radeonsi: avoid stale state pointers

Signed-off-by: Christian König <christian.koenig@amd.com>
Reviewed-by: Marek Olšák <marek.olsak@amd.com>

radeonsi: avoid stale pointers in si_delete_shader_selector

Signed-off-by: Christian König <christian.koenig@amd.com>
Reviewed-by: Marek Olšák <marek.olsak@amd.com>

Revert "winsys/radeon: if there's VRAM-only usage, keep it"

This reverts commit 67aef6dafa29fed008ea6065c425a6a92a651be9.

It caused GPU hangs. The question is why.

Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=75900

radeon/vce: fix memory leak

Signed-off-by: Christian König <christian.koenig@amd.com>
Reviewed-by: Marek Olšák <marek.olsak@amd.com>

glcpp: Do not remove spaces to preserve locations.

After preprocessing by glcpp all adjacent spaces were replaced by
single one and glsl parser received column-shifted shader source.
It negatively affected ast location set up and produced wrong error
messages for heavily-spaced shaders.

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

glsl: Change locations from yylloc to appropriate tokens positions.

Reviewed-by: Carl Worth <cworth@cworth.org>

glsl: Add ast_node method to set location range.

Reviewed-by: Carl Worth <cworth@cworth.org>

glsl: Make ast_node location comments more informative.

Reviewed-by: Carl Worth <cworth@cworth.org>

glsl: Extend ast location structure to hande end token position.

Reviewed-by: Carl Worth <cworth@cworth.org>

glsl: Update lexers in glsl and glcpp to hande end position of token.

Reviewed-by: Carl Worth <cworth@cworth.org>

scons: Add drivers/common/meta_generate_mipmap.c to src/mesa/SConscript.

This patch fixes this SCons build error introduced with commit
70e7905608b374f65d1f4b657f8ab61808c76ef6.

build/linux-x86_64-debug/mesa/libmesa.a(driverfuncs.os): In function `_mesa_init_driver_functions':
src/mesa/drivers/common/driverfuncs.c:99: undefined reference to `_mesa_meta_GenerateMipmap'

Signed-off-by: Vinson Lee <vlee@freedesktop.org>

meta: Support GenerateMipmaps on 1DArray textures.

I don't know how many people care about this case, but it's easy enough
to do, so we may as well.  The tricky part is that for some reason Mesa
stores the number of array slices in Height, not Depth.

I thought the easiest way to handle that here was to make Height = 1
(the actual height), and srcDepth = srcImage->Height.  This requires
some munging when calling _mesa_prepare_mipmap_level, so I created a
wrapper that sorts it out for us.

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

meta: Use srcWidth/Height/Depth rather than srcImage->Width and such.

This is equivalent for now, and will differ once we add 1DArray support.

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

meta: Support GenerateMipmaps on 2DArray textures.

This is largely a matter of looping over the number of slices/layers,
and not minifying depth (presumably that code exists for the unfinished
3D texture support).

Normally, I would have made the loop over array slices the outermost
loop.  I suspect that would make it trickier to support 3D textures
someday, though, so I didn't.  The advantage is that we would only have
one BufferData call per slice, rather than one per miplevel and slice.

However, a GenerateMipmaps microbenchmark indicates that either way is
basically just as fast.  So I'm not sure it's worth bothering.

Improves performance in a GenerateMipmaps microbenchmark by nearly 5x.

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

meta: Add a 'layer' argument to bind_fbo_image().

For array textures and 3D textures, this represents the layer to use.
Just pass 0 for now.

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

meta: Refactor code for binding a texture image to the FBO.

Almost the exact same code appeared twice, and it needs to expand to
handle additional texture targets.  Refactor it to tidy up the code and
avoid duplicating more work in the future.

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

meta: Use minify() in GenerateMipmaps code.

This is what the macro is for.

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

meta: Drop redundant FBO creation code in GenerateMipmaps.

fallback_required() already creates the FBO in order to check whether we
can render to the format.  So it's guaranteed to exist.

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

meta: Replace GLboolean with bool in fallback_required().

This doesn't interact with the GL API, so we shouldn't use GL types.

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

meta: Make _mesa_meta_check_generate_mipmap_fallback static.

This was only ever used in one place; there's no reason for it to be
non-static.

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

meta: Split GenerateMipmap() into its own file.

Putting the implementation of each GL function in its own file makes it
much easier not to get lost.

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

meta: De-static setup_texture_coords().

This will be used in multiple files soon.

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

glapi: Add KHR_debug.xml

mesa: add missing DebugMessageControl types

Signed-off-by: Timothy Arceri <t_arceri@yahoo.com.au>
Reviewed-by: Brian Paul <brianp@vmware.com>

mesa: make ARB_debug_output functions an alias of
 KHR_debug

Also update dispatch sanity removing ARB_debug_output checks and
removing KHR_debug placeholders as the checks have already been added

V2: Make sure we exit case statements with conditional breaks rather than
just dropping through.

Signed-off-by: Timothy Arceri <t_arceri@yahoo.com.au>
Reviewed-by: Brian Paul <brianp@vmware.com>

glapi: move KHR_debug into its own file

Signed-off-by: Timothy Arceri <t_arceri@yahoo.com.au>
Reviewed-by: Brian Paul <brianp@vmware.com>

glx_pbuffer: Refactor GetDrawableAttribute

Move the pdraw != NULL check out so that they don't
have to be duplicated.

Signed-off-by: Adel Gadllah <adel.gadllah@gmail.com>
Reviewed-by: Ian Romanick <ian.d.romanick@intel.com>

glx: Update glxext.h to revision 25407

Signed-off-by: Adel Gadllah <adel.gadllah@gmail.com>
Reviewed-by: Ian Romanick <ian.d.romanick@intel.com>

radeon: Include radeon_elf_util.c in the list of LLVM_C_FILES v2

This fixes the a build breakage caused by
6974eb907600b9d0176d3158ff0fd30ac3e56a55 on build configurations where
all the following are true:

1. radeonsi is not being built
2. r600g is being built
3. opencl is disabled
4. --enable-r600-llvm-compiler is not being used
5. libelf is not installed

v2:
  - Add $(RADEON_CFLAGS) to libllvmradeon_la_CFLAGS

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

st/mesa: only mark framebuffer as sRGB capable if Mesa supports the format

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

radeon/llvm: Factor elf parsing code out into its own function

Reviewed-by: Michel Dänzer <michel.daenzer@amd.com>

radeon: Rename struct radeon_llvm_binary to radeon_shader_binary v2

And move its definition into r600_pipe_common.h;  This struct is a just
a container for shader code and has nothing to do with LLVM.

v2:
  - Drop unrelated Makefile change

Reviewed-by: Michel Dänzer <michel.daenzer@amd.com>

gallium: rename R4A4 and A4R4 formats to match their swizzle

Like L4A4.

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

mesa: fix the format of glEdgeFlagPointer

Softpipe expects a float in the vertex shader, which is what glEdgeFlag
generates.

This fixes piglit/gl-2.0-edgeflag.

Cc: mesa-stable@lists.freedesktop.org
Reviewed-by: Brian Paul <brianp@vmware.com>
Reviewed-by: Ian Romanick <ian.d.romanick@intel.com>

radeonsi: fix blit compressed texture workaround to support 2D arrays

We don't have a piglit test for this, but I think it's correct.

Reviewed-by: Michel Dänzer <michel.daenzer@amd.com>

r600g: fix blitting the last 2 mipmap levels for Evergreen

This fixes a lot of compressedteximage piglit tests.

R600-R700 don't have this issue.

Cc: mesa-stable@lists.freedesktop.org

r600g: fix texelFetchOffset GLSL functions

Cc: mesa-stable@lists.freedesktop.org

winsys/radeon: if there's VRAM-only usage, keep it

radeon: Use upload manager for buffer downloads

Using DMA for reads is much faster.

Signed-off-by: Niels Ole Salscheider <niels_ole@salscheider-online.de>
Signed-off by: Marek Olšák <marek.olsak@amd.com>

glapi: use 'Mesa' in error messages

A user would have no idea what "_glthread_" is.  This removes the
last remaining instance of the _glthread_ string in Mesa.

Reviewed-by: Chia-I Wu <olv@lunarg.com>

st/mesa: add test_format_conversion() debug function

To check that the st_mesa_format_to_pipe_format() and
st_pipe_format_to_mesa_format() functions correctly convert
all corresponding Mesa/Gallium formats.

This found that MESA_FORMAT_YCBCR_REV was missing in
st_mesa_format_to_pipe_format().  Fixed that too.

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

st/mesa: add MESA_FORMAT_R8G8B8A8_SRGB in st_mesa_format_to_pipe_format()

v2: rename patch after rebasing on top of Jose's changes.

Reviewed-by: Chia-I Wu <olvaffe@gmail.com>

mesa/st: Fix PIPE_FORMAT_R8G8B8A8_SRGB -> MESA_FORMAT_ conversion.

Copy'n'past typo introduced in my
1d8e3067fd9ce72b3711fc208409ac4a7cff4e53 commit.

This fixes swapped RB channels I was seeing in my test machines.

Trivial.

st/vdpau: Add rotation v2

v2: add static asserts

Signed-off-by: Kusanagi Kouichi <slash@ac.auone-net.jp>
Reviewed-by: Christian König <christian.koenig@amd.com>

vl: Add rotation v3

v2: rotate in gen_rect_verts instead
v3: clear rotate in vl_compositor_clear_layers,
    update calc_drawn_area as well

Signed-off-by: Kusanagi Kouichi <slash@ac.auone-net.jp>
Signed-off-by: Christian König <christian.koenig@amd.com>

st/omx/enc: fix crash on destruction

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

mesa: Drop unused hash_table::mem_ctx field.

It's never used, and it's equivalent to ralloc_parent(ht) if you really
need it.

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

clover: Fix build against LLVM SVN r203065 or newer

llvm/Linker.h was moved to llvm/Linker/Linker.h.

Reviewed-by: Tom Stellard <thomas.stellard@amd.com>

mesa: add MESA_FORMAT_R8G8B8A8_SRGB

To match PIPE_FORMAT_R8G8B8A8_SRGB.
v2: fix component name copy&paste bugs

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

mesa: Wrap SSE4.1 code in #ifdef __SSE4_1__.

Because people insist on doing things like explicitly disabling SSE 4.1.

Cc: "10.0 10.1" <mesa-stable@lists.freedesktop.org>
Tested-by: David Heidelberger <david.heidelberger@ixit.cz>
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=71547

i965: Fix render-to-texture in non-FinishRenderTexture cases.

We've had several problems now with FinishRenderTexture not getting called
enough, and we're ready to just give up on it ever doing what we need.  In
particular, an upcoming Steam title had rendering bugs that could be fixed
by always_flush_cache=true.

Instead of hoping Mesa core can figure out when we need to flush our
caches, just track what BOs we've rendered to in a set, and when we render
from a BO in that set, emit a flush and clear the set.

There's some overhead to keeping this set, but most of that is just
hashing the pointer -- it turns out our set never even gets very large,
because cache flushes are so common (even on cairo-gl).

No statistically significant performance difference in cairo-gl (n=100),
despite spending ~.5% CPU in these set operations.

v1: (Original patch by Eric Anholt.)
v2: (Changes by Ken Graunke.)
  - Rebase forward from May 7th 2013 -> March 4th 2014.
  - Drop the FinishRenderTexture hook entirely; after rebasing the
    patch, the hook was just an empty function.
  - Move the brw_render_cache_set_clear() call from
    intel_batchbuffer_emit_flush() to brw_emit_pipe_control_flush().
    In theory, this could catch more cases where we've flushed.
  - Consider stencil as a possible texturing source.
v3: (changes by anholt):
  - Move set_clear() back to emit_mi_flush() -- it means we can drop
    more forced flushes from the code.  In the previous location, it
    wouldn't have been called when we wanted pre-gen6.
  - Move the set clear from batch init to reset -- it should be empty at
    the start of every batch, since the kernel handled any inter-batch
    flush for us.
v4: Drop the debug code in set.c that I accidentally committed.

Signed-off-by: Eric Anholt <eric@anholt.net>
Signed-off-by: Kenneth Graunke <kenneth@whitecape.org>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Tested-by: Dylan Baker <baker.dylan.c@gmail.com> [v2]

mesa: fix copy & paste bugs in pack_ubyte_SRGB8()

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

mesa: fix copy & paste bugs in pack_ubyte_SARGB8()

Cc: "10.0" "10.1" <mesa-stable@lists.freedesktop.org>
Reviewed-by: Jose Fonseca <jfonseca@vmware.com>

gallium/util: Fix memory leak

Fix a leaked vertex shader in u_blitter.c

Signed-off-by: Aaron Watry <awatry@gmail.com>
Reviewed-by: Brian Paul <brianp@vmware.com>
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
CC: "10.1" <mesa-stable@lists.freedesktop.org>

st/mesa: Add R8G8B8A8_SRGB case to st_pipe_format_to_mesa_format.

With the recent SRGB changes all my automated OpenGL llvmpipe tests
(piglit, conform, glretrace) start asserting with the backtrace below.

I'm hoping this change will fix it.  I'm not entirely sure, as this
doesn't happen in my development machine (the bug probably depends on
the exact X visual).

Anyway, it seems the sensible thing to do here.

   Program terminated with signal 5, Trace/breakpoint trap.
   #0  _debug_assert_fail (expr=expr@entry=0x7fa324df2ed7 "0", file=file@entry=0x7fa324e3fc30 "src/mesa/state_tracker/st_format.c", line=line@entry=758, function=function@entry=0x7fa324e40160 <__func__.34798> "st_pipe_format_to_mesa_format") at src/gallium/auxiliary/util/u_debug.c:281
   #0  _debug_assert_fail (expr=expr@entry=0x7fa324df2ed7 "0", file=file@entry=0x7fa324e3fc30 "src/mesa/state_tracker/st_format.c", line=line@entry=758, function=function@entry=0x7fa324e40160 <__func__.34798> "st_pipe_format_to_mesa_format") at src/gallium/auxiliary/util/u_debug.c:281
   No locals.
   #1  0x00007fa3241d22b3 in st_pipe_format_to_mesa_format (format=format@entry=PIPE_FORMAT_R8G8B8A8_SRGB) at src/mesa/state_tracker/st_format.c:758
           __func__ = "st_pipe_format_to_mesa_format"
   #2  0x00007fa3241c8ec5 in st_new_renderbuffer_fb (format=format@entry=PIPE_FORMAT_R8G8B8A8_SRGB, samples=0, sw=<optimised out>) at src/mesa/state_tracker/st_cb_fbo.c:295
           strb = 0x19e8420
   #3  0x00007fa32409d355 in st_framebuffer_add_renderbuffer (stfb=stfb@entry=0x19e7fa0, idx=<optimised out>) at src/mesa/state_tracker/st_manager.c:314
           rb = <optimised out>
           format = PIPE_FORMAT_R8G8B8A8_SRGB
           sw = <optimised out>
   #4  0x00007fa32409e635 in st_framebuffer_create (st=0x19e7fa0, st=0x19e7fa0, stfbi=0x19e7a30) at src/mesa/state_tracker/st_manager.c:458
           stfb = 0x19e7fa0
           mode = {rgbMode = 1 '\001', floatMode = 0 '\000', colorIndexMode = 0 '\000', doubleBufferMode = 0, stereoMode = 0, haveAccumBuffer = 0 '\000', haveDepthBuffer = 1 '\001', haveStencilBuffer = 1 '\001', redBits = 8, greenBits = 8, blueBits = 8, alphaBits = 8, redMask = 0, greenMask = 0, blueMask = 0, alphaMask = 0, rgbBits = 32, indexBits = 0, accumRedBits = 0, accumGreenBits = 0, accumBlueBits = 0, accumAlphaBits = 0, depthBits = 24, stencilBits = 8, numAuxBuffers = 0, level = 0, visualRating = 0, transparentPixel = 0, transparentRed = 0, transparentGreen = 0, transparentBlue = 0, transparentAlpha = 0, transparentIndex = 0, sampleBuffers = 0, samples = 0, maxPbufferWidth = 0, maxPbufferHeight = 0, maxPbufferPixels = 0, optimalPbufferWidth = 0, optimalPbufferHeight = 0, swapMethod = 0, bindToTextureRgb = 0, bindToTextureRgba = 0, bindToMipmapTexture = 0, bindToTextureTargets = 0, yInverted = 0, sRGBCapable = 1}
           idx = <optimised out>
   #5  st_framebuffer_reuse_or_create (st=st@entry=0x19dfce0, fb=<optimised out>, stfbi=stfbi@entry=0x19e7a30) at src/mesa/state_tracker/st_manager.c:728
   No locals.
   #6  0x00007fa32409e8cc in st_api_make_current (stapi=<optimised out>, stctxi=0x19dfce0, stdrawi=0x19e7a30, streadi=0x19e7a30) at src/mesa/state_tracker/st_manager.c:747
           st = 0x19dfce0
           stdraw = 0x640064
           stread = 0x1300000006
           ret = <optimised out>
   #7  0x00007fa324074a20 in XMesaMakeCurrent2 (c=c@entry=0x195bb00, drawBuffer=0x19e7e90, readBuffer=0x19e7e90) at src/gallium/state_trackers/glx/xlib/xm_api.c:1194
   No locals.
   #8  0x00007fa3240783c8 in glXMakeContextCurrent (dpy=0x194e900, draw=8388610, read=8388610, ctx=0x195bac0) at src/gallium/state_trackers/glx/xlib/glx_api.c:1177
           drawBuffer = <optimised out>
           readBuffer = <optimised out>
           xmctx = 0x195bb00
           glxCtx = 0x195bac0
           firsttime = 0 '\000'
           no_rast = 0 '\000'
   #9  0x00007fa32407852f in glXMakeCurrent (dpy=<optimised out>, drawable=<optimised out>, ctx=<optimised out>) at src/gallium/state_trackers/glx/xlib/glx_api.c:1211
   No locals.

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

glapi: remove u_mutex wrapper code, use c99 thread mutexes directly

v2: fix initializer mistake spotted by Chia-I Wu.

Reviewed-by: Chia-I Wu <olv@lunarg.com>

glapi: rename u_current dispatch table functions

Put "table" in the names to make things more understandable.

Reviewed-by: Chia-I Wu <olv@lunarg.com>

glapi: replace 'user' with 'context' in u_current.[ch] code

To make the functions more understandable.

Reviewed-by: Chia-I Wu <olv@lunarg.com>

glsl: fix compiler warnings in link_uniforms.cpp

With a non-debug build, gcc has two complaints:
1. 'found' var not used.  Silence with '(void) found;'
2. 'id' not initialized.  It's assigned by the UniformHash->get()
   call, actually.  But init it to zero to silence gcc.

Reviewed-by: Matt Turner <mattst88@gmail.com>

mesa/st: only compare the one scissor

sizeof(scissor) returns the size of the full array rather than a single
element. Fix it to consider just the one element.

Fixes: 0705fa35 ("st/mesa: add support for GL_ARB_viewport_array (v0.2)")
Signed-off-by: Ilia Mirkin <imirkin@alum.mit.edu>
Reviewed-by: Dave Airlie <airlied@redhat.com>

st/mesa: make winsys fbo sRGB-capable when supported

The texture formats of winsys fbo are always linear becase the st manager
(st/dri for example) could not know the colorspace used.  But it does not mean
that we cannot make the fbo sRGB-capable.  By

 - setting rb->Visual.sRGBCapable to GL_TRUE when the pipe driver supports the
   format in sRGB colorspace,
 - giving rb an sRGB internal format, and
 - updating code to check rb->Format instead of strb->texture->format,

we should be good.

Fixed bug 75226 for at least llvmpipe and ilo, with no piglit regression.

v2: do not set rb->Visual.sRGBCapable for GLES contexts to avoid surprises

Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=75226

Reviewed-by: Brian Paul <brianp@vmware.com>
Tested-by: Michel Dänzer <michel.daenzer@amd.com>

st/mesa: add mappings for MESA_FORMAT_B8G8R8X8_SRGB

The format is mapped to PIPE_FORMAT_B8G8R8X8_SRGB.

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

mesa: add MESA_FORMAT_B8G8R8X8_SRGB

The format is needed to represent an RGB-only winsys framebuffer that is
sRGB-capable.

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

mesa: fix packing/unpacking for MESA_FORMAT_A4R4G4B4_UNORM

Spotted by Chia-I Wu.
v2: also fix unpack_ubyte_ARGB4444_REV()

Reviewed-by: Roland Scheidegger <sroland@vmware.com>
Reviewed-by: Chia-I Wu <olv@lunarg.com>
Reviewed-by: Anuj Phogat <anuj.phogat@gmail.com>

i965: Fix predicated-send-based discards with MRT.

We need the header setup to not be predicated on which pixels are
undiscarded.  I'm not sure originally if I had thought that the mask
disable implied predicate disable, or if I had just misread the mask
disable as predicate disable.  Either way, I know I had spent more time
thinking about this in the gen8 generator than the gen7 generator.

Plus, it turns out that I had mis-implemented the "the GPU will use the
predicate unless this header is present" comment, by skipping setting up
the pixel mask when the header was present.

Fixes GPU hangs in piglit glsl-fs-discard-mrt, Trine, Trine 2 and
preusmably MLL.

Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=75207
Tested-by: Tapani Pälli <tapani.palli@intel.com>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>

configure: Fix bashism.

/bin/sh defaults to dash on debian.

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

docs: update 10.2 release notes

mesa: remove remaining uses of _glthread_GetID()

It was really only used in the radeon driver for a debug printf.
And evidently, libGL.so referenced it just to work around some sort
of linker issue.

This patch removes the two calls to the function and the function
itself.

Fixes undefined _glthread_GetID symbol in libGL reported by 'nm'.
Though, the missing symbol doesn't cause any issues on my system but
it does cause glxinfo to fail on one of our test systems.

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

mesa: new init_teximage_fields_ms() function to init MS texture images

Before, it was kind of ugly to set the multisample fields with
assignments after we called _mesa_init_teximage_fields().

Reviewed-by: Anuj Phogat <anuj.phogat@gmail.com>

WIP: freedreno/a3xx: incorrect scissor for binning pass

If scissor optimization is used (to avoid bringing scissored portions of
the render target into GMEM and then back out to system memory) in
combination with hw binning pass, the result would be a scissor mismatch
between binning pass and rendering pass.  This would cause rendering
bugs in some scenarios with (for example) gnome-shell.

I would have expected that simply using the correct screen-scissor
during the binning pass would be enough, but seems like there is
something else missing.  So for now disable binning pass if scissor
optimization is used.

i965: Mark invariants in backend_visitor as constants

Reviewed-by: Matt Turner <mattst88@gmail.com>
Signed-off-by: Topi Pohjolainen <topi.pohjolainen@intel.com>

i965: Merge resolving of shader program source

Reviewed-by: Matt Turner <mattst88@gmail.com>
Signed-off-by: Topi Pohjolainen <topi.pohjolainen@intel.com>

i965: Merge initialisation of backend_visitor

Reviewed-by: Matt Turner <mattst88@gmail.com>
Signed-off-by: Topi Pohjolainen <topi.pohjolainen@intel.com>

i965/wm: Use resolved miptree consistently in surface setup

Most of the logic refers to the local variable 'mt' directly but
a few cases use 'intelObj->mt' instead. These are the same for
now but will be different once stencil miptree gets used.

v2 (Ian): fixed also indentation in surrounding lines

Reviewed-by: Matt Turner <mattst88@gmail.com>
Reviewed-by: Ian Romanick <ian.d.romanick@intel.com>
Signed-off-by: Topi Pohjolainen <topi.pohjolainen@intel.com>

i965/vec4: Mark invariant members as constants in vec4_visitor

Reviewed-by: Matt Turner <mattst88@gmail.com>
Reviewed-by: Ian Romanick <ian.d.romanick@intel.com>
Signed-off-by: Topi Pohjolainen <topi.pohjolainen@intel.com>

i965: Mark sources for offset getters as constants

Reviewed-by: Matt Turner <mattst88@gmail.com>
Reviewed-by: Ian Romanick <ian.d.romanick@intel.com>
Signed-off-by: Topi Pohjolainen <topi.pohjolainen@intel.com>

docs: Import 10.1 release notes, add news item.

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

nv50,nvc0: add 11f_11f_10f vertex support

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

i965: Implement ARB_stencil_texturing on Gen8+.

On earlier hardware, we had to implement math in the shader to translate
Y-tiled or untiled coordinates to W-tiled coordinates (which is what
BLORP does today in order to texture from stencil buffers).

On Broadwell, we can simply state that it's W-tiled in SURFACE_STATE,
and adjust the pitch.  This is much easier.

In the surface state code, I chose to handle the "should we sample depth
or stencil?" question separately from the setup for sampling from
stencil.  This should make it work with the BindRenderbufferTexImage
hook as well, and hopefully be reusable for GL_ARB_texture_stencil8
someday.

v2: Update docs/GL3.txt (caught by Matt).

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

mesa: Add core API support for GL_ARB_stencil_texturing (from 4.3).

While the GL_ARB_stencil_texturing extension does not allow the creation
of stencil textures, it does allow shaders to sample stencil values
stored in packed depth/stencil textures.

Specifically, applications can call glTexParameter* with a pname of
GL_DEPTH_STENCIL_TEXTURE_MODE and value of either GL_DEPTH_COMPONENT or
GL_STENCIL_INDEX to select which component they wish to sample.  The
default value is GL_DEPTH_COMPONENT (for traditional depth sampling).

Shaders should use an unsigned integer sampler (presumably usampler2D)
to access stencil data.  Otherwise, results are undefined.  Using shadow
samplers with GL_STENCIL_INDEX selected also is undefined behavior.

This patch creates a new gl_texture_object field, StencilSampling, to
indicate that stencil should be sampled rather than depth.  (I chose to
use a boolean since I figured it would be more convenient for drivers.)
It also introduces the [Get]TexParameter code to get and set the value,
and of course the extension plumbing.

v2: Also consider textures incomplete when sampling stencil with
    non-NEAREST min/mag filters (caught by Eric Anholt).

Signed-off-by: Kenneth Graunke <kenneth@whitecape.org>
Reviewed-by: Matt Turner <mattst88@gmail.com>
Reviewed-by: Eric Anholt <eric@anholt.net>

radeon/uvd: fix typo in documentation

s/grap/grab/

Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

dri: Require libudev-dev for building DRI on Linux.

The loader infrastructure for everything but DRI2 requires that udev be
present, so we can figure out an appropriate driver from the fd.  We don't
have a portable solution yet, but presumably it will have similar lookup
based on the device node.

It will also be even more required for krh's udev-based hwdb support,
which lets us have a loader that actually loads DRI drivers not included
in the loader's source distribution.

Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=75212
Reviewed-by: Matt Turner <mattst88@gmail.com>

clover: Use correct LLVM version in #if for DataLayout construction

Spotted by Michel Dänzer.

translate: fix buffer overflows

Because in draw we always inject position at slot 0 whenever
fragment shader would take the maximum number of inputs (32) it
meant that we had PIPE_MAX_ATTRIBS + 1 slots to translate, which
meant that we were crashing with fragment shaders that took
the maximum number of attributes as inputs. The actual max number
of attributes we need to translate thus is PIPE_MAX_ATTRIBS + 1.

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

draw/llvm: fix generation of the VS with GS present

draw_current_shader_* functions return a final output when considering
both the geometry shader and the vertex shader. But when code generating
vertex shader we can not be using output slots from the geometry shader
because, obviously, those can be completely different. This fixes a
number of very non-obvious crashes.
A side-effect of this bug was that sometimes the vertex shading code
could save some random outputs as position/clip when the geometry
shader was writing them and vertex shader had different outputs at
those slots (sometimes writing garbage and sometimes something correct).

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

mesa: Allow GL_DEPTH_COMPONENT and GL_DEPTH_STENCIL combinations in glTexImage{123}D()

From OpenGL 3.3 spec, page 141:
   "Textures with a base internal format of DEPTH_COMPONENT or DEPTH_STENCIL
    require either depth component data or depth/stencil component data.
    Textures with other base internal formats require RGBA component data.
    The error INVALID_OPERATION is generated if one of the base internal
    format and format is DEPTH_COMPONENT or DEPTH_STENCIL, and the other
    is neither of these values."

Fixes Khronos OpenGL CTS test failure: proxy_textures_invalid_size

Cc: <mesa-stable@lists.freedesktop.org>
Signed-off-by: Anuj Phogat <anuj.phogat@gmail.com>
Reviewed-by: Brian Paul <brianp@vmware.com>

mesa: Use clear_teximage_fields() in place of _mesa_init_teximage_fields()

This patch makes no functional changes to the code.

Signed-off-by: Anuj Phogat <anuj.phogat@gmail.com>
Reviewed-by: Brian Paul <brianp@vmware.com>

mesa: Set initial internal format of a texture to GL_RGBA

From OpenGL 4.0 spec, page 398:
   "The initial internal format of a texel array is RGBA
    instead of 1. TEXTURE_COMPONENTS is deprecated; always
    use TEXTURE_INTERNAL_FORMAT."

Fixes Khronos OpenGL CTS test failure: proxy_textures_invalid_size

Cc: <mesa-stable@lists.freedesktop.org>
Signed-off-by: Anuj Phogat <anuj.phogat@gmail.com>
Reviewed-by: Brian Paul <brianp@vmware.com>

scons: Build with C++11 with LLVM >= 3.5.

Starting with llvm-3.5svn r202574, LLVM expects C+11 mode.

commit f8bc17fadc8f170c1126328d203f0dab78960137
Author: Chandler Carruth <chandlerc@gmail.com>
Date:   Sat Mar 1 06:31:00 2014 +0000

    [C++11] Turn off compiler-based detection of R-value references, relying
    on the fact that we now build in C++11 mode with modern compilers. This
    should flush out any issues. If the build bots are happy with this, I'll
    GC all the code for coping without R-value references.

    git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@202574 91177308-0d34-0410-b5e6-96231b3b80d8

Signed-off-by: Vinson Lee <vlee@freedesktop.org>

st/osmesa: check buffer size when searching for buffers

Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=75543
Cc: "10.1" <mesa-stable@lists.freedesktop.org>

configure: s/--with-llvm-shared-libs/--enable-llvm-shared-libs/

`--enable-llvm-shared-libs` option was recently renamed as
`--with-llvm-shared-libs`, but several error messages still mention the
old option, causing confusing.

Trivial.

c11/threads: Don't implement thrd_current on Windows.

GetCurrentThread() returns a pseudo-handle (a constant which only makes
sense when used within the calling thread) and not a real handle.

DuplicateHandle() will return a real handle, but it will create a new
handle every time we call.  Calling DuplicateHandle() here means we will
leak handles, which can cause serious problems.

In short, the Windows implementation of thrd_t needs a thorough make
over, and it won't be pretty.  It looks like C11 committee
over-simplified things: it would be much better to have seperate objects
for threads and thread IDs like C++11 does.

For now, just comment out the thrd_current() implementation, so we get
build errors if anybody tries to use it.

Thanks to Brian Paul for spotting and diagnosing this problem.

Cc: "10.0" "10.1" <mesa-stable@lists.freedesktop.org>
Reviewed-by: Brian Paul <brianp@vmware.com>

mapi/u_thread: Use GetCurrentThreadId

u_thread_self() expects thrd_current() to return a unique numeric ID
for the current thread, but this is not feasible on Windows.

Cc: "10.0" "10.1" <mesa-stable@lists.freedesktop.org>
Reviewed-by: Brian Paul <brianp@vmware.com>

c11/threads: Fix nano to milisecond conversion.

Per https://gist.github.com/yohhoy/2223710/#comment-710118

Cc: "10.0" "10.1" <mesa-stable@lists.freedesktop.org>
Reviewed-by: Brian Paul <brianp@vmware.com>
Reviewed-by: Michel Dänzer <michel@daenzer.net>

r600g: implement edge flags

Cc: mesa-stable@lists.freedesktop.org
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>

r600g: port color buffer format conversion from radeonsi

r600_translate_colorformat is rewritten to look like radeonsi.
r600_translate_colorswap is shared with radeonsi.
r600_colorformat_endian_swap is consolidated.

This adds some formats which were missing. Future "plain" formats will
automatically be supported.

Cc: mesa-stable@lists.freedesktop.org
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>

radeonsi: move translate_colorswap to common code

Also translate the Y__X swizzle.

Cc: mesa-stable@lists.freedesktop.org
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>

Revert "configure: use enable_dri_glx local variable"

This reverts commit dfe8cb48fc13587afb2d829c29432c0eacd0aba3.

Accidently pushed this commit, over 1bb23abe065(configure: disable
shared glapi when building xlib powered glx).

configure: disable shared glapi when building xlib powered glx

With commit 0432aa064b(configure: use shared-glapi when more than one
gl* API is used) we removed "disable shared-glapi when building without
dri" hunk.

In the good old days of classic mesa, dri and xlib-glx were mutually
exclusive thus the hunk made sense.

Currently enable-dri is used as a synonym for a range of things thus
it's more appropriate to handle xlib-glx explicitly.

Fixes a missing symbol '_glapi_Dispatch' in a xlib powered libGL,
build using the following

 ./autogen.sh --enable-xlib-glx --disable-dri --with-gallium-drivers=swrast

Cc: Brian Paul <brianp@vmware.com>
Reported-by: Brian Paul <brianp@vmware.com>
Signed-off-by: Emil Velikov <emil.l.velikov@gmail.com>

mesa: remove unneeded glthread.c file

The _glthread_GetID() function is also defined in mapi_glapi.c

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