wrong before, and are at least close now.

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

Since wayland 4bde293ff8 we cant
lookup visuals, as we dont receive the visual token events.

The format for pixmap-images thus has to default to argb for now.

GLES uses GL_APIENTRYP instead of GLAPIENTRYP, which breaks with the
latest API table generation code. This fixes the issue by emitting a
definition for GL_APIENTRYP when generating the GLES files.

This prevents the error

    prog: for the -disable-mmx option: may only occur zero or one times!

when creating a new context after XCloseDisplay with DRI drivers linked
with a shared LLVM 2.8 library.

In particular, this fixes the case where a vertex shader only uses
generic vertex attributes (non-0th).  Before, we were no-op'ing the
glDrawArrays/Elements().

This fixes the new piglit pos-array test.

NOTE: This is a candidate for the 7.10 branch.

Previously, always did unorm8->float/nonlinear-to-linear conversion (using
lookup table), then convert back to nonlinear (using the expensive math
func pow among others), and finally convert back to int (assuming caller
wants unorm8), because the float texture fetch function is used for getting
the actual texel values. This should probably all be changed at some point,
but for now simply enable the memcpy path also for srgb formats (but if for
instance swizzling is required, still the whole conversion will be done).

In order to have the benefit of type checking, and detect missing
GLAPIENTRY keywords on public entrypoints.

NOTE: this is a candidate for the 7.10 branch.

Clip distance is calculated each time vertex position is written
which is suboptiomal is some cases but very safe.
User clip planes are an obsolete feature anyway.

Every time number of clip planes increases, the vertex program
is recompiled.
That ensures no overhead in normal case (no user clip planes)
and reasonable overhead otherwise.

Fixes 3D windows in compiz, and reflection effect in neverball.
Also fixes compiz expo plugin when windows were dragged and each
window shown 3 times.

Fixes
GL_ARB_depth_texture/fbo-clear-formats
GL_EXT_packed_depth_stencil/fbo-clear-formats

This was going to get in the way of separate depth/stencil (which
wants to know about both, and whether they are the same rb), and also
wasn't a sufficient flag for the fix in the following commit.

Fixes glean pointAtten.

Improves glbenchmark egypt performance 0.6% +/- 0.4% (n=6).

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

In the 16-wide rework, I missed that we were setting some things to be
SIMD16 mode (corresponding to their setup in emit_texture_gen4()).

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

This removes the stupid strict-conformance fallback code I broke when
adding ARB_sampler_objects.

Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=36572
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org> (v1)

These fields are documented to be in the payload, and though the FB
write docs say they *aren't* in the payload, for all other fields the
payload and header is structured so that no overwriting is required
except for non-default options.

It turns out there's nothing in the hardware preventing this.  It
appears that it ought to work on pre-gen6 as well, but just produces
GPU hangs.

Improves glbenchmark Egypt framerate 4.4% +/- 0.3% (n=3), and Pro by
2.6% +/- 0.6% (n=3).

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

As of gen6, alt-mode (which we use) MOVs of floats are not raw --
they'll modify infs/nans.  This broke discard and alpha test in
16-wide, where apparently the upper 8 bits of the pixel enables being
set were causing the whole value to get trashed upon being moved.
Treating the values as UD instead of float makes sure they get
preserved.  While I'm here, replace the two 8-wide moves of the halves
of the header with a single compressed move.

Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=36648
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>

This is part of fixing fbo-alphatest-nocolor -- a regression in
35e8fe5c99 after the initial regression,
that had us using a garbage BLEND_STATE[0] (in particular, the alpha
test enable) if no color buffer was bound.

I thought I was thwarted initially when I couldn't do conditional mod
on a MOV, and couldn't use two immediate constants in one instruction.
But g0 != g0 is also a way to produce a failing comparison.

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

Fixes fdo 36738.

Anisotropic filtering extension for swrast intended to be used by osmesa
to create high quality renderings.
Based on Higher Quality Elliptical Weighted Avarage Filter (EWA).
A 2nd implementation using footprint assembly is also provided.

Signed-off-by: Brian Paul <brianp@vmware.com>

Fixes failed assertion when calling _mesa_print_arrays() debug function.

Fixes https://bugs.freedesktop.org/show_bug.cgi?id=36651

NOTE: This is a candidate for the 7.10 branch.

This should help distinguish Sandybridge GT1/GT2 from Ivybridge GT1/GT2.

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

Correctly links against selinux library when MESA is built with --enable-selinux option.

Fixes bug #36333 in Freedesktop bugzilla

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

this seems like a logical thing to do and sets the correct st flags
for vertex textures.

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

This function was taking a lot more CPU than required due to it memsetting
a bunch of memory that didn't require it from what I can see.

We should only memset here when we are about to fill out the sampler,
otherwise we end up doing a bunch of memsets for everytime this function
is called, basically setting 0 memory to 0.

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

this is taken from a patch from Mathias Froehlich, just going to
stage it in a few pieces.

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

The change for GPU hanging in 13bab58f04
fell back even when rb == NULL, which is wrong for GLES2 and caused
segfaulting in GLES2 conformance.  For the GPU hang case (where the
broken 2D driver failed to allocate a BO for the window system
renderbuffer), it also would assertion fail/segfault immediately after
the fallback setup when the renderbuffer map failed.

Fixes GLES2 conformance packed_depth_stencil.

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

Texture LOD Bias is now S4.8 instead of S4.6;
Min LOD, and Max LOD are now U4.8 instead of U4.6.

Fixes piglit test tex-miplevel-selection.

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

Volume 5c 1.13.7 lists it as [PreDevILK] and [DevIVB+].

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

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

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

The data port messages for this are rather different.  For now, fail to
compile rather than hanging the GPU.

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

On gen4/5, the RNDZ and RNDE instructions return floor(x), but set special
"round increment bits" in the flag register; a predicated ADD (+1) fixes
the result.

The documentation still lists '.r' as existing, and says that the
predicated add is necessary, but it apparently lies.  According to the
simulator, BRW_CONDITIONAL_R (7) is not a valid conditional modifier
and the RNDZ and RNDE instructions simply produce the correct value.

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