mesa.git
9 years agonir/build: add nir_vec() helper
Rob Clark [Fri, 18 Sep 2015 17:23:36 +0000 (13:23 -0400)]
nir/build: add nir_vec() helper

Signed-off-by: Rob Clark <robclark@freedesktop.org>
Reviewed-by: Connor Abbott <cwabbott0@gmail.com>
9 years agofreedreno/ir3: lower txp/clamp in NIR
Rob Clark [Wed, 16 Sep 2015 17:42:21 +0000 (13:42 -0400)]
freedreno/ir3: lower txp/clamp in NIR

Signed-off-by: Rob Clark <robclark@freedesktop.org>
9 years agonir/lower_tex: add support to clamp texture coords
Rob Clark [Fri, 18 Sep 2015 14:44:27 +0000 (10:44 -0400)]
nir/lower_tex: add support to clamp texture coords

Some hardware needs to clamp texture coordinates to [0.0, 1.0] in the
shader to emulate GL_CLAMP.  This is added to lower_tex_proj since, in
the case of projected coords, the clamping needs to happen *after*
projection.

v2: comments/suggestions from Ilia and Eric, use txs to get texture size
and clamp RECT textures to their dimensions rather than [0.0, 1.0] to
avoid having to lower RECT textures to 2D.

Signed-off-by: Rob Clark <robclark@freedesktop.org>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
9 years agonir/lower_tex: support for lowering RECT textures
Rob Clark [Wed, 16 Sep 2015 20:49:14 +0000 (16:49 -0400)]
nir/lower_tex: support for lowering RECT textures

v2: comments/suggestions from Ilia and Eric, split out get_texture_size()
helper so we can use it in the next commit for clamping RECT textures.

Signed-off-by: Rob Clark <robclark@freedesktop.org>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
9 years agonir/lower_tex: support projector lowering per sampler type
Rob Clark [Wed, 16 Sep 2015 16:56:58 +0000 (12:56 -0400)]
nir/lower_tex: support projector lowering per sampler type

Some hardware, such as adreno a3xx, supports txp on some but not all
sampler types.  In this case we want more fine grained control over
which texture projectors get lowered.

v2: split out nir_lower_tex_options struct to make it easier to
add the additional parameters coming in the following patches

Signed-off-by: Rob Clark <robclark@freedesktop.org>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
9 years agonir/lower_tex: split out project_src() helper
Rob Clark [Wed, 16 Sep 2015 16:53:12 +0000 (12:53 -0400)]
nir/lower_tex: split out project_src() helper

Split this out to reduce noise in later patches.

Signed-off-by: Rob Clark <robclark@freedesktop.org>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
9 years agonir: rename nir_lower_tex_projector
Rob Clark [Thu, 17 Sep 2015 11:54:35 +0000 (07:54 -0400)]
nir: rename nir_lower_tex_projector

Since the following patches will add additional tex-lowering related
functionality, which doesn't make sense to split out into a separate
pass (as they would require duplication of the projector lowering
logic), let's give this pass a more generic name.

Signed-off-by: Rob Clark <robclark@freedesktop.org>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
9 years agoi965/vec4: Change types as needed to propagate source modifiers using current instruction
Alejandro Piñeiro [Wed, 16 Sep 2015 08:26:55 +0000 (10:26 +0200)]
i965/vec4: Change types as needed to propagate source modifiers using current instruction

SEL and MOV instructions, as long as they don't have source modifiers, are
just copying bits around.  So those kind of instruction could be propagated
even if there are type mismatches. This is needed because NIR generates
integer SEL and MOV instructions whenever it doesn't know what else to
generate.

This commit adds support for copy propagation using current instruction
as reference.

Equivalent to commit 472ef9 but for vec4.

v2: include check for saturate, as Jason Ekstrand suggested
v3: check that the dst.type and the src type are the same, in order to
    solve (among others) the following deqp regression with v2:
    dEQP-GLES3.functional.shaders.operator.unary_operator.minus.lowp_uint_vertex

Reviewed-by: Jason Ekstrand <jason.ekstrand@intel.com>
9 years agoi965/fs: Fix comparison between signed and unsigned integer expressions
Iago Toral Quiroga [Fri, 18 Sep 2015 09:02:34 +0000 (11:02 +0200)]
i965/fs: Fix comparison between signed and unsigned integer expressions

brw_fs_visitor.cpp: In member function 'void fs_visitor::emit_urb_writes()':
brw_fs_visitor.cpp:977:58: warning: comparison between signed and unsigned integer expressions [-Wsign-compare]

Reviewed-by: Tapani Pälli <tapani.palli@intel.com>
9 years agomesa: fix errors when reading depth with glReadPixels
Tapani Pälli [Tue, 15 Sep 2015 06:17:20 +0000 (09:17 +0300)]
mesa: fix errors when reading depth with glReadPixels

OpenGL ES 3.0 spec 3.7.2 "Transfer of Pixel Rectangles" specifies
DEPTH_COMPONENT, UNSIGNED_INT as a valid couple, validation for
internal format is checked by is_float_depth().

Fix regression caused by 81d2fd91a90e5b2fd9fd74792a7a7c329f0e4d29 in:
   ES3-CTS.gtf.GL3Tests.packed_pixels.packed_pixels

Test uses GL_DEPTH_COMPONENT, UNSIGNED_INT only when GL_NV_read_depth
extension is present.

v2: change check in _mesa_error_check_format_and_type to be explicit
    for ES 2.0+, desktop OpenGL does not allow this behaviour + uses
    this function for both glReadPixels and glDrawPixels validation.
    (No Piglit regressions seen with v2.)

Signed-off-by: Tapani Pälli <tapani.palli@intel.com>
Reviewed-by: Anuj Phogat <anuj.phogat@gmail.com>
Reviewed-by: Samuel Iglesias Gonsálvez <siglesias@igalia.com> [v1]
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=92009
Cc: "10.6 11.0" <mesa-stable@lists.freedesktop.org>
9 years agonir/builder: fix c++11 compiler warning
Rob Clark [Fri, 18 Sep 2015 01:07:41 +0000 (21:07 -0400)]
nir/builder: fix c++11 compiler warning

Fixes:

   In file included from nir/nir_lower_samplers.cpp:27:0:
   nir/nir_builder.h: In function 'nir_ssa_def* nir_channel(nir_builder*, nir_ssa_def*, int)':
   nir/nir_builder.h:222:37: warning: narrowing conversion of 'c' from 'int' to 'unsigned int' inside { } is ill-formed in C++11 [-Wnarrowing]
       unsigned swizzle[4] = {c, c, c, c};

Signed-off-by: Rob Clark <robclark@freedesktop.org>
9 years agonir: really actually fix comment this time
Rob Clark [Fri, 18 Sep 2015 01:06:11 +0000 (21:06 -0400)]
nir: really actually fix comment this time

Signed-off-by: Rob Clark <robclark@freedesktop.org>
9 years agonir/print: print variable names
Rob Clark [Thu, 17 Sep 2015 22:18:45 +0000 (18:18 -0400)]
nir/print: print variable names

Signed-off-by: Rob Clark <robclark@freedesktop.org>
Reviewed-by: Timothy Arceri <t_arceri@yahoo.com.au>
9 years agonir: some comment fixups
Rob Clark [Thu, 17 Sep 2015 22:18:19 +0000 (18:18 -0400)]
nir: some comment fixups

Signed-off-by: Rob Clark <robclark@freedesktop.org>
Reviewed-by: Timothy Arceri <t_arceri@yahoo.com.au>
9 years agofreedreno/ir3: add --gpu arg to cmdline compiler
Rob Clark [Wed, 16 Sep 2015 17:57:26 +0000 (13:57 -0400)]
freedreno/ir3: add --gpu arg to cmdline compiler

Signed-off-by: Rob Clark <robclark@freedesktop.org>
9 years agofreedreno/a4xx: wire up ucp support
Rob Clark [Sat, 12 Sep 2015 15:15:32 +0000 (11:15 -0400)]
freedreno/a4xx: wire up ucp support

Signed-off-by: Rob Clark <robclark@freedesktop.org>
9 years agofreedreno/ir3: add support for ucp
Rob Clark [Thu, 10 Sep 2015 20:09:13 +0000 (16:09 -0400)]
freedreno/ir3: add support for ucp

Use nir_lower_clip pass for adding the VS/FS instructions to handle
user-clip-planes and CLIPDIST.  Wire up support for load_user_clip_plane
intrinsic to fetch ucp[plane] values as driver-params (passed as const's
to the shader).

Signed-off-by: Rob Clark <robclark@freedesktop.org>
9 years agonir: add lowering stage for user-clip-planes / clipdist
Rob Clark [Wed, 9 Sep 2015 18:57:15 +0000 (14:57 -0400)]
nir: add lowering stage for user-clip-planes / clipdist

The vertex shader lowering adds calculation for CLIPDIST, if needed
(ie. user-clip-planes), and the frag shader lowering adds conditional
kills based on CLIPDIST value (which should be treated as a normal
interpolated varying by the driver).

Note that this won't quite do the right thing in the face of MSAA plus
user-clip-planes, since all the samples would be killed or not (rather
than potentially only a portion of them).  But it's better than no UCP
support at all for drivers that don't have this in hw.

Signed-off-by: Rob Clark <robclark@freedesktop.org>
Reviewed-by: Eric Anholt <eric@anholt.net>
9 years agonir: add sysval for user-clip-planes
Rob Clark [Thu, 27 Aug 2015 21:42:40 +0000 (17:42 -0400)]
nir: add sysval for user-clip-planes

For lowering user-clip-planes, we need a way to pass the enabled/used
user-clip-planes in to shader.

Signed-off-by: Rob Clark <robclark@freedesktop.org>
Reviewed-by: Jason Ekstrand <jason@jlekstrand.net>
9 years agofreedreno/ir3: convert from tgsi semantic/index to varying-slot
Rob Clark [Fri, 11 Sep 2015 21:20:48 +0000 (17:20 -0400)]
freedreno/ir3: convert from tgsi semantic/index to varying-slot

Signed-off-by: Rob Clark <robclark@freedesktop.org>
9 years agoglsl: add SYSTEM_VALUE_VERTEX_CNT
Rob Clark [Fri, 11 Sep 2015 21:01:23 +0000 (17:01 -0400)]
glsl: add SYSTEM_VALUE_VERTEX_CNT

Used internally in freedreno/ir3 to calc stream-out position.  Seems
like a generic enough way to implement stream-out (using str instrs),
plus it avoids compiler warnings by sneaking in a non-enum value in
switch statements.

Signed-off-by: Rob Clark <robclark@freedesktop.org>
9 years agofreedreno/ir3: switch to shader_enums.h interp constants
Rob Clark [Thu, 10 Sep 2015 21:25:18 +0000 (17:25 -0400)]
freedreno/ir3: switch to shader_enums.h interp constants

A small step towards un-TGSI'ifying ir3.

Signed-off-by: Rob Clark <robclark@freedesktop.org>
9 years agonv50,nvc0: flush texture cache in presence of coherent bufs
Ilia Mirkin [Thu, 17 Sep 2015 02:17:18 +0000 (22:17 -0400)]
nv50,nvc0: flush texture cache in presence of coherent bufs

This fixes the newly-added arb_texture_buffer_object-bufferstorage
piglit test.

Signed-off-by: Ilia Mirkin <imirkin@alum.mit.edu>
Cc: "11.0" <mesa-stable@lists.freedesktop.org>
9 years agonv50,nvc0: detect underlying resource changes and update tic
Ilia Mirkin [Tue, 15 Sep 2015 05:32:40 +0000 (01:32 -0400)]
nv50,nvc0: detect underlying resource changes and update tic

When updating texture buffers, we might end up replacing the whole
buffer. Check that the tic address matches the resource address, and if
not, update the tic and reupload it.

This fixes:
  arb_direct_state_access-texture-buffer
  arb_texture_buffer_object-data-sync

Signed-off-by: Ilia Mirkin <imirkin@alum.mit.edu>
Cc: "11.0" <mesa-stable@lists.freedesktop.org>
9 years agovc4: Try to pair up instructions when only one of them has PM bit
Boyan Ding [Sun, 30 Aug 2015 07:07:33 +0000 (15:07 +0800)]
vc4: Try to pair up instructions when only one of them has PM bit

Instructions with difference in PM field can actually be paired up if
the one without PM doesn't do packing/unpacking and non-NOP
packing/unpacking operations from PM instruction aren't added to the
other without PM.

total instructions in shared programs: 48209 -> 47460 (-1.55%)
instructions in affected programs:     11688 -> 10939 (-6.41%)

Signed-off-by: Boyan Ding <boyan.j.ding@gmail.com>
Reviewed-by: Eric Anholt <eric@anholt.net>
9 years agoi965/vec4: Use nir_move_vec_src_uses_to_dest
Jason Ekstrand [Tue, 8 Sep 2015 23:45:57 +0000 (16:45 -0700)]
i965/vec4: Use nir_move_vec_src_uses_to_dest

The idea here is not that it gives register coalescing a little bit of a
helping hand.  It doesn't actually fix the coalescing problems, but it
seems to help a good bit.

Shader-db results for vec4 programs on Haswell:

   total instructions in shared programs: 1746280 -> 1683959 (-3.57%)
   instructions in affected programs:     1259166 -> 1196845 (-4.95%)
   helped:                                11363
   HURT:                                  148

v2 (Jason Ekstrand):
 - Run nir_move_vec_src_uses_to_dest after going out of SSA
 - New shader-db numbers

Reviewed-by: Eduardo Lima Mitev <elima@igalia.com>
9 years agonir: Add a pass to rewrite uses of vecN sources to the vecN destination
Jason Ekstrand [Tue, 8 Sep 2015 22:18:01 +0000 (15:18 -0700)]
nir: Add a pass to rewrite uses of vecN sources to the vecN destination

v2 (Jason Ekstrand):
 - Handle non-SSA sources and destinations

Reviewed-by: Eduardo Lima Mitev <elima@igalia.com>
9 years agonir: Add comments to nir_index_instrs and nir_index_ssa_defs
Jason Ekstrand [Mon, 14 Sep 2015 19:25:28 +0000 (12:25 -0700)]
nir: Add comments to nir_index_instrs and nir_index_ssa_defs

The provided indices have the very nice property that if A dominates B then
A->index <= B->index.  We should document that somewhere.

Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
9 years agonir: Add a generic instruction index
Jason Ekstrand [Tue, 8 Sep 2015 23:43:51 +0000 (16:43 -0700)]
nir: Add a generic instruction index

Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
9 years agomesa: Fix texture compression on big-endian systems
Ulrich Weigand [Tue, 15 Sep 2015 13:23:26 +0000 (15:23 +0200)]
mesa: Fix texture compression on big-endian systems

Various pieces of code to create compressed textures will first
generate an uncompressed RGBA texture into a temporary buffer,
and then read from that buffer while creating the final compressed
texture in the requested format.

The code reading from the temporary buffer assumes the buffer is
formatted as an array of bytes in RGBA order.  However, the buffer
is filled using a _mesa_texstore call with MESA_FORMAT_R8G8B8A8_UNORM
format -- this is defined as an array of *integers* holding the
RGBA values in packed format (least-significant to most-significant).
This means incorrect bytes are accessed on big-endian systems.

This patch fixes this by using the MESA_FORMAT_A8B8G8R8_UNORM format
instead on big-endian systems when filling the buffer.  This fixes
about 100 piglit test case failures on s390x for me.

Signed-off-by: Ulrich Weigand <ulrich.weigand@de.ibm.com>
Tested-by: Oded Gabbay <oded.gabbay@gmail.com>
Cc: "10.6" "11.0" <mesa-stable@lists.freedesktop.org>
Signed-off-by: Dave Airlie <airlied@gmail.com>
9 years agost/xa: Use PIPE_FORMAT_R8_UNORM when available
Thomas Hellstrom [Wed, 16 Sep 2015 12:53:13 +0000 (05:53 -0700)]
st/xa: Use PIPE_FORMAT_R8_UNORM when available

XA has been using L8_UNORM for a8 and yuv component surfaces.
This commit instead makes XA prefer R8_UNORM since it's assumed to have a
higher availability.

Also neither of these formats are suitable as destination formats using
destination alpha blending, so reject those operations.

Signed-off-by: Thomas Hellstrom <thellstrom@vmware.com>
Reviewed-by: Brian Paul <brianp@vmware.com>
9 years agomesa: return initial value for VALIDATE_STATUS if pipe not bound
Tapani Pälli [Tue, 1 Sep 2015 10:53:44 +0000 (13:53 +0300)]
mesa: return initial value for VALIDATE_STATUS if pipe not bound

From OpenGL 4.5 Core spec (7.13):

    "If pipeline is a name that has been generated (without subsequent
    deletion) by GenProgramPipelines, but refers to a program pipeline
    object that has not been previously bound, the GL first creates a
    new state vector in the same manner as when BindProgramPipeline
    creates a new program pipeline object."

I interpret this as "If GetProgramPipelineiv gets called without a
bound (but valid) pipeline object, the state should reflect initial
state of a new pipeline object." This is also expected behaviour by
ES31-CTS.sepshaderobjs.PipelineApi conformance test.

Signed-off-by: Tapani Pälli <tapani.palli@intel.com>
Reviewed-by: Marta Lofstedt <marta.lofstedt@intel.com>
9 years agomesa: return initial value for PROGRAM_SEPARABLE when not linked
Tapani Pälli [Tue, 1 Sep 2015 10:53:43 +0000 (13:53 +0300)]
mesa: return initial value for PROGRAM_SEPARABLE when not linked

From OpenGL ES 3.1 spec (7.12):

    "Most properties set within program objects are specified not to
    take effect until the next call to LinkProgram or ProgramBinary.
    Some properties further require a successful call to either of
    these commands before taking effect. GetProgramiv returns the
    properties currently in effect for program, which may differ from
    the properties set within program since the most recent call to
    LinkProgram or ProgramBinary, which have not yet taken effect. If
    there has been no such call putting changes to pname into effect,
    initial values are returned."

Signed-off-by: Tapani Pälli <tapani.palli@intel.com>
Reviewed-by: Marta Lofstedt <marta.lofstedt@intel.com>
9 years agomesa: enable query of PROGRAM_PIPELINE_BINDING for ES 3.1
Tapani Pälli [Mon, 14 Sep 2015 07:46:01 +0000 (10:46 +0300)]
mesa: enable query of PROGRAM_PIPELINE_BINDING for ES 3.1

Specified in OpenGL ES 3.1 spec, Table 23.32: Program Object State.

Signed-off-by: Tapani Pälli <tapani.palli@intel.com>
Reviewed-by: Marta Lofstedt <marta.lofstedt@intel.com>
9 years agonir: support indirect indexing samplers in struct arrays
Timothy Arceri [Wed, 26 Aug 2015 12:18:36 +0000 (22:18 +1000)]
nir: support indirect indexing samplers in struct arrays

As a bonus we get indirect support for arrays of arrays for free.

V5: couple of small clean-ups suggested by Jason.

V4: fix struct member location caclulation, use nir_ssa_def rather than
nir_src for the indirect as suggested by Jason

V3: Use nir_instr_rewrite_src() with empty src rather then clearing
the use_link list directly for the old indirects as suggested by Jason

V2: Fixed validation error in debug build

Reviewed-by: Jason Ekstrand <jason.ekstrand@intel.com>
9 years agoglsl: add helper for calculating offsets for struct members
Timothy [Fri, 11 Sep 2015 21:33:27 +0000 (07:33 +1000)]
glsl: add helper for calculating offsets for struct members

V2: update comments

Reviewed-by: Jason Ekstrand <jason.ekstrand@intel.com>
9 years agoglsl: make variables private
Timothy Arceri [Tue, 1 Sep 2015 05:52:10 +0000 (15:52 +1000)]
glsl: make variables private

Reviewed-by: Jason Ekstrand <jason.ekstrand@intel.com>
9 years agoglsl: store uniform slot id in var location field
Timothy Arceri [Sun, 30 Aug 2015 02:50:34 +0000 (12:50 +1000)]
glsl: store uniform slot id in var location field

This will allow us to access the uniform later on without resorting to
building a name string and looking it up in UniformHash.

V3: remove line wrap change from this patch

V2: store slot number for all non-UBO uniforms to make code more
consitent, renamed explicit_binding to explicit_location and added
comment about what it does. Store the location at every shader stage.
Updated data.location comments in ir/nir.h.

Reviewed-by: Jason Ekstrand <jason.ekstrand@intel.com>
9 years agoglsl: assign hidden uniforms their slot id earlier
Timothy Arceri [Wed, 2 Sep 2015 01:29:11 +0000 (11:29 +1000)]
glsl: assign hidden uniforms their slot id earlier

This is required so that the next patch can safely assign the slot id
to the var.

The ids are now assigned in the order we want before allocating storage
so there is no need to sort the storage array and move things around.

V2: rename variable to make code easier to follow as suggested by Jason

Reviewed-by: Jason Ekstrand <jason.ekstrand@intel.com>
9 years agoglsl: order indices for samplers inside a struct array
Timothy Arceri [Sun, 30 Aug 2015 02:49:46 +0000 (12:49 +1000)]
glsl: order indices for samplers inside a struct array

This allows the correct offset to be easily calculated for indirect
indexing when a struct array contains multiple samplers, or any crazy
nesting.

The indices for the folling struct will now look like this:
Sampler index: 0 Name: s[0].tex
Sampler index: 1 Name: s[1].tex
Sampler index: 2 Name: s[0].si.tex
Sampler index: 3 Name: s[1].si.tex
Sampler index: 4 Name: s[0].si.tex2
Sampler index: 5 Name: s[1].si.tex2

Before this change it looked like this:
Sampler index: 0 Name: s[0].tex
Sampler index: 3 Name: s[1].tex
Sampler index: 1 Name: s[0].si.tex
Sampler index: 4 Name: s[1].si.tex
Sampler index: 2 Name: s[0].si.tex2
Sampler index: 5 Name: s[1].si.tex2

struct S_inner {
   sampler2D tex;
   sampler2D tex2;
};

struct S {
   sampler2D tex;
   S_inner si;
};

uniform S s[2];

V3: Update comments with suggestions from Jason

V2: rename struct array counter to have better name

Reviewed-by: Jason Ekstrand <jason.ekstrand@intel.com>
9 years agoRevert "mesa/extensions: restrict GL_OES_EGL_image to GLES"
Dave Airlie [Wed, 16 Sep 2015 20:55:31 +0000 (06:55 +1000)]
Revert "mesa/extensions: restrict GL_OES_EGL_image to GLES"

This reverts commit 48961fa3ba37999a6f8fd812458b735e39604a95.

glamor/Xwayland use this, the spec saying something when it
was written, and the fact that the comment says Mesa relies on it
hasn't changed.

I also don't have a copy of this patch in my mail archive, which
seems wierd, did it get posted to mesa-dev?

Signed-off-by: Dave Airlie <airlied@redhat.com>
9 years agovc4: Only build in simulator mode if we find pkg-config for it.
Eric Anholt [Wed, 16 Sep 2015 19:51:00 +0000 (15:51 -0400)]
vc4: Only build in simulator mode if we find pkg-config for it.

This will let other developers build it x86 for build-testing purposes.

9 years agofreedreno/a3xx: use NUM_USER_CLIP_PLANES helper instead of magic number
Ilia Mirkin [Wed, 16 Sep 2015 18:19:21 +0000 (14:19 -0400)]
freedreno/a3xx: use NUM_USER_CLIP_PLANES helper instead of magic number

Use the helper from the newly-updated generated header file.

Signed-off-by: Ilia Mirkin <imirkin@alum.mit.edu>
9 years agofreedreno/a3xx: fix blending of L8 format
Ilia Mirkin [Mon, 14 Sep 2015 05:59:01 +0000 (01:59 -0400)]
freedreno/a3xx: fix blending of L8 format

Even though luminance formats don't have alpha, we still want the alpha
output to go to the blender. This fixes the luminance blending tests.

Signed-off-by: Ilia Mirkin <imirkin@alum.mit.edu>
Cc: "11.0" <mesa-stable@lists.freedesktop.org>
9 years agofreedreno/a3xx: add support for dual-source blending
Ilia Mirkin [Sun, 13 Sep 2015 23:50:45 +0000 (19:50 -0400)]
freedreno/a3xx: add support for dual-source blending

Signed-off-by: Ilia Mirkin <imirkin@alum.mit.edu>
9 years agovc4: convert from tgsi semantic/index to varying-slot
Eric Anholt [Wed, 9 Sep 2015 17:23:55 +0000 (13:23 -0400)]
vc4: convert from tgsi semantic/index to varying-slot

(originally part of previous patch, split out to separate patch by Rob)

v2: squash in some fixes from Eric
v3: Another fix from Eric for point coords.

Signed-off-by: Rob Clark <robclark@freedesktop.org>
9 years agogallium/ttn: Convert to using VARYING_SLOT_* / FRAG_RESULT_*.
Eric Anholt [Tue, 4 Aug 2015 21:28:02 +0000 (14:28 -0700)]
gallium/ttn: Convert to using VARYING_SLOT_* / FRAG_RESULT_*.

This avoids exceeding the size of the .index bitfield since it got
truncated, and should make our NIR look more like the NIR that the rest of
the NIR developers are working on.

v2: split out vc4 updates, first patch uses varying_slot_to_tgsi_semantic()
    helper, and second patch does the actual conversion.
v3: add frag_result_to_tgsi_semantic() helper and don't try to map
    frag_results to semantic name/index as if they were varying_slot's
v4: use VERT_ATTRIB_ for VS inputs
v5: Fix vc4 build.

Signed-off-by: Rob Clark <robclark@freedesktop.org>
9 years agonv50, nvc0: fix max texture buffer size to 128M elements
Ilia Mirkin [Tue, 15 Sep 2015 23:39:25 +0000 (19:39 -0400)]
nv50, nvc0: fix max texture buffer size to 128M elements

This is what the hardware supports, there never was any sort of 64K
limit.

Signed-off-by: Ilia Mirkin <imirkin@alum.mit.edu>
Cc: "10.6 11.0" <mesa-stable@lists.freedesktop.org>
9 years agost/mesa: avoid integer overflows with buffers >= 512MB
Ilia Mirkin [Tue, 15 Sep 2015 23:32:10 +0000 (19:32 -0400)]
st/mesa: avoid integer overflows with buffers >= 512MB

This fixes failures with the newly-submitted max-size texture buffer
piglit test for GPUs exposing >= 128M max texels.

Signed-off-by: Ilia Mirkin <imirkin@alum.mit.edu>
Cc: "10.6 11.0" <mesa-stable@lists.freedesktop.org>
Reviewed-by: Glenn Kennard <glenn.kennard@gmail.com>
9 years agomesa: move GL_APPLE_object_purgeable functions to new file
Brian Paul [Tue, 15 Sep 2015 20:28:38 +0000 (14:28 -0600)]
mesa: move GL_APPLE_object_purgeable functions to new file

Move this code out of bufferobj.c since it's not strongly connected to
buffer objects.

Acked-by: Matt Turner <mattst88@gmail.com>
9 years agomesa: remove trailing whitespace in bufferobj.c
Brian Paul [Tue, 15 Sep 2015 20:04:58 +0000 (14:04 -0600)]
mesa: remove trailing whitespace in bufferobj.c

Trivial.

9 years agomesa: whitespace, line wrap fixes in varray.c
Brian Paul [Tue, 15 Sep 2015 20:03:04 +0000 (14:03 -0600)]
mesa: whitespace, line wrap fixes in varray.c

Trivial.

9 years agonir/print: print symbolic names from shader-enum
Rob Clark [Tue, 15 Sep 2015 22:55:48 +0000 (18:55 -0400)]
nir/print: print symbolic names from shader-enum

v2: split out moving of FILE *fp into state structure into it's own
(more complete patch) to reduce the noise in this one

Signed-off-by: Rob Clark <robclark@freedesktop.org>
Reviewed-by: Connor Abbott <cwabbott0@gmail.com>
Reviewed-by: Iago Toral Quiroga <itoral@igalia.com>
9 years agonir/print: bit of state refactoring
Rob Clark [Tue, 15 Sep 2015 22:50:41 +0000 (18:50 -0400)]
nir/print: bit of state refactoring

Rename print_var_state to print_state, and stuff FILE ptr into the state
object.  This avoids passing around an extra parameter everywhere.

v2: even more extensive conversion.. use state *everywhere* instead of
FILE ptr, and convert nir_print_instr() to use state as well

Signed-off-by: Rob Clark <robclark@freedesktop.org>
Reviewed-by: Iago Toral Quiroga <itoral@igalia.com>
Reviewed-by: Connor Abbott <cwabbott0@gmail.com>
9 years agoglsl: shader-enum to name debug fxns
Rob Clark [Fri, 11 Sep 2015 16:48:05 +0000 (12:48 -0400)]
glsl: shader-enum to name debug fxns

Signed-off-by: Rob Clark <robclark@freedesktop.org>
Reviewed-by: Iago Toral Quiroga <itoral@igalia.com>
9 years agofreedreno: one screen to rule them all
Rob Clark [Fri, 4 Sep 2015 15:35:33 +0000 (11:35 -0400)]
freedreno: one screen to rule them all

Similar to fee0686c21c631d96d6042741267a3c218c23ffc, but in this case to
ensure that drm_gralloc and libGLES_mesa are sharing a single screen.

Bumps libdrm_freedreno version dependency, as it requires the new
fd_device_fd() API.

Signed-off-by: Rob Clark <robclark@freedesktop.org>
9 years agofreedreno/ir3: use NIR to lower ffract instead of tgsi_lowering
Rob Clark [Mon, 14 Sep 2015 15:54:05 +0000 (11:54 -0400)]
freedreno/ir3: use NIR to lower ffract instead of tgsi_lowering

Signed-off-by: Rob Clark <robclark@freedesktop.org>
9 years agonir: add lowering for ffract
Rob Clark [Mon, 14 Sep 2015 15:13:19 +0000 (11:13 -0400)]
nir: add lowering for ffract

Signed-off-by: Rob Clark <robclark@freedesktop.org>
Reviewed-by: Iago Toral Quiroga <itoral@igalia.com>
9 years agoi965/fs: The barrier send uses only 1 payload register
Jordan Justen [Tue, 15 Sep 2015 21:01:17 +0000 (14:01 -0700)]
i965/fs: The barrier send uses only 1 payload register

When preparing the barrier payload, the instructions should operate in
simd8 mode since we only use 1 payload register.

fs_inst::regs_read is also updated to indicate that it only reads one
register for SHADER_OPCODE_BARRIER.

These issues were flagged by:

commit cadd7dd384b33a779d46bd664f456bed4a21a5b7
Author: Jason Ekstrand <jason.ekstrand@intel.com>
Date:   Thu Jul 2 15:41:02 2015 -0700

    i965/fs: Add a very basic validation pass

Signed-off-by: Jordan Justen <jordan.l.justen@intel.com>
Reviewed-by: Jason Ekstrand <jason.ekstrand@intel.com>
9 years agonir/builder: Use a normal temporary array in nir_channel
Jason Ekstrand [Tue, 15 Sep 2015 19:09:06 +0000 (12:09 -0700)]
nir/builder: Use a normal temporary array in nir_channel

C++ gets cranky if we take references of temporaries.  This isn't a problem
yet in master because nir_builder is never used from C++.  However, it will
be in the future so we should fix it now.

Reviewed-by: Rob Clark <robclark@freedesktop.org>
9 years agofreedreno/a4xx: more texture formats
Rob Clark [Mon, 14 Sep 2015 19:15:06 +0000 (15:15 -0400)]
freedreno/a4xx: more texture formats

Signed-off-by: Rob Clark <robclark@freedesktop.org>
9 years agofreedreno/a4xx: border-color support
Rob Clark [Tue, 15 Sep 2015 21:25:47 +0000 (17:25 -0400)]
freedreno/a4xx: border-color support

Signed-off-by: Rob Clark <robclark@freedesktop.org>
9 years agofreedreno/a4xx: wire up texture clamp lowering
Rob Clark [Tue, 15 Sep 2015 21:25:25 +0000 (17:25 -0400)]
freedreno/a4xx: wire up texture clamp lowering

Signed-off-by: Rob Clark <robclark@freedesktop.org>
9 years agofreedreno: helper for a3xx/a4xx border-colors
Rob Clark [Tue, 15 Sep 2015 13:23:21 +0000 (09:23 -0400)]
freedreno: helper for a3xx/a4xx border-colors

Both use the same layout for the buffer containing border-color values,
so rather than duplicating the logic in a4xx, split it out into a
helper.

Signed-off-by: Rob Clark <robclark@freedesktop.org>
9 years agofreedreno: update generated headers
Rob Clark [Mon, 14 Sep 2015 20:59:36 +0000 (16:59 -0400)]
freedreno: update generated headers

Signed-off-by: Rob Clark <robclark@freedesktop.org>
9 years agonir/lower_vec_to_movs: Coalesce into destinations of fdot instructions
Jason Ekstrand [Thu, 10 Sep 2015 00:18:55 +0000 (17:18 -0700)]
nir/lower_vec_to_movs: Coalesce into destinations of fdot instructions

Now that we have a replicating fdot instruction, we can actually coalesce
into the destinations of vec4 instructions.  We couldn't really do this
before because, if the destination had to end up in .z, we couldn't
reswizzle the instruction.  With a replicated destination, the result ends
up in all channels so we can just set the writemask and we're done.

Shader-db results for vec4 programs on Haswell:

   total instructions in shared programs: 1747753 -> 1746280 (-0.08%)
   instructions in affected programs:     143274 -> 141801 (-1.03%)
   helped:                                667
   HURT:                                  0

It turns out that dot-products matter...

Reviewed-by: Eduardo Lima Mitev <elima@igalia.com>
9 years agoi965/vec4: Use the replicated fdot instruction in NIR
Jason Ekstrand [Thu, 10 Sep 2015 18:08:15 +0000 (11:08 -0700)]
i965/vec4: Use the replicated fdot instruction in NIR

Reviewed-by: Connor Abbott <cwabbott0@gmail.com>
Reviewed-by: Eduardo Lima Mitev <elima@igalia.com>
9 years agonir: Add a fdot instruction that replicates the result to a vec4
Jason Ekstrand [Thu, 10 Sep 2015 17:51:46 +0000 (10:51 -0700)]
nir: Add a fdot instruction that replicates the result to a vec4

Fortunately, nir_constant_expr already auto-splats if "dst" never shows up
in the constant expression field so we don't need to do anything there.

Reviewed-by: Connor Abbott <cwabbott0@gmail.com>
Reviewed-by: Eduardo Lima Mitev <elima@igalia.com>
9 years agonir/lower_vec_to_movs: Coalesce movs on-the-fly when possible
Jason Ekstrand [Wed, 9 Sep 2015 21:40:06 +0000 (14:40 -0700)]
nir/lower_vec_to_movs: Coalesce movs on-the-fly when possible

The old pass blindly inserted a bunch of moves into the shader with no
concern for whether or not it was really needed.  This adds code to try and
coalesce into the destination of the instruction providing the value.

Shader-db results for vec4 shaders on Haswell:

   total instructions in shared programs: 1754420 -> 1747753 (-0.38%)
   instructions in affected programs:     231230 -> 224563 (-2.88%)
   helped:                                1017
   HURT:                                  2

This approach is heavily based on a different patch by Eduardo Lima Mitev
<elima@igalia.com>.  Eduardo's patch did this in a separate pass as opposed
to integrating it into nir_lower_vec_to_movs.

Reviewed-by: Eduardo Lima Mitev <elima@igalia.com>
9 years agonir/lower_vec_to_movs: Get rid of start_idx and swizzle compacting
Jason Ekstrand [Wed, 9 Sep 2015 21:47:28 +0000 (14:47 -0700)]
nir/lower_vec_to_movs: Get rid of start_idx and swizzle compacting

Previously, we did this thing with keeping track of a separate start_idx
which was different from the iteration variable.  I think this was a relic
of the way that GLSL IR implements writemasks.  In NIR, if a given bit in
the writemask is unset then that channel is just "unused", not missing.  In
particular, a vec4 operation with a writemask of 0xd will use sources 0, 2,
and 3 and leave source 1 alone.  We can simplify things a good deal (and
make them correct) by removing this "compacting" step.

Reviewed-by: Eduardo Lima Mitev <elima@igalia.com>
Reviewed-by: Eric Anholt <eric@anholt.net>
9 years agoi965/vec4_nir: Use partial SSA form rather than full non-SSA
Jason Ekstrand [Wed, 9 Sep 2015 20:55:39 +0000 (13:55 -0700)]
i965/vec4_nir: Use partial SSA form rather than full non-SSA

We made this switch in the FS backend some time ago and it seems to make a
number of things a bit easier.  In particular, supporting SSA values takes
very little work in the backend and allows us to take advantage of the
majority of the SSA information even after we've gotten rid of Phi nodes.

Reviewed-by: Eduardo Lima Mitev <elima@igalia.com>
9 years agonir/lower_vec_to_movs: Handle partially SSA shaders
Jason Ekstrand [Wed, 9 Sep 2015 20:42:14 +0000 (13:42 -0700)]
nir/lower_vec_to_movs: Handle partially SSA shaders

v2 (Jason Ekstrand):
 - Use nir_instr_rewrite_dest
 - Pass the impl directly into lower_vec_to_movs_block

Reviewed-by: Eduardo Lima Mitev <elima@igalia.com>
9 years agonir/lower_vec_to_movs: Pass the shader around directly
Jason Ekstrand [Wed, 9 Sep 2015 19:58:58 +0000 (12:58 -0700)]
nir/lower_vec_to_movs: Pass the shader around directly

Previously, we were passing the shader around, we were just calling it
"mem_ctx".  However, the nir_shader is (and must be for the purposes of
mark-and-sweep) the mem_ctx so we might as well pass it around explicitly.

Reviewed-by: Eduardo Lima Mitev <elima@igalia.com>
9 years agoi965/fs: Add a very basic validation pass
Jason Ekstrand [Thu, 2 Jul 2015 22:41:02 +0000 (15:41 -0700)]
i965/fs: Add a very basic validation pass

Currently the validation pass only validates that regs_read and
regs_written are consistent with the sizes of VGRF's.  We can add more as
we find it to be useful.

Reviewed-by: Matt Turner <mattst88@gmail.com>
9 years agoi965/fs_surface_builder: Only apply predicate to components that exist
Jason Ekstrand [Mon, 14 Sep 2015 22:36:24 +0000 (15:36 -0700)]
i965/fs_surface_builder: Only apply predicate to components that exist

In certain conditions, we have to do bounds-checking in the shader for
image_load_store.  The way this works for image loads is that we do a
predicated load and then emit a series of selects, one per component,
that gives us 0 or the loaded value depending on whether or not you're
in bounds.  However, we were hard-coding 4 components which may not be
correct.  Instead, we should be using size which is the number of
components read.

Reviewed-by: Francisco Jerez <currojerez@riseup.net>
9 years agoi965/fs: Only read output_components many components when writing an output
Jason Ekstrand [Mon, 14 Sep 2015 21:18:13 +0000 (14:18 -0700)]
i965/fs: Only read output_components many components when writing an output

Reviewed-by: Kristian Høgsberg <krh@bitplanet.net>
9 years agoi965/fs: Set output_components for lowered clip distance outputs
Jason Ekstrand [Mon, 14 Sep 2015 22:09:00 +0000 (15:09 -0700)]
i965/fs: Set output_components for lowered clip distance outputs

Reviewed-by: Kristian Høgsberg <krh@bitplanet.net>
9 years agomesa/teximage: restrict GL_ETC1_RGB8_OES support to GLES
Nanley Chery [Thu, 27 Aug 2015 23:05:22 +0000 (16:05 -0700)]
mesa/teximage: restrict GL_ETC1_RGB8_OES support to GLES

According to the extensions table and our glext headers,
OES_compressed_ETC1_RGB8_texture is only supported in
GLES1 and GLES2. Since we may give users a GLES3 context
when a GLES2 context is requested, we also allow this
extension for GLES3 as well.

Reviewed-by: Anuj Phogat <anuj.phogat@gmail.com>
Signed-off-by: Nanley Chery <nanley.g.chery@intel.com>
9 years agomesa/extensions: restrict GL_OES_EGL_image to GLES
Nanley Chery [Thu, 10 Sep 2015 17:48:46 +0000 (10:48 -0700)]
mesa/extensions: restrict GL_OES_EGL_image to GLES

Driver vendors do this as well. The extension specification
lists GLES 1.1 or 2.0 as requirements.

Reviewed-by: Chad Versace <chad.versace@intel.com>
Signed-off-by: Nanley Chery <nanley.g.chery@intel.com>
9 years agomesa/extensions: restrict luminance alpha formats to API_OPENGL_COMPAT
Nanley Chery [Thu, 27 Aug 2015 23:05:22 +0000 (16:05 -0700)]
mesa/extensions: restrict luminance alpha formats to API_OPENGL_COMPAT

According the GL 3.1 spec, luminance alpha formats are deprecated.

Reviewed-by: Anuj Phogat <anuj.phogat@gmail.com>
Signed-off-by: Nanley Chery <nanley.g.chery@intel.com>
9 years agogallium/svga: Enable PIPE_FORMAT_L8_UNORM for vgpu10
Thomas Hellstrom [Tue, 15 Sep 2015 06:40:07 +0000 (23:40 -0700)]
gallium/svga: Enable PIPE_FORMAT_L8_UNORM for vgpu10

It's extensively used by XA for a8- and planar yuv component surfaces.
This fixes broken XA yuv blits using vgpu10 contexts.

Signed-off-by: Thomas Hellstrom <thellstrom@vmware.com>
Reviewed-by: Brian Paul <brianp@vmware.com>
9 years agoegl/dri2: don't leak the fd on dri2_terminate
Emil Velikov [Thu, 10 Sep 2015 13:41:38 +0000 (14:41 +0100)]
egl/dri2: don't leak the fd on dri2_terminate

Currently the check was incorrect as it did not consider the (unlikely)
case of fd == 0. In order to fix this we should first correctly
initialize it to -1, as the swrast implementations leave it set to zero
(props to calloc()).

Signed-off-by: Emil Velikov <emil.l.velikov@gmail.com>
Reviewed-by: Boyan Ding <boyan.j.ding@gmail.com>
9 years agoegl/dri2/drm: compact existing device mgmt
Emil Velikov [Mon, 7 Sep 2015 08:53:53 +0000 (09:53 +0100)]
egl/dri2/drm: compact existing device mgmt

Move the fcntl(dupfd_cloexec) to the else branch where it belongs.
Otherwise it's not immediately obvious that the code is hit, only when
an existing device is used.

Signed-off-by: Emil Velikov <emil.l.velikov@gmail.com>
Reviewed-by: Boyan Ding <boyan.j.ding@gmail.com>
9 years agoegl/dri2: Close file descriptor on error.
Matt Turner [Wed, 15 Jul 2015 16:00:41 +0000 (09:00 -0700)]
egl/dri2: Close file descriptor on error.

v2: [Emil Velikov]
Rework the error path to a common goto, close only if we own the fd.
v3; [Emil Velikov]
Always close the fd (we either opened the device or dup'd) (Boyan, Ian)

Signed-off-by: Emil Velikov <emil.l.velikov@gmail.com>
Reviewed-by: Boyan Ding <boyan.j.ding@gmail.com>
9 years agogbm: convert gbm bo format to fourcc format on dma-buf import
Ray Strode [Fri, 28 Aug 2015 18:50:21 +0000 (14:50 -0400)]
gbm: convert gbm bo format to fourcc format on dma-buf import

At the moment if a gbm buffer is imported and the gbm buffer
has an old-style GBM_BO_FORMAT format, the import will crash,
since it's passed directly to DRI functions that expect
a fourcc format (as provided by the newer GBM_FORMAT
definitions)

This commit addresses the problem in two ways:

1) it prevents invalid formats from leading to a crash by
returning EINVAL if the image couldn't be created

2) it translates GBM_BO_FORMAT formats into the comparable
GBM_FORMAT formats.

Reference: https://bugzilla.gnome.org/show_bug.cgi?id=753531
CC: "10.6 11.0" <mesa-stable@lists.freedesktop.org>
Reviewed-by: Emil Velikov <emil.l.velikov@gmail.com>
Reviewed-by: Kristian Høgsberg <krh@bitplanet.net>
9 years agodocs: document INTEL_DEBUG 'optimizer' envvar
Alejandro Piñeiro [Mon, 14 Sep 2015 18:16:25 +0000 (20:16 +0200)]
docs: document INTEL_DEBUG 'optimizer' envvar

Reviewed-by: Matt Turner <mattst88@gmail.com>
9 years agoi965: Move perf_debug code to brw_codegen_*_prog()
Kristian Høgsberg Kristensen [Sat, 5 Sep 2015 00:09:40 +0000 (17:09 -0700)]
i965: Move perf_debug code to brw_codegen_*_prog()

We're trying to avoid a libdrm dependency in the core compiler, so let's
move the perf_debug code one level up from the brw_*_emit() helpers to
the brw_codegen_*_prog() helpers.

Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Reviewed-by: Jordan Justen <jordan.l.justen@intel.com>
Signed-off-by: Kristian Høgsberg Kristensen <krh@bitplanet.net>
9 years agoi965: Move brw_fs_precompile() to brw_wm.c
Kristian Høgsberg Kristensen [Fri, 4 Sep 2015 23:55:03 +0000 (16:55 -0700)]
i965: Move brw_fs_precompile() to brw_wm.c

All other precompile functions live in the brw_<stage>.c files, make fs
follow the convention.

Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Reviewed-by: Jordan Justen <jordan.l.justen@intel.com>
Signed-off-by: Kristian Høgsberg Kristensen <krh@bitplanet.net>
9 years agoi965: Move compute shader code around
Kristian Høgsberg Kristensen [Fri, 4 Sep 2015 23:35:34 +0000 (16:35 -0700)]
i965: Move compute shader code around

This moves the compute shader code around in order to make the way the
code is split up more consistent. There should be no functional changes.
Typically we have a few files per stage:

    brw_vs.c, brw_wm.c brw_gs.c:

        code to drive code generation and implement precompiling and
        cache search.

    genX_<stage>_state.c

        gen specific implementation of the state emission for the shader
        stage.

The brw_*_emit() functions are all in the same files as the visitor
classes they use (with the exception of VS, which may use either vec4 or
fs).

To make compute follow this convention, we move the brw_cs_emit()
function into brw_fs.cpp. We can then rename brw_cs.cpp to brw_cs.c and
do this in C like the other similar files.  Finally, move state setup
and atoms to gen7_cs_state.c.

Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Reviewed-by: Jordan Justen <jordan.l.justen@intel.com>
Signed-off-by: Kristian Høgsberg Kristensen <krh@bitplanet.net>
9 years agometa: Abort meta pbo path if TexSubImage need signed unsigned conversion
Anuj Phogat [Fri, 24 Jul 2015 22:53:58 +0000 (15:53 -0700)]
meta: Abort meta pbo path if TexSubImage need signed unsigned conversion

See similar fix for Readpixels in mesa commit 0d20790. Jason suggested
we need that for TexSubImage as well.

Cc: <mesa-stable@lists.freedesktop.org>
Signed-off-by: Anuj Phogat <anuj.phogat@gmail.com>
Reviewed-by: Jason Ekstrand <jason.ekstrand@intel.com>
9 years agonvc0/ir: start offset at texBindBase for txq, like regular texturing
Ilia Mirkin [Fri, 11 Sep 2015 03:58:17 +0000 (23:58 -0400)]
nvc0/ir: start offset at texBindBase for txq, like regular texturing

Curiously this has no actual effect. I think it's because the first 8
textures are bound in multiple slots for some reason. However seems
prudent to use these the same way as regular texturing, esp in the case
where there are more than 8 textures bound.

Signed-off-by: Ilia Mirkin <imirkin@alum.mit.edu>
9 years agovc4: Fix build from recent NIR cleanups.
Eric Anholt [Mon, 14 Sep 2015 15:21:07 +0000 (11:21 -0400)]
vc4: Fix build from recent NIR cleanups.

9 years agoi965/vec4_nir: Load constants as integers
Antia Puentes [Mon, 14 Sep 2015 07:50:59 +0000 (09:50 +0200)]
i965/vec4_nir: Load constants as integers

Loads constants using integer as their register type, like it is
done in FS backend.

No shader-db changes in HSW.

Cc: "10.6 11.0" <mesa-stable@lists.freedesktop.org>
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=91716
Reviewed-by: Jason Ekstrand <jason.ekstrand@intel.com>
9 years agoi965/vec4: Fix saturation errors when coalescing registers
Antia Puentes [Wed, 5 Aug 2015 13:57:33 +0000 (15:57 +0200)]
i965/vec4: Fix saturation errors when coalescing registers

If the register types do not match and the instruction
that contains the final destination is saturated, register
coalescing generated non-equivalent code.

This did not happen when using IR because types usually
matched, but it is visible in nir-vec4.

For example,
   mov      vgrf7:D vgrf2:D
   mov.sat  m4:F vgrf7:F

is coalesced to:
   mov.sat  m4:D vgrf2:D

The patch prevents coalescing in such scenario, unless the
instruction we want to coalesce into is a MOV (without type
conversion implied). In that case, the patch sets the register
types to the type of the final destination.

Shader-db results in HSW (only vec4 instructions shown):

total instructions in shared programs: 1754415 -> 1754416 (0.00%)
instructions in affected programs:     74 -> 75 (1.35%)
helped:                                0
HURT:                                  1
GAINED:                                0
LOST:                                  0

Only one extra instruction in one of the shaders, that comes from
eliminating a saturation error by preventing register coalesce.

Cc: "10.6 11.0" <mesa-stable@lists.freedesktop.org>
Reviewed-by: Jason Ekstrand <jason.ekstrand@intel.com>
9 years agodocs: cleanups + mark some work as done
Tapani Pälli [Mon, 14 Sep 2015 05:50:51 +0000 (08:50 +0300)]
docs: cleanups + mark some work as done

Signed-off-by: Tapani Pälli <tapani.palli@intel.com>
Reviewed-by: Ilia Mirkin <imirkin@alum.mit.edu>
9 years agodocs: only astc ldr required for ES3.2, not hdr
Ilia Mirkin [Mon, 14 Sep 2015 05:07:05 +0000 (01:07 -0400)]
docs: only astc ldr required for ES3.2, not hdr

Signed-off-by: Ilia Mirkin <imirkin@alum.mit.edu>
Reviewed-by: Tapani Pälli <tapani.palli@intel.com>
9 years agost/mesa: emit TXQS, support ARB_shader_texture_image_samples
Ilia Mirkin [Fri, 11 Sep 2015 01:44:45 +0000 (21:44 -0400)]
st/mesa: emit TXQS, support ARB_shader_texture_image_samples

The image component of the ext is a no-op since there is no image support
in gallium (yet).

Signed-off-by: Ilia Mirkin <imirkin@alum.mit.edu>
Reviewed-by: Edward O'Callaghan <eocallaghan@alterapraxis.com>
Reviewed-by: Brian Paul <brianp@vmware.com>
9 years agor600g: add support for TXQS tgsi opcode
Ilia Mirkin [Fri, 11 Sep 2015 02:33:34 +0000 (22:33 -0400)]
r600g: add support for TXQS tgsi opcode

Signed-off-by: Ilia Mirkin <imirkin@alum.mit.edu>
Reviewed-by: Glenn Kennard <glenn.kennard@gmail.com>
9 years agonv50/ir: add support for TXQS tgsi opcode
Ilia Mirkin [Fri, 11 Sep 2015 02:07:27 +0000 (22:07 -0400)]
nv50/ir: add support for TXQS tgsi opcode

Signed-off-by: Ilia Mirkin <imirkin@alum.mit.edu>
9 years agogallium: add PIPE_CAP_TGSI_TXQS to let st know if TXQS is supported
Ilia Mirkin [Fri, 11 Sep 2015 21:29:49 +0000 (17:29 -0400)]
gallium: add PIPE_CAP_TGSI_TXQS to let st know if TXQS is supported

Signed-off-by: Ilia Mirkin <imirkin@alum.mit.edu>
Reviewed-by: Glenn Kennard <glenn.kennard@gmail.com>