i965/vec4: Use the replicated fdot instruction in NIR

Reviewed-by: Connor Abbott <cwabbott0@gmail.com>
Reviewed-by: Eduardo Lima Mitev <elima@igalia.com>

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>

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>

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>

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>

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>

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>

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>

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>

i965/fs: Only read output_components many components when writing an output

Reviewed-by: Kristian Høgsberg <krh@bitplanet.net>

i965/fs: Set output_components for lowered clip distance outputs

Reviewed-by: Kristian Høgsberg <krh@bitplanet.net>

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>

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>

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>

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>

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>

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>

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>

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>

docs: document INTEL_DEBUG 'optimizer' envvar

Reviewed-by: Matt Turner <mattst88@gmail.com>

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>

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>

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>

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>

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>

vc4: Fix build from recent NIR cleanups.

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>

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>

docs: cleanups + mark some work as done

Signed-off-by: Tapani Pälli <tapani.palli@intel.com>
Reviewed-by: Ilia Mirkin <imirkin@alum.mit.edu>

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>

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>

r600g: add support for TXQS tgsi opcode

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

nv50/ir: add support for TXQS tgsi opcode

Signed-off-by: Ilia Mirkin <imirkin@alum.mit.edu>

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>

tgsi: add a TXQS opcode to retrieve the number of texture samples

Signed-off-by: Ilia Mirkin <imirkin@alum.mit.edu>
Reviewed-by: Roland Scheidegger <sroland@vmware.com>
Reviewed-by: Edward O'Callaghan <eocallaghan@alterapraxis.com>

glsl/cs: Initialize gl_LocalInvocationIndex in main()

We initialize gl_LocalInvocationIndex based on the extension spec
formula:

    gl_LocalInvocationIndex =
        gl_LocalInvocationID.z * gl_WorkGroupSize.x * gl_WorkGroupSize.y +
        gl_LocalInvocationID.y * gl_WorkGroupSize.x +
        gl_LocalInvocationID.x;

https://www.opengl.org/registry/specs/ARB/compute_shader.txt

Signed-off-by: Jordan Justen <jordan.l.justen@intel.com>
Reviewed-by: Tapani Pälli <tapani.palli@intel.com>

glsl/cs: Exclude gl_LocalInvocationIndex from builtin variable stripping

We lower gl_LocalInvocationIndex based on the extension spec formula:

    gl_LocalInvocationIndex =
        gl_LocalInvocationID.z * gl_WorkGroupSize.x * gl_WorkGroupSize.y +
        gl_LocalInvocationID.y * gl_WorkGroupSize.x +
        gl_LocalInvocationID.x;

https://www.opengl.org/registry/specs/ARB/compute_shader.txt

We need to set this variable in main(), even if gl_LocalInvocationIndex
is not referenced by the shader. (It may be used by a linked shader.)
Therefore, we can't eliminate it as a dead variable.

Signed-off-by: Jordan Justen <jordan.l.justen@intel.com>
Reviewed-by: Tapani Pälli <tapani.palli@intel.com>

glsl/cs: Initialize gl_GlobalInvocationID in main()

We initialize gl_GlobalInvocationID based on the extension spec
formula:

    gl_GlobalInvocationID =
        gl_WorkGroupID * gl_WorkGroupSize + gl_LocalInvocationID

https://www.opengl.org/registry/specs/ARB/compute_shader.txt

Signed-off-by: Jordan Justen <jordan.l.justen@intel.com>
Cc: Ilia Mirkin <imirkin@alum.mit.edu>
Reviewed-by: Tapani Pälli <tapani.palli@intel.com>

glsl: Move link_get_main_function_signature to a common location

Also rename to _mesa_get_main_function_signature.

We will call it near the end of compilation to insert some code into
main for initializing some compute shader global variables.

Signed-off-by: Jordan Justen <jordan.l.justen@intel.com>
Reviewed-by: Tapani Pälli <tapani.palli@intel.com>
Reviewed-by: Eduardo Lima Mitev <elima@igalia.com>

glsl/cs: Don't strip gl_GlobalInvocationID and dependencies

We lower gl_GlobalInvocationID based on the extension spec formula:

    gl_GlobalInvocationID =
        gl_WorkGroupID * gl_WorkGroupSize + gl_LocalInvocationID

https://www.opengl.org/registry/specs/ARB/compute_shader.txt

We need to set this variable in main(), even if gl_GlobalInvocationID
is not referenced by the shader. (It may be used by a linked shader.)
Therefore, we can't eliminate these as dead variables.

Signed-off-by: Jordan Justen <jordan.l.justen@intel.com>
Reviewed-by: Tapani Pälli <tapani.palli@intel.com>

i965/nir: Support gl_WorkGroupID variable

Signed-off-by: Jordan Justen <jordan.l.justen@intel.com>
Reviewed-by: Kristian Høgsberg <krh@bitplanet.net>

i965/cs: Initialize gl_WorkGroupID variable from payload

Signed-off-by: Jordan Justen <jordan.l.justen@intel.com>
Reviewed-by: Kristian Høgsberg <krh@bitplanet.net>

nir: Add gl_WorkGroupID system variable

Signed-off-by: Jordan Justen <jordan.l.justen@intel.com>
Reviewed-by: Kristian Høgsberg <krh@bitplanet.net>

glsl/cs: Add gl_WorkGroupID variable

Signed-off-by: Jordan Justen <jordan.l.justen@intel.com>
Reviewed-by: Kristian Høgsberg <krh@bitplanet.net>

i965/nir: Support gl_LocalInvocationID variable

Signed-off-by: Jordan Justen <jordan.l.justen@intel.com>
Reviewed-by: Kristian Høgsberg <krh@bitplanet.net>

i965/cs: Initialize gl_LocalInvocationID from payload

Signed-off-by: Jordan Justen <jordan.l.justen@intel.com>
Reviewed-by: Kristian Høgsberg <krh@bitplanet.net>

i965/cs: Initialize gl_LocalInvocationID in push constant data

Signed-off-by: Jordan Justen <jordan.l.justen@intel.com>
Reviewed-by: Kristian Høgsberg <krh@bitplanet.net>

i965/cs: Reserve local invocation id in payload regs

Signed-off-by: Jordan Justen <jordan.l.justen@intel.com>
Reviewed-by: Kristian Høgsberg <krh@bitplanet.net>
Reviewed-by: Iago Toral Quiroga <itoral@igalia.com>

nir: Add gl_LocalInvocationID variable

Signed-off-by: Jordan Justen <jordan.l.justen@intel.com>
Reviewed-by: Iago Toral Quiroga <itoral@igalia.com>
Reviewed-by: Kristian Høgsberg <krh@bitplanet.net>

glsl/cs: Add gl_LocalInvocationID variable

Signed-off-by: Jordan Justen <jordan.l.justen@intel.com>
Reviewed-by: Iago Toral Quiroga <itoral@igalia.com>
Reviewed-by: Kristian Høgsberg <krh@bitplanet.net>

softpipe: Change faces type to uint

This is to avoid needless float<->int conversions, since all
face-related computations are made on integers. Spotted by Emil
Velikov.

Reviewed-by: Brian Paul <brianp@vmware.com>

freedreno/ir3: fix compile warn after 1807a08e

New enum to add to switch so compiler doesn't complain.

   commit 1807a08e4f35b014f2a80d1e88dd74a9f096d7a5
   Author:     Ilia Mirkin <imirkin@alum.mit.edu>
   AuthorDate: Thu Aug 27 23:05:03 2015 -0400
   Commit:     Ilia Mirkin <imirkin@alum.mit.edu>
   CommitDate: Thu Sep 10 17:38:33 2015 -0400

       nir: add nir_texop_texture_samples and convert from glsl

Signed-off-by: Rob Clark <robclark@freedesktop.org>

freedreno/ir3: fix compile break after a4aa25be

Following commit dropped the unused memctx arg:

   commit a4aa25be1e0a27b1a6a6b0bcf576beb9dfe1ea7a
   Author:     Jason Ekstrand <jason.ekstrand@intel.com>
   AuthorDate: Wed Sep 9 13:24:35 2015 -0700
   Commit:     Jason Ekstrand <jason.ekstrand@intel.com>
   CommitDate: Fri Sep 11 09:21:20 2015 -0700

       nir: Remove the mem_ctx parameter from ssa_def_rewrite_uses

Signed-off-by: Rob Clark <robclark@freedesktop.org>

nir: add nir_channel() to get at single components of vec's

Rather than make yet another copy of channel(), let's move it into nir.

Signed-off-by: Rob Clark <robclark@freedesktop.org>
Reviewed-by: Jason Ekstrand <jason.ekstrand@intel.com>

tgsi/scan: add support to figure out max nesting depth

Sometimes a useful thing for compilers (or, for example, tgsi_to_nir) to
know.  And pretty trivial for scan to figure this out for us.

Signed-off-by: Rob Clark <robclark@freedesktop.org>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Reviewed-by: Brian Paul <brianp@vmware.com>

r600: Fix llvm build since const buffer changes

In commit f9caabe8f1bff86d19b53d9ecba5c72b238d9e23:

One place in r600_llvm.c was forgotten when replacing
R600_UCP_CONST_BUFFER with R600_BUFFER_INFO_CONST_BUFFER.

Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=91985
Signed-off-by: Kai Wasserbäch <kai@dev.carbon-project.org>
Signed-off-by: Dave Airlie <airlied@gmail.com>

i965/vec4: Don't reswizzle hardware registers

Cc: "11.0 10.6" <mesa-stable@lists.freedesktop.org>
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=91719
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>

i965/emit: Add assertions for accumulator restrictions

Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>

docs: add news item and link release notes for 11.0.0

Signed-off-by: Emil Velikov <emil.l.velikov@gmail.com>

docs: add sha256 checksums for 11.0.0

Signed-off-by: Emil Velikov <emil.l.velikov@gmail.com>
(cherry picked from commit c4bae5792bb5515da42e23f166f5ba5d68f79615)

docs: Update 11.0.0 release notes

Signed-off-by: Emil Velikov <emil.l.velikov@gmail.com>
(cherry picked from commit 4f1e500150be2e82a2d7eb954f7198cc0c5cbec1)

r600: Enable fp64 on chips with native support

Cypress/Cayman/Aruba, earlier r6xx/r7xx chips only support a subset
of the needed fp64 ops, and don't do GL4 anyway.

Signed-off-by: Glenn Kennard <glenn.kennard@gmail.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>

r600g: Support I2D/U2D/D2I/D2U

Only for Cypress/Cayman/Aruba, older chips have only partial fp64 support.
Uses float intermediate values so only accurate for int24 range, which
matches what the blob does.

Signed-off-by: Glenn Kennard <glenn.kennard@gmail.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>

r600g: lower number of driver const buffers

I'm going to want a driver constant buffer for tess to coordinate
LDS storage, so before I go tackling that I decided to merge the
clip/samplepos and texture info buffers into one. So I can steal
the spare one.

This creates a single constant buffer between the two, with
clip/samplepos taking up a reserved 128 bytes at the start.

Reviewed-by: Edward O'Callaghan <eocallaghan@alterapraxis.com>
Reviewed-by: Glenn Kennard <glenn.kennard@gmail.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>

r600: define some values for the fetch constant offsets.

This just puts these in one place and #defines them.

Reviewed-by: Glenn Kennard <glenn.kennard@gmail.com>
Reviewed-by: Edward O'Callaghan <eocallaghan@alterapraxis.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>

docs: Update with GLES3.2 entries and status

V2: -Change to "not started" for most entries
    -Add status for multisample_2d_array
    -Change shader_multisample_interpolation to "not_stared"

V3 (idr): Move the GLES 3.2 section after the "Additional functions"
section from GLES 3.1.  Note that GL_KHR_texture_compression_astc_hdr is
done for i965 on gen9+ hardware.  Note that GL_OES_shader_io_blocks is
based on some features from GLSL 1.50.

Signed-off-by: Thomas Helland <thomashelland90@gmail.com>
Signed-off-by: Ian Romanick <ian.d.romanick@intel.com>
Reviewed-by: Tapani Pälli <tapani.palli@intel.com> [v2]
Reviewed-by: Ian Romanick <ian.d.romanick@intel.com>

softpipe: Constify variables

This commit makes a lot of variables constant - this is basically done
by moving the computation to variable definition. Some of them are
moved into lower scopes (like in img_filter_2d_ewa).

Reviewed-by: Brian Paul <brianp@vmware.com>
Reviewed-by: Roland Scheidegger <sroland@vmware.com>

softpipe: Constify sp_tgsi_sampler

Add a small inline function doing the casting - this is to make sure
we don't do a cast from some completely unrelated type. This commit
does not make tgsi_sampler parameters const in vfuncs themselves for
now - probably llvmpipe would need looking at before making such a
change.

Reviewed-by: Brian Paul <brianp@vmware.com>
Reviewed-by: Roland Scheidegger <sroland@vmware.com>

softpipe: Constify sampler and view parameters in mip filters

Those functions actually could always take them as constants.

Reviewed-by: Brian Paul <brianp@vmware.com>
Reviewed-by: Roland Scheidegger <sroland@vmware.com>

softpipe: Constify sampler and view parameters in img filters

Those functions actually could always take them as constants.

Reviewed-by: Brian Paul <brianp@vmware.com>
Reviewed-by: Roland Scheidegger <sroland@vmware.com>

tgsi, softpipe: Constify tgsi_sampler in query_lod vfunc

A followup from previous commit - since all functions called by
query_lod take pointers to const sp_sampler_view and const sp_sampler,
which are taken from tgsi_sampler subclass, we can the tgsi_sampler as
const itself now.

Reviewed-by: Brian Paul <brianp@vmware.com>
Reviewed-by: Roland Scheidegger <sroland@vmware.com>

softpipe: Constify some sampler and view parameters

This is to prepare for making tgsi_sampler parameter in query_lod a
const too. These functions do not modify anything in either sampler or
view anymore.

Reviewed-by: Brian Paul <brianp@vmware.com>
Reviewed-by: Roland Scheidegger <sroland@vmware.com>

softpipe: Move the faces array from view to filter_args

With that, sp_sampler_view instances are not abused anymore as a local
storage, so we can later make them constant.

Reviewed-by: Brian Paul <brianp@vmware.com>
Reviewed-by: Roland Scheidegger <sroland@vmware.com>

nir/from_ssa: Use instr_rewrite_dest

Reviewed-by: Eduardo Lima Mitev <elima@igalia.com>

nir: Add a function for rewriting instruction destinations

Reviewed-by: Eduardo Lima Mitev <elima@igalia.com>

nir: Only unlink sources that are actually valid

Reviewed-by: Thomas Helland <thomashelland90@gmail.com>

nir: Remove the mem_ctx parameter from ssa_def_rewrite_uses

Reviewed-by: Thomas Helland <thomashelland90@gmail.com>

nir: Fix a bunch of ralloc parenting errors

As of a10d4937, we would really like things associated with an instruction
to be allocated out of that instruction and not out of the shader.  In
particular, you should be passing the instruction that will ultimately be
holding the source into nir_src_copy rather than an arbitrary memory
context.

We also change the prototypes of nir_dest_copy and nir_alu_src/dest_copy to
explicitly take an instruction so we catch this earlier in the future.

Cc: "11.0" <mesa-stable@lists.freedesktop.org>
Reviewed-by: Thomas Helland <thomashelland90@gmail.com>

nir/lower_outputs_to_temporaries: Reparent the output name

We copy the output, make the old output the temporary, and give the
temporary a new name.  The copy keeps the pointer to the old name.  This
works just fine up until the point where we lower things to SSA and delete
the old variable and, with it, the name.  Instead, we should re-parent to
the copy.

Reviewed-by: Eduardo Lima Mitev <elima@igalia.com>

i965/vec4: check writemask when bailing out at register coalesce

opt_register_coalesce stopped to check previous instructions to
coalesce with if somebody else was writing on the same
destination. This can be optimized to check if somebody else was
writing to the same channels of the same destination using the
writemask.

Shader DB results (taking into account only vec4):

total instructions in shared programs: 1781593 -> 1734957 (-2.62%)
instructions in affected programs:     1238390 -> 1191754 (-3.77%)
helped:                                12782
HURT:                                  0
GAINED:                                0
LOST:                                  0

v2: removed some parenthesis, fixed indentation, as suggested by
    Matt Turner
v3: added brackets, for consistency, as suggested by Eduardo Lima

Reviewed-by: Matt Turner <mattst88@gmail.com>

tgsi,softpipe: capitalize the tgsi_sampler_control enum values

We use capitalized enum values everywhere else.
This improves understanding a bit too.

Reviewed-by: Roland Scheidegger <sroland@vmware.com>

nir: Store some geometry shader data in nir_shader.

This makes it possible for NIR shaders to know the number of output
vertices and the number of invocations.  Drivers could also access
these directly without going through gl_program.

We should probably add InputType and OutputType here too, but currently
those are stored as GL_* enums, and I wanted to avoid using those in
NIR, as I suspect Vulkan/SPIR-V will use different enums.  (We should
probably make our own.)

We could add VerticesIn, but it's easily computable from the input
topology, so I'm not sure whether it's worth it.  It's also currently
not stored in gl_shader (only gl_shader_program), which would require
changes to the glsl_to_nir interface or require us to store it there.

This is a bit of duplication of data...ideally, we would factor these
substructs out of gl_program, gl_shader_program, and nir_shader, creating
a gl_geometry_info class...but it would need to go in a new place (in
src/glsl?) that isn't mtypes.h nor nir.h.

Signed-off-by: Kenneth Graunke <kenneth@whitecape.org>
Reviewed-by: Jason Ekstrand <jason.ekstrand@intel.com>

nir/builder: Add nir_load_var() and nir_store_var() helpers.

These provide a convenient way to do simple variable loads and stores.

Signed-off-by: Kenneth Graunke <kenneth@whitecape.org>
Reviewed-by: Jason Ekstrand <jason.ekstrand@intel.com>

glsl: Use hash tables for opt_constant_propagation() kill sets.

Cuts compile/link time of the fragment shader in #91857 by 19%
(16.28 -> 13.05).

I didn't bother with the acp sets because they're smaller, but it
might be worth doing as well.

Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=91857
Signed-off-by: Kenneth Graunke <kenneth@whitecape.org>
Reviewed-by: Thomas Helland <thomashelland90@gmail.com>
Reviewed-by: Timothy Arceri <t_arceri@yahoo.com.au>
Tested-by: Tapani Pälli <tapani.palli@intel.com>

i965: Use hash tables for brw_fs_vector_splitting().

Cuts compile/link time of the fragment shader in #91857 by 25%
(21.64 -> 16.28).

v2: Drop unnecessary _mesa_hash_table_destroy call, and use
    refs.ht->entries == 0 rather than ad-hoc checking (suggested by
    Timothy Arceri).

Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=91857
Signed-off-by: Kenneth Graunke <kenneth@whitecape.org>
Reviewed-by: Thomas Helland <thomashelland90@gmail.com>
Reviewed-by: Timothy Arceri <t_arceri@yahoo.com.au>
Tested-by: Tapani Pälli <tapani.palli@intel.com>

glsl: Use hash tables in opt_constant_variable().

Cuts compile/link time of the fragment shader in bug #91857 by 31%
(31.79 -> 21.64).  It has over 8,000 variables so linked lists are
terrible.

Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=91857
Signed-off-by: Kenneth Graunke <kenneth@whitecape.org>
Reviewed-by: Thomas Helland <thomashelland90@gmail.com>
Reviewed-by: Timothy Arceri <t_arceri@yahoo.com.au>
Tested-by: Tapani Pälli <tapani.palli@intel.com>

meta: Use result of texture coordinate clamping operation

Previously the result of the complicated clamp() expression just dropped
on the floor: clamp does not modify any of its parameters.  Looking at
the surrounding code, I believe this is supposed to modify the value of
tex_coord.

This change (along with a change to avoid the use of
brw_blorp_framebuffer) does not affect any existing piglit tests.  I'm
not sure what this clamp is trying to accomplish, so I'm not sure how to
write a test to exercise this path.

I also noticed another bug in this code.  There is no way the array
texture case could possibly work.  This will generate code for the
TEXEL_FETCH macro like:

    #define TEXEL_FETCH(coord) texelFetch(texSampler, ivec3(coord), sample_map[int(2 * fract(coord.x))]);

Since the coord parameter of this macro is a vec2 at all invocations, no
expansion of this macro will even compile.

Signed-off-by: Ian Romanick <ian.d.romanick@intel.com>
Reviewed-by: Anuj Phogat <anuj.phogat@gmail.com>
Cc: Topi Pohjolainen <topi.pohjolainen@intel.com>
Cc: Jordan Justen <jordan.l.justen@intel.com>

meta: Always bind the texture

We may have been called from glGenerateTextureMipmap with CurrentUnit
still set to 0, so we don't know when we can skip binding the texture.
Assume that _mesa_BindTexture will be fast if we're rebinding the same
texture.

v2: Remove currentTexUnitSave because it is now unused.  Suggested by
both Neil and Anuj.

Signed-off-by: Ian Romanick <ian.d.romanick@intel.com>
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=91847
Cc: "11.0" <mesa-stable@lists.freedesktop.org>
Reviewed-by: Neil Roberts <neil@linux.intel.com>
Reviewed-by: Anuj Phogat <anuj.phogat@gmail.com>

i915, i965: Silence unused parameter warnings in intel_batchbuffer_advance

These only occurred in release builds, but they occurred in every file
that included intel_batchbuffer.h.  Lots of spam. :(

intel_batchbuffer.h: In function 'intel_batchbuffer_advance':
intel_batchbuffer.h:153:47: warning: unused parameter 'brw' [-Wunused-parameter]
 intel_batchbuffer_advance(struct brw_context *brw)
                                               ^

Signed-off-by: Ian Romanick <ian.d.romanick@intel.com>
Reviewed-by: Anuj Phogat <anuj.phogat@gmail.com>

i915: Silence unused parameter warning in intel_miptree_create_layout

The for_bo parameter of intel_miptree_create_layout appears to be unused
since 27eedca when Eric removed some Gen5 code (after the i915 and i965
drivers parted ways).

intel_mipmap_tree.c: In function 'old_intel_miptree_create_layout':
intel_mipmap_tree.c:77:35: warning: unused parameter 'for_bo' [-Wunused-parameter]
                             bool for_bo)
                                   ^

Signed-off-by: Ian Romanick <ian.d.romanick@intel.com>
Reviewed-by: Anuj Phogat <anuj.phogat@gmail.com>

i915, i965: Silence unused parameter warnings in intel_miptree_unmap_gtt

intel_mipmap_tree.c: In function 'intel_miptree_unmap_gtt':
intel_mipmap_tree.c:777:34: warning: unused parameter 'map' [-Wunused-parameter]
    struct intel_miptree_map *map,
                                  ^
intel_mipmap_tree.c:778:17: warning: unused parameter 'level' [-Wunused-parameter]
    unsigned int level,
                 ^
intel_mipmap_tree.c:779:17: warning: unused parameter 'slice' [-Wunused-parameter]
    unsigned int slice)
                 ^

Signed-off-by: Ian Romanick <ian.d.romanick@intel.com>
Reviewed-by: Anuj Phogat <anuj.phogat@gmail.com>

i915: Silence unused parameter warnings

intel_mipmap_tree.c: In function 'old_intel_miptree_unmap_raw':
intel_mipmap_tree.c:726:51: warning: unused parameter 'intel' [-Wunused-parameter]
 intel_miptree_unmap_raw(struct intel_context *intel,
                                                   ^

Signed-off-by: Ian Romanick <ian.d.romanick@intel.com>
Reviewed-by: Anuj Phogat <anuj.phogat@gmail.com>

i915: Remove prototype for nonexistent brw_miptree_layout

Hasn't existed in the i915 source since the i915 and i965 drivers parted
ways.

Signed-off-by: Ian Romanick <ian.d.romanick@intel.com>
Reviewed-by: Anuj Phogat <anuj.phogat@gmail.com>

i965: Make intel_miptree_map_raw static

This hasn't been used outside intel_mipmap_tree.c since d5d4ba9 started
using meta instead of the blitter for PBO TexSubImage.  While we're
here, remove the unused brw parameter from the function formerly known
as intel_miptree_unmap_raw.

Signed-off-by: Ian Romanick <ian.d.romanick@intel.com>
Reviewed-by: Anuj Phogat <anuj.phogat@gmail.com>

i915, i965: Silence unused parameter warnings in intel_mipmap_tree.h

These only occurred in release builds, but they occurred in every file
that included intel_mipmap_tree.h.  Lots of spam. :(

intel_mipmap_tree.h: In function 'intel_miptree_check_level_layer':
intel_mipmap_tree.h:595:59: warning: unused parameter 'mt' [-Wunused-parameter]
 intel_miptree_check_level_layer(struct intel_mipmap_tree *mt,
                                                           ^
intel_mipmap_tree.h:596:42: warning: unused parameter 'level' [-Wunused-parameter]
                                 uint32_t level,
                                          ^
intel_mipmap_tree.h:597:42: warning: unused parameter 'layer' [-Wunused-parameter]
                                 uint32_t layer)
                                          ^

Signed-off-by: Ian Romanick <ian.d.romanick@intel.com>
Reviewed-by: Anuj Phogat <anuj.phogat@gmail.com>

i965: Silence unused parameter warnings in intel_mipmap_tree.c

The target parameter of compute_msaa_layout appears to be unused since
83b83fb when support for CMS textures was added for Gen7.

The brw parameter of intel_get_non_msrt_mcs_alignment appears to be
unused since e92fbdc when the GEN check (along with the "can we fast
clear" decision) was moved to a different function.

intel_mipmap_tree.c: In function 'compute_msaa_layout':
intel_mipmap_tree.c:62:73: warning: unused parameter 'target' [-Wunused-parameter]
 compute_msaa_layout(struct brw_context *brw, mesa_format format, GLenum target,
                                                                         ^
intel_mipmap_tree.c: In function 'intel_get_non_msrt_mcs_alignment':
intel_mipmap_tree.c:143:54: warning: unused parameter 'brw' [-Wunused-parameter]
 intel_get_non_msrt_mcs_alignment(struct brw_context *brw,
                                                      ^

Signed-off-by: Ian Romanick <ian.d.romanick@intel.com>
Reviewed-by: Anuj Phogat <anuj.phogat@gmail.com>
Cc: Ben Widawsky <benjamin.widawsky@intel.com>

i965: Silence unused parameter warnings in intel_fbo.c

intel_fbo.c: In function 'intel_alloc_window_storage':
intel_fbo.c:415:48: warning: unused parameter 'ctx' [-Wunused-parameter]
 intel_alloc_window_storage(struct gl_context * ctx, struct gl_renderbuffer *rb,
                                                ^
intel_fbo.c: In function 'intel_nop_alloc_storage':
intel_fbo.c:428:74: warning: unused parameter 'rb' [-Wunused-parameter]
 intel_nop_alloc_storage(struct gl_context * ctx, struct gl_renderbuffer *rb,
                                                                          ^
intel_fbo.c:429:32: warning: unused parameter 'internalFormat' [-Wunused-parameter]
                         GLenum internalFormat, GLuint width, GLuint height)
                                ^
intel_fbo.c:429:55: warning: unused parameter 'width' [-Wunused-parameter]
                         GLenum internalFormat, GLuint width, GLuint height)
                                                       ^
intel_fbo.c:429:69: warning: unused parameter 'height' [-Wunused-parameter]
                         GLenum internalFormat, GLuint width, GLuint height)
                                                                     ^
intel_fbo.c: In function 'intel_blit_framebuffer_with_blitter':
intel_fbo.c:790:61: warning: unused parameter 'filter' [-Wunused-parameter]
                                     GLbitfield mask, GLenum filter)
                                                             ^

Signed-off-by: Ian Romanick <ian.d.romanick@intel.com>
Reviewed-by: Anuj Phogat <anuj.phogat@gmail.com>

st/mesa: set the vbuffer to NULL if we are skipping it

If we skip a vbuffer we need to make sure we NULL out
the contents, otherwise when it gets passed to the driver
it will get confused.

This was hit by:
GL41-CTS.gpu_shader_fp64.varyings

Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>

i965/cs: Enable barrier in MEDIA_INTERFACE_DESCRIPTOR

Enable barrier in MEDIA_INTERFACE_DESCRIPTOR if the program uses the
barrier() GLSL function.

On Ivy Bridge and Haswell, this allows the piglit test
tests/spec/arb_compute_shader/execution/simple-barrier-atomics.shader_test
to pass. On gen8, this enables a similar test with a local group size
of 896 to pass.

Signed-off-by: Jordan Justen <jordan.l.justen@intel.com>
Reviewed-by: Kristian Høgsberg <krh@bitplanet.net>

i965/cs: Emit texture surfaces to enable CS sampling

Signed-off-by: Jordan Justen <jordan.l.justen@intel.com>
Reviewed-by: Kristian Høgsberg <krh@bitplanet.net>