mesa.git
11 years agobuild: Remove GALLIUM_DIRS
Matt Turner [Sat, 19 Jan 2013 08:09:36 +0000 (00:09 -0800)]
build: Remove GALLIUM_DIRS

It's always constant anyway.

Tested-by: Emil Velikov <emil.l.velikov@gmail.com>
Reviewed-and-Tested-by: Andreas Boll <andreas.boll.dev@gmail.com>
Reviewed-by: Jordan Justen <jordan.l.justen@intel.com>
11 years agobuild: Get rid of SRC_DIRS
Matt Turner [Sat, 19 Jan 2013 07:59:49 +0000 (23:59 -0800)]
build: Get rid of SRC_DIRS

Tested-by: Emil Velikov <emil.l.velikov@gmail.com>
Reviewed-and-Tested-by: Andreas Boll <andreas.boll.dev@gmail.com>
Reviewed-by: Jordan Justen <jordan.l.justen@intel.com>
11 years agobuild: Get rid of CORE_DIRS
Matt Turner [Sat, 19 Jan 2013 07:43:05 +0000 (23:43 -0800)]
build: Get rid of CORE_DIRS

A step toward working make dist/distcheck.

Tested-by: Emil Velikov <emil.l.velikov@gmail.com>
Reviewed-and-Tested-by: Andreas Boll <andreas.boll.dev@gmail.com>
Reviewed-by: Jordan Justen <jordan.l.justen@intel.com>
11 years agobuild: Move src/mapi/mapi/* to src/mapi/
Matt Turner [Mon, 21 Jan 2013 18:27:25 +0000 (10:27 -0800)]
build: Move src/mapi/mapi/* to  src/mapi/

Tested-by: Emil Velikov <emil.l.velikov@gmail.com>
Reviewed-and-Tested-by: Andreas Boll <andreas.boll.dev@gmail.com>
Reviewed-by: Jordan Justen <jordan.l.justen@intel.com>
11 years agobuild: Rename sources.mak -> Makefile.sources
Matt Turner [Mon, 21 Jan 2013 18:25:31 +0000 (10:25 -0800)]
build: Rename sources.mak -> Makefile.sources

For the sake of consistency.

Tested-by: Emil Velikov <emil.l.velikov@gmail.com>
Reviewed-and-Tested-by: Andreas Boll <andreas.boll.dev@gmail.com>
Reviewed-by: Jordan Justen <jordan.l.justen@intel.com>
11 years agoradeonsi: Read config values from the .AMDGPU.config ELF section
Tom Stellard [Thu, 4 Apr 2013 20:21:06 +0000 (16:21 -0400)]
radeonsi: Read config values from the .AMDGPU.config ELF section

Instead of emitting configuration values (e.g. number of gprs used) in a
predefined order, the LLVM backend now emits these values in
register/value pairs.  The first dword contains the register address and
the second dword contians the value to write.

Reviewed-by: Michel Dänzer <michel.daenzer@amd.com>
11 years agoradeon/llvm: Handle ELF formatted binary output from the LLVM backend
Tom Stellard [Thu, 4 Apr 2013 20:02:51 +0000 (13:02 -0700)]
radeon/llvm: Handle ELF formatted binary output from the LLVM backend

11 years agoradeon/llvm: Use a struct for storing compiled code
Tom Stellard [Thu, 4 Apr 2013 16:57:13 +0000 (09:57 -0700)]
radeon/llvm: Use a struct for storing compiled code

11 years agogallivm: fix small but severe bug in handling multiple lod level strides
Roland Scheidegger [Mon, 15 Apr 2013 01:57:23 +0000 (03:57 +0200)]
gallivm: fix small but severe bug in handling multiple lod level strides

Inserting the value for the second quad in the wrong place for the
following shuffle. This meant the row or image stride was undefined which is
quite catastrophic, can lead to bogus texels fetched or just segfault.
This code is only hit for SoA path currently, still surprising it
didn't crash more or caused more visible issues (I think llvm used a
broadcast shuffle for the undefined parts of the vector, hence the undefined
value for the second quad was just the same as that from the first quad,
so as long as both quads hit the same mip level everything was fine, and since
lower mips always have the same large stride it made it less likely to
hit out-of-bound memory in case of differing lods).

Note: this is a candidate for stable branches.

Reviewed-by: Jose Fonseca <jfonseca@vmware.com>
11 years agoclover: Fix usage of incorrect object as destination in clEnqueueCopyBufferToImage.
Francisco Jerez [Sun, 7 Apr 2013 16:31:06 +0000 (18:31 +0200)]
clover: Fix usage of incorrect object as destination in clEnqueueCopyBufferToImage.

Signed-off-by: Francisco Jerez <currojerez@riseup.net>
11 years agoclover: Define platform class and merge with device_registry.
Francisco Jerez [Sat, 6 Apr 2013 12:35:00 +0000 (14:35 +0200)]
clover: Define platform class and merge with device_registry.

Null platform IDs are OK according to the spec, but some applications have
been reported to get paranoid and assume that our NULL platform is unusable.

As it doesn't hurt to have device enumeration separate from the rest of the
device code (quite the opposite, it makes the code cleaner), make the API use
an actual platform object that keeps track of the available devices instead of
the former NULL pointer.

Reported-and-reviewed-by: Tom Stellard <thomas.stellard@amd.com>
Signed-off-by: Francisco Jerez <currojerez@riseup.net>
11 years agoclover: Add missing fields to the module serializer.
Francisco Jerez [Sat, 13 Apr 2013 10:17:32 +0000 (12:17 +0200)]
clover: Add missing fields to the module serializer.

Signed-off-by: Francisco Jerez <currojerez@riseup.net>
11 years agoi965: Shut up the last release build warning.
Eric Anholt [Thu, 11 Apr 2013 17:46:02 +0000 (10:46 -0700)]
i965: Shut up the last release build warning.

I don't see a sensible value to use in this path, but we shouldn't ever
hit this outside of developer new-texture-target enabling.

Reviewed-by: Matt Turner <mattst88@gmail.com>
11 years agoi965: Silence one more compile warning.
Eric Anholt [Thu, 11 Apr 2013 17:38:04 +0000 (10:38 -0700)]
i965: Silence one more compile warning.

We don't want to store this thing in the class, and we do need the
definition to be at the top of the function and held onto until the end
here, so there's not much to do besides (void) reference it.

Reviewed-by: Matt Turner <mattst88@gmail.com>
11 years agoi965: Fix a warning in the release build.
Eric Anholt [Thu, 11 Apr 2013 17:37:13 +0000 (10:37 -0700)]
i965: Fix a warning in the release build.

This was copy and pasted from can_reswizzle_dst(), and we can just fold it
in instead to avoid the warning.

Reviewed-by: Matt Turner <mattst88@gmail.com>
11 years agoi965: Fix an unused variable warning in the release build.
Eric Anholt [Thu, 11 Apr 2013 17:34:59 +0000 (10:34 -0700)]
i965: Fix an unused variable warning in the release build.

I think this actually clarifies what's going on in the asserts a bit,
given how many regions we've got floating around.

Reviewed-by: Matt Turner <mattst88@gmail.com>
11 years agoi965: Fix an unused variable warning in the release build.
Eric Anholt [Thu, 11 Apr 2013 17:33:00 +0000 (10:33 -0700)]
i965: Fix an unused variable warning in the release build.

It's used in an assert, but we have this as a member of the class anyway.

Reviewed-by: Matt Turner <mattst88@gmail.com>
11 years agointel: Return failure properly in the texsubimage blit path.
Eric Anholt [Thu, 11 Apr 2013 17:30:51 +0000 (10:30 -0700)]
intel: Return failure properly in the texsubimage blit path.

We assert that failure doesn't happen, but it fixes a warning in the
release build and it would at least give working behavior for a user by
falling back to the normal texsubimage path.

Reviewed-by: Matt Turner <mattst88@gmail.com>
11 years agointel: Fix a warning in the release build.
Eric Anholt [Thu, 11 Apr 2013 17:16:26 +0000 (10:16 -0700)]
intel: Fix a warning in the release build.

This was silly -- checking that we didn't overflow the array by dividing
the array size by 2 and then multiplying it back up by 2.

Reviewed-by: Matt Turner <mattst88@gmail.com>
11 years agointel: Fix an unused variable warning in the release build.
Eric Anholt [Thu, 11 Apr 2013 17:14:58 +0000 (10:14 -0700)]
intel: Fix an unused variable warning in the release build.

Reviewed-by: Matt Turner <mattst88@gmail.com>
11 years agointel: Improve diagnostics for emit_linear_blit failure path.
Eric Anholt [Thu, 11 Apr 2013 17:11:50 +0000 (10:11 -0700)]
intel: Improve diagnostics for emit_linear_blit failure path.

This fixes unused variable warnings in the release build, and should be
more useful if it ever triggers.

Reviewed-by: Matt Turner <mattst88@gmail.com>
11 years agoi965: Fix error path for MCS allocation.
Eric Anholt [Thu, 11 Apr 2013 17:08:56 +0000 (10:08 -0700)]
i965: Fix error path for MCS allocation.

Asserts don't stop execution in release builds, so we would continue on to
use an uninitialized format value.  Just take the failure path, which
appears to continue up the call stack for a while.

Reviewed-by: Matt Turner <mattst88@gmail.com>
11 years agoi830: Move assert-only code into the assert.
Eric Anholt [Thu, 11 Apr 2013 17:06:06 +0000 (10:06 -0700)]
i830: Move assert-only code into the assert.

The call has no side effects, and moving it into the assert cleans up a
compile warning in the release build.

Reviewed-by: Matt Turner <mattst88@gmail.com>
11 years agoi965/fs: Fix some untriggered optimization bugs with uncompressed/sechalf.
Eric Anholt [Thu, 11 Apr 2013 16:54:41 +0000 (09:54 -0700)]
i965/fs: Fix some untriggered optimization bugs with uncompressed/sechalf.

We have this support for firsthalf/sechalf instructions, which would be
called in the !has_compr4 (aka original gen4) 16-wide case.  We currently
only support 16-wide for gen5+, so we weren't tripping over this, but it
would have been a problem if we ever try to enable it.

Reviewed-by: Matt Turner <mattst88@gmail.com>
11 years agoi965/fs: Add basic-block-level dead code elimination.
Eric Anholt [Tue, 9 Apr 2013 00:43:06 +0000 (17:43 -0700)]
i965/fs: Add basic-block-level dead code elimination.

This is a poor substitute for proper global dead code elimination that
could replace both our current paths, but it was very easy to write.  It
particularly helps with Valve's shaders that are translated out of DX
assembly, which has been register allocated and thus have a bunch of
unrelated uses of the same variable (some of which get copy-propagated
from and then left for dead).

shader-db results:
total instructions in shared programs: 1735753 -> 1731698 (-0.23%)
instructions in affected programs:     492620 -> 488565 (-0.82%)

v2: Fix comment typo

Reviewed-by: Matt Turner <mattst88@gmail.com>
11 years agoi965/fs: Remove incorrect note of writing attr in centroid workaround.
Eric Anholt [Tue, 9 Apr 2013 01:46:23 +0000 (18:46 -0700)]
i965/fs: Remove incorrect note of writing attr in centroid workaround.

This instruction doesn't update its IR destination, it just moves from
payload to f0.  This caused the dead code elimination pass I'm adding to
dead-code-eliminate the first step of interpolation.

Reviewed-by: Matt Turner <mattst88@gmail.com>
11 years agoi965/fs: Add a helper function for checking for partial register updates.
Eric Anholt [Mon, 4 Jun 2012 15:59:00 +0000 (08:59 -0700)]
i965/fs: Add a helper function for checking for partial register updates.

These checks were all over, and every time I wrote one I had to try to
decide again what the cases were for partial updates.

v2: Fix inadvertent reladdr check removal.
Reviewed-by: Matt Turner <mattst88@gmail.com>
11 years agomesa: Add a macro to bitset for determining bitset size.
Eric Anholt [Thu, 4 Apr 2013 16:47:03 +0000 (09:47 -0700)]
mesa: Add a macro to bitset for determining bitset size.

Reviewed-by: Matt Turner <mattst88@gmail.com>
11 years agoi965: Fix compiler warnings since the introduction of texture multisample.
Eric Anholt [Fri, 5 Apr 2013 20:43:52 +0000 (13:43 -0700)]
i965: Fix compiler warnings since the introduction of texture multisample.

Reviewed-by: Matt Turner <mattst88@gmail.com>
11 years agomesa: Don't leak gl_context::BeginEnd at context destruction
Ian Romanick [Wed, 10 Apr 2013 14:50:49 +0000 (07:50 -0700)]
mesa: Don't leak gl_context::BeginEnd at context destruction

The other dispatch tables (Exec and Save) are freed, but BeginEnd is
never freed.  This was found by inspection why investigating the leak of
shared state in _mesa_initialize_context.

NOTE: This is a candidate for stable branches

Signed-off-by: Ian Romanick <ian.d.romanick@intel.com>
Reviewed-by: Brian Paul <brianp@vmware.com>
11 years agomesa: Don't leak shared state when context initialization fails
Ian Romanick [Wed, 10 Apr 2013 14:47:08 +0000 (07:47 -0700)]
mesa: Don't leak shared state when context initialization fails

Back up at line 1017 (not shown in patch), we add a reference to the
shared state.  Several places after that may divert to the error
handler, but, as far as I can tell, nothing ever unreferences the shared
state.

Fixes issue identified by Klocwork analysis:

    Resource acquired to 'shared->TexMutex' at line 1012 may be lost
    here. Also there is one similar error on line 1087.

NOTE: This is a candidate for the stable branches.

Signed-off-by: Ian Romanick <ian.d.romanick@intel.com>
Reviewed-by: Brian Paul <brianp@vmware.com>
11 years agoegl/dri2: NULL check value returned by dri2_create_surface
Ian Romanick [Tue, 9 Apr 2013 03:00:30 +0000 (20:00 -0700)]
egl/dri2: NULL check value returned by dri2_create_surface

dri2_create_surface can fail for a variety of reasons, including bad
input data.  Dereferencing the NULL pointer and crashing is not okay.

Fixes issue identified by Klocwork analysis:

    Pointer 'surf' returned from call to function 'dri2_create_surface'
    at line 285 may be NULL and will be dereferenced at line 291.

NOTE: This is a candidate for the stable branches.

Signed-off-by: Ian Romanick <ian.d.romanick@intel.com>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
11 years agomesa: NULL check the pointer before trying to dereference it
Ian Romanick [Tue, 9 Apr 2013 02:55:18 +0000 (19:55 -0700)]
mesa: NULL check the pointer before trying to dereference it

Duh.

Fixes issues identified by Klocwork analysis:

    Pointer 'table' returned from call to function 'calloc' at line 115
    may be NULL and will be dereferenced at line 117.

and

    Suspicious dereference of pointer 'table' before NULL check at line
    119.

NOTE: This is a candidate for the stable branches.

Signed-off-by: Ian Romanick <ian.d.romanick@intel.com>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
11 years agoglsl: Fix hypothetical NULL dereference related to process_array_type
Ian Romanick [Mon, 8 Apr 2013 23:53:46 +0000 (16:53 -0700)]
glsl: Fix hypothetical NULL dereference related to process_array_type

Ensure that process_array_type never returns NULL, and let
process_array_type handle the case where the supplied base type is NULL.

Fixes issues identified by Klocwork analysis:

    Pointer 'type' returned from call to function 'get_type' at line
    1907 may be NULL and may be dereferenced at line 1912.

and

    Pointer 'field_type' checked for NULL at line 4160 will be
    dereferenced at line 4165. Also there is one similar error on line
    4174.

Signed-off-by: Ian Romanick <ian.d.romanick@intel.com>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
11 years agoglsl: Fix hypothetical NULL dereference in ast_process_structure_or_interface_block
Ian Romanick [Mon, 8 Apr 2013 23:37:04 +0000 (16:37 -0700)]
glsl: Fix hypothetical NULL dereference in ast_process_structure_or_interface_block

Fixes issue identified by Klocwork analysis:

    Pointer 'field_type' returned from call to function 'glsl_type' at
    line 4126 may be NULL and may be dereferenced at line 4139.  Also
    there are 2 similar errors on line(s) 4165, 4174.

In practice, it should be impossible to actually get NULL in here
because a syntax error would have already caused compilation to halt.

Signed-off-by: Ian Romanick <ian.d.romanick@intel.com>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
11 years agor300g: Fix bug in OMOD optimization
Tom Stellard [Wed, 20 Mar 2013 02:39:31 +0000 (22:39 -0400)]
r300g: Fix bug in OMOD optimization

https://bugs.freedesktop.org/show_bug.cgi?id=60503

NOTE: This is a candidate for the stable branches.

11 years agonvc0: set ret variable if launch desc allocation failed
Emil Velikov [Thu, 11 Apr 2013 23:22:33 +0000 (00:22 +0100)]
nvc0: set ret variable if launch desc allocation failed

Pointed out by gcc

nve4_compute.c: In function 'nve4_launch_grid':
nve4_compute.c:511:7: warning: 'ret' may be used uninitialized in
 this function [-Wmaybe-uninitialized]
    if (ret)
       ^

Signed-off-by: Emil Velikov <emil.l.velikov@gmail.com>
Edit by Christoph Bumiller:
Set it to -1 to indicate failure and only when it's actually required.

11 years agonvc0: bail out early during nve4_compute_setup()
Emil Velikov [Thu, 11 Apr 2013 23:22:32 +0000 (00:22 +0100)]
nvc0: bail out early during nve4_compute_setup()

Exit gracefully rather than trying to create a random object, whenever the
chipset is unknown

Signed-off-by: Emil Velikov <emil.l.velikov@gmail.com>
11 years agonvc0: compile nve4_cache_split_name() only in debug build
Emil Velikov [Thu, 11 Apr 2013 23:22:31 +0000 (00:22 +0100)]
nvc0: compile nve4_cache_split_name() only in debug build

As otherwise it is unused - pointed out by gcc

nve4_compute.c:586:20: warning: 'nve4_cache_split_name' defined but not used [-Wunused-function]
 static const char *nve4_cache_split_name(unsigned value)
                    ^

Signed-off-by: Emil Velikov <emil.l.velikov@gmail.com>
11 years agonv50/codegen: do not emitATOM() if the subOp is unknown
Emil Velikov [Thu, 11 Apr 2013 23:22:30 +0000 (00:22 +0100)]
nv50/codegen: do not emitATOM() if the subOp is unknown

For debug build we'll hit the assert, for release we are going to emit random data
as subOp is used uninitilised. Spotted by gcc

codegen/nv50_ir_emit_nv50.cpp: In member function 'void nv50_ir::CodeEmitterNV50::emitATOM(const nv50_ir::Instruction*)':
codegen/nv50_ir_emit_nv50.cpp:1554:12: warning: 'subOp' may be used uninitialized in this function [-Wmaybe-uninitialized]
    uint8_t subOp;
            ^

Signed-off-by: Emil Velikov <emil.l.velikov@gmail.com>
11 years agonvc0: implement multisample textures
Christoph Bumiller [Sat, 6 Apr 2013 12:52:05 +0000 (14:52 +0200)]
nvc0: implement multisample textures

11 years agonvc0: patch up TEX cases with 5 or 6 sources on nve4
Christoph Bumiller [Sat, 6 Apr 2013 15:40:02 +0000 (17:40 +0200)]
nvc0: patch up TEX cases with 5 or 6 sources on nve4

Hackishly fixes alignment requirement of 2nd tuple for now.

11 years agonvc0: fix 2D engine MS2 resolve
Christoph Bumiller [Sat, 6 Apr 2013 12:15:54 +0000 (14:15 +0200)]
nvc0: fix 2D engine MS2 resolve

11 years agonv50,nvc0: add RGBX16/32_FLOAT formats
Christoph Bumiller [Mon, 8 Apr 2013 11:47:26 +0000 (13:47 +0200)]
nv50,nvc0: add RGBX16/32_FLOAT formats

11 years agoi965/vs: Print error if vertex shader fails to compile.
Matt Turner [Thu, 11 Apr 2013 16:57:23 +0000 (09:57 -0700)]
i965/vs: Print error if vertex shader fails to compile.

Reviewed-by: Eric Anholt <eric@anholt.net>
11 years agoi965: NULL check prog on shader compilation failure.
Matt Turner [Thu, 11 Apr 2013 16:55:42 +0000 (09:55 -0700)]
i965: NULL check prog on shader compilation failure.

Also change if (shader) to if (prog) for consistency.

Reviewed-by: Eric Anholt <eric@anholt.net>
11 years agoscons: Add st_cb_msaa.c to source list.
José Fonseca [Thu, 11 Apr 2013 21:37:34 +0000 (22:37 +0100)]
scons: Add st_cb_msaa.c to source list.

11 years agor600g: add get_sample_position support (v3)
Dave Airlie [Sun, 3 Mar 2013 20:19:07 +0000 (06:19 +1000)]
r600g: add get_sample_position support (v3)

v2: I rewrote this to use the sample positions properly.
v3: rewrite properly to use bitfield to cast back to signed ints

Signed-off-by: Dave Airlie <airlied@redhat.com>
11 years agost/mesa: add support for ARB_texture_multisample (v3)
Dave Airlie [Sun, 3 Mar 2013 20:17:11 +0000 (06:17 +1000)]
st/mesa: add support for ARB_texture_multisample (v3)

This adds support to the mesa state tracker for ARB_texture_multisample.

hardware doesn't seem to use a different texture instructions, so
I don't think we need to create one for TGSI at this time.

Thanks to Marek for fixes to sample number picking.

v2: idr pointed out a bug in how we picked the max sample counts,
use new internal format chooser interface to pick proper answers.
v3: use st_choose_format directly, it was okay, fix anding of masks.

Reviewed-by: Marek Olšák <maraeo@gmail.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>
11 years agost/mesa: add support for get sample position
Dave Airlie [Sun, 3 Mar 2013 20:14:18 +0000 (06:14 +1000)]
st/mesa: add support for get sample position

This just calls into the gallium interface.

Reviewed-by: Marek Olšák <maraeo@gmail.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>
11 years agogallium: add get_sample_position interface
Dave Airlie [Sun, 3 Mar 2013 20:11:07 +0000 (06:11 +1000)]
gallium: add get_sample_position interface

This is to be used to implement glGet GL_SAMPLE_POSITION.

Reviewed-by: Marek Olšák <maraeo@gmail.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>
11 years agor600g: fix two issues in compressed msaa reading code
Dave Airlie [Thu, 11 Apr 2013 04:51:35 +0000 (05:51 +0100)]
r600g: fix two issues in compressed msaa reading code

I've no idea when sample_chan would ever be 4 here, but 4 is most
definitely wrong, array textures have it as 3 as well.

Also the cayman code though unused is obviously wrong.

Signed-off-by: Dave Airlie <airlied@redhat.com>
11 years agoi965/vs: Don't hardcode DEBUG_VS in generic vec4 code.
Paul Berry [Sat, 23 Mar 2013 04:55:03 +0000 (21:55 -0700)]
i965/vs: Don't hardcode DEBUG_VS in generic vec4 code.

Since the vec4_visitor and vec4_generator classes are going to be
re-used for geometry shaders, we can't enable their debug
functionality based on (INTEL_DEBUG & DEBUG_VS) anymore.  Instead, add
a debug_flag boolean to these two classes, so that when they're
instantiated the caller can specify whether debug dumps are needed.

Reviewed-by: Jordan Justen <jordan.l.justen@intel.com>
Reviewed-by: Eric Anholt <eric@anholt.net>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
11 years agoi965/vs: Generalize computation of array strides in preparation for GS.
Paul Berry [Fri, 22 Mar 2013 18:26:34 +0000 (11:26 -0700)]
i965/vs: Generalize computation of array strides in preparation for GS.

Geometry shader inputs are arrays, but they use an unusual array
layout: instead of all array elements for a given geometry shader
input being stored consecutively, all geometry shader inputs are
interleaved into one giant array.  As a result, the array stride we
use to access geometry shader inputs must be equal to the size of the
input VUE, rather than the size of the array element.

This patch introduces a new virtual function,
vec4_visitor::compute_array_stride(), which will allow geometry shader
compilation to specialize the computation of array stride to account
for the unusual layout of geometry shader input arrays.  It also
renames the local variable that the ir_dereference_array visitor uses
to store the stride, to avoid confusion.

Reviewed-by: Jordan Justen <jordan.l.justen@intel.com>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
11 years agoi965/vs: Generalize attribute setup code in preparation for GS.
Paul Berry [Fri, 22 Mar 2013 15:47:49 +0000 (08:47 -0700)]
i965/vs: Generalize attribute setup code in preparation for GS.

This patch introduces a new function,
vec4_visitor::lower_attributes_to_hw_regs(), which replaces registers
of type ATTR in the instruction stream with the hardware registers
that store those attributes.  This logic will need to be common
between the vertex and geometry shaders.

Reviewed-by: Jordan Justen <jordan.l.justen@intel.com>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
11 years agoi965/vs: Generalize vertex emission code in preparation for GS.
Paul Berry [Fri, 22 Mar 2013 13:59:52 +0000 (06:59 -0700)]
i965/vs: Generalize vertex emission code in preparation for GS.

This patch introduces a new function, vec4_visitor::emit_vertex(),
which contains the code for emitting vertices that will need to be
common between the vertex and geometry shaders.

Geometry shaders will need to use a different message header, and a
different opcode, for their URB writes, so we introduce virtual
functions emit_urb_write_header() and emit_urb_write_opcode() to take
care of the GS-specific behaviours.

Also, since vertex emission happens at the end of the VS, but in the
middle of the GS, we need to be sure to only call
emit_shader_time_end() during VS vertex emission.  We accomplish this
by moving the call to emit_shader_time_end() into the VS
implementation of emit_urb_write_opcode().

Reviewed-by: Jordan Justen <jordan.l.justen@intel.com>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
11 years agoi965/vs: rename vec4_generator::generate_vs_instruction.
Paul Berry [Sun, 17 Feb 2013 19:34:05 +0000 (11:34 -0800)]
i965/vs: rename vec4_generator::generate_vs_instruction.

Since this function is going to get used for geometry shaders too, it
deserves a more generic name: generate_vec4_instruction.

Reviewed-by: Jordan Justen <jordan.l.justen@intel.com>
Reviewed-by: Eric Anholt <eric@anholt.net>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
11 years agoi965/vs: Generalize data structures pointed to by vec4_generator.
Paul Berry [Sun, 17 Feb 2013 19:25:37 +0000 (11:25 -0800)]
i965/vs: Generalize data structures pointed to by vec4_generator.

This patch removes the following field from vec4_generator, since it
is not used:

- struct brw_vs_compile *c

And changes the following field:

- struct gl_vertex_program *vp => struct gl_program *prog

With these changes, vec4_generator no longer refers to any VS-specific
data structures.  This will pave the way for re-using it for geometry
shaders.

Reviewed-by: Jordan Justen <jordan.l.justen@intel.com>
v2: Use the name "prog" rather than "p".

Reviewed-by: Eric Anholt <eric@anholt.net>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
11 years agoi965/vs: Rename vec4_generator::prog to shader_prog.
Paul Berry [Tue, 9 Apr 2013 21:31:28 +0000 (14:31 -0700)]
i965/vs: Rename vec4_generator::prog to shader_prog.

The next patch is going to change the type of vec4_generator::vp from
struct gl_vertex_program * to struct gl_program *, and rename it.  The
sensible name to change it to is vec4_generator::prog.  However, prog
is already used.  Since the existing vec4_generator::prog is of type
struct gl_shader_program, it makes sense to rename it to shader_prog.

Reviewed-by: Eric Anholt <eric@anholt.net>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
11 years agoi965/vs: move VS-specific data members to vs_vec4_visitor.
Paul Berry [Sun, 17 Feb 2013 17:09:45 +0000 (09:09 -0800)]
i965/vs: move VS-specific data members to vs_vec4_visitor.

This patch moves the following data structures from vec4_visitor to
vec4_vs_visitor, since they contain VS-specific data:

- struct brw_vs_compile *c (renamed to vs_compile)
- struct brw_vs_prog_data *prog_data (renamed to vs_prog_data)
- src_reg *vp_temp_regs
- src_reg vp_addr_reg

Since brw_vs_compile and brw_vs_prog_data also contain vec4-generic
data, the following pointers are added to the base class, to allow it
to access the vec4-generic portions of these data structures:

- struct brw_vec4_compile *c
- struct brw_vec4_prog_key *key
- struct brw_vec4_prog_data *prog_data

Reviewed-by: Jordan Justen <jordan.l.justen@intel.com>
Reviewed-by: Eric Anholt <eric@anholt.net>
v2: Use shorter names in the base class and longer names in the
derived class.

Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
11 years agoi965/vs: move ARB_vertex_program functions to vec4_vs_visitor.
Paul Berry [Sat, 16 Feb 2013 04:27:03 +0000 (20:27 -0800)]
i965/vs: move ARB_vertex_program functions to vec4_vs_visitor.

This patch moves functions from vec4_visitor to vec4_vs_visitor that
deal with ARB (assembly) vertex programs.  There's no point in having
these functions in the base class since we don't intend to support
assembly programs for the GS stage.  The following functions are
moved:

- setup_vp_regs
- get_vp_dst_reg
- get_vp_src_reg

Reviewed-by: Jordan Justen <jordan.l.justen@intel.com>
Reviewed-by: Eric Anholt <eric@anholt.net>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
11 years agoi965/vs: Add virtual function make_reg_for_system_value().
Paul Berry [Sun, 17 Feb 2013 16:05:52 +0000 (08:05 -0800)]
i965/vs: Add virtual function make_reg_for_system_value().

The system values handled by vec4_visitor::visit(ir_variable *) are
VS-specific (vertex ID and instance ID).  This patch moves the
handling of those values into a new virtual function,
make_reg_for_system_value(), so that this VS-specific code won't be
inherited by geomtry shaders.

Reviewed-by: Jordan Justen <jordan.l.justen@intel.com>
Reviewed-by: Eric Anholt <eric@anholt.net>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
11 years agoi965/vs: Make some vec4_visitor functions virtual.
Paul Berry [Sun, 17 Feb 2013 16:05:52 +0000 (08:05 -0800)]
i965/vs: Make some vec4_visitor functions virtual.

This patch makes the following vec4_visitor functions virtual, since
they will need to be implemented differently for vertex and geometry
shaders.  Some of the functions are renamed to reflect their generic
purpose, rather than their VS-specific behaviour:

- setup_attributes
- emit_attribute_fixups (renamed to emit_prolog)
- emit_vertex_program_code (renamed to emit_program_code)
- emit_urb_writes (renamed to emit_thread_end)

Reviewed-by: Jordan Justen <jordan.l.justen@intel.com>
Reviewed-by: Eric Anholt <eric@anholt.net>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
11 years agoi965/vs: Make vec4_vs_visitor class derived from vec4_visitor.
Paul Berry [Sat, 16 Feb 2013 04:12:13 +0000 (20:12 -0800)]
i965/vs: Make vec4_vs_visitor class derived from vec4_visitor.

This patch just creates the derived class; later patches will migrate
VS-specific functions and data structures from the base class into the
derived class.

Reviewed-by: Jordan Justen <jordan.l.justen@intel.com>
Reviewed-by: Eric Anholt <eric@anholt.net>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
11 years agoi965/vs: split brw_vs_prog_data into generic and VS-specific parts.
Paul Berry [Sun, 17 Feb 2013 15:48:21 +0000 (07:48 -0800)]
i965/vs: split brw_vs_prog_data into generic and VS-specific parts.

This will allow the generic parts to be re-used for geometry shaders.

Reviewed-by: Jordan Justen <jordan.l.justen@intel.com>
v2: Put urb_read_length and urb_entry_size in the generic struct.

Reviewed-by: Eric Anholt <eric@anholt.net>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
11 years agoi965/vs: split brw_vs_prog_key into generic and VS-specific parts.
Paul Berry [Sat, 16 Feb 2013 16:45:09 +0000 (08:45 -0800)]
i965/vs: split brw_vs_prog_key into generic and VS-specific parts.

This will allow the generic parts to be re-used for geometry shaders.

Reviewed-by: Jordan Justen <jordan.l.justen@intel.com>
Reviewed-by: Eric Anholt <eric@anholt.net>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
11 years agoi965/vs: split brw_vs_compile into generic and VS-specific parts.
Paul Berry [Sat, 16 Feb 2013 04:33:31 +0000 (20:33 -0800)]
i965/vs: split brw_vs_compile into generic and VS-specific parts.

This will allow the generic parts to be re-used for geometry shaders.

Reviewed-by: Jordan Justen <jordan.l.justen@intel.com>
Reviewed-by: Eric Anholt <eric@anholt.net>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
11 years agoi965/vs: Remove brw_vs_prog_data pointer from brw_vs_compile.
Paul Berry [Sat, 16 Feb 2013 17:49:11 +0000 (09:49 -0800)]
i965/vs: Remove brw_vs_prog_data pointer from brw_vs_compile.

In patches that follow, we'll be splitting structs brw_vs_prog_data
and brw_vs_compile into a vec4-generic base struct and a VS-specific
derived struct (this will allow the vec4-generic code to be re-used
for geometry shaders).  Having brw_vs_compile point to
brw_vs_prog_data makes it difficult to do this cleanly.

Fortunately most of the functions that use brw_vs_compile (those in
the vec4_visitor class) already have access to brw_vs_prog_data
through a separate pointer (vec4_visitor::prog_data).  So all we have
to do is use that pointer consistently, and plumb prog_data through
the few remaining functions that need access to it.

Reviewed-by: Jordan Justen <jordan.l.justen@intel.com>
Reviewed-by: Eric Anholt <eric@anholt.net>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
11 years agoi965: Generalize computation of VUE map in preparation for GS.
Paul Berry [Fri, 22 Mar 2013 19:39:40 +0000 (12:39 -0700)]
i965: Generalize computation of VUE map in preparation for GS.

This patch modifies the arguments to brw_compute_vue_map() so that
they no longer bake in the assumption that we are generating a VUE map
for vertex shader outputs.  It also makes the function non-static so
that we can re-use it for geometry shader outputs.

Reviewed-by: Jordan Justen <jordan.l.justen@intel.com>
Reviewed-by: Eric Anholt <eric@anholt.net>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
11 years agoi965/vs: Make type of vec4_visitor::vp more generic.
Paul Berry [Sat, 16 Feb 2013 04:19:23 +0000 (20:19 -0800)]
i965/vs: Make type of vec4_visitor::vp more generic.

The vec4_visitor functions don't use any VS specific data from
vec4_visitor::vp.  So rename it to "prog" and change its type from
struct gl_vertex_program * to struct gl_program *.  This will allow
the code to be re-used for geometry shaders.

Reviewed-by: Jordan Justen <jordan.l.justen@intel.com>
v2: Use the name "prog" rather than "p".

Reviewed-by: Eric Anholt <eric@anholt.net>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
11 years agoi965: Rename backend_visitor::prog to shader_prog.
Paul Berry [Tue, 9 Apr 2013 00:17:44 +0000 (17:17 -0700)]
i965: Rename backend_visitor::prog to shader_prog.

The next patch is going to change the type of vec4_visitor::vp from
struct gl_vertex_program * to struct gl_program *, and rename it.  The
sensible name to change it to is vec4_visitor::prog.  However, prog is
already used in backend_visitor (which vec4_visitor derives from).
Since backend_visitor::prog is of type struct gl_shader_program *, it
makes sense to rename it to shader_prog.

Reviewed-by: Eric Anholt <eric@anholt.net>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
11 years agoglsl: Fix (and validate) comment above glsl_type::name.
Paul Berry [Wed, 10 Apr 2013 03:36:30 +0000 (20:36 -0700)]
glsl: Fix (and validate) comment above glsl_type::name.

The comment above glsl_type::name claimed that it could sometimes be
NULL.  This was wrong--it is never NULL.  Many error handling paths
would segfault if it were.  (Anonymous structs are assigned names like
"#anon_struct_0001"--see the ast_struct_specifier constructor in
glsl_parser_extras.cpp.)

Fix the comment and add assertions to validate that it really is never
NULL.

Reviewed-by: Ian Romanick <ian.d.romanick@intel.com>
11 years agoradeon/uvd: add UVD implementation v5
Christian König [Wed, 3 Apr 2013 08:18:35 +0000 (10:18 +0200)]
radeon/uvd: add UVD implementation v5

Just everything you need for UVD with r600g and radeonsi.

v2: move UVD code to radeon subdir, clean up build system additions,
    remove an unused SI function, disable tiling on SI for now.
v3: some minor indentation fix and rebased
v4: dpb size calculation fixed
v5: implement proper fall-back in case the kernel doesn't support UVD,
    based on patches from Andreas Boll but cleaned up a bit more.

Signed-off-by: Christian König <christian.koenig@amd.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
11 years agoradeon/winsys: add uvd ring support to winsys v3
Christian König [Mon, 8 Apr 2013 14:41:01 +0000 (16:41 +0200)]
radeon/winsys: add uvd ring support to winsys v3

Separated from UVD patch for clarity.

v2: sync with next tree for 3.10
v3: as pointed out by Andreas Bool check for drm minor >= 32

http://cgit.freedesktop.org/~agd5f/linux/log/?h=drm-next-3.10-wip

Signed-off-by: Christian König <christian.koenig@amd.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Reviewed-by: Andreas Boll <andreas.boll.dev@gmail.com>
11 years agost/mesa: fix UBO offsets.
Dave Airlie [Thu, 11 Apr 2013 05:20:19 +0000 (15:20 +1000)]
st/mesa: fix UBO offsets.

Reported and tested by degasus on #radeon.

Note: This is a candidate for the 9.1 branch

Signed-off-by: Dave Airlie <airlied@redhat.com>
11 years agoegl/x11: Fix initialisation of swap_interval
Ralf Jung [Tue, 9 Apr 2013 12:09:50 +0000 (14:09 +0200)]
egl/x11: Fix initialisation of swap_interval

The EGLConfig attributes EGL_MIN/MAX_SWAP_INTERVAL were incorrectly set to
0 and 0. This prevented clients from setting the swap interval to a
reasonable value, like 1 or 2.

Swap interval worked correctly in Mesa 9.0. The commit below introduced
the bug.

    commit 7e9bd2b2ed35a440a96362417100a7e43715d606
    Author: Eric Anholt <eric@anholt.net>
    Date:   Tue Sep 25 14:05:30 2012 -0700
egl: Add support for driconf control of swapinterval.

Note: This is a candidate for the 9.1 branch.
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=63078
[chadv: Wrote commit message]
Reviewed-by: Chad Versace <chad.versace@linux.intel.com>
Reviewed-by: Eric Anholt <eric@anholt.net>
11 years agointel: Fall back to X-tiling when larger than estimated aperture size.
Kenneth Graunke [Wed, 10 Apr 2013 20:49:16 +0000 (13:49 -0700)]
intel: Fall back to X-tiling when larger than estimated aperture size.

If a region is larger than the estimated aperture size, we map/unmap it
by copying with the BLT engine.  Which means we can't use Y-tiling.

Fixes Piglit max-texture-size and tex3d-maxsize, which regressed in my
recent change to use Y-tiling by default on Gen6+.  This was due to a
botched merge conflict resolution.

v2: Return a mask of valid tilings from intel_miptree_select_tiling.
    This allows us to avoid the X-tiling fallback if Y-tiling is actually
    mandatory.

Signed-off-by: Kenneth Graunke <kenneth@whitecape.org>
Reviewed-by: Chad Versace <chad.versace@linux.intel.com>
Reviewed-by: Eric Anholt <eric@anholt.net>
11 years agointel: Refactor code in intel_miptree_choose_tiling().
Kenneth Graunke [Wed, 10 Apr 2013 20:46:10 +0000 (13:46 -0700)]
intel: Refactor code in intel_miptree_choose_tiling().

This reduces the nesting level slightly, and in my opinion, makes it a
bit easier to follow.

Signed-off-by: Kenneth Graunke <kenneth@whitecape.org>
Reviewed-by: Chad Versace <chad.versace@linux.intel.com>
Reviewed-by: Eric Anholt <eric@anholt.net>
11 years agointel: Move the max_gtt_map_object_size estimation to intel_context.
Kenneth Graunke [Tue, 9 Apr 2013 18:11:19 +0000 (11:11 -0700)]
intel: Move the max_gtt_map_object_size estimation to intel_context.

We need know this in order to decide what tiling mode to use.

Signed-off-by: Kenneth Graunke <kenneth@whitecape.org>
Reviewed-by: Chad Versace <chad.versace@linux.intel.com>
Reviewed-by: Eric Anholt <eric@anholt.net>
11 years agor600g: Add support for GL_ARB_texture_buffer_range
Fredrik Höglund [Fri, 22 Mar 2013 16:14:43 +0000 (17:14 +0100)]
r600g: Add support for GL_ARB_texture_buffer_range

Reviewed-by: Marek Olšák <maraeo@gmail.com>
11 years agoi965/blorp: Remove unnecessary test in gen7_blorp_emit_depth_stencil_config.
Paul Berry [Tue, 9 Apr 2013 12:58:43 +0000 (05:58 -0700)]
i965/blorp: Remove unnecessary test in gen7_blorp_emit_depth_stencil_config.

gen7_blorp_emit_depth_stencil_config() is only called when
params->depth.mt is non-null.  Therefore, it's not necessary to do an
"if (params->depth.mt)" test inside it.  The presence of this if test
was misleading static analysis tools (and briefly, me) into thinking
that gen7_blorp_emit_depth_stencil_config() might sometimes access
uninitialized data and dereference a null pointer.

Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
11 years agor600g: fix valgrind warning on Cayman
Marek Olšák [Fri, 5 Apr 2013 12:22:42 +0000 (14:22 +0200)]
r600g: fix valgrind warning on Cayman

Warning: "Conditional jump or move depends on uninitialised value(s)".

11 years agogallivm/tgsi: handle untyped moves
Zack Rusin [Tue, 9 Apr 2013 12:11:45 +0000 (05:11 -0700)]
gallivm/tgsi: handle untyped moves

both mov and ucmp can be used to move variables of any type.
correctly note that about ucmp in the tgsi_info and make
sure gallivm can handle that by correctly casting the untyped
moves.

Signed-off-by: Zack Rusin <zackr@vmware.com>
Reviewed-by: Jose Fonseca <jfonseca@vmware.com>
Reviewed-by: Roland Scheidegger <sroland@vmware.com>
11 years agogallivm: fix loops and conditionals within GS
Zack Rusin [Tue, 9 Apr 2013 12:01:54 +0000 (05:01 -0700)]
gallivm: fix loops and conditionals within GS

We were using simple temporaries, without using alloca or phi
nodes which meant that on every iteration of the loop our
temporaries, which were holding the number of vertices and
primitives which were emitted, were being reset to zero. Now
we're using alloca to allocate those variables to preserve
them across conditionals.

Signed-off-by: Zack Rusin <zackr@vmware.com>
Reviewed-by: Jose Fonseca <jfonseca@vmware.com>
Reviewed-by: Roland Scheidegger <sroland@vmware.com>
11 years agollvmpipe: implement PIPE_QUERY_SO_STATISTICS
Zack Rusin [Tue, 9 Apr 2013 10:50:32 +0000 (03:50 -0700)]
llvmpipe: implement PIPE_QUERY_SO_STATISTICS

We were missing the implementation of PIPE_QUERY_SO_STATISTICS
query, this change implements it on top of the existing
facilities.

Signed-off-by: Zack Rusin <zackr@vmware.com>
Reviewed-by: Jose Fonseca <jfonseca@vmware.com>
Reviewed-by: Roland Scheidegger <sroland@vmware.com>
11 years agogallivm: fix unsigned divide and remainder opcodes
Zack Rusin [Tue, 9 Apr 2013 10:28:48 +0000 (03:28 -0700)]
gallivm: fix unsigned divide and remainder opcodes

We want to both make sure we never divide by zero to not generate
sigfpe and that divide by zero is guaranteed to return 0xffffffff.
Based on José idea.

Signed-off-by: Zack Rusin <zackr@vmware.com>
Reviewed-by: Jose Fonseca <jfonseca@vmware.com>
Reviewed-by: Roland Scheidegger <sroland@vmware.com>
11 years agogallivm: fix breakc
Zack Rusin [Thu, 4 Apr 2013 22:06:14 +0000 (15:06 -0700)]
gallivm: fix breakc

we break when the mask values are 0 not, 1, plus it's bit comparison
not a floating point comparison. This fixes both.

Signed-off-by: Zack Rusin <zackr@vmware.com>
Reviewed-by: Jose Fonseca <jfonseca@vmware.com>
Reviewed-by: Roland Scheidegger <sroland@vmware.com>
11 years agointel/hsw: Enable hiz (v2)
Chad Versace [Fri, 5 Apr 2013 23:35:47 +0000 (16:35 -0700)]
intel/hsw: Enable hiz (v2)

Enable hiz by setting intel_context::has_hiz.  However, to work around
a hardware bug, we selectively enable hiz for only nicely aligned miptree
slices.

No Piglit regressions on Haswell 0x0d26 rev07 when based atop
mesa-master-4ad3601.

Improves the performance of GLB27_TRex_C24Z16_FixedTimeStep by 18.52%
(hsw-0x0d26-rev07; kernel-3.9.0-rc1; GLBenchmark 2.7.0 Release a68901;
samples=3).

v2: Replace the check for IS_HASWELL(devid) in intel_miptree_slice_has_hiz()
    with a conditional set of has_hiz. [for anholt]

Reviewed-by: Eric Anholt <eric@anholt.net>
Signed-off-by: Chad Versace <chad.versace@linux.intel.com>
11 years agoi965: Remove brw_context::depthstencil::hiz_mt
Chad Versace [Fri, 5 Apr 2013 21:51:31 +0000 (14:51 -0700)]
i965: Remove brw_context::depthstencil::hiz_mt

After recent refactorings, the field is written but no longer read.

Reviewed-by: Eric Anholt <eric@anholt.net>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Signed-off-by: Chad Versace <chad.versace@linux.intel.com>
11 years agointel: Replace checks for hiz_mt with intel_has*hiz()
Chad Versace [Mon, 8 Apr 2013 20:45:49 +0000 (13:45 -0700)]
intel: Replace checks for hiz_mt with intel_has*hiz()

When appropriate, replace each check `hiz_mt != NULL` with either a call
to intel_miptree_slice_has_hiz() or intel_renderbuffer_has_hiz().  No
behavioral change.

This prepares for selectively enabling hiz on individual miptree slices
for Haswell.

This refactoring had several side effects.

  1. To prevent new warnings about discarding the const qualifier,
     I removed 'const' from some variable declarations in
     intel_validate_framebuffer().  The alternative was to add const
     qualifiers to multiple function signatures in the
     intel_renderbuffer_has_hiz call graph. Since the dominant convention
     in the Intel code is to not qualify function parameters as const,
     I chose to remove rather than add const qualifiers.

  2. I changed the signature of brw_emit_depth_stencil_hiz() by replacing
     `struct intel_mipmap_tree *hiz_mt` with `bool hiz`. The function used
     hiz_mt mostly as a boolean indicator of the presence of hiz, so the
     signature change is consistent with the patch's goal.

Reviewed-by: Eric Anholt <eric@anholt.net>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Signed-off-by: Chad Versace <chad.versace@linux.intel.com>
11 years agoi965: Change signature of brw_get_depthstencil_tile_masks()
Chad Versace [Fri, 5 Apr 2013 21:29:53 +0000 (14:29 -0700)]
i965: Change signature of brw_get_depthstencil_tile_masks()

Add new parameters `depth_level` and `depth_layer`, which specify depth
miptree's slice of interest.  A following patch will pass the new
parameters through to intel_miptree_slice_has_hiz().

Reviewed-by: Eric Anholt <eric@anholt.net>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Signed-off-by: Chad Versace <chad.versace@linux.intel.com>
11 years agoi965/blorp: Add fields brw_blorp_mip_info::level,layer
Chad Versace [Wed, 20 Mar 2013 00:44:50 +0000 (17:44 -0700)]
i965/blorp: Add fields brw_blorp_mip_info::level,layer

The new fields define the 2D miptree slice to be used. A following patch
will pass the new fields through to intel_miptree_slice_has_hiz().

Reviewed-by: Eric Anholt <eric@anholt.net>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Signed-off-by: Chad Versace <chad.versace@linux.intel.com>
11 years agointel: Add field intel_mipmap_slice::has_hiz
Chad Versace [Thu, 21 Feb 2013 03:18:40 +0000 (19:18 -0800)]
intel: Add field intel_mipmap_slice::has_hiz

On Haswell, HiZ will selectively be enabled on individual miptree slices
to workaround a hardware bug. The new field 'has_hiz' indicates if HiZ is
enabled for a given slice.

Also add two new accessor functions for this field.
  intel_miptree_slice_has_hiz
  intel_renderbuffer_has_hiz

The new field and accessor functions are not yet used. Also, this patch
introduces no behavioral change because, in this patch,
intel_miptree_alloc_hiz() sets has_hiz for all slices.

Reviewed-by: Eric Anholt <eric@anholt.net>
Signed-off-by: Chad Versace <chad.versace@linux.intel.com>
11 years agoi965/blorp: Align rectangle primitive for hiz ops
Chad Versace [Tue, 12 Mar 2013 02:21:46 +0000 (19:21 -0700)]
i965/blorp: Align rectangle primitive for hiz ops

The hardware docs and the simulator require that the rectangle primitive
emitted during fast depth clears and hiz resolves must be aligned to 8x4
pixels.

Reviewed-by: Eric Anholt <eric@anholt.net>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Signed-off-by: Chad Versace <chad.versace@linux.intel.com>
11 years agoi965/vs: Use GRFs for pull constant offsets on gen7.
Eric Anholt [Thu, 4 Apr 2013 21:10:18 +0000 (14:10 -0700)]
i965/vs: Use GRFs for pull constant offsets on gen7.

This allows the computation of the offset to get written directly into the
message source.

shader-db results:
total instructions in shared programs: 3308390 -> 3283025 (-0.77%)
instructions in affected programs:     442998 -> 417633 (-5.73%)

No difference in GLB2.7 low res (n=9).

Reviewed-by: Matt Turner <mattst88@gmail.com>
11 years agoi965/vs: When asked to make a dst_reg for a src.xxxx, just write to src.x.
Eric Anholt [Thu, 4 Apr 2013 23:30:49 +0000 (16:30 -0700)]
i965/vs: When asked to make a dst_reg for a src.xxxx, just write to src.x.

We have several places in our pull constant handling where we make a
temporary src_reg for an int, and then turn it into a dst.  In doing so,
we were writing to the dst.xyzw, so we never register coalesced it with a
later mov from dst.x to real_dst.x.

These extra channels written would be removed if we had channel-wise DCE
in the backend, but we don't.  Fix it for now by just not writing these
extra channels that won't get used.

Reviewed-by: Matt Turner <mattst88@gmail.com>
11 years agoi965/gen6: Reduce updates of transform feedback offsets with HW contexts.
Eric Anholt [Tue, 9 Apr 2013 20:34:08 +0000 (13:34 -0700)]
i965/gen6: Reduce updates of transform feedback offsets with HW contexts.

The software-tracked transform feedback offsets (svbi_0_starting_index)
are incorrect in the presence of primitive restart, so we were actually
updating it with a bogus value if the batch wrapped and we emitted the
packet again during a single transform feedback.  By reducing state
emission, we avoid the bug.

Fixes piglit OpenGL 3.1/primitive-restart-xfb flush
Reviewed-by: Paul Berry <stereotype441@gmail.com>
NOTE: This is a candidate for the 9.1 branch.

11 years agoi965/gen7: Skip resetting SOL offsets at batch start with HW contexts.
Eric Anholt [Mon, 8 Apr 2013 17:56:23 +0000 (10:56 -0700)]
i965/gen7: Skip resetting SOL offsets at batch start with HW contexts.

The software-tracked transform feedback offsets (svbi_0_starting_index)
are incorrect in the presence of primitive restart, so we can't reliably
compute offsets for our buffer pointers after a batch flush.  Thanks to HW
contexts, our transform feedback offsets are now saved, so we can just
keep using the ones from before the batch wrap.

Fixes piglit OpenGL 3.1/primitive-restart-xfb flush
Reviewed-by: Paul Berry <stereotype441@gmail.com>
NOTE: This is a candidate for the 9.1 branch.

11 years agoradeonsi: remove sampler writemask v3
Christian König [Tue, 26 Mar 2013 14:09:27 +0000 (15:09 +0100)]
radeonsi: remove sampler writemask v3

v2: fix instrinsic name as well
v3: LLVM revision incremented as well

Signed-off-by: Christian König <christian.koenig@amd.com>
Reviewed-by: Michel Dänzer <michel.daenzer@amd.com>
11 years agopipe-loader: Fix out of source build
Niels Ole Salscheider [Sun, 24 Feb 2013 22:00:04 +0000 (23:00 +0100)]
pipe-loader: Fix out of source build

Signed-off-by: Niels Ole Salscheider <niels_ole@salscheider-online.de>