| Age | Commit message (Collapse) | Author |
|---|
|
Improve performance when compiling small shaders.
Avoid copying witness table entries that are not getting used during linking.
Avoid copying auto-diff related decorations and derivative functions during linking, if the user modules doesn't use autodiff.
Cache operator overload resolution results on global session, so each new Session doesn't need to repetitively run through overload resolution from scratch.
|
|
input values are equal, the gradient is split evenly between both inputs. (#6411)
Co-authored-by: Yong He <yonghe@outlook.com>
|
|
Partially fix #6378
* Fix invalid access mode for texture_buffer
* Fix texture view create issue in metal
In newTextureView, levelRange should represent the mipmap level range,
while sliceRange should represent the texture layer range for texture
array. But the implement inverse those two wrongly.
|
|
* Simplify implicit cast ctors for vector & matrix.
* Fix formatting.
* Fix tests.
* Fix Falcor test.
* Mark __builtin_cast as internal.
|
|
* HLSL: Add 'f' suffix to float literals in code generation
Fixes #6078
1) Previously, Slang would emit HLSL float literals without a suffix (e.g.,"1.5"),
which caused DXC to interpret them as 64-bit doubles by default unless
the -HV 202x flag was used. This could cause validation errors when these literals
were used with intrinsics that only accept 32-bit floats (like ddx, ddy).
This change modifies the HLSL emitter to explicitly add 'f' suffix to
32-bit float literals, ensuring they are correctly typed regardless of DXC's version or flags.
For example:
- "1.5" becomes "1.5f"
- "(0.0 / 0.0)" becomes "(0.0f / 0.0f)" for NaN
- "float4(1.0, 2.0, 3.0, 4.0)" becomes "float4(1.0f, 2.0f, 3.0f, 4.0f)"
2) Added a test case to verify the behavior with various float literal scenarios
including basic literals, intrinsic calls, and vector construction.
3) Remove some tests that were marked as known failures
* Add dxil test as suggested by jkwak-work
|
|
* Support for dynamic array of textures access for Texture footprint.
* Do a check for DXIL for dynamic array/footprint.
* format code (#40)
---------
Co-authored-by: jsmall-nvidia <jsmall@nvidia.com>
Co-authored-by: slangbot <ellieh+slangbot@nvidia.com>
Co-authored-by: Yong He <yonghe@outlook.com>
|
|
This commit inlines functions that takes InputPatch and OutputPatch as
the function parameter.
Co-authored-by: Yong He <yonghe@outlook.com>
|
|
|
|
* Fix zero size array handling in slangc
Fixes #2890
1. Fix zero size array handling in slangc
2. Add new zero size array diagnostic test.
* format code
* fix review comments
---------
Co-authored-by: slangbot <186143334+slangbot@users.noreply.github.com>
Co-authored-by: Ellie Hermaszewska <ellieh@nvidia.com>
|
|
|
|
|
|
Fixes #4251
When binding attributes (like [[vk::binding]]) are specified on uniforms
that get packed into the default constant buffer, these binding
attributes are effectively ignored since the uniform will always be
placed at descriptor set 0, binding 0. This can be confusing for users
who expect their explicit bindings to take effect.
This change adds a new warning (71) that informs users when their
binding attributes on uniforms will be ignored, and suggests declaring
the uniform inside a constant buffer to preserve the explicit binding.
The warning helps users understand:
1. Why their binding attribute isn't having the expected effect
2. That the uniform is being packed into the default constant buffer
3. How to fix it by using a constant buffer declaration
Added test case in tests/bugs/binding-attribute-ignored.slang to verify
the warning behavior.
Co-authored-by: Ellie Hermaszewska <ellieh@nvidia.com>
|
|
|
|
The GLSL interface block implementation accidentally
removed global variables which are still in use.
|
|
Adds support for input GLSL interface blocks.
closes #5535
Co-authored-by: slangbot <186143334+slangbot@users.noreply.github.com>
|
|
* Fix modifier parsing.
* Fix.
* Fix.
* Fix.
|
|
* Neaten metal parameter block checking
* Disallow only resources in constant buffers in parameterblocks on metal
closes https://github.com/shader-slang/slang/issues/6200
* add unit test for metal parameterblock cbuffer
---------
Co-authored-by: Yong He <yonghe@outlook.com>
|
|
* Fix 6317.
* Fixes #6316.
* Fix cmake preset.
---------
Co-authored-by: Ellie Hermaszewska <ellieh@nvidia.com>
|
|
* Allow LHS of `where` to be any type.
* Register free-form extensions when loading precompiled module.
* Fix test.
* Fix.
* Fix `as<IRType>`.
* try fix precompiled module test.
|
|
* Fix crash when handling uninitialized resource type
close #6328.
When declare a var with struct type, if the struct has a resource
type field and it doesn't provide explicit constructor, because
slang won't implicit construct such variable at declare site if user
doesn't explicitly call the initializer list, we should report the
resource type field uninitialized as an error to prevent future possible
crash when legalize any use of such variable.
|
|
Closes https://github.com/shader-slang/slang/issues/6339
|
|
* Publicly link core module sources
Closes https://github.com/shader-slang/slang/issues/6037
* format
* format
* remove unnecessary conditional
* comments
* neaten
* Only build embedded core modules when necessary
* Make dependencies on core module source more precise
* more precice dependencies on core module source
* Neaten
|
|
|
|
Co-authored-by: Yong He <yonghe@outlook.com>
Co-authored-by: Ellie Hermaszewska <ellieh@nvidia.com>
|
|
* Add raypayload decoration to ray payload structs
Closes https://github.com/shader-slang/slang/issues/6104
* Disable PAQs when compiling with DXC
See https://github.com/shader-slang/slang/issues/3448
|
|
fs. (#6314)
Co-authored-by: Ellie Hermaszewska <ellieh@nvidia.com>
|
|
* Remove the docs/proposals directory
This directory will get added to the spec repository in the following PR:
https://github.com/shader-slang/spec/pull/6
This closes #6155.
* Remove entry from .github/CODEOWNERS file
* Redirect some proposal references
---------
Co-authored-by: Yong He <yonghe@outlook.com>
|
|
* initial work for metal subgroups
* add glsl intrinsics
* enable wave tests
* enable glsl subgroup tests, glsl barrier fixes
* minor fixes
* fix incorrect test target
* disable some glsl functional tests
* disable failing glsl test
---------
Co-authored-by: Yong He <yonghe@outlook.com>
|
|
The musl libc replacement removed fopen64, fgetpos64, and other 64-bit
variants because it's 64-bit only. However, it does have the following
in its headers:
#define fgetpos64 fgetpos
Just importing <stdio.h> is enough to get slang compiling on musl
systems like Alpine Linux.
Fixes #6330.
Co-authored-by: Yong He <yonghe@outlook.com>
|
|
* Add WGSL to list of formats supporting multiple entry points
* Add a test for generating multiple WGSL entrypoints
|
|
Propagate the kIROp_MemoryQualifierSetDecoration
memory decorations when converting from structures
to flat global resources
struct Tex
{
globallycoherent float4 bias = { 0.1, 0.3, 0.9, 1.5};
};
uniform Tex gTex;
should have the following code generated in GLSL:
coherent layout(rgba32f)
layout(binding = 2)
uniform image2D globalParams_gTex_tex_0;
This closes #6079
|
|
* Use and() and or() functions for logical-AND and OR
With this commit, Slang will emit function calls to `and()` and `or()`
for the logical-AND and logical-OR when the operands are non-scalar and
the target profile is SM6.0 and above. This is required change from
SM6.0.
For WGSL, there is no operator overloadings of `&&` and `||` when the
operands are non-scalar. Unlike HLSL, WGSL also don't have `and()` nor
`or()`. Alternatively, we can use `select()`.
|
|
* Support stage_switch.
* Update proposal status.
* Fix gl_InstanceID.
* Fix.
|
|
(#6292)
When a module consists of multiple source files, the module scope gets over-written for
each source file that's parsed into the module.
The result is that if you do something like the following, where source1.slang contains
an import statement, then the imported module will get imported into the module scope
corresponding to source2.slang, but won't be found from the scope of source1.slang.
slangc source1.slang source2.slang # 1 module from 2 source files
This patch fixes this problem by not over-writing existing container decl scope
when parsing new source files into the container.
This closes $6221.
|
|
* Allow tuples to work with initiailizer list.
* Update definition of C-Style types.
|
|
* maxtessfactor attribute should take a floating point value
* Support integer value on maxtessfactor
|
|
* Fix DCE for calls to functions that have associations
* Update slang-ir-util.cpp
* Update slang-ir-util.cpp
|
|
* Use two-stage parsing to disambiguate generic app and comparison.
* Typo fix.
* Update doc.
|
|
* SP004: implement initialize list translation to ctor
- We synthesize a member-wise constructor for each struct follow
the rules described in SP004.
- Add logic to translate the initialize list to constructor invoke
- Add cuda-host decoration for the synthesized constructor
- Remove the default constructor when we have a valid member init constructor
- Disable -zero-initialize option, will re-implement it in followup (#6109).
- Fix the overload lookup issue
When creating invoke expression for ctor, we need to call
ResolveInvoke() to find us the best candidates, however
the existing lookup logic could find us the base constructor
for child struct, we should eliminate this case by providing
the LookupOptions::IgnoreInheritance to lookup, this requires
us to create a subcontext on SemanticsVisitor to indicate that
we only want to use this option on looking the constructor.
- Do not implicit initialize a struct that doesn't have explicit default
constructor.
Co-authored-by: slangbot <186143334+slangbot@users.noreply.github.com>
|
|
* Fix autodiff issue for vector<T, N>
Close #6154
We didn't implement correctly for vector<T, N> regarding the
differentiablity. As we check differentiable before specialization,
however according to the definition of vector, it has to be specialized
to IFloat to know it's conformed to IDifferential type. Therefore for
parameter type vector<T, N> will become no_diff.
Therefore, we change the implementation a to make it explicit conform
to IDifferential type.
* fix typo
|
|
|
|
* Fix warning about push constants, shaderRecordEXT
These resources should not warn about missing vk::binding.
* format code
---------
Co-authored-by: slangbot <186143334+slangbot@users.noreply.github.com>
|
|
entrypoint. (#6260)
* Auto enable `-fvk-use-entrypoint-name` when there is more than one entrypoint.
* Fix.
|
|
* initial work
* more work
* more work on glsl intrinsics
* add subgroup broadcast for glsl
* wip add wgsl extension tracking
* enable tests, enable extensions and added some todos
* format and warning fixes
* fix wgsl extension tracker
---------
Co-authored-by: Yong He <yonghe@outlook.com>
|
|
|
|
* Base compiler options for targets on target-specific compiler options
Before this change, the target compiler options were based on the linkage-wide compiler
options, which where later again inherited from (basically a no-op).
With this change, the target-specific compiler options are added first, and then
the linkage-wide comnpiler options are inherited from.
* Remove debug instructions if target-specific setting is NONE
This helps to address #6092.
* Make sure the linkage debug info level is sufficient for each target
This closes #6092.
|
|
* Distribute slang-glsl-module.bin in release packages.
* Fix.
* fix2.
* Build glsl modlue into a dll.
|
|
* AD: Docs Update
* More documentation
* More documentation
* More docs fixes
* Cleanup documentation
* More docs polish. Add docs for the [Differentiable] attributes
* Fixup code sections
* Fixup
* Address review comments
* regenerate documentation Table of Contents
* Update docs with more playground links
---------
Co-authored-by: slangbot <186143334+slangbot@users.noreply.github.com>
Co-authored-by: Yong He <yonghe@outlook.com>
|
|
|
|
|
