LoopDawg [Mon, 31 Jul 2017 19:41:42 +0000 (13:41 -0600)]
HLSL: add methods to track user structure in texture return type.
Some languages allow a restricted set of user structure types returned from texture sampling
operations. Restrictions include the total vector size of all components may not exceed 4,
and the basic types of all members must be identical.
This adds underpinnings for that ability. Because storing a whole TType or even a simple
TTypeList in the TSampler would be expensive, the structure definition is held in a
table outside the TType. The TSampler contains a small bitfield index, currently 4 bits
to support up to 15 separate texture template structure types, but that can be adjusted
up or down. Vector returns are handled as before.
There are abstraction methods accepting and returning a TType (such as may have been parsed
from a grammar). The new methods will accept a texture template type and set the
sampler to the structure if possible, checking a range of error conditions such as whether
the total structure vector components exceed 4, or whether their basic types differe, or
whether the struct contains non-vector-or-scalar members. Another query returns the
appropriate TType for the sampler.
High level summary of design:
In the TSampler, this holds an index into the texture structure return type table:
unsigned int structReturnIndex : structReturnIndexBits;
These are the methods to set or get the return type from the TSampler. They work for vector or structure returns, and potentially could be expanded to handle other things (small arrays?) if ever needed.
bool setTextureReturnType(TSampler& sampler, const TType& retType, const TSourceLoc& loc);
void getTextureReturnType(const TSampler& sampler, const TType& retType, const TSourceLoc& loc) const;
The ``convertReturn`` lambda in ``HlslParseContext::decomposeSampleMethods`` is greatly expanded to know how to copy a vec4 sample return to whatever the structure type should be. This is a little awkward since it involves introducing a comma expression to return the proper aggregate value after a set of memberwise copies.
John Kessenich [Tue, 15 Aug 2017 16:18:32 +0000 (10:18 -0600)]
HLSL: Add fall-back for opaque initializers to just generate long-term expected code.
This generated code needs an optimization pass to eliminate the assignments
to the opaque members.
John Kessenich [Tue, 15 Aug 2017 04:16:36 +0000 (13:16 +0900)]
Merge pull request #1013 from KhronosGroup/flatten-nonarrayed
HLSL: Flatten structs for all non-arrayed I/O interfaces.
John Kessenich [Tue, 15 Aug 2017 04:10:00 +0000 (22:10 -0600)]
SPV: Correct selection of storage-image capabilities. Fixes #986.
Code was reflecting an old historical use of sampled as a SPIR-V
2-valued operand, instead of its current 3 values.
John Kessenich [Fri, 11 Aug 2017 06:17:26 +0000 (00:17 -0600)]
HLSL: Flatten structs for all non-arrayed I/O interfaces.
John Kessenich [Fri, 11 Aug 2017 06:14:46 +0000 (00:14 -0600)]
HLSL: Make fresh array sizes for TessLevelOuter and TessLevelInner arrays.
This prevents potentional sharing from inadvertently affecting other arrays.
John Kessenich [Thu, 10 Aug 2017 17:41:11 +0000 (11:41 -0600)]
HLSL Test: Expand test, adding a user-patch-constant signature.
John Kessenich [Wed, 9 Aug 2017 20:29:29 +0000 (14:29 -0600)]
HLSL: Move debug naming to a simpler, more consistent, scheme.
This will help in expanding flattening and reducing splitting.
John Kessenich [Wed, 9 Aug 2017 02:02:21 +0000 (20:02 -0600)]
HLSL: Include built-in processing for vertex input and fragment output flattening.
John Kessenich [Tue, 8 Aug 2017 05:40:05 +0000 (23:40 -0600)]
Non-functional HLSL: Factor out built-ins from splitting and related simplifications.
John Kessenich [Mon, 7 Aug 2017 21:35:29 +0000 (06:35 +0900)]
Merge pull request #1011 from LoopDawg/pragma-pack-matrix
HLSL: implement #pragma pack_matrix(layout)
LoopDawg [Mon, 7 Aug 2017 18:08:50 +0000 (12:08 -0600)]
HLSL: implement #pragma pack_matrix(layout)
This adds support for #pragma pack_matrix() to the HLSL front end.
The pragma sets the default matrix layout for subsequent unqualified matrices
in structs or buffers. Explicit qualification overrides the pragma value. Matrix
layout is not permitted at the structure level in HLSL, so only leaves which are
matrix types can be so qualified.
Note that due to the semantic (not layout) difference in first matrix indirections
between HLSL and SPIR-V, the sense of row and column major are flipped. That's
independent of this PR: just a factor to note. A column_major qualifier appears
as a RowMajor member decoration in SPIR-V modules, and vice versa.
John Kessenich [Mon, 7 Aug 2017 01:42:42 +0000 (19:42 -0600)]
HLSL: Correct which things flattening tracks for linkage, based on caller, not type.
Includes related trackLinkage() code removal and name improvements.
John Kessenich [Sat, 5 Aug 2017 18:03:05 +0000 (03:03 +0900)]
Merge pull request #1010 from LoopDawg/mat-rowcol-fix-1
HLSL: fix qualifier propagation from user struct types to block defin…
LoopDawg [Fri, 4 Aug 2017 21:40:53 +0000 (15:40 -0600)]
HLSL: fix qualifier propagation from user struct types to block definitions.
The HLSL FE tracks four versions of a declared type to avoid losing information, since it
is not (at type-decl time) known how the type will be used downstream. If such a type
was used in a cbuffer declaration, the cbuffer type's members should have been using
the uniform form of the original user structure type, but were not.
This would manifest as matrix qualifiers (and other things, such as pack offsets) on user struct
members going missing in the SPIR-V module if the struct type was a member of a cbuffer, like so:
struct MyBuffer
{
row_major float4x4 mat1;
column_major float4x4 mat2;
};
cbuffer Example
{
MyBuffer g_MyBuffer;
};
Fixes: #789
John Kessenich [Fri, 4 Aug 2017 19:51:54 +0000 (13:51 -0600)]
HLSL: Correct use of isPerVertexBuiltIn() to be isClipOrCullDistance().
This allows removal of isPerVertexBuiltIn(). It also leads to
removal of addInterstageIoToLinkage(), which is no longer needed.
Includes related name improvements.
John Kessenich [Fri, 4 Aug 2017 19:41:32 +0000 (13:41 -0600)]
HLSL: Test change: Geometry shaders can't return values from main.
John Kessenich [Fri, 4 Aug 2017 18:04:44 +0000 (12:04 -0600)]
HLSL: Flatten more I/O: non-arrayed user-only structures.
The goal is to flatten all I/O, but there are multiple categories and
steps to complete, likely including a final unification of splitting
and flattening.
John Kessenich [Fri, 4 Aug 2017 00:41:48 +0000 (18:41 -0600)]
HLSL: Stop including empty structures in the I/O interface. Fix #785.
John Kessenich [Wed, 2 Aug 2017 23:08:43 +0000 (17:08 -0600)]
Non-functional: HLSL: Simplify I/O logic for splitting.
John Kessenich [Tue, 1 Aug 2017 17:11:52 +0000 (02:11 +0900)]
Merge pull request #1006 from KhronosGroup/4.60
GLSL: Add GLSL 4.60 features
John Kessenich [Mon, 24 Jul 2017 05:31:44 +0000 (23:31 -0600)]
GLSL 4.6: Implement shader group vote.
John Kessenich [Mon, 24 Jul 2017 05:11:41 +0000 (23:11 -0600)]
GLSL 4.6: Implement draw parameters.
John Kessenich [Sun, 23 Jul 2017 22:08:26 +0000 (16:08 -0600)]
GLSL 4.6: Implement atomic counter ops and SPV_KHR_shader_atomic_counter_ops.
John Kessenich [Sun, 23 Jul 2017 20:44:59 +0000 (14:44 -0600)]
GLSL: Initiate version GLSL 460, including accept extraneous semicolons.
John Kessenich [Mon, 31 Jul 2017 07:19:26 +0000 (01:19 -0600)]
HLSL Tests: Fix two tests to be valid under FXC.
John Kessenich [Mon, 31 Jul 2017 05:39:48 +0000 (23:39 -0600)]
Non-functional: HLSL further simplications to base I/O flattening on.
John Kessenich [Sun, 30 Jul 2017 22:54:02 +0000 (16:54 -0600)]
Non-functional: HLSL: clean up dead code for splitting.
Most of this was obsoleted by entry-point wrapping.
Some other is just unnecessary.
Also, includes some spelling/name improvements.
This is to help lay ground work for flattening user I/O.
John Kessenich [Sat, 29 Jul 2017 21:23:11 +0000 (06:23 +0900)]
Merge pull request #1005 from LoopDawg/remove-unused
HLSL: Non-functional: warning fix + remove unused member
LoopDawg [Sat, 29 Jul 2017 00:41:53 +0000 (18:41 -0600)]
HLSL: Non-functional: warning fix, remove unused member.
Two non-functional changes:
1. Remove flattenLevel, which is unneeded since at or around
d1be7545c6a23657587dfe3bc4b149c91290a66f.
2. Fix build warining about unused variable in executeInitializer.
John Kessenich [Fri, 28 Jul 2017 23:33:58 +0000 (17:33 -0600)]
GLSL reflection: Fix #985: reflect runtime sized arrays having no constant index.
John Kessenich [Fri, 28 Jul 2017 23:32:27 +0000 (17:32 -0600)]
Nonfunctional: Add reflect test case, and fix long lines in reflection code.
John Kessenich [Fri, 28 Jul 2017 22:20:13 +0000 (16:20 -0600)]
Nonfunctional: Shorten some lines to the coding standard, to retrigger failed test run.
John Kessenich [Thu, 27 Jul 2017 16:35:42 +0000 (01:35 +0900)]
Merge pull request #1002 from amdrexu/bugfix
SPV: Memory qualifiers should decorate top-level block members
Rex Xu [Thu, 27 Jul 2017 06:33:16 +0000 (14:33 +0800)]
SPV: Memory qualifiers should decorate top-level block members
John Kessenich [Wed, 26 Jul 2017 23:38:03 +0000 (08:38 +0900)]
Merge pull request #997 from LoopDawg/clipcull-semantic
HLSL: handle multiple clip/cull semantic IDs
LoopDawg [Wed, 5 Jul 2017 17:33:06 +0000 (11:33 -0600)]
HLSL: handle multiple clip/cull semantic IDs
HLSL allows several variables to be declared. There are packing rules involved:
e.g, a float3 and a float1 can be packed into a single array[4], while for a
float3 and another float3, the second one will skip the third array entry to
avoid straddling
This is implements that ability. Because there can be multiple variables involved,
and the final output array will often be a different type altogether (to fuse
the values into a single destination), a new variable is synthesized, unlike the prior
clip/cull support which used the declared variable. The new variable name is
taken from one of the declared ones, so the old tests are unchanged.
Several new tests are added to test various packing scenarios.
Only two semantic IDs are supported: 0, and 1, per HLSL rules. This is
encapsulated in
static const int maxClipCullRegs = 2;
and the algorithm (probably :) ) generalizes to larger values, although there
are a few issues around how HLSL would pack (e.g, would 4 scalars be packed into
a single HLSL float4 out reg? Probably, and this algorithm assumes so).
John Kessenich [Wed, 26 Jul 2017 03:03:29 +0000 (21:03 -0600)]
SPV: Update to latest 1.0 headers, removing redundancies in GLSL.ext.AMD.h.
John Kessenich [Tue, 25 Jul 2017 07:36:17 +0000 (16:36 +0900)]
Merge pull request #1000 from LoopDawg/samplecmpzero-cubearray-fix
Fix dref explicit LOD form of sample with cube texture arrays
LoopDawg [Tue, 25 Jul 2017 00:45:37 +0000 (18:45 -0600)]
Fix dref explicit LOD form of sample with cube texture arrays
The dref parameter was being used as the LOD. Now it it's properly the dref.
John Kessenich [Sun, 23 Jul 2017 19:54:15 +0000 (13:54 -0600)]
GLSL: Implement version 320 for ES.
John Kessenich [Sun, 23 Jul 2017 17:49:42 +0000 (11:49 -0600)]
GLSL: Add version-number checking.
John Kessenich [Fri, 21 Jul 2017 19:36:33 +0000 (13:36 -0600)]
SPV/OpenGL: Require locations on non-opaque uniform variables.
John Kessenich [Fri, 21 Jul 2017 16:59:42 +0000 (01:59 +0900)]
Merge pull request #991 from LoopDawg/resource-set-binding-fix
HLSL: Fix crash with --resource-set-binding [n] (global form, not per-register form)
John Kessenich [Thu, 20 Jul 2017 22:44:17 +0000 (16:44 -0600)]
Reflection: Fix #977: Expose getBinding(), use in new getUniformBinding().
LoopDawg [Fri, 14 Jul 2017 21:15:47 +0000 (15:15 -0600)]
Fix crash with --resource-set-binding [n] (common set form, not per-register form)
--resource-set-binding has a mode which allows per-register assignments of
bindings and descriptor sets on the command line, and another accepting a
single descriptor set value to assign to all variables.
The former worked, but the latter would crash when assigning the values.
This fixes it, and makes the former case a bit more robust against premature
termination of the pre-register values, which must come in (regname,set,binding)
triples.
This also allows the form "--resource-set-binding stage setnum", which was
mentioned in the usage message, but did not parse.
The operation of the per-register form of this option is unchanged.
John Kessenich [Tue, 18 Jul 2017 09:51:02 +0000 (03:51 -0600)]
Merge pull request #996 from KhronosGroup/decorate-parameters
SPV: Decorate parameters
John Kessenich [Tue, 18 Jul 2017 08:58:06 +0000 (02:58 -0600)]
SPV: Fix #995: Include memory decorations on parameters.
John Kessenich [Tue, 18 Jul 2017 08:35:46 +0000 (02:35 -0600)]
SPV: Non-functional: support lists of decorations per parameter.
John Kessenich [Sun, 16 Jul 2017 12:01:38 +0000 (06:01 -0600)]
Merge pull request #994 from KhronosGroup/opaque-init
HLSL: Fix #980: flatten opaque initializers to use aliases.
John Kessenich [Sun, 16 Jul 2017 11:46:13 +0000 (05:46 -0600)]
Fix #980: flatten opaque initializers to use aliases.
John Kessenich [Sun, 16 Jul 2017 07:16:55 +0000 (01:16 -0600)]
Merge pull request #992 from polarina/null-deref-fix
Fix NULL pointer dereference in TParseContext::builtInOpCheck
Gabríel Arthúr Pétursson [Sat, 15 Jul 2017 15:21:12 +0000 (15:21 +0000)]
Fix NULL pointer dereference in TParseContext::builtInOpCheck
John Kessenich [Fri, 14 Jul 2017 11:52:31 +0000 (05:52 -0600)]
SPV: Address #989: Don't add Location to built-ins when automapping.
John Kessenich [Fri, 14 Jul 2017 11:45:50 +0000 (05:45 -0600)]
Merge pull request #987 from KhronosGroup/global-std140
Address part B of #982: $Global will use std140 instead of HLSL offsets.
John Kessenich [Thu, 13 Jul 2017 17:39:16 +0000 (11:39 -0600)]
Address part A of #982: $Global will use std140 instead of HLSL offsets.
From comment about this:
Adjust alignment for HLSL rules
TODO: make this consistent in early phases of code: adjusting this late means inconsistencies with earlier code, which for reflection is an issue.
Until reflection is brought in sync with these adjustments, don't apply to $Global,
which is the most likely to rely on reflection, and least likely to rely
implicit layouts.
John Kessenich [Thu, 13 Jul 2017 14:29:53 +0000 (08:29 -0600)]
Merge pull request #978 from LoopDawg/global-const-init-fix
HLSL: support global const initializers from non-constant rvalues
John Kessenich [Thu, 13 Jul 2017 13:14:26 +0000 (07:14 -0600)]
Merge pull request #983 from LoopDawg/warning-fix
Fix StandAlone.cpp compilation warnings
LoopDawg [Wed, 12 Jul 2017 15:04:39 +0000 (09:04 -0600)]
Fix StandAlone.cpp compilation warnings
find_first_of returns size_t, was overflowing int.
LoopDawg [Mon, 10 Jul 2017 21:43:40 +0000 (15:43 -0600)]
WIP: HLSL: support global const initializers from non-constant rvalues
Semantic test left over from other source languages is removed, since this is permitted by HLSL.
Also, to support the functionality, a targeted test is performed for this case and it is
turned into a EvqGlobal qualifier to create an AST initialization segment when needed.
Constness is now propagated up aggregate chains during initializer construction. This
handles hierarchical cases such as the distinction between:
static const float2 a[2] = { { 1, 2 }, { 3, 4} };
vs
static const float2 a[2] = { { 1, 2 }, { cbuffer_member, 4} };
The first of which can use a first class constant initalization, and the second cannot.
John Kessenich [Fri, 7 Jul 2017 16:22:25 +0000 (10:22 -0600)]
Build: Fix #975: shut up a compiler warning (code was fine).
John Kessenich [Thu, 6 Jul 2017 17:58:12 +0000 (11:58 -0600)]
Nonfunctional, infrastructure: Clean up Test directory droppings.
The multithreaded outputs in particular can be troublesome when searching.
John Kessenich [Thu, 6 Jul 2017 06:44:32 +0000 (00:44 -0600)]
Merge pull request #971 from amdrexu/bugfix
HLSL: Translate directive [flatten] and [branch] to SPV control mask.
Rex Xu [Tue, 4 Jul 2017 15:23:40 +0000 (23:23 +0800)]
HLSL: Translate directive [flatten] and [branch] to SPV control mask.
John Kessenich [Thu, 6 Jul 2017 00:34:15 +0000 (18:34 -0600)]
Tests: Add a test for concurrent use of std430 and push_constant.
John Kessenich [Wed, 5 Jul 2017 20:48:19 +0000 (14:48 -0600)]
Merge pull request #965 from chaoc/spv-khr-post-depth-coverage
Implement SPV_KHR_post_depth_coverage
chaoc [Sat, 1 Jul 2017 00:14:30 +0000 (17:14 -0700)]
Implement SPV_KHR_post_depth_coverage
Added support for both extension GL_ARB_post_depth_coverage and GL_EXT_post_depth_coverage.
John Kessenich [Wed, 5 Jul 2017 19:12:08 +0000 (13:12 -0600)]
Merge pull request #974 from LoopDawg/anyall-types
HLSL: handle type conversion for any/all intrinsics
LoopDawg [Wed, 5 Jul 2017 18:07:16 +0000 (12:07 -0600)]
HLSL: handle type conversion for any/all intrinsics
HLSL allows float/etc types for any/all intrinsics, while the
SPIR-V opcode requires bool. This adds a simple decomposition
to type convert the argument. It could get a little more clever
in some of the type cases if it ever had to.
John Kessenich [Tue, 4 Jul 2017 17:00:42 +0000 (11:00 -0600)]
Merge pull request #969 from d3x0r/patch-1
Add option to skip installation
John Kessenich [Tue, 4 Jul 2017 17:00:23 +0000 (11:00 -0600)]
Merge pull request #970 from rohith10/master
Fix CMake scripts to set the POSITION_INDEPENDENT_CODE property correctly
Rohith Chandran [Tue, 4 Jul 2017 14:53:45 +0000 (10:53 -0400)]
Fix CMake scripts: The set_property script can be used to set only a single property, so now setting the POSITION_INDEPENDENT_CODE property correctly.
d3x0r [Tue, 4 Jul 2017 13:01:09 +0000 (06:01 -0700)]
Update CMakeLists.txt
d3x0r [Tue, 4 Jul 2017 13:00:35 +0000 (06:00 -0700)]
Update CMakeLists.txt
d3x0r [Tue, 4 Jul 2017 13:00:04 +0000 (06:00 -0700)]
Update CMakeLists.txt
d3x0r [Tue, 4 Jul 2017 12:59:34 +0000 (05:59 -0700)]
Update CMakeLists.txt
d3x0r [Tue, 4 Jul 2017 12:59:03 +0000 (05:59 -0700)]
Update CMakeLists.txt
d3x0r [Tue, 4 Jul 2017 12:58:07 +0000 (05:58 -0700)]
Update CMakeLists.txt
d3x0r [Tue, 4 Jul 2017 12:57:47 +0000 (05:57 -0700)]
Update CMakeLists.txt
d3x0r [Tue, 4 Jul 2017 12:57:08 +0000 (05:57 -0700)]
Update CMakeLists.txt
d3x0r [Tue, 4 Jul 2017 12:54:57 +0000 (05:54 -0700)]
Add option to skip installation
As a static target, when included in other cmake projects, it is not needed to install these libraries and headers as part of this build, and just need to link to this.
John Kessenich [Tue, 4 Jul 2017 04:14:09 +0000 (22:14 -0600)]
Merge pull request #967 from dsrbecky/multiview
Multiview extension: Accept layout(num_views) qualifier
John Kessenich [Thu, 29 Jun 2017 23:43:31 +0000 (17:43 -0600)]
HLSL: Non-functional: Move partial flattened access into symbol node.
Lays the groundwork for fixing issue #954.
Partial flattenings were previously tracked through a stack of active subsets
in the parse context, but full functionality needs AST nodes to represent
this across time, removing the need for parsecontext tracking.
John Kessenich [Fri, 30 Jun 2017 00:23:31 +0000 (18:23 -0600)]
HLSL: Non-functional: some coding convention tweaks (120 columns, nullptr).
This commit, and next one, are specifically to make a future commit
handling partial dereferences of flattening objects easier to see.
John Kessenich [Tue, 4 Jul 2017 03:48:58 +0000 (21:48 -0600)]
Merge pull request #966 from TiemoJung/io_remapper_update
io resolver improvements
John Kessenich [Tue, 4 Jul 2017 03:47:39 +0000 (21:47 -0600)]
Merge pull request #959 from TiemoJung/texture_upgrade
Fix for not transforming all image symbols into sampled images symbols
David Srbecky [Mon, 3 Jul 2017 14:48:49 +0000 (15:48 +0100)]
Multiview extension: Accept layout(num_views) qualifier
t.jung [Sat, 1 Jul 2017 08:26:23 +0000 (10:26 +0200)]
io resolver improvements
- adds stage parameter to end notify
- adds begin notify
- adds begin resolve
- adds end resolve
Change-Id: Ib134fe42a1a6d996d81dca59475cce6b4a9ff865
John Kessenich [Fri, 30 Jun 2017 22:39:32 +0000 (16:39 -0600)]
Merge pull request #964 from dsrbecky/multiview
Implement extensions GL_OVR_multiview and GL_OVR_multiview2
David Srbecký [Fri, 30 Jun 2017 17:33:51 +0000 (18:33 +0100)]
Implement extensions GL_OVR_multiview and GL_OVR_multiview2
They are almost identical to the already supported GL_EXT_multiview
t.jung [Thu, 29 Jun 2017 13:29:32 +0000 (15:29 +0200)]
Fix for not transforming all image symbols into sampled images symbols
Change-Id: I10f9bcea2e97f36912a81214552b93db4805b61a
John Kessenich [Thu, 29 Jun 2017 21:28:52 +0000 (15:28 -0600)]
Merge pull request #957 from amdrexu/feature
Implement extension GL_ARB_shader_stencil_export
John Kessenich [Thu, 29 Jun 2017 21:20:59 +0000 (15:20 -0600)]
Merge pull request #961 from benvanik/fix_order
Fixing initialization order to fix -Wreorder warning in clang.
Ben Vanik [Thu, 29 Jun 2017 18:13:35 +0000 (11:13 -0700)]
Fixing initialization order to fix -Wreorder warning in clang.
John Kessenich [Thu, 29 Jun 2017 16:46:47 +0000 (10:46 -0600)]
GLSL: validation of early_fragment_tests, others, on an object.
Some layout qualifiers must not be on an object and had a TBD for testing
them: early_fragment_tests, cw/ccw, points, point_mode, and
fractional_even_spacing/fractional_odd_spacing/equal_spacing.
Rex Xu [Thu, 29 Jun 2017 09:46:34 +0000 (17:46 +0800)]
Implement extension GL_ARB_shader_stencil_export
John Kessenich [Tue, 27 Jun 2017 21:17:38 +0000 (15:17 -0600)]
HLSL: Fix #942: Map SV_TargetN to SPV Location N.
John Kessenich [Tue, 27 Jun 2017 01:56:41 +0000 (19:56 -0600)]
HLSL: Fix hull-shader test.
John Kessenich [Mon, 26 Jun 2017 23:56:54 +0000 (17:56 -0600)]
HLSL: Don't pin down which tessellation stage must error check input primitive.
John Kessenich [Mon, 26 Jun 2017 23:52:22 +0000 (17:52 -0600)]
HLSL: support point mode.