mesa.git
8 years agoi965: enable ARB_framebuffer_no_attachments for Gen7+
Kevin Rogovin [Wed, 17 Jun 2015 10:29:58 +0000 (13:29 +0300)]
i965: enable ARB_framebuffer_no_attachments for Gen7+

Enable GL_ARB_framebuffer_no_attachments in i965 for Gen7 and higher.

Reviewed-by: Ian Romanick <ian.d.romanick@intel.com>
Signed-off-by: Kevin Rogovin <kevin.rogovin@intel.com>
8 years agoi965: execution of frag-shader when it has atomic buffer
Kevin Rogovin [Wed, 17 Jun 2015 10:29:57 +0000 (13:29 +0300)]
i965: execution of frag-shader when it has atomic buffer

Ensure that the GPU spawns the fragment shader thread for those
fragment shaders with atomic buffer access.

Reviewed-by: Tapani Pälli <tapani.palli@intel.com>
Signed-off-by: Kevin Rogovin <kevin.rogovin@intel.com>
8 years agomesa: function for testing if current frag-shader has atomics
Kevin Rogovin [Wed, 17 Jun 2015 10:29:56 +0000 (13:29 +0300)]
mesa: function for testing if current frag-shader has atomics

Add helper function that checks if current fragment shader active
of gl_context has atomic buffer access.

Reviewed-by: Ian Romanick <ian.d.romanick@intel.com>
Signed-off-by: Kevin Rogovin <kevin.rogovin@intel.com>
8 years agoi965: Use _mesa_geometric_ functions appropriately
Kevin Rogovin [Wed, 17 Jun 2015 10:29:55 +0000 (13:29 +0300)]
i965: Use _mesa_geometric_ functions appropriately

Change references to gl_framebuffer::Width, Height, MaxNumLayers
and Visual::samples to use the _mesa_geometry_ convenience functions
for those places where the geometry of the gl_framebuffer is needed
(in contrast to the geometry of the intersection of the attachments
of the gl_framebuffer).

This patch is to pave the way to enable GL_ARB_framebuffer_no_attachments
on Gen7 and higher in i965.

Reviewed-by: Ian Romanick <ian.d.romanick@intel.com>
Signed-off-by: Kevin Rogovin <kevin.rogovin@intel.com>
8 years agomesa: helper function for scissor box of gl_framebuffer
Kevin Rogovin [Wed, 17 Jun 2015 10:29:54 +0000 (13:29 +0300)]
mesa: helper function for scissor box of gl_framebuffer

Add helper convenience function that intersects the scissor values
against a passed bounding box. In addition, to avoid replicated code,
make the function _mesa_scissor_bounding_box() use this new function.

Reviewed-by: Ian Romanick <ian.d.romanick@intel.com>
Signed-off-by: Kevin Rogovin <kevin.rogovin@intel.com>
8 years agomesa: add helper functions for geometry of gl_framebuffer
Kevin Rogovin [Wed, 17 Jun 2015 10:29:53 +0000 (13:29 +0300)]
mesa: add helper functions for geometry of gl_framebuffer

Add convenience helper functions for fetching geometry of gl_framebuffer
that return the geometry of the gl_framebuffer instead of the geometry of
the buffers of the gl_framebuffer when then the gl_framebuffer has no
attachments.

Reviewed-by: Ian Romanick <ian.d.romanick@intel.com>
Signed-off-by: Kevin Rogovin <kevin.rogovin@intel.com>
8 years agoPATCH 03/10] mesa: Complete ARB_framebuffer_no_attachments in Mesa core
Kevin Rogovin [Wed, 17 Jun 2015 10:29:52 +0000 (13:29 +0300)]
PATCH 03/10] mesa: Complete ARB_framebuffer_no_attachments in Mesa core

Implement GL_ARB_framebuffer_no_attachments in Mesa core
 - changes to conditions for framebuffer completenss
 - implement set/get functions for framebuffers for
   new functions in GL_ARB_framebuffer_no_attachments

Reviewed-by: Ian Romanick <ian.d.romanick@intel.com>
Signed-off-by: Kevin Rogovin <kevin.rogovin@intel.com>
8 years agomesa: Constants and functions for ARB_framebuffer_no_attachments
Kevin Rogovin [Wed, 17 Jun 2015 10:29:51 +0000 (13:29 +0300)]
mesa: Constants and functions for ARB_framebuffer_no_attachments

Define the enumeration constants, function entry points and
glGet for the GL_ARB_framebuffer_no_attachments.

Reviewed-by: Ian Romanick <ian.d.romanick@intel.com>
Signed-off-by: Kevin Rogovin <kevin.rogovin@intel.com>
8 years agomesa: Define infrastructure for ARB_framebuffer_no_attachments
Kevin Rogovin [Wed, 17 Jun 2015 10:29:50 +0000 (13:29 +0300)]
mesa: Define infrastructure for ARB_framebuffer_no_attachments

Define the infrastructure for the extension GL_ARB_framebuffer_no_attachments:
 - extension table
 - additions to gl_framebuffer

Reviewed-by: Ian Romanick <ian.d.romanick@intel.com>
Signed-off-by: Kevin Rogovin <kevin.rogovin@intel.com>
8 years agovc4: Make sure that direct texture clamps have a minimum value of 0.
Eric Anholt [Mon, 15 Jun 2015 22:05:36 +0000 (15:05 -0700)]
vc4: Make sure that direct texture clamps have a minimum value of 0.

I was thinking of the MIN opcode in terms of unsigned math, but it's
signed, so if you used a negative array index, you could read before the
UBO.  Fixes segfaults under simulation in piglit array indexing tests with
mprotect-based guard pages.

8 years agovc4: Swap around which src we spill to ra31/rb31.
Eric Anholt [Tue, 16 Jun 2015 00:47:12 +0000 (17:47 -0700)]
vc4: Swap around which src we spill to ra31/rb31.

I wanted to assert that src1 came from a non-unspilled register in shader
validation, and this easily gets us that.  And, as a bonus:

total instructions in shared programs: 93347 -> 92723 (-0.67%)
instructions in affected programs:     60524 -> 59900 (-1.03%)

8 years agovc4: R4 is not a valid register for clamped direct texturing.
Eric Anholt [Tue, 16 Jun 2015 19:03:10 +0000 (12:03 -0700)]
vc4: R4 is not a valid register for clamped direct texturing.

Our array only goes to R3, and R4 is a special case that shouldn't be
used.

8 years agovc4: Factor out the live clamp register getter.
Eric Anholt [Mon, 15 Jun 2015 21:54:26 +0000 (14:54 -0700)]
vc4: Factor out the live clamp register getter.

8 years agovc4: Drop the unused "stride" field of surfaces.
Eric Anholt [Mon, 15 Jun 2015 18:41:06 +0000 (11:41 -0700)]
vc4: Drop the unused "stride" field of surfaces.

We're always looking at the slice anyway, when we would have needed it.

8 years agovc4: Handle refcounting the exec BO like we do in the kernel.
Eric Anholt [Fri, 12 Jun 2015 19:47:47 +0000 (12:47 -0700)]
vc4: Handle refcounting the exec BO like we do in the kernel.

This reduces the diff to the kernel, and will be useful when I make the
kernel allocate more BOs as part of validation.

8 years agovc4: Use VC4_SET/GET_FIELD for some RCL packets.
Eric Anholt [Thu, 11 Jun 2015 23:08:11 +0000 (16:08 -0700)]
vc4: Use VC4_SET/GET_FIELD for some RCL packets.

8 years agovc4: Make symbolic values for packet sizes.
Eric Anholt [Wed, 10 Jun 2015 20:20:25 +0000 (13:20 -0700)]
vc4: Make symbolic values for packet sizes.

8 years agovc4: Use symbolic values in texture ptype validation.
Eric Anholt [Wed, 10 Jun 2015 19:58:47 +0000 (12:58 -0700)]
vc4: Use symbolic values in texture ptype validation.

8 years agovc4: Move vc4_packet.h to the kernel/ directory, since it's also shared.
Eric Anholt [Wed, 10 Jun 2015 19:47:56 +0000 (12:47 -0700)]
vc4: Move vc4_packet.h to the kernel/ directory, since it's also shared.

I want to notice discrepancies when I diff -u between Mesa and the kernel.

8 years agoi965/gen9: Disable Mip Tail for YF/YS tiled surfaces
Anuj Phogat [Wed, 15 Apr 2015 05:06:50 +0000 (22:06 -0700)]
i965/gen9: Disable Mip Tail for YF/YS tiled surfaces

Disabling miptails fixed the buffer corruption happening in FBO
which use YF/YS tiled renderbuffer or texture as color attachment.

Spec recommends disabling mip tails only for non-mip-mapped surfaces.
But, without disabling miptails I couldn't get correct data out of
mipmapped YF/YS tiled surface.

We need better understanding of miptails before start using them.
For now this patch helps move things forward.

Signed-off-by: Anuj Phogat <anuj.phogat@gmail.com>
Reviewed-by: Ben Widawsky <ben@bwidawsk.net>
8 years agoi965/gen9: Set vertical and horizontal surface alignments
Anuj Phogat [Wed, 15 Apr 2015 05:06:48 +0000 (22:06 -0700)]
i965/gen9: Set vertical and horizontal surface alignments

Patch sets the alignments for texture and renderbuffer surfaces.

V3: Make changes inside horizontal_alignment() and
    vertical_alignment() (Topi)

Signed-off-by: Anuj Phogat <anuj.phogat@gmail.com>
Reviewed-by: Topi Pohjolainen <topi.pohjolainen@intel.com>
8 years agoi965: Use BRW_SURFACE_* in place of GL_TEXTURE_*
Anuj Phogat [Wed, 15 Apr 2015 05:06:48 +0000 (22:06 -0700)]
i965: Use BRW_SURFACE_* in place of GL_TEXTURE_*

Makes no functional changes in the code.

Signed-off-by: Anuj Phogat <anuj.phogat@gmail.com>
Reviewed-by: Chris Forbes <chrisf@ijw.co.nz>
Reviewed-by: Topi Pohjolainen <topi.pohjolainen@intel.com>
8 years agoi965: Rename use_linear_1d_layout() and make it global
Anuj Phogat [Wed, 15 Apr 2015 05:06:48 +0000 (22:06 -0700)]
i965: Rename use_linear_1d_layout() and make it global

This function will be utilised in later patches.

V2: Make both pointers constants (Topi)

Signed-off-by: Anuj Phogat <anuj.phogat@gmail.com>
Reviewed-by: Topi Pohjolainen <topi.pohjolainen@intel.com>
8 years agoi965/gen9: Set tiled resource mode in surface state
Anuj Phogat [Wed, 15 Apr 2015 05:06:48 +0000 (22:06 -0700)]
i965/gen9: Set tiled resource mode in surface state

This patch sets the tiled resource mode for texture and renderbuffer
surfaces.

Signed-off-by: Anuj Phogat <anuj.phogat@gmail.com>
Reviewed-by: Topi Pohjolainen <topi.pohjolainen@intel.com>
8 years agoegl/dri2: implement platform_surfaceless
Haixia Shi [Fri, 12 Jun 2015 17:10:58 +0000 (10:10 -0700)]
egl/dri2: implement platform_surfaceless

The surfaceless platform is for off-screen rendering only. Render node support
is required.

Only consider the render nodes. Do not use normal nodes as they require
auth hooks.

v3: change platform_null to platform_surfaceless
v4: make libdrm required for surfaceless
v5: remove modified include guards with defined(HAVE_SURFACELESS_PLATFORM)
v6: use O_CLOEXEC for drm fd

Signed-off-by: Haixia Shi <hshi@chromium.org>
Signed-off-by: Zach Reizner <zachr@google.com>
Reviewed-by: Emil Velikov <emil.l.velikov@gmail.com>
Reviewed-by: Chad Versace <chad.versace@intel.com>
8 years agoi965/vec4: Fix the source register for indexed samplers
Neil Roberts [Thu, 28 May 2015 18:35:44 +0000 (19:35 +0100)]
i965/vec4: Fix the source register for indexed samplers

Previously when setting up the sample instruction for an indirect
sampler the vec4 backend was directly passing the pseudo opcode's
src0. However vec4_visitor::visit(ir_texture *) doesn't set the
texture operation's src0 -- it's left as BAD_FILE, which when
translated into a brw_reg gives the null register. In brw_SAMPLE,
gen6_resolve_implied_move() inserts a MOV from the inst->base_mrf and
sets the src0 appropriately. The indirect sampler case did not have a
call to gen6_resolve_implied_move().

The fs backend avoids this because the platforms that support dynamic
indexing of samplers (IVB+) have been converted to not use the
fake-MRF hack, and instead send from proper GRFs.

This patch makes it call gen6_resolve_implied_move before setting up
the indirect message. This is similar to what is done for constant
sampler numbers in brw_SAMPLE.

The Piglit tests for sampler array indexing didn't pick this up
because they were using a texture with a solid colour so it didn't
matter what texture coordinates were actually used. The tests have now
been changed to be more thorough in this commit:

http://cgit.freedesktop.org/piglit/commit/?id=4f9caf084eda7

With that patch the tests for gs and vs are currently failing on
Ivybridge, but this patch fixes them. There are no other changes to a
Piglit run on Ivybridge.

On Skylake the gs tests were failing even without the Piglit patch
because Skylake needs the source registers to work correctly in order
to send a message header to select SIMD4x2 mode.

(The explanation in the commit message is partially written by Matt
Turner)

Tested-by: Anuj Phogat <anuj.phogat@gmail.com>
Reviewed-by: Matt Turner <mattst88@gmail.com>
8 years agost/mesa: improve assertions in vp/fp translation
Marek Olšák [Mon, 6 Apr 2015 23:10:17 +0000 (01:10 +0200)]
st/mesa: improve assertions in vp/fp translation

Reviewed-by: Brian Paul <brianp@vmware.com>
8 years agomesa: don't rebind constant buffers after every state change if GS is active
Marek Olšák [Sun, 14 Jun 2015 14:37:02 +0000 (16:37 +0200)]
mesa: don't rebind constant buffers after every state change if GS is active

Reviewed-by: Brian Paul <brianp@vmware.com>
8 years agomesa: generalize sso stage interleaving check
Chris Forbes [Sun, 21 Sep 2014 00:07:55 +0000 (12:07 +1200)]
mesa: generalize sso stage interleaving check

For tessellation.

v2: cleanup by Marek Olšák

Reviewed-by: Brian Paul <brianp@vmware.com>
8 years agomesa: remove unused variables from gl_program
Marek Olšák [Sat, 13 Jun 2015 11:02:20 +0000 (13:02 +0200)]
mesa: remove unused variables from gl_program

Reviewed-by: Brian Paul <brianp@vmware.com>
8 years agoglsl: add ir reader support for ir_barrier
Chris Forbes [Sun, 7 Sep 2014 07:24:15 +0000 (19:24 +1200)]
glsl: add ir reader support for ir_barrier

Picked from the tessellation branch.

Reviewed-by: Brian Paul <brianp@vmware.com>
8 years agoglsl: print locations of variables
Marek Olšák [Thu, 19 Mar 2015 22:28:25 +0000 (23:28 +0100)]
glsl: print locations of variables

Reviewed-by: Brian Paul <brianp@vmware.com>
8 years agoconfigure.ac: rename LLVM_VERSION_PATCH to avoid conflict with llvm-config.h
Marek Olšák [Sat, 6 Jun 2015 11:24:11 +0000 (13:24 +0200)]
configure.ac: rename LLVM_VERSION_PATCH to avoid conflict with llvm-config.h

Reviewed-by: Tom Stellard <thomas.stellard@amd.com>
8 years agoRevert "glsl: remove restriction on unsized arrays in GLSL ES 3.10"
Timothy Arceri [Mon, 15 Jun 2015 11:00:47 +0000 (21:00 +1000)]
Revert "glsl: remove restriction on unsized arrays in GLSL ES 3.10"

This reverts commit adee54f8269c5e9f4fde91d19f0e465afc8f14d8.

Further down in the GLSL ES 3.10 spec it say:

"If an array is declared as the last member of a shader storage block
and the size is not specified at compile-time, it is sized at run-time.
In all other cases, arrays are sized only at compile-time."

Reviewed-by: Samuel Iglesias Gonsálvez <siglesias@igalia.com>
8 years agomesa: set override_version per api version override
Tapani Pälli [Tue, 16 Jun 2015 10:46:47 +0000 (13:46 +0300)]
mesa: set override_version per api version override

Before 9b5e92f get_gl_override was called only once, but now it is
called for multiple APIs (GLES2, GL), version needs to be set always.

Signed-off-by: Tapani Pälli <tapani.palli@intel.com>
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=90797
Reviewed-by: Jordan Justen <jordan.l.justen@intel.com>
Reviewed-by: Martin Peres <martin.peres@linux.intel.com>
Tested-by: Martin Peres <martin.peres@linux.intel.com>
8 years agoi965: Fix aligning to the block size in intel_miptree_copy_slice
Neil Roberts [Thu, 11 Jun 2015 15:59:07 +0000 (16:59 +0100)]
i965: Fix aligning to the block size in intel_miptree_copy_slice

This function was trying to align the width and height to a multiple
of the block size for compressed textures. It was using align_w/h as a
shortcut to get the block size as up until Gen9 this always happens to
match. However in Gen9+ the alignment values are expressed as
multiples of the block size so in effect the alignment values are
always 4 for compressed textures as that is the minimum value we can
pick. This happened to work for most compressed formats because the
block size is also 4, but for FXT1 this was breaking because it has a
block width of 8.

This fixes some Piglit tests testing FXT1 such as

spec@3dfx_texture_compression_fxt1@fbo-generatemipmap-formats

Reviewed-by: Nanley Chery <nanley.g.chery@intel.com>
8 years agonv50,nvc0: clamp uniform size to 64k
Ilia Mirkin [Mon, 15 Jun 2015 19:48:58 +0000 (15:48 -0400)]
nv50,nvc0: clamp uniform size to 64k

The state tracker will pass through requests from buggy applications
which will have the buffer size larger than the max allowed (64k). Clamp
the size to 64k so that we don't get errors when uploading the constbuf
data.

Signed-off-by: Ilia Mirkin <imirkin@alum.mit.edu>
Cc: "10.5 10.6" <mesa-stable@lists.freedesktop.org>
8 years agonvc0/ir: fix collection of first uses for texture barrier insertion
Ilia Mirkin [Fri, 12 Jun 2015 14:09:05 +0000 (16:09 +0200)]
nvc0/ir: fix collection of first uses for texture barrier insertion

One of the places we have to insert texbars is in situations where the
result of the tex gets overwritten by a different instruction (e.g. in a
conditional statement). However in some situations it can actually
appear as though the original tex itself is an overwriting instruction.
This can naturally never really happen, so just ignore the tex
instruction when it comes up.

Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=90347
Signed-off-by: Ilia Mirkin <imirkin@alum.mit.edu>
Cc: "10.5 10.6" <mesa-stable@lists.freedesktop.org>
8 years agoegl: Drop check for driver != NULL.
Eric Anholt [Tue, 9 Jun 2015 19:16:19 +0000 (12:16 -0700)]
egl: Drop check for driver != NULL.

Back in 2013, a patch was added (with 2 reviewers!) at the end of the
block to early exit the loop in this case, without noticing that the loop
already did.  I added another early exit case, again without noticing, but
Rob caught me.  Just drop the loop condition that apparently surprises
most of us, instead of leaving the end of the loop conspicuously not
exiting on success.

Reviewed-by: Ian Romanick <ian.d.romanick@intel.com>
Reviewed-by: Rob Clark <robdclark@gmail.com>
8 years agogallium: Drop the gallium-specific Android sw winsys.
Eric Anholt [Tue, 9 Jun 2015 18:45:05 +0000 (11:45 -0700)]
gallium: Drop the gallium-specific Android sw winsys.

This was part of gallium_egl, and we now have the normal libEGL Android
winsys support to handle it.

Reviewed-by: Emil Velikov <emil.l.velikov@gmail.com>
8 years agovc4: Add support for building on Android.
Eric Anholt [Wed, 3 Jun 2015 17:15:31 +0000 (10:15 -0700)]
vc4: Add support for building on Android.

v2: Add a comment explaining why we link libmesa_glsl.  Drop warning
    option from freedreno.  Add vc4 to the documentation for
    BOARD_GPU_DRIVERS.

Reviewed-by: Emil Velikov <emil.l.velikov@gmail.com>
8 years agogallium: Enable build of NIR support on Android.
Eric Anholt [Sun, 7 Jun 2015 18:57:46 +0000 (11:57 -0700)]
gallium: Enable build of NIR support on Android.

v2: Add a comment explaining why we link libmesa_glsl.

Reviewed-by: Emil Velikov <emil.l.velikov@gmail.com>
8 years agoegl/dri2: Fix Android Lollipop build on ARM.
Eric Anholt [Sun, 7 Jun 2015 23:47:25 +0000 (16:47 -0700)]
egl/dri2: Fix Android Lollipop build on ARM.

Reviewed-by: Emil Velikov <emil.l.velikov@gmail.com>
8 years agometa: Abort texture upload if pixels == null and no pixel unpack buffer set
Anuj Phogat [Fri, 15 May 2015 13:01:15 +0000 (06:01 -0700)]
meta: Abort texture upload if pixels == null and no pixel unpack buffer set

in case of glTexImage{1,2,3}D(). Texture has already been allocated
at this point and we have no data to upload. With out this patch,
with create_pbo = true, we end up creating a temporary pbo and then
uploading uninitialzed texture data.

Signed-off-by: Anuj Phogat <anuj.phogat@gmail.com>
Reviewed-by: Neil Roberts <neil@linux.intel.com>
8 years agometa: Abort meta path if ReadPixels need rgb to luminance conversion
Anuj Phogat [Tue, 12 May 2015 11:17:04 +0000 (04:17 -0700)]
meta: Abort meta path if ReadPixels need rgb to luminance conversion

After recent addition of pbo testing in piglit test getteximage-luminance,
it fails on i965. This patch makes a sub test pass.

Signed-off-by: Anuj Phogat <anuj.phogat@gmail.com>
Cc: <mesa-stable@lists.freedesktop.org>
Reviewed-by: Tapani Pälli <tapani.palli@intel.com>
8 years agomesa: Turn need_rgb_to_luminance_conversion() in to a global function
Anuj Phogat [Fri, 1 May 2015 07:05:18 +0000 (00:05 -0700)]
mesa: Turn need_rgb_to_luminance_conversion() in to a global function

This will be used by _mesa_meta_pbo_GetTexSubImage() in a later patch.

Signed-off-by: Anuj Phogat <anuj.phogat@gmail.com>
Cc: <mesa-stable@lists.freedesktop.org>
Reviewed-by: Tapani Pälli <tapani.palli@intel.com>
8 years agomesa: Use helper function need_rgb_to_luminance_conversion()
Anuj Phogat [Fri, 1 May 2015 06:36:18 +0000 (23:36 -0700)]
mesa: Use helper function need_rgb_to_luminance_conversion()

Signed-off-by: Anuj Phogat <anuj.phogat@gmail.com>
Cc: <mesa-stable@lists.freedesktop.org>
Reviewed-by: Tapani Pälli <tapani.palli@intel.com>
Reviewed-by: Iago Toral Quiroga <itoral@igalia.com>
8 years agomesa: Handle integer formats in need_rgb_to_luminance_conversion()
Anuj Phogat [Fri, 1 May 2015 06:35:20 +0000 (23:35 -0700)]
mesa: Handle integer formats in need_rgb_to_luminance_conversion()

Signed-off-by: Anuj Phogat <anuj.phogat@gmail.com>
Cc: <mesa-stable@lists.freedesktop.org>
Reviewed-by: Tapani Pälli <tapani.palli@intel.com>
Reviewed-by: Iago Toral Quiroga <itoral@igalia.com>
8 years agometa: Use is_power_of_two() helper function
Anuj Phogat [Mon, 1 Jun 2015 16:32:55 +0000 (09:32 -0700)]
meta: Use is_power_of_two() helper function

Signed-off-by: Anuj Phogat <anuj.phogat@gmail.com>
Reviewed-by: Tapani Pälli <tapani.palli@intel.com>
8 years agoi965: Check for miptree pitch alignment before using intel_miptree_map_movntdqa()
Anuj Phogat [Tue, 5 May 2015 06:10:28 +0000 (23:10 -0700)]
i965: Check for miptree pitch alignment before using intel_miptree_map_movntdqa()

We have an assert() in intel_miptree_map_movntdqa() which expects
the pitch to be 16 byte aligned.

Signed-off-by: Anuj Phogat <anuj.phogat@gmail.com>
Reviewed-by: Tapani Pälli <tapani.palli@intel.com>
8 years agoi965: Remove break after return
Anuj Phogat [Thu, 28 May 2015 21:48:51 +0000 (14:48 -0700)]
i965: Remove break after return

Signed-off-by: Anuj Phogat <anuj.phogat@gmail.com>
Reviewed-by: Tapani Pälli <tapani.palli@intel.com>
8 years agonv50/ir: OP_JOIN is a flow instruction
Jürgen Rühle [Sat, 6 Jun 2015 16:37:20 +0000 (18:37 +0200)]
nv50/ir: OP_JOIN is a flow instruction

OP_JOIN instructions are assumed to be flow instructions and mercilessly
casted to FlowInstruction.

This patch fixes an instance where an OP_JOIN is created as a plain
instruction. This can cause crashes in the ir printer.

[imirkin: add ->fixed = 1]
Reviewed-by: Ilia Mirkin <imirkin@alum.mit.edu>
8 years agodocs: add news item and link release notes for mesa 10.6.0
Emil Velikov [Sun, 14 Jun 2015 15:43:21 +0000 (16:43 +0100)]
docs: add news item and link release notes for mesa 10.6.0

Signed-off-by: Emil Velikov <emil.l.velikov@gmail.com>
8 years agodocs: Add sha256sums for the 10.6.0 release
Emil Velikov [Sun, 14 Jun 2015 15:40:00 +0000 (16:40 +0100)]
docs: Add sha256sums for the 10.6.0 release

Signed-off-by: Emil Velikov <emil.l.velikov@gmail.com>
(cherry picked from commit 5d327b373531861f86a726db669b3d656f1b5f8d)

8 years agodocs: Update 10.6.0 release notes
Emil Velikov [Sun, 14 Jun 2015 15:26:40 +0000 (16:26 +0100)]
docs: Update 10.6.0 release notes

Signed-off-by: Emil Velikov <emil.l.velikov@gmail.com>
(cherry picked from commit 3b9cde5c8138fb5cc45c652f2a5c15c5fa222bd7)

8 years agoilo: add ilo_state_raster_{line,poly}_stipple
Chia-I Wu [Mon, 15 Jun 2015 03:24:47 +0000 (11:24 +0800)]
ilo: add ilo_state_raster_{line,poly}_stipple

Initialize hardware stipple states on bound instead of on emission.

8 years agoilo: add ilo_state_sample_pattern
Chia-I Wu [Mon, 15 Jun 2015 04:01:29 +0000 (12:01 +0800)]
ilo: add ilo_state_sample_pattern

Move sample pattern initialization from ilo_render to
ilo_state_sample_pattern.

8 years agoilo: add 3DSTATE_AA_LINE_PARAMETERS to ilo_state_raster
Chia-I Wu [Mon, 15 Jun 2015 03:57:10 +0000 (11:57 +0800)]
ilo: add 3DSTATE_AA_LINE_PARAMETERS to ilo_state_raster

Utilize ilo_state_raster to avoid redundant state change.

8 years agogallium/util: add util_last_bit64
Marek Olšák [Sun, 10 May 2015 18:35:15 +0000 (20:35 +0200)]
gallium/util: add util_last_bit64

This will be needed by radeonsi.

Reviewed-by: Ilia Mirkin <imirkin@alum.mit.edu>
8 years agoglsl: fix "tesselation" typo
Marek Olšák [Sat, 6 Jun 2015 12:12:34 +0000 (14:12 +0200)]
glsl: fix "tesselation" typo

Trivial.

8 years agor600g: handle TGSI input/output array declarations correctly
Marek Olšák [Fri, 5 Jun 2015 17:09:21 +0000 (19:09 +0200)]
r600g: handle TGSI input/output array declarations correctly

Most of this code could be removed if r600g used tgsi_shader_info.

8 years agoilo: merge ilo_state_3d*.[ch] to ilo_state.[ch]
Chia-I Wu [Wed, 10 Jun 2015 23:36:28 +0000 (07:36 +0800)]
ilo: merge ilo_state_3d*.[ch] to ilo_state.[ch]

With most code replaced to ilo_state_*, what was left did not belong there
anymore.

8 years agoilo: add ilo_state_ps to ilo_shader_cso
Chia-I Wu [Fri, 12 Jun 2015 07:08:02 +0000 (15:08 +0800)]
ilo: add ilo_state_ps to ilo_shader_cso

8 years agoilo: add ilo_state_{vs,hs,ds,gs} to ilo_shader_cso
Chia-I Wu [Fri, 12 Jun 2015 06:47:02 +0000 (14:47 +0800)]
ilo: add ilo_state_{vs,hs,ds,gs} to ilo_shader_cso

8 years agoilo: embed ilo_state_sbe in ilo_shader
Chia-I Wu [Tue, 2 Jun 2015 15:09:53 +0000 (23:09 +0800)]
ilo: embed ilo_state_sbe in ilo_shader

8 years agoilo: embed ilo_state_vf in ilo_ve_state
Chia-I Wu [Tue, 2 Jun 2015 06:57:48 +0000 (14:57 +0800)]
ilo: embed ilo_state_vf in ilo_ve_state

8 years agoilo: embed ilo_state_urb in ilo_state_vector
Chia-I Wu [Thu, 28 May 2015 05:43:56 +0000 (13:43 +0800)]
ilo: embed ilo_state_urb in ilo_state_vector

8 years agoilo: embed ilo_state_sol in ilo_shader
Chia-I Wu [Fri, 29 May 2015 07:25:13 +0000 (15:25 +0800)]
ilo: embed ilo_state_sol in ilo_shader

8 years agoilo: embed ilo_state_cc in ilo_blend_state
Chia-I Wu [Mon, 11 May 2015 11:48:52 +0000 (19:48 +0800)]
ilo: embed ilo_state_cc in ilo_blend_state

8 years agoilo: embed ilo_state_raster in ilo_rasterizer_state
Chia-I Wu [Fri, 5 Jun 2015 02:23:24 +0000 (10:23 +0800)]
ilo: embed ilo_state_raster in ilo_rasterizer_state

8 years agoilo: embed ilo_state_viewport in ilo_viewport_state
Chia-I Wu [Sun, 17 May 2015 16:00:37 +0000 (00:00 +0800)]
ilo: embed ilo_state_viewport in ilo_viewport_state

8 years agoilo: replace ilo_sampler_cso with ilo_state_sampler
Chia-I Wu [Thu, 21 May 2015 09:18:37 +0000 (17:18 +0800)]
ilo: replace ilo_sampler_cso with ilo_state_sampler

8 years agoilo: replace ilo_view_surface with ilo_state_surface
Chia-I Wu [Wed, 20 May 2015 13:44:30 +0000 (21:44 +0800)]
ilo: replace ilo_view_surface with ilo_state_surface

8 years agoilo: replace ilo_zs_surface with ilo_state_zs
Chia-I Wu [Mon, 18 May 2015 15:32:10 +0000 (23:32 +0800)]
ilo: replace ilo_zs_surface with ilo_state_zs

8 years agoilo: add ilo_state_ps
Chia-I Wu [Fri, 12 Jun 2015 06:56:56 +0000 (14:56 +0800)]
ilo: add ilo_state_ps

We want to make ilo_shader_cso a union of ilo_state_{vs,hs,ds,gs,ps}.

8 years agoilo: add ilo_state_{vs,hs,ds,gs}
Chia-I Wu [Fri, 29 May 2015 16:58:51 +0000 (00:58 +0800)]
ilo: add ilo_state_{vs,hs,ds,gs}

We want to make ilo_shader_cso a union of ilo_state_{vs,hs,ds,gs} and ps
payload.

8 years agoilo: add ilo_state_sbe
Chia-I Wu [Fri, 12 Jun 2015 06:02:37 +0000 (14:02 +0800)]
ilo: add ilo_state_sbe

We want to replace ilo_kernel_routing with ilo_state_sbe.

8 years agoilo: add ilo_state_vf
Chia-I Wu [Sat, 30 May 2015 16:00:49 +0000 (00:00 +0800)]
ilo: add ilo_state_vf

We want to replace ilo_ve_state with ilo_state_vf.

8 years agoilo: add ilo_state_urb
Chia-I Wu [Thu, 28 May 2015 05:21:02 +0000 (13:21 +0800)]
ilo: add ilo_state_urb

8 years agoilo: add ilo_state_sol
Chia-I Wu [Fri, 29 May 2015 05:08:18 +0000 (13:08 +0800)]
ilo: add ilo_state_sol

8 years agoilo: add ilo_state_cc
Chia-I Wu [Mon, 11 May 2015 06:23:49 +0000 (14:23 +0800)]
ilo: add ilo_state_cc

We want to replace ilo_dsa_state and ilo_blend_state with ilo_state_cc.

8 years agoilo: add ilo_state_raster
Chia-I Wu [Sun, 10 May 2015 05:52:21 +0000 (13:52 +0800)]
ilo: add ilo_state_raster

We want to replace ilo_rasterizer_state with ilo_state_raster.

8 years agoilo: add ilo_state_viewport
Chia-I Wu [Tue, 12 May 2015 15:43:50 +0000 (23:43 +0800)]
ilo: add ilo_state_viewport

We want to replace ilo_viewport_cso and ilo_scissor_state with
ilo_state_viewport.

8 years agoilo: add ilo_state_sampler
Chia-I Wu [Wed, 13 May 2015 05:10:54 +0000 (13:10 +0800)]
ilo: add ilo_state_sampler

We want to replace ilo_sampler_cso with ilo_state_sampler.

8 years agoilo: add ilo_state_surface
Chia-I Wu [Thu, 14 May 2015 01:46:42 +0000 (09:46 +0800)]
ilo: add ilo_state_surface

We want to replace ilo_view_surface with ilo_state_surface.

8 years agoilo: add ilo_state_zs
Chia-I Wu [Sat, 16 May 2015 00:27:24 +0000 (08:27 +0800)]
ilo: add ilo_state_zs

We want to replace ilo_zs_surface with ilo_state_zs.  One noteworthy
difference is that ilo_state_zs always aligns level 0 to 8x4 when HiZ is
enabled.  HiZ will not be enabled for 1D surfaces as a result.

8 years agoilo: update genhw headers
Chia-I Wu [Sat, 9 May 2015 13:39:34 +0000 (21:39 +0800)]
ilo: update genhw headers

Generate these new enums

  enum gen_reorder_mode;
  enum gen_clip_mode;
  enum gen_front_winding;
  enum gen_fill_mode;
  enum gen_cull_mode;
  enum gen_pixel_location;
  enum gen_sample_count;
  enum gen_inputattr_select;
  enum gen_msrast_mode;
  enum gen_prefilter_op;

Correct the type of GEN6_SAMPLER_DW0_BASE_LOD.  Rename gen_logicop_function,
gen_sampler_mip_filter, gen_sampler_map_filter, gen_sampler_aniso_ratio, and
others.

8 years agoilo: add ilo_image_disable_aux()
Chia-I Wu [Fri, 22 May 2015 06:21:22 +0000 (14:21 +0800)]
ilo: add ilo_image_disable_aux()

When aux bo allocation fails, ilo_image_disable_aux() should be called to
disable aux buffer.

8 years agoilo: add array_size and level_count to ilo_image
Chia-I Wu [Tue, 26 May 2015 07:46:44 +0000 (15:46 +0800)]
ilo: add array_size and level_count to ilo_image

We will use them for bound checking.

8 years agoilo: add pipe_texture_target to ilo_image
Chia-I Wu [Sun, 17 May 2015 03:55:05 +0000 (11:55 +0800)]
ilo: add pipe_texture_target to ilo_image

Save the target in ilo_image instead of passing it around.

8 years agoilo: fix "Render Cache Read Write Mode"
Chia-I Wu [Fri, 15 May 2015 02:39:05 +0000 (10:39 +0800)]
ilo: fix "Render Cache Read Write Mode"

It needs be set to R/W only when using certain messages via DP render cache.
Since we only use RT wrties with the render cache, we never need to set it.

8 years agoilo: avoid resource owning in core
Chia-I Wu [Thu, 21 May 2015 08:30:03 +0000 (16:30 +0800)]
ilo: avoid resource owning in core

It is up to the users whether to reference count the BOs or not.

8 years agoilo: assert core objects are zero-initialized
Chia-I Wu [Fri, 22 May 2015 05:49:20 +0000 (13:49 +0800)]
ilo: assert core objects are zero-initialized

Core objects are usually embedded inside calloc()'ed objects and we expect
them to be zero-initialized.

8 years agoradeon/llvm: Handle LLVM backend rename from R600 to AMDGPU
Tom Stellard [Thu, 11 Jun 2015 15:42:25 +0000 (15:42 +0000)]
radeon/llvm: Handle LLVM backend rename from R600 to AMDGPU

Reviewed-by: Marek Olšák <marek.olsak@amd.com>
8 years agogallivm: Only build lp_profile() body when PROFILE is defined
Tom Stellard [Wed, 27 May 2015 23:51:43 +0000 (16:51 -0700)]
gallivm: Only build lp_profile() body when PROFILE is defined

The only use of lp_profile() is wrapped in #if defined(PROFILE),
so there is no reason to build it unless this macro is defined.

Reviewed-by: Jose Fonseca <jfonseca@vmware.com>
8 years agoglsl: fix compile error message
Timothy Arceri [Wed, 10 Jun 2015 08:35:08 +0000 (18:35 +1000)]
glsl: fix compile error message

Reviewed-by: Anuj Phogat <anuj.phogat@gmail.com>
8 years agoi965/gen8+: Add aux buffer alignment assertions
Ben Widawsky [Thu, 14 May 2015 16:28:37 +0000 (09:28 -0700)]
i965/gen8+: Add aux buffer alignment assertions

This helped find the incorrect HALIGN values from the previous patches.

v2: Add PRM references for assertions (Chad)

v3: Remove duplicated part of commit message, assert num_samples > 1, instead of
num_samples > 0. (Chad)

Signed-off-by: Ben Widawsky <ben@bwidawsk.net>
Reviewed-by: Chad Versace <chad.versace@intel.com>
Reviewed-by: Anuj Phogat <anuj.phogat@gmail.com>
Reviewed-by: Jordan Justen <jordan.l.justen@intel.com>
8 years agoi965/gen9: Set HALIGN_16 for all aux buffers
Ben Widawsky [Fri, 22 May 2015 22:57:37 +0000 (15:57 -0700)]
i965/gen9: Set HALIGN_16 for all aux buffers

Just like the previous patch, but for the GEN9 constraints.

v2:
bugfix: Gen9 HALIGN was being set for all miptree buffers (Chad). To address
this, move the check to where the gen8 check is, and do the appropriate
conditional there.

v3:
Remove stray whitespace introduced in v2 (Chad)
Rework comment to show AUX_CCS and AUX_MCS specifically. Remove misworded part
about gen7 (Chad).

Signed-off-by: Ben Widawsky <ben@bwidawsk.net>
Reviewed-by: Anuj Phogat <anuj.phogat@gmail.com> (v1)
Reviewed-by: Jordan Justen <jordan.l.justen@intel.com> (v1)
Reviewed-by: Chad Versace <chad.versace@intel.com>
8 years agoi965/gen8: Correct HALIGN for AUX surfaces
Ben Widawsky [Thu, 14 May 2015 16:30:02 +0000 (09:30 -0700)]
i965/gen8: Correct HALIGN for AUX surfaces

This restriction was attempted in this commit:
commit 47053464630888f819ef8cc44278f1a1220159b9
Author: Anuj Phogat <anuj.phogat@gmail.com>
Date:   Fri Feb 13 11:21:21 2015 -0800

   i965/gen8: Use HALIGN_16 if MCS is enabled for non-MSRT

However, the commit itself doesn't achieve the desired goal as determined by the
asserts which the next patch adds. mcs_mt is NULL (never set) we're in the
process of allocating the mcs_mt miptree when we get to this function. I didn't
check, but perhaps this would work with blorp, however, meta clears allocate the
miptree structure (which AFAICT needs the alignment also) way before it
allocates using meta clears where the renderbuffer is allocated way before the
aux buffer.

The restriction is referenced in a few places, but the most concise one [IMO]
from the spec is for Gen9. Gen8 loosens the restriction in that it only requires
this for non-msrt surface.

   When Auxiliary Surface Mode is set to AUX_CCS_D or AUX_CCS_E, HALIGN 16 must
   be used.

With the code before the miptree layout flag rework (patches preceding this),
accomplishing this workaround is very difficult.

v2:
bugfix: Don't set HALIGN16 for gens before 8 (Chad)

v3:
non-trivial rebase

Signed-off-by: Ben Widawsky <ben@bwidawsk.net>
Cc: Neil Roberts <neil@linux.intel.com>
Reviewed-by: Jordan Justen <jordan.l.justen@intel.com>
Reviewed-by: Anuj Phogat <anuj.phogat@gmail.com>
Reviewed-by: Chad Versace <chad.versace@intel.com>
8 years agoi965: Extract tiling from fast clear decision
Ben Widawsky [Fri, 22 May 2015 05:47:37 +0000 (22:47 -0700)]
i965: Extract tiling from fast clear decision

There are several constraints when determining if one can fast clear a surface.
Some of these are alignment, pixel density, tiling formats, and others that vary
by generation. The helper function which exists today does a suitable job,
however it conflates "BO properties" with "Miptree properties" when using
tiling. I consider the former to be attributes of the physical surface, things
which are determined through BO allocation, and the latter being attributes
which are derived from the API, and having nothing to do with the underlying
surface.

Determining tiling properties and creating miptrees are related operations
(when we allocate a BO for a miptree) with some disjoint constraints. By
extracting the decisions into two distinct choices (tiling vs. miptree
properties), we gain flexibility throughout the code to make determinations
about when we can or cannot fast clear strictly on the miptree.

To signify this change, I've also renamed the function to indicate it is a
distinction made on the miptree. I am torn as to whether or not it was a good
idea to remove "non_msrt" since it's a really nice thing for grep.

v2:
Reword some comments (Chad)
intel_is_non_msrt_mcs_tile_supported->intel_tiling_supports_non_msrt_mcs (Chad)
Make full if ladder for gens in above function (Chad)

Signed-off-by: Ben Widawsky <ben@bwidawsk.net>
Cc: Topi Pohjolainen <topi.pohjolainen@intel.com>
Reviewed-by: Jordan Justen <jordan.l.justen@intel.com>
Reviewed-by: Anuj Phogat <anuj.phogat@gmail.com>
Reviewed-by: Chad Versace <chad.versace@linux.intel.com>