Revert "glapi/es1: Don't mark core functions as static_dispatch=false"

This reverts commit f6e290f80cc6728647e9cee35546190f081197e2.

To fix the broken build.  The DRI-enabled build seems OK after reverting.
Th non-DRI/gallium build is still suffering from an unrelated issue in
the pipe-loader code.

mesa: Allow setting GL_TEXTURE_MAX_LEVEL to 0 with GL_TEXTURE_RECTANGLE.

Currently, we raise an error when doing this which breaks a conformance
test from the OpenGL samples pack. Even if this is a bit silly it is not
an error.

From http://www.opengl.org/wiki/Rectangle_Texture:

"Rectangle textures contain exactly one image; they cannot have mipmaps.
Therefore, any texture parameters that depend on LODs are irrelevant
when used with rectangle textures; attempting to set these parameters to
any value other than 0 will result in an error."

Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=76496

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

gallium/docs: fix silent math failures due to ~ and &

Signed-off-by: Ilia Mirkin <imirkin@alum.mit.edu>
Reviewed-by: Brian Paul <brianp@vmware.com>

gallium/docs: line up some of the equations

Signed-off-by: Ilia Mirkin <imirkin@alum.mit.edu>
Reviewed-by: Brian Paul <brianp@vmware.com>

gallium/docs: fix incorrect/missing references

Signed-off-by: Ilia Mirkin <imirkin@alum.mit.edu>
Reviewed-by: Brian Paul <brianp@vmware.com>

gallium/docs: fix use of _ in math sections

Signed-off-by: Ilia Mirkin <imirkin@alum.mit.edu>
Reviewed-by: Brian Paul <brianp@vmware.com>

gallium/docs: add format to index

Signed-off-by: Ilia Mirkin <imirkin@alum.mit.edu>
Reviewed-by: Brian Paul <brianp@vmware.com>

gallium/docs: fix a lot of bad formatting

Signed-off-by: Ilia Mirkin <imirkin@alum.mit.edu>
Reviewed-by: Brian Paul <brianp@vmware.com>

glsl: remove UBO fields from _mesa_glsl_parse_state

They are not needed since 514f8c7ec7cc1ab18be93cebb5b9bf970b1955a9.

Signed-off-by: Chia-I Wu <olvaffe@gmail.com>
Reviewed-by: Matt Turner <mattst88@gmail.com>

nv50: implement clear_buffer to accelerate ARB_clear_buffer_object

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

mesa/st: Accelerate ARB_clear_buffer_object with clear_buffer

Signed-off-by: Ilia Mirkin <imirkin@alum.mit.edu>
Reviewed-by: Marek Olšák <marek.olsak@amd.com>

gallium: add interface to clear buffers

Signed-off-by: Ilia Mirkin <imirkin@alum.mit.edu>
Reviewed-by: Marek Olšák <marek.olsak@amd.com>

mapi_abi: Remove ABI-check work arounds for functions that are no longer exported

The previous commit stopped exporting 21 libGLESv2 and 88 libGLESv1_CM
functions.  This removes the work-arounds for those functions from
ABI-check.

Signed-off-by: Ian Romanick <ian.d.romanick@intel.com>
Reviewed-by: Matt Turner <mattst88@gmail.com>
Acked-by: Chad Versace <chad.versace@linux.intel.com>

mapi_abi: Make ES1 and ES2 static_dispatch=false functions hidden

This has been a long standing issue with the ES libraries.  Functions
marked in the XML with 'static_dispatch=false' were still incorrectly
exported.  ABI-check is supposed to detect this case, but we have to
paper over failures every time a new extension is added.

This change will cause a big pile of functions to disappear from
libGLESv2 and libGLESv1_CM.

libGLESv2 loses (20 functions):

    glBindVertexArrayOES
    glCompressedTexImage3DOES
    glCompressedTexSubImage3DOES
    glCopyTexSubImage3DOES
    glDeleteVertexArraysOES
    glDiscardFramebufferEXT
    glDrawBuffersNV
    glFlushMappedBufferRangeEXT
    glFramebufferTexture3DOES
    glGenVertexArraysOES
    glGetBufferPointervOES
    glGetProgramBinaryOES
    glIsVertexArrayOES
    glMapBufferOES
    glMapBufferRangeEXT
    glProgramBinaryOES
    glReadBufferNV
    glTexImage3DOES
    glTexSubImage3DOES
    glUnmapBufferOES

libGLESv1_CM loses (88 functions):

    glAlphaFuncxOES
    glBindFramebufferOES
    glBindRenderbufferOES
    glBlendEquationOES
    glBlendEquationSeparateOES
    glBlendFuncSeparateOES
    glCheckFramebufferStatusOES
    glClearColorxOES
    glClearDepthfOES
    glClearDepthxOES
    glClipPlanefOES
    glClipPlanexOES
    glColor4xOES
    glDeleteFramebuffersOES
    glDeleteRenderbuffersOES
    glDepthRangefOES
    glDepthRangexOES
    glDiscardFramebufferEXT
    glDrawTexfOES
    glDrawTexfvOES
    glDrawTexiOES
    glDrawTexivOES
    glDrawTexsOES
    glDrawTexsvOES
    glDrawTexxOES
    glDrawTexxvOES
    glFlushMappedBufferRangeEXT
    glFogxOES
    glFogxvOES
    glFramebufferRenderbufferOES
    glFramebufferTexture2DOES
    glFrustumfOES
    glFrustumxOES
    glGenerateMipmapOES
    glGenFramebuffersOES
    glGenRenderbuffersOES
    glGetBufferPointervOES
    glGetClipPlanefOES
    glGetClipPlanexOES
    glGetFixedvOES
    glGetFramebufferAttachmentParameterivOES
    glGetLightxvOES
    glGetMaterialxvOES
    glGetRenderbufferParameterivOES
    glGetTexEnvxvOES
    glGetTexGenfvOES
    glGetTexGenivOES
    glGetTexGenxvOES
    glGetTexParameterxvOES
    glIsFramebufferOES
    glIsRenderbufferOES
    glLightModelxOES
    glLightModelxvOES
    glLightxOES
    glLightxvOES
    glLineWidthxOES
    glLoadMatrixxOES
    glMapBufferOES
    glMapBufferRangeEXT
    glMaterialxOES
    glMaterialxvOES
    glMultiTexCoord4xOES
    glMultMatrixxOES
    glNormal3xOES
    glOrthofOES
    glOrthoxOES
    glPointParameterxOES
    glPointParameterxvOES
    glPointSizePointerOES
    glPointSizexOES
    glPolygonOffsetxOES
    glQueryMatrixxOES
    glRenderbufferStorageOES
    glRotatexOES
    glSampleCoveragexOES
    glScalexOES
    glTexEnvxOES
    glTexEnvxvOES
    glTexGenfOES
    glTexGenfvOES
    glTexGeniOES
    glTexGenivOES
    glTexGenxOES
    glTexGenxvOES
    glTexParameterxOES
    glTexParameterxvOES
    glTranslatexOES
    glUnmapBufferOES

Signed-off-by: Ian Romanick <ian.d.romanick@intel.com>
Reviewed-by: Matt Turner <mattst88@gmail.com>
Reviewed-by: Chad Versace <chad.versace@linux.intel.com>
Reviewed-by: Chia-I Wu <olv@lunarg.com>
Cc: Paul Berry <stereotype441@gmail.com>

mapi: Hack around glGetInternalformativ not being hidden in GLES

This is hella ugly.  The same-named function in desktop OpenGL is
hidden, but it needs to be exposed by libGLESv2 for OpenGL ES 3.0.
There's no way to express in the XML that a function should be be hidden
in one API but exposed in another.

This won't affect any change now, but it will prevent a regression in a
later patch.

Signed-off-by: Ian Romanick <ian.d.romanick@intel.com>
Reviewed-by: Matt Turner <mattst88@gmail.com>
Reviewed-by: Chad Versace <chad.versace@linux.intel.com>

glapi/es1: Don't mark core functions as static_dispatch=false

Functions that are part of OpenGL ES 1.0 or 1.1 should have static
dispatch functions in libGLESv1_CM.  This doesn't affect any change yet,
but it will prevent later regressions.

Signed-off-by: Ian Romanick <ian.d.romanick@intel.com>
Reviewed-by: Matt Turner <mattst88@gmail.com>
Acked-by: Chad Versace <chad.versace@linux.intel.com>

glapi: Mark all GL_ARB_separate_shader_objects functions with static_dispatch=false

This prevents the entrypoints from being (incorrectly) advertised by
libGL.

Signed-off-by: Ian Romanick <ian.d.romanick@intel.com>
Reviewed-by: Matt Turner <mattst88@gmail.com>
Acked-by: Chad Versace <chad.versace@linux.intel.com>

glapi: Remove some duplicate ignore="true" lines

It looks like these were added accidentally by Paul in commit 1a1db174.
From the commit message and the look of the patch, I think this was just
some sed-job left overs.

Signed-off-by: Ian Romanick <ian.d.romanick@intel.com>
Reviewed-by: Matt Turner <mattst88@gmail.com>
Reviewed-by: Chad Versace <chad.versace@linux.intel.com>

i965/vec4: Don't trim writemasks of texture instructions.

It was my understanding that the writemask works in SIMD4x2 mode for
texturing instructions and doesn't require a message header. Some bit of
this logic must be wrong, so disable it until it's understood.

Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=76617
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>

automake: ask the linker to do garbage collection

By doing GC the linker removes all the symbols that are not referenced
and/or used by the final library. This results in a saving of ~100K
up-to ~600K per (stripped) binary (classic vs gallium drivers).

If interested one can ask the compiler to print the sections that are
removed using -Wl,--print-gc-sections.

v2: Check if ld supports the flag before using it.

Signed-off-by: Emil Velikov <emil.l.velikov@gmail.com>
Acked-by: Matt Turner <mattst88@gmail.com> (v1)

automake: add -Wl,--no-undefined to all libraries

... apart from the dri drivers.
With this final change we can build mesa without fear that
the resulting libraries will have unresolved symbols.

Signed-off-by: Emil Velikov <emil.l.velikov@gmail.com>
Reviewed-by: Matt Turner <mattst88@gmail.com>

gallium/targets: add missing library dependencies

Signed-off-by: Emil Velikov <emil.l.velikov@gmail.com>
Reviewed-by: Matt Turner <mattst88@gmail.com>

pipe-loader: reorder PIPE_LIBS

Reorder -lm, -lrt, -lpthreads and -ldl to be consistent with the
rest of mesa.

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

pipe-loader: use PTHREAD_LIBS over -lpthread

Signed-off-by: Emil Velikov <emil.l.velikov@gmail.com>
Reviewed-by: Matt Turner <mattst88@gmail.com>

dri/i965: use CLOCK_LIBS over -lrt

Signed-off-by: Emil Velikov <emil.l.velikov@gmail.com>
Reviewed-by: Matt Turner <mattst88@gmail.com>

automake: consistently use -no-undefined

Set the flag for all but the dri targets. They have missing
glapi symbols which are required for the normal operation with
the X server.

Jon, I fear that you'll need to carry the "no-undefined" hunk
locally when building the dri drivers under cygwin.

Cc: Jon TURNEY <jon.turney@dronecode.org.uk>
Signed-off-by: Emil Velikov <emil.l.velikov@gmail.com>
Reviewed-by: Matt Turner <mattst88@gmail.com>

targets/egl-static: move the common LDFLAGS into AM_LDFLAGS

Signed-off-by: Emil Velikov <emil.l.velikov@gmail.com>
Reviewed-by: Matt Turner <mattst88@gmail.com>

targets/omx: do not link against the trace driver

Unused due to the missing GALLIUM_TRACE define.

Requested-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Emil Velikov <emil.l.velikov@gmail.com>
Reviewed-by: Christian König <christian.koenig@amd.com>

gallium/targets: explicitly include a dummy.cpp and remove all the LINK mayhem

Explicitly setting the linker variable was required for old and broken
build toolchains. At this point this should no longer be needed, and
setting the sources lists will trigger generation of the correct LINK
variables.

Explicitly include dummy.cpp to use g++ to link the static library which
in most cases is based upon C++ code.

v2: Reword commit message.

Signed-off-by: Emil Velikov <emil.l.velikov@gmail.com>
Reviewed-by: Matt Turner <mattst88@gmail.com>

gallium/targets: move LLVM_LIBS handling inside Automake.inc

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

gallium/targets: fold LLVM_LDFLAGS inside Automake.inc

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

targets/omx: use GALLIUM_OMX_LINKER_FLAGS

Signed-off-by: Emil Velikov <emil.l.velikov@gmail.com>
Reviewed-by: Christian König <christian.koenig@amd.com>

targets/omx: introduce GALLIUM_OMX_LIB_DEPS

Signed-off-by: Emil Velikov <emil.l.velikov@gmail.com>
Reviewed-by: Christian König <christian.koenig@amd.com>

targets/pipe-loader: move LLVM_LIBS handling inside PIPE_LIBS

This lets us have only one if HAVE_MESA_LLVM block, rather than
one for each driver.

Signed-off-by: Emil Velikov <emil.l.velikov@gmail.com>
Reviewed-by: Matt Turner <mattst88@gmail.com>

targets/pipe-loader: include dummy.cpp irrespective of HAVE_MESA_LLVM

Signed-off-by: Emil Velikov <emil.l.velikov@gmail.com>
Reviewed-by: Matt Turner <mattst88@gmail.com>

targets/pipe-loader: compact duplicating LDFLAGS

Every library uses the same libtool/linker flags. Compact those
into AM_LDFLAGS and append the version script to it.

Signed-off-by: Emil Velikov <emil.l.velikov@gmail.com>
Reviewed-by: Matt Turner <mattst88@gmail.com>

pipe-loader/swrast: add soft/llvmpipe defines

Or it compiles them in, but pretends they don't exist

v2: Rebase (Emil)

Signed-off-by: Emil Velikov <emil.l.velikov@gmail.com>
Reviewed-by: Matt Turner <mattst88@gmail.com>

targets/xa: drop libudev references from automake build

Mesa does _not_ link against libudev. Additionally the only place
that deals with it is the loader, thus we can drop the CFLAGS.

Signed-off-by: Emil Velikov <emil.l.velikov@gmail.com>
Reviewed-by: Matt Turner <mattst88@gmail.com>

dri/common: LIBDRM_LIBS is not a linker/libtool flag, add it to LIBADD

Signed-off-by: Emil Velikov <emil.l.velikov@gmail.com>
Reviewed-by: Matt Turner <mattst88@gmail.com>

drivers/x11: GL_LIB_DEPS is not a linker/libtool flag, add it to LIBADD

Signed-off-by: Emil Velikov <emil.l.velikov@gmail.com>
Reviewed-by: Matt Turner <mattst88@gmail.com>

configure: autodetect video state-trackers when non swrast driver is present

It makes little sense to enable the vdpau, xvmc and omx state-trackers
as they do not make use of (don't work with) the software driver.

Signed-off-by: Emil Velikov <emil.l.velikov@gmail.com>
Reviewed-by: Matt Turner <mattst88@gmail.com>

configure: use grep in quiet mode, rather than piping stderr/stdout to /dev/null

grep -q is easier to read and consistent with the rest of configure.ac.

Signed-off-by: Emil Velikov <emil.l.velikov@gmail.com>
Reviewed-by: Matt Turner <mattst88@gmail.com>

configure: error out when building gallium-osmesa without softpipe

Gallium osmesa links against the softpipe driver, thus the build
will fail if it's missing.

Signed-off-by: Emil Velikov <emil.l.velikov@gmail.com>
Acked-by: Matt Turner <mattst88@gmail.com>

Partially revert "automake: allow only shared builds"

Evidently at least static OSMesa is still used as shared one
causes substantial increase in the load time for some programs
that use it (from seconds up-to ~30min).

Rather than forcing everyone to use shared mesa, revert commit
a6efbac9fb502c4f0166e7a0680b6828e1f6926c and default to shared
build when both shared and static are disabled.

v2: Whitespace cleanup, drop silly comment.

Reported-by: Burlen Loring <burlen.loring@gmail.com>
Reviewed-by: Matt Turner <mattst88@gmail.com>

configure: enable dri3 only for linux

Currently only linux can make use of dri3, so it would make sense to
enable it explicitly for the platform.
Drop a duplicated libudev check while we're at it.

v3: Properly handle dri3 and reword commit message.

Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=76377
Cc: "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>
Reviewed-by: Matt Turner <mattst88@gmail.com>

mesa: Fix format matching checks for GL_INTENSITY* internalformats.

GL_INTENSITY has never been valid as a pixel format -- to get the memcpy
pack/unpack paths, the app needs to specify GL_RED as the pixel format
(or GL_RED_INTEGER for the integer formats).

Note: This was briefly merged before, but exposed some breakage in gallium, so
was reverted. Hopefully it will stick this time.

Signed-off-by: Chris Forbes <chrisf@ijw.co.nz>
Reviewed-by: Brian Paul <brianp@vmware.com>

st: fix st_choose_matching_format to ignore intensity

_mesa_format_matches_format_and_type() returns true for
GL_RED/GL_RED_INTEGER (with an appropriate type) into an intensity
mesa_format.

We want the `red`-based format instead, regardless of the order we find
them in our walk of the mesa formats list.

Signed-off-by: Chris Forbes <chrisf@ijw.co.nz>
Reviewed-by: Brian Paul <brianp@vmware.com>
Reviewed-by: Marek Olšák <marek.olsak@amd.com>

mesa: fix texstore for MESA_FORMAT_R8G8B8A8_SRGB

The case for this was in the wrong function, and this format's store
func was not set in the table at all.

Signed-off-by: Chris Forbes <chrisf@ijw.co.nz>
Reviewed-by: Brian Paul <brianp@vmware.com>

freedreno/a3xx/compiler: fix RECT textures

Whether or not the coords are normalized is handled in the texture
state.  But we otherwise need to treat RECT sample instructions as 2D.

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

freedreno/a3xx/compiler: avoid negative register ids

In some cases, we need a register to be assigned up to three components
before the base.  Since we can't have negative register #'s, just shift
everything up.  May increase register usage for trivial shaders, but I
don't think we are shader limited in those cases.  A proper solution is
going to require a better register assignment algorithm (which is on the
TODO list), this is just a hack to get us by until then.

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

freedreno/a3xx: missing wfi

RB_FRAME_BUFFER_DIMENSION is not a banked context register, so we need
to wait for the GPU to idle before updating it.  But we'd rather not
have unnecessary WFI's, so actually keep track if we need to emit it or
not.

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

freedreno/a3xx: little extra debug

Catch things which should not happen in debug builds.

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

freedreno: handle null sampler

This is something that XA triggers.  In some cases it will only use
SAMP[1] (composite mask) but not SAMP[0] (composite src).

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

i965: Add Cherryview support.

Based on a patch by Ville Syrjälä.

As usual, these are placeholder values; actual values will come later.

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

glsl: Clean up "unused parameter" warnings

../../src/glsl/builtin_functions.cpp:72:1: warning: unused parameter 'state' [-Wunused-parameter]

../../src/glsl/ir_clone.cpp:31:1: warning: unused parameter 'ht' [-Wunused-parameter]

../../src/glsl/ir_equals.cpp:44:1: warning: unused parameter 'ir' [-Wunused-parameter]
../../src/glsl/ir_equals.cpp:50:1: warning: unused parameter 'ignore' [-Wunused-parameter]
../../src/glsl/ir_equals.cpp:68:1: warning: unused parameter 'ignore' [-Wunused-parameter]

../../src/glsl/ir_print_visitor.cpp:149:6: warning: unused parameter 'ir' [-Wunused-parameter]
../../src/glsl/ir_print_visitor.cpp:556:1: warning: unused parameter 'ir' [-Wunused-parameter]
../../src/glsl/ir_print_visitor.cpp:562:1: warning: unused parameter 'ir' [-Wunused-parameter]

../../src/glsl/link_uniforms.cpp:213:1: warning: unused parameter 'record_type' [-Wunused-parameter]

../../src/glsl/loop_analysis.cpp:225:1: warning: unused parameter 'ir' [-Wunused-parameter]

../../src/glsl/loop_unroll.cpp:73:30: warning: unused parameter 'ir' [-Wunused-parameter]
../../src/glsl/loop_unroll.cpp:79:30: warning: unused parameter 'ir' [-Wunused-parameter]
../../src/glsl/loop_unroll.cpp:85:30: warning: unused parameter 'ir' [-Wunused-parameter]

../../src/glsl/opt_copy_propagation_elements.cpp:189:1: warning: unused parameter 'ir' [-Wunused-parameter]

../../src/glsl/opt_cse.cpp:402:1: warning: unused parameter 'ir' [-Wunused-parameter]

../../src/glsl/opt_dead_code_local.cpp:117:30: warning: unused parameter 'ir' [-Wunused-parameter]

../../src/glsl/opt_redundant_jumps.cpp:53:1: warning: unused parameter 'ir' [-Wunused-parameter]

../../src/glsl/opt_vectorize.cpp:301:1: warning: unused parameter 'ir' [-Wunused-parameter]

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

mesa: Clean up "unused parameter" warnings

program/ir_to_mesa.cpp:2008:1: warning: unused parameter 'ir' [-Wunused-parameter]
program/ir_to_mesa.cpp:2272:1: warning: unused parameter 'ir' [-Wunused-parameter]
program/ir_to_mesa.cpp:2278:1: warning: unused parameter 'ir' [-Wunused-parameter]

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

mesa/program: Constify find_variable_storage

Also clean up an old whitespace blooper.

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

glsl: Move Doxygen block closing ot the correct place

This is the closing for the "\defgroup IR Intermediate representation
nodes" all the way at the top of the file.

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

i965: Make sure we always compute valid index bounds before drawing.

When doing software rendering (i.e. rendering to the selection buffer) we need
to make sure that we have valid index bounds before calling _tnl_draw_prims(),
otherwise we can crash.

Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=59455
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>

glsl: remove {add,get}_type_ast from glsl_symbol_table

They are not needed since 0da1a2cc369052643ccaea75a1722cc37652d82a.

Signed-off-by: Chia-I Wu <olvaffe@gmail.com>
Reviewed-by: Matt Turner <mattst88@gmail.com>

mesa: fix glMultiDrawArrays inside a display list

The underlying glDrawArrays() calls weren't getting compiled into
the display list.  We simply need to use the current dispatch table
so the CALL_DrawArrays() is routed to the display list save function.

This patch also fixes glMultiModeDrawArraysIBM and
glMultiModeDrawElementsIBM.

Fixes the new piglit gl-1.4-dlist-multidrawarrays test.

Cc: "10.0" "10.1" <mesa-stable@lists.freedesktop.org>
Reviewed-by: José Fonseca <jfonseca@vmware.com>
Reviewed-by: Ian Romanick <ian.d.romanick@intel.com>

st/mesa: overhaul texture / sample swizzle code

Previously we only examined the GL_DEPTH_MODE state to determine the
sampler view swizzle for depth textures.  Now we also consider the
texture base format for color textures too.

The basic idea is if we're sampling from a RGB texture we always
want to get A=1, even if the actual hardware format might be RGBA.
We had assumed that the texture's A values were always one since that's
what Mesa's texstore code does.  But if we render to the RGBA texture,
the A values might not be 1.  Subsequent sampling didn't return the
right values.

Now we examine the user-specified texture base format vs. the actual
gallium format to determine the right swizzle.

Fixes several fbo-blending-formats, fbo-clear-formats and fbo-tex-rgbx
failures with VMware/svga driver (and possibly other drivers).
No other piglit regressions with softpipe or VMware/svga.

Reviewed-by: Marek Olšák <maraeo@gmail.com>

st/mesa: simplify apply_depthmode()

In preparation for following changes.

I used a temporary test harness to compare the old code to the new
for all possible swizzle inputs.  No change in results.

i965: Use intel_upload_space() for pull constant uploads.

This also happens to fix a leak of the current GS pull constant BO on
context destroy, by just not holding on to the pull const bos after the
surface state is generated.

No statistically significant performance difference on GLB2.7 on HSW at
1024x768 (n=40) or 320x240 (n=44), or on BYT at 320x240 (n=47).

v2: Rebase on intel_upload simplification.

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

i965: Massively simplify the intel_upload implementation.

The implementation kept a page-sized area for uploading data, and
uploaded chunks from that to a 64kb-sized streamed buffer.  This wasted
cache footprint (and extra state tracking to do so) when we want to just
write our data into the buffer immediately.

Instead, build it around an interface like brw_state_batch() that just
gets you a pointer to BO memory to upload your stuff immediately.

Improves OpenArena on HSW by 1.62209% +/- 0.355299% (n=61) and on BYT by
1.7916% +/- 0.415743% (n=31).

v2: Rebase on Mesa master, drop old prototypes.  Re-do performance
    comparison on a kernel that doesn't punish CPU efficiency
    improvements.

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

draw/llvm: improve debugging output a bit

it's useful to know what the llvmbuildstore arguments are going to
be before executing it because it can crash and make sure to
print out the inputs only if we're not generating a gs because
it fetches inputs differently.

Signed-off-by: Zack Rusin <zackr@vmware.com>
Reviewed-by: Jose Fonseca <jfonseca@vmware.com>
Reviewed-by: Roland Scheidegger <sroland@vmware.com>

draw/gs: reduce the size of the gs output buffer

We used to overallocate the output buffer sometimes running out
of memory with applications rendering large geometries. The actual
maximum number of vertices out is simply the maximum number of
primitives in (number of gs invocations) multiplied by the maximum
number of output vertices per gs input primitive (i.e. gs invocation).

Signed-off-by: Zack Rusin <zackr@vmware.com>
Reviewed-by: Jose Fonseca <jfonseca@vmware.com>
Reviewed-by: Roland Scheidegger <sroland@vmware.com>

svga: add work-around for Sauerbraten Z fighting issue

Reviewed-by: Jose Fonseca <jfonseca@vmware.com>

svga: null out query's hwbuf pointer after destroying

Just to be extra safe.

Reviewed-by: Jose Fonseca <jfonseca@vmware.com>

svga: add some debug_printf() calls in the query object code

To help debug failures.

Reviewed-by: Jose Fonseca <jfonseca@vmware.com>

st/mesa: add null pointer checking in query object functions

Don't pass null query object pointers into gallium functions.
This avoids segfaulting in the VMware driver (and others?) if the
pipe_context::create_query() call fails and returns NULL.

Cc: "10.0" "10.1" <mesa-stable@lists.freedesktop.org>
Reviewed-by: Roland Scheidegger <sroland@vmware.com>

svga: fix a comment (sampler vs. sampler_view)

mesa: fix unpack_Z32_FLOAT_X24S8() / unpack_Z32_FLOAT() mix-up

And use the z32f_x24s8 helper struct in unpack_Z32_FLOAT_X24S8().
Cc: "10.0" "10.1" <mesa-stable@lists.freedesktop.org>
Reviewed-by: Roland Scheidegger <sroland@vmware.com>

mesa: fix indentation, formatting, etc in fbobject.c

mesa: rename format_(un)pack.c functions to match format names (pt. 7)

sed commands:
s/z_Z24_S8\b/S8_UINT_Z24_UNORM/g
s/z_S8_Z24\b/Z24_UNORM_S8_UINT/g
s/z_Z16\b/Z_UNORM16/g
s/z_Z32\b/Z_UNORM32/g
s/z_Z32_FLOAT/Z_FLOAT32/g

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

mesa: rename format_(un)pack.c functions to match format names (pt. 6)

sed commands:
s/ARGB2101010_UINT\b/B10G10R10A2_UINT/g
s/ABGR2101010_UINT\b/R10G10B10A2_UINT/g

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

mesa: rename format_(un)pack.c functions to match format names (pt. 5)

sed commands:
s/SIGNED_R_UNORM8\b/R_SNORM8/g
s/SIGNED_RG88_REV\b/R8G8_SNORM/g
s/SIGNED_RGBX8888\b/X8B8G8R8_SNORM/g
s/SIGNED_A8B8G8R8_UNORM\b/A8B8G8R8_SNORM/g
s/SIGNED_R8G8B8A8_UNORM\b/R8G8B8A8_SNORM/g
s/SIGNED_R_UNORM16\b/R_SNORM16/g
s/SIGNED_R16G16_UNORM\b/R16G16_SNORM/g
s/SIGNED_RGB_16\b/RGB_SNORM16/g
s/SIGNED_RGBA_16\b/RGBA_SNORM16/g
s/SIGNED_A_UNORM8\b/A_SNORM8/g
s/SIGNED_L_UNORM8\b/L_SNORM8/g
s/SIGNED_L8A8_UNORM\b/L8A8_SNORM/g
s/SIGNED_L_UNORM8\b/I_SNORM8/g
s/SIGNED_A_UNORM16\b/A_SNORM16/g
s/SIGNED_L_UNORM16\b/L_SNORM16/g
s/SIGNED_L16A16_UNORM\b/LA_SNORM16/g
s/SIGNED_L_UNORM16\b/I_SNORM16/g
s/XBGR16161616_SNORM\b/RGBX_SNORM16/g
s/SIGNED_G8R8_UNORM\b/G8R8_SNORM/g
s/SIGNED_G16R16_UNORM\b/G16R16_SNORM/g
s/SIGNED_I_UNORM8\b/I_SNORM8/g
s/SIGNED_I_UNORM16\b/I_SNORM16/g

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

mesa: rename format_(un)pack.c functions to match format names (pt. 4)

sed commands:
s/SRGBA_UNORM8\b/A8B8G8R8_SRGB/g
s/SABGR_UNORM8\b/R8G8B8A8_SRGB/g
s/SARGB8\b/B8G8R8A8_SRGB/g
s/XBGR8888_SRGB\b/R8G8B8X8_SRGB/g
s/XRGB8888_SRGB\b/B8G8R8X8_SRGB/g
s/SL_UNORM8\b/L_SRGB8/g
s/SLA_UNORM8\b/L8A8_SRGB/g

manually changed SRGB8 -> BGR_SRGB8

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

mesa: rename format_(un)pack.c functions to match format names (pt. 3)

sed commands:
s/LUMINANCE_FLOAT32\b/L_FLOAT32/g
s/LUMINANCE_FLOAT16\b/L_FLOAT16/g
s/LUMINANCE_ALPHA_FLOAT32\b/LA_FLOAT32/g
s/LUMINANCE_ALPHA_FLOAT16\b/LA_FLOAT16/g
s/ALPHA_FLOAT32\b/A_FLOAT32/g
s/ALPHA_FLOAT16\b/A_FLOAT16/g
s/XBGR32323232_FLOAT\b/RGBX_FLOAT32/g
s/RGB9_E5_FLOAT\b/R9G9B9E5_FLOAT/g
s/R11_G11_B10_FLOAT\b/R11G11B10_FLOAT/g
s/INTENSITY_FLOAT16\b/I_FLOAT16/g
s/INTENSITY_FLOAT32\b/I_FLOAT32/g

v2: removed a few redundant/no-op substitutions

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

mesa: rename format_(un)pack.c functions to match format names (pt. 2)

sed commands:
s/ABGR2101010\b/R10G10B10A2_UNORM/g
s/XRGB2101010_UNORM\b/B10G10R10X2_UNORM/g
s/XBGR16161616_UNORM\b/RGBX_UNORM16/g
s/ABGR2101010\b/R10G10B10A2_UNORM/g
s/I8\b/I_UNORM8/g
s/I16\b/I_UNORM16/g

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

mesa: rename format_(un)pack.c functions to match format names (pt. 1)

sed commands:
s/RGBA8888\b/A8B8G8R8_UNORM/g
s/RGBA8888_REV\b/R8G8B8A8_UNORM/g
s/ARGB8888\b/B8G8R8A8_UNORM/g
s/ARGB8888_REV\b/A8R8G8B8_UNORM/g
s/RGBA8888\b/X8B8G8R8_UNORM/g
s/RGBA8888_REV\b/R8G8B8X8_UNORM/g
s/XRGB8888\b/B8G8R8X8_UNORM/g
s/XRGB8888_REV\b/X8R8G8B8_UNORM/g
s/RGB888\b/BGR_UNORM8/g
s/BGR888\b/RGB_UNORM8/g
s/RGB565\b/B5G6R5_UNORM/g
s/RGB565_REV\b/R5G6B5_UNORM/g
s/ARGB4444\b/B4G4R4A4_UNORM/g
s/ARGB4444_REV\b/A4R4G4B4_UNORM/g
s/RGBA5551\b/A1B5G5R5_UNORM/g
s/ARGB1555\b/B5G5R5A1_UNORM/g
s/ARGB1555_REV\b/A1R5G5B5_UNORM/g
s/AL44\b/L4A4_UNORM/g
s/AL88\b/L8A8_UNORM/g
s/AL88_REV\b/A8L8_UNORM/g
s/AL1616\b/L16A16_UNORM/g
s/AL1616_REV\b/A16L16_UNORM/g
s/RGB332\b/B2G3R3_UNORM/g
s/A8\b/A_UNORM8/g
s/A16\b/A_UNORM16/g
s/L8\b/L_UNORM8/g
s/L16\b/L_UNORM16/g
s/L8\b/I_UNORM8/g
s/L16\b/I_UNORM16/g
s/R8\b/R_UNORM8/g
s/GR88\b/R8G8_UNORM/g
s/RG88\b/G8R8_UNORM/g
s/R16\b/R_UNORM16/g
s/GR1616\b/R16G16_UNORM/g
s/RG1616\b/G16R16_UNORM/g
s/ARGB2101010\b/B10G10R10A2_UNORM/g

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

llvmpipe: Fix llvmpipe_create_gs_state.

Revert unintended behaviour change from commit
b995a010e688bc4d4557e973e5e28091c378e881.

Tested-by: José Fonseca <jfonseca@vmware.com>

st/omx/dec: fix possible segfault at eos

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

mapi/glapi: Use ElementTree instead of libxml2.

It is quite hard to meet the dependency of the libxml2 python bindings
outside Linux, and in particularly on MacOSX; whereas ElementTree is
part of Python's standard library.  ElementTree is more limited than
libxml2: no DTD verification, defaults from DTD, or XInclude support,
but none of these limitations is serious enough to justify using
libxml2.

In fact, it was easier to refactor the code to use ElementTree than to
try to get libxml2 python bindings.

In the process, gl_item_factory class was refactored so that there is
one method for each kind of object to be created, as it simplifies
things substantially.

I confirmed that precisely the same output is generated for GL/GLX/GLES.

v2: Remove m4/ax_python_module.m4 as suggested by Matt Turner.

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

mapi/glapi: Remove glX_doc.py.

As suggested by Ian Romanick, given it's no longer used.

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

st/mesa: fix sampler view handling with shared textures v4

Release the references to the sampler views before
destroying the pipe context.

v2: remove TODO and unrelated change
v3: move to st_texture.[ch], rename callback, add comment
v4: fix rebase mess up and add further cleanups

Signed-off-by: Christian König <christian.koenig@amd.com>
Reviewed-by: Brian Paul <brianp@vmware.com>
Cc: "10.0 10.1" <mesa-stable@lists.freedesktop.org>

gallivm: fix no-op n:n lp_build_resize()

This can get called in some circumstances if both src type and dst type
have same width (seen with float32->unorm32). While this particular case
was bogus anyway let's just fix that as it can work trivially (due to the
way it was called it actually worked anyway apart from the assert).

Reviewed-by: Zack Rusin <zackr@vmware.com>
Reviewed-by: Jose Fonseca <jfonseca@vmware.com>

i965: For fast color clears, only check the color of live channels.

When deciding if a clear color is suitable for fast clear,
take into account if a color channel is active in the
buffer format.

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

i965: Set Broadwell MOCS values everywhere it's possible.

This patch introduces two pre-canned MOCS values: BDW_MOCS_WB
(write-back, all caches) and BDW_MOCS_WT (write-through, all caches).

We use write-through caching for render targets, and write-back for
all other data.  (At least on Haswell, I believe write-back LLC/eLLC
didn't work for scan-out buffers, while write-through did.)

No performance analysis has been done on the impact of this patch.

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

mesa: In core profile, refuse to draw unless a VAO is bound.

Core profile requires a non-default VAO to be bound.  Currently, calls
to glVertexAttribPointer raise INVALID_OPERATION unless a VAO is bound,
and we never actually get any vertex data set.  Trying to draw without
any vertex data can only cause problems.  In i965, it causes a crash.

Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=76400
Signed-off-by: Kenneth Graunke <kenneth@whitecape.org>
Reviewed-by: Ian Romanick <ian.d.romanick@intel.com>
Cc: mesa-stable@lists.freedesktop.org

Revert "build: llvm libs may not be in system search path, add rpath"

This reverts commit d9b983519c63b9072677364a6e399d213a1855e5.

Unfortunately it seems like rpath is evaluated before LD_LIBRARY_PATH,
so this breaks e.g. steam, as well as any other user of that env var,
if the llvm path happens to be where other libs also reside.

Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=76082
Reviewed-by: Tom Stellard <thomas.stellard@amd.com>
Signed-off-by: Ilia Mirkin <imirkin@alum.mit.edu>

Revert "mesa: Fix format matching checks for GL_INTENSITY* internalformats."

This reverts commit 40d7b5195351d3e4199e7a840615a595a6dbaefc.

mesa: move GLbitfield any_valid_stages declaration before code

To fix MSVC build.

glsl: Clean up "unused parameter" warnings

../../src/glsl/ir_constant_expression.cpp:486:1: warning: unused parameter 'variable_context' [-Wunused-parameter]
../../src/glsl/ir_constant_expression.cpp:1633:1: warning: unused parameter 'variable_context' [-Wunused-parameter]
../../src/glsl/ir_constant_expression.cpp:1752:1: warning: unused parameter 'variable_context' [-Wunused-parameter]
../../src/glsl/ir_constant_expression.cpp:1761:1: warning: unused parameter 'variable_context' [-Wunused-parameter]
../../src/glsl/ir_constant_expression.cpp:1769:1: warning: unused parameter 'variable_context' [-Wunused-parameter]

Signed-off-by: Ian Romanick <ian.d.romanick@intel.com>
Reviewed-by: Topi Pohjolainen <topi.pohjolainen@intel.com>

glsl: Minor clean ups in constant_referenced

These could probably be squashed into one of the previous commits.

Signed-off-by: Ian Romanick <ian.d.romanick@intel.com>
Reviewed-by: Topi Pohjolainen <topi.pohjolainen@intel.com>

glsl: Remove ir_dereference::constant_referenced

All of the functionality is implemented in a private function in the one
file where it is used.

Signed-off-by: Ian Romanick <ian.d.romanick@intel.com>
Reviewed-by: Topi Pohjolainen <topi.pohjolainen@intel.com>

glsl: Fold implementation of ir_dereference_array::constant_referenced into wrapper

Signed-off-by: Ian Romanick <ian.d.romanick@intel.com>
Reviewed-by: Topi Pohjolainen <topi.pohjolainen@intel.com>

glsl: Fold implementation of ir_dereference_record::constant_referenced into wrapper

Signed-off-by: Ian Romanick <ian.d.romanick@intel.com>
Reviewed-by: Topi Pohjolainen <topi.pohjolainen@intel.com>

glsl: Fold implementation of ir_dereference_variable::constant_referenced into wrapper

Signed-off-by: Ian Romanick <ian.d.romanick@intel.com>
Reviewed-by: Topi Pohjolainen <topi.pohjolainen@intel.com>

glsl: Add wrapper function that calls ir_dereference::constant_referenced

Signed-off-by: Ian Romanick <ian.d.romanick@intel.com>
Reviewed-by: Topi Pohjolainen <topi.pohjolainen@intel.com>