Axel Davy [Sun, 15 Feb 2015 15:22:43 +0000 (16:22 +0100)]
st/nine: Avoid useless updates in SetSamplerState
Check for redundant settings
Signed-off-by: Axel Davy <axel.davy@ens.fr>
Reviewed-by: David Heidelberg <david@ixit.cz>
Axel Davy [Tue, 24 Mar 2015 09:16:59 +0000 (10:16 +0100)]
st/nine: Rework rasterizer states
Separate state preparation and state commit
Signed-off-by: Axel Davy <axel.davy@ens.fr>
Axel Davy [Fri, 15 May 2015 21:52:09 +0000 (23:52 +0200)]
st/nine: Reorder DSA state settings
Separate state preparation and state commit
Signed-off-by: Axel Davy <axel.davy@ens.fr>
Axel Davy [Tue, 24 Mar 2015 08:35:18 +0000 (09:35 +0100)]
st/nine: Reorder nine_state.
Instead of mixing state preparation (filling pipe_****)
and state commit (pipe->set_*****),
begin doing so in two separate functions.
This will allow to implement efficient Stateblocks,
and eventually lead to optimisation where the complete
pipe_*** structure is only partially updated.
Signed-off-by: Axel Davy <axel.davy@ens.fr>
Axel Davy [Sun, 15 Feb 2015 08:19:16 +0000 (09:19 +0100)]
st/nine: Remove group_mask argument from nine_update_state
It was only used to discriminate update framebuffer vs update
everything. Instead use two functions.
Signed-off-by: Axel Davy <axel.davy@ens.fr>
Tiziano Bacocco [Fri, 30 Jan 2015 19:10:38 +0000 (20:10 +0100)]
st/nine: Implement TEXBEM,TEXBEML and BEM
Signed-off-by: Tiziano Bacocco <tizbac2@gmail.com>
Patrick Rudolph [Sun, 17 May 2015 10:46:42 +0000 (12:46 +0200)]
st/nine: Fix use of uninitialized values
Set all values to 0 after allocation. Found using valgrind.
Reviewed-by: Axel Davy <axel.davy@ens.fr>
Signed-off-by: Patrick Rudolph <siro@das-labor.org>
Patrick Rudolph [Sat, 16 May 2015 17:47:00 +0000 (19:47 +0200)]
st/nine: Prevent possible crash
In case NineBaseTexture9_ctor returns an error
This->surfaces[l] might be NULL.
Reviewed-by: Axel Davy <axel.davy@ens.fr>
Signed-off-by: Patrick Rudolph <siro@das-labor.org>
Patrick Rudolph [Tue, 19 May 2015 18:18:29 +0000 (20:18 +0200)]
st/nine: Return correct error codes in NineDevice9_Reset
Allow more than two errors, and return D3DERR_INVALIDCALL
for failed display resolution changes.
Reviewed-by: Axel Davy <axel.davy@ens.fr>
Signed-off-by: Patrick Rudolph <siro@das-labor.org>
Patrick Rudolph [Wed, 13 May 2015 17:51:30 +0000 (19:51 +0200)]
st/nine: Fail on D3DUSAGE_DYNAMIC for D3DPOOL_SCRATCH textures
Texture with pool D3DPOOL_SCRATCH and D3DPOOL_MANAGED
cannot be used with flag D3DUSAGE_DYNAMIC.
Signed-off-by: Patrick Rudolph <siro@das-labor.org>
Axel Davy [Thu, 14 May 2015 15:22:26 +0000 (17:22 +0200)]
st/nine: Fix Lock Checks for Compressed textures
Signed-off-by: Axel Davy <axel.davy@ens.fr>
Axel Davy [Thu, 14 May 2015 15:01:40 +0000 (17:01 +0200)]
st/nine: Impose restrictions on DXTN texture sizes
This is the expected behaviour.
Fixes wine tests.
Signed-off-by: Axel Davy <axel.davy@ens.fr>
Reviewed-by: David Heidelberg <david@ixit.cz>
Patrick Rudolph [Wed, 13 May 2015 17:43:04 +0000 (19:43 +0200)]
st/nine: Return NULL pointer in lock error cases
Tests showed, that in case of errors, the pBits pointer is set to NULL.
The pBits field isn't set to NULL in case of an already locked object.
Reviewed-by: Axel Davy <axel.davy@ens.fr>
Signed-off-by: Patrick Rudolph <siro@das-labor.org>
Patrick Rudolph [Wed, 13 May 2015 05:19:11 +0000 (07:19 +0200)]
st/nine: Fix resource SetPriority/GetPriority
Return 0 for non MANAGED textures and surfaces.
Fixes failing wine d3d9 tests device.c test_resource_priority.
Reviewed-by: Axel Davy <axel.davy@ens.fr>
Signed-off-by: Patrick Rudolph <siro@das-labor.org>
Patrick Rudolph [Tue, 12 May 2015 18:28:17 +0000 (20:28 +0200)]
st/nine: Clean GetPrivateData
Move the assert to return error codes in the correct order.
Always set the pSizeOfData to the required buffer size.
Fixes failing wine test device.c test_private_data()
Reviewed-by: Axel Davy <axel.davy@ens.fr>
Signed-off-by: Patrick Rudolph <siro@das-labor.org>
Patrick Rudolph [Mon, 20 Apr 2015 17:26:55 +0000 (19:26 +0200)]
st/nine: Allow lock coordinates outside range
This fixes wine test device.c test_lockrect_invalid()
Mimic WindowsXp behaviour and allow negative values in the rectangle passed.
Add comment to point out behaviour used.
Reviewed-by: Axel Davy <axel.davy@ens.fr>
Signed-off-by: Patrick Rudolph <siro@das-labor.org>
Patrick Rudolph [Tue, 12 May 2015 18:07:09 +0000 (20:07 +0200)]
st/nine: Fix GenerateMipSubLevels potential crash
For the case of D3DPOOL_MANAGED textures, This->base.resource can be NULL
at the start of the function. In This case, UploadSelf will take care
of the defining. Assign resource after the UploadSelf call
to prevent NULL pointer exception.
Reviewed-by: Axel Davy <axel.davy@ens.fr>
Signed-off-by: Patrick Rudolph <siro@das-labor.org>
Axel Davy [Sat, 16 May 2015 16:40:02 +0000 (18:40 +0200)]
st/nine: Fix FillColor Flag check
IT is better check if the surface was created with RT flag,
instead of checking capability (llvmpipe was complaining)
Signed-off-by: Axel Davy <axel.davy@ens.fr>
Axel Davy [Thu, 7 May 2015 23:29:35 +0000 (01:29 +0200)]
st/nine: Fix StretchRect checks
Fixes Wine tests
Signed-off-by: Axel Davy <axel.davy@ens.fr>
Axel Davy [Fri, 8 May 2015 14:54:53 +0000 (16:54 +0200)]
st/nine: Implement EvictManagedResources
EvictManagedResources is used by apps to free
the gpu memory of MANAGED textures (which have
a cpu memory backing)
Signed-off-by: Axel Davy <axel.davy@ens.fr>
Axel Davy [Wed, 6 May 2015 22:03:21 +0000 (00:03 +0200)]
st/nine: Track managed textures
Signed-off-by: Axel Davy <axel.davy@ens.fr>
Axel Davy [Tue, 5 May 2015 18:40:12 +0000 (20:40 +0200)]
st/nine: Only update dirty rect for UpdateTexture
UpdateTexture is supposed to optimise by uploading only for the
dirty region of the source (d3d9 doc, wine tests).
This patch adds the behaviour for surfaces, but not entirely for
volumes.
Signed-off-by: Axel Davy <axel.davy@ens.fr>
Axel Davy [Sun, 3 May 2015 22:17:09 +0000 (00:17 +0200)]
st/nine: Textures start dirty
According to the spec all textures start
dirty.
Signed-off-by: Axel Davy <axel.davy@ens.fr>
Axel Davy [Sun, 3 May 2015 21:53:48 +0000 (23:53 +0200)]
st/nine: Track dirty region for SYSTEMMEM too
Dirty regions should be tracked for both MANAGED
and SYSTEMMEM.
Until now we didn't bother to track for SYSTEMMEM,
because we hadn't implemented using the dirty region
to avoid some copies
Signed-off-by: Axel Davy <axel.davy@ens.fr>
Axel Davy [Sun, 3 May 2015 21:46:14 +0000 (23:46 +0200)]
st/nine: Add missing BASETEX_REGISTER_UPDATE calls
If the texture is bound and dirty_mip is true,
BASETEX_REGISTER_UPDATE adds the texture to the list
of things to update before the next draw call.
Some calls to it were missing.
Signed-off-by: Axel Davy <axel.davy@ens.fr>
Axel Davy [Sun, 3 May 2015 21:28:25 +0000 (23:28 +0200)]
st/nine: SetAutoGenFilterType should regenerate the sublevels
It should regenerate the sublevels according to the spec
Signed-off-by: Axel Davy <axel.davy@ens.fr>
Axel Davy [Sun, 3 May 2015 20:57:13 +0000 (22:57 +0200)]
st/nine: Simplify NineVolume9_CopyVolume
We had only one usage for this function.
Signed-off-by: Axel Davy <axel.davy@ens.fr>
Axel Davy [Sun, 3 May 2015 20:40:38 +0000 (22:40 +0200)]
st/nine: Split NineSurface9_CopySurface
NineSurface9_CopySurface was supporting more cases than what
we needed, and doing checks that were innapropriate for
some NineSurface9_CopySurface use cases.
This patch splits it into two for the two use cases, and moves
the checks to the caller.
This patch also adds a few checks to NineDevice9_UpdateSurface
Signed-off-by: Axel Davy <axel.davy@ens.fr>
Axel Davy [Tue, 5 May 2015 21:51:24 +0000 (23:51 +0200)]
st/nine: Simplify Volume9 dirty region tracking
Similar to what was done for Surface9, track the dirty region
only in VolumeTexture9.
Signed-off-by: Axel Davy <axel.davy@ens.fr>
Marek Olšák [Sat, 8 Nov 2014 12:02:47 +0000 (13:02 +0100)]
util/u_blitter: implement alpha blending for pipe->blit
Christoph Bumiller [Sat, 31 May 2014 21:13:24 +0000 (23:13 +0200)]
gallium: Add blending to pipe blit
This type of blending is used for gallium nine software cursor
Signed-off-by: David Heidelberg <david@ixit.cz>
Axel Davy [Sat, 21 Mar 2015 21:28:59 +0000 (22:28 +0100)]
st/nine: Revert to sw cursor in case of failure to set hw cursor
Signed-off-by: Axel Davy <axel.davy@ens.fr>
Reviewed-by: David Heidelberg <david@ixit.cz>
Axel Davy [Sat, 21 Mar 2015 21:23:56 +0000 (22:23 +0100)]
st/nine: Do not call ID3DPresent_GetCursorPos for sw cursor
For sw cursor we do not tell wine the cursor position (the app
tells us directly). We shouldn't use ID3DPresent_GetCursorPos.
device->cursor.pos already contains the coordinates the app
gave us.
Signed-off-by: Axel Davy <axel.davy@ens.fr>
Reviewed-by: David Heidelberg <david@ixit.cz>
Axel Davy [Sun, 22 Mar 2015 17:48:07 +0000 (18:48 +0100)]
st/nine: Force hw cursor for Windowed mode
According to the spec, Windowed mode must
have hw cursor
Signed-off-by: Axel Davy <axel.davy@ens.fr>
Reviewed-by: David Heidelberg <david@ixit.cz>
Axel Davy [Sat, 21 Mar 2015 21:21:14 +0000 (22:21 +0100)]
st/nine: Hide hardware cursor when we don't use it
We have either hardware cursor or software cursor.
When we use software cursor, we should hide the hardware
cursor.
Signed-off-by: Axel Davy <axel.davy@ens.fr>
Reviewed-by: David Heidelberg <david@ixit.cz>
Axel Davy [Sun, 15 Feb 2015 20:30:44 +0000 (21:30 +0100)]
st/nine: fix D3DRS_DITHERENABLE wrong state group
D3DRS_DITHERENABLE was assigned to the rasterizer state
group, but it was used for the blend group.
Assign it to the blend group.
Signed-off-by: Axel Davy <axel.davy@ens.fr>
Patrick Rudolph [Sun, 19 Apr 2015 08:14:30 +0000 (10:14 +0200)]
st/nine: Account POINTSIZE_MIN and POINTSIZE_MAX for point size
When using D3DRS_POINTSIZE make sure the value is at least
D3DRS_POINTSIZE_MIN but not greater than D3DRS_POINTSIZE_MAX.
Fixes some Wine tests.
Reviewed-by: Axel Davy <axel.davy@ens.fr>
Signed-off-by: Patrick Rudolph <siro@das-labor.org>
Patrick Rudolph [Tue, 12 May 2015 05:27:37 +0000 (07:27 +0200)]
st/nine: Align texture memory
Align texture memory on 32 byte boundry to allow
SSE/AVX memcpy to work on locked rects.
This fixes some crashes with games using SSE.
Reviewed-by: David Heidelberg <david@ixit.cz>
Reviewed-by: Axel Davy <axel.davy@ens.fr>
Signed-off-by: Patrick Rudolph <siro@das-labor.org>
Axel Davy [Sat, 16 May 2015 16:41:51 +0000 (18:41 +0200)]
st/nine: Always set point_quad_rasterization to 1
Both Points and Point Sprites are rasterized like quads,
according to d3d9 doc and gallium rasterizer doc.
Signed-off-by: Axel Davy <axel.davy@ens.fr>
Axel Davy [Sat, 16 May 2015 20:41:26 +0000 (22:41 +0200)]
st/nine: Fix Swizzle for ATI2 format
We had red and green in the wrong channels
for the ATI2 format (RGTC2).
Found thanks to wine tests.
Signed-off-by: Axel Davy <axel.davy@ens.fr>
Reviewed-by: David Heidelberg <david@ixit.cz>
Patrick Rudolph [Mon, 25 May 2015 08:36:21 +0000 (10:36 +0200)]
target/d3dadapter9: Return Windows like card names
Add support for multiple cards and fill in Win
like card name, driver name and version info.
Use fallback for unknown vendors and unknown card names.
Reviewed-by: Axel Davy <axel.davy@ens.fr>
Signed-off-by: Patrick Rudolph <siro@das-labor.org>
David Heidelberg [Fri, 10 Apr 2015 22:13:53 +0000 (00:13 +0200)]
st/nine: Require gcc >= 4.6
Nine code uses some C11 features, and this
leads to compile error on gcc <= 4.5
Another way would have been to use the
-fms-extensions CFLAG
Signed-off-by: David Heidelberg <david@ixit.cz>
Cc: "10.4 10.5 10.6" <mesa-stable@lists.freedesktop.org>
Ilia Mirkin [Fri, 21 Aug 2015 19:08:15 +0000 (15:08 -0400)]
glsl: fix error message when validating tcs output decls
Signed-off-by: Ilia Mirkin <imirkin@alum.mit.edu>
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Rob Clark [Mon, 10 Aug 2015 21:27:19 +0000 (17:27 -0400)]
relnote updates
Signed-off-by: Rob Clark <robclark@freedesktop.org>
Ilia Mirkin [Fri, 21 Aug 2015 00:06:50 +0000 (20:06 -0400)]
st/mesa: pass through 4th opcode argument in bitmap/pixel visitors
Signed-off-by: Ilia Mirkin <imirkin@alum.mit.edu>
Cc: "10.6" <mesa-stable@lists.freedesktop.org>
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Ilia Mirkin [Thu, 20 Aug 2015 23:59:04 +0000 (19:59 -0400)]
st/mesa: fix assignments with 4-operand arguments (i.e. BFI)
Signed-off-by: Ilia Mirkin <imirkin@alum.mit.edu>
Cc: "10.6" <mesa-stable@lists.freedesktop.org>
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Martin Peres [Fri, 21 Aug 2015 13:25:14 +0000 (16:25 +0300)]
i965: allow image_size on float images
This got missed because the piglit test only tested int images to avoid a
combinatiorial explosion of format, targets, stages and sizes which
takes more than 5 minutes to test on nvidia's driver.
This patch also drops the IMAGE_FUNCTION_AVAIL_ATOMIC which is not applicable
to the image_size codepath but was not hurting in any way.
Signed-off-by: Martin Peres <martin.peres@linux.intel.com>
Reviewed-by: Francisco Jerez <currojerez@riseup.net>
Zoltan Gilian [Wed, 19 Aug 2015 09:56:08 +0000 (11:56 +0200)]
clover: fix llvm 3.5 build error
There is no MDOperand in llvm 3.5.
v2: Check if kernel metadata is present to avoid crash (EdB).
v3: Second attempt to avoid crash: switch off metadata query for llvm < 3.6.
Reviewed-by: Serge Martin (EdB) <edb+mesa@sigluy.net>
Reviewed-by: Francisco Jerez <currojerez@riseup.net>
Tapani Pälli [Thu, 20 Aug 2015 07:25:59 +0000 (10:25 +0300)]
mesa: update fbo state in glTexStorage
We have to re-validate FBOs rendering to the texture like is done
with TexImage and CopyTexImage.
Signed-off-by: Tapani Pälli <tapani.palli@intel.com>
Reviewed-by: Anuj Phogat <anuj.phogat@gmail.com>
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=91673
Cc: "10.6" <mesa-stable@lists.freedesktop.org>
Eric Anholt [Wed, 19 Aug 2015 05:19:12 +0000 (22:19 -0700)]
vc4: Add algebraic opt for rcp(1.0).
We're generating rcps as part of backend lowering of the packed coordinate
in the CS, and we don't want to lower them in NIR because of the extra
newton-raphson steps in the common case. However, GLB2.7 is moving a
vertex attribute with a 1.0 W component to the position, and that makes us
produce some silly RCPs.
total instructions in shared programs: 97590 -> 97580 (-0.01%)
instructions in affected programs: 74 -> 64 (-13.51%)
Eric Anholt [Wed, 19 Aug 2015 05:07:47 +0000 (22:07 -0700)]
vc4: Allow unpack_8[abcd]_f's src to stay in r4.
I had QPU emit code to do it, but forgot to flag the register class.
total instructions in shared programs: 97974 -> 97590 (-0.39%)
instructions in affected programs: 25291 -> 24907 (-1.52%)
Eric Anholt [Wed, 19 Aug 2015 04:26:05 +0000 (21:26 -0700)]
vc4: Pack the unorm-packing bits into a src MUL instruction when possible.
Now that we do non-SSA QIR instructions, we can take a NIR SSA src that's
only used by the unorm packing and just stuff the pack bits into it.
total instructions in shared programs: 98136 -> 97974 (-0.17%)
instructions in affected programs: 4149 -> 3987 (-3.90%)
Eric Anholt [Wed, 19 Aug 2015 04:43:42 +0000 (21:43 -0700)]
vc4: Add a QIR helper for whether the op is a MUL type.
Eric Anholt [Wed, 19 Aug 2015 03:18:51 +0000 (20:18 -0700)]
vc4: Drop an unused algebraic op.
NIR now handles this optimization for us.
Eric Anholt [Thu, 6 Aug 2015 03:54:02 +0000 (20:54 -0700)]
vc4: Switch QPU_PACK_SCALED to be two non-SSA instructions.
total instructions in shared programs: 98159 -> 98136 (-0.02%)
instructions in affected programs: 12279 -> 12256 (-0.19%)
Eric Anholt [Thu, 6 Aug 2015 03:31:21 +0000 (20:31 -0700)]
vc4: Make the pack-to-unorm instructions be non-SSA.
This helps ensure that the register allocator doesn't force the later pack
operations to insert extra MOVs.
total instructions in shared programs: 98170 -> 98159 (-0.01%)
instructions in affected programs: 2134 -> 2123 (-0.52%)
Eric Anholt [Tue, 4 Aug 2015 02:25:47 +0000 (19:25 -0700)]
vc4: Allow QIR registers to be non-SSA.
Now that we have NIR, most of the optimization we still need to do is
peepholes on instruction selection rather than general dataflow
operations. This means we want to be able to have QIR be a lot closer to
the actual QPU instructions, just with virtual registers. Allowing
multiple instructions writing the same register opens up a lot of
possibilities.
Eric Anholt [Thu, 6 Aug 2015 03:11:07 +0000 (20:11 -0700)]
vc4: We can now move TEX_RESULT accesses across other r4 ops.
No difference on shader-db.
Timothy Arceri [Wed, 27 May 2015 10:12:42 +0000 (20:12 +1000)]
glsl: fix binding validation for interface blocks
V2: rebase on SSBO changes
Reviewed-by: Ian Romanick <ian.d.romanick@intel.com>
Timothy Arceri [Sun, 16 Aug 2015 04:26:23 +0000 (14:26 +1000)]
glsl: interleave constant propagation and folding
The constant folding pass can take a long time to complete
so rather than running through the entire pass each time
a new constant is propagated (and vice versa) interleave them.
This change helps ES31-CTS.arrays_of_arrays.InteractionFunctionCalls1
go from around 2 min -> 23 sec.
Reviewed-by: Ian Romanick <ian.d.romanick@intel.com>
Ilia Mirkin [Fri, 21 Aug 2015 02:13:48 +0000 (22:13 -0400)]
nv50/ir: pre-compute BFE arg when both bits and offset are imm
Due to a quirk in how the nv50 opt passes run, the algebraic
optimization that looks for these BFE's happens before the constant
folding pass. Rearranging these passes isn't a great idea, but this is
easy enough to fix. Allows a following cvt to eliminate the bfe in
certain situations.
Signed-off-by: Ilia Mirkin <imirkin@alum.mit.edu>
Ilia Mirkin [Wed, 19 Aug 2015 22:43:47 +0000 (18:43 -0400)]
glsl: expose textureQueryLod in GLSL 4.00+ fragment shaders
See issue from the ARB_texture_query_lod spec for LOD vs Lod confusion:
(3) The core specification uses the "Lod" spelling, not "LOD". Should
this extension be modified to use "Lod"?
RESOLVED: The "Lod" spelling is the correct spelling for the core
specification and the preferred spelling for use. However, use of
"LOD" also exists, as the extension predated the core specification,
so this extension won't remove use of "LOD".
Signed-off-by: Ilia Mirkin <imirkin@alum.mit.edu>
Reviewed-by: Timothy Arceri <t_arceri@yahoo.com.au>
Nanley Chery [Fri, 21 Aug 2015 01:00:20 +0000 (18:00 -0700)]
Revert "mesa/formats: refactor by collapsing cases in switch statement by type"
This reverts commit
ffe6c6ad5f719dedd1b6b95e8590e3f20b23d340.
_mesa_format_num_components() does not include the padding bits in mesa formats
containing 'X' channels. This could cause mipmap generation for certain
uncompressed formats to underestimate the number of channels in the source
image by 1.
Signed-off-by: Nanley Chery <nanley.g.chery@intel.com>
Glenn Kennard [Thu, 13 Aug 2015 18:30:07 +0000 (20:30 +0200)]
r600g: Fix handling of TGSI_OPCODE_ARR with SB
FLT_TO_INT goes in the vector pipes on evergreen/NI,
not the trans unit as on earlier chips.
Signed-off-by: Glenn Kennard <glenn.kennard@gmail.com>
Reviewed-by: Dave Airlie <airlied@redhat.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>
Edward O'Callaghan [Wed, 19 Aug 2015 08:58:47 +0000 (18:58 +1000)]
r600: Turn 'r600_shader_key' struct into union
This struct was getting a bit crowded, following the lead of
radeonsi, mirror the idea of having sub-structures for each
shader type. Turning 'r600_shader_key' into an union saves
some trivial memory and CPU cycles for the shader keys.
[airlied: drop as_ls, and reorder so larger fields at start.]
Signed-off-by: Edward O'Callaghan <eocallaghan@alterapraxis.com>
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>
Edward O'Callaghan [Wed, 19 Aug 2015 08:58:46 +0000 (18:58 +1000)]
r600: Rewrite r600_shader_selector_key() to use a switch stmt
Signed-off-by: Edward O'Callaghan <eocallaghan@alterapraxis.com>
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>
Jason Ekstrand [Mon, 3 Aug 2015 15:17:42 +0000 (08:17 -0700)]
i965: Use NIR by default for vertex shaders
Shader-db results for vec4 on i965:
total instructions in shared programs: 1499894 -> 1502261 (0.16%)
instructions in affected programs: 1414224 -> 1416591 (0.17%)
helped: 2434
HURT: 10543
GAINED: 1
LOST: 0
Shader-db results for vec4 on g4x:
total instructions in shared programs: 1437411 -> 1439779 (0.16%)
instructions in affected programs: 1362402 -> 1364770 (0.17%)
helped: 2434
HURT: 10544
GAINED: 0
LOST: 0
Shader-db results for vec4 on Iron Lake:
total instructions in shared programs: 1437214 -> 1439593 (0.17%)
instructions in affected programs: 1362205 -> 1364584 (0.17%)
helped: 2433
HURT: 10544
GAINED: 1
LOST: 0
Shader-db results for vec4 on Sandy Bridge:
total instructions in shared programs: 2022092 -> 1941570 (-3.98%)
instructions in affected programs: 1886838 -> 1806316 (-4.27%)
helped: 7510
HURT: 10737
GAINED: 0
LOST: 0
Shader-db results for vec4 on Ivy Bridge:
total instructions in shared programs: 1853749 -> 1804960 (-2.63%)
instructions in affected programs: 1686736 -> 1637947 (-2.89%)
helped: 6735
HURT: 11101
GAINED: 0
LOST: 0
Shader-db results for vec4 on Haswell:
total instructions in shared programs: 1853749 -> 1804960 (-2.63%)
instructions in affected programs: 1686736 -> 1637947 (-2.89%)
helped: 6735
HURT: 11101
GAINED: 0
LOST: 0
Signed-off-by: Jason Ekstrand <jason.ekstrand@intel.com>
Acked-by: Kenneth Graunke <kenneth@whitecape.org>
Acked-by: Matt Turner <mattst88@gmail.com>
Kai Wasserbäch [Fri, 14 Aug 2015 12:49:43 +0000 (14:49 +0200)]
glsl: check if return_deref in lower_subroutine_visitor::visit_leave isn't NULL
Fixes a crash in Piglit's
spec@arb_shader_subroutine@linker@no-mutual-recursion.vert for me.
Signed-off-by: Kai Wasserbäch <kai@dev.carbon-project.org>
Reviewed-by: Dave Airlie <airlied@redhat.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>
Tobias Klausmann [Sun, 11 Jan 2015 21:40:22 +0000 (22:40 +0100)]
nv50/ir: Handle OP_CVT when folding constant expressions
[imirkin: handle more type combinations, use macro]
Signed-off-by: Ilia Mirkin <imirkin@alum.mit.edu>
Ilia Mirkin [Wed, 19 Aug 2015 03:16:32 +0000 (23:16 -0400)]
nvc0/ir: undo more shifts still by allowing a pre-SHL to occur
This happens with unpackSnorm lowering. There's yet another
bitfield-extract behind it, but there's too much variation to be worth
cutting through.
Signed-off-by: Ilia Mirkin <imirkin@alum.mit.edu>
Ilia Mirkin [Wed, 19 Aug 2015 02:53:11 +0000 (22:53 -0400)]
nvc0/ir: don't require AND when the high byte is being addressed
unpackUnorm* lowering doesn't AND the high byte/word as it's
unnecessary. Detect that situation as well.
Signed-off-by: Ilia Mirkin <imirkin@alum.mit.edu>
Ilia Mirkin [Wed, 19 Aug 2015 01:09:12 +0000 (21:09 -0400)]
nvc0/ir: detect i2f/i2i which operate on specific bytes/words
Some Unigine shaders have been observed to unpack bytes out of 32-bit
integers and convert them to floats. I2F/I2I can handle this sort of
thing directly. Detect the handleable situations.
This misses 16-bit word capabilities in nv50, but I haven't seen shaders
that would actually make use of that.
Signed-off-by: Ilia Mirkin <imirkin@alum.mit.edu>
Ilia Mirkin [Wed, 19 Aug 2015 01:07:33 +0000 (21:07 -0400)]
nvc0/ir: detect AND/SHR pairs and convert into EXTBF
Some shaders appear to extract bits using shift/and combos. Detect
(some) of those and convert to EXTBF instead.
Signed-off-by: Ilia Mirkin <imirkin@alum.mit.edu>
Chih-Wei Huang [Fri, 19 Jun 2015 18:00:15 +0000 (02:00 +0800)]
nv50/ir: support different unordered_set implementations
If build with C++11 standard, use std::unordered_set.
Otherwise if build on old Android version with stlport,
use std::tr1::unordered_set with a wrapper class.
Otherwise use std::tr1::unordered_set.
Signed-off-by: Chih-Wei Huang <cwhuang@linux.org.tw>
Reviewed-by: Ilia Mirkin <imirkin@alum.mit.edu>
Martin Peres [Thu, 20 Aug 2015 12:15:56 +0000 (15:15 +0300)]
i965: Fix "handle nir_intrinsic_image_size"
I pushed a half-baked version of "i965: handle nir_intrinsic_image_size" by
accident. Not having the Reviewed-by: tags on the last two commits should
have been a red flag but I somehow missed it after the QA check.
This patch should fix image-size for non-int images. I will add support to
the piglit test for all the other image types.
Sorry for the noise.
Signed-off-by: Martin Peres <martin.peres@linux.intel.com>
Reviewed-by: Francisco Jerez <currojerez@riseup.net>
Martin Peres [Wed, 29 Apr 2015 09:42:16 +0000 (12:42 +0300)]
i965: enable GL_ARB_shader_image_size
Signed-off-by: Martin Peres <martin.peres@linux.intel.com>
Martin Peres [Wed, 29 Apr 2015 09:39:16 +0000 (12:39 +0300)]
i965: handle nir_intrinsic_image_size
v2, Review from Francisco Jerez:
- avoid the camelCase for the booleans
- init the booleans using the sampler type
- force the initialization of all the components of the output register
v3:
- Rename a variable from CubeMapArray to CubeArray to re-use GLSL's name (Ilia)
- Fix some indentation and drop parenthesis (Topi)
- Fix a signed/unsigned comparaison warning
Signed-off-by: Martin Peres <martin.peres@linux.intel.com>
Martin Peres [Tue, 11 Aug 2015 14:42:12 +0000 (17:42 +0300)]
nir: convert the glsl intrinsic image_size to nir_intrinsic_image_size
v2, review from Francisco Jerez:
- make the destination variable as large as what the nir instrinsic
defines (4) instead of the size of the return variable of glsl. This
is still safe for the already existing code because all the intrinsics
affected returned the same amount of components as expected by glsl IR.
In the case of image_size, it is not possible to do so because the
returned number of component depends on the image type and this case
is not well handled by nir.
v3:
- Style fix
Signed-off-by: Martin Peres <martin.peres@linux.intel.com>
Reviewed-by: Francisco Jerez <currojerez@riseup.net>
Martin Peres [Mon, 27 Apr 2015 16:25:34 +0000 (19:25 +0300)]
glsl: add support for the imageSize builtin
The code is heavily inspired from Francisco Jerez's code supporting the
image_load_store extension.
Backends willing to support this builtin should handle
__intrinsic_image_size.
v2: Based on the review of Ilia Mirkin
- Enable the extension for GLES 3.1
- Fix indentation
- Fix the return type (float to int, number of components for CubeImages)
- Add a warning related to GLES 3.1
v3: Based on the review of Francisco Jerez
- Refactor the code to share both add_image_function and _image with the other
image-related functions
v4: Based on Topi Pohjolainen's comments
- Do not add parenthesis for the return value
v5: based on Francisco Jerez's comments:
- Fix a few indent issues
- Reduce the size of a condition by testing the dimension and array properties
instead of enumerating all the formats.
Signed-off-by: Martin Peres <martin.peres@linux.intel.com>
Reviewed-by: Francisco Jerez <currojerez@riseup.net>
Martin Peres [Mon, 27 Apr 2015 17:05:14 +0000 (20:05 +0300)]
main: add extension GL_ARB_shader_image_size
Reviewed-by: Edward O'Callaghan <eocallaghan@alterapraxis.com>
Reviewed-by: Francisco Jerez <currojerez@riseup.net>
Signed-off-by: Martin Peres <martin.peres@linux.intel.com>
Francisco Jerez [Thu, 20 Aug 2015 10:46:53 +0000 (13:46 +0300)]
docs: Mark GLES 3.1 image load/store as done on i965.
Francisco Jerez [Wed, 19 Aug 2015 11:42:50 +0000 (14:42 +0300)]
mesa: Add ES31 API tag for the extension table.
I'll mark the OES_shader_image_atomic extension entry with this tag to
make sure that we don't expose it on earlier GLES API versions
accidentally, because according to the extension:
"OpenGL ES 3.1 and GLSL ES 3.10 are required."
Reviewed-by: Tapani Pälli <tapani.palli@intel.com>
Francisco Jerez [Sun, 16 Aug 2015 22:47:50 +0000 (01:47 +0300)]
glsl: Parse the allowed image format qualifiers in GLSL ES 3.1.
This includes the minimum required desktop/ES GLSL version in the
format qualifier table in anticipation of new GLSL versions extending
the set of supported image formats. According to section 4.4.7 of the
GLSL ES 3.1 spec:
"The format layout qualifier identifiers for image variable
declarations are:
[...]
rgba32f
rgba16f
r32f
rgba8
rgba8_snorm
[...]
rgba32i
rgba16i
rgba8i
r32i
[...]
rgba32ui
rgba16ui
rgba8ui
r32ui"
Reviewed-by: Tapani Pälli <tapani.palli@intel.com>
Francisco Jerez [Mon, 17 Aug 2015 16:12:00 +0000 (19:12 +0300)]
glsl: Recognise image memory qualifiers in GLSL ES 3.1.
Reviewed-by: Tapani Pälli <tapani.palli@intel.com>
Francisco Jerez [Mon, 17 Aug 2015 14:42:30 +0000 (17:42 +0300)]
glsl: Define image-related built-in constants required by GLSL ES 3.1.
Reviewed-by: Tapani Pälli <tapani.palli@intel.com>
Francisco Jerez [Sun, 16 Aug 2015 22:39:38 +0000 (01:39 +0300)]
glsl: Remove duplicate definition of gl_MaxTess*ImageUniforms built-in constants.
These seem to have been re-added at some point during the
ARB_tessellation_shader implementation work. AFAICT the second
(correct) definition of each constant would have had no effect because
the symbols were already defined.
Reviewed-by: Tapani Pälli <tapani.palli@intel.com>
Francisco Jerez [Sun, 16 Aug 2015 22:38:00 +0000 (01:38 +0300)]
glsl: Accept atomic_uint type in GLSL ES 3.1.
Reviewed-by: Tapani Pälli <tapani.palli@intel.com>
Francisco Jerez [Sun, 16 Aug 2015 22:37:12 +0000 (01:37 +0300)]
glsl: Accept supported image types in GLSL ES 3.1.
These are a subset of the image types supported by desktop GL,
excluding 1D, 1D array, rectangle, buffer, cube array, 2D MS and 2D
MS array texture targets.
Reviewed-by: Tapani Pälli <tapani.palli@intel.com>
Francisco Jerez [Sun, 16 Aug 2015 22:34:41 +0000 (01:34 +0300)]
glsl: Expose image load and store built-ins in GLSL ES 3.1.
Reviewed-by: Tapani Pälli <tapani.palli@intel.com>
Francisco Jerez [Sun, 16 Aug 2015 22:34:13 +0000 (01:34 +0300)]
glsl: Use a separate availability class for image atomic built-ins.
These are not part of unextended GLSL ES 3.1.
Reviewed-by: Timothy Arceri <t_arceri@yahoo.com.au>
Reviewed-by: Tapani Pälli <tapani.palli@intel.com>
Francisco Jerez [Sun, 16 Aug 2015 22:28:57 +0000 (01:28 +0300)]
glsl: Allow precision qualifiers on general opaque types.
From the GLSL ES 3.1 spec, section 4.7.3:
"Any floating point, integer, opaque type declaration can have the
type preceded by one of these precision qualifiers: [...] highp
[...], mediump [...], lowp [...]."
Reviewed-by: Timothy Arceri <t_arceri@yahoo.com.au>
Reviewed-by: Tapani Pälli <tapani.palli@intel.com>
Francisco Jerez [Sun, 16 Aug 2015 22:27:43 +0000 (01:27 +0300)]
glsl: Implement GLSL ES restriction on images being either readonly or writeonly.
Reviewed-by: Tapani Pälli <tapani.palli@intel.com>
Francisco Jerez [Sun, 16 Aug 2015 22:26:40 +0000 (01:26 +0300)]
glsl: Require that all image uniforms have a format qualifier in GLSL ES.
Note that this is slightly more permissive than the spec language
requires: "Any image variable must specify a format layout qualifier."
The GLSL ES spec seems really sketchy regarding format layout
qualifiers on function formal parameters -- On the one hand they are
required, but on the other hand it doesn't provide any syntax to
specify them (see section 6.1.1), they don't participate in parameter
type matching for overload resolution, and are in fact explictly
forbidden ("Layout qualifiers cannot be used on formal function
parameters"). Of course none of the image built-in functions defined
by the spec specify format layout qualifiers (and they probably
couldn't sensibly), to contradict its own requirement.
This probably qualifies for a spec bug, but in the meantime do the
sensible thing and require layout qualifiers on uniforms *only*.
Reviewed-by: Timothy Arceri <t_arceri@yahoo.com.au>
Francisco Jerez [Sun, 16 Aug 2015 22:25:11 +0000 (01:25 +0300)]
glsl: Add support for image binding qualifiers.
Support for binding an image to an image unit explicitly in the shader
source is required by both GLSL 4.2 and GLSL ES 3.1, but not by the
original ARB_shader_image_load_store extension.
Reviewed-by: Timothy Arceri <t_arceri@yahoo.com.au>
Francisco Jerez [Sun, 16 Aug 2015 22:21:01 +0000 (01:21 +0300)]
glsl: Forbid non-constant image array indexing in GLSL ES 3.1.
Reviewed-by: Timothy Arceri <t_arceri@yahoo.com.au>
Francisco Jerez [Sun, 16 Aug 2015 23:05:43 +0000 (02:05 +0300)]
mesa: Refuse to bind image uniforms using glUniform in GLES.
The GLES 3.1 spec removed support for updating the image unit bound to
an image uniform using glUniform1i() calls.
Reviewed-by: Tapani Pälli <tapani.palli@intel.com>
Reviewed-by: Ian Romanick <ian.d.romanick@intel.com>
Francisco Jerez [Sun, 16 Aug 2015 23:02:17 +0000 (02:02 +0300)]
mesa: Refuse to bind a mutable texture object to an image unit in GLES.
Reviewed-by: Ian Romanick <ian.d.romanick@intel.com>
Reviewed-by: Timothy Arceri <t_arceri@yahoo.com.au>
Francisco Jerez [Sun, 16 Aug 2015 23:01:40 +0000 (02:01 +0300)]
mesa: Initialize image unit state to different defaults in GLES.
There is no GL_R8 image format in GLES, according to the state table
20.32 of the GLES 3.1 spec the default value should be GL_R32UI. The
ES31-CTS.shader_image_load_store.basic-api-bind Khronos conformance
test checks that this is the case.
Reviewed-by: Ian Romanick <ian.d.romanick@intel.com>
Francisco Jerez [Sun, 16 Aug 2015 23:00:48 +0000 (02:00 +0300)]
mesa: Reset image unit state to the default values when a bound image is deleted.
The ES31-CTS.shader_image_load_store.basic-api-bind conformance test
expects the whole image unit state to be reset when the bound texture
object is deleted. The ARB_shader_image_load_store extension is
rather vague regarding what should happen with image unit state other
than the texture object in that case, but the GL 4.2 and GLES 3.1
specifications (section "Automatic Unbinding of Deleted Objects")
explicitly require it to be reset to the default values.
Reviewed-by: Tapani Pälli <tapani.palli@intel.com>
Reviewed-by: Ian Romanick <ian.d.romanick@intel.com>
Francisco Jerez [Sun, 16 Aug 2015 22:58:53 +0000 (01:58 +0300)]
mesa: Reject image formats not supported by GLES.
Reviewed-by: Timothy Arceri <t_arceri@yahoo.com.au>
Reviewed-by: Tapani Pälli <tapani.palli@intel.com>
Reviewed-by: Ian Romanick <ian.d.romanick@intel.com>