John Kessenich [Sun, 12 Nov 2017 22:55:22 +0000 (15:55 -0700)]
Memory: Do process and 1st thread tear down.
Addresses #928, #389, and 1st item in #976. Overlaps #916.
This had been dropped, when moving away from the old Win32 DLL model.
Issue: per-thread tear down for other threads.
John Kessenich [Sun, 12 Nov 2017 22:43:03 +0000 (15:43 -0700)]
Memory: Move to a normal model of ownership of memory pools, for new/delete.
Addresses step 4 of #976, overlaps #916.
For each pool, now, it is newed, remembered, and freed by the same entity,
rather than having a mix (thread finalize freeing current pool) that could
lead to double freeing of the same pool. It is quite rational and simple now.
This will enable reinstalling process and thread tear down.
John Kessenich [Sun, 12 Nov 2017 22:28:58 +0000 (15:28 -0700)]
Memory: Non-Functional: Rationalize and improve encapsulation of TLS usage.
This will make the next (functional) commit easier to see.
John Kessenich [Fri, 10 Nov 2017 20:52:00 +0000 (13:52 -0700)]
Merge pull request #1142 from greg-lunarg/kg3
Update spirv-tools known good.
John Kessenich [Fri, 10 Nov 2017 02:19:56 +0000 (19:19 -0700)]
Merge pull request #1145 from antiagainst/travis-auto-deploy
Travis: auto deploy build artifacts to GitHub Releases
Lei Zhang [Thu, 9 Nov 2017 02:32:01 +0000 (21:32 -0500)]
Travis: auto deploy build artifacts to GitHub Releases
Pushing a commit to the master branch will trigger a build on
Travis. If the build is successful, the artifacts will be
collected and pushed to GitHub Releases, under the "master-tot"
release.
John Kessenich [Thu, 9 Nov 2017 19:10:24 +0000 (12:10 -0700)]
Merge pull request #1114 from LoopDawg/validator-script
Add script to crank test shaders through spirv-val
John Kessenich [Thu, 9 Nov 2017 19:08:28 +0000 (12:08 -0700)]
Merge pull request #1116 from LoopDawg/reverse-shift-args
Reverse order of setShiftBindingForSet parameters
John Kessenich [Thu, 9 Nov 2017 19:07:42 +0000 (12:07 -0700)]
Merge pull request #1144 from antiagainst/appveyor-auto-deploy
Appveyor: auto deploy build artifacts to GitHub Releases
LoopDawg [Sat, 21 Oct 2017 16:46:39 +0000 (10:46 -0600)]
Reverse order of setShiftBindingForSet parameters
Per feedback on PR #1111, this reverses the order of the parameters for the setShiftBinding API.
It is now:
void TShader::setShiftBindingForSet(TResourceType res, unsigned int base, unsigned int set);
LoopDawg [Thu, 19 Oct 2017 18:09:56 +0000 (12:09 -0600)]
Add script to crank test shaders through spirv-val
This script will crank a supplied set of glslang test shaders through the
spirv-val tool, reporting on the results.
There are some important things to note:
* Like 'runtests', this must be invoked from the 'Test' subdirectory.
* This is mostly useful on the hlsl.* tests, although it is not strictly
limited to those. The reason is that most of the glsl tests either contain
validation error cases, and so fail to compile, or are not using a #version
compatible with producing SPIR-V modules.
* Some tests, such as negative tests, or most of the glsl tests, have
intentional compilation errors. This script treats that as OK. Failures
are successfully compiling shaders which proceed to fail spirv-val.
* spirv-val is looked for in either the External directory, or if not
found there, in a sibling directory of glslang, and if not found there
either, in /usr/local/bin.
* There are a bunch of command line options. ./validate-shaders.sh --help
will describe them.
Some examples to try:
./validate-shaders.sh hlsl.* # exercise all hlsl.* tests.
./validate-shaders.sh --terse hlsl.* # same, but tersely.
# dump validator results for problems in something.frag:
./validate-shaders.sh --quiet --dump-val something.frag
Lei Zhang [Wed, 8 Nov 2017 22:26:51 +0000 (17:26 -0500)]
Appveyor: auto deploy build artifacts to GitHub Releases
Pushing a commit to the master branch will trigger a build on
Appveyor. If the build is successful, the artifacts will be
collected and pushed to GitHub Releases, under the "master-tot"
release.
GregF [Wed, 8 Nov 2017 20:56:08 +0000 (13:56 -0700)]
Update spirv-tools known good.
John Kessenich [Fri, 3 Nov 2017 09:39:48 +0000 (03:39 -0600)]
SPV: Don't support noise*() when generating SPV.
John Kessenich [Fri, 3 Nov 2017 04:48:15 +0000 (22:48 -0600)]
Errors: Have a whole set of tool-level error messages report to stderr.
John Kessenich [Fri, 3 Nov 2017 04:32:14 +0000 (22:32 -0600)]
Versioning: Update some version strings. Still need better overall version.
John Kessenich [Fri, 3 Nov 2017 03:53:25 +0000 (21:53 -0600)]
Merge pull request #1139 from greg-lunarg/remap2
Remapper: Fix strip algo when strip ranges overlap
GregF [Thu, 2 Nov 2017 20:01:57 +0000 (14:01 -0600)]
Remapper: Fix strip algo when strip ranges overlap
John Kessenich [Thu, 2 Nov 2017 12:48:32 +0000 (06:48 -0600)]
Fix #1079: don't give error when macro() name used without open (.
John Kessenich [Thu, 2 Nov 2017 06:05:53 +0000 (00:05 -0600)]
John Kessenich [Thu, 2 Nov 2017 04:33:45 +0000 (22:33 -0600)]
Fix #1065: don't validate Vulkan binding numbers.
To handle sparse assignments, don't check maxCombinedTextureImageUnits.
John Kessenich [Sat, 28 Oct 2017 20:42:44 +0000 (14:42 -0600)]
Non-functional: copyright update, to help with non-determinism test.
John Kessenich [Sat, 28 Oct 2017 18:55:02 +0000 (12:55 -0600)]
Tests: Non-determinancy: Add back in the IntMin tests.
John Kessenich [Sat, 28 Oct 2017 08:03:09 +0000 (02:03 -0600)]
Merge pull request #1133 from greg-lunarg/kg2
Add python3 compatibility to update_glslang_sources.py
GregF [Fri, 27 Oct 2017 21:39:45 +0000 (15:39 -0600)]
Add python3 compatibility to update_glslang_sources.py
John Kessenich [Fri, 27 Oct 2017 10:15:57 +0000 (04:15 -0600)]
Non-determinant tests: Remove recent tests, to see if things stabelize.
Both debug and release clang builds have segfaulted on recent
changes, non deterministically, while doing the single/multi-thread
test all test files. Removing recent test files, to see if it gives
a clue.
John Kessenich [Fri, 27 Oct 2017 09:10:21 +0000 (03:10 -0600)]
Non-functional: Add {} expected by convention; retrigger failed bot tests.
John Kessenich [Fri, 27 Oct 2017 07:30:03 +0000 (01:30 -0600)]
Merge pull request #1129 from xorgy/const-fold-int-min-modulo-negative-one
GLSL: Fold constant SHRT_MIN/INT_MIN/LLONG_MIN % -1 to 0.
Aaron Muir Hamilton [Wed, 25 Oct 2017 00:11:53 +0000 (00:11 +0000)]
GLSL: Fold constant SHRT_MIN/INT_MIN/LLONG_MIN % -1 to 0.
John Kessenich [Tue, 24 Oct 2017 14:18:09 +0000 (08:18 -0600)]
Merge pull request #1124 from xorgy/check-inner-implicit-atomic-uint
GLSL: Check for implicit inner dimension in array of atomic_uint.
John Kessenich [Tue, 24 Oct 2017 14:14:38 +0000 (08:14 -0600)]
Merge pull request #1127 from xorgy/preprocessor-int-min-mod-negative-one
Preprocessor: Evaluate INT_MIN % -1 to 0.
Aaron Muir Hamilton [Tue, 24 Oct 2017 10:59:01 +0000 (10:59 +0000)]
Preprocessor: Evaluate INT_MIN % -1 to 0.
John Kessenich [Tue, 24 Oct 2017 04:11:42 +0000 (22:11 -0600)]
Merge pull request #1125 from greg-lunarg/kg1
Update spirv-tools known-good
GregF [Mon, 23 Oct 2017 22:48:42 +0000 (16:48 -0600)]
Update spirv-tools known-good
Aaron Muir Hamilton [Mon, 23 Oct 2017 15:31:10 +0000 (15:31 +0000)]
GLSL: Check for implicit inner dimension in array of atomic_uint.
John Kessenich [Mon, 23 Oct 2017 05:28:22 +0000 (23:28 -0600)]
Merge pull request #1118 from xorgy/only-swizzle-numbers-and-bools
Only try swizzles on vectors, numbers, and booleans.
John Kessenich [Mon, 23 Oct 2017 05:19:02 +0000 (23:19 -0600)]
Merge branch 'only-parse-inf-constant-in-hlsl' of https://github.com/xorgy/glslang into xorgy-only-parse-inf-constant-in-hlsl
John Kessenich [Mon, 23 Oct 2017 05:17:18 +0000 (23:17 -0600)]
Tests: Add test for GLSL +-#INF.
John Kessenich [Mon, 23 Oct 2017 05:09:32 +0000 (23:09 -0600)]
Merge pull request #1120 from xorgy/overlong-hex-literal
Check for hexadecimal literals exceeding MaxTokenLength.
Aaron Muir Hamilton [Mon, 23 Oct 2017 02:56:27 +0000 (02:56 +0000)]
GLSL: Only parse [-]1.#INF in HLSL mode.
Aaron Muir Hamilton [Sun, 22 Oct 2017 17:41:13 +0000 (17:41 +0000)]
Check for hexadecimal literals exceeding MaxTokenLength.
Aaron Muir Hamilton [Sat, 21 Oct 2017 19:13:42 +0000 (19:13 +0000)]
Only try swizzles on vectors, numbers, and booleans.
John Kessenich [Fri, 20 Oct 2017 22:31:26 +0000 (16:31 -0600)]
Merge pull request #1115 from LoopDawg/stdarray-binding-set
Nonfunctional: minor: use std::array for per-set shifts, fix warning.
LoopDawg [Fri, 20 Oct 2017 18:02:38 +0000 (12:02 -0600)]
Nonfunctional: minor: use std::array for per-set shifts, fix warning.
Two unrelated, minor tweaks:
(1) Use std::array for shiftBindingForSet. Now matches shiftBinding.
(2) Add parens in shouldFlatten() to make compiler warning happy.
John Kessenich [Fri, 20 Oct 2017 17:35:29 +0000 (11:35 -0600)]
Merge pull request #1111 from LoopDawg/per-set-binding-offsets
Add per-descriptor-set IO mapping shift values.
John Kessenich [Fri, 20 Oct 2017 16:56:50 +0000 (10:56 -0600)]
SPV: Don't emit StorageImageMultisample capability for subpass images.
John Kessenich [Fri, 20 Oct 2017 16:56:20 +0000 (10:56 -0600)]
Non-functional: missing comments from missed git add.
John Kessenich [Thu, 19 Oct 2017 22:54:25 +0000 (16:54 -0600)]
HLSL: Fix #1106. Support shader setting of binding/set for $Global.
LoopDawg [Wed, 18 Oct 2017 01:27:14 +0000 (19:27 -0600)]
Add per-descriptor-set IO mapping shift values.
This PR adds the ability to provide per-descriptor-set IO mapping shift
values. If a particular binding does not land into a per-set value,
then it falls back to the prior behavior (global shifts per resource class).
Because there were already 6 copies of many different methods and internal
variables and functions, and this PR would have added 6 more, a new API is
introduced to cut down on replication and present a cleaner interface.
For the global (non-set-specific) API, the old entry points still exist
for backward compatibility, but are phrased internally in terms of the
following.
// Resource type for IO resolver
enum TResourceType {
EResSampler,
EResTexture,
EResImage,
EResUbo,
EResSsbo,
EResUav,
EResCount
};
Methods on TShader:
void setShiftBinding(TResourceType res, unsigned int base);
void setShiftBindingForSet(TResourceType res, unsigned int set, unsigned int base);
The first method replaces the 6 prior entry points of various spellings, which
exist now in depreciated form. The second provides per-resource-set functionality.
Both accept an enum from the list above.
From the command line, the existing options can accept either a single shift value as
before, or a series of 1 or more [set offset] pairs. Both can be provided, as in:
... --stb 20 --stb 2 25 3 30 ...
which will use the offset 20 for anything except descriptor set 2 (which uses 25) and
3 (which uses 30).
John Kessenich [Thu, 19 Oct 2017 08:07:30 +0000 (02:07 -0600)]
GLSL->SPIR-V: Put precision decorations on imageLoad().
John Kessenich [Thu, 19 Oct 2017 06:26:39 +0000 (00:26 -0600)]
GLSL: Make sampling operations have agnostic precision qualifiers for desktop.
Desktop defaults to highp for samplers, but this should not apply to the built-in
functions, so make it appy only to user declarations.
John Kessenich [Thu, 19 Oct 2017 06:24:05 +0000 (00:24 -0600)]
Merge pull request #1113 from xorgy/preprocessor-int-min-over-negative-one
Preprocessor: Evaluate INT_MIN / -1 to 0.
John Kessenich [Wed, 18 Oct 2017 20:53:53 +0000 (14:53 -0600)]
Merge pull request #1102 from KhronosGroup/partially-flatten
HLSL: Partially flatten hierarchies, instead of all or nothing.
Aaron Muir Hamilton [Wed, 18 Oct 2017 19:33:00 +0000 (19:33 +0000)]
Preprocessor: Evaluate INT_MIN / -1 to 0.
This division is undefined behaviour which raises SIGFPE on x86.
Most C++ preprocessors evaluate this silently to 0.
John Kessenich [Tue, 17 Oct 2017 18:27:51 +0000 (12:27 -0600)]
Build: Fix build warning.
John Kessenich [Tue, 17 Oct 2017 15:03:17 +0000 (09:03 -0600)]
Merge pull request #1109 from xorgy/never-try-to-swizzle-void
GLSL: Don't try to swizzle void.
Aaron Muir Hamilton [Tue, 17 Oct 2017 08:11:33 +0000 (08:11 +0000)]
Don't try to swizzle void.
GregF [Mon, 16 Oct 2017 16:53:53 +0000 (10:53 -0600)]
Update spirv-tools known-good
John Kessenich [Mon, 16 Oct 2017 21:29:07 +0000 (15:29 -0600)]
PP: Fix #1104: Missing check for #if overflow.
Also, rationalized this to generally make it safer and more readable.
It could use a more modern approach, at some point...
John Kessenich [Mon, 16 Oct 2017 19:11:53 +0000 (13:11 -0600)]
HLSL: More clip fix: It is more involved than previous commit. Complete.
John Kessenich [Mon, 16 Oct 2017 17:42:35 +0000 (11:42 -0600)]
Fix #1103: clip() works on int/uint.
John Kessenich [Mon, 16 Oct 2017 17:35:11 +0000 (11:35 -0600)]
Tests: Add clip(int) tests.
John Kessenich [Fri, 13 Oct 2017 05:12:40 +0000 (23:12 -0600)]
Infrastructure: Add a TShader method to publicly expose the intermediate.
Fixes #1088.
John Kessenich [Wed, 11 Oct 2017 20:03:45 +0000 (14:03 -0600)]
HLSL: Partially flatten hierarchies, instead of all or nothing.
Fixes #1092. Allows arrays of opaques to keep arrayness, unless
needed by uniform array flattening.
Can handle assignments of mixed amounts of flattening.
John Kessenich [Wed, 11 Oct 2017 03:05:06 +0000 (06:05 +0300)]
Merge pull request #1098 from LoopDawg/attribute-gets
HLSL: nonfunctional: add helper access methods to TAttributeMap
John Kessenich [Tue, 10 Oct 2017 22:54:06 +0000 (01:54 +0300)]
Merge pull request #1097 from SWIFTingAround/master
Set cmake-policy CMP0048 to NEW
John Kessenich [Tue, 10 Oct 2017 22:48:49 +0000 (01:48 +0300)]
Merge pull request #1094 from LoopDawg/split-shadow-tx
HLSL: nonfunctional: rename setId -> switchId, add comment
John Kessenich [Tue, 10 Oct 2017 22:48:04 +0000 (01:48 +0300)]
Merge pull request #1080 from hedejing/master
Add getAsLoopNode() method for TIntermLoop
LoopDawg [Tue, 10 Oct 2017 20:59:40 +0000 (14:59 -0600)]
HLSL: add helper access methods to TAttributeMap
There was some code replication around getting string and integer
values out of an attribute map. This adds new methods to the
TAttributeMap class to encapsulate some accessor details.
Tim Diekmann [Tue, 10 Oct 2017 14:26:31 +0000 (16:26 +0200)]
Set cmake-policy CMP0048 to NEW
He Dejing [Mon, 2 Oct 2017 18:41:23 +0000 (02:41 +0800)]
Add getAsLoopNode() method for Loop node
follow local conventions of spaces
LoopDawg [Sat, 7 Oct 2017 22:42:37 +0000 (16:42 -0600)]
HLSL: nonfunctional: rename setId -> switchId, add comment
Method rename, add comment about its intended use. No
test diffs.
John Kessenich [Sat, 7 Oct 2017 19:23:09 +0000 (13:23 -0600)]
HLSL: Validate implicit initializer assignment to opaque members.
Fixes #1091.
John Kessenich [Sat, 7 Oct 2017 17:33:11 +0000 (11:33 -0600)]
SPV: GeneratorVersion: bump version number because of atomic decrement change.
John Kessenich [Sat, 7 Oct 2017 04:19:21 +0000 (07:19 +0300)]
Merge pull request #1090 from tafuri/#1084-fix-segfault
Only track variables in the global scope
John Kessenich [Sat, 7 Oct 2017 03:21:48 +0000 (21:21 -0600)]
SPV: Correct semantics of atomic-counter decrement.
The semantics differ between GLSL/HLSL and SPIR-V.
Translate between these.
John Kessenich [Sat, 7 Oct 2017 03:01:07 +0000 (21:01 -0600)]
HLSL: Tests: Add subset of flattened opaque test to legalize results.
Sebastian Tafuri [Fri, 6 Oct 2017 22:47:54 +0000 (00:47 +0200)]
Added test for local structured buffer variable
Sebastian Tafuri [Fri, 6 Oct 2017 22:12:50 +0000 (00:12 +0200)]
Fixed formatting
John Kessenich [Fri, 6 Oct 2017 21:33:20 +0000 (00:33 +0300)]
Merge pull request #1089 from LoopDawg/split-shadow-tx
HLSL: split textures used for both shadow and non-shadow modes
Sebastian Tafuri [Fri, 6 Oct 2017 20:45:00 +0000 (22:45 +0200)]
Only track variables in the global scope
LoopDawg [Thu, 5 Oct 2017 22:25:52 +0000 (16:25 -0600)]
HLSL: split textures used for both shadow and non-shadow modes
A single texture can statically appear in code mixed with a shadow sampler
and a non-shadow sampler. This would be create invalid SPIR-V, unless
one of them is provably dead.
The previous detection of this happened before DCE, so some shaders would
trigger the error even though they wouldn't after DCE. To handle that
case, this PR splits the texture into two: one with each mode. It sets
"needsLegalization" (if that happens for any texture) to warn that this shader
will need post-compilation legalization.
If the texture is only used with one of the two modes, behavior is as it
was before.
John Kessenich [Fri, 6 Oct 2017 06:24:41 +0000 (00:24 -0600)]
Readme: Update with additional Windows build information for SPIRV-tools build.
John Kessenich [Wed, 4 Oct 2017 20:41:08 +0000 (23:41 +0300)]
Merge pull request #1085 from KhronosGroup/flatten-assign-subset
HLSL: Fix #954: Track/access subsets of flattened multi-level aggrega…
John Kessenich [Wed, 4 Oct 2017 19:27:43 +0000 (13:27 -0600)]
HLSL: Fix #954: Track/access subsets of flattened multi-level aggregates.
Works in conjuction with d1be754 to represent and modify a partially
dereferenced multi-level flattened aggregate.
John Kessenich [Wed, 4 Oct 2017 19:01:25 +0000 (22:01 +0300)]
Merge pull request #1081 from xxxbxxx/for-upstream-1
HLSL: Fix crash when flattening both side of assignement simultaneously.
Unknown [Tue, 3 Oct 2017 07:10:26 +0000 (09:10 +0200)]
HLSL: Fix crash when flattening both side of assignement simultaneously.
John Kessenich [Mon, 2 Oct 2017 22:47:48 +0000 (01:47 +0300)]
Merge pull request #1075 from LoopDawg/subpass-input
HLSL: add subpass input types and methods
LoopDawg [Wed, 27 Sep 2017 15:04:43 +0000 (09:04 -0600)]
HLSL: add subpass input types and methods
Add support for Subpass Input proposal of issue #1069.
Subpass input types are given as:
layout(input_attachment_index = 1) SubpassInput<float4> subpass_f;
layout(input_attachment_index = 2) SubpassInput<int4> subpass_i;
layout(input_attachment_index = 3) SubpassInput<uint4> subpass_u;
layout(input_attachment_index = 1) SubpassInputMS<float4> subpass_ms_f;
layout(input_attachment_index = 2) SubpassInputMS<int4> subpass_ms_i;
layout(input_attachment_index = 3) SubpassInputMS<uint4> subpass_ms_u;
The input attachment may also be specified using attribute syntax:
[[vk::input_attachment_index(7)]] SubpassInput subpass_2;
The template type may be a shorter-than-vec4 vector, but currently user
structs are not supported. (An unimplemented error will be issued).
The load operations are methods on objects of the above type:
float4 result = subpass_f.SubpassLoad();
int4 result = subpass_i.SubpassLoad();
uint4 result = subpass_u.SubpassLoad();
float4 result = subpass_ms_f.SubpassLoad(samp);
int4 result = subpass_ms_i.SubpassLoad(samp);
uint4 result = subpass_ms_u.SubpassLoad(samp);
Additionally, the AST printer could not print EOpSubpass* nodes. Now it can.
Fixes #1069
John Kessenich [Sat, 30 Sep 2017 20:54:18 +0000 (14:54 -0600)]
Build: Fix a couple build issues.
John Kessenich [Sat, 30 Sep 2017 20:34:50 +0000 (14:34 -0600)]
HLSL: Additional attribute support: [[]], namespace, parameters:
- support C++11 style brackets [[...]]
- support namespaces [[vk::...]]
- support these on parameter declarations in functions
- support location, binding/set, input attachments
John Kessenich [Fri, 29 Sep 2017 23:51:52 +0000 (17:51 -0600)]
Fix #1060: Could crash if using --source-entry-point with -e; fixed.
John Kessenich [Fri, 29 Sep 2017 17:57:30 +0000 (20:57 +0300)]
Merge pull request #1078 from greg-lunarg/addopt14
Remove opaque workarounds
GregF [Fri, 29 Sep 2017 16:08:20 +0000 (10:08 -0600)]
Legalization: Fix tests after workarounds removed
John Kessenich [Sun, 10 Sep 2017 21:21:05 +0000 (15:21 -0600)]
SPV: rationalize parameter handling for "original" and "writable" parameters.
John Kessenich [Fri, 8 Sep 2017 08:38:07 +0000 (02:38 -0600)]
HLSL: Pass opaques by local copy, instead of by interface original.
Also, remove assumption that if something is opaque that it
must be in the UniformConstant storage class.
This allows function declarations to know all parameters will
be in the Function storage class.
John Kessenich [Thu, 7 Sep 2017 10:33:11 +0000 (04:33 -0600)]
HLSL: Remove workarounds for assigning to opaques.
This assumes it will be combined with optimizing transforms
that eliminate assignments to opaques.
John Kessenich [Fri, 29 Sep 2017 02:07:50 +0000 (05:07 +0300)]
Merge pull request #1076 from LoopDawg/shadow-tx-types
HLSL: force textures to shadow modes from combined samplers
LoopDawg [Wed, 27 Sep 2017 15:12:51 +0000 (09:12 -0600)]
HLSL: force textures to shadow modes from combined samplers
Texture shadow mode must match the state of the sampler they are
combined with. This change does that, both for the AST and the
symbol table. Note that the texture cannot easily be *created*
the right way, because this may not be known at that time. Instead,
the texture is subsequently patched.
This cannot work if a single texture is used with both a shadow and
non-shadow sampler, so that case is detected and generates an error.
This is permitted by the HLSL language, however. See #1073 discussion.
Fixed one test source that was using a texture with both shadow and
non-shadow samplers.
John Kessenich [Thu, 28 Sep 2017 19:38:29 +0000 (22:38 +0300)]
Merge pull request #1077 from greg-lunarg/addopt12
Legalization: Fix warnings and disable tests when spirv-tools not present