mesa.git
9 years agonir/spirv: Add support for constants
Jason Ekstrand [Thu, 30 Apr 2015 01:14:11 +0000 (18:14 -0700)]
nir/spirv: Add support for constants

9 years agonir/spirv: Add basic support for types
Jason Ekstrand [Wed, 29 Apr 2015 21:36:01 +0000 (14:36 -0700)]
nir/spirv: Add basic support for types

9 years agonir/types: Add more helpers for creating types
Jason Ekstrand [Wed, 29 Apr 2015 21:28:37 +0000 (14:28 -0700)]
nir/types: Add more helpers for creating types

9 years agoglsl/types: Expose the function_param and struct_field structs to C
Jason Ekstrand [Wed, 29 Apr 2015 21:48:12 +0000 (14:48 -0700)]
glsl/types: Expose the function_param and struct_field structs to C

Previously, they were hidden behind a #ifdef __cplusplus so C wouldn't find
them.  This commit simpliy moves the ifdef.

9 years agoglsl/types: Add support for function types
Jason Ekstrand [Wed, 25 Feb 2015 00:29:33 +0000 (16:29 -0800)]
glsl/types: Add support for function types

9 years agoglsl: Add GLSL_TYPE_FUNCTION to the base types enums
Jason Ekstrand [Wed, 25 Feb 2015 00:27:32 +0000 (16:27 -0800)]
glsl: Add GLSL_TYPE_FUNCTION to the base types enums

9 years agonir/spirv: Rework the way values are added
Jason Ekstrand [Wed, 29 Apr 2015 21:34:06 +0000 (14:34 -0700)]
nir/spirv: Rework the way values are added

Instead of having functions to add values and set various things, we just
have a function that does a few asserts and then returns the value.  The
caller is then responsible for setting the various fields.

9 years agonir/spirv: Add stub support for extension instructions
Jason Ekstrand [Wed, 29 Apr 2015 21:32:55 +0000 (14:32 -0700)]
nir/spirv: Add stub support for extension instructions

9 years agoREVERT: Add a simple helper program for testing SPIR-V -> NIR translation
Jason Ekstrand [Wed, 29 Apr 2015 21:29:38 +0000 (14:29 -0700)]
REVERT: Add a simple helper program for testing SPIR-V -> NIR translation

9 years agoglsl/compiler: Move the error_no_memory stub to standalone_scaffolding.cpp
Jason Ekstrand [Wed, 29 Apr 2015 21:30:22 +0000 (14:30 -0700)]
glsl/compiler: Move the error_no_memory stub to standalone_scaffolding.cpp

9 years agonir: Add the start of a SPIR-V to NIR translator
Jason Ekstrand [Wed, 29 Apr 2015 00:43:16 +0000 (17:43 -0700)]
nir: Add the start of a SPIR-V to NIR translator

At the moment, it can handle the very basics of strings and can ignore
debug instructions.  It also has basic support for decorations.

9 years agonir: Import the revision 30 SPIR-V header from Khronos
Jason Ekstrand [Tue, 28 Apr 2015 04:13:44 +0000 (21:13 -0700)]
nir: Import the revision 30 SPIR-V header from Khronos

9 years agonir: Fix aggressive typos in nir_from_ssa.c.
Kenneth Graunke [Sat, 9 May 2015 02:36:19 +0000 (19:36 -0700)]
nir: Fix aggressive typos in nir_from_ssa.c.

s/agressive/aggressive/g

Trivial.

9 years agonir/search: Save/restore the variables_seen bitmask when matching
Jason Ekstrand [Fri, 8 May 2015 16:48:33 +0000 (09:48 -0700)]
nir/search: Save/restore the variables_seen bitmask when matching

Shader-db results on Broadwell:

   total instructions in shared programs: 7152330 -> 7137006 (-0.21%)
   instructions in affected programs:     1330548 -> 1315224 (-1.15%)
   helped:                                5797
   HURT:                                  76
   GAINED:                                0
   LOST:                                  8

Reviewed-by: Connor Abbott <cwabbott0@gmail.com>
9 years agonir/search: Assert that variable id's are in range
Jason Ekstrand [Fri, 8 May 2015 16:42:05 +0000 (09:42 -0700)]
nir/search: Assert that variable id's are in range

Reviewed-by: Connor Abbott <cwabbott0@gmail.com>
9 years agonir/search: handle explicitly sized sources in match_value
Jason Ekstrand [Fri, 8 May 2015 15:33:01 +0000 (08:33 -0700)]
nir/search: handle explicitly sized sources in match_value

Previously, this case was being handled in match_expression prior to
calling match_value.  However, there is really no good reason for this
given that match_value has all of the information it needs.  Also, they
weren't being handled properly in the commutative case and putting it in
match_value gives us that for free.

Reviewed-by: Connor Abbott <cwabbott0@gmail.com>
9 years agonir/nir: Use a linked list instead of a hash set for use/def sets
Jason Ekstrand [Fri, 24 Apr 2015 17:16:27 +0000 (10:16 -0700)]
nir/nir: Use a linked list instead of a hash set for use/def sets

This commit switches us from the current setup of using hash sets for
use/def sets to using linked lists.  Doing so should save us quite a bit of
memory because we aren't carrying around 3 hash sets per register and 2 per
SSA value.  It should also save us CPU time because adding/removing things
from use/def sets is 4 pointer manipulations instead of a hash lookup.

Running shader-db 50 times with USE_NIR=0, NIR, and NIR + use/def lists:

   GLSL IR Only:        586.4 +/- 1.653833
   NIR with hash sets:  675.4 +/- 2.502108
   NIR + use/def lists: 641.2 +/- 1.557043

I also ran a memory usage experiment with Ken's patch to delete GLSL IR and
keep NIR.  This patch cuts an aditional 42.9 MiB of ralloc'd memory over
and above what we gained by deleting the GLSL IR on the same dota trace.

On the code complexity side of things, some things are now much easier and
others are a bit harder.  One of the operations we perform constantly in
optimization passes is to replace one source with another.  Due to the fact
that an instruction can use the same SSA value multiple times, we had to
iterate through the sources of the instruction and determine if the use we
were replacing was the only one before removing it from the set of uses.
With this patch, uses are per-source not per-instruction so we can just
remove it safely.  On the other hand, trying to iterate over all of the
instructions that use a given value is more difficult.  Fortunately, the
two places we do that are the ffma peephole where it doesn't matter and GCM
where we already gracefully handle duplicates visits to an instruction.

Another aspect here is that using linked lists in this way can be tricky to
get right.  With sets, things were quite forgiving and the worst that
happened if you didn't properly remove a use was that it would get caught
in the validator.  With linked lists, it can lead to linked list corruption
which can be harder to track.  However, we do just as much validation of
the linked lists as we did of the sets so the validator should still catch
these problems.  While working on this series, the vast majority of the
bugs I had to fix were caught by assertions.  I don't think the lists are
going to be that much worse than the sets.

Reviewed-by: Connor Abbott <cwabbott0@gmail.com>
9 years agoutil/list: Add a list validation function
Jason Ekstrand [Tue, 28 Apr 2015 03:40:11 +0000 (20:40 -0700)]
util/list: Add a list validation function

Acked-by: Connor Abbott <cwabbott0@gmail.com>
Reviewed-by: Rob Clark <robclark@freedesktop.org>
9 years agoutil/list: Add list_empty and list_length functions
Jason Ekstrand [Tue, 28 Apr 2015 03:39:37 +0000 (20:39 -0700)]
util/list: Add list_empty and list_length functions

v2: Don't use C99 when iterating over the list

Acked-by: Connor Abbott <cwabbott0@gmail.com>
Reviewed-by: Rob Clark <robclark@freedesktop.org>
9 years agoutil/list: Add C99-based iterator macros
Jason Ekstrand [Tue, 28 Apr 2015 01:56:02 +0000 (18:56 -0700)]
util/list: Add C99-based iterator macros

v2: Use LIST_ENTRY instead of container_of in iterators

Acked-by: Connor Abbott <cwabbott0@gmail.com>
Reviewed-by: Rob Clark <robclark@freedesktop.org>
9 years agoutil: Move gallium's linked list to util
Jason Ekstrand [Tue, 28 Apr 2015 00:41:27 +0000 (17:41 -0700)]
util: Move gallium's linked list to util

The linked list in gallium is pretty much the kernel list and we would like
to have a C-based linked list for all of mesa.  Let's not duplicate and
just steal the gallium one.

Acked-by: Connor Abbott <cwabbott0@gmail.com>
Reviewed-by: Rob Clark <robclark@freedesktop.org>
9 years agogallium/double_list: s/INLINE/inline and remove the p_compiler include
Jason Ekstrand [Mon, 27 Apr 2015 23:58:29 +0000 (16:58 -0700)]
gallium/double_list: s/INLINE/inline and remove the p_compiler include

Acked-by: Connor Abbott <cwabbott0@gmail.com>
Reviewed-by: Rob Clark <robclark@freedesktop.org>
9 years agonir: Use nir_instr_rewrite_src in copy propagation
Jason Ekstrand [Tue, 21 Apr 2015 19:12:26 +0000 (12:12 -0700)]
nir: Use nir_instr_rewrite_src in copy propagation

We were rolling our own rewrite_src variant in copy-propagation.  Let's
stop doing that and use the ones in core NIR.

Reviewed-by: Connor Abbott <cwabbott0@gmail.com>
9 years agonir: Add a function for rewriting the condition of an if statement
Jason Ekstrand [Fri, 24 Apr 2015 17:34:30 +0000 (10:34 -0700)]
nir: Add a function for rewriting the condition of an if statement

Reviewed-by: Connor Abbott <cwabbott0@gmail.com>
9 years agonir: Add and use initializer #defines for nir_src and nir_dest
Jason Ekstrand [Wed, 22 Apr 2015 01:00:21 +0000 (18:00 -0700)]
nir: Add and use initializer #defines for nir_src and nir_dest

Reviewed-by: Connor Abbott <cwabbott0@gmail.com>
9 years agonir: Modernize the out-of-SSA pass
Jason Ekstrand [Tue, 21 Apr 2015 18:16:04 +0000 (11:16 -0700)]
nir: Modernize the out-of-SSA pass

The out-of-SSA pass was one of the first passes written when getting SSA
up-and-going (for obvious reasons).  As such, it came before a lot of the
nifty SSA-based helpers were introduced.  This commit modernizes it so that
we're no longer doing nearly as much manual banging on use/def sets.

Reviewed-by: Connor Abbott <cwabbott0@gmail.com>
9 years agonir/validate: Validate SSA def parent instructions
Jason Ekstrand [Fri, 24 Apr 2015 00:10:42 +0000 (17:10 -0700)]
nir/validate: Validate SSA def parent instructions

Reviewed-by: Connor Abbott <cwabbott0@gmail.com>
9 years agonv50/ir: only propagate saturate up if some actual folding took place
Ilia Mirkin [Fri, 8 May 2015 22:54:08 +0000 (18:54 -0400)]
nv50/ir: only propagate saturate up if some actual folding took place

The former logic would copy the saturate up to any mul with an immediate
if there was a subsequent mul with a saturate. However we only want to
do that if we collapsed 2 muls by multiplying their immediates (or were
able to put the immediate in as a post-multiplier).

Signed-off-by: Ilia Mirkin <imirkin@alum.mit.edu>
Cc: mesa-stable@lists.freedesktop.org
9 years agonir: Delete all traces of nir_op_flog
Ian Romanick [Thu, 7 May 2015 00:04:15 +0000 (17:04 -0700)]
nir: Delete all traces of nir_op_flog

Nothing produces it, and nothing can consume it.

Signed-off-by: Ian Romanick <ian.d.romanick@intel.com>
Reviewed-by: Matt Turner <mattst88@gmail.com>
Acked-by: Jason Ekstrand <jason.ekstrand@intel.com>
9 years agonir: Don't produce nir_op_flog from GLSL IR
Ian Romanick [Thu, 7 May 2015 00:01:37 +0000 (17:01 -0700)]
nir: Don't produce nir_op_flog from GLSL IR

All paths that produce GLSL IR for NIR lower ir_unop_log.  All paths
that consume NIR will explode if they geta nir_op_flog.

Signed-off-by: Ian Romanick <ian.d.romanick@intel.com>
Reviewed-by: Matt Turner <mattst88@gmail.com>
Acked-by: Jason Ekstrand <jason.ekstrand@intel.com>
9 years agonir: Delete all traces of nir_op_fexp
Ian Romanick [Thu, 7 May 2015 00:00:18 +0000 (17:00 -0700)]
nir: Delete all traces of nir_op_fexp

Nothing produces it, and nothing can consume it.

Signed-off-by: Ian Romanick <ian.d.romanick@intel.com>
Reviewed-by: Matt Turner <mattst88@gmail.com>
Acked-by: Jason Ekstrand <jason.ekstrand@intel.com>
9 years agonir: Don't produce nir_op_fexp from GLSL IR
Ian Romanick [Wed, 6 May 2015 23:57:22 +0000 (16:57 -0700)]
nir: Don't produce nir_op_fexp from GLSL IR

All paths that produce GLSL IR for NIR lower ir_unop_exp.  All paths
that consume NIR will explode if they geta nir_op_fexp.

Signed-off-by: Ian Romanick <ian.d.romanick@intel.com>
Reviewed-by: Matt Turner <mattst88@gmail.com>
Acked-by: Jason Ekstrand <jason.ekstrand@intel.com>
9 years agoprog_to_nir: OPCODE_EXP is not nir_op_fexp
Ian Romanick [Wed, 6 May 2015 23:54:06 +0000 (16:54 -0700)]
prog_to_nir: OPCODE_EXP is not nir_op_fexp

It's a weird thing that provides some values related to 2**x.  It's also
already handled by a case in the switch.

Signed-off-by: Ian Romanick <ian.d.romanick@intel.com>
Reviewed-by: Matt Turner <mattst88@gmail.com>
Acked-by: Jason Ekstrand <jason.ekstrand@intel.com>
9 years agoi965/fs: Improve a comment about stripping trailing zeroes
Neil Roberts [Fri, 8 May 2015 15:13:52 +0000 (16:13 +0100)]
i965/fs: Improve a comment about stripping trailing zeroes

Originally I wrote that removing the first parameter doesn't work but
I didn't know why. I now found a mention of this in the PRM so it's
probably worthing adding it to the comment.

9 years agodocs: Update the ARB_direct_state_access status
Fredrik Höglund [Wed, 18 Mar 2015 19:50:06 +0000 (20:50 +0100)]
docs: Update the ARB_direct_state_access status

Reviewed-by: Laura Ekstrand <laura@jlekstrand.net>
9 years agomesa: Implement GetVertexArrayIndexed[64]iv
Fredrik Höglund [Mon, 2 Mar 2015 18:07:52 +0000 (19:07 +0100)]
mesa: Implement GetVertexArrayIndexed[64]iv

v2: Fix the name of the entry point in the error messages.

Reviewed-by: Laura Ekstrand <laura@jlekstrand.net>
9 years agomesa: Add support for querying GL_VERTEX_ATTRIB_ARRAY_LONG
Fredrik Höglund [Mon, 2 Mar 2015 18:04:16 +0000 (19:04 +0100)]
mesa: Add support for querying GL_VERTEX_ATTRIB_ARRAY_LONG

This parameter was added in OpenGL 4.3 and GL_ARB_direct_state_access.

Reviewed-by: Laura Ekstrand <laura@jlekstrand.net>
9 years agomesa: Add a vao parameter to get_vertex_array_attrib
Fredrik Höglund [Mon, 2 Mar 2015 17:58:36 +0000 (18:58 +0100)]
mesa: Add a vao parameter to get_vertex_array_attrib

This is needed to implement glGetVertexArrayIndexediv and
glGetVertexArrayIndexed64iv.

v2: Make the vao parameter const.

Reviewed-by: Laura Ekstrand <laura@jlekstrand.net>
9 years agomesa: Implement GetVertexArrayiv
Fredrik Höglund [Mon, 2 Mar 2015 17:56:31 +0000 (18:56 +0100)]
mesa: Implement GetVertexArrayiv

Reviewed-by: Laura Ekstrand <laura@jlekstrand.net>
9 years agomesa: Implement VertexArrayBindingDivisor
Fredrik Höglund [Mon, 2 Mar 2015 17:52:36 +0000 (18:52 +0100)]
mesa: Implement VertexArrayBindingDivisor

Reviewed-by: Laura Ekstrand <laura@jlekstrand.net>
9 years agomesa: Add a vao parameter to vertex_binding_divisor
Fredrik Höglund [Mon, 2 Mar 2015 17:50:49 +0000 (18:50 +0100)]
mesa: Add a vao parameter to vertex_binding_divisor

This is needed to implement VertexArrayBindingDivisor.

Reviewed-by: Laura Ekstrand <laura@jlekstrand.net>
9 years agomesa: Implement VertexArrayAttribBinding
Fredrik Höglund [Mon, 2 Mar 2015 17:49:06 +0000 (18:49 +0100)]
mesa: Implement VertexArrayAttribBinding

Reviewed-by: Laura Ekstrand <laura@jlekstrand.net>
9 years agomesa: Add a vao parameter to vertex_attrib_binding
Fredrik Höglund [Mon, 2 Mar 2015 17:46:42 +0000 (18:46 +0100)]
mesa: Add a vao parameter to vertex_attrib_binding

This is needed to implement VertexArrayAttribBinding.

Reviewed-by: Laura Ekstrand <laura@jlekstrand.net>
9 years agomesa: Implement VertexArrayAttrib[I|L]Format
Fredrik Höglund [Mon, 2 Mar 2015 17:44:00 +0000 (18:44 +0100)]
mesa: Implement VertexArrayAttrib[I|L]Format

Reviewed-by: Laura Ekstrand <laura@jlekstrand.net>
9 years agomesa: Add a vao parameter to update_array_format
Fredrik Höglund [Mon, 2 Mar 2015 17:41:09 +0000 (18:41 +0100)]
mesa: Add a vao parameter to update_array_format

This is needed to implement VertexArrayAttrib*Format.

Reviewed-by: Laura Ekstrand <laura@jlekstrand.net>
9 years agomesa: Refactor VertexAttrib[I|L]Format
Fredrik Höglund [Mon, 2 Mar 2015 17:39:50 +0000 (18:39 +0100)]
mesa: Refactor VertexAttrib[I|L]Format

The only difference between these functions is the legal types and
sizes, so consolidate the code into a single vertex_attrib_format()
function and call it from all three entry points.

Reviewed-by: Laura Ekstrand <laura@jlekstrand.net>
9 years agomesa: Implement VertexArrayVertexBuffers
Fredrik Höglund [Mon, 2 Mar 2015 17:37:27 +0000 (18:37 +0100)]
mesa: Implement VertexArrayVertexBuffers

Reviewed-by: Laura Ekstrand <laura@jlekstrand.net>
9 years agomesa: Implement VertexArrayVertexBuffer
Fredrik Höglund [Mon, 2 Mar 2015 17:35:10 +0000 (18:35 +0100)]
mesa: Implement VertexArrayVertexBuffer

Reviewed-by: Laura Ekstrand <laura@jlekstrand.net>
9 years agomesa: Add a vao parameter to bind_vertex_buffer
Fredrik Höglund [Mon, 2 Mar 2015 17:33:14 +0000 (18:33 +0100)]
mesa: Add a vao parameter to bind_vertex_buffer

This is needed to implement VertexArrayVertexBuffer and
VertexArrayVertexBuffers.

Reviewed-by: Laura Ekstrand <laura@jlekstrand.net>
9 years agomesa: Implement VertexArrayElementBuffer
Fredrik Höglund [Mon, 2 Mar 2015 17:30:12 +0000 (18:30 +0100)]
mesa: Implement VertexArrayElementBuffer

v2: Add a doxygen comment.

Reviewed-by: Laura Ekstrand <laura@jlekstrand.net>
9 years agomesa: Implement EnableVertexArrayAttrib
Fredrik Höglund [Mon, 2 Mar 2015 17:27:58 +0000 (18:27 +0100)]
mesa: Implement EnableVertexArrayAttrib

Reviewed-by: Laura Ekstrand <laura@jlekstrand.net>
9 years agomesa: Implement DisableVertexArrayAttrib
Fredrik Höglund [Mon, 2 Mar 2015 17:27:18 +0000 (18:27 +0100)]
mesa: Implement DisableVertexArrayAttrib

Reviewed-by: Laura Ekstrand <laura@jlekstrand.net>
9 years agomesa: Keep track of the last looked-up VAO
Fredrik Höglund [Mon, 2 Mar 2015 17:25:45 +0000 (18:25 +0100)]
mesa: Keep track of the last looked-up VAO

This saves the cost of repeated hash table lookups when the same
vertex array object is referenced in a sequence of calls such as:

    glVertexArrayAttribFormat(vao, ...);
    glVertexArrayAttribBinding(vao, ...);
    glEnableVertexArrayAttrib(vao, ...);
    ...

Note that VAO's are container objects that are not shared between
contexts.

Reviewed-by: Laura Ekstrand <laura@jlekstrand.net>
9 years agomesa: Add _mesa_lookup_vao_err
Fredrik Höglund [Mon, 2 Mar 2015 17:24:36 +0000 (18:24 +0100)]
mesa: Add _mesa_lookup_vao_err

This is a convenience function that generates GL_INVALID_OPERATION
when the array object doesn't exist.

Reviewed-by: Laura Ekstrand <laura@jlekstrand.net>
9 years agomesa: Implement CreateVertexArrays
Fredrik Höglund [Mon, 2 Mar 2015 17:22:50 +0000 (18:22 +0100)]
mesa: Implement CreateVertexArrays

v2: Update the documentation for gen_vertex_arrays().

Reviewed-by: Laura Ekstrand <laura@jlekstrand.net>
9 years agoi965/skl: In opt_sampler_eot always set destination register to null
Neil Roberts [Thu, 7 May 2015 13:20:17 +0000 (14:20 +0100)]
i965/skl: In opt_sampler_eot always set destination register to null

opt_sampler_eot enables a direct write to framebuffer from a sample.
In order to do this the sample message needs to have a message header
so if there wasn't one already then the function adds one. In addition
the function sets the destination register to null because it's no
longer used. However it was only doing this in cases where it was
adding a message header. This patch just moves setting the destination
so that it happens even if there's a messge header. In practice this
doesn't seem to make any difference but it's a bit cleaner.

Reviewed-by: Anuj Phogat <anuj.phogat@gmail.com>
9 years agoi965/fs: Set the header_size on LOAD_PAYLOAD in opt_sampler_eot
Neil Roberts [Thu, 7 May 2015 17:55:55 +0000 (18:55 +0100)]
i965/fs: Set the header_size on LOAD_PAYLOAD in opt_sampler_eot

Commit 94ee908448 added a header size parameter to the function to
create the LOAD_PAYLOAD instruction. However this broke
opt_sampler_eot which manually constructs the instruction and so
wasn't setting the header_size. This ends up making the parameters for
the send message all have the wrong location and it all falls apart.

Reviewed-by: Jason Ekstrand <jason.ekstrand@intel.com>
Reviewed-by: Anuj Phogat <anuj.phogat@gmail.com>
9 years agodocs: document the LIBGL_DRI3_DISABLE environment variable
Martin Peres [Thu, 7 May 2015 13:57:48 +0000 (16:57 +0300)]
docs: document the LIBGL_DRI3_DISABLE environment variable

Suggested-by: Axel Davy <axel.davy@ens.fr>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Signed-off-by: Martin Peres <martin.peres@linux.intel.com>
9 years agodocs: update ARB_vertex_attrib_64bit status
Dave Airlie [Fri, 24 Apr 2015 02:50:21 +0000 (12:50 +1000)]
docs: update ARB_vertex_attrib_64bit status

Add to GL3.txt and release notes.

Signed-off-by: Dave Airlie <airlied@redhat.com>
9 years agost/mesa: add double input support including lowering (v3.1)
Dave Airlie [Fri, 20 Feb 2015 01:42:19 +0000 (11:42 +1000)]
st/mesa:  add double input support including lowering (v3.1)

This takes a different approach to previously, we cannot index into the
inputMapping with anything but the mesa attribute index, so we can't use
the just add one to index trick, we need more info to add one to it
after we've mapped the input.

(Fixed copy propgation and cleaned up a little)

v2: drop float64 format check, just attr->Doubles.
merge enable patch.
v3: cleanup code a bit.
v3.1: minor review fixups (comment, newline) (Ilia)

Reviewed-by: Ilia Mirkin <imirkin@alum.mit.edu>
Signed-off-by: Dave Airlie <airlied@redhat.com>
9 years agomesa/vbo: add support for 64-bit vertex attributes. (v1)
Dave Airlie [Fri, 20 Feb 2015 01:41:01 +0000 (11:41 +1000)]
mesa/vbo: add support for 64-bit vertex attributes. (v1)

This adds support in the vbo and array code to handle
double vertex attributes.

v0.2: merge code to handle doubles in vbo layer.
v1: don't use v0, merge api_array elt code.

Acked-by: Ilia Mirkin <imirkin@alum.mit.edu>
Reviewed-by: Ian Romanick <ian.d.romanick@intel.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>
9 years agoglsl: check total count of multi-slot double vertex attribs
Dave Airlie [Thu, 30 Apr 2015 00:42:06 +0000 (10:42 +1000)]
glsl: check total count of multi-slot double vertex attribs

The spec is vague all over the place about this, but this seems
to be the intent, we can probably make this optional later if
someone makes hw that cares and writes a driver.

Basically we need to double count some of the d types but
only for totalling not for slot number assignment.

Reviewed-by: Ilia Mirkin <imirkin@alum.mit.edu>
Signed-off-by: Dave Airlie <airlied@redhat.com>
9 years agoglsl: track which program inputs are doubles
Dave Airlie [Wed, 8 Apr 2015 04:38:19 +0000 (14:38 +1000)]
glsl: track which program inputs are doubles

instead of doing the attempts at dual slot handling here,
let the backend do it.

Reviewed-by: Ilia Mirkin <imirkin@alum.mit.edu>
Signed-off-by: Dave Airlie <airlied@redhat.com>
9 years agoglsl: add ARB_vertex_attrib_64bit support. (v2)
Dave Airlie [Fri, 20 Feb 2015 01:38:12 +0000 (11:38 +1000)]
glsl: add ARB_vertex_attrib_64bit support. (v2)

Just more boilerplate stuff.

v2:
bad fallthrough on versioning,
this is my ugly but self contained solution (Ian)

Reviewed-by: Ilia Mirkin <imirkin@alum.mit.edu>
Signed-off-by: Dave Airlie <airlied@redhat.com>
9 years agomesa: add ARB_vertex_attrib_64bit to extensions. (v2)
Dave Airlie [Fri, 20 Feb 2015 01:32:55 +0000 (11:32 +1000)]
mesa: add ARB_vertex_attrib_64bit to extensions. (v2)

Just add the boilerplate bits.

v2: add to version.c

Reviewed-by: Ilia Mirkin <imirkin@alum.mit.edu>
Reviewed-by: Ian Romanick <ian.d.romanick@intel.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>
9 years agomapi: add GL_ARB_vertex_attrib_64bit support
Dave Airlie [Fri, 20 Feb 2015 01:26:33 +0000 (11:26 +1000)]
mapi: add GL_ARB_vertex_attrib_64bit support

This just adds the glapi bits.

Reviewed-by: Ilia Mirkin <imirkin@alum.mit.edu>
Reviewed-by: Ian Romanick <ian.d.romanick@intel.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>
9 years agost/glsl_to_tgsi: fix ir_assignment hack doing bad things for doubles
Dave Airlie [Fri, 20 Feb 2015 03:58:18 +0000 (13:58 +1000)]
st/glsl_to_tgsi: fix ir_assignment hack doing bad things for doubles

This hack for fixing gl_FragDepth apparantly caused a GLSL shader
outputting a single double to try and output a dvec4, but we hadn't
assigned outputs for the secondary bit.

This avoids going into the hack code for scalar doubles.

Reviewed-by: Ilia Mirkin <imirkin@alum.mit.edu>
Signed-off-by: Dave Airlie <airlied@redhat.com>
9 years agoi965/wm/gen6: Add option for disabling statistics collection
Topi Pohjolainen [Thu, 7 May 2015 13:07:15 +0000 (16:07 +0300)]
i965/wm/gen6: Add option for disabling statistics collection

Normally this is always needed but for internal blits and clears
we need to be able to disable it.

Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Signed-off-by: Topi Pohjolainen <topi.pohjolainen@intel.com>
9 years agoi965/wm/gen6: Refactor state setup
Topi Pohjolainen [Wed, 6 May 2015 17:31:49 +0000 (20:31 +0300)]
i965/wm/gen6: Refactor state setup

Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Signed-off-by: Topi Pohjolainen <topi.pohjolainen@intel.com>
9 years agoi965: Remove unused variables
Anuj Phogat [Wed, 6 May 2015 22:11:42 +0000 (15:11 -0700)]
i965: Remove unused variables

Signed-off-by: Anuj Phogat <anuj.phogat@gmail.com>
Reviewed-by: Jordan Justen <jordan.l.justen@intel.com>
9 years agoi965: Change the order of conditions tested in if
Anuj Phogat [Wed, 15 Apr 2015 18:46:53 +0000 (11:46 -0700)]
i965: Change the order of conditions tested in if

Reduces the number of conditions tested in if to one in case of
non-integer formats. Makes no functional changes.

Signed-off-by: Anuj Phogat <anuj.phogat@gmail.com>
Reviewed-by: Matt Turner <mattst88@gmail.com>
9 years agonir: Allow feq/fne/ieq/ine to be optimized with inot.
Matt Turner [Wed, 6 May 2015 05:54:59 +0000 (22:54 -0700)]
nir: Allow feq/fne/ieq/ine to be optimized with inot.

instructions in affected programs:     380 -> 376 (-1.05%)
helped:                                2

Reviewed-by: Jordan Justen <jordan.l.justen@intel.com>
Reviewed-by: Glenn Kennard <glenn.kennard@gmail.com>
9 years agonir: Recognize (a < c || b < c) as min(a, b) < c.
Matt Turner [Wed, 6 May 2015 03:25:07 +0000 (20:25 -0700)]
nir: Recognize (a < c || b < c) as min(a, b) < c.

... and (a >= c) || (b >= c) as max(a, b) >= c.

Similar to commit 97e6c1b9.

total instructions in shared programs: 6182276 -> 6182180 (-0.00%)
instructions in affected programs:     6400 -> 6304 (-1.50%)
helped:                                68
HURT:                                  4

Reviewed-by: Jordan Justen <jordan.l.justen@intel.com>
Reviewed-by: Glenn Kennard <glenn.kennard@gmail.com>
9 years agonir: Recognize trivial min/max.
Matt Turner [Wed, 6 May 2015 03:20:30 +0000 (20:20 -0700)]
nir: Recognize trivial min/max.

No changes, but does prevent some regressions in the next commit.

Reviewed-by: Jordan Justen <jordan.l.justen@intel.com>
Reviewed-by: Glenn Kennard <glenn.kennard@gmail.com>
9 years agonir: Recognize i2b(b2i(x)) as x.
Matt Turner [Wed, 6 May 2015 02:56:27 +0000 (19:56 -0700)]
nir: Recognize i2b(b2i(x)) as x.

Helps the same set of programs as the previous commit.

instructions in affected programs:     4490 -> 4346 (-3.21%)
helped:                                8

Reviewed-by: Jordan Justen <jordan.l.justen@intel.com>
Reviewed-by: Glenn Kennard <glenn.kennard@gmail.com>
9 years agonir: Recognize imul(b2i(a), b2i(b)) as a logical AND.
Matt Turner [Wed, 6 May 2015 02:50:58 +0000 (19:50 -0700)]
nir: Recognize imul(b2i(a), b2i(b)) as a logical AND.

Four shaders in Unreal 4's Sun Temple are helped, and gain SIMD16
because we avoid an integer multiplication.

instructions in affected programs:     2353 -> 2245 (-4.59%)
helped:                                4
GAINED:                                4

Reviewed-by: Jordan Justen <jordan.l.justen@intel.com>
Reviewed-by: Glenn Kennard <glenn.kennard@gmail.com>
9 years agoi965/sync: Implement DRI2_Fence extension
Chad Versace [Wed, 6 May 2015 02:05:32 +0000 (19:05 -0700)]
i965/sync: Implement DRI2_Fence extension

This enables EGL_KHR_fence_sync and EGL_KHR_wait_sync.

Below is the difference in piglit results, before and after this patch.
No regressions and several tests improve from 'skip' to 'pass'. Out of
EGL_KHR_fence_sync tests, two of the multithreaded tests skip; all other
tests pass.

  cmdline: piglit run -p gbm -t sync tests/quick.py
  mesa: master@1ac7db0
  piglit: 4069bec
  hw: Ivybridge

        | before after
  ------+-------------
   pass |     32    46
   fail |      0     0
  crash |      0     0
   skip |     35    21
  total |     67    67

v2:
  - Set fence->signalled = true in brw_fence_has_completed() too.

Reviewed-by: Daniel Stone <daniels@collabora.com>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
9 years agoi965/sync: Replace prefix 'intel_sync' -> 'intel_gl_sync'
Chad Versace [Wed, 6 May 2015 02:05:32 +0000 (19:05 -0700)]
i965/sync: Replace prefix 'intel_sync' -> 'intel_gl_sync'

I'm about to implement DRI2_Fenc in intel_syncobj.c.  To prevent
madness, we need to prefix functions for GL_ARB_sync with 'gl' and
functions for DRI2_Fence with 'dri'. Otherwise, the file will become
a jumble of similiarly named functions.

For example:
    old-name:      intel_client_wait_sync()
    new-name:      intel_gl_client_wait_sync()
    soon-to-come:  intel_dri_client_wait_sync()

I wrote this renaming commit separately from the commit that implements
DRI2_Fence because I wanted the latter diff to be reviewable.

Reviewed-by: Daniel Stone <daniels@collabora.com>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
9 years agoi915/sync: Return early when calloc fails
Chad Versace [Wed, 6 May 2015 02:05:31 +0000 (19:05 -0700)]
i915/sync: Return early when calloc fails

Reviewed-by: Daniel Stone <daniels@collabora.com>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
9 years agoi965/sync: Return NULL when calloc fails
Chad Versace [Wed, 6 May 2015 02:05:30 +0000 (19:05 -0700)]
i965/sync: Return NULL when calloc fails

Reviewed-by: Daniel Stone <daniels@collabora.com>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
9 years agoi915/sync: Don't crash when deleting sync object
Chad Versace [Wed, 6 May 2015 02:05:29 +0000 (19:05 -0700)]
i915/sync: Don't crash when deleting sync object

Don't pass NULL to drm_intel_bo_unreference(). It doesn't like that.

Bug found by code inspection.

Reviewed-by: Daniel Stone <daniels@collabora.com>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
9 years agoi965/sync: Don't crash when deleting sync object
Chad Versace [Wed, 6 May 2015 02:05:28 +0000 (19:05 -0700)]
i965/sync: Don't crash when deleting sync object

Don't pass NULL to drm_intel_bo_unreference(). It doesn't like that.

Bug found by code inspection.

Reviewed-by: Daniel Stone <daniels@collabora.com>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
9 years agoegl/dri2: Fix codestyle in a comment
Chad Versace [Thu, 7 May 2015 15:09:07 +0000 (08:09 -0700)]
egl/dri2: Fix codestyle in a comment

Pointed out by Kenneth Graunke. Trivial fix.

9 years agoglx: report which DRI version is used when in verbose debug mode
Martin Peres [Wed, 6 May 2015 10:31:30 +0000 (13:31 +0300)]
glx: report which DRI version is used when in verbose debug mode

This should make it more obvious in bug reports while also removing
any sort of guesswork for developers.

Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Signed-off-by: Martin Peres <martin.peres@linux.intel.com>
9 years agoglapi: Add positional argument specifier.
Vinson Lee [Wed, 6 May 2015 19:39:09 +0000 (12:39 -0700)]
glapi: Add positional argument specifier.

Fix build error introduced with commit 1c5a57a "glapi/es3.1: Add support
for GLES versions > 3.0" with Python < 2.7.

  File "src/mapi/glapi/gen/gl_genexec.py", line 230, in <module>
    printer.Print(api)
  File "src/mapi/glapi/gen/gl_XML.py", line 120, in Print
    self.printBody(api)
  File "src/mapi/glapi/gen/gl_genexec.py", line 187, in printBody
    condition_parts.append('(ctx->API == API_OPENGLES2 && ctx->Version >= {})'.format(int(f.api_map['es2'] * 10)))
ValueError: zero length field name in format

Signed-off-by: Vinson Lee <vlee@freedesktop.org>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Reviewed-by: Ian Romanick <ian.d.romanick@intel.com>
9 years agonv50/ir: add SHL to the list of U32 opcodes
Ilia Mirkin [Thu, 7 May 2015 00:48:40 +0000 (20:48 -0400)]
nv50/ir: add SHL to the list of U32 opcodes

Having the wrong inferred type prevents a number of optimizations,
including constant propagation (since float immediates work differently
than integer immediates).

Signed-off-by: Ilia Mirkin <imirkin@alum.mit.edu>
9 years agoi965: Sort extension enable lists
Ian Romanick [Tue, 21 Apr 2015 04:43:32 +0000 (13:43 +0900)]
i965: Sort extension enable lists

Sort by GEN, then sort by extension name.

Signed-off-by: Ian Romanick <ian.d.romanick@intel.com>
Reviewed-by: Matt Turner <mattst88@gmail.com>
9 years agor600g: Fix Clang return-type build error.
Vinson Lee [Wed, 6 May 2015 17:09:38 +0000 (10:09 -0700)]
r600g: Fix Clang return-type build error.

Fix Clang return-type error introduced with commit
96f164f6f047833091eb98a73aa80c31dc94f962 "gallium: make
pipe_context::begin_query return a boolean".

  CC       r600_query.lo
r600_query.c:443:3: error: non-void function 'r600_begin_query' should return a value [-Wreturn-type]
                return;
                ^

Signed-off-by: Vinson Lee <vlee@freedesktop.org>
Reviewed-by: Tom Stellard <thomas.stellard@amd.com>
9 years agoi965/fs: Allow copy propagation on ATTR file registers.
Kenneth Graunke [Tue, 10 Mar 2015 11:18:06 +0000 (04:18 -0700)]
i965/fs: Allow copy propagation on ATTR file registers.

This especially helps with NIR because we currently emit MOVs at the top
of the shader to copy from various ATTR registers to a giant VGRF array
of all inputs.  (This could potentially be done better, but since
there's only ever one write to each register, it should be trivial to
copy propagate away...)

With NIR - only vertex shaders:
total instructions in shared programs: 3129373 -> 2889581 (-7.66%)
instructions in affected programs:     3119717 -> 2879925 (-7.69%)
helped:                                20833

Without NIR - only vertex shaders:
total instructions in shared programs: 2745901 -> 2724483 (-0.78%)
instructions in affected programs:     693426 -> 672008 (-3.09%)
helped:                                3516

Signed-off-by: Kenneth Graunke <kenneth@whitecape.org>
Reviewed-by: Matt Turner <mattst88@gmail.com>
Reviewed-by: Jason Ekstrand <jason.ekstrand@intel.com>
9 years agoi965/fs_inst: Get rid of the effective_width field
Jason Ekstrand [Thu, 2 Apr 2015 01:15:42 +0000 (18:15 -0700)]
i965/fs_inst: Get rid of the effective_width field

The effective_width field was an ill-concieved hack to get around issues in
the LOAD_PAYLOAD instruction.  Now that the LOAD_PAYLOAD instruction is far
more sane, this field can die.

Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
9 years agoi965/fs: Rework the fs_visitor LOAD_PAYLOAD instruction
Jason Ekstrand [Wed, 25 Mar 2015 00:00:04 +0000 (17:00 -0700)]
i965/fs: Rework the fs_visitor LOAD_PAYLOAD instruction

The newly reworked instruction is far more straightforward than the
original.  Before, the LOAD_PAYLOAD instruction was lowered by a the
complicated and broken-by-design pile of heuristics to try and guess
force_writemask_all, exec_size, and a number of other factors on the
sources.

Instead, we use the header_size on the instruction to denote which sources
are "header sources".  Header sources are required to be a single physical
hardware register that is copied verbatim.  The registers that follow are
considered the actual payload registers and have a width that correspond's
to the LOAD_PAYLOAD's exec_size and are treated as being per-channel.  This
gives us a fairly straightforward lowering:

 1) All header sources are copied directly using force_writemask_all and,
    since they are guaranteed to be a single register, there are no
    force_sechalf issues.

 2) All non-header sources are copied using the exact same force_sechalf
    and force_writemask_all modifiers as the LOAD_PAYLOAD operation itself.

 3) In order to accommodate older gens that need interleaved colors,
    lower_load_payload detects when the destination is a COMPR4 register
    and automatically interleaves the non-header sources.  The
    lower_load_payload pass does the right thing here regardless of whether
    or not the hardware actually supports COMPR4.

This patch commit itself is made up of a bunch of smaller changes squashed
together.  Individual change descriptions follow:

i965/fs: Rework fs_visitor::LOAD_PAYLOAD

   We rework LOAD_PAYLOAD to verify that all of the sources that count as
   headers are, indeed, exactly one register and that all of the non-header
   sources match the destination width.  We then take the exec_size for
   LOAD_PAYLOAD directly from the destination width.

i965/fs: Make destinations of load_payload have the appropreate width

i965/fs: Rework fs_visitor::lower_load_payload

   v2: Don't allow the saturate flag on LOAD_PAYLOAD instructions

i965/fs_cse: Support the new-style LOAD_PAYLOAD

i965/fs_inst::is_copy_payload: Support the new-style LOAD_PAYLOAD

i965/fs: Simplify setup_color_payload

   Previously, setup_color_payload was a a big helper function that did a
   lot of gen-specific special casing for setting up the color sources of
   the LOAD_PAYLOAD instruction.  Now that lower_load_payload is much more
   sane, most of that complexity isn't needed anymore.  Instead, we can do
   a simple fixup pass for color clamps and then just stash sources
   directly in the LOAD_PAYLOAD.  We can trust lower_load_payload to do the
   right thing with respect to COMPR4.

Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
9 years agoi965/fs: Make LOAD_PAYLOAD take a header size
Jason Ekstrand [Tue, 24 Mar 2015 17:37:33 +0000 (10:37 -0700)]
i965/fs: Make LOAD_PAYLOAD take a header size

Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
9 years agoi965/fs: Make emit_single_fb_write take an explicit exec_size
Jason Ekstrand [Tue, 24 Mar 2015 20:55:10 +0000 (13:55 -0700)]
i965/fs: Make emit_single_fb_write take an explicit exec_size

Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
9 years agoi965/fs_inst: Add an is_copy_payload helper
Jason Ekstrand [Wed, 1 Apr 2015 22:38:23 +0000 (15:38 -0700)]
i965/fs_inst: Add an is_copy_payload helper

This commit adds a new is_copy_payload helper to fs_inst that takes the
place of the similarly named functions in cse and register coalesce.  The
two is_copy_payload functions in CSE and register coalesce were subtly
different and potentially subtly broken.  The new version unifies the two
and should be more correct.

Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
9 years agoi965: Change header_present to header_size in backend_instruction
Jason Ekstrand [Tue, 24 Mar 2015 17:17:32 +0000 (10:17 -0700)]
i965: Change header_present to header_size in backend_instruction

Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
9 years agoi965/fs_cse: Factor out code to create copy instructions
Jason Ekstrand [Tue, 24 Mar 2015 22:06:24 +0000 (15:06 -0700)]
i965/fs_cse: Factor out code to create copy instructions

v2: Get rid of the block parameter and make src a const reference

Reviewed-by: Topi Pohjolainen <topi.pohjolainen@intel.com>
Reviewed-by: Matt Turner <mattst88@gmail.com>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
9 years agoi965/fs: Make half(fs_reg, unsigned) handle register files more explicitly
Jason Ekstrand [Tue, 5 May 2015 22:57:11 +0000 (15:57 -0700)]
i965/fs: Make half(fs_reg, unsigned) handle register files more explicitly

Previously, we had a special case for uniforms and immediates and then a
bunch of asserts for various other pessimal things.  This commit changes it
so that it explicitly does something on each register file.  Some of them
are disallowed and others are treated properly.

Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
9 years agoi965/fs: Fix passing an immediate to half().
Francisco Jerez [Wed, 22 Apr 2015 12:01:24 +0000 (15:01 +0300)]
i965/fs: Fix passing an immediate to half().

Immediates are generally uniform, they yield the same value to both
halves of any instruction.

Reviewed-by: Matt Turner <mattst88@gmail.com>
9 years agoswrast: Build fix for darwin
Jeremy Huddleston Sequoia [Fri, 2 Jan 2015 03:48:40 +0000 (19:48 -0800)]
swrast: Build fix for darwin

Fixes regression from commit 64b1dc44495890cbc2c7c5509cb830264020998c

Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=90147
Signed-off-by: Jeremy Huddleston Sequoia <jeremyhu@apple.com>
Reviewed-by: Brian Paul <brianp@vmware.com>
CC: Emil Velikov <emil.l.velikov@gmail.com>
CC: jon.turney@dronecode.org.uk
CC: ionic@macports.org
9 years agoegl/dri2: Check return value of __DRI2fence::create_fence()
Chad Versace [Wed, 6 May 2015 02:05:20 +0000 (19:05 -0700)]
egl/dri2: Check return value of __DRI2fence::create_fence()

If it returns NULL, then return early with an error.

Reviewed-by: Marek Olšák <marek.olsak@amd.com>