mesa.git
9 years agoegl_dri2: Remove trailing whitespaces
Boyan Ding [Sat, 13 Jun 2015 07:36:27 +0000 (15:36 +0800)]
egl_dri2: Remove trailing whitespaces

Signed-off-by: Boyan Ding <boyan.j.ding@gmail.com>
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
9 years agoi965/skl: Fix aligning mt->total_width to the block size
Neil Roberts [Tue, 16 Jun 2015 12:53:40 +0000 (13:53 +0100)]
i965/skl: Fix aligning mt->total_width to the block size

brw_miptree_layout_2d tries to ensure that mt->total_width is a
multiple of the compressed block size, presumably because it wouldn't
be possible to make an image that has a fraction of a block. However
it was doing this by aligning mt->total_width to align_w. Previously
align_w has been used as a shortcut for getting the block width
because before Gen9 the block width was always equal to the alignment.
Commit 4ab8d59a2 tried to fix these cases to use the block width
instead of the alignment but it missed this case.

I think in practice this probably won't make any difference because
the buffer for the texture will be allocated to be large enough to
contain the entire pitch and libdrm aligns the pitch to the tile width
anyway. However I think the patch is worth having to make the
intention clearer.

Reviewed-by: Anuj Phogat <anuj.phogat@gmail.com>
Reviewed-by: Ben Widawsky <ben@bwidawsk.net>
9 years agomesa: Enable subdir-objects globally.
Matt Turner [Wed, 10 Jun 2015 23:30:56 +0000 (16:30 -0700)]
mesa: Enable subdir-objects globally.

Reviewed-by: Emil Velikov <emil.l.velikov@gmail.com>
9 years agomesa: fold duplicated GL/GL_CORE/GLES3 entry in get_hash_params.py
Emil Velikov [Wed, 24 Jun 2015 13:06:33 +0000 (14:06 +0100)]
mesa: fold duplicated GL/GL_CORE/GLES3 entry in get_hash_params.py

Signed-off-by: Emil Velikov <emil.l.velikov@gmail.com>
Reviewed-by: Matt Turner <mattst88@gmail.com>
9 years agoilo: define ILO_IMAGE_MAX_LEVEL_COUNT
Chia-I Wu [Fri, 26 Jun 2015 03:38:46 +0000 (11:38 +0800)]
ilo: define ILO_IMAGE_MAX_LEVEL_COUNT

Define ILO_IMAGE_MAX_LEVEL_COUNT for ilo_image and remove unnecessary header
includes.

9 years agoilo: replace pipe_format by gen_surface_format
Chia-I Wu [Thu, 25 Jun 2015 14:27:04 +0000 (22:27 +0800)]
ilo: replace pipe_format by gen_surface_format

Replace pipe_format by gen_surface_format in ilo_image.  Change how depth
format is specified in ilo_state_zs.

9 years agoilo: always use the specified image format
Chia-I Wu [Wed, 24 Jun 2015 23:43:47 +0000 (07:43 +0800)]
ilo: always use the specified image format

Move silent promotion of PIPE_FORMAT_ETC1_RGB8 or combined depth/stencil out
of core.

9 years agoilo: replace pipe_texture_target by gen_surface_type
Chia-I Wu [Wed, 24 Jun 2015 14:47:30 +0000 (22:47 +0800)]
ilo: replace pipe_texture_target by gen_surface_type

Replace pipe_texture_target by gen_surface_type in ilo_image.  Change how
GEN6_SURFTYPE_CUBE is specified in ilo_state_surface and ilo_state_zs.

9 years agoilo: initialize ilo_image from ilo_image_info
Chia-I Wu [Wed, 24 Jun 2015 14:46:36 +0000 (22:46 +0800)]
ilo: initialize ilo_image from ilo_image_info

Convert pipe_resource to ilo_image_info for image initialization.

9 years agoilo: remove ilo_image_disable_aux()
Chia-I Wu [Wed, 24 Jun 2015 23:18:31 +0000 (07:18 +0800)]
ilo: remove ilo_image_disable_aux()

Fail resource creation when aux bo allocation fails.

9 years agoilo: improve SURFTYPE_BUFFER validations
Chia-I Wu [Wed, 24 Jun 2015 04:57:57 +0000 (12:57 +0800)]
ilo: improve SURFTYPE_BUFFER validations

Reorganize the validations to make them more systematic.

9 years agoilo: remove ilo_buffer
Chia-I Wu [Fri, 26 Jun 2015 05:08:32 +0000 (13:08 +0800)]
ilo: remove ilo_buffer

Since the addition of ilo_vma, it was used only to pad a bo for sampling
engine surfaces.  Replace it entirely with these functions

  ilo_state_surface_buffer_size()
  ilo_state_vertex_buffer_size()
  ilo_state_index_buffer_size()
  ilo_state_sol_buffer_size()

9 years agoilo: introduce ilo_vma
Chia-I Wu [Wed, 24 Jun 2015 05:59:31 +0000 (23:59 -0600)]
ilo: introduce ilo_vma

This cleans up the code a bit and makes ilo_state_vector_resource_renamed()
simpler and more robust.  It also allows a single bo to back mulitple VMAs.

9 years agomesa: remove unnecessary checks in _mesa_readpixels_needs_slow_path
Iago Toral Quiroga [Tue, 23 Jun 2015 06:42:14 +0000 (08:42 +0200)]
mesa: remove unnecessary checks in _mesa_readpixels_needs_slow_path

readpixels_can_use_memcpy will later call _mesa_format_matches_format_and_type
which does much tighter checks than these to decide if we can use
memcpy for readpixels.

Also, the checks do not seem to be extensive enough anyway, since we are
checking for signed/unsigned conversion only when the framebuffer has integers,
but the same checks could be done for other types anyway, since as long as
there is a signed/unsigned conversion we can't memcpy.

No regressions observed on i965/llvmpipe.

Reviewed-by: Anuj Phogat <anuj.phogat@gmail.com>
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
9 years agoi965/vec4_live_variables: Do liveness analysis bottom-to-top
Jason Ekstrand [Thu, 25 Jun 2015 15:08:27 +0000 (08:08 -0700)]
i965/vec4_live_variables: Do liveness analysis bottom-to-top

From Muchnick's Advanced Compiler Design and Implementation:

"To determine which variables are live at each point in a flowgraph, we
perform a backward data-flow analysis"

Previously, we were walking the blocks forwards and updating the livein and
then the liveout.  However, the livein calculation depends on the liveout
and the liveout depends on the successor blocks.  The net result is that it
takes one full iteration to go from liveout to livein and then another
full iteration to propagate to the predecessors.  This works out to an
O(n^2) computation where n is the number of blocks.  If we run things in
the other order, it's O(nl) where l is the maximum loop depth which is
practically bounded by 3.

In b2c6ba0c4b21391dc35018e1c8c4f7f7d8952bea, we made this same change in
the FS backend to great effect.  Might as well keep it consistent and make
the same change for vec4.  Also, this took the time to run the test:

ES31-CTS.arrays_of_arrays.InteractionFunctionCalls1

from 6:49.62 to 3:31.40 on Timothy Arceri's machine.

Reviewed-by: Matt Turner <mattst88@gmail.com>
9 years agoi965/skl: Use more compact hiz dimensions
Ben Widawsky [Thu, 25 Jun 2015 03:07:54 +0000 (20:07 -0700)]
i965/skl: Use more compact hiz dimensions

gen8 had some special restrictions which don't seem to carry over to gen9.
Quoting the spec for SKL:
"The Z_Height and Z_Width values must equal those present in
3DSTATE_DEPTH_BUFFER incremented by one."

This fixes nothing in piglit (and regresses nothing).

Signed-off-by: Ben Widawsky <ben@bwidawsk.net>
Reviewed-by: Anuj Phogat <anuj.phogat@gmail.com>
Reviewed-by: Jordan Justen <jordan.l.justen@intel.com>
9 years agoradeonsi: don't fail in si_shader_io_get_unique_index
Marek Olšák [Thu, 25 Jun 2015 12:58:37 +0000 (14:58 +0200)]
radeonsi: don't fail in si_shader_io_get_unique_index

Trivial. Picked from my tessellation branch.

9 years agoi965: Drop brw->depthstencil.stencil_offset from gen8_depth_state.c.
Kenneth Graunke [Wed, 24 Jun 2015 07:04:11 +0000 (00:04 -0700)]
i965: Drop brw->depthstencil.stencil_offset from gen8_depth_state.c.

This is always 0 - only brw_workaround_depthstencil_alignment ever sets
it, and that doesn't run on Gen6+.  My initial Broadwell depth state
commit had this mistake.

Signed-off-by: Kenneth Graunke <kenneth@whitecape.org>
Reviewed-by: Anuj Phogat <anuj.phogat@gmail.com>
9 years agonir: Recognize max(min(a, 1.0), 0.0) as fsat(a).
Kenneth Graunke [Thu, 11 Jun 2015 08:59:44 +0000 (01:59 -0700)]
nir: Recognize max(min(a, 1.0), 0.0) as fsat(a).

We already recognize min(max(a, 0.0), 1.0) as a saturate, but neglected
this variant (which is also handled by the GLSL IR pass).

shader-db results on Broadwell:
total instructions in shared programs: 7363046 -> 7362788 (-0.00%)
instructions in affected programs:     11928 -> 11670 (-2.16%)
helped:                                64
HURT:                                  0

Signed-off-by: Kenneth Graunke <kenneth@whitecape.org>
Reviewed-by: Iago Toral Quiroga <itoral@igalia.com>
9 years agosoftpipe,llvmpipe: fix PIPE_SHADER_CAP_MAX_INPUTS value
Marek Olšák [Wed, 24 Jun 2015 22:56:32 +0000 (00:56 +0200)]
softpipe,llvmpipe: fix PIPE_SHADER_CAP_MAX_INPUTS value

PIPE_MAX_SHADER_INPUTS was recently bumped to 80 because of tessellation.

Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=91099
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=91101

Reviewed-by: Brian Paul <brianp@vmware.com>
Reviewed-by: Roland Scheidegger <sroland@vmware.com>
9 years agoi965/bxt: Add basic Broxton infrastructure
Ben Widawsky [Wed, 17 Jun 2015 22:50:11 +0000 (15:50 -0700)]
i965/bxt: Add basic Broxton infrastructure

The thread counts and URB information are all speculative numbers that were
based on some CHV numbers at the time.

v2:
Originally this patch had PCI IDs. I've moved that to a new patch at the end of
the series.
Remove is_cherryview hack.
Add PCI ids. These match the ones defined in the kernel. The only one tested by
us is 0x0a84.
Capitalize the hex string (Mark)

Signed-off-by: Ben Widawsky <ben@bwidawsk.net>
Tested-by: "Lecluse, Philippe" <Philippe.Lecluse@intel.com>
Reviewed-by: Mark Janes <mark.a.janes@intel.com>
9 years agoradeon: Advertise correct GL_QUERY_COUNTER_BITS/GL_SAMPLES_PASSED value
Ian Romanick [Mon, 22 Jun 2015 18:09:27 +0000 (11:09 -0700)]
radeon: Advertise correct GL_QUERY_COUNTER_BITS/GL_SAMPLES_PASSED value

Commit b765119c changed the default value of all the counter bits to
64.  However, older hardware only has 32 counter bits.

This has only been build-tested.  We don't have any tests that verify
the advertised value against implementation behavior, so I don't know
what additional testing could be done.

NOTE: It appears that many Gallium drivers (at least r300 and i915g)
have the same problem, but I don't see a way for the state-tracker to
determine the counter size.  Marek says, "For Gallium, a new PIPE_CAP or
new get_xxx_param function will be needed."

Signed-off-by: Ian Romanick <ian.d.romanick@intel.com>
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Cc: Alex Deucher <alexander.deucher@amd.com>
9 years agoi965/fs_live_variables: Do liveness analysis bottom-to-top
Jason Ekstrand [Mon, 8 Jun 2015 23:03:19 +0000 (16:03 -0700)]
i965/fs_live_variables: Do liveness analysis bottom-to-top

From Muchnick's Advanced Compiler Design and Implementation:

"To determine which variables are live at each point in a flowgraph, we
perform a backward data-flow analysis"

Previously, we were walking the blocks forwards and updating the livein and
then the liveout.  However, the livein calculation depends on the liveout
and the liveout depends on the successor blocks.  The net result is that it
takes one full iteration to go from liveout to livein and then another
full iteration to propagate to the predecessors.  This works out to an
O(n^2) computation where n is the number of blocks.  If we run things in
the other order, it's O(nl) where l is the maximum loop depth which is
practically bounded by 3.

On my HSW desktop, one particular shadertoy test gets a 20% improvement in
compile times:

N           Min           Max        Median           Avg        Stddev
x  10        15.965        16.884        16.026       16.1822    0.34736846
+  10        12.813        13.052        12.876       12.8891    0.06913666
Difference at 95.0% confidence
        -3.2931 +/- 0.235316
        -20.3501% +/- 1.45417%
        (Student's t, pooled s = 0.250444)

Reviewed-by: Matt Turner <mattst88@gmail.com>
9 years agoi965: Delete linked GLSL IR when using NIR.
Tapani Pälli [Thu, 11 Jun 2015 07:41:53 +0000 (10:41 +0300)]
i965: Delete linked GLSL IR when using NIR.

This is based on Kenneth's patch to delete 'most of the IR'. Due to
linker changes to clone variables, we can now free all of IR.

Saves 58MB of memory when replaying a Dota 2 trace on Broadwell.

Signed-off-by: Tapani Pälli <tapani.palli@intel.com>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Cc: mesa-stable@lists.freedesktop.org
9 years agoglsl: clone inputs and outputs during linking
Tapani Pälli [Thu, 11 Jun 2015 07:41:52 +0000 (10:41 +0300)]
glsl: clone inputs and outputs during linking

This increases memory pressure during linking but makes it easier
for backend to free IR after it is not needed anymore.

v2: use resource list as ralloc context in case of relink (Kenneth)

Signed-off-by: Tapani Pälli <tapani.palli@intel.com>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Cc: mesa-stable@lists.freedesktop.org
9 years agoi965: Rename intel_emit* to reflect their new location in brw_pipe_control
Chris Wilson [Thu, 30 Apr 2015 16:04:51 +0000 (17:04 +0100)]
i965: Rename intel_emit* to reflect their new location in brw_pipe_control

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
9 years agoi965: Transplant PIPE_CONTROL routines to brw_pipe_control
Chris Wilson [Thu, 30 Apr 2015 15:53:12 +0000 (16:53 +0100)]
i965: Transplant PIPE_CONTROL routines to brw_pipe_control

Start trimming the fat from intel_batchbuffer.c. First by moving the set
of routines for emitting PIPE_CONTROLS (along with the lore concerning
hardware workarounds) to a separate brw_pipe_control.c

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
9 years agonir: Use a switch statement for detecting move-like operations.
Kenneth Graunke [Wed, 24 Jun 2015 06:17:53 +0000 (23:17 -0700)]
nir: Use a switch statement for detecting move-like operations.

Suggested by Jason Ekstrand.

Signed-off-by: Kenneth Graunke <kenneth@whitecape.org>
Reviewed-by: Samuel Iglesias Gonsálvez <siglesias@igalia.com>
9 years agosvga: silence warnings about unexpected shader type
Brian Paul [Wed, 24 Jun 2015 16:41:52 +0000 (10:41 -0600)]
svga: silence warnings about unexpected shader type

Trivial.

9 years agost/mesa: remove unneeded pipe_surface_release() in st_render_texture()
Brian Paul [Mon, 22 Jun 2015 20:04:09 +0000 (14:04 -0600)]
st/mesa: remove unneeded pipe_surface_release() in st_render_texture()

This caused us to always free the pipe_surface for the renderbuffer.
The subsequent call to st_update_renderbuffer_surface() would typically
just recreate it.  Remove the call to pipe_surface_release() and let
st_update_renderbuffer_surface() take care of freeing the old surface
if it needs to be replaced (because of change to mipmap level, etc).

This can save quite a few calls to pipe_context::create_surface() and
surface_destroy().

Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Reviewed-by: Jose Fonseca <jfonseca@vmware.com>
9 years agost/wgl: add stw_nopfuncs.h to the sources lists
Emil Velikov [Wed, 24 Jun 2015 11:59:55 +0000 (12:59 +0100)]
st/wgl: add stw_nopfuncs.h to the sources lists

Signed-off-by: Emil Velikov <emil.l.velikov@gmail.com>
9 years agoloader: move loader_open_device out of HAVE_LIBUDEV block
Julien Isorce [Tue, 23 Jun 2015 21:47:05 +0000 (22:47 +0100)]
loader: move loader_open_device out of HAVE_LIBUDEV block

Fixes the following build issue, when building without libudev.

CCLD   libGL.la
./.libs/libglx.a(dri2_glx.o): In function `dri2CreateScreen':
src/glx/dri2_glx.c:1186: undefined reference to `loader_open_device'
collect2: ld returned 1 exit status

CCLD     libEGL.la
Undefined symbols for architecture x86_64:
"_loader_open_device", referenced from:
  _dri2_initialize_x11_dri2 in libegl_dri2.a(platform_x11.o)

Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=91077
Signed-off-by: Julien Isorce <j.isorce@samsung.com>
Reviewed-by: Emil Velikov <emil.l.velikov@gmail.com>
9 years agowinsys/radeon: reduce BO cache timeout
Grigori Goronzy [Wed, 24 Jun 2015 01:40:38 +0000 (03:40 +0200)]
winsys/radeon: reduce BO cache timeout

1000 ms is an extreme value for typical interactive loads. A large
cache has some disadvantages. Search for reusable BOs can take a long
time and memory might get exhausted.

Let's be rather conservative and use half of the old value,
500ms. This is beneficial to some loads on my test system and there
are no regressions.

Reviewed-by: Marek Olšák <marek.olsak@amd.com>
9 years agowinsys/radeon: align BO size to page size
Grigori Goronzy [Wed, 24 Jun 2015 01:38:02 +0000 (03:38 +0200)]
winsys/radeon: align BO size to page size

This is the basic granularity for BO allocations. The alignment also
helps with BO reuse by the cached bufmgr.

This results in a huge 45% speedup in Metro 2033 Redux on my test
system. The game relies on buffer orphaning with very small buffers
(hundreds of bytes in size) and that did not work efficiently
before. This change may also affect other applications and games.

Reviewed-by: Marek Olšák <marek.olsak@amd.com>
9 years agoglsl: remove cross validation of interpolation qualifier with GLSL 4.40
Tapani Pälli [Tue, 9 Jun 2015 08:06:56 +0000 (11:06 +0300)]
glsl: remove cross validation of interpolation qualifier with GLSL 4.40

Signed-off-by: Tapani Pälli <tapani.palli@intel.com>
Reviewed-by: Timothy Arceri <t_arceri@yahoo.com.au>
9 years agoi965: Fix whitespace error in gen8_depth_state.c
Kenneth Graunke [Wed, 24 Jun 2015 06:15:22 +0000 (23:15 -0700)]
i965: Fix whitespace error in gen8_depth_state.c

Trivial.

9 years agoi965: Don't count NIR instructions for shader-db.
Kenneth Graunke [Mon, 22 Jun 2015 21:20:20 +0000 (14:20 -0700)]
i965: Don't count NIR instructions for shader-db.

Matt, Jason, and I haven't found this useful in a long time.

Signed-off-by: Kenneth Graunke <kenneth@whitecape.org>
Reviewed-by: Matt Turner <mattst88@gmail.com>
9 years agowinsys/radeon: Unmap GPU VM address range when destroying BO
Michel Dänzer [Thu, 21 May 2015 01:49:05 +0000 (10:49 +0900)]
winsys/radeon: Unmap GPU VM address range when destroying BO

But only when doing so is safe according to the
RADEON_INFO_VA_UNMAP_WORKING kernel query.

This avoids kernel GPU VM address range conflicts when the BO has other
references than the GEM handle being closed, e.g. when the BO is shared.

Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=90537
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=90873

Cc: "10.5 10.6" <mesa-stable@lists.freedesktop.org>
Reviewed-by: Christian König <christian.koenig@amd.com>
9 years agovc4: Also dump VC4_PACKET_LOAD_TILE_BUFFER_GENERAL.
Eric Anholt [Tue, 23 Jun 2015 00:38:14 +0000 (17:38 -0700)]
vc4: Also dump VC4_PACKET_LOAD_TILE_BUFFER_GENERAL.

9 years agovc4: Add dumping for VC4_PACKET_LOAD/STORE_FULL_RES_TILE_BUFFER.
Eric Anholt [Tue, 23 Jun 2015 00:34:24 +0000 (17:34 -0700)]
vc4: Add dumping for VC4_PACKET_LOAD/STORE_FULL_RES_TILE_BUFFER.

9 years agovc4: Don't try to CSE color reads.
Eric Anholt [Wed, 24 Jun 2015 01:08:49 +0000 (18:08 -0700)]
vc4: Don't try to CSE color reads.

It returns a new value for each sample in the TLB.  We've already avoided
trying to get the same index's color multiple times at the vc4_program.c
level, so we're not losing anything by doing this.

9 years agovc4: Make a helper for TLB color writes, too.
Eric Anholt [Wed, 24 Jun 2015 01:04:00 +0000 (18:04 -0700)]
vc4: Make a helper for TLB color writes, too.

We've done so for all the other QIR instruction generation in this file.

9 years agovc4: Pull the blending operation out to a separate function.
Eric Anholt [Wed, 24 Jun 2015 00:53:07 +0000 (17:53 -0700)]
vc4: Pull the blending operation out to a separate function.

It's fairly separate from the rest of the TLB operations at frag end time,
and we'll need to run it multiple times to support MSAA blending.

9 years agovc4: Clarify size calculation for Z/S writes.
Eric Anholt [Mon, 22 Jun 2015 18:45:27 +0000 (11:45 -0700)]
vc4: Clarify size calculation for Z/S writes.

It's the same value for loads and stores, because they're basically the
same packet.

9 years agovc4: Add an "args" temporary for RCL setup.
Eric Anholt [Mon, 22 Jun 2015 20:14:57 +0000 (13:14 -0700)]
vc4: Add an "args" temporary for RCL setup.

9 years agovc4: Reuse (and extend) the packet.h sizes for dumping.
Eric Anholt [Tue, 23 Jun 2015 00:31:24 +0000 (17:31 -0700)]
vc4: Reuse (and extend) the packet.h sizes for dumping.

9 years agovc4: Fix printfs for blit fallbacks.
Eric Anholt [Sat, 20 Jun 2015 22:30:19 +0000 (15:30 -0700)]
vc4: Fix printfs for blit fallbacks.

9 years agotgsi_to_nir: Fix translation of TXF on MSAA targets.
Eric Anholt [Tue, 23 Jun 2015 18:02:12 +0000 (11:02 -0700)]
tgsi_to_nir: Fix translation of TXF on MSAA targets.

Noticed while trying to add GL_ARB_texture_multisample support to vc4.

Reviewed-by: Ilia Mirkin <imirkin@alum.mit.edu>
9 years agoi965/fs: Get rid of an unused variable in emit_barrier()
Jason Ekstrand [Tue, 23 Jun 2015 22:39:42 +0000 (15:39 -0700)]
i965/fs: Get rid of an unused variable in emit_barrier()

Reviewed-by: Jordan Justen <jordan.l.justen@intel.com>
9 years agoi965: Remove the brw_context from the visitors
Jason Ekstrand [Tue, 23 Jun 2015 00:17:56 +0000 (17:17 -0700)]
i965: Remove the brw_context from the visitors

As of this commit, nothing actually needs the brw_context.

Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Reviewed-by: Chris Forbes <chrisf@ijw.co.nz>
9 years agoi965/vec4_vs: Add an explicit use_legacy_snorm_formula flag
Jason Ekstrand [Tue, 23 Jun 2015 00:30:23 +0000 (17:30 -0700)]
i965/vec4_vs: Add an explicit use_legacy_snorm_formula flag

This way we can stop doing is_gles3 checks inside of the compiler.

Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Reviewed-by: Chris Forbes <chrisf@ijw.co.nz>
9 years agoi965/vec4: Turn some _mesa_problem calls into asserts
Jason Ekstrand [Tue, 23 Jun 2015 00:17:51 +0000 (17:17 -0700)]
i965/vec4: Turn some _mesa_problem calls into asserts

Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Reviewed-by: Chris Forbes <chrisf@ijw.co.nz>
9 years agoi965/vs: Pass the current set of clip planes through run() and run_vs()
Jason Ekstrand [Sat, 20 Jun 2015 00:29:42 +0000 (17:29 -0700)]
i965/vs: Pass the current set of clip planes through run() and run_vs()

Previously, these were pulled out of the GL context conditionally based on
whether we were running ff/ARB or a GLSL program.  Now, we just pass them
in so that the visitor doesn't have to grab them itself.

Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Reviewed-by: Chris Forbes <chrisf@ijw.co.nz>
9 years agoi965/fs: Add a do_rep_send flag to run_fs
Jason Ekstrand [Sat, 20 Jun 2015 00:25:28 +0000 (17:25 -0700)]
i965/fs: Add a do_rep_send flag to run_fs

Previously, we were pulling it from brw->do_rep_send

Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Reviewed-by: Chris Forbes <chrisf@ijw.co.nz>
9 years agoi965: Pull calls to get_shader_time_index out of the visitor
Jason Ekstrand [Fri, 19 Jun 2015 22:40:09 +0000 (15:40 -0700)]
i965: Pull calls to get_shader_time_index out of the visitor

Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Reviewed-by: Chris Forbes <chrisf@ijw.co.nz>
9 years agoi965: Use a single index per shader for shader_time.
Jason Ekstrand [Fri, 19 Jun 2015 21:46:03 +0000 (14:46 -0700)]
i965: Use a single index per shader for shader_time.

Previously, each shader took 3 shader time indices which were potentially
at arbirary points in the shader time buffer.  Now, each shader gets a
single index which refers to 3 consecutive locations in the buffer.  This
simplifies some of the logic at the cost of having a magic 3 a few places.

Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Reviewed-by: Chris Forbes <chrisf@ijw.co.nz>
9 years agoi965: Add compiler options to brw_compiler
Jason Ekstrand [Thu, 16 Apr 2015 22:28:17 +0000 (15:28 -0700)]
i965: Add compiler options to brw_compiler

This creates the options at screen cration time and then we just copy them
into the context at context creation time.  We also move is_scalar to the
brw_compiler structure.

We also end up manually setting some values that the core would have set by
default for us.  Fortunately, there are only two non-zero shader compiler
option defaults that we aren't overriding anyway so this isn't a big deal.

Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Reviewed-by: Chris Forbes <chrisf@ijw.co.nz>
9 years agoi965/fs: Plumb compiler debug logging through brw_compiler
Jason Ekstrand [Tue, 23 Jun 2015 00:01:22 +0000 (17:01 -0700)]
i965/fs: Plumb compiler debug logging through brw_compiler

Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Reviewed-by: Chris Forbes <chrisf@ijw.co.nz>
9 years agoi965/fs: Do the no16 perf logging directly in fs_visitor::no16()
Jason Ekstrand [Mon, 22 Jun 2015 23:32:06 +0000 (16:32 -0700)]
i965/fs: Do the no16 perf logging directly in fs_visitor::no16()

While we're at it, we'll drop the note about 10-20% performance loss.

Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Reviewed-by: Chris Forbes <chrisf@ijw.co.nz>
9 years agoi965/fs: Make no16 non-variadic
Jason Ekstrand [Mon, 22 Jun 2015 23:30:04 +0000 (16:30 -0700)]
i965/fs: Make no16 non-variadic

We never used the fact that it was variadic anyway.

Reviewed-by: Chris Forbes <chrisf@ijw.co.nz>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
9 years agoi965: Move INTEL_DEBUG variable parsing to screen creation time
Jason Ekstrand [Fri, 17 Apr 2015 00:39:13 +0000 (17:39 -0700)]
i965: Move INTEL_DEBUG variable parsing to screen creation time

v2: Do bufmgr set_debug and set_aub_dump at screen time as well.

Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
9 years agoi965: Remove the dependance on brw_context from the generators
Jason Ekstrand [Thu, 16 Apr 2015 21:34:04 +0000 (14:34 -0700)]
i965: Remove the dependance on brw_context from the generators

Reviewed-by: Chris Forbes <chrisf@ijw.co.nz>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
9 years agoi965: Plumb compiler debug logging through a function pointer in brw_compiler
Jason Ekstrand [Thu, 16 Apr 2015 21:13:52 +0000 (14:13 -0700)]
i965: Plumb compiler debug logging through a function pointer in brw_compiler

v2 (Ken): Make shader_debug_log a printf-like function.
v3 (Jason): Add a void * to pass the brw_context through

Reviewed-by: Chris Forbes <chrisf@ijw.co.nz>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
9 years agomesa: Add a va_args variant of _mesa_gl_debug().
Kenneth Graunke [Sat, 18 Apr 2015 19:23:33 +0000 (12:23 -0700)]
mesa: Add a va_args variant of _mesa_gl_debug().

This will be useful for wrapper functions.

Signed-off-by: Kenneth Graunke <kenneth@whitecape.org>
Reviewed-by: Jason Ekstrand <jason.ekstrand@intel.com>
Reviewed-by: Chris Forbes <chrisf@ijw.co.nz>
9 years agoi965: Replace some instances of brw->gen with devinfo->gen
Jason Ekstrand [Sat, 20 Jun 2015 00:19:38 +0000 (17:19 -0700)]
i965: Replace some instances of brw->gen with devinfo->gen

9 years agoi965: Initialize backend_shader::mem_ctx in its constructor.
Matt Turner [Mon, 22 Jun 2015 18:42:15 +0000 (11:42 -0700)]
i965: Initialize backend_shader::mem_ctx in its constructor.

We were initializing it in each subclasses' constructors for some
reason.

Reviewed-by: Jordan Justen <jordan.l.justen@intel.com>
9 years agoi965: Assert that the GL primitive isn't out of range.
Matt Turner [Mon, 22 Jun 2015 18:20:32 +0000 (11:20 -0700)]
i965: Assert that the GL primitive isn't out of range.

Coverity sees the if (mode >= BRW_PRIM_OFFSET (128)) test and assumes
that the else-branch might execute for mode to up 127, which out be out
of bounds.

Reviewed-by: Jordan Justen <jordan.l.justen@intel.com>
9 years agoi965/cfg: Assert that cur_do/while/if pointers are non-NULL.
Matt Turner [Mon, 22 Jun 2015 18:09:49 +0000 (11:09 -0700)]
i965/cfg: Assert that cur_do/while/if pointers are non-NULL.

Coverity sees that the functions immediately below the new assertions
dereference these pointers, but is unaware that an ENDIF always follows
an IF, etc.

Reviewed-by: Jordan Justen <jordan.l.justen@intel.com>
9 years agomesa: Delete unused ICEIL().
Matt Turner [Sat, 20 Jun 2015 03:40:15 +0000 (20:40 -0700)]
mesa: Delete unused ICEIL().

Can't find any uses of it in git history.

Reviewed-by: Jordan Justen <jordan.l.justen@intel.com>
9 years agoi965/fs: Don't mess up stride for uniform integer multiplication.
Matt Turner [Mon, 22 Jun 2015 17:59:33 +0000 (10:59 -0700)]
i965/fs: Don't mess up stride for uniform integer multiplication.

If the stride is 0, the source is a uniform and we should not modify the
stride.

Cc: "10.6" <mesa-stable@lists.freedesktop.org>
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=91047
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
9 years agoegl/x11: Remove duplicate call to dri2_x11_add_configs_for_visuals
Boyan Ding [Sat, 13 Jun 2015 07:33:20 +0000 (15:33 +0800)]
egl/x11: Remove duplicate call to dri2_x11_add_configs_for_visuals

The call to dri2_x11_add_configs_for_visuals (previously
dri2_add_configs_for_visuals) was moved downwards in commit f8c5b8a1,
but appeared again in its original position after its rename in
d019cd81. Remove it.

Cc: "10.5 10.6" <mesa-stable@lists.freedesktop.org>
Signed-off-by: Boyan Ding <boyan.j.ding@gmail.com>
Reviewed-by: Samuel Iglesias Gonsálvez <siglesias@igalia.com>
Reviewed-by: Emil Velikov <emil.l.velikov@gmail.com>
Reviewed-by: Chad Versace <chad.versace@intel.com>
9 years agoi965/gen9: Don't use encrypted MOCS
Ben Widawsky [Wed, 17 Jun 2015 22:50:13 +0000 (15:50 -0700)]
i965/gen9: Don't use encrypted MOCS

On gen9+ MOCS is an index into a table. It is 7 bits, and AFAICT, bit 0 is for
doing encrypted reads.

I don't recall how I decided to do this for BXT. I don't know this patch was
ever needed, since it seems nothing is broken today on SKL. Furthermore, this
patch may no longer be needed because of the ongoing changes with MOCS setup. It
is what is being used/tested, so it's included in the series.

The chosen values are the old values left shifted. That was also an arbitrary
choice.

v2: Use shift in MOCS to make it clear what we're doing. (Ken)

Signed-off-by: Ben Widawsky <ben@bwidawsk.net>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
9 years agonv50,nvc0: make sure to pushbuf_refn before putting bo into pushbuf_data
Ilia Mirkin [Sun, 21 Jun 2015 23:03:35 +0000 (19:03 -0400)]
nv50,nvc0: make sure to pushbuf_refn before putting bo into pushbuf_data

Without first running the bo through pushbuf_refn, the nouveau drm
library will have uninitialized structures regarding this bo, and will
insert incorrect data.

This fixes supertuxkart 0.9 crash on start (where it ends up doing a lot
of indirect draws).

Signed-off-by: Ilia Mirkin <imirkin@alum.mit.edu>
Cc: "10.5 10.6" <mesa-stable@lists.freedesktop.org>
9 years agonvc0: always put all tfb bufs into bufctx
Ilia Mirkin [Sun, 21 Jun 2015 19:00:16 +0000 (15:00 -0400)]
nvc0: always put all tfb bufs into bufctx

Since we clear the TFB bufctx binding point above, we need to put all of
the active tfb's back in, even if they haven't changed since last time.
Otherwise the tfb may get moved into sysmem and the underlying mapping
will generate write errors.

Signed-off-by: Ilia Mirkin <imirkin@alum.mit.edu>
Cc: "10.5 10.6" <mesa-stable@lists.freedesktop.org>
9 years agoglsl: binding point is a texture unit, which is a combined space
Ilia Mirkin [Tue, 23 Jun 2015 04:16:59 +0000 (00:16 -0400)]
glsl: binding point is a texture unit, which is a combined space

This fixes compilation failures in Dota 2 Reborn where a texture unit
binding point was used that was numerically higher than the max
per stage.

Signed-off-by: Ilia Mirkin <imirkin@alum.mit.edu>
Reviewed-by: Chris Forbes <chrisf@ijw.co.nz>
Reviewed-by: Timothy Arceri <t_arceri@yahoo.com.au>
Tested-by: Nick Sarnie <commendsarnex@gmail.com>
Cc: "10.5 10.6" <mesa-stable@lists.freedesktop.org>
9 years agoandroid: egl: do not link against libglapi
Emil Velikov [Fri, 19 Jun 2015 18:35:19 +0000 (19:35 +0100)]
android: egl: do not link against libglapi

The only reason we touch glapi is to dlopen it in order to:
 - make sure that the unresolved _glapi* symbols in the dri modules are
provided.
 - fetch glFlush() and use it at various stages in the dri2 driver.

Cc: Chih-Wei Huang <cwhuang@linux.org.tw>
Cc: Eric Anholt <eric@anholt.net>
Signed-off-by: Emil Velikov <emil.l.velikov@gmail.com>
9 years agogbm: do not (over)link against libglapi.so
Emil Velikov [Fri, 19 Jun 2015 18:22:38 +0000 (19:22 +0100)]
gbm: do not (over)link against libglapi.so

The whole of GBM does not rely on even a single symbol from the GL
dispatch library, unsuprisingly. The only need for it comes from the
unresolved symbols in the DRI modules, which are now correctly handled
with Frank's commit.

Cc: "10.5 10.6" <mesa-stable@lists.freedesktop.org>
Signed-off-by: Emil Velikov <emil.l.velikov@gmail.com>
9 years agogbm: dlopen libglapi so gbm_create_device works
Frank Henigman [Thu, 6 Nov 2014 21:29:26 +0000 (16:29 -0500)]
gbm: dlopen libglapi so gbm_create_device works

Dri driver libs are not linked to pull in libglapi so gbm_create_device()
fails when it tries to dlopen them (unless the application is linked
with something that does pull in libglapi, like libGL).
Until dri drivers can be fixed properly, dlopen libglapi before trying
to dlopen them.

Cc: "10.5 10.6" <mesa-stable@lists.freedesktop.org>
Signed-off-by: Frank Henigman <fjhenigman@google.com>
[Emil Velikov: Drop misleading bugzilla link, mention that libname differs]
Reviewed-by: Emil Velikov <emil.l.velikov@gmail.com>
9 years agoconfigure: drop unused variable GBM_BACKEND_DIRS
Emil Velikov [Fri, 19 Jun 2015 16:48:30 +0000 (17:48 +0100)]
configure: drop unused variable GBM_BACKEND_DIRS

Signed-off-by: Emil Velikov <emil.l.velikov@gmail.com>
Reviewed-by: Eric Anholt <eric@anholt.net>
9 years agoconfigure: error out when building libEGL without shared-glapi
Emil Velikov [Fri, 19 Jun 2015 16:46:41 +0000 (17:46 +0100)]
configure: error out when building libEGL without shared-glapi

The latter is a hard requirement and without it we'll error out later
on in the build.

Cc: "10.5 10.6" <mesa-stable@lists.freedesktop.org>
Signed-off-by: Emil Velikov <emil.l.velikov@gmail.com>
Reviewed-by: Eric Anholt <eric@anholt.net>
9 years agoconfigure: error out when building backend-less libEGL
Emil Velikov [Fri, 19 Jun 2015 16:44:02 +0000 (17:44 +0100)]
configure: error out when building backend-less libEGL

Cc: "10.5 10.6" <mesa-stable@lists.freedesktop.org>
Signed-off-by: Emil Velikov <emil.l.velikov@gmail.com>
Reviewed-by: Eric Anholt <eric@anholt.net>
9 years agodrivers/x11: drop unneeded HAVE_X11_DRIVER check
Emil Velikov [Fri, 19 Jun 2015 16:37:18 +0000 (17:37 +0100)]
drivers/x11: drop unneeded HAVE_X11_DRIVER check

Already handled in the Makefile which includes the drivers/x11 subdir.

Signed-off-by: Emil Velikov <emil.l.velikov@gmail.com>
Reviewed-by: Eric Anholt <eric@anholt.net>
9 years agoconfigure: allow building shared-glapi powered libgl-xlib
Emil Velikov [Fri, 19 Jun 2015 18:43:23 +0000 (19:43 +0100)]
configure: allow building shared-glapi powered libgl-xlib

Cc: Brian Paul <brianp@vmware.com>
Cc: Adam Jackson <ajax@redhat.com>
Signed-off-by: Emil Velikov <emil.l.velikov@gmail.com>
Reviewed-by: Eric Anholt <eric@anholt.net>
Acked-by: Jose Fonseca <jfonseca@vmware.com>
9 years agotargets/libgl-xlib: fix the build against shared_glapi
Emil Velikov [Fri, 19 Jun 2015 16:30:13 +0000 (17:30 +0100)]
targets/libgl-xlib: fix the build against shared_glapi

Cc: Brian Paul <brianp@vmware.com>
Signed-off-by: Emil Velikov <emil.l.velikov@gmail.com>
Reviewed-by: Eric Anholt <eric@anholt.net>
Acked-by: Jose Fonseca <jfonseca@vmware.com>
9 years agodrivers/x11: fix the build against shared_glapi
Emil Velikov [Fri, 19 Jun 2015 16:28:25 +0000 (17:28 +0100)]
drivers/x11: fix the build against shared_glapi

Cc: Brian Paul <brianp@vmware.com>
Signed-off-by: Emil Velikov <emil.l.velikov@gmail.com>
Reviewed-by: Eric Anholt <eric@anholt.net>
Acked-by: Jose Fonseca <jfonseca@vmware.com>
9 years agoconfigure: warn about shared_glapi & xlib-glx only when both are set
Emil Velikov [Fri, 19 Jun 2015 16:19:46 +0000 (17:19 +0100)]
configure: warn about shared_glapi & xlib-glx only when both are set

Printing out the message when shared_glapi is disabled only leads to
confusion.

Cc: "10.5 10.6" <mesa-stable@lists.freedesktop.org>
Signed-off-by: Emil Velikov <emil.l.velikov@gmail.com>
Reviewed-by: Eric Anholt <eric@anholt.net>
9 years agoglapi: remap_helper.py: remove unused argument 'es'
Emil Velikov [Thu, 18 Jun 2015 22:30:34 +0000 (23:30 +0100)]
glapi: remap_helper.py: remove unused argument 'es'

Identical to the previous commit - unused by neither the Autotools,
Android or SCons build.

Signed-off-by: Emil Velikov <emil.l.velikov@gmail.com>
Reviewed-by: Dylan Baker <baker.dylan.c@gmail.com>
Reviewed-by: Jose Fonseca <jfonseca@vmware.com>
9 years agoglapi: gl_table.py: remove unused variable 'es'
Emil Velikov [Thu, 18 Jun 2015 22:28:05 +0000 (23:28 +0100)]
glapi: gl_table.py: remove unused variable 'es'

None of the three build systems ever set it, as such we can clear things
up a bit.

Signed-off-by: Emil Velikov <emil.l.velikov@gmail.com>
Reviewed-by: Dylan Baker <baker.dylan.c@gmail.com>
Reviewed-by: Jose Fonseca <jfonseca@vmware.com>
9 years agoegl: Use the loader_open_device() helper to do open with CLOEXEC
Derek Foreman [Wed, 17 Jun 2015 16:28:51 +0000 (11:28 -0500)]
egl: Use the loader_open_device() helper to do open with CLOEXEC

We've moved the open with CLOEXEC idiom into a helper function, so
call it instead of duplicating the code.

This also replaces a couple of opens that didn't properly do CLOEXEC.

Signed-off-by: Derek Foreman <derekf@osg.samsung.com>
Reviewed-by: Kristian Høgsberg <krh@bitplanet.net>
Reviewed-by: Emil Velikov <emil.l.velikov@gmail.com>
9 years agoglx: Use loader_open_device() helper
Derek Foreman [Wed, 17 Jun 2015 16:28:50 +0000 (11:28 -0500)]
glx: Use loader_open_device() helper

We've moved the open with CLOEXEC idiom into a helper function, so
call it instead of duplicating the code here.

Signed-off-by: Derek Foreman <derekf@osg.samsung.com>
Reviewed-by: Kristian Høgsberg <krh@bitplanet.net>
Reviewed-by: Emil Velikov <emil.l.velikov@gmail.com>
9 years agoloader: Rename drm_open_device() to loader_open_device() and share it
Derek Foreman [Wed, 17 Jun 2015 16:28:49 +0000 (11:28 -0500)]
loader: Rename drm_open_device() to loader_open_device() and share it

This is already our common idiom for opening files with CLOEXEC and
it's a little ugly, so let's share this one implementation.

Signed-off-by: Derek Foreman <derekf@osg.samsung.com>
Reviewed-by: Kristian Høgsberg <krh@bitplanet.net>
Reviewed-by: Emil Velikov <emil.l.velikov@gmail.com>
9 years agoegl/drm: Duplicate fd with F_DUPFD_CLOEXEC to prevent leak
Derek Foreman [Wed, 17 Jun 2015 16:28:48 +0000 (11:28 -0500)]
egl/drm: Duplicate fd with F_DUPFD_CLOEXEC to prevent leak

Replacing dup() with fcntl F_DUPFD_CLOEXEC creates the duplicate
file descriptor with CLOEXEC so it won't be leaked to child
processes if the process fork()s later.

Signed-off-by: Derek Foreman <derekf@osg.samsung.com>
Reviewed-by: Kristian Høgsberg <krh@bitplanet.net>
Reviewed-by: Emil Velikov <emil.l.velikov@gmail.com>
9 years agodraw,tgsi: Assume TGSI_PROPERTY_GS_INVOCATIONS default of 1.
Jose Fonseca [Tue, 23 Jun 2015 11:18:06 +0000 (12:18 +0100)]
draw,tgsi: Assume TGSI_PROPERTY_GS_INVOCATIONS default of 1.

If the shader doesn't specify number of invocations, assume one.

This fixes geometry shaders on state trackers other than Mesa (and
probably graw tests too.)

Trivial.

9 years agoglsl: Specify the shader stage in linker errors due to too many in/outputs.
Jose Fonseca [Fri, 19 Jun 2015 12:53:46 +0000 (13:53 +0100)]
glsl: Specify the shader stage in linker errors due to too many in/outputs.

Reviewed-by: Ian Romanick <ian.d.romanick@intel.com>
Reviewed-by: Brian Paul <brianp@vmware.com>
9 years agodocs: update GL3 with softpipe/llvmpipe gpu_shader5 pieces.
Dave Airlie [Tue, 23 Jun 2015 05:55:30 +0000 (15:55 +1000)]
docs: update GL3 with softpipe/llvmpipe gpu_shader5 pieces.

This just updates the bits I've added in the previous few patches.

Signed-off-by: Dave Airlie <airlied@redhat.com>
9 years agodraw/gallivm: add invocation ID support for llvmpipe.
Dave Airlie [Mon, 22 Jun 2015 03:59:25 +0000 (13:59 +1000)]
draw/gallivm: add invocation ID support for llvmpipe.

This extends the draw code to add support for invocations.

Reviewed-by: Roland Scheidegger <sroland@vmware.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>
9 years agodraw/tgsi: implement geom shader invocation support.
Dave Airlie [Wed, 20 May 2015 00:32:32 +0000 (10:32 +1000)]
draw/tgsi: implement geom shader invocation support.

This is just for softpipe, llvmpipe won't work without
some changes.

Reviewed-by: Roland Scheidegger <sroland@vmware.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>
9 years agotgsi: handle indirect sampler arrays. (v2)
Dave Airlie [Wed, 27 May 2015 08:37:17 +0000 (18:37 +1000)]
tgsi: handle indirect sampler arrays. (v2)

This is required for ARB_gpu_shader5 support in softpipe.

v2: add support to txd/txf/txq paths.

Reviewed-by: Roland Scheidegger <sroland@vmware.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>
9 years agonir: Allow vec2/vec3/vec4 instructions in the select peephole pass.
Kenneth Graunke [Wed, 10 Jun 2015 07:52:07 +0000 (00:52 -0700)]
nir: Allow vec2/vec3/vec4 instructions in the select peephole pass.

These are basically just moves, so they should be safe as well.

When disabling i965's GLSL IR level scalarizer (channel expressions)
pass, I started seeing NIR code like this:

        if ssa_21 {
                block block_1:
                /* preds: block_0 */
                vec4 ssa_120 = vec4 ssa_82, ssa_83, ssa_84, ssa_30
                /* succs: block_3 */
        } else {
                block block_2:
                /* preds: block_0 */
                /* succs: block_3 */
        }
        block block_3:
        /* preds: block_1 block_2 */
        vec4 ssa_33 = phi block_1: ssa_120, block_2: ssa_2

Previously, the GLSL IR scalarizer pass would break the vec4 into a
series of fmovs, which were allowed by the peephole pass.  But with
the vec4 operation, they were not.  We want to keep getting selects.

Normal i965 on Broadwell:
instructions in affected programs:     200 -> 176 (-12.00%)
helped:                                4

With brw_fs_channel_expressions() disabled:
instructions in affected programs:     1832 -> 1646 (-10.15%)
helped:                                30

Signed-off-by: Kenneth Graunke <kenneth@whitecape.org>
Reviewed-by: Jason Ekstrand <jason.ekstrand@intel.com>
Reviewed-by: Connor Abbott <cwabbott0@gmail.com>
9 years agoi965: Add and fix comments in brw_vue_map.c.
Kenneth Graunke [Fri, 15 May 2015 16:58:42 +0000 (09:58 -0700)]
i965: Add and fix comments in brw_vue_map.c.

Signed-off-by: Kenneth Graunke <kenneth@whitecape.org>
Reviewed-by: Topi Pohjolainen <topi.pohjolainen@intel.com>