mesa.git
8 years agoanv/state: Allow levelCount to be 0
Jason Ekstrand [Thu, 17 Dec 2015 01:27:35 +0000 (17:27 -0800)]
anv/state: Allow levelCount to be 0

This can happen if the client is creating an image view of a textureable
surface and they only ever intend to render to that view.

8 years agonir/spirv: Add a missing break statement
Jason Ekstrand [Wed, 16 Dec 2015 01:24:18 +0000 (17:24 -0800)]
nir/spirv: Add a missing break statement

8 years agoanv/pipeline: Allow the user to pass a null MultisampleCreateInfo
Jason Ekstrand [Wed, 16 Dec 2015 00:26:08 +0000 (16:26 -0800)]
anv/pipeline: Allow the user to pass a null MultisampleCreateInfo

According to section 5.2 of the Vulkan spec, this is allowed for color-only
rendering pipelines.

8 years agoanv/descriptor_set: Initialize immutable_samplers to NULL
Jason Ekstrand [Wed, 16 Dec 2015 00:24:19 +0000 (16:24 -0800)]
anv/descriptor_set: Initialize immutable_samplers to NULL

Previously this wasn't a problem.  However, with the new API update,
descriptor sets can now be sparse so the client doesn't have to provide an
entry for every binding.  This means that it's possible for a binding to be
uninitialized other than the memset.  In that case, we want to have a null
array of immutable samplers.

8 years agoanv/device: Bump the size of the instruction block pool
Jason Ekstrand [Tue, 15 Dec 2015 19:49:26 +0000 (11:49 -0800)]
anv/device: Bump the size of the instruction block pool

Some CTS test shaders were failing to compile.  At some point soon, we
really need to make a real pipeline cache and stop using a block pool for
this.

8 years agoanv/pipeline: Properly set IncludeVertexHandles in 3DSTATE_GS
Jason Ekstrand [Tue, 15 Dec 2015 19:36:55 +0000 (11:36 -0800)]
anv/pipeline: Properly set IncludeVertexHandles in 3DSTATE_GS

8 years agonir/spirv: Add support for more CS intrinsics
Jason Ekstrand [Tue, 15 Dec 2015 04:33:46 +0000 (20:33 -0800)]
nir/spirv: Add support for more CS intrinsics

8 years agonir/lower_system_values: Add support for computed builtins.
Jason Ekstrand [Tue, 15 Dec 2015 04:22:09 +0000 (20:22 -0800)]
nir/lower_system_values: Add support for computed builtins.

In particular, this commit adds support for computing gl_GlobalInvocationID
and gl_LocalInvocationIndex from other intrinsics.

8 years agoshader_enums: Add enums for gl_GlobalInvocationID and gl_LocalInvocationIndex
Jason Ekstrand [Tue, 15 Dec 2015 04:20:58 +0000 (20:20 -0800)]
shader_enums: Add enums for gl_GlobalInvocationID and gl_LocalInvocationIndex

8 years agonir/lower_system_values: Refactor and use the builder.
Jason Ekstrand [Tue, 15 Dec 2015 03:51:51 +0000 (19:51 -0800)]
nir/lower_system_values: Refactor and use the builder.

Now that we have a helper in the builder for system values and a helper in
core NIR to get the intrinsic opcode, there's really no point in having
things split out into a helper function.  This commit "modernizes" this
pass to use helpers better and look more like newer passes.

8 years agonir/builder: Add a load_system_value helper
Jason Ekstrand [Tue, 15 Dec 2015 03:51:50 +0000 (19:51 -0800)]
nir/builder: Add a load_system_value helper

While we're at it, go ahead and make nir_lower_clip use it.

Cc: Rob Clark <robclark@gmail.com>
8 years agonir/lower_system_values: Stop supporting non-SSA
Jason Ekstrand [Tue, 15 Dec 2015 03:51:49 +0000 (19:51 -0800)]
nir/lower_system_values: Stop supporting non-SSA

The one user of this (i965) only ever calls it while in SSA form.

8 years agoisl: Add func isl_surf_get_image_offset_sa
Chad Versace [Sat, 12 Dec 2015 01:14:52 +0000 (17:14 -0800)]
isl: Add func isl_surf_get_image_offset_sa

The function calculates the offset to a subimage within the surface, in
units of surface samples.

All unit tests pass with `make check`. (Admittedly, though, there are
too few unit tests).

8 years agoisl: Fix calculation of array pitch for layout GEN4_2D
Chad Versace [Mon, 14 Dec 2015 20:31:55 +0000 (12:31 -0800)]
isl: Fix calculation of array pitch for layout GEN4_2D

The height of the miptree's right half was not large enough.

Found by `make check` in test_isl_surf_get_offset, which is added in the
next commit.

8 years agoisl: Move it a standalone directory
Chad Versace [Mon, 14 Dec 2015 18:28:18 +0000 (10:28 -0800)]
isl: Move it a standalone directory

The plan all along was to eventualyl move isl out of the Vulkan
directory, because I intended i965 and anvil to share it.

A small problem I encountered when attempting to write unit tests for
isl precipitated the move.  I discovered that it's easier to get isl
unit tests to build if I remove the extra, unneeded dependencies
injected by src/vulkan/Makefile.am. And the easiest way to remove those
unneeded dependencies is to move isl out of src/vulkan. (Unit tests come
in subsequent commits).

8 years agovec4/generator: Actually pass the sampler into generate_tex
Jason Ekstrand [Tue, 15 Dec 2015 05:10:26 +0000 (21:10 -0800)]
vec4/generator: Actually pass the sampler into generate_tex

This is an artifact of the way the separate samplers/textures series ended
up getting sent out and rebased.  This should fix a number of CTS tests
involving geometry shaders.

8 years agoanv: Rename gs_vec4 to gs_kernel
Jordan Justen [Tue, 15 Dec 2015 02:16:38 +0000 (18:16 -0800)]
anv: Rename gs_vec4 to gs_kernel

The code generated may be vec4 or simd8 depending on how we start the
compiler.

To run the GS in SIMD8, set the INTEL_SCALAR_GS environment variable.
This was added in:

    commit 36fd65381756ed1b8f774f7fcdd555941a3d39e1
    Author: Kenneth Graunke <kenneth@whitecape.org>
    Date:   Wed Mar 11 23:14:31 2015 -0700

        i965: Add scalar geometry shader support.

Signed-off-by: Jordan Justen <jordan.l.justen@intel.com>
8 years agonir/spirv_to_nir: Use a minimum of 1 for GS invocations
Jordan Justen [Tue, 15 Dec 2015 02:12:57 +0000 (18:12 -0800)]
nir/spirv_to_nir: Use a minimum of 1 for GS invocations

glslang is giving us 0, which causes the SIMD8 GS compile to hit an
assert.

Signed-off-by: Jordan Justen <jordan.l.justen@intel.com>
8 years agoanv: Fix CUBE storage images
Jason Ekstrand [Tue, 15 Dec 2015 00:59:59 +0000 (16:59 -0800)]
anv: Fix CUBE storage images

8 years agoanv: Add support for storage texel buffers
Jason Ekstrand [Tue, 15 Dec 2015 00:51:12 +0000 (16:51 -0800)]
anv: Add support for storage texel buffers

8 years agoanv: Pass an isl_format into fill_buffer_surface_state
Jason Ekstrand [Tue, 15 Dec 2015 00:14:20 +0000 (16:14 -0800)]
anv: Pass an isl_format into fill_buffer_surface_state

8 years agoanv/cmd_buffer: Gen 8 requires 64 byte alignment for push constant data
Jordan Justen [Mon, 14 Dec 2015 23:24:11 +0000 (15:24 -0800)]
anv/cmd_buffer: Gen 8 requires 64 byte alignment for push constant data

See MEDIA_CURBE_LOAD, CURBE Data Start Address & CURBE Total Data Length

Signed-off-by: Jordan Justen <jordan.l.justen@intel.com>
8 years agoanv: Add initial support for cube maps
Jason Ekstrand [Mon, 14 Dec 2015 22:52:57 +0000 (14:52 -0800)]
anv: Add initial support for cube maps

This fixes 486 cubemap CTS tests.

8 years agoanv/buffer_view: Store a bo + offset instead of buffer pointer
Jason Ekstrand [Mon, 14 Dec 2015 22:09:35 +0000 (14:09 -0800)]
anv/buffer_view: Store a bo + offset instead of buffer pointer

This is what image_view does.  Also, we really need to do this so that we
can properly handle the combined offsets from the buffer and from
pCreateInfo.

This fixes some of the nonzero offset buffer view CTS tests.

8 years agoanv: Remove anv_image::surface_type
Chad Versace [Thu, 10 Dec 2015 00:59:02 +0000 (16:59 -0800)]
anv: Remove anv_image::surface_type

When building RENDER_SURFACE_STATE, the driver set
SurfaceType = anv_image::surface_type, which was calculated during
anv_image_init(). This was bad because the value of
anv_image::surface_type was taken from a gen-specific header,
gen8_pack.h, even though the anv_image structure is used for all gens.

Replace anv_image::surface_type with a gen-specific lookup function,
anv_surftype(), defined in gen${x}_state.c.

The lookup function contains some useful asserts that caught some nasty
bugs in anv meta, which were fixed in the previous commit.

8 years agoanv/meta: Fix VkImageViewType
Chad Versace [Thu, 10 Dec 2015 01:03:14 +0000 (17:03 -0800)]
anv/meta: Fix VkImageViewType

Meta unconditionally used VK_IMAGE_VIEW_TYPE_2D in the functions below.
This caused some out-of-bound memory accesses.
  anv_CmdCopyImage
  anv_CmdBlitImage
  anv_CmdCopyBufferToImage
  anv_CmdClearColorImage

Fix it by adding a new function, anv_meta_get_view_type().

8 years agoisl: Rename s/lod_align/image_align/ for consistency
Chad Versace [Mon, 14 Dec 2015 16:48:19 +0000 (08:48 -0800)]
isl: Rename s/lod_align/image_align/ for consistency

Regarding the subimages within a surface, sometimes isl called them
"images" and sometimes "LODs". This patch make isl consistently refer to
them as "images".  I choose the term "image" over "LOD" because LOD is
an misnomer when applied to 3D surfaces. The alignment applies to each
individual 2D subimage, not to the LOD as a whole.

This patch changes no behavior. It's just a manually performed,
case-insensitive, replacement s/lod/image/ that maintains correct
indentation.  any behavior.

8 years agoanv/tests: gitignore block_pool_no_free
Chad Versace [Mon, 14 Dec 2015 16:27:59 +0000 (08:27 -0800)]
anv/tests: gitignore block_pool_no_free

8 years agoanv: Fix build for unit tests
Chad Versace [Mon, 14 Dec 2015 16:23:17 +0000 (08:23 -0800)]
anv: Fix build for unit tests

Clearly no one has been running `make check`, because the unittestbuild
has been broken for a long time. After this buildfix, all tests now
pass.

8 years agoanv: Add initial support for texel buffers
Jason Ekstrand [Sun, 13 Dec 2015 00:11:23 +0000 (16:11 -0800)]
anv: Add initial support for texel buffers

8 years agoi965/nir: Provide a default LOD for buffer textures
Jason Ekstrand [Sun, 13 Dec 2015 00:09:51 +0000 (16:09 -0800)]
i965/nir: Provide a default LOD for buffer textures

Our hardware requires an LOD for all texelFetch commands even if they are
on buffer textures.  GLSL IR gives us an LOD of 0 in that case, but the LOD
is really rather meaningless.  This commit allows other NIR producers to be
more lazy and not provide one at all.

8 years agoMerge remote-tracking branch 'mesa-public/master' into vulkan
Jason Ekstrand [Fri, 11 Dec 2015 22:29:13 +0000 (14:29 -0800)]
Merge remote-tracking branch 'mesa-public/master' into vulkan

This pulls in a shared local memory fix.

8 years agoi965/fs: Use the correct source for local memory load offsets
Jason Ekstrand [Fri, 11 Dec 2015 19:59:53 +0000 (11:59 -0800)]
i965/fs: Use the correct source for local memory load offsets

The offset for loads is in src[0].  This was a copy+paste error in the
nir_intrinsic_load/store refactoring.  This commit fixes a segfault in
ES31-CTS.compute_shader.work-group-size.  I have no idea how piglit failed
to catch this...

Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=93348
Reviewed-by: Kristian Høgsberg <krh@bitplanet.net>
8 years agogen8/pipeline: Support vec4 vertex shaders
Jason Ekstrand [Fri, 11 Dec 2015 21:25:15 +0000 (13:25 -0800)]
gen8/pipeline: Support vec4 vertex shaders

In order to actually get them, you need INTEL_DEBUG=vec4.

8 years agoi965: Add Gen8+ tessellation control shader state (3DSTATE_HS).
Kenneth Graunke [Thu, 12 Nov 2015 07:22:06 +0000 (23:22 -0800)]
i965: Add Gen8+ tessellation control shader state (3DSTATE_HS).

Signed-off-by: Kenneth Graunke <kenneth@whitecape.org>
Reviewed-by: Jordan Justen <jordan.l.justen@intel.com>
8 years agoi965: Add Gen7+ tessellation engine state (3DSTATE_TE).
Kenneth Graunke [Thu, 12 Nov 2015 07:15:23 +0000 (23:15 -0800)]
i965: Add Gen7+ tessellation engine state (3DSTATE_TE).

Signed-off-by: Kenneth Graunke <kenneth@whitecape.org>
Reviewed-by: Jordan Justen <jordan.l.justen@intel.com>
8 years agoi965: Add Gen8+ tessellation evaluation shader state (3DSTATE_DS).
Kenneth Graunke [Tue, 10 Nov 2015 22:35:27 +0000 (14:35 -0800)]
i965: Add Gen8+ tessellation evaluation shader state (3DSTATE_DS).

Signed-off-by: Kenneth Graunke <kenneth@whitecape.org>
Reviewed-by: Jordan Justen <jordan.l.justen@intel.com>
8 years agoi965: Add tessellation shader push constant support.
Kenneth Graunke [Thu, 1 Oct 2015 00:04:23 +0000 (17:04 -0700)]
i965: Add tessellation shader push constant support.

Based on a patch by Chris Forbes.

Signed-off-by: Kenneth Graunke <kenneth@whitecape.org>
Reviewed-by: Jordan Justen <jordan.l.justen@intel.com>
8 years agoi965: Add tessellation shader sampler support.
Kenneth Graunke [Sat, 10 Oct 2015 00:07:23 +0000 (17:07 -0700)]
i965: Add tessellation shader sampler support.

Based on code by Chris Forbes and Fabian Bieler.

Signed-off-by: Kenneth Graunke <kenneth@whitecape.org>
Reviewed-by: Jordan Justen <jordan.l.justen@intel.com>
8 years agoi965: Add tessellation shader surface support.
Kenneth Graunke [Wed, 30 Sep 2015 22:42:54 +0000 (15:42 -0700)]
i965: Add tessellation shader surface support.

This is brw_gs_surface_state.c copy and pasted twice with search and
replace.

brw_binding_table.c code is similarly copy and pasted.

v2: Drop dword_pitch related fields.

Signed-off-by: Kenneth Graunke <kenneth@whitecape.org>
Acked-by: Jason Ekstrand <jason.ekstrand@intel.com>
8 years agoi965: Make fs_visitor::emit_urb_writes set EOT for TES as well.
Kenneth Graunke [Wed, 2 Dec 2015 01:52:58 +0000 (17:52 -0800)]
i965: Make fs_visitor::emit_urb_writes set EOT for TES as well.

Tessellation evaluation shaders work almost identically to vertex
shaders - we have a set of URB writes at the end of the program, and the
last one should terminate it.

Geometry shaders really are the special case, where multiple
EmitVertex() calls trigger URB writes in the middle of the program.

Signed-off-by: Kenneth Graunke <kenneth@whitecape.org>
Reviewed-by: Jason Ekstrand <jason.ekstrand@intel.com>
Reviewed-by: Kristian Høgsberg <krh@bitplanet.net>
8 years agoi965: Don't hardcode g1 for URB handles in fs_visitor::emit_urb_writes().
Kenneth Graunke [Wed, 2 Dec 2015 01:51:33 +0000 (17:51 -0800)]
i965: Don't hardcode g1 for URB handles in fs_visitor::emit_urb_writes().

Tessellation evaluation shaders will use g4 instead.  For now, make an
fs_reg called urb_handle and use that in place of hardcoding g1.

Signed-off-by: Kenneth Graunke <kenneth@whitecape.org>
Reviewed-by: Jason Ekstrand <jason.ekstrand@intel.com>
Reviewed-by: Kristian Høgsberg <krh@bitplanet.net>
8 years agoi965: Make brw_set_message_descriptor() non-static.
Kenneth Graunke [Wed, 2 Dec 2015 02:20:54 +0000 (18:20 -0800)]
i965: Make brw_set_message_descriptor() non-static.

I want to use this directly from brw_vec4_generator.cpp.

Signed-off-by: Kenneth Graunke <kenneth@whitecape.org>
Reviewed-by: Jason Ekstrand <jason.ekstrand@intel.com>
Reviewed-by: Kristian Høgsberg <krh@bitplanet.net>
8 years agoRevert "i965/HACK: Build brw_cs into libcompiler"
Kristian Høgsberg Kristensen [Fri, 11 Dec 2015 21:09:42 +0000 (13:09 -0800)]
Revert "i965/HACK: Build brw_cs into libcompiler"

This reverts commit 6df7963531b9c33caf96a68e8e6cd4796677b8aa.

8 years agoMerge ../mesa into vulkan
Kristian Høgsberg Kristensen [Fri, 11 Dec 2015 21:09:06 +0000 (13:09 -0800)]
Merge ../mesa into vulkan

8 years agoi965: Move brw_cs_fill_local_id_payload() to libi965_compiler
Kristian Høgsberg Kristensen [Fri, 11 Dec 2015 19:18:18 +0000 (11:18 -0800)]
i965: Move brw_cs_fill_local_id_payload() to libi965_compiler

This is a helper function for setting up the local invocation ID
payload according to the cs_prog_data generated by the compiler. It's
intended to be available to users of libi965_compiler so move it there.

8 years agovc4: Add quick algebraic optimization for clamping of unpacked values.
Eric Anholt [Fri, 11 Dec 2015 05:54:41 +0000 (21:54 -0800)]
vc4: Add quick algebraic optimization for clamping of unpacked values.

GL likes to saturate your incoming color, but if that color's coming from
unpacking from unorms, there's no point.  Ideally we'd have a range
propagation pass that cleans these up in NIR, but that doesn't seem to be
going to land soon.  It seems like we could do a one-off optimization in
nir_opt_algebraic, except that doesn't want to operate on expressions
involving unpack_unorm_4x8, since it's sized.

total instructions in shared programs: 87879 -> 87761 (-0.13%)
instructions in affected programs:     6044 -> 5926 (-1.95%)
total estimated cycles in shared programs: 349457 -> 349252 (-0.06%)
estimated cycles in affected programs:     6172 -> 5967 (-3.32%)

No SSPD on openarena (which had the biggest gains, in its VS/CSes), n=15.

8 years agovc4: When doing algebraic optimization into a MOV, use the right MOV.
Eric Anholt [Fri, 11 Dec 2015 06:02:30 +0000 (22:02 -0800)]
vc4: When doing algebraic optimization into a MOV, use the right MOV.

If there were src unpacks, changing to the integer MOV instead of float
(for example) would change the unpack operation.

8 years agovc4: Fix handling of src packs on in qir_follow_movs().
Eric Anholt [Fri, 11 Dec 2015 06:23:10 +0000 (22:23 -0800)]
vc4: Fix handling of src packs on in qir_follow_movs().

The caller isn't going to expect it from a return, so it would probably
get misinterpreted.  If the caller had an unpack in its reg, that's fine,
but don't lose track of it.

8 years agovc4: Add missing progress note in opt_algebraic.
Eric Anholt [Fri, 11 Dec 2015 05:51:03 +0000 (21:51 -0800)]
vc4: Add missing progress note in opt_algebraic.

8 years agovc4: Add debugging of the estimated time to run the shader to shader-db.
Eric Anholt [Wed, 9 Dec 2015 01:55:36 +0000 (17:55 -0800)]
vc4: Add debugging of the estimated time to run the shader to shader-db.

8 years agovc4: Fix handling of sample_mask output.
Eric Anholt [Wed, 9 Dec 2015 01:18:37 +0000 (17:18 -0800)]
vc4: Fix handling of sample_mask output.

I apparently broke this in a late refactor, in such a way that I decided
its tests were some of those interminable ones that I should just
blacklist from my testing.  As a result, the refactors related to it were
totally wrong.

8 years agosoftpipe: enable GL_ARB_viewport_array support, update GL3.txt doc
Edward O'Callaghan [Fri, 11 Dec 2015 11:43:31 +0000 (22:43 +1100)]
softpipe: enable GL_ARB_viewport_array support, update GL3.txt doc

Signed-off-by: Edward O'Callaghan <eocallaghan@alterapraxis.com>
Reviewed-by: Roland Scheidegger <sroland@vmware.com>
8 years agosoftpipe: implement some support for multiple viewports
Edward O'Callaghan [Fri, 11 Dec 2015 11:43:30 +0000 (22:43 +1100)]
softpipe: implement some support for multiple viewports

Mostly related to making sure the rasterizer can correctly
pick out the correct scissor box for the current viewport.

Signed-off-by: Edward O'Callaghan <eocallaghan@alterapraxis.com>
Reviewed-by: Roland Scheidegger <sroland@vmware.com>
8 years agodraw: don't assume fixed offset for data in struct vertex_info
Roland Scheidegger [Fri, 11 Dec 2015 03:53:21 +0000 (04:53 +0100)]
draw: don't assume fixed offset for data in struct vertex_info

Otherwise, if struct vertex_info is changed, you're in for some surprises...

Reviewed-by: Jose Fonseca <jfonseca@vmware.com>
8 years agoi965/gen9: Don't do fast clears when GL_FRAMEBUFFER_SRGB is enabled
Neil Roberts [Wed, 25 Nov 2015 11:14:37 +0000 (12:14 +0100)]
i965/gen9: Don't do fast clears when GL_FRAMEBUFFER_SRGB is enabled

When GL_FRAMEBUFFER_SRGB is enabled any single-sampled renderbuffers
are resolved in intel_update_state because the hardware can't cope
with fast clears on SRGB buffers. In that case it's pointless to do a
fast clear because it will just be immediately resolved.

Reviewed-by: Topi Pohjolainen <topi.pohjolainen@intel.com>
8 years agoi965/gen9: Allow fast clears for non-MSRT SRGB buffers
Neil Roberts [Tue, 24 Nov 2015 15:34:46 +0000 (16:34 +0100)]
i965/gen9: Allow fast clears for non-MSRT SRGB buffers

SRGB buffers are not marked as losslessly compressible so previously
they would not be used for fast clears. However in practice the
hardware will never actually see that we are using SRGB buffers for
fast clears if we use the linear equivalent format when clearing and
make sure to resolve the buffer as a linear format before sampling
from it.

This is an important use case because by default the window system
framebuffers are created as SRGB so without this fast clears won't be
used there.

Reviewed-by: Topi Pohjolainen <topi.pohjolainen@intel.com>
8 years agoi965/gen9: Resolve SRGB color buffers when GL_FRAMEBUFFER_SRGB enabled
Neil Roberts [Tue, 24 Nov 2015 18:23:14 +0000 (19:23 +0100)]
i965/gen9: Resolve SRGB color buffers when GL_FRAMEBUFFER_SRGB enabled

SKL can't cope with the CCS buffer for SRGB buffers. Normally the
hardware won't see the SRGB formats because when GL_FRAMEBUFFER_SRGB
is disabled these get mapped to their linear equivalents. In order to
avoid relying on the CCS buffer when it is enabled this patch now
makes it flush the renderbuffers.

Reviewed-by: Topi Pohjolainen <topi.pohjolainen@intel.com>
8 years agoi965/gen8+: Don't upload the MCS buffer for single-sampled textures
Neil Roberts [Tue, 24 Nov 2015 16:59:28 +0000 (17:59 +0100)]
i965/gen8+: Don't upload the MCS buffer for single-sampled textures

For single-sampled textures the MCS buffer is only used to implement
fast clears. However the surface always needs to be resolved before
being used as a texture anyway so the the MCS buffer doesn't actually
achieve anything. This is important for Gen9 because in that case SRGB
surfaces are not supported for fast clears and we don't want the
hardware to see the MCS buffer in that case.

Reviewed-by: Topi Pohjolainen <topi.pohjolainen@intel.com>
8 years agoi965/meta-fast-clear: Disable GL_FRAMEBUFFER_SRGB during clear
Neil Roberts [Tue, 24 Nov 2015 16:01:03 +0000 (17:01 +0100)]
i965/meta-fast-clear: Disable GL_FRAMEBUFFER_SRGB during clear

Adds MESA_META_FRAMEBUFFER_SRGB to the meta save state so that
GL_FRAMEBUFFER_SRGB will be disabled when performing the fast clear.
That way the render surface state will be programmed with the linear
equivalent format during the clear. This is important for Gen9 because
the SRGB formats are not marked as losslessly compressible so in
theory they aren't support for fast clears. It shouldn't make any
difference whether GL_FRAMEBUFFER_SRGB is enabled for the fast clear
operation because the color is not actually written to the framebuffer
so there is no chance for the hardware to apply the SRGB conversion on
it anyway.

Reviewed-by: Topi Pohjolainen <topi.pohjolainen@intel.com>
8 years agowinsys/amdgpu: clear the buffer cache on mmap failure and try again
Marek Olšák [Wed, 9 Dec 2015 21:45:56 +0000 (22:45 +0100)]
winsys/amdgpu: clear the buffer cache on mmap failure and try again

Reviewed-by: Michel Dänzer <michel.daenzer@amd.com>
8 years agowinsys/radeon: clear the buffer cache on mmap failure and try again
Marek Olšák [Wed, 9 Dec 2015 21:45:56 +0000 (22:45 +0100)]
winsys/radeon: clear the buffer cache on mmap failure and try again

Reviewed-by: Michel Dänzer <michel.daenzer@amd.com>
8 years agowinsys/amdgpu: clear the buffer cache on allocation failure and try again
Marek Olšák [Wed, 9 Dec 2015 21:36:26 +0000 (22:36 +0100)]
winsys/amdgpu: clear the buffer cache on allocation failure and try again

Reviewed-by: Michel Dänzer <michel.daenzer@amd.com>
8 years agowinsys/radeon: clear the buffer cache on allocation failure and try again
Marek Olšák [Wed, 9 Dec 2015 21:36:26 +0000 (22:36 +0100)]
winsys/radeon: clear the buffer cache on allocation failure and try again

Reviewed-by: Michel Dänzer <michel.daenzer@amd.com>
8 years agogallium/radeon: remove radeon_winsys_cs_handle
Marek Olšák [Sun, 6 Dec 2015 23:00:59 +0000 (00:00 +0100)]
gallium/radeon: remove radeon_winsys_cs_handle

"radeon_winsys_cs_handle *cs_buf" is now equivalent to "pb_buffer *buf".

Reviewed-by: Nicolai Hähnle <nicolai.haehnle@amd.com>
Reviewed-by: Edward O'Callaghan <eocallaghan@alterapraxis.com>
Reviewed-by: Michel Dänzer <michel.daenzer@amd.com>
8 years agowinsys/radeon: use pb_cache instead of pb_cache_manager
Marek Olšák [Sun, 6 Dec 2015 19:57:05 +0000 (20:57 +0100)]
winsys/radeon: use pb_cache instead of pb_cache_manager

This is a prerequisite for the removal of radeon_winsys_cs_handle.

Reviewed-by: Nicolai Hähnle <nicolai.haehnle@amd.com>
Reviewed-by: Edward O'Callaghan <eocallaghan@alterapraxis.com>
Reviewed-by: Michel Dänzer <michel.daenzer@amd.com>
8 years agowinsys/radeon: use radeon_bomgr less
Marek Olšák [Sun, 6 Dec 2015 21:48:45 +0000 (22:48 +0100)]
winsys/radeon: use radeon_bomgr less

Reviewed-by: Nicolai Hähnle <nicolai.haehnle@amd.com>
Reviewed-by: Edward O'Callaghan <eocallaghan@alterapraxis.com>
Reviewed-by: Michel Dänzer <michel.daenzer@amd.com>
8 years agowinsys/radeon: rename radeon_bomgr_init_functions
Marek Olšák [Sun, 6 Dec 2015 21:34:01 +0000 (22:34 +0100)]
winsys/radeon: rename radeon_bomgr_init_functions

Reviewed-by: Nicolai Hähnle <nicolai.haehnle@amd.com>
Reviewed-by: Edward O'Callaghan <eocallaghan@alterapraxis.com>
Reviewed-by: Michel Dänzer <michel.daenzer@amd.com>
8 years agowinsys/radeon: move variables from radeon_bomgr to radeon_drm_winsys
Marek Olšák [Sun, 6 Dec 2015 21:32:33 +0000 (22:32 +0100)]
winsys/radeon: move variables from radeon_bomgr to radeon_drm_winsys

radeon_bomgr is going away.

Reviewed-by: Nicolai Hähnle <nicolai.haehnle@amd.com>
Reviewed-by: Edward O'Callaghan <eocallaghan@alterapraxis.com>
Reviewed-by: Michel Dänzer <michel.daenzer@amd.com>
8 years agowinsys/radeon: remove redundant radeon_bomgr::va
Marek Olšák [Sun, 6 Dec 2015 21:10:04 +0000 (22:10 +0100)]
winsys/radeon: remove redundant radeon_bomgr::va

Reviewed-by: Nicolai Hähnle <nicolai.haehnle@amd.com>
Reviewed-by: Edward O'Callaghan <eocallaghan@alterapraxis.com>
Reviewed-by: Michel Dänzer <michel.daenzer@amd.com>
8 years agowinsys/amdgpu: don't use the "rws" abbreviation for amdgpu_winsys
Marek Olšák [Sun, 6 Dec 2015 21:19:38 +0000 (22:19 +0100)]
winsys/amdgpu: don't use the "rws" abbreviation for amdgpu_winsys

Reviewed-by: Nicolai Hähnle <nicolai.haehnle@amd.com>
Reviewed-by: Edward O'Callaghan <eocallaghan@alterapraxis.com>
Reviewed-by: Michel Dänzer <michel.daenzer@amd.com>
8 years agowinsys/amdgpu: use pb_cache instead of pb_cache_manager
Marek Olšák [Sun, 6 Dec 2015 19:57:05 +0000 (20:57 +0100)]
winsys/amdgpu: use pb_cache instead of pb_cache_manager

This is a prerequisite for the removal of radeon_winsys_cs_handle.

Reviewed-by: Nicolai Hähnle <nicolai.haehnle@amd.com>
Reviewed-by: Edward O'Callaghan <eocallaghan@alterapraxis.com>
Reviewed-by: Michel Dänzer <michel.daenzer@amd.com>
8 years agogallium/pb_bufmgr_cache: use the new pb_cache module
Marek Olšák [Sun, 6 Dec 2015 23:23:06 +0000 (00:23 +0100)]
gallium/pb_bufmgr_cache: use the new pb_cache module

Reviewed-by: Nicolai Hähnle <nicolai.haehnle@amd.com>
Reviewed-by: Edward O'Callaghan <eocallaghan@alterapraxis.com>
Acked-by: Michel Dänzer <michel.daenzer@amd.com>
8 years agogallium/pb_cache: add a copy of cache bufmgr independent of pb_manager
Marek Olšák [Sun, 6 Dec 2015 18:38:26 +0000 (19:38 +0100)]
gallium/pb_cache: add a copy of cache bufmgr independent of pb_manager

This simplified (basically duplicated) version of pb_cache_manager will
allow removing some ugly hacks from radeon and amdgpu winsyses and
flatten simplify their design.

The difference is that winsyses must manually add buffers to the cache
in "destroy" functions and the cache doesn't know about the buffers before
that. The integration is therefore trivial and the impact on the winsys
design is negligible.

Reviewed-by: Nicolai Hähnle <nicolai.haehnle@amd.com>
Reviewed-by: Edward O'Callaghan <eocallaghan@alterapraxis.com>
Acked-by: Michel Dänzer <michel.daenzer@amd.com>
8 years agoradeonsi: implement fast stencil clear
Marek Olšák [Thu, 10 Dec 2015 00:37:39 +0000 (01:37 +0100)]
radeonsi: implement fast stencil clear

Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
8 years agoradeonsi: re-enable Hyper-Z for stencil
Marek Olšák [Tue, 8 Dec 2015 16:33:55 +0000 (17:33 +0100)]
radeonsi: re-enable Hyper-Z for stencil

Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
8 years agor600g: remove a Hyper-Z workaround that's likely not needed anymore
Marek Olšák [Thu, 10 Dec 2015 00:46:17 +0000 (01:46 +0100)]
r600g: remove a Hyper-Z workaround that's likely not needed anymore

FORCE_OFF == 0, no need to set that

Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
8 years agor600g: re-enable Hyper-Z for stencil on Evergreen & Cayman
Marek Olšák [Thu, 10 Dec 2015 00:40:14 +0000 (01:40 +0100)]
r600g: re-enable Hyper-Z for stencil on Evergreen & Cayman

Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
8 years agogallium/radeon: fix Hyper-Z hangs by programming PA_SC_MODE_CNTL_1 correctly
Marek Olšák [Wed, 9 Dec 2015 19:26:21 +0000 (20:26 +0100)]
gallium/radeon: fix Hyper-Z hangs by programming PA_SC_MODE_CNTL_1 correctly

This is the recommended setting according to hw people and it makes Hyper-Z
stable. Just the two magic states.

This fixes Evergreen, Cayman, SI, CI, VI (using the Cayman code).

Cc: 11.0 11.1 <mesa-stable@lists.freedesktop.org>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
8 years agoradeonsi: don't use the CP DMA workaround on Fiji and newer
Marek Olšák [Fri, 4 Dec 2015 20:24:46 +0000 (21:24 +0100)]
radeonsi: don't use the CP DMA workaround on Fiji and newer

Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
8 years agoradeonsi: apply the streamout workaround to Fiji as well
Marek Olšák [Fri, 4 Dec 2015 20:24:21 +0000 (21:24 +0100)]
radeonsi: apply the streamout workaround to Fiji as well

Cc: 11.0 11.1 <mesa-stable@lists.freedesktop.org>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
8 years agoradeonsi: also print hexadecimal values for register fields in the IB parser
Marek Olšák [Wed, 9 Dec 2015 22:39:45 +0000 (23:39 +0100)]
radeonsi: also print hexadecimal values for register fields in the IB parser

Reviewed-by: Michel Dänzer <michel.daenzer@amd.com
Reviewed-by: Edward O'Callaghan <eocallaghan@alterapraxis.com>
8 years agoradeonsi: implement RB+ for Stoney (v2)
Marek Olšák [Tue, 1 Dec 2015 13:56:54 +0000 (14:56 +0100)]
radeonsi: implement RB+ for Stoney (v2)

v2: fix dual source blending

Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
8 years agoradeonsi: don't call of u_prims_for_vertices for patches and rectangles
Marek Olšák [Wed, 9 Dec 2015 21:14:32 +0000 (22:14 +0100)]
radeonsi: don't call of u_prims_for_vertices for patches and rectangles

Both caused a crash due to a division by zero in that function.
This is an alternative fix.

Cc: 11.0 11.1 <mesa-stable@lists.freedesktop.org>
Reviewed-by: Michel Dänzer <michel.daenzer@amd.com>
Reviewed-by: Edward O'Callaghan <eocallaghan@alterapraxis.com>
8 years agoradeonsi: use tgsi_shader_info::colors_written
Marek Olšák [Thu, 10 Dec 2015 12:16:58 +0000 (13:16 +0100)]
radeonsi: use tgsi_shader_info::colors_written

Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
8 years agor600g: write all MRTs only if there is exactly one output (fixes a hang)
Marek Olšák [Thu, 10 Dec 2015 12:19:45 +0000 (13:19 +0100)]
r600g: write all MRTs only if there is exactly one output (fixes a hang)

This fixes a hang in
piglit/arb_blend_func_extended-fbo-extended-blend-pattern_gles2 on REDWOOD.

Cc: 11.0 11.1 <mesa-stable@lists.freedesktop.org>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
8 years agotgsi/scan: add flag colors_written
Marek Olšák [Thu, 10 Dec 2015 12:15:50 +0000 (13:15 +0100)]
tgsi/scan: add flag colors_written

This is a prerequisite for the following r600g fix.

Cc: 11.0 11.1 <mesa-stable@lists.freedesktop.org>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
8 years agoRevert "radeonsi: disable DCC on Stoney"
Marek Olšák [Fri, 11 Dec 2015 14:08:00 +0000 (15:08 +0100)]
Revert "radeonsi: disable DCC on Stoney"

This reverts commit 32f05fadbbdf2a3fb60055e610bbbdcd82dd3ce5.

It turned out the problem with Stoney was caused by incorrect handling of
a non-power-two VRAM size in the kernel driver.
This is an optional BIOS setting and can be worked around by choosing
a different VRAM size in the BIOS.

Cc: 11.1 <mesa-stable@lists.freedesktop.org>
8 years agonir: silence uninitialized warning
Timothy Arceri [Fri, 11 Dec 2015 00:07:14 +0000 (11:07 +1100)]
nir: silence uninitialized warning

Reviewed-by: Rob Clark <robdclark@gmail.com>
8 years agoanv/pipeline: Get rid of the no kernel input parameters hack
Jason Ekstrand [Fri, 11 Dec 2015 06:37:27 +0000 (22:37 -0800)]
anv/pipeline: Get rid of the no kernel input parameters hack

Previously, meta would pass null shaders in for the VS when it intended to
disable the VS.  However, this meant that we didn't know what inputs we had
and would dead-code things in the FS.  In order to solve this, we
hard-coded a number.  Now meta passes in a VS even if it plans to disable
the stage so this is no longer needed.

8 years agoanv/apply_pipeline_layout: Multiply uniform sizes by 4
Jason Ekstrand [Fri, 11 Dec 2015 06:36:47 +0000 (22:36 -0800)]
anv/apply_pipeline_layout: Multiply uniform sizes by 4

This is because uniforms are now in terms of bytes everywhere.

8 years agoi965/HACK: Build brw_cs into libcompiler
Jason Ekstrand [Fri, 11 Dec 2015 06:36:05 +0000 (22:36 -0800)]
i965/HACK: Build brw_cs into libcompiler

We need it for CS push constants

8 years agomesa/shader: return correct attribute location for double matrix arrays
Dave Airlie [Thu, 10 Dec 2015 01:44:34 +0000 (11:44 +1000)]
mesa/shader: return correct attribute location for double matrix arrays

If we have a dmat2[4], then dmat2[0] is at 17, dmat2[1] at 19,
dmat2[2] at 21 etc. The old code was returning 17,18,19.

I think this code is also wrong for float matricies as well.

There is now a piglit for the float case.

This partly fixes:
GL41-CTS.vertex_attrib_64bit.limits_test

[airlied: update with Tapani suggestion to clean it up].

Cc: "11.0 11.1" <mesa-stable@lists.freedesktop.org>
Reviewed-by: Timothy Arceri <timothy.arceri@collabora.com>
Reviewed-by: Tapani Pälli <tapani.palli@intel.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>
8 years agogen8/cmd_buffer: Don't push CS constants if there aren't any
Jason Ekstrand [Fri, 11 Dec 2015 02:56:15 +0000 (18:56 -0800)]
gen8/cmd_buffer: Don't push CS constants if there aren't any

Issuing MEDIA_CURB_LOAD with a size of zero causes GPU hangs on BDW.

8 years agoanv: Use 4 instead of sizeof(gl_constant_value)
Jason Ekstrand [Fri, 11 Dec 2015 02:55:14 +0000 (18:55 -0800)]
anv: Use 4 instead of sizeof(gl_constant_value)

We no longer have access to gl_constant_value and, really, it's 4 because
our uniform layout code works entirely in dwords.

8 years agonir/spirv: Put SSBO store writemasks in the right index
Jason Ekstrand [Fri, 11 Dec 2015 02:54:42 +0000 (18:54 -0800)]
nir/spirv: Put SSBO store writemasks in the right index

It moved with the nir_intrinsic_load/store update.

8 years agoMerge remote-tracking branch 'mesa-public/master' into vulkan
Jason Ekstrand [Fri, 11 Dec 2015 00:58:24 +0000 (16:58 -0800)]
Merge remote-tracking branch 'mesa-public/master' into vulkan

This pulls in nir_intrinsic_load/store changes and the switch of all
uniforms in i965 to bytes.  This accounts for the Vulkan changes.

8 years agodraw: fix clipping with linear interpolated values and gl_ClipVertex
Roland Scheidegger [Fri, 11 Dec 2015 01:21:17 +0000 (02:21 +0100)]
draw: fix clipping with linear interpolated values and gl_ClipVertex

Discovered this when working on other clip code, apparently didn't work
correctly - the combination of linear interpolated values and using
gl_ClipVertex produced wrong values (failing all such combinations
in piglits glsl-1.30 interpolation tests, named
interpolation-noperspective-XXX-vertex).
Use the pre-clip-pos values when determining the interpolation factor to
fix this.
Noone really understands this code well, but everybody agrees this looks
sane... This fixes all those failing tests (10 in total) both with
the llvm and non-llvm draw paths, with no piglit regressions.

Reviewed-by: Brian Paul <brianp@vmware.com>
Reviewed-by: Dave Airlie <airlied@redhat.com>
8 years agor600: add missing return value check.
Dave Airlie [Thu, 10 Dec 2015 23:36:36 +0000 (09:36 +1000)]
r600: add missing return value check.

Pointed out by coverity scan.

Signed-off-by: Dave Airlie <airlied@redhat.com>
8 years agoanv/icd: Advertise the right ABI version
Jason Ekstrand [Thu, 10 Dec 2015 20:27:13 +0000 (12:27 -0800)]
anv/icd: Advertise the right ABI version