Axel Davy [Sat, 7 May 2016 09:33:24 +0000 (11:33 +0200)]
st/nine: Turn on thread_submit by default when on different device
The last remaining issues with thread_submit have been resolved,
thus turn it when on a different device (the case where is is
beneficial).
Signed-off-by: Axel Davy <axel.davy@ens.fr>
Axel Davy [Sun, 3 Apr 2016 11:04:39 +0000 (13:04 +0200)]
st/nine: Fix usage of rasterizer multisample bit.
pipe_rasterizer multisample bit should be enabled only when really
wanting to do multisampling, thus we should disable when not having
msaa render target.
This fixes some depth calculation precision issues on radeon.
Also disable it when depth and stencil tests are disabled, since in that
case multisampling is same as not multisampled.
Signed-off-by: Axel Davy <axel.davy@ens.fr>
Axel Davy [Sun, 3 Apr 2016 08:52:22 +0000 (10:52 +0200)]
st/nine: ATOC has effect only with ALPHATESTENABLE
ATOC extension does something only when alpha test is enabled.
Use a second bit to encode the difference with ATIATOC.
Signed-off-by: Axel Davy <axel.davy@ens.fr>
Axel Davy [Sat, 7 May 2016 09:20:47 +0000 (11:20 +0200)]
st/nine: Add debug string for ATOC
We were missing a debug string for this format.
Signed-off-by: Axel Davy <axel.davy@ens.fr>
Axel Davy [Sat, 19 Mar 2016 18:27:34 +0000 (19:27 +0100)]
st/nine: Add asserts for output/input packing
Nine doesn't support vs output/ps input packing.
We haven't found any application requiring that,
and implementing it properly is complex.
Add asserts for now.
Signed-off-by: Axel Davy <axel.davy@ens.fr>
Axel Davy [Mon, 14 Mar 2016 20:29:53 +0000 (21:29 +0100)]
st/nine: Use correct PIPE_HANDLE_USAGE flag for frontbuffer copy
When taking screenshots we do a copy from the frontbuffer
to an allocated buffer (which we then copy to a ram buffer).
Signed-off-by: Axel Davy <axel.davy@ens.fr>
Axel Davy [Sat, 12 Mar 2016 11:24:51 +0000 (12:24 +0100)]
st/nine: Fix output shift calculation
We were getting it wrong for negative values.
Signed-off-by: Axel Davy <axel.davy@ens.fr>
Axel Davy [Fri, 11 Mar 2016 22:30:05 +0000 (23:30 +0100)]
st/nine: Fix CheckDeviceFormat advertising for surfaces
Signed-off-by: Axel Davy <axel.davy@ens.fr>
Axel Davy [Fri, 11 Mar 2016 22:03:56 +0000 (23:03 +0100)]
st/nine: Improve buffer placement
Signed-off-by: Axel Davy <axel.davy@ens.fr>
Axel Davy [Fri, 11 Mar 2016 21:22:10 +0000 (22:22 +0100)]
st/nine: Fix buffer bind flags
Signed-off-by: Axel Davy <axel.davy@ens.fr>
Axel Davy [Fri, 11 Mar 2016 21:17:25 +0000 (22:17 +0100)]
st/nine: Fix buffer locking flags handling
Our behaviour was not entirely similar to what
the docs and our tests describe.
Drop d3dlock_buffer_to_pipe_transfer_usage.
Signed-off-by: Axel Davy <axel.davy@ens.fr>
Patrick Rudolph [Wed, 9 Mar 2016 14:15:46 +0000 (15:15 +0100)]
st/nine: Improve logging
Add missing DBG calls in dtors.
Signed-off-by: Patrick Rudolph <siro@das-labor.org>
Reviewed-by: Axel Davy <axel.davy@ens.fr>
Patrick Rudolph [Mon, 22 Feb 2016 18:49:19 +0000 (19:49 +0100)]
st/nine: Use WINE thread for threadpool
Use present interface 1.2 function ID3DPresent_CreateThread
to create the thread for threadpool.
Creating the thread with WINE prevents some rarely occuring crashes.
Signed-off-by: Patrick Rudolph <siro@das-labor.org>
Reviewed-by: Axel Davy <axel.davy@ens.fr>
Patrick Rudolph [Wed, 10 Feb 2016 18:41:12 +0000 (19:41 +0100)]
st/nine: Don't present if window is occluded
The problem is that if one d3d present call fails,
because of our occlusion check in present method,
the next presentation call will send the same pixmap to the Xserver again,
without waiting it is released, which is wrong.
Move the present call after occlusion check to return and prevent
Xpixmaps errors.
Signed-off-by: Patrick Rudolph <siro@das-labor.org>
Reviewed-by: Axel Davy <axel.davy@ens.fr>
Patrick Rudolph [Wed, 10 Feb 2016 18:30:20 +0000 (19:30 +0100)]
st/nine: Use new function to query for resolution mismatch
Any third party app might change the current screen resolution.
Poll for resolution mismatch to force a device reset.
Required for non ex devices only.
Signed-off-by: Patrick Rudolph <siro@das-labor.org>
Reviewed-by: Axel Davy <axel.davy@ens.fr>
Patrick Rudolph [Sat, 27 Feb 2016 08:58:19 +0000 (09:58 +0100)]
st/nine: Implement IPresent version 1.2
Implement presentation interface version 1.2:
* ID3DPresent_ResolutionMismatch
Poll for resolution mismatch.
A third party app might have changed resolution,
which requires a device reset.
* ID3DPresent_CreateThread
Create a thread in WINE to allow nine to use Windows API
functions. Required for multi-threaded presentation.
In single-threaded presentation mode the calling thread is
already known to WINE.
* ID3DPresent_WaitForThread
Wait for a wine thread to terminate.
Signed-off-by: Patrick Rudolph <siro@das-labor.org>
Reviewed-by: Axel Davy <axel.davy@ens.fr>
Axel Davy [Mon, 7 Mar 2016 22:36:44 +0000 (23:36 +0100)]
st/nine: Implement BumpEnvMap for ff
Signed-off-by: Axel Davy <axel.davy@ens.fr>
Axel Davy [Mon, 7 Mar 2016 22:03:25 +0000 (23:03 +0100)]
st/nine: Format conversion for volumes in UpdateTexture
We were doing the conversion for surfaces, but not yet
volumes. Now that volumes can do conversion, use it.
Signed-off-by: Axel Davy <axel.davy@ens.fr>
Axel Davy [Mon, 7 Mar 2016 22:01:02 +0000 (23:01 +0100)]
st/nine: Remove one useless function output
Signed-off-by: Axel Davy <axel.davy@ens.fr>
Axel Davy [Mon, 7 Mar 2016 21:30:05 +0000 (22:30 +0100)]
st/nine: Add support for X8L8V8U8
X8L8V8U8 support should be common. Some more recent cards
do support this format, but not L6V5U5.
Add fallback for this format to have it alwaus supported.
L6V5U5 conversion rule apparently differs a bit from the normal
spec, and thus the gallium equivalent format leads to slightly
wrong colors. Since some recent cards do not support it, do not
support it either.
Signed-off-by: Axel Davy <axel.davy@ens.fr>
Axel Davy [Mon, 7 Mar 2016 21:18:16 +0000 (22:18 +0100)]
st/nine: Add format fallback with conversion to volumes
Signed-off-by: Axel Davy <axel.davy@ens.fr>
Axel Davy [Mon, 7 Mar 2016 20:55:30 +0000 (21:55 +0100)]
st/nine: Add format fallback with conversion to surfaces
Signed-off-by: Axel Davy <axel.davy@ens.fr>
Axel Davy [Mon, 7 Mar 2016 20:35:59 +0000 (21:35 +0100)]
gallium/util: Implement util_format_translate_3d
This is the equivalent of util_format_translate, but for volumes.
Signed-off-by: Axel Davy <axel.davy@ens.fr>
Axel Davy [Sun, 6 Mar 2016 14:54:50 +0000 (15:54 +0100)]
st/nine: Fix Pointsize in programmable shader
Signed-off-by: Axel Davy <axel.davy@ens.fr>
Axel Davy [Sun, 6 Mar 2016 14:14:30 +0000 (15:14 +0100)]
st/nine: Fix ff pointscale computation
Signed-off-by: Axel Davy <axel.davy@ens.fr>
Axel Davy [Sat, 5 Mar 2016 18:06:29 +0000 (19:06 +0100)]
st/nine: Fix header of GetIndices
There is a mistake in the online documentation,
the function only has 2 arguments.
Signed-off-by: Axel Davy <axel.davy@ens.fr>
Axel Davy [Sat, 5 Mar 2016 11:17:06 +0000 (12:17 +0100)]
st/nine: Increase minor d3dadapter9drm ABI
Version 0.1 allows to assume that the second
element of the IDirect3D* structures will
be a pointer to the internal nine vtable.
This is useful if the gallium nine user wants
to wrap some interfaces.
Signed-off-by: Axel Davy <axel.davy@ens.fr>
Axel Davy [Fri, 12 Feb 2016 18:08:42 +0000 (19:08 +0100)]
st/nine: Fix leak after ctor failures
Previously ctor failures would not unreference
the device.
Signed-off-by: Axel Davy <axel.davy@ens.fr>
Axel Davy [Tue, 9 Feb 2016 22:51:39 +0000 (23:51 +0100)]
st/nine: Add ColorFill test for compressed textures
ColorFill should contain alignment checks
for compressed textures.
Signed-off-by: Axel Davy <axel.davy@ens.fr>
Axel Davy [Tue, 9 Feb 2016 22:33:51 +0000 (23:33 +0100)]
st/nine: PositionT and Tessfactor are forbidden as PS input
According to wine tests, they are forbidden as PS input,
which makes sense.
Signed-off-by: Axel Davy <axel.davy@ens.fr>
Axel Davy [Tue, 9 Feb 2016 22:15:21 +0000 (23:15 +0100)]
st/nine: Fix some shader failures not triggering error
Some failures during shader translation would not
raise errors before this patch.
Signed-off-by: Axel Davy <axel.davy@ens.fr>
Axel Davy [Sun, 21 Feb 2016 15:38:23 +0000 (16:38 +0100)]
st/nine: Forbid POSITION0 for PS3.0
POSITION0 input is forbidden for PS3.0 apparently.
Signed-off-by: Axel Davy <axel.davy@ens.fr>
Axel Davy [Tue, 9 Feb 2016 21:35:27 +0000 (22:35 +0100)]
st/nine: Rework UpdateTexture Checks
Our code did match the user documentation of the function
quite well (except for format check).
However the DDI documentation and wine tests show that
documentation was not correct. Thus adapt our code to
fit the best possible to the -real- spec.
Signed-off-by: Axel Davy <axel.davy@ens.fr>
Axel Davy [Tue, 9 Feb 2016 07:20:34 +0000 (08:20 +0100)]
st/nine: Use bufs instead of Flags for Clear
bufs doesn't contain depthstencil if
there is z buffer mismatch. This is the behaviour
we want.
Signed-off-by: Axel Davy <axel.davy@ens.fr>
Axel Davy [Sun, 10 Apr 2016 21:55:19 +0000 (23:55 +0200)]
d3dadapter9: Add ddebug, rbug and trace support
Add support for ddebug, rbug and trace
Signed-off-by: Axel Davy <axel.davy@ens.fr>
Axel Davy [Sun, 3 Apr 2016 20:36:01 +0000 (22:36 +0200)]
radeon: Change AA sample locations for EG+
This sets the AA location to the d3d11
spec.
EG/NI 8X MSAA is left as is. Not sure
why it was set different to Cayman, so
lets it as is.
Signed-off-by: Axel Davy <axel.davy@ens.fr>
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Axel Davy [Mon, 7 Mar 2016 21:26:22 +0000 (22:26 +0100)]
radeonsi: Mixed colorbuffer formats are unsupported
Besides depth/stencil, the hardware doesn't support
mixed formats.
The GL state tracker doesn't make use of them.
Signed-off-by: Axel Davy <axel.davy@ens.fr>
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Axel Davy [Sat, 19 Mar 2016 18:57:00 +0000 (19:57 +0100)]
radeonsi: Change default behaviour for undefined COLOR0
d3d 9 needs COLOR0 to be 1.0 on all channels when
undefined. 0.0 for the others is fine.
GL behaviour is undefined.
Signed-off-by: Axel Davy <axel.davy@ens.fr>
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Axel Davy [Sun, 3 Apr 2016 19:47:11 +0000 (21:47 +0200)]
r600g: Change default behaviour for undefined COLOR0
d3d 9 needs COLOR0 to be 1.0 on all channels when
undefined. 0.0 for the others is fine.
GL behaviour is undefined.
Signed-off-by: Axel Davy <axel.davy@ens.fr>
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Axel Davy [Sat, 19 Mar 2016 18:55:24 +0000 (19:55 +0100)]
r600: Change default behaviour for undefined COLOR0
d3d 9 needs COLOR0 to be 1.0 on all channels when
undefined. 0.0 for the others is fine.
GL behaviour is undefined.
Signed-off-by: Axel Davy <axel.davy@ens.fr>
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Christian Schmidbauer [Fri, 29 Apr 2016 19:53:30 +0000 (20:53 +0100)]
st/nine: Clean up WINAPI definition
As Emil pointed out, only gcc, clang and MSVC compatibility is required.
Hence the check for GNUC can be skipped, as __i386__ and __x86_64__ are
only defined for gcc/clang, not for MSVC.
Remove the #undef which has been there for historic reasons, when wine
dlls for nine have been built inside mesa. Instead use #ifndef in order
to avoid redefining WINAPI from MSVC's headers.
Reviewed-by: Emil Velikov <emil.velikov@collabora.com>
Acked-by: Axel Davy <axel.davy@ens.fr>
Brian Paul [Wed, 18 May 2016 19:01:03 +0000 (13:01 -0600)]
svga: add another debug_printf() in svga_screen_create()
Signed-off-by: Brian Paul <brianp@vmware.com>
Brian Paul [Wed, 18 May 2016 01:28:37 +0000 (19:28 -0600)]
spirv: add switch case for nir_texop_txf_ms_mcs in vtn_handle_texture()
Mark it as unreachable. Silences a compiler warning:
spirv/spirv_to_nir.c:1397:4: warning: enumeration value
'nir_texop_txf_ms_mcs' not handled in switch [-Wswitch]
switch (instr->op) {
^
Reviewed-by: Jason Ekstrand <jason.ekstrand@intel.com>
Matt Turner [Wed, 18 May 2016 19:48:20 +0000 (12:48 -0700)]
Revert "i965/urb: fixes division by zero"
This reverts commit
2a8aa1e3deb99a1ae16d942318da648c1327ece5.
Ardinartsev Nikita [Mon, 16 May 2016 23:27:22 +0000 (02:27 +0300)]
i965/urb: fixes division by zero
Fixes regression introduced by
af5ca43f2676bff7499f93277f908b681cb821d0
Reviewed-by: Matt Turner <mattst88@gmail.com>
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=95419
Matt Turner [Mon, 16 May 2016 22:31:00 +0000 (15:31 -0700)]
mesa: fclose() filename on error.
Pretty useless, as it's in debugging code. Found by Coverity (CID
1257016).
Matt Turner [Mon, 16 May 2016 22:22:59 +0000 (15:22 -0700)]
i965/fs: Assert that nir_op_extract_*'s src1 is a constant.
Matt Turner [Mon, 16 May 2016 21:49:38 +0000 (14:49 -0700)]
glsl: Check that layout is non-null before dereferencing.
layout should only be null for structs, but it's checked everywhere else
and confuses Coverity (CID
1358495).
Reviewed-by: Timothy Arceri <timothy.arceri@collabora.com>
Matt Turner [Mon, 16 May 2016 21:43:26 +0000 (14:43 -0700)]
egl/dri2: Don't check return result of mtx_unlock().
Coverity (CID
1358496) warns that the cleanup code doesn't unlock the
mutex (which is arguably kind of stupid, since the only case that can
happen is when mtx_unlock() failed!). But, mtx_unlock() isn't going to
fail -- the mutex was locked by this thread just a few lines above it.
Matt Turner [Mon, 16 May 2016 19:48:00 +0000 (12:48 -0700)]
spirv: Properly size the src[] array.
Operations like nir_op_bitfield_insert have four arguments, and Coverity
isn't privy to the fact that 4-argument operations aren't possible here,
so it thinks this can lead to memory corruption. Just increase the size
of the array to quell any fears.
Matt Turner [Mon, 16 May 2016 18:12:24 +0000 (11:12 -0700)]
isl: Mark default cases in switch unreachable.
To silence -Wmaybe-uninitialized warnings.
Ian Romanick [Tue, 17 May 2016 20:49:11 +0000 (13:49 -0700)]
glsl/linker: Ensure the first stage of an SSO pipeline has input locs assigned
Previously an SSO pipeline containing only a tessellation control shader
and a tessellation evaluation shader would not get locations assigned
for the TCS inputs. This would lead to assertion failures in some
piglit tests, such as arb_program_interface_query-resource-query.
That piglit test still fails on some tessellation related subtests.
Specifically, these subtests fail:
'GL_PROGRAM_INPUT(tcs) active resources' expected 2 but got 3
'GL_PROGRAM_INPUT(tcs) max length name' expected 12 but got 16
'GL_PROGRAM_INPUT(tcs,tes) active resources' expected 2 but got 3
'GL_PROGRAM_INPUT(tcs,tes) max length name' expected 12 but got 16
'GL_PROGRAM_OUTPUT(tcs) active resources' expected 15 but got 3
'GL_PROGRAM_OUTPUT(tcs) max length name' expected 23 but got 12
Signed-off-by: Ian Romanick <ian.d.romanick@intel.com>
Reviewed-by: Timothy Arceri <timothy.arceri@collabora.com>
Cc: mesa-stable@lists.freedesktop.org
Ian Romanick [Tue, 17 May 2016 20:30:46 +0000 (13:30 -0700)]
glsl/linker: Don't include interface name for built-in blocks
Commit
11096ec introduced a regression in some piglit tests (e.g.,
arb_program_interface_query-resource-query). I did not notice this
regression because other (unrelated) problems caused failed assertions
in those same tests on my system... so they crashed before getting to
the new failure.
v2: Use is_gl_identifier. Suggested by Tim.
Signed-off-by: Ian Romanick <ian.d.romanick@intel.com>
Reviewed-by: Timothy Arceri <timothy.arceri@collabora.com>
Cc: mesa-stable@lists.freedesktop.org
Ian Romanick [Tue, 17 May 2016 01:01:10 +0000 (18:01 -0700)]
glsl: Assert that inputs have a location assigned
This catches a problem previously undetected until deep in the backend.
Signed-off-by: Ian Romanick <ian.d.romanick@intel.com>
Reviewed-by: Timothy Arceri <timothy.arceri@collabora.com>
Ian Romanick [Tue, 17 May 2016 22:01:19 +0000 (15:01 -0700)]
glsl/linker: Fix trivial typos in comments
Signed-off-by: Ian Romanick <ian.d.romanick@intel.com>
Reviewed-by: Timothy Arceri <timothy.arceri@collabora.com>
Ian Romanick [Tue, 17 May 2016 00:58:55 +0000 (17:58 -0700)]
glsl/linker: Fix some formatting to match current coding conventions
Signed-off-by: Ian Romanick <ian.d.romanick@intel.com>
Reviewed-by: Timothy Arceri <timothy.arceri@collabora.com>
Ian Romanick [Mon, 16 May 2016 19:48:06 +0000 (12:48 -0700)]
glsl/linker: Silence unused parameter warning
The use of the parameter was removed in
d6b92028.
glsl/link_varyings.cpp:1390:39: warning: unused parameter ‘separate_shader’ [-Wunused-parameter]
bool separate_shader)
^
Signed-off-by: Ian Romanick <ian.d.romanick@intel.com>
Reviewed-by: Timothy Arceri <timothy.arceri@collabora.com>
Ian Romanick [Tue, 17 May 2016 20:54:09 +0000 (13:54 -0700)]
glsl/linker: Silence unused parameter warning
The parameter appears to have been unused since the function was added
in commit
12ba6cfb. Remove it.
glsl/linker.cpp:2886:60: warning: unused parameter ‘prog’ [-Wunused-parameter]
match_explicit_outputs_to_inputs(struct gl_shader_program *prog,
^
Signed-off-by: Ian Romanick <ian.d.romanick@intel.com>
Reviewed-by: Timothy Arceri <timothy.arceri@collabora.com>
Ian Romanick [Fri, 13 May 2016 00:44:46 +0000 (17:44 -0700)]
i965: Silence unused parameter warnings
The only place that actually used the type parameter was the GS visitor,
and it was always passed glsl_type::int. Just remove the parameter.
brw_vec4_vs_visitor.cpp:38:61: warning: unused parameter ‘type’ [-Wunused-parameter]
const glsl_type *type)
^
Signed-off-by: Ian Romanick <ian.d.romanick@intel.com>
Reviewed-by: Timothy Arceri <timothy.arceri@collabora.com>
Daniel Scharrer [Tue, 19 Apr 2016 14:57:06 +0000 (16:57 +0200)]
mesa: Don't advertise GLES 3.1 without compute support
The MaxComputeWorkGroupInvocations constant is used in
compute_version_es2() instead of extensions->ARB_compute_shader
as ES has lower requirements than desktop GL.
Both i965 and gallium set this constant before enabling compute support.
Signed-off-by: Daniel Scharrer <daniel@constexpr.org>
Signed-off-by: Marek Olšák <marek.olsak@amd.com>
Rob Clark [Wed, 18 May 2016 13:19:00 +0000 (09:19 -0400)]
mesa/st: don't leak name
Pointed out by coverity.
Signed-off-by: Rob Clark <robclark@freedesktop.org>
Brian Paul [Fri, 6 May 2016 15:46:29 +0000 (09:46 -0600)]
svga: null out all sampler views if start=num=0
Because the CSO module handles sampler views for fragment shaders
differently than vertex/geom shaders, VS/GS shader sampler views
aren't explicitly unbound like for FS sampler vers. This code
checks for the case of start=num=0 and nulls out the sampler views.
Fixes a assert regression in piglit's arb_texture_multisample-
sample-position test.
Reviewed-by: Charmaine Lee <charmainel@vmware.com>
Brian Paul [Tue, 10 May 2016 19:01:26 +0000 (13:01 -0600)]
st/mesa: remove unused st_context::default_texture
The code which used this was removed quite a while ago.
Reviewed-by: Ilia Mirkin <imirkin@alum.mit.edu>
Reviewed-by: Sinclair Yeh <syeh@vmware.com>
Reviewed-by: Charmaine Lee <charmainel@vmware.com>
Brian Paul [Fri, 6 May 2016 15:20:23 +0000 (09:20 -0600)]
cso: remove / add some comments
Signed-off-by: Brian Paul <brianp@vmware.com>
Eric Anholt [Tue, 17 May 2016 19:57:44 +0000 (12:57 -0700)]
vc4: Add support for vertex color clamping in the rasterizer.
This gets us precompile of vertex shaders at the state tracker level as
well.
Eric Anholt [Tue, 17 May 2016 19:46:27 +0000 (12:46 -0700)]
vc4: Move tgsi_to_nir to precompile time.
Now we have an immutable nir shader in our shader's CSO that we can clone
and lower/optimize.
Eric Anholt [Tue, 17 May 2016 19:15:54 +0000 (12:15 -0700)]
vc4: Mark the driver as supporting fragment color clamping in rast.
We always clamp fragment colors, since they're always 8-bit unorm, so
there's no need to have us compile separate shaders based on
GL_ARB_color_buffer_float. This gives us precompilation of fragment
programs to the vc4_shader_state_create() level.
Eric Anholt [Tue, 17 May 2016 19:24:06 +0000 (12:24 -0700)]
vc4: Enable sharing shaders across contexts.
This allows the same pipe_shader_state to be referenced from multiple
contexts. Since our pipe_shader_state is treated as immutable (other than
the variant number) within the driver, this is no problem.
Eric Anholt [Tue, 17 May 2016 22:30:59 +0000 (15:30 -0700)]
vc4: Switch to using nir_load_front_face.
This will be generated by glsl_to_nir, and it turns out that this is a
more code-efficient path than the floating point math, anyway.
No change on shader-db, but drops an instruction in piglit's
glsl-fs-frontfacing.
Eric Anholt [Tue, 17 May 2016 19:35:44 +0000 (12:35 -0700)]
vc4: Drop the dead export_linkage array.
This came from deriving from freedreno.
Eric Anholt [Tue, 17 May 2016 21:06:39 +0000 (14:06 -0700)]
vc4: Fix a -Wformat-security warning.
This is apparently enabled as an error in Android builds, and the compiler
can't tell that the return value is safe.
Alex Deucher [Tue, 17 May 2016 21:10:50 +0000 (17:10 -0400)]
radeonsi: add new polaris11 pci ids
Reviewed-by: Nicolai Hähnle <nicolai.haehnle@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Alex Deucher [Tue, 17 May 2016 21:06:26 +0000 (17:06 -0400)]
radeonsi: add new polaris10 pci ids
Reviewed-by: Nicolai Hähnle <nicolai.haehnle@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Kenneth Graunke [Fri, 13 May 2016 06:04:10 +0000 (23:04 -0700)]
i965: Make brw_reg_from_fs_reg() halve exec_size when compressed.
In
a5d7e144eaf43fee37e6ff9e2de194407087632b, Connor generalized the
exec_size halving code to handle more cases. As part of this, he made
it not halve anything if the region accessed falls completely in a
single register.
Unfortunately, it started producing some invalid regions:
-add(16) g6<1>F g10<8,8,1>UW -g1<0,1,0>F { align1 compr };
-add(16) g8<1>F g12<8,8,1>UW -g1.1<0,1,0>F { align1 compr };
+add(16) g6<1>F g10<16,16,1>UW -g1<0,1,0>F { align1 compr };
+add(16) g8<1>F g12<16,16,1>UW -g1.1<0,1,0>F { align1 compr };
Here, the UW source region completely fits within a register. However,
we have to use instruction compression because the destination region
spans two registers. <16,16,1> is invalid because it's compressed.
To handle this, skip the "everything fits in one register" case and
fall through to the exec_size halving case when compressed.
Fixes hundreds of Piglit regressions on GM965.
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=95370
Signed-off-by: Kenneth Graunke <kenneth@whitecape.org>
Reviewed-by: Matt Turner <mattst88@gmail.com>
Kenneth Graunke [Mon, 16 May 2016 22:09:17 +0000 (15:09 -0700)]
i965: Move compression decisions before brw_reg_from_fs_reg().
brw_reg_from_fs_reg() needs to know whether the instruction will be
compressed or not.
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=95370
Signed-off-by: Kenneth Graunke <kenneth@whitecape.org>
Reviewed-by: Matt Turner <mattst88@gmail.com>
Kenneth Graunke [Mon, 4 Apr 2016 06:03:34 +0000 (23:03 -0700)]
i965: Enable ES 3.2 sample shading extensions.
This enables:
- GL_OES_sample_shading
- GL_OES_sample_variables
- GL_OES_shader_multisample_interpolation
On Gen8, we pass all the CTS tests, and all but 4 of the dEQP-GLES31
tests (dealing with 1x/2x MSAA at half rate sampling). We believe
those 4 dEQP-GLES31 tests are incorrect.
Signed-off-by: Kenneth Graunke <kenneth@whitecape.org>
Reviewed-by: Jason Ekstrand <jason@jlekstrand.net>
Reviewed-by: Ian Romanick <ian.d.romanick@intel.com>
Jordan Justen [Tue, 17 May 2016 21:06:39 +0000 (14:06 -0700)]
anv: Fix warning: unused variable ‘cs_prog_data’
This was introduced in
8a80af282091e692da7bf4e412918ba2362dfb4f.
Reported-by: Jason Ekstrand <jason@jlekstrand.net>
Signed-off-by: Jordan Justen <jordan.l.justen@intel.com>
Mauro Rossi [Tue, 17 May 2016 20:48:36 +0000 (22:48 +0200)]
android: fix building error in libmesa_st_mesa
Fixes the following building error due to libmesa_nir dependency:
In file included from external/mesa/src/mesa/state_tracker/st_glsl_to_nir.cpp:44:0:
external/mesa/src/compiler/nir/nir.h:42:25: fatal error: nir_opcodes.h: No such file or directory
#include "nir_opcodes.h"
^
compilation terminated.
build/core/binary.mk:706: recipe for target 'out/target/product/x86/obj/STATIC_LIBRARIES/libmesa_st_mesa_intermediates/state_tracker/st_glsl_to_nir.o' failed
make: *** [out/target/product/x86/obj/STATIC_LIBRARIES/libmesa_st_mesa_intermediates/state_tracker/st_glsl_to_nir.o] Error 1
make: *** Waiting for unfinished jobs....
Reviewed-by: Rob Herring <robh@kernel.org>
Signed-off-by: Rob Clark <robclark@freedesktop.org>
Nicolai Hähnle [Thu, 12 May 2016 17:44:44 +0000 (12:44 -0500)]
radeonsi: force level zero on image instructions in non-fragment shaders (v2)
Section 8.9 (Texture Functions) of the OpenGL Shading Language 4.5
specification:
However, automatic level of detail is computed only for fragment shaders.
Other shaders operate as though the base level of detail were computed as
zero.
and Section 8.9.3 (Texture Gather Functions):
When performing a texture gather operation, the minification and
magnification filters are ignored, and the rules for LINEAR filtering in
the OpenGL Specification are applied to the base level of the texture
image to identify the four texels i_0 j_1, i_1 j_1, i_1 j_0, and i_0 j_0.
Of course, explicit LOD or derivative variants work in all shader types.
This fixes several GL4x-CTS.texture_gather.* tests.
v2: TG4 is always level zero (thanks, Ilia)
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Nicolai Hähnle [Thu, 12 May 2016 16:52:36 +0000 (11:52 -0500)]
radeonsi: emit TXQ in separate functions
TXQ is sufficiently different that having in it in the same code path as
texture sampling/fetching opcodes doesn't make much sense.
v2: guard against NULL pointer dereferences
Reviewed-by: Marek Olšák <marek.olsak@amd.com> (v1)
Nicolai Hähnle [Fri, 6 May 2016 03:08:39 +0000 (22:08 -0500)]
winsys/amdgpu: cleanup error handling in amdgpu_ctx_create
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Nicolai Hähnle [Thu, 5 May 2016 21:35:09 +0000 (16:35 -0500)]
winsys/amdgpu: avoid ioctl call when fence_wait is called without timeout
When user fences are used, we don't need the kernel for polling.
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Nicolai Hähnle [Fri, 6 May 2016 22:02:30 +0000 (17:02 -0500)]
gallium/radeon: add radeon_emitted to check for non-trivial IBs
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Nicolai Hähnle [Fri, 6 May 2016 22:07:10 +0000 (17:07 -0500)]
gallium/radeon: use radeon_emit_array
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Nicolai Hähnle [Fri, 6 May 2016 21:42:03 +0000 (16:42 -0500)]
gallium/radeon: use radeon_emit
Mostly generated using a sed-script, with manual fix-up for multi-line
statements.
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Nicolai Hähnle [Fri, 13 May 2016 05:12:37 +0000 (00:12 -0500)]
st/mesa: fix reversed copyimage canonical format
The format_desc swizzle describes where in the array each color channel
comes from - but the existing code was written as if each entry in the
swizzle described the meaning of an array element.
Fixes piglit's arb_copy_image-format-swizzle.
Cc: "11.1 11.2" <mesa-stable@lists.freedesktop.org>
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Jordan Justen [Thu, 12 May 2016 01:03:15 +0000 (18:03 -0700)]
Revert "HACK: Don't re-configure L3$ in render stages pre-BDW"
This reverts commit
41af9b2e517dd0c17e519490ca915b96f6898390.
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=94468
Reviewed-by: Jason Ekstrand <jason@jlekstrand.net>
Jordan Justen [Sat, 2 Apr 2016 08:34:40 +0000 (01:34 -0700)]
anv: Port L3 cache programming from i965
Signed-off-by: Jordan Justen <jordan.l.justen@intel.com>
Acked-by: Jason Ekstrand <jason@jlekstrand.net>
Jordan Justen [Tue, 12 Apr 2016 17:15:38 +0000 (10:15 -0700)]
anv/gen7: Add memory barrier to vkCmdWaitEvents call
We also have this barrier call for gen8 vkCmdWaitEvents.
We don't implement waiting on events for gen7 yet, but this barrier at
least helps to not regress CTS cases when data caching is enabled.
Without this, the tests would intermittently report a failure when the
data cache was enabled.
Signed-off-by: Jordan Justen <jordan.l.justen@intel.com>
Reviewed-by: Jason Ekstrand <jason@jlekstrand.net>
Jordan Justen [Sat, 2 Apr 2016 20:44:55 +0000 (13:44 -0700)]
anv: Keep track of whether the data cache should be enabled in L3
If images or shader buffers are used, we will enable the data cache in
the the L3 config.
Signed-off-by: Jordan Justen <jordan.l.justen@intel.com>
Reviewed-by: Jason Ekstrand <jason@jlekstrand.net>
Jordan Justen [Sat, 2 Apr 2016 08:25:05 +0000 (01:25 -0700)]
genxml/hsw: Add L3 cache control registers
These were added to the i965 driver in
5912da45a69923afa1b7f2eb5bb371d848813c41.
Signed-off-by: Jordan Justen <jordan.l.justen@intel.com>
Reviewed-by: Jason Ekstrand <jason@jlekstrand.net>
Jan Vesely [Tue, 17 May 2016 13:25:44 +0000 (09:25 -0400)]
Treewide: Remove Elements() macro
Signed-off-by: Jan Vesely <jano.vesely@gmail.com>
Reviewed-by: Brian Paul <brianp@vmware.com>
Jan Vesely [Tue, 17 May 2016 17:29:40 +0000 (13:29 -0400)]
r600g,sb: Don't use standard macro name
Signed-off-by: Jan Vesely <jan.vesely@rutgers.edu>
Jason Ekstrand [Mon, 16 May 2016 18:02:57 +0000 (11:02 -0700)]
anv/formats: Add support for VK_FORMAT_B4G4R4A4_UNORM pre-gen8
Jason Ekstrand [Mon, 16 May 2016 17:25:54 +0000 (10:25 -0700)]
anv: Add a devinfo argument to the get_format functions
Jason Ekstrand [Mon, 16 May 2016 04:50:47 +0000 (21:50 -0700)]
anv/formats: Set the swizzle to RGB1 when using an RGBA format to fake RGB
This way we get correct sampling from RGB formats that are faked as RGBA.
This should also cause it to disable rendering and blending on those
formats. We should be able to render to them and, on Broadwell and above,
we can blend on them with work-arounds. However, we'll add support for
that more properly later when it's deemed useful. For now, disabling
rendering and blending should be safe.
Jason Ekstrand [Mon, 16 May 2016 04:46:05 +0000 (21:46 -0700)]
anv/formats: Refactor anv_get_format
The new code removes the switch statement and instead handles depth/stencil
as up-front special cases. This allows for potentially more complicated
color format handling in the future.
Jason Ekstrand [Mon, 16 May 2016 04:41:55 +0000 (21:41 -0700)]
anv: Use 16 bits for the isl_format in anv_format
This way the entire anv_format structure fits in 32 bits
Jason Ekstrand [Mon, 16 May 2016 04:31:38 +0000 (21:31 -0700)]
anv/formats: Use the isl_channel_select enum for the swizzle
Jason Ekstrand [Mon, 16 May 2016 04:15:59 +0000 (21:15 -0700)]
anv/formats: Add an anv_get_format helper
This commit removes anv_format_for_vk_format and adds an anv_get_format
helper. The anv_get_format helper returns the anv_format by-value. Unlike
anv_format_for_vk_format the format returned by anv_get_format is 100%
accurate and includes any tweaks needed for tiled vs. linear.
anv_get_isl_format is now just a wrapper around anv_get_format that picks
off just the isl_format.