llvmpipe: Simplify and fix system variables fetch.

The system array values concept doesn't really because it expects the
system values to be fixed per call, which is wrong for gl_VertexID and
iffy for gl_SampleID.  So this patch does two things:

- kill the array, have emit_fetch_system_value directly pick the
  values it needs (only gl_InstanceID for now, as the previous code)

- correctly handle the expected type in emit_fetch_system_value

Signed-off-by: Olivier Galibert <galibert@pobox.com>
Reviewed-by: Brian Paul <brianp@vmware.com>

draw: fix flat shading and screen-space linear interpolation in clipper

This includes:
- picking up correctly which attributes are flatshaded and which are
  noperspective

- copying the flatshaded attributes when needed, including the
  non-built-in ones

- correctly interpolating the noperspective attributes in screen-space
  instead than in a 3d-correct fashion.

Signed-off-by: Olivier Galibert <galibert@pobox.com>
Reviewed-by: Brian Paul <brianp@vmware.com>

softpipe: Offset is not to be applied to the layer parameter of array texture fetches.

Signed-off-by: Olivier Galibert <galibert@pobox.com>
Reviewed-by: Brian Paul <brianp@vmware.com>

st/mesa: clamp glDrawPixels size to max texture size

st/mesa: move st_validate_state() call earlier in st_DrawPixels()

r600g: fix z/stencil texture creation v2

z or stencil texture should not be created with the z/stencil
flags for surface creation as they are intended to be bound
as texture.

v2: remove broken code

Signed-off-by: Jerome Glisse <jglisse@redhat.com>

radeon/llvm: Fix CR/LF in Processors.td

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

radeon/llvm: Fix sin/cos codegen on R700

Based on https://bugs.freedesktop.org/show_bug.cgi?id=50317#c4

Fixes:
https://bugs.freedesktop.org/show_bug.cgi?id=50316
https://bugs.freedesktop.org/show_bug.cgi?id=50317

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

docs: update GL3.txt for ARB_base_instance

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

st/mesa: Add support for GL_ARB_base_instance

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

gallium: Add PIPE_CAP_START_INSTANCE

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

mesa: Add support for GL_ARB_base_instance

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

scons: Do not build svga if using Solaris Studio C compiler.

Solaris Studio C compiler does not support anonymous structs and
anonymous unions.

Signed-off-by: Vinson Lee <vlee@freedesktop.org>
Reviewed-by: Brian Paul <brianp@vmware.com>

i965: Fix brw_swap_cmod() for LE/GE comparisons.

The idea here is to rewrite comparisons like 2 >= x with x <= 2; we want
to simply exchange arguments, not negate the condition.  If equality was
part of the original comparison, it should remain part of the swapped
version.

This is the true cause of bug #50298.  It didn't manifest itself on
Sandybridge because we embed the conditional modifier in the IF
instruction rather than emitting a CMP.  All other platforms use CMP.

It also didn't manifest itself on the master branch because commit
be5f27a84d ("glsl: Refine the loop instruction counting.") papered over
the problem.

NOTE: This is a candidate for stable release branches.

Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=50298
Signed-off-by: Kenneth Graunke <kenneth@whitecape.org>
Reviewed-by: Eric Anholt <eric@anholt.net>

docs: start release notes file for 8.1

radeon/llvm: Update comment in AMDGPU.td

radeon/llvm: Remove unused AMDIL TableGen definitons

radeon/llvm: Eliminate getRegClassFromType() function

We can use TargetLowering::getRegClassFor() instead.

radeon/llvm: Remove deadcode from AMDILISelLowering.cpp

gallium: Add support for Solaris Studio C++ compiler.

Signed-off-by: Vinson Lee <vlee@freedesktop.org>
Reviewed-by: Brian Paul <brianp@vmware.com>

llvmpipe: Implement cylindrical wrapping.

Tested against mesa demos cylwrap and dx9 DCT address.exe which now passes 100%.

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

st/glx: Do not undefine _R, _G, and _B.

Fixes build error on Cygwin and Solaris. _R, _G, and _B are used in
ctype.h on those platforms.

Signed-off-by: Vinson Lee <vlee@freedesktop.org>
Reviewed-by: Brian Paul <brianp@vmware.com>

svga: fix synchronization bug between sampler views and surfaces

This fixes a bug where a sampler view was using stale texture/resource
data when the texture was modified through a surface (render to texture).
Bumping the texture and layer ages triggers sampler view revalidation.

Fixes piglit fbo-blit failure.

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

gles2: Add GL_NV_read_buffer extension

This lets us select the front buffer for reading under GLES2.

Reviewed-by: Brian Paul <brianp@vmware.com>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>

get.c: Rename EXTRA_VERSION_ES2 to EXTRA_API_ES2

This extra condition checks the API not the version of the API, so rename
to reflect that.

Reviewed-by: Brian Paul <brianp@vmware.com>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>

docs/relnotes: comment out bug template

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

docs/relnotes: replace tbd with release date

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

docs/relnotes: fix validation errors

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

docs/relnotes: consolidate html header

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

draw: Ensure that the vertex_header type size matches expectation.

This is failing sometimes, probably because TargetData keeps a structure layout
cache, which can becomes bogus, ever since the InvalidateStructLayoutInfo API
was removed in LLVM r135245.

This change merely makes the problem easier to diagnose (an assertion
failure instead of a random crash).

r600g: enable streamout by default on r7xx and DRM 2.17.0

Now that it's in Linus's tree.

Has anyone had a chance to test streamout on Cayman recently?

st/mesa: properly allocate MSAA renderbuffers

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

st/mesa: make unsupported renderbuffer formats always fail as FBO incomplete

instead of failing to allocate a renderbuffer.

This also fixes piglit/get-renderbuffer-internalformat with non-renderable
formats.

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

st/mesa: separate sw renderbuffer allocation from hw one

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

mesa: if AllocStorage doesn't choose a format, report FRAMEBUFFER_UNSUPPORTED

This allows drivers not to do any allocation in AllocStorage if the storage
cannot be allocated because of an unsupported internalformat + samples combo.

The little ugliness is that AllocStorage is expected to return TRUE in this
case.

Reviewed-by: Brian Paul <brianp@vmware.com>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>

i915g: More ops commute.

This allows using the optimizations more broadly.

r600g: fix lockups with streamout on r7xx

This requires the latest streamout kernel patches.

Streamout is disabled by default on r7xx, so this patch is safe for regular
users.

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

r600g: compute CS space for streamout correctly, add comments

SET_CONTEXT_REG was not counted in.

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

r600g: set SMX_ACTION_ENA to fix streamout cache flushes on some chipsets

It helps on R7xx.

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

clover: Fix build with LLVM libs installed to non-standard directories

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

st/mesa: don't do srgb->linear conversion in decompress_with_blit

This fixes piglit/getteximage-formats on r600g.

NOTE: This is a candidate for stable branches.

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

glsl: Use ir_unop_f2u to convert floats to uints.

Fixes piglit tests
spec/glsl-1.30/execution/{vs,fs}-float-uint-conversion on i965.

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

gallium: Add TGSI_OPCODE_F2U to gallivm backend.

Note: for the moment TGSI_OPCODE_F2U is implemented using
lp_build_itrunc() (the same function used to implement
TGSI_OPCODE_F2I).  In the long run, we should create an
lp_build_utrunc() function to do the proper conversion.  But this
should allow us to limp along with mostly correct behaviour for now.

gallium: Add support for ir_unop_f2u to tgsi backend.

Reviewed-by: José Fonseca <jfonseca@vmware.com>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>

ir_to_mesa: Add support for ir_unop_f2u to ir_to_mesa backend.

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

i965: Add support for ir_unop_f2u to i965 backend.

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

glsl: Add support for ir_unop_f2u to constant folding.

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

glsl: Add unary operation ir_unop_f2u.

Previously, we performed conversions from float->uint by a two step
process: float->int->uint.  However, on platforms that use saturating
conversions (e.g. i965), this didn't work, because if the source value
was larger than the maximum representable int (0x7fffffff), then
converting it to an int would clamp it to 0x7fffffff.

This patch just adds the new opcode; further patches will adapt
optimization passes and back-ends to use it, and then finally the
ast_to_hir logic will be modified to emit the new opcode.

Reviewed-by: Brian Paul <brianp@vmware.com>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>

i965/blorp: Implement source clipping.

This patch modifies blorp blits (which are used for MSAA) to properly
account for clipping of source coordinates.  Previously, if we
detected the possibility of source clipping, we would fall back to the
blit meta-op, which doesn't support MSAA and is very slow for depth
and stencil buffers.

Fixes piglit tests
"EXT_framebuffer_multisample/clip-and-scissor-blit" on i965/Gen6+.

Also substantially speeds up the Humble Bundle V game "Psychonauts" on
Gen6+ (without this patch, the game's depth buffer blits use the slow
blit meta-op).

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

scons: add st_atom_array.c to the build

winsys/radeon: enable IB submission to compute rings v2

This allows to submit things to the compute only
rings on cayman+

v2: rebased on current master and actually make use
    of the new flag in evergreen_compute.c

Signed-off-by: Christian König <deathsimple@vodafone.de>
Reviewed-by: Marek Olšák <maraeo@gmail.com>

st/mesa: atomize vertex array state

This moves the state validation to where all the other states are validated.

winsys/radeon: Remove unnecessary pipe_thread_destroy in radeon_drm_cs_destroy

Fixes crash bug introduced with 210ddf0819b5 fd.o #49198
pthread_detach after a pthread_join is unneeded.

Signed-off-by: Maarten Lankhorst <m.b.lankhorst@gmail.com>
Signed-off-by: Marek Olšák <maraeo@gmail.com>

nv50,nvc0: fix stream output target buffer leak

It manifests at exit as:
"WARNING: destroying GPU memory cache with some buffers still in use"

nv50: disable stream output before reconfiguring it

If we don't, the GPU will just throw an ILLEGAL_OPERATION error.

nv50/ir: handle NEG,ABS modifiers for short RCP encoding

st/mesa: fix glDrawPixels(GL_DEPTH_COMPONENT) color output

When drawing a depth image the fragment shader also needs to emit the
current raster color.

The new piglit drawpix-z test exercises this.

NOTE: This is a candiate for the 8.0 branch.

docs: add info about shortlog_mesa.sh script

glx/tests and mesa/tests: Update .gitignore files.

This patch updates .gitignore files to account for the new build
artifacts introduced by the following commits:

ae376f0 glx/tests: Rename test as glx-test
8fecdcc mesa/tests: Add tests for _mesa_lookup_enum_by_{name,nr} functions
a29ad2b mesa/tests: Add tests for the generated dispatch table

st/vdpau: fix YCbCr down/up-loads for buffers larger than requested

When the video buffer turns out to be larger than
requested by the application we shouldn't upload
or download more data into / from it original requested.

Fixes: https://bugs.freedesktop.org/show_bug.cgi?id=39309
Signed-off-by: Christian König <deathsimple@vodafone.de>

scons: Fix Haiku binary optimizations

Haiku targets the Pentium or higher processor.
To ensure compatibility we can do march 586 and
mtune 686. Mesa will still use sse however if
the cpu supports it (and the stack is properly
aligned). These flags only effect the internal
compiler optimizations.

mesa: fix html in shortlog_mesa.sh script

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

mesa: added Ian's shortlog_mesa.sh script in bin/

svga: make svga_surface_needs_propagation() surface const

svga: add svga_surface_const() cast wrapper

svga: fix comment typo

rbug: fix make process on Linux Mint 13 x64.

Previously, rbug_*.c would fail to compile with incomplete prototype
errors when make was run from the command line on my machine. My IDE
always built fine, and still does after this patch (Netbeans 7.1.2).

Most of the includes from files in gallium/auxiliary/rbug/* were
assuming an rbug/ subdirectory, while the headers are actually in the
same directory as the .c files.

The build error was also previously a problem for me on Ubuntu 11.10
and Mint 12.

Fixes build for the following configuration: ./autogen.sh
--enable-debug --enable-texture-float --with-gallium-drivers=r600
--with-dri-drivers=radeon --enable-r600-llvm-compiler

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

windows/gdi: Remove GL_NV_register_combiners and GL_NV_vertex_array_range exports

glsl: Fix pi/2 constant in acos built-in function

In single precision, 1.5707963 becomes 1.5707962513 which is too
small.  However, 1.5707964 becomes 1.5707963705 which is just right.
The value 1.5707964 is already used in asin.ir.

NOTE: This is a candidate for stable release branches.

Signed-off-by: Ian Romanick <ian.d.romanick@intel.com>
Reviewed-by: Olivier Galibert <galibert@pobox.com>
Reviewed-by: Paul Berry <stereotype441@gmail.com>

glapi: Remove GL_NV_vertex_array_range from the dispatch table

There is no GLX protocol for these functions.  Open-source Linux
driver have not supported this extension for many years, and it seems
unlikely at this point that this support will return.  There's no
reason to have slots for these functions in the dispatch table.

The unit tests (GetProcAddress::TableDidntShrink and others) are also updated.

Signed-off-by: Ian Romanick <ian.d.romanick@intel.com>
Reviewed-by: Brian Paul <brianp@vmware.com>

glapi: Remove GL_NV_fence from the dispatch table

There is no GLX protocol for these functions.  No open-source Linux
driver has ever supported this extension, and it seems unlikely at
this point that one ever will.  There's no reason to have slots for
these functions in the dispatch table.

The unit tests (GetProcAddress::TableDidntShrink and others) are also updated.

Signed-off-by: Ian Romanick <ian.d.romanick@intel.com>
Reviewed-by: Brian Paul <brianp@vmware.com>

glapi: Remove GL_NV_register_combiners from the dispatch table

There is no GLX protocol for these functions.  No open-source Linux
driver has ever supported this extension, and it seems unlikely at
this point that one ever will.  There's no reason to have slots for
these functions in the dispatch table.

The unit tests (GetProcAddress::TableDidntShrink and others) are also updated.

Signed-off-by: Ian Romanick <ian.d.romanick@intel.com>
Reviewed-by: Brian Paul <brianp@vmware.com>

glapi: Remove GL_APPLE_texture_range from the dispatch table

There is no GLX protocol for these functions, and no Linux driver has
ever supported this extension.  There's no reason to have slots for
these functions in the dispatch table.

The unit tests (GetProcAddress::TableDidntShrink and others) are also updated.

Signed-off-by: Ian Romanick <ian.d.romanick@intel.com>
Reviewed-by: Brian Paul <brianp@vmware.com>

glapi: Remove GL_SGIX_pixel_texture from the dispatch table

There is no GLX protocol for this function.  Open-source Linux driver
have not supported this extension for many years, and it seems
unlikely at this point that this support will return.  There's no
reason to have slots for this function in the dispatch table.

The unit tests (GetProcAddress::TableDidntShrink and others) are also updated.

Signed-off-by: Ian Romanick <ian.d.romanick@intel.com>
Reviewed-by: Brian Paul <brianp@vmware.com>

glapi: Remove GL_SGIS_pixel_texture from the dispatch table

There is no GLX protocol for these functions, and no Linux driver has
ever supported this extension.  There's no reason to have slots for
these functions in the dispatch table.

The unit tests (GetProcAddress::TableDidntShrink and others) are also updated.

Signed-off-by: Ian Romanick <ian.d.romanick@intel.com>
Reviewed-by: Brian Paul <brianp@vmware.com>

mesa/tests: Add tests for the generated dispatch table

Signed-off-by: Ian Romanick <ian.d.romanick@intel.com>
Reviewed-by: Brian Paul <brianp@vmware.com>

mesa/tests: Add tests for _mesa_lookup_enum_by_{name,nr} functions

Signed-off-by: Ian Romanick <ian.d.romanick@intel.com>
Reviewed-by: Brian Paul <brianp@vmware.com>

glapi: Add missing GL_EXT_texture_sRGB_decode enums

Signed-off-by: Ian Romanick <ian.d.romanick@intel.com>
Reviewed-by: Brian Paul <brianp@vmware.com>

glapi: Add missing GL_EXT_framebuffer_sRGB enums

Signed-off-by: Ian Romanick <ian.d.romanick@intel.com>
Reviewed-by: Brian Paul <brianp@vmware.com>

glapi: Add missing GL_EXT_packed_float enums

Signed-off-by: Ian Romanick <ian.d.romanick@intel.com>
Reviewed-by: Brian Paul <brianp@vmware.com>

glapi: Add missing framebuffer sRGB enum

Signed-off-by: Ian Romanick <ian.d.romanick@intel.com>
Reviewed-by: Brian Paul <brianp@vmware.com>

glapi: Add uniform buffer object enums

These are from OpenGL 3.1 and ARB_uniform_buffer_object.  I only added
them to 3.1 because that required the least work.

Signed-off-by: Ian Romanick <ian.d.romanick@intel.com>
Reviewed-by: Brian Paul <brianp@vmware.com>

glapi: Add missing enums for GL_NV_fragment_program

Signed-off-by: Ian Romanick <ian.d.romanick@intel.com>
Reviewed-by: Brian Paul <brianp@vmware.com>

glapi: Add missing enums for GL_ARB_occlusion_query2

Signed-off-by: Ian Romanick <ian.d.romanick@intel.com>
Reviewed-by: Brian Paul <brianp@vmware.com>

glapi: Remove extraneous GL_ from TEXTURE_IMMUTABLE_FORMAT

Signed-off-by: Ian Romanick <ian.d.romanick@intel.com>
Reviewed-by: Brian Paul <brianp@vmware.com>

glapi: Add missing enums for GL_ATI_fragment_shader

Signed-off-by: Ian Romanick <ian.d.romanick@intel.com>
Reviewed-by: Brian Paul <brianp@vmware.com>

glapi: Add texture swizzle enums

These are from OpenGL 3.3, ARB_texture_swizzle, and
EXT_texture_swizzle (with different names).  I only added them to 3.3
because that required the least work.

Signed-off-by: Ian Romanick <ian.d.romanick@intel.com>
Reviewed-by: Brian Paul <brianp@vmware.com>

glapi: Add a couple missing 3.0 enums

Signed-off-by: Ian Romanick <ian.d.romanick@intel.com>
Reviewed-by: Brian Paul <brianp@vmware.com>

glapi: Add missing _NV extension on COMBINE4

Signed-off-by: Ian Romanick <ian.d.romanick@intel.com>
Reviewed-by: Brian Paul <brianp@vmware.com>

glapi: Add missing enums for GL_EXT_vertex_array

Signed-off-by: Ian Romanick <ian.d.romanick@intel.com>
Reviewed-by: Brian Paul <brianp@vmware.com>

glapi: Add missing enums for GL_EXT_compiled_vertex_array

Signed-off-by: Ian Romanick <ian.d.romanick@intel.com>
Reviewed-by: Brian Paul <brianp@vmware.com>

glx/tests: Add unit tests for generated code in indirect_init.c

Signed-off-by: Ian Romanick <ian.d.romanick@intel.com>
Reviewed-by: Brian Paul <brianp@vmware.com>

glx/tests: Add unit tests for generated code in indirect_size.c

Signed-off-by: Ian Romanick <ian.d.romanick@intel.com>
Reviewed-by: Brian Paul <brianp@vmware.com>

glx/tests: Rename test as glx-test

This matches the existing test in src/glsl/tests.

Signed-off-by: Ian Romanick <ian.d.romanick@intel.com>
Reviewed-by: Brian Paul <brianp@vmware.com>

glx: Move tests from tests/glx to src/glx/tests

This matches the organization of other unit tests in Mesa.

Signed-off-by: Ian Romanick <ian.d.romanick@intel.com>
Reviewed-by: Brian Paul <brianp@vmware.com>

util: add some comments, fix indentation

glsl: Transform dot product by a basis vector into a swizzle

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

glsl: Add is_basis function

Determines whether it's a basis vector, i.e., a vector with one element
equal to 1 and all other elements equal to 0.

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

glsl: Check for zero vectors in ir_binop_dot

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

mesa: move variable declaration out of loop to fix MSVC build