mesa.git
11 years agoglx: Centralize the code for context flushing.
Eric Anholt [Fri, 28 Dec 2012 19:28:38 +0000 (11:28 -0800)]
glx: Centralize the code for context flushing.

Reviewed-by: Marek Olšák <maraeo@gmail.com>
11 years agoglx: Add a little comment about what dri2FlushFrontBuffer() does.
Eric Anholt [Fri, 28 Dec 2012 20:13:07 +0000 (12:13 -0800)]
glx: Add a little comment about what dri2FlushFrontBuffer() does.

Reviewed-by: Marek Olšák <maraeo@gmail.com>
11 years agoradeonsi: Handle scaled and integer formats for samplers and vertex elements.
Michel Dänzer [Tue, 5 Feb 2013 16:24:51 +0000 (17:24 +0100)]
radeonsi: Handle scaled and integer formats for samplers and vertex elements.

Also, add assertions to stress that render targets don't support scaled
formats.

20 more little piglits.

11 years agoradeonsi: Don't advertise PIPE_FORMAT_L8A8_SRGB support.
Michel Dänzer [Tue, 5 Feb 2013 16:30:56 +0000 (17:30 +0100)]
radeonsi: Don't advertise PIPE_FORMAT_L8A8_SRGB support.

The hardware can't do it.

11 years agoradeonsi: Remove incorrect (and dead) assignment in tex_fetch_args().
Michel Dänzer [Wed, 6 Feb 2013 16:14:26 +0000 (17:14 +0100)]
radeonsi: Remove incorrect (and dead) assignment in tex_fetch_args().

The proper return type is assigned at the end of the function.

11 years agoradeonsi: Use unique names for referring to texture sampling intrinsics.
Michel Dänzer [Wed, 6 Feb 2013 14:43:10 +0000 (15:43 +0100)]
radeonsi: Use unique names for referring to texture sampling intrinsics.

Append the overloaded vector type used for passing in the addressing
parameters.

Without this, LLVM uses the same function signature for all those types,
which cannot work.

Fixes problems e.g. with FlightGear and Red Eclipse.

11 years agor300g: put textures with usage=staging in GTT and make them linear
Marek Olšák [Thu, 7 Feb 2013 03:29:42 +0000 (04:29 +0100)]
r300g: put textures with usage=staging in GTT and make them linear

11 years agor600g: fix slice tile max for compressed texture and async dma
Jerome Glisse [Wed, 6 Feb 2013 18:54:02 +0000 (13:54 -0500)]
r600g: fix slice tile max for compressed texture and async dma

Was using the pixel size instead of the number of block for the slice
tile max computation which resulted in dma writing at wrong address.

Signed-off-by: Jerome Glisse <jglisse@redhat.com>
11 years agoradeonsi: use new RGBX formats
Marek Olšák [Fri, 25 Jan 2013 23:28:17 +0000 (00:28 +0100)]
radeonsi: use new RGBX formats

11 years agor300g: fix blending and alpha-test with RGBX16F and enable MSAA for it
Marek Olšák [Fri, 1 Feb 2013 20:45:15 +0000 (21:45 +0100)]
r300g: fix blending and alpha-test with RGBX16F and enable MSAA for it

11 years agor300g: use new RGBX formats
Marek Olšák [Fri, 25 Jan 2013 23:27:27 +0000 (00:27 +0100)]
r300g: use new RGBX formats

11 years agor600g: use new RGBX formats
Marek Olšák [Fri, 25 Jan 2013 23:28:00 +0000 (00:28 +0100)]
r600g: use new RGBX formats

11 years agost/mesa: use new RGBX formats
Marek Olšák [Fri, 25 Jan 2013 23:26:50 +0000 (00:26 +0100)]
st/mesa: use new RGBX formats

Reviewed-by: Brian Paul <brianp@vmware.com>
11 years agomesa: add RGBX formats for existing GL RGB texture formats
Marek Olšák [Mon, 28 Jan 2013 15:33:10 +0000 (16:33 +0100)]
mesa: add RGBX formats for existing GL RGB texture formats

v2: fix compilation of swrast

11 years agogallium: add RGBX formats for existing GL RGB texture formats
Marek Olšák [Sat, 19 Jan 2013 14:48:00 +0000 (15:48 +0100)]
gallium: add RGBX formats for existing GL RGB texture formats

Reviewed-by: Brian Paul <brianp@vmware.com>
11 years agoi965/blorp: Support blits between ARGB and XRGB formats.
Kenneth Graunke [Tue, 29 Jan 2013 06:26:31 +0000 (22:26 -0800)]
i965/blorp: Support blits between ARGB and XRGB formats.

Now that we have support for overriding alpha to 1.0, we can handle
blitting between these formats in either direction.

For now, we only support two XRGB formats: MESA_FORMAT_XRGB8888 and
MESA_FORMAT_RGBX8888_REV.  Most places only appear to worry about the
former, so ignore the latter for now.  We can always add it later.

NOTE: This is a candidate for the 9.1 branch.

Signed-off-by: Kenneth Graunke <kenneth@whitecape.org>
Reviewed-by: Ian Romanick <ian.d.romanick@intel.com>
Tested-by: Martin Steigerwald <martin@lichtvoll.de>
11 years agoi965/blorp: Support overriding destination alpha to 1.0.
Kenneth Graunke [Tue, 29 Jan 2013 06:03:18 +0000 (22:03 -0800)]
i965/blorp: Support overriding destination alpha to 1.0.

Currently, Blorp requires the source and destination formats to be
equal.  However, we'd really like to be able to blit between XRGB and
ARGB formats; our BLT engine paths have supported this for a long time.

For ARGB -> XRGB, nothing needs to occur: the missing alpha is already
interpreted as 1.0.  For XRGB -> ARGB, we need to smash the alpha
channel to 1.0 when writing the destination colors.  This is fairly
straightforward with blending.

For now, this code is never used, as the source and destination formats
still must be equal.  The next patch will relax that restriction.

NOTE: This is a candidate for the 9.1 branch.

Signed-off-by: Kenneth Graunke <kenneth@whitecape.org>
Reviewed-by: Ian Romanick <ian.d.romanick@intel.com>
Tested-by: Martin Steigerwald <martin@lichtvoll.de>
11 years agoi965: Implement CopyTexSubImage2D via BLORP (and use it by default).
Kenneth Graunke [Wed, 16 Jan 2013 06:17:23 +0000 (22:17 -0800)]
i965: Implement CopyTexSubImage2D via BLORP (and use it by default).

The BLT engine has many limitations.  Currently, it can only blit
X-tiled buffers (since we don't have a kernel API to whack the BLT
tiling mode register), which means all depth/stencil operations get
punted to meta code, which can be very CPU-intensive.

Even if we used the BLT engine, it can't blit between buffers with
different tiling modes, such as an X-tiled non-MSAA ARGB8888 texture
and a Y-tiled CMS ARGB8888 renderbuffer.  This is a fundamental
limitation, and the only way around that is to use BLORP.

Previously, BLORP only handled BlitFramebuffer.  This patch adds an
additional frontend for doing CopyTexSubImage.  It also makes it the
default.  This is partly to increase testing and avoid hiding bugs,
and partly because the BLORP path can already handle more cases.  With
trivial extensions, it should be able to handle everything the BLT can.

This helps PlaneShift massively, which tries to CopyTexSubImage2D
between depth buffers whenever a player casts a spell.  Since these
are Y-tiled, we hit meta and software ReadPixels paths, eating 99% CPU
while delivering ~1 FPS.  This is particularly bad in an MMO setting
because people cast spells all the time.

It also helps Xonotic in 4X MSAA mode.  At default power management
settings, I measured a 6.35138% +/- 0.672548% performance boost (n=5).
(This data is from v1 of the patch.)

No Piglit regressions on Ivybridge (v3) or Sandybridge (v2).

v2: Create a fake intel_renderbuffer to wrap the destination texture
    image and then reuse do_blorp_blit rather than reimplementing most
    of it.  Remove unnecessary clipping code and conditional rendering
    check.

v3: Reuse formats_match() to centralize checks; delete temporary
    renderbuffers.  Reorganize the code.

v4: Actually copy stencil when dealing with separate stencil buffers but
    packed depth/stencil formats.  Tested by a new Piglit test.

NOTE: This is a candidate for the 9.1 branch.

Signed-off-by: Kenneth Graunke <kenneth@whitecape.org>
Reviewed-by: Paul Berry <stereotype441@gmail.com> [v4]
Reviewed-by: Ian Romanick <ian.d.romanick@intel.com> [v3]
Reviewed-and-tested-by: Carl Worth <cworth@cworth.org> [v2]
Tested-by: Martin Steigerwald <martin@lichtvoll.de> [v3]
11 years agomesa: Put extern "C" guards in renderbuffer.h.
Kenneth Graunke [Tue, 29 Jan 2013 06:56:35 +0000 (22:56 -0800)]
mesa: Put extern "C" guards in renderbuffer.h.

I need to use this from C++ code.

NOTE: This is a candidate for the 9.1 branch.

Signed-off-by: Kenneth Graunke <kenneth@whitecape.org>
Reviewed-by: Ian Romanick <ian.d.romanick@intel.com>
11 years agollvmpipe: remove extraneous const qualifier
Brian Paul [Wed, 6 Feb 2013 16:16:58 +0000 (09:16 -0700)]
llvmpipe: remove extraneous const qualifier

11 years agogallium/util: remove duplicated function util_format_is_rgb_no_alpha
Marek Olšák [Fri, 25 Jan 2013 20:21:22 +0000 (21:21 +0100)]
gallium/util: remove duplicated function util_format_is_rgb_no_alpha

It only checks if alpha is present, so it's the same as util_format_has_alpha.

Reviewed-by: Brian Paul <brianp@vmware.com>
11 years agost/mesa: get rid of GET_CURRENT_CONTEXT in st_choose_format
Marek Olšák [Thu, 24 Jan 2013 20:14:30 +0000 (21:14 +0100)]
st/mesa: get rid of GET_CURRENT_CONTEXT in st_choose_format

Reviewed-by: Brian Paul <brianp@vmware.com>
11 years agost/mesa: adjust texture format selection to try the closest base format first
Marek Olšák [Sat, 19 Jan 2013 13:11:08 +0000 (14:11 +0100)]
st/mesa: adjust texture format selection to try the closest base format first

Reviewed-by: Brian Paul <brianp@vmware.com>
11 years agost/mesa: put RGBX8 and RGBA8 in the default format lists
Marek Olšák [Thu, 24 Jan 2013 19:54:02 +0000 (20:54 +0100)]
st/mesa: put RGBX8 and RGBA8 in the default format lists

Reviewed-by: Brian Paul <brianp@vmware.com>
11 years agost/mesa: add the rest of RGB8 format/type combos to exact_format_mapping tables
Marek Olšák [Thu, 24 Jan 2013 19:33:29 +0000 (20:33 +0100)]
st/mesa: add the rest of RGB8 format/type combos to exact_format_mapping tables

These formats were added a few months after these tables were committed.
No idea why we have the table though. AFAIK, texstore always takes the slow path
for GL_RGBn.

Reviewed-by: Brian Paul <brianp@vmware.com>
11 years agomesa: fixup inconsistent naming of RG16 formats
Marek Olšák [Thu, 24 Jan 2013 20:06:50 +0000 (21:06 +0100)]
mesa: fixup inconsistent naming of RG16 formats

Reviewed-by: Brian Paul <brianp@vmware.com>
11 years agor600g: report correct control flow depth
Marek Olšák [Thu, 31 Jan 2013 18:39:41 +0000 (19:39 +0100)]
r600g: report correct control flow depth

11 years agoglsl: fix incorrect comment about do_common_optimization
Marek Olšák [Thu, 31 Jan 2013 22:29:53 +0000 (23:29 +0100)]
glsl: fix incorrect comment about do_common_optimization

11 years agost/mesa: emit saturates in the vertex shader if Shader Model 3.0 is supported
Marek Olšák [Thu, 31 Jan 2013 21:30:44 +0000 (22:30 +0100)]
st/mesa: emit saturates in the vertex shader if Shader Model 3.0 is supported

v2: change the requirement from GLSL 1.30 to SM 3.0 (R500 can do this)

11 years agost/mesa: advertise ARB_shading_language_packing for GLSL >= 1.30
Marek Olšák [Thu, 31 Jan 2013 22:52:14 +0000 (23:52 +0100)]
st/mesa: advertise ARB_shading_language_packing for GLSL >= 1.30

Reviewed-by: Brian Paul <brianp@vmware.com>
11 years agost/mesa: do most of GLSL lowering outside of the optimization do-while loop
Marek Olšák [Thu, 31 Jan 2013 18:44:57 +0000 (19:44 +0100)]
st/mesa: do most of GLSL lowering outside of the optimization do-while loop

based on the intel driver

Reviewed-by: Brian Paul <brianp@vmware.com>
11 years agost/mesa: remove dead code depending on EmitCondCodes
Marek Olšák [Thu, 31 Jan 2013 18:43:50 +0000 (19:43 +0100)]
st/mesa: remove dead code depending on EmitCondCodes

EmitCondCodes is always false.

Reviewed-by: Brian Paul <brianp@vmware.com>
11 years agor300g: try to use color varyings for texcoords if max texcoord limit is exceeded
Marek Olšák [Sat, 2 Feb 2013 05:20:25 +0000 (06:20 +0100)]
r300g: try to use color varyings for texcoords if max texcoord limit is exceeded

+35 piglits

11 years agor300/compiler: copy-propagate saturate mode when possible
Marek Olšák [Sun, 3 Feb 2013 15:54:59 +0000 (16:54 +0100)]
r300/compiler: copy-propagate saturate mode when possible

Reviewed-by: Tom Stellard <thomas.stellard@amd.com>
11 years agor300/compiler: add support for saturate output modifier in r500 vertex shaders
Marek Olšák [Sat, 2 Feb 2013 03:59:27 +0000 (04:59 +0100)]
r300/compiler: add support for saturate output modifier in r500 vertex shaders

The GLSL compiler can simplify clamp(v,0,1) to saturate. The state tracker
doesn't use it yet, but it will.

Reviewed-by: Tom Stellard <thomas.stellard@amd.com>
11 years agor300g: fix blending with RGBX formats
Marek Olšák [Fri, 1 Feb 2013 20:09:31 +0000 (21:09 +0100)]
r300g: fix blending with RGBX formats

Change DST_ALPHA to ONE.

11 years agor300g: fix blending with blend color and RGBA formats
Marek Olšák [Fri, 1 Feb 2013 19:59:38 +0000 (20:59 +0100)]
r300g: fix blending with blend color and RGBA formats

NOTE: This is a candidate for the stable branches.

11 years agoegl/dri: Don't invoke dri2_dpy->flush if it's NULL.
José Fonseca [Tue, 5 Feb 2013 13:43:01 +0000 (13:43 +0000)]
egl/dri: Don't invoke dri2_dpy->flush if it's NULL.

I'd like to test Mesa OpenGL ES along side with NVIDIA libGL drivers. But
without this change, I get a NULL pointer dereference.

Reviewed-by: Brian Paul <brianp@vmware.com>
11 years agoglsl: Initialize ast_parameter_declarator member variables.
Vinson Lee [Sat, 2 Feb 2013 07:27:34 +0000 (23:27 -0800)]
glsl: Initialize ast_parameter_declarator member variables.

Fixes uninitialized pointer field defect reported by Coverity.

Signed-off-by: Vinson Lee <vlee@freedesktop.org>
Reviewed-by: Brian Paul <brianp@vmware.com>
11 years agosvga: fix sRGB rendering
Brian Paul [Tue, 5 Feb 2013 19:20:39 +0000 (12:20 -0700)]
svga: fix sRGB rendering

We weren't emitting the SVGA_RS_OUTPUTGAMMA state so sRGB rendering
didn't work properly.

Fixes piglit's framebuffer-srgb test.

Note: This is a candidate for the stable branches.

Reviewed-by: Jose Fonseca <jfonseca@vmware.com>
11 years agor600g/compute: Fix segfault caused by new shader disassembler
Tom Stellard [Mon, 4 Feb 2013 19:31:54 +0000 (19:31 +0000)]
r600g/compute: Fix segfault caused by new shader disassembler

11 years agoRequire libdrm_radeon 2.4.42 for radeonsi.
Michel Dänzer [Tue, 5 Feb 2013 14:12:14 +0000 (15:12 +0100)]
Require libdrm_radeon 2.4.42 for radeonsi.

It has new PCI IDs and an important tiled surface layout fix.

11 years agoi965: Disable write masking when setting up texturing m0.
Eric Anholt [Thu, 30 Aug 2012 18:07:52 +0000 (11:07 -0700)]
i965: Disable write masking when setting up texturing m0.

v2/Kayden: Also disable write masking in the vec4 backend.

Fixes 78 oglconform glsl-bif-tex-* subcases.

Signed-off-by: Kenneth Graunke <kenneth@whitecape.org>
Reviewed-by: Paul Berry <stereotype441@gmail.com> [v1]
Reviewed-by: Eric Anholt <eric@anholt.net> [v2]
11 years agointel: Fix regression in intel_create_image_from_name stride handling
Tapani Pälli [Mon, 28 Jan 2013 06:53:56 +0000 (08:53 +0200)]
intel: Fix regression in intel_create_image_from_name stride handling

Strangely, the DRIimage interface we have passes the pitch in pixels
instead of bytes, which anholt missed in the change to using bytes for
region pitch.

Signed-off-by: Tapani Pälli <tapani.palli@intel.com>
Reviewed-by: Eric Anholt <eric@anholt.net>
11 years agoi965: Fix segfaults from 45a28a927ab7f29ff325b9326d386a39ba538c18
Eric Anholt [Mon, 4 Feb 2013 17:32:06 +0000 (09:32 -0800)]
i965: Fix segfaults from 45a28a927ab7f29ff325b9326d386a39ba538c18

If you look up a level that isn't in the miptree, you crash.

Reviewed-by: Chad Versace <chad.versace@linux.intel.com>
11 years agoradeonsi: add Oland pci ids
Alex Deucher [Fri, 25 Jan 2013 00:11:28 +0000 (19:11 -0500)]
radeonsi: add Oland pci ids

Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Note: this is a candidate for the 9.1 branch.

11 years agoradeonsi: default PA_SC_RASTER_CONFIG to 0
Alex Deucher [Fri, 25 Jan 2013 00:09:48 +0000 (19:09 -0500)]
radeonsi: default PA_SC_RASTER_CONFIG to 0

That should work in all cases.

Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Note: this is a candidate for the 9.1 branch.

11 years agoradeonsi: add support for Oland chips
Alex Deucher [Fri, 25 Jan 2013 00:08:07 +0000 (19:08 -0500)]
radeonsi: add support for Oland chips

Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Note: this is a candidate for the 9.1 branch

11 years agoglsl: Support transform feedback of varying structs.
Paul Berry [Tue, 29 Jan 2013 15:34:52 +0000 (07:34 -0800)]
glsl: Support transform feedback of varying structs.

Since transform feedback needs to be able to access individual fields
of varying structs, we can no longer match up the arguments to
glTransformFeedbackVaryings() with variables in the vertex shader.

Instead, we build up a hashtable which records information about each
possible name that is a candidate for transform feedback, and then
match up the arguments to glTransformFeedbackVaryings() with the
contents of that hashtable.

Populating the hashtable uses the program_resource_visitor
infrastructure, so the logic is shared with how we handle uniforms.

NOTE: This is a candidate for the 9.1 branch.

Reviewed-by: Jordan Justen <jordan.l.justen@intel.com>
Reviewed-by: Matt Turner <mattst88@gmail.com>
11 years agoglsl: Use parse_program_resource_name to parse transform feedback varyings.
Paul Berry [Tue, 29 Jan 2013 13:52:47 +0000 (05:52 -0800)]
glsl: Use parse_program_resource_name to parse transform feedback varyings.

Previously, transform feedback varyings were parsed in an ad-hoc
fashion that wasn't compatible with structs (or array of structs).
This patch makes it use parse_program_resource_name(), which correctly
handles both.

Note that parse_program_resource_name()'s technique for handling
mal-formed input strings is to simply let them through and rely on the
fact that a future name lookup will fail.  Because of this,
tfeedback_decl::init() no longer needs to return a boolean error
code--it always succeeds, and if the input was mal-formed the error
will be detected later.

NOTE: This is a candidate for the 9.1 branch.

Reviewed-by: Jordan Justen <jordan.l.justen@intel.com>
Reviewed-by: Matt Turner <mattst88@gmail.com>
11 years agoglsl: Rename uniform_field_visitor to program_resource_visitor.
Paul Berry [Mon, 28 Jan 2013 22:43:03 +0000 (14:43 -0800)]
glsl: Rename uniform_field_visitor to program_resource_visitor.

There's actually nothing uniform-specific in uniform_field_visitor.
It is potentially useful for all kinds of program resources (in
particular, future patches will use it for transform feedback
varyings).

This patch renames it to program_resource_visitor, and clarifies
several comments, to reflect the fact that it is useful for more than
just uniforms.

NOTE: This is a candidate for the 9.1 branch.

Reviewed-by: Jordan Justen <jordan.l.justen@intel.com>
Reviewed-by: Matt Turner <mattst88@gmail.com>
11 years agomesa/glsl: Separate parsing logic from _mesa_get_uniform_location.
Paul Berry [Mon, 28 Jan 2013 22:21:59 +0000 (14:21 -0800)]
mesa/glsl: Separate parsing logic from _mesa_get_uniform_location.

The parsing logic is moved to a new function in the GLSL module,
parse_program_resource_name().  This name was chosen because it should
eventually be useful for handling everything that OpenGL 4.3 calls
"program resources" (e.g. uniforms, vertex inputs, fragment outputs,
and transform feedback varyings).

Future patches will make use of this function for linking transform
feedback varyings.

NOTE: This is a candidate for the 9.1 branch.

Reviewed-by: Jordan Justen <jordan.l.justen@intel.com>
Reviewed-by: Matt Turner <mattst88@gmail.com>
11 years agogallium/egl: Fix include dirs for VPATH build
Quentin Glidic [Sat, 2 Feb 2013 17:08:34 +0000 (18:08 +0100)]
gallium/egl: Fix include dirs for VPATH build

NOTE: This is a candidate for the 9.1 branch.
Reviewed-by: Matt Turner <mattst88@gmail.com>
Signed-off-by: Quentin Glidic <sardemff7+git@sardemff7.net>
11 years agointel: make sure to setup image dimension in image_from_planar setup
Abdiel Janulgue [Mon, 4 Feb 2013 11:30:06 +0000 (13:30 +0200)]
intel: make sure to setup image dimension in image_from_planar setup

Fixes: https://bugs.freedesktop.org/show_bug.cgi?id=60212
Tested-by: Scott Moreau <oreaus@gmail.com>
Tested-by: Tiago Vignatti <tiago.vignatti@intel.com>
Reviewed-by: Chad Versace <chad.versace@linux.intel.com>
Signed-off-by: Abdiel Janulgue <abdiel.janulgue@linux.intel.com>
11 years agobuiltin_compiler/build: Don't use *_FOR_BUILD when not cross compiling
Matt Turner [Wed, 30 Jan 2013 19:50:55 +0000 (11:50 -0800)]
builtin_compiler/build: Don't use *_FOR_BUILD when not cross compiling

Previously we were relying on CFLAGS_FOR_BUILD to be the same as CFLAGS
when not cross compiling, but this assumption didn't take into
consideration 32-bit builds on 64-bit systems. More generally, not
honoring CFLAGS is bad.

Automake is evidently too stupid to accept

if CROSS_COMPILING
CC = @CC_FOR_BUILD@
...
else
CC = @CC@
endif

without warning that CC has been already defined. The warnings are
harmless, but I'd prefer to avoid future reports about them, so define
proxy variables, which are assigned inside the conditional and then
unconditionally assigned to CC et al.

NOTE: This is a candidate for the 9.1 branch.
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=59737
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=60038

11 years agost/mesa: emit SQRT opcode when driver supports it
Brian Paul [Fri, 1 Feb 2013 18:17:50 +0000 (11:17 -0700)]
st/mesa: emit SQRT opcode when driver supports it

11 years agogallium/drivers: handle PIPE_SHADER_CAP_TGSI_SQRT_SUPPORTED query
Brian Paul [Fri, 1 Feb 2013 18:16:54 +0000 (11:16 -0700)]
gallium/drivers: handle PIPE_SHADER_CAP_TGSI_SQRT_SUPPORTED query

Initially, only softpipe/llvmpipe support SQRT.

11 years agogallivm: implement support for SQRT opcode
Brian Paul [Fri, 1 Feb 2013 18:16:22 +0000 (11:16 -0700)]
gallivm: implement support for SQRT opcode

11 years agotgsi: add support for new SQRT opcode
Brian Paul [Fri, 1 Feb 2013 18:15:43 +0000 (11:15 -0700)]
tgsi: add support for new SQRT opcode

11 years agogallium: add SQRT shader opcode
Brian Paul [Fri, 1 Feb 2013 17:59:43 +0000 (10:59 -0700)]
gallium: add SQRT shader opcode

The glsl-to-tgsi translater will emit SQRT to implement GLSL's sqrt()
and distance() functions if the PIPE_SHADER_CAP_TGSI_SQRT_SUPPORTED
query says it's supported by the driver.

Otherwise, sqrt(x) is implemented with x*rsq(x).  The problem with
this is sqrt(0) must be handled specially because rsq(0) might be
Inf/NaN/undefined (and then 0*rsq(0) is Inf/Nan/undefined).  In the
glsl-to-tgsi code we use an extra CMP to check if x is zero and then
replace the result of x*rsq(x) with zero.

In the end, this makes sqrt() generate much more reasonable code for
drivers that can do square roots.

Note that many of piglit's generated shader tests use the GLSL
distance() function.

11 years agoradeonsi: Remove spurious traces of R16G16B16 support.
Michel Dänzer [Fri, 25 Jan 2013 15:04:37 +0000 (16:04 +0100)]
radeonsi: Remove spurious traces of R16G16B16 support.

The hardware can't do it, and these were causing warnings in some piglit tests.

NOTE: This is a candidate for the 9.1 branch.

11 years agoradeonsi: Enable texture arrays.
Michel Dänzer [Fri, 25 Jan 2013 17:30:30 +0000 (18:30 +0100)]
radeonsi: Enable texture arrays.

28/30 piglit tests pass.

NOTE: This is a candidate for the 9.1 branch.

11 years agoradeonsi: Improve packing of texture address parameters.
Michel Dänzer [Fri, 25 Jan 2013 11:10:11 +0000 (12:10 +0100)]
radeonsi: Improve packing of texture address parameters.

In particular, the LOD bias and depth comparison values are packed before the
'normal' texture coordinates, and the array slice and LOD values are appended.

NOTE: This is a candidate for the 9.1 branch.

11 years agoradeonsi: Adapt to sample intrinsics changes.
Michel Dänzer [Thu, 24 Jan 2013 17:54:51 +0000 (18:54 +0100)]
radeonsi: Adapt to sample intrinsics changes.

Fix up intrinsic names, and bitcast texture address parameters to integers.

NOTE: This is a candidate for the 9.1 branch.

11 years agost/mesa: simplify the update_single_texture() function
Brian Paul [Sat, 2 Feb 2013 01:00:45 +0000 (18:00 -0700)]
st/mesa: simplify the update_single_texture() function

In particular, rework the sRGB/linear format selection code.
There's no reason to mess with the Mesa format.
Just do everything in terms of the gallium pipe_format.

Reviewed-by: Marek Olšák <maraeo@gmail.com>
11 years agost/mesa: merge st_ChooseTextureFormat_renderable() into st_ChooseTextureFormat()
Brian Paul [Fri, 1 Feb 2013 18:44:35 +0000 (11:44 -0700)]
st/mesa: merge st_ChooseTextureFormat_renderable() into st_ChooseTextureFormat()

That was the only place it was being called from.

11 years agost/mesa: improve the format choosing code for DrawPixels
Brian Paul [Fri, 1 Feb 2013 18:38:28 +0000 (11:38 -0700)]
st/mesa: improve the format choosing code for DrawPixels

The code before was getting a pipe format, then calling
st_pipe_format_to_mesa_format() and then converting back again with
st_mesa_format_to_pipe_format().  This removes one conversion step.

11 years agogallium: handle unhandled PIPE_CAP_TEXTURE_BUFFER_OFFSET_ALIGNMENT
Andreas Boll [Thu, 31 Jan 2013 08:35:14 +0000 (09:35 +0100)]
gallium: handle unhandled PIPE_CAP_TEXTURE_BUFFER_OFFSET_ALIGNMENT

Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=60098

Signed-off-by: Brian Paul <brianp@vmware.com>
11 years agost/mesa: don't choose DXT formats if we can't do DXT compression
Brian Paul [Fri, 1 Feb 2013 01:52:57 +0000 (18:52 -0700)]
st/mesa: don't choose DXT formats if we can't do DXT compression

If we call gl[Copy]TexImage2D() with a generic compression format
(e.g. intFormat=GL_COMPRESSED_RGBA) we can't choose a DXT format if
we don't have the external DXT compression library.

We weren't actually enforcing this before since the
pipe_screen::is_format_supported(DXT) query has no dependency on
the DXT compression library.

Now if we're given a generic compressed format and we can't do DXT
compression we'll fall back to a non-compressed format.

v2: use util_format_is_s3tc() function and add more comments about
the allow_dxt parameter.

Note: This is a candidate for the stable branches.

Reviewed-by: Jose Fonseca <jfonseca@vmware.com>
11 years agomesa: don't use format chooser code for glCompressedTexImage
Brian Paul [Fri, 1 Feb 2013 01:49:01 +0000 (18:49 -0700)]
mesa: don't use format chooser code for glCompressedTexImage

When glCompressedTexImage is called the internalFormat is a specific
format for the incoming image and the the hardware format should be
the same (since we never do format transcoding).  So use the simpler
_mesa_glenum_to_compressed_format() function.  This change is also
needed for the next patch.

Note: This is a candidate for the stable branches.

11 years agoi965: Fix the SF Vertex URB Read Length calculation for Gen7 platforms.
Kenneth Graunke [Sat, 2 Feb 2013 20:46:57 +0000 (12:46 -0800)]
i965: Fix the SF Vertex URB Read Length calculation for Gen7 platforms.

Ivybridge doesn't appear to have the same errata as Sandybridge; no
corruption was observed by setting it to more than the minimal correct
value.  It's possible that we were simply lucky, since the URB entries
are 1024-bit on Ivybridge vs. 512-bit Sandybridge.  Or perhaps the
underlying hardware issue is fixed.

Either way, we may as well program the minimum value since it's now
readily available, likely to be more efficient, and possibly more
correct.

v2: Use GEN7_SBE_* defines rather than GEN6_SF_*.  (A copy and paste
    mistake.)  They're the same, but using the right names is better.

NOTE: This is a candidate for all stable branches.
Reviewed-by: Paul Berry <stereotype441@gmail.com>
Signed-off-by: Kenneth Graunke <kenneth@whitecape.org>
11 years agoi965: Fix the SF Vertex URB Read Length calculation for Sandybridge.
Kenneth Graunke [Sat, 2 Feb 2013 20:46:49 +0000 (12:46 -0800)]
i965: Fix the SF Vertex URB Read Length calculation for Sandybridge.

(This commit message was primarily written by Paul Berry, who explained
 what's going on far better than I would have.)

Previous to this patch, we thought that the only restrictions on
3DSTATE_SF's URB read length were (a) it needs to be large enough to
read all the VUE data that the SF needs, and (b) it can't be so large
that it tries to read VUE data that doesn't exist.  Since the VUE map
already tells us how much VUE data exists, we didn't bother worrying
about restriction (a); we just did the easy thing and programmed the
read length to satisfy restriction (b).

However, we didn't notice this erratum in the hardware docs: "[errata]
Corruption/Hang possible if length programmed larger than recommended".
Judging by the context surrounding this erratum, it's pretty clear that
it means "URB read length must be exactly the size necessary to read all
the VUE data that the SF needs, and no larger".  Which means that we
can't program the read length based on restriction (b)--we have to
program it based on restriction (a).

The URB read size needs to precisely match the amount of data that the
SF consumes; it doesn't work to simply base it on the size of the VUE.

Thankfully, the PRM contains the precise formula the hardware expects.

Fixes random UI corruption in Steam's "Big Picture Mode", random terrain
corruption in PlaneShift, and Piglit's fbo-5-varyings test.

NOTE: This is a candidate for all stable branches.
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=56920
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=60172
Tested-by: Jordan Justen <jordan.l.justen@intel.com> (v1/Piglit)
Tested-by: Martin Steigerwald <martin@lichtvoll.de> (PlaneShift)
Reviewed-by: Paul Berry <stereotype441@gmail.com>
Signed-off-by: Kenneth Graunke <kenneth@whitecape.org>
11 years agoi965: Compute the maximum SF source attribute.
Kenneth Graunke [Sat, 2 Feb 2013 20:40:54 +0000 (12:40 -0800)]
i965: Compute the maximum SF source attribute.

The maximum SF source attribute is necessary to compute the Vertex URB
read length properly, which will be done in the next commit.

NOTE: This is a candidate for all stable branches.
Reviewed-by: Paul Berry <stereotype441@gmail.com>
Tested-by: Martin Steigerwald <martin@lichtvoll.de>
Signed-off-by: Kenneth Graunke <kenneth@whitecape.org>
11 years agoi965: Refactor Gen6+ SF attribute override code.
Kenneth Graunke [Sat, 2 Feb 2013 20:33:21 +0000 (12:33 -0800)]
i965: Refactor Gen6+ SF attribute override code.

The next patch will benefit from easy access to the source attribute
number and whether or not we're swizzling.  It doesn't want the final
attr_override DWord form, however.

NOTE: This is a candidate for all stable branches.
Reviewed-by: Paul Berry <stereotype441@gmail.com>
Tested-by: Martin Steigerwald <martin@lichtvoll.de>
Signed-off-by: Kenneth Graunke <kenneth@whitecape.org>
11 years agoglsl: Remove hash table from ir_set_program_inouts pass.
Kenneth Graunke [Thu, 31 Jan 2013 06:11:17 +0000 (22:11 -0800)]
glsl: Remove hash table from ir_set_program_inouts pass.

Back when ir_var_in and ir_var_out signified both function parameters
and shader input/outputs, we had trouble distinguishing the two when
looking at a dereference.  Now that we have separate ir_var_shader_in
and ir_var_shader_out modes, we can determine this easily.

Removing the hash table saves memory and CPU overhead.

Signed-off-by: Kenneth Graunke <kenneth@whitecape.org>
11 years agoi965: Remove dead field brw_wm_prog_data::error.
Kenneth Graunke [Thu, 31 Jan 2013 05:39:00 +0000 (21:39 -0800)]
i965: Remove dead field brw_wm_prog_data::error.

11 years agoi965: Remove dead field brw_context::constant_map.
Kenneth Graunke [Thu, 31 Jan 2013 05:36:38 +0000 (21:36 -0800)]
i965: Remove dead field brw_context::constant_map.

This was used by the old VS backend, but that's long gone.

11 years agor600g: Fix memory leak.
Vinson Lee [Fri, 1 Feb 2013 07:45:13 +0000 (23:45 -0800)]
r600g: Fix memory leak.

Fixes resource leak defect reported by Coverity.

Signed-off-by: Vinson Lee <vlee@freedesktop.org>
Reviewed-by: Marek Olšák <maraeo@gmail.com>
Reviewed-by: Brian Paul <brianp@vmware.com>
11 years agoegl/dri2: Fix memory leak.
Vinson Lee [Fri, 1 Feb 2013 07:41:14 +0000 (23:41 -0800)]
egl/dri2: Fix memory leak.

Fixes resource leak defect reported by Coverity.

Signed-off-by: Vinson Lee <vlee@freedesktop.org>
Reviewed-by: Brian Paul <brianp@vmware.com>
11 years agonv30: Fix memory leak.
Vinson Lee [Fri, 1 Feb 2013 07:37:41 +0000 (23:37 -0800)]
nv30: Fix memory leak.

Fixes resource leak defect reported by Coverity.

Signed-off-by: Vinson Lee <vlee@freedesktop.org>
Reviewed-by: Brian Paul <brianp@vmware.com>
11 years agonv50: Fix memory leak.
Vinson Lee [Fri, 1 Feb 2013 07:35:26 +0000 (23:35 -0800)]
nv50: Fix memory leak.

Fixes resource leak defect reported by Coverity.

Signed-off-by: Vinson Lee <vlee@freedesktop.org>
Reviewed-by: Brian Paul <brianp@vmware.com>
11 years agonvc0: Fix memory leak.
Vinson Lee [Fri, 1 Feb 2013 07:32:41 +0000 (23:32 -0800)]
nvc0: Fix memory leak.

Fixes resource leak defect reported by Coverity.

Signed-off-by: Vinson Lee <vlee@freedesktop.org>
Reviewed-by: Brian Paul <brianp@vmware.com>
11 years agoswrast: Fix memory leak.
Vinson Lee [Fri, 1 Feb 2013 07:28:41 +0000 (23:28 -0800)]
swrast: Fix memory leak.

Fixes resource leak defect reported by Coverity.

Signed-off-by: Vinson Lee <vlee@freedesktop.org>
Reviewed-by: Brian Paul <brianp@vmware.com>
11 years agoconfigure.ac: Fix --with-llvm-shared-libs
Quentin Glidic [Fri, 25 Jan 2013 14:02:04 +0000 (15:02 +0100)]
configure.ac: Fix --with-llvm-shared-libs

The third argument of AC_ARG_WITH is evaluated for any provided value,
not only on --with-, so it must not force-enable the feature
Also, setting $with_llvm_shared_libs in the opencl check was overriding
the user switch

https://bugs.freedesktop.org/show_bug.cgi?id=59851

Signed-off-by: Quentin Glidic <sardemff7+git@sardemff7.net>
11 years agor600g/llvm: Select the correct GPU type for RV670
Tom Stellard [Fri, 1 Feb 2013 16:03:56 +0000 (16:03 +0000)]
r600g/llvm: Select the correct GPU type for RV670

RV670 belongs in the R600 chip class

https://bugs.freedesktop.org/show_bug.cgi?id=58666

NOTE: This is a candidate for the 9.1 branch

11 years agointel: implement create image from texture
Abdiel Janulgue [Wed, 28 Nov 2012 11:30:18 +0000 (13:30 +0200)]
intel: implement create image from texture

Save miptree level info to DRIImage:
- Appropriately-aligned base offset pointing to the image
- Additional x/y adjustment offsets from above.

v8:  -Bump intelImageExtension version
v9:  -Don't use internal _eglError but implement error reporting in new DRI inteface
      instead. This fixes Android build problems based on feedback from
      Adrian M Negreanu and Chad Versace.
     -Move the non-tile-aligned check and error-reporting to intel_set_texture_image_region
v10: -Don't #include "egl/main/eglcurrent.h". [chadv]

Reviewed-by: Eric Anholt <eric@anholt.net> (v6)
Acked-by: Chad Versace <chad.versace@linux.intel.com> (v10)
Signed-off-by: Abdiel Janulgue <abdiel.janulgue@linux.intel.com>
11 years agointel: Account for mt->offset in intel_miptree_map
Abdiel Janulgue [Fri, 25 Jan 2013 13:37:00 +0000 (15:37 +0200)]
intel: Account for mt->offset in intel_miptree_map

We need to take account the offset from original bo when using glTexSubImage()
and other functions that manipulate the subregion of an exported texture.
Offsets are appended to mapped region address and when blitting from a source
region.

Reviewed-by: Eric Anholt <eric@anholt.net>
Reviewed-by: Chad Versace <chad.versace@linux.intel.com>
Signed-off-by: Abdiel Janulgue <abdiel.janulgue@linux.intel.com>
11 years agointel: Create a miptree using offsets in intel_set_texture_image_region
Abdiel Janulgue [Wed, 28 Nov 2012 13:10:33 +0000 (15:10 +0200)]
intel: Create a miptree using offsets in intel_set_texture_image_region

When binding a region to a texture image, re-create the miptree base-level
considering the offset and dimension information exported by DRIImage.

v8: - Move the alignment surface address checks from the image-from-texture
      code to the texture-from-image side. This allows the error reporting to conform to
      OES_EGL_Image and to prevent mixing up EGL and GL errors. Reported by Chad Versace.
    - Addressed an existing issue in renderbuffer case where there is a
      a possibility of creating EGL images out of depthstencil textures which isn't
      really possible. This was spotted by Eric earlier.

Reviewed-by: Eric Anholt <eric@anholt.net> (v6)
Reviewed-by: Chad Versace <chad.versace@linux.intel.com> (v8)
Signed-off-by: Abdiel Janulgue <abdiel.janulgue@linux.intel.com>
11 years agoi965: Account for offsets when updating SURFACE_STATE.
Abdiel Janulgue [Wed, 28 Nov 2012 10:34:21 +0000 (12:34 +0200)]
i965: Account for offsets when updating SURFACE_STATE.

If the offsets are present, this lets us specify a particular level and slice
in a shared region using the base level of an exported mip-map tree.

Reviewed-by: Eric Anholt <eric@anholt.net>
Reviewed-by: Chad Versace <chad.versace@linux.intel.com>
Signed-off-by: Abdiel Janulgue <abdiel.janulgue@linux.intel.com>
11 years agointel: add pixel offset calculator for miptree levels
Abdiel Janulgue [Fri, 7 Dec 2012 11:40:35 +0000 (13:40 +0200)]
intel: add pixel offset calculator for miptree levels

Add helper to calculate fine-grained x and y adjustment pixels
to an image within a miptree level for tiled regions.

Reviewed-by: Eric Anholt <eric@anholt.net>
Reviewed-by: Chad Versace <chad.versace@linux.intel.com>
Signed-off-by: Abdiel Janulgue <abdiel.janulgue@linux.intel.com>
11 years agointel: Expose intel_miptree_create_internal as intel_miptree_create_layout.
Abdiel Janulgue [Thu, 17 Jan 2013 15:04:01 +0000 (17:04 +0200)]
intel: Expose intel_miptree_create_internal as intel_miptree_create_layout.

Reviewed-by: Eric Anholt <eric@anholt.net>
Reviewed-by: Chad Versace <chad.versace@linux.intel.com>
Signed-off-by: Abdiel Janulgue <abdiel.janulgue@linux.intel.com>
11 years agointel: expose dimensions and offsets of a miptree level in DRIImage
Abdiel Janulgue [Thu, 17 Jan 2013 15:04:27 +0000 (17:04 +0200)]
intel: expose dimensions and offsets of a miptree level in DRIImage

v8: - Append has_depthstencil field in DRIImage structure.

Reviewed-by: Eric Anholt <eric@anholt.net> (v6)
Reviewed-by: Chad Versace <chad.versace@linux.intel.com> (v8)
Signed-off-by: Abdiel Janulgue <abdiel.janulgue@linux.intel.com>
11 years agodri2: Create image from texture
Abdiel Janulgue [Tue, 27 Nov 2012 13:19:54 +0000 (15:19 +0200)]
dri2: Create image from texture

Add create image from texture extension and bump version.

v8: - Add appropriate image errors codes in DRI interface so we don't
      have to use internal EGL functions in driver. Suggested by Chad Versace.

Reviewed-by: Eric Anholt <eric@anholt.net> (v6)
Reviewed-by: Chad Versace <chad.versace@linux.intel.com> (v8)
Signed-off-by: Abdiel Janulgue <abdiel.janulgue@linux.intel.com>
11 years agoradeonsi: Fix draws using user index buffer.
Michel Dänzer [Fri, 1 Feb 2013 17:49:07 +0000 (18:49 +0100)]
radeonsi: Fix draws using user index buffer.

Was broken since commit bf469f4edc60bd1c5fd770cb231b8d5ab801427f
('gallium: add void *user_buffer in pipe_index_buffer').

Fixes 11 piglit tests and lots of missing geometry e.g. in TORCS.

NOTE: This is a candidate for the 9.1 branch.

11 years agost/mesa: whitespace/indentation fix
Brian Paul [Fri, 1 Feb 2013 00:57:24 +0000 (17:57 -0700)]
st/mesa: whitespace/indentation fix

11 years agosvga: check for NaN shader immediates
Brian Paul [Thu, 31 Jan 2013 00:44:25 +0000 (17:44 -0700)]
svga: check for NaN shader immediates

The svga device doesn't handle them.  Replace with zeros.
Fixes several piglit tests, such as "glsl-const-builtin-inversesqrt".

Reviewed-by: Reviewed-by: José Fonseca <jfonseca@vmware.com>
11 years agosvga: add, use SVGA3D_SURFACE_HINT_VOLUME flag
Brian Paul [Thu, 31 Jan 2013 00:43:57 +0000 (17:43 -0700)]
svga: add, use SVGA3D_SURFACE_HINT_VOLUME flag

Reviewed-by: Reviewed-by: José Fonseca <jfonseca@vmware.com>
11 years agotrace: measure time for each gallium call
Brian Paul [Wed, 30 Jan 2013 22:45:08 +0000 (15:45 -0700)]
trace: measure time for each gallium call

To get a rough idea of how much time is spent in each gallium driver
function.  The time is measured in microseconds.

11 years agotrace: add void to function definition
Brian Paul [Tue, 29 Jan 2013 23:49:58 +0000 (16:49 -0700)]
trace: add void to function definition

11 years agotrace: allow GALLIUM_TRACE=stdout/stderr
Brian Paul [Tue, 29 Jan 2013 23:48:25 +0000 (16:48 -0700)]
trace: allow GALLIUM_TRACE=stdout/stderr