| Commit message (Collapse) | Author | Age |
|---|
| ... | |
| |
|
|
|
|
|
|
|
| |
* Fix crash when processing nested switch.
* Clean up.
---------
Co-authored-by: Yong He <yhe@nvidia.com>
|
| |
|
|
|
|
|
|
|
| |
* Reimplement address elimination pass.
* Fix error.
* Update test references.
Co-authored-by: Yong He <yhe@nvidia.com>
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
* Fixes for crash when inlining at global scope
Recent changes to the way inlining is implemented in the Slang compiler
have broken certain scenarios involving `static const` declarations.
The basic problem is that the initial-value expression for a `static const`
gets lowered into IR code at the global scope of a module, and if
that code includes `call`s to stdlib operations marked `forceInlineEarly`,
then we end up trying to apply inlining to code at module scope.
The current inlining operation assumes that all `call`s are in basic
blocks, and that the correct way to do inlining involves splitting
those blocks.
This change adds logic to detect when the callee at a call site to
be inlined consists of a single basic block ending in a `return`,
and in that case it invokes specialized inlining logic that doesn't
split basic blocks and doesn't need to care if the original `call`
is in a basic block.
Thus we are able to inline calls to single-basic-block `forceInlineEarly`
functions called as part of the initialization for global-scope
`static const` variables.
This logic does *not* solve the problem of calls to multi-block
`forceInlineEarly` functions from the global scope. Such calls cannot
really be inlined.
A secondary problem that arises when inlining such calls is that the
callee might include local temporaries (`var` instructions) that are
read and written (`load`s and `store`s), and none of those instructions
should be allowed at the global scope.
In the case of the functions being inlined here, the `load`/`store`
operations are superfluous, and should be cleaned up by our SSA pass.
The only reason that they seem to *not* be getting cleaned up in the
case that was been triggering crashes is that the callee is a generic.
The current logic for the SSA pass was skipping the bodies of generic
functions, so they would not be cleaned up. This change enables the SSA
pass to apply to the bodies of generic functions, and also ensures that
SSA cleanups are applied *before* any `forceInlineEarly` functions get
inlined.
* fixup: liveness test outputs
|
| |
|
| |
Co-authored-by: Yong He <yhe@nvidia.com>
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
* #include an absolute path didn't work - because paths were taken to always be relative.
* WIP inlining of functions that take or return string related types on GPU targets.
* Small fixes.
* Added a test.
* Add checking for any getStringHash insts are valid.
* Support getStringHash on CUDA.
* Tweak diagnostic.
|
| |
|
|
|
|
|
|
|
|
|
|
|
| |
argument. (#2536)
* Fix non-static generic func call issue.
* Add test case.
* Revert unnecessary change.
* Update test comment.
Co-authored-by: Yong He <yhe@nvidia.com>
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
* #include an absolute path didn't work - because paths were taken to always be relative.
* Add handling for root paths.
* Fixes around absolute paths.
* Add SimplifyStyle
* Remove unrequire include.
* Fix some details around RelativeFileSystem canonical paths.
* For MemoryFileSystem make sure "/a" and "a" maps to same canonical path.
* Add test for canonicalPath.
* Improve comment.
* More testing around canonical paths.
* Fix for user attribute lookup issue.
* Add a test.
* Small improvements in test.
* Improve the comments around lookup workaround.
|
| |
|
| |
Co-authored-by: Yong He <yhe@nvidia.com>
|
| |
|
|
|
|
|
| |
* Fix regression in check-overload.
* Make sure language server supports partiallyAppliedGenericExpr.
Co-authored-by: Yong He <yhe@nvidia.com>
|
| |
|
|
|
|
|
| |
* Allow interface requirements to reference to the interface type itself.
* add comment explaining the change.
Co-authored-by: Yong He <yhe@nvidia.com>
|
| |
|
|
|
|
|
|
|
| |
failure (#2396)
* Report diagnostic when dynamic dispatch failed instead of crashing.
* Specialize and SSA in a loop. Explicit specialization only interface.
Co-authored-by: Yong He <yhe@nvidia.com>
|
| |
|
| |
Co-authored-by: Yong He <yhe@nvidia.com>
|
| |
|
|
| |
(#2388)
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
* Warning on bool to float conversion.
* Fix test cases.
* Improve.
* LanguageServer: don't show constant value for non constant variables.
* Fix tests.
* Fix warnings in tests.
Co-authored-by: Yong He <yhe@nvidia.com>
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
* Add gfx interface definition in Slang.
- add gfx interface definitons in Slang.
- fix slang compiler to correctly type-check `out` interface argument.
- modify gfx interface to be fully COM compatible
- add convenient ShaderProgram creation methods to gfx.
* Fix compile errors and warnings.
* Update project files
* Fix cuda.
* Properly implement queryInterface in command encoder impls.
Co-authored-by: Yong He <yhe@nvidia.com>
|
| | |
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
* Major language server features.
* Include slangd in binary release.
* Fix compiler issues.
* Fix compiler error.
* Completion resolve.
* Various improvements.
* Update diagnostic test expected output.
* Bug fix for source locations.
* Adjust diagnostic update frequency.
* Update github actions to store artifacts.
* Fix infinite parser loop.
* Fix parser recovery.
* Fix parser recovery.
* Update test.
* Fix test.
* Disable IR gen for language server.
* Allow commit characters in auto completion.
* Fix lookup for invoke exprs.
* More parser robustness fixes.
* update solution file
Co-authored-by: Yong He <yhe@nvidia.com>
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
* Use IR pass to eliminate phi nodes
"Phi nodes" are one of the key contrivances that makes SSA (Static
Single Assignment) form work. Because SSA is so great for compiler
IRs, we kind of need to deal with phi nodes, but they also get in
the way because they don't have a direct analog in most lower-level
machine ISAs or execution models, nor in most of the high-level
languages a transpiler wants to emit. As a result a compiler like
ours needs to be able to eliminate the phi nodes from a program as
part of generating output code.
(For any clever people noting that SPIR-V supports phi nodes
directly: yes, it does. It doesn't need to and it probably *shouldn't*.
Anybody involved in the decision-making knows my reasoning, and
anybody else should feel free to ask me if they want the lecture.
Anyway...)
The basic idea of elimiating phi nodes is simple enough. We replace
each phi node with a temporary variable. Uses of the phi use values
loaded from the temporary. The operation of the phi itself
(assigning a value based on the branch taken) amounts to an assignment
into the temporary.
Previously, the Slang compiler dealt with phi nodes very late in
the process of generating code: in the middle of emitting strings
of source code in a high-level language like HLSL or GLSL. Doing the
work that late in compilation has two big drawbacks:
1. Our ability to emit clean and/or optimal code is limited because
we may not be able to make certain changes to the IR, or because we
cannot make use of additional information like a dominator tree that
might be available at other points in compilation.
2. Any other IR passes that relate to temporary variables won't be
able to see the variables that we generate for phi nodes. This could
raise issues with correctness (e.g., if we want to compute live-range
information for *all* temporary variables), or performance (we have no
way to run additional IR optimization passes after phis are eliminated).
This change addresses these problems by making the elimination of
phi nodes an explicit IR pass. Additional optimizations can easily be
run after this pass (although we'd need to be careful not to run
passes that could end up introducing new phis). The pass makes use
of the information available to it to try to produce code that will
emit to "clean" HLSL/GLSL.
The core of the pass is in `slang-ir-eliminate-phis.cpp`, and is
heavily commented, so I won't describe the approach in detail here.
There are two related issues that came up, though:
First, it turned out that our emit logic for local variables (`IRVar`
instructions) wasn't using the function we'd defined named `emitVar()`.
One worrying consequence of that oversight was that the `precise`
modifier would impact generated HLSL/GLSL for variables that turned
into SSA values (including phi nodes), but *not* for local variables
that had not been SSA'd (or that had been SSA'd and then de-SSA'd).
This change also fixes that bug; it is unclear how widespread the
impact of the original issue might be.
Second, generating explicit IR temporaries for phi nodes exposed a
pre-existing bug in the `slang-ir-restructure-scoping` pass. That pass
basically detects cases where we have an instruction `I` with a use
`U` such that the use follows the rules of SSA form ("def dominates
use," meaning `I` dominations `U`), but does not follow the more
restrictive scoping rules of high-level-language output (where a value
computed "inside" a loop is not automatically visible to code outside
the loop just because it dominates that code). That pass did not
correctly account for the case where `I` was a temporary variable.
It seems that case could not arise before now because we didn't have
any passes that would move `var`, `load`, or `store` operations out
of the basic block they started in. The fix for that pass was relatively
simple, and will make the whole thing more robust in case we add more
aggressive optimizations later.
* fixup: expected test output
|
| |
|
|
|
|
|
| |
* #include an absolute path didn't work - because paths were taken to always be relative.
* Fix for overloaded name lookup.
* Small improvements.
|
| |
|
|
|
|
|
|
|
| |
* #include an absolute path didn't work - because paths were taken to always be relative.
* Meta-2 test works.
* Add new generic test for static const variable in a function in a generic.
* Generic function with static const variable doesn't work.
|
| |
|
|
|
| |
* #include an absolute path didn't work - because paths were taken to always be relative.
* Added sample-grad-clamp-lod sample.
|
| |
|
|
|
|
|
|
|
|
|
|
| |
* #include an absolute path didn't work - because paths were taken to always be relative.
* Fix for = {} initialization with a field that is generic type parameter.
* Handling for if a non type is passed to a generic parameter which requires a type.
* Small comment improvements.
Fix some tab issues.
* This fixes the matrix.slang issue. Move the matrix.slang test into bugs as generic-default-matrix.slang
|
| |
|
|
| |
`ImageSubscript` for GLSL (#2146)
|
| |
|
|
|
| |
* #include an absolute path didn't work - because paths were taken to always be relative.
* Fix bool handling in constant folding for generic parameters.
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Fixes #1990
The underlying problem here is in the `ExtractExistentialType` AST node class.
An "existential" in current Slang is typically a value of interface type. When such a value is used in an operation, the type-checker "opens" the extistential so that subsequent type-checking steps can work with the (statically unknown) specific type of the value stored inside. The `ExtractExistentialType` AST node represents the type of an existential that has been "opened" in this way.
When the front-end performs lookup "into" a value with one of these types, it nees to use a reference to the original interface declaration with a "this-type substitution" that refers to the "opened" type (a this-type substitution tells the compiler the concrete type it should use in place of `This` in signatures within the interface; it allows compiler to "see" the right associated type definitions to use in a context).
Prior to this change, the implementation would store the specialized reference to the original interface declaration in the `ExtractExistentialType` node as part of its state. The catch there is that the specialized interface reference indirectly refers to the `ExtractExistentialType` AST node itself, creating a circularity. As soon as the front-end performs any operation that tries to recurse over that structure, it would go into an infinite loop.
The fix here sounds kind of like a hack, but seems to be pretty nice in practice. Instead of always storing the specialized interface reference, we instead store the few values that are needed to construct it, and then create and cache the actual reference on-demand. The on-demand created fields are not considered part of the state of the AST node for any kind of recursion or serialization, so they avoid the original problem.
A single test case was added that represents the original bug, and confirms the fix.
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
* Format list updated with additional formats supported by both D3D and Vulkan; D3DUtil::getMapFormat() and VkUtil::getVkFormat() updated to include additional formats; GFX_FORMAT() updated with all additional formats (BC compression unfinished)
* Finished updating GFX_FORMAT with newly added formats and sizes; Pixel size is now tracked using the FormatPixelSize struct containing the values for bytes per block and pixels per block to accomodate BC formats; Updated gfxGetFormatSize and associated sub-calls to return FormatPixelSize instead of uint8_t; Most calls to gfxGetFormatSize() updated to reflect changes, a couple calls still unupdated
* Changes to accommodate new formats finished, debugging slang-literal unit test
* First format unit test working
* One test added for BC1Unorm and RGBA8Unorm_SRGB, both passing
* Refactored format testing code to merge BC1Unorm and RGBA8Unorm SRGB into a single file
* All unit tests added for BC and Srgb formats
* Most tests added and working; Added five additional formats (still need tests) and made the appropriate changes to support these; createTextureView() modified for D3D11, D3D12, and Vulkan to take into account the format specified in the texture view desc when the texture's format is typeless
* Format enums renamed to more closely match their D3D counterparts; Added a universal float and uint buffer and buffer view for use across all Format tests
* Remaining tests added; D3D12 tests pass, but Vulkan crashes in BC1_UNORM and D3D11 spits out a bunch of D3D11 Errors (but supposedly passes)
* re-run premake
* Added Sint versions of test shaders; Vulkan and D3D11 tests also pass
* Size struct for format unit tests no longer use initializer lists
* Fixed a Size struct missed in the previous pass
* Fixed minor bugs causing tests to fail
* Added documentation detailing all currently unsupported formats
* Skip tests causing unsupported format warnings due to swiftshader
* updated several test using old Format enum names
* Revert change to compareComputeResult() that was added for debugging purposes
* DEBUGGING: Added prints to identify which formats are failing on CI
* Reverted attempted debugging changes; Fixed texture2d-gather.hlsl to use updated Format enums
* Fixed incorrect array sizes in d3d11 _initSrvDesc()
* Commented out further tests that produce unexpected results when tested for Vulkan with swiftshader
* Revert "Merge branch 'expanded-format-support' of https://github.com/lucy96chen/slang into expanded-format-support"
This reverts commit 20008f0d3ecc3b1405ecac8c138edaa3cd37ed6b, reversing
changes made to 6081e95827315fee50e18409394d5abd62fac787.
* Added a fuzzy comparison function for use with floats
* submodule update
* Revert messed up changes caused by previous revert after automatically merging on github
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
* Work to mitigate SPIR-V bloat
SPIR-V is not an especially compact format, but some patterns in how Slang generates code and then runs it through `spirv-opt` lead to many redundant field-by-field copy operations being emitted. This change attempts to address some of the resulting bloat from the Slang side of things.
Note: experimentation shows that the bloat is less pronounced when running either *no* SPIR-V optimizations or *full* SPIR-V optimizations, so it is also likely that the bloat should be addressed by changing which `spirv-opt` passes the Slang compiler runs in default (`-O1`) builds. Such changes should come as a distinct pull request.
This change primarily does two things:
First, the code generation strategy for passing arguments to `out` and `inout` parameters has been changed. In the past, the compiler would *always* copy the argument value into a temporary, then pass the address of the temporary, and then write back the value after the call. The new code generation strategy attempts to identify when an argument value already has a simple address in memory and passes that address directly when possible. This eliminates many copy operations that occur before/after calls to functions with `out`/`inout` parameters.
Second, we introduce an IR optimization pass that detects call sites where the entire contents of a buffer (usually a constant buffer) is being passed to a callee function, such that many bytes are loaded and then passed even if only very few are used in the callee. The pass moves the load operations from the caller to a specialized version of the the callee where possible (e.g., when the constant buffer in question is a global shader parameter). Doing this eliminates another major category of copies.
Notes:
* The IR lowering logic is complicated by the fact that several kinds of l-values (values that are usable as the desitnation of assignment, or for `out`/`inout` arguments) are not actually addressable. An easy example is a non-contiguous swizzle like `v.xwz` on a `float4`, where the value occupies 12 bytes, but not 12 consecutive bytes with a single address. There are many more corner cases like that and the IR lowering pass carries a lot of complexity to deal with them. A more systematic overhaul is due some time soon.
* The IR representation of `out` and `inout` parameters deserves some careful scrutiny when making these kinds of changes. The official semantics of `inout` in HLSL has been "copy in copy out" (and `out` is just "copy out") which is observably different from any solution that passes in the address of an l-value directly. By making this change we are saying that Slang's semantics are not precisely those of legacy HLSL, and that our semantics for `inout` parameters are closer to those of `inout` in Swift or of a mutable borrow in Rust. In the Swift case the implementation can freely pass the underlying storage of an l-value or the address of a temporary, and valid programs may not observe the different. It is thus illegal to observe the value in a storage local while a mutation to that location is "in flight." All of this is way more detailed and technical than 99% of Slang users will ever care about, but importantly it gives us semantic cover to eliminate these copies in the IR, and also to emit output C++ code that implements `out` and `inout` as by-reference parameter passing.
* There was an exsting generic pass for specializing functions based on call sites that uses a "template method" style of pattern to customize its behavior. That pass needed to be generalized to handle this use case because it had previously operated on the assumption that the "desire" to specialize a callee function must be driven by the parameter declarations of that function, and not on the argument values passed in. The code has been slightly refactored to allow the policy for specialization to consider both parameters and arguments.
* Unsurprisingly, a bunch of the GLSL (and thus SPIR-V) generated has changed with this work, so several baseline `.slang.glsl` files needed to be updated.
* This change is incomplete in that it does not address broader cases of buffer loads, including both partial loads from constant buffers (just loading one field, but a field that uses a "large" structure type), and loads from multi-element buffers (a lot from a structured buffer where the element type is "large"). The main question in each of those cases is how to define how "large" a structure needs to be before we decide to try and sink loads into callee functions like this. In the worst case, sinking loads in this way may actually create *more* memory traffic (because the same values get loaded in multiple callee functions).
* fixup: run premake
* fixup: typo
|
| | |
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
* Include a "stack trace" with nested-import errors
When errors occur in nested `#include` files it is often helpful to have a "stack trace" / traceback of the `#include` chain that led from a root translation unit to the file with an error.
This change implements a similar feature for `import`s.
It is worth noting that `import`s don't really *require* this kind of compiler support the way `#include`s do because the intention is that the meaning of an `import`ed file does not depend on the order or nesting of `import`s. As such, when trying to *fix* an error in an `import`ed file, you usually don't care how it came to be `import`ed into your shaders.
The use case here is somebody adapting a large body of Slang code to use in a different codebase, such that they have certain `.slang` files they don't actually intend to have compile correctly, and they want to be able to diagnose how they came to include those files when/if they cause problems.
The actual feature implementation is pretty simple because we already track a stack of active `import`s so that we can detect and diagnose recursive `import`s. This change simply changes the disagnostics when there is an error in imported code so that instead of just noting the inner-most `import` site it lists all the `import` sites that were active at the time.
The change includes a test case to confirm that the behavior works (at least for the case of a parse error).
* fixup: test outputs
Co-authored-by: Yong He <yonghe@outlook.com>
Co-authored-by: jsmall-nvidia <jsmall@nvidia.com>
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
* Overhaul the preprocessor
The old Slang preprocessor was based on a simple mental model that tried to unify two parts of macro expansion:
* Scanning for macro invocations in a sequence of tokens
* Producing the expanded tokens for a macro expansion by substituting arguments into its body
The basic was that substitution of macro arguments into a macro definition is superficially similar to top-level macro expansion, just with an environment where the macro arguments act like `#define`s for the corresponding parameter names. That approach was "clever" and could conceivably have been extended to include a lot of advanced preprocessor features (e.g., a preprocessor-level `lambda` would be easy to support!), but it was basically impossible to make it correctly handle all the corner cases of the full C/C++ preprocessor.
The fundamental problem with the old approach was that it conflated the two parts of expansion listed above into one implementation, while the various special cases of the C/C++ preprocessor rely on treating the two cases very differently. The new approach here (which is somewhere between a refactor and a full rewrite of the preprocessor) changes things up in a few key ways:
* The abstraction still cares a lot about streams of tokens, but it now treats the top level streams (`InputFile`s) as fairly different from the lower-level streams (`InputStream`s)
* Macro expansion is handled as a dedicated type of stream that wraps another stream. This allows macro expansion to be applied to anything, and supports cases where multiple rounds of macro expansion are required by the spec.
* Macro *invocations* and the substitution of their arguments are now handled by a completely new system.
* Macro arguments are no longer treated as if they were `#define`s
* The macro body/definition is analyzed at definition time to detect various kinds of issues, and to derive a list of "ops" that make it easier to "play back" the definition at substitution time
* Token pasting and stringizing are now only handled in macro definitions (rather than being allowed anywhere), and their use cases are restricted to only those that make sense (e.g., you can't stringize anythign except a macro parameter, because anything else wouldn't make sense)
The key new types here are the `ExpansionInputStream` which handles scanning for macro invocations, and the `MacroInvocation` type, which handles playing back the macro body with substitutions.
The `ExpansionInputStream` is the easier of the two to understand. By refactoring it to use a single token of lookahead, the one major detail it had to deal with before (abandoning expansion of a function-like macro if the macro name was not followed by `(`) is significantly easier to manage.
The more subtle part is the `MacroInvocation` type, and most of the complexity there is around handling of token pasting, and the fact that either or both of the operands to a token paste might be empty.
Many of the test cases that exposed the problems in the preprocessor have been moved from `current-bugs` to `preprocessor` since they now work correctly.
* debugging: enable extractor command line dump
* fixup
* fixup
|
| |
|
|
|
|
|
|
|
|
|
| |
* #include an absolute path didn't work - because paths were taken to always be relative.
* Pass LexerFlags to all lexing functionality that may output a diagnostic.
* Add test for lexing disabling issue.
* Improve tests.
Co-authored-by: Yong He <yonghe@outlook.com>
|
| |
|
|
|
|
|
|
|
| |
* #include an absolute path didn't work - because paths were taken to always be relative.
* Fix bug causing infinite loop in parser.
* Add a faster path for LookAheadToken when the offset is 0.
* Fix typo introduced in last commit.
|
| |
|
|
|
|
|
|
|
| |
* #include an absolute path didn't work - because paths were taken to always be relative.
* Made vector init combinations explicit.
* Add test for vector initialization.
* Small comment change - really just to retrigger TC.
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
This change originated as an attempt to re-enable a test case, but it has ended up disabling more tests (for good reasons) than it re-enables.
The main change here is a significant overhaul of the way that the D3D12 render path extracts information from the Slang reflection API to produce a root signature.
There were also some supporting fixes in the reflection information to make sure it returns what the D3D12 back-end needed.
The big picture here is that the D3D12 path now uses the descriptor ranges stored in the reflection data more or less directly.
It still needs to use register/space offset information queried via the "old" reflection API, but it only does so at the top level now, for the program and entry points themselves.
All other layout information is derived directly from what Slang provides.
Smaller changes:
* The "flat" reflection API was expanded to include `getBindingRangeDescriptorRangeCount()` which was clearly missing.
* The "flat" reflection results for a constant buffer or parameter block that didn't contain any uniform data and was mapped to a plain constant buffer needed to be fixed up. That logic is still way to subtle to be trusted.
* Several additional tests were disabled that relied on static specialization, global/entry-point generi type parameters, structured buffers of interfaces or other features we don't officially support with shader objects right now. All of the affected tests were somehow passing by sheer luck and because they often passed in specialization arguments via explicit `TEST_INPUT` lines.
* The `inteface-shader-param` test is re-enabled now that we can properly describe its input with the new `set` mode on `TEST_INPUT`
* `ShaderCursor::getElement()` can now be used on structure types (in addition to arrays) to support by-index access to fields
* The `TEST_INPUT` system was expanded to support both by-name and by-index setting of structure fields for aggregates
* The `TEST_INPUT` system was expanded to allow an `out` prefix to mark parts of an expression as outputs on a `set` lines
* The `TEST_INPUT` system was expanded so that anything that would be allowed on a `TEST_INPUT` line by itself (like `ubuffer(...)`) can now be used as a sub-expression on a `set` line
Co-authored-by: Yong He <yonghe@outlook.com>
|
| |
|
|
|
|
|
|
|
|
|
|
|
| |
* Refactor `gfx` to surface `CommandBuffer` interface.
* Fixes.
* Fix code review issues, and make vulkan runnable on devices without VK_EXT_extended_dynamic_states.
* Update solution files
* Move out-of-date examples to examples/experimental
Co-authored-by: Yong He <yhe@nvidia.com>
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
| |
* #include an absolute path didn't work - because paths were taken to always be relative.
* First pass at handling 'names' that are too long in GLSL output.
* Test to check functionality with very long func name.
* Add access a long names buffer.
* Fix typo in assert.
Fix issue with coercion error for 1.0f / 0x7fffffff
Co-authored-by: Tim Foley <tfoleyNV@users.noreply.github.com>
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
* #include an absolute path didn't work - because paths were taken to always be relative.
* WIP: First pass in supporting output of line error information.
* Add support for lexing to better be able to indicate SourceLocation information.
* Fix lexer usage in DiagnosticSink in C++ extractor.
* Update diagnostics tests to have line location info.
* Fixed test expected output that now have source location information in them.
* Better handling of tab.
* Fix test expected results for tabbing change.
* DiagnosticLexer -> DiagnosticSink::SourceLocationLexer
Added line continuation tests.
* Fix typo.
* Added String::appendRepeatedChar
* Change to rerun tests.
Co-authored-by: Tim Foley <tfoleyNV@users.noreply.github.com>
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
* #include an absolute path didn't work - because paths were taken to always be relative.
* Work around for issue with obfuscation (and lack of name hints) leading to names in output not being correctly uniquified.
* Improve appendChar
Remove unrequired memory juggling to scrub names.
* Remove test code.
* Small fixes in comments and method called.
* Remove linkage decoration on functions that are specialized.
* Obfuscation naming with specialization test.
* Fix instruction deletion.
Co-authored-by: Tim Foley <tfoleyNV@users.noreply.github.com>
|
| |
|
|
|
|
|
| |
* Fix existential specialization of mutable buffer loads.
* fix
Co-authored-by: Yong He <yhe@nvidia.com>
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
This change converts a large number of our existing tests to use the `ShaderObject` support that was added to the `gfx` layer.
In many cases, tests were just updated to pass `-shaderobj` and the result Just Worked.
In other cases, a `name` attribute had to be added to one or more `TEST_INPUT` lines.
For tests that did not work with shader objects "out of the box," I spent a little bit of time trying to get them work, but fell back to letting those tests run in the older mode.
Future changes to the infrastructure will be needed to get those additional tests working in the new path.
Along with the changes to test files, the following implementation changes were made to get additional tests working:
* Because the shader object mode uses explicit register bindings (from reflection), the hacky logic that was offseting `u` registers for D3D12 based on the number of render targets gets disabled (by another hack).
* The "flat" reflection information coming from Slang was not correctly reporting "binding ranges" for things that consumed only uniform data (which would be everything on CUDA/CPU), so it was refactored to properly include binding ranges for anything where the type of the field/variable implied a binding range should be created (even if the `LayoutResourceKind` was `::Uniform`).
* A few fixes were made to the CUDA implementation of `Renderer`, in order to get additional tests up and running. Most of these changes had to do with texture bindings, which hadn't really been tested previously.
In addition, a few changes were made that were attempts at getting more tests working, but didn't actually help. These could be dropped if requested:
* As a quality-of-life feature (not being used) the `object` style of `TEST_INPUT` line is upgraded to support inferring the type to use from the type of the input being set.
* Any `object` shader input lines get ignored in non-shader-object mode.
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
* Use "capability" system to select VKRT extension
Slang currently supports translation of ray tracing shader code to Vulkan GLSL code that uses the `GL_NV_ray_tracing` extension. A multi-vendor equivalent of that extension has been released as `GL_EXT_ray_tracing` and we want Slang to support that extension as well.
At the simplest, making the change from one extension to the other is just a matter of changing a few strings, since it does not appear that anything of significance was changed at the GLSL level (or even in SPIR-V). Where this gets trickier is when we have users who want us to support *both* extensions, and to be able to switch between them.
The solution we've implemented here more or less amounts to:
* If you don't tell the compiler which extension to use, it will default to `GL_EXT_ray_tracing` (the newer multi-vendor one).
* If you explicitly want the older extension, you can opt into it using the `-profile` option or via a new API for explicitly adding capabilities to your target.
Making that work required a few different kinds of changes:
* The options parsing and public API needed ways to add optional capabilities to a target.
* During GLSL code emit, we can check the capabilities that were added to the target to see if the `GL_NV_ray_tracing` extension was explicitly enabled and, if not, default to using the `GL_EXT_ray_tracing` names for things. This step is needed because some of the modifiers/attributes involved in the extension have to be handled explicitly in the code generator rather than implicitly as part of mapping intrinsic functions.
* We add two different translations to the relevant operatiosn in the stdlib, one marked with each of the extensions. If profile/capability-based overload resolution can be relied on to pick the right one, this should Just Work.
* Next, a bunch of work had to go into making capability-based overloading Just Work for the purposes of this change. There's been a nearly complete reworking of the implementation of `CapabilitySet` here to make it more suitable for our needs.
* The tests that were using ray tracing translation for Vulkan needed to be updated. For some of them I updated their baselines to use `GL_EXT_ray_tracing` so that they can test the new path. For others, I updated the command line for the test case so that it explicitly opts into using `GL_NV_ray_tracing`. The result is that we have some coverage of each extension. I would have liked to have each test run in both modes, but our pass-through glslang support doesn't support `-D` options, so I couldn't take that step easily.
This change does *not* add support for `GL_EXT_ray_query`, the extension that supports "DXR 1.1" style queries under Vulkan. Adding support for that extension should hopefully be a smaller step because it doesn't have the same multiple-extensions issue.
This change does *not* address a lot of possible avenues for improvement or cleanup around the capability system. It focuses only on those changes that are necessary to make the ray tracing feature work and leaves the rest for future work.
* fixup: infinite loop
* Comment-only change to retrigger TC build
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
* "Shader Toy" example and related fixes
This change introduces a new `shader-toy` example program that is primarily designed to show how Slang's features for type-based encapsulation and modularity can be applied to modularity for effects along the lines of those from `shadertoy.com`.
The Example
-----------
The example is being checked in with an example "toy" effect that I hastily put together, so that it would not be encumbered with any IP concerns. I wrote the effect using the shadertoy.com editor, so I can be sure it is valid GLSL. During bringup of the application I used a pre-existing and larger effect for testing, so some of the support code that was added is not being used at present.
The big-picture idea here is to have an exmaple that shows how to modularize things using Slang interfaces and generics, and then to use the Slang compiler API to manage the compilation, composition, specialization, and linking steps. For better or worse this leads to the sequence of API calls involved being much longer than what was in something like the `hello-world` example.
Future Work (Example)
---------------------
There is a lot of room for improvement and expansion here, so this should be viewed as a checkpoint of work in progress rather than something I'm claiming as a finalized demonstration of all we'd like to achieve. Areas for future work include:
* We need to copy the integration of "Dear, IMGUI" that was already done for the `model-viewer` example so that this example can have a UI.
* Now that the compilation flow is broken into all these additional steps, it should be possible to have the application load multiple effects as distinct modules, and then provide a UI for switching between them. The chosen effect module would be used to specialize the top-level shader(s) before kernel generation.
* The checked-in logic includes a compute shader that can execute an effect, but that hasn't been tested nor has it been wired up to any kind of UI. We should have a way to switch between multiple execution methods, with a goal of eventually including CPU execution.
* The "GLSL compatibility" code needs a lot of improvements before it is likely to be usable for a nontrivial number of shaders. Some of that work is waiting on Slang compiler fixes, though.
* We should consider allowing the individual "toy" effects to define their own uniform parameters and expose those via a UI and reflection. The catch in this case is not that this would be difficult to do, but that it would be a semantic change to how shader toy effects currently work.
The Compiler Fixes
------------------
Doing this work exposed a few bugs in Slang, and this change includes fixes for the ones that were quick to address.
We already had logic in `slang-check-shader.cpp` that was validating the entry points in a compile request - either by checking the explicitly-listed entry points, or by scanning for `[shader("...")]` attributes. The problem is that the routine that did that checking was not being invoked on all compiles. The logic that handled entry points was only being run for manual compiles using `SlangCompileRequest`, while anything using `import` or `loadModule` would ignore entry points. I refactored the relevant code into a subroutine that will be invoked in all compilation scenarios.
There were already `TODO` comments in `SpecializedComponentType` which made the point about how a specialized entry point like `myShader<YourType>` would need to properly show that it has dependencies on both the module that defines `myShader` *and* the module that defines `YourType`, while only the former was being handled at present. I went ahead and implemented the logic to scan the generic arguments for a specialized compoment type in order to determine what module(s) the arguments depend on (both type arguments and witness tables). With that change, using `IComponentType::link` on a specialized component will properly pull in the module(s) that the generic arguments come from.
In `slang-ir-legalize-types.cpp` we could run into assertion failures in debug builds because of code trying to legalize layout `IRAttr`s for fields or parameters with types that need legalization. In practice it is safe to skip these layout attributes, because legalization of the fields/parameters they pertain to would result in creation of entirely new layout attributes, and the old ones would then be unreferenced.
Future Work (Fixes)
-------------------
There are other compiler bugs that this work exposed, but which this change does not address. These will need to be resolved as part of subsequent changes:
* Slang allows for default-initialization of variables of a generic type. That is, given `<T : ISomething>` a user is allowed to declare `T x = {};` and the Slang front-end does not complain. Instead, this leads to an internal compiler error during IR lowering.
* The Slang `__init()` feature probably needs to be upgraded to a properly supported feature, and we probably need a way to make implementing default-initialization an easy thing (e.g., any `struct` type that has initial-value expressions for all its fields should automatically and implicitly satsify an `init();` requirement declared in an interface)
* Iniside an `__init()` definition, code has mutable access to members of the enclosing type, but for some reason the front-end is incorrectly treating `this` as immutable in those contexts. As a result you can write to `someField` but not `this.someField`.
* User-defined operator overloads flat out don't work (which isn't surprising given that no clients have decided to use them yet, and we have no test coverage for them). This is actually due to the shadowing rules being used for lookup right now, so a fix for this issue is going to have far-reaching consequences around what overloads are visible where (and anything that impacts overload resolution is a big can of worms, including around performance).
* fixup: test case had missing main function
|
| |
|
|
|
|
|
|
|
| |
* Add shader object parameter binding to renderer_test.
* remove multiple-definitions.hlsl
* Fix cuda implementation.
Co-authored-by: Tim Foley <tfoleyNV@users.noreply.github.com>
|
| |
|
|
|
|
|
| |
Slang generates code that turns the implicit `this` parameter of a method into an explicit parameter. The logic that decides whether that parameter should be `inout` is a bit involved, and there was a bug where a generic method would lead to the use of an `in` modifier (the default) and override the `inout` modifier that was requested by the method itself.
This change fixes the logic to treat generic declarations in the parent chain of a leaf method as having no bearing on whether an implicit `this` parameter should be `inout` or not.
A test case is included that breaks with the old behavior, and demonstrates that a generic `[mutating]` method can now work correctly.
|
| |
|
|
|
|
|
| |
The basic problem here is that when a function has multiple declarations with matching signatures (e.g., a forward declaration and then a later definition with a body), the IR lowering logic would lower all declarations whenever the first one was encountered, but then would only register an IR value as the lowered version of the first declaration. Other matching declarations would then run the risk of being lowered again, and in the case where they included features like loops with break/continue labels, that would create the risk of keys getting inserted into certain dictionaries more than one, leading to exceptions.
This change ensures that when lowering a function that has multiple matching declarations to IR, we register an IR value for all of those declarations and not just the first.
I have added a test case that leads to a crash without this change, to ensure that we don't introduce a regression down the line.
|
| |
|
|
|
| |
* Enable all dynamic dispatch tests on CUDA.
* Fix expected cross-compile test results.
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
During semantic checking, the compiler used to link together `ExtensionDecl`s into a singly-linked list dangling off of the `AggTypeDecl` that they applied to. This approach made lookup relatively easy, because given a `DeclRef` to an `AggTypeDecl` one could easily find and walk the list of candidate extensions.
Unfortunately, the simple approach has two major strikes against it:
* First, as we recently ran into, it creates a lifetime/ownership problem, in cases where the `ExtensionDecl` is outlived by the `AggTypeDecl` it applies to. This creates the one and only place in the compiler today where an "old" AST node might point to a "new" AST node, and it resulted in use-after-free problems in client code.
* Second, the scoping of `extension`s ends up being completely wrong. All of the `extension` methods on a type end up being visible in all cases, instead of just in the context of modules where the `extension` itself is visible. The comparable feature in C# (static extension methods) is careful to not make scoping mistakes like this. The Swift langauge has loose scoping for `extension` more akin to what we have in Slang today, but the maintainers seem to consider it a misfeature.
This change attempts to clean up both issues by changing the way that extension declarations are stored. There are two main pieces:
1. The primary "source of truth" for extension lookup has been moved to the `ModuleDecl`, where a module is responsible for storing a cache of the extensions declared within that module (keyed by the declaration of the type being extended). This cache is updated at the same point where the old code would mutate the AST node being depended on.
2. A secondary aggregated cache is added to the `SharedSemanticsContext` used during semantic checking. This cache includes entries from across multiple modules, and is intended to be invalidated and rebuilt on demand if new modules are added during checking.
Access to the candidate extensions has now been put behind subroutines that require a semantics-checking context to be passed in (there was always one available in contexts that care about extensions).
In addition, the operation for looking up members including those from extensions was refactored heavily to involve internal rather than external iteration and, more importantly, was changed so that it actually tests whether the `ExtensionDecl`s it loops over apply to the type in question, rather than blindly letting extensions members be looked up in ways that don't make sense.
There are three test cases added here to confirm aspects of the fix:
* First, I added a test that reproduces the crash that was being seen, so that we have a regression test for the fix.
* Second, I added a basic semantic-checking test to confirm that an `extension` from an `import`ed module is still visible/usable, to confirm that I didn't break existing valid uses of extensions.
* Third, I added a diagnostic test that ensures that we correctly ignore extensions that should not be visible in a given context as a result of `import` declarations.
Co-authored-by: jsmall-nvidia <jsmall@nvidia.com>
|
| |
|
|
|
|
|
|
|
| |
contains an array (#1448)
* This code is disabled, it was part of the optimization `Specialize function calls involving array arguments. (#1389)` on github.
It is disabled here because it causes a problem when a struct is passed to a function that contains a structured buffer *and* an array. It is specialized on the struct type, and so those types become parameters to the function. If the struct contains a structured buffer this is a problem on GLSL/VK based targets because currently structured buffers cannot be function parameters.
The fix for now is to just disable this optimization.
* Fix typo in name of test expected values.
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
If somebody defines two `struct` types with the same name:
```hlsl
struct A {}
// ...
struct A {}
```
and then tries to use that name when specializing a generic function:
```hlsl
void doThing<T>() { ... }
// ...
doThing<A>();
```
then the Slang front-end currently crashes, which leads to it not diagnosing the original problem (the conflicting declarations of `A`).
This change fixes up the checking of generic arguments so that it properly fills in dummy "error" arguments in place of missing or incorrect arguments, and thus guarantees that the generic substitution it creates will at least be usable for the next steps of checking (rather than leaving null pointers in the substitution).
This change also fixes up the error message for the case where a generic application like `F<A>` is formed where `F` is not a generic. We already had a more refined diagnostic defined for that case, but for some reason the site in the code where we ought to use it was still issuing an internal compiler error around an unimplemented feature.
This chagne includes a diagnostic test case to cover both of the above fixes.
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
The Slang compiler was bit by a known issue when translating from SSA form back to straight-line code. Give code like the following:
int x = 0;
int y = 1;
while(...)
{
...
int t = x;
x = y;
y = t;
}
...
The SSA construction pass will eliminate the temporary `t` and yield code something like:
br(b, 0, 1);
block b(param x : Int, param y : Int):
...
br(b, y, x);
The loop-dependent variables have become parameters of the loop block, and the branchs to the top of the loop pass the appropriate values for the next iteration (e.g., the jump that starts the loop sends in `0` and `1`).
The problem comes up when translating the back-edge the continues the loop out of SSA form. Our generated code will re-introduce temporaries for `x` and `y`:
int x;
int y;
// jump into loop becomes:
x = 0;
y = 1;
for(;;)
{
...
// back-edge becomes
x = y;
y = x;
continue;
}
The problem there is that we've naively translated a branch like `br(b, <a>, <b>)` into `x = <a>; y = <b>;` but that doesn't work correctly in the case where `<b>` is `x`, because we will have already clobbered the value of `x` with `<a>`.
The simplest fix is to introduce a temporary (just like the input code had), and generate:
// back-edge becomes
int t = x;
x = y;
y = t;
This change modifies the `emitPhiVarAssignments()` function so that it detects bad cases like the above and emits temporaries to work around the problem. A new test case is included that produced incorrect output before the change, and now produces the expected results.
A secondary change is folded in here that tries to guard against a more subtle version of the problem:
for(...)
{
...
int x1 = x + 1;
int y1 = y + 1;
x = y1;
y = x1;
}
In this more complicated case, each of `x` and `y` is being assigned to a value derived from the other, but neither is being set using a block parameter directly, so the changes to `emitPhiVarAssignments()` do not apply.
The problem in this case would be if the `shouldFoldInstIntoUseSites()` logic decided to fold the computation of `x1` or `y1` into the branch instruction, resulting in:
x = y + 1;
y = x + 1;
which would again violate the semantics of the original code, because now there is an assignment to `x` before the computation of `x + 1`.
Right now it seems impossible to force this case to arise in practice, due to implementation details in how we generate IR code for loops. In particular, the block that computes the `x+1` and `y+1` values is currently always distinct from the block that branches back to the top of the loop, and we do not allow "folding" of sub-expressions from different blocks. It is possible, however, that future changes to the compiler could change the form of the IR we generate and make it possible for this problem to arise.
The right fix for this issue would be to say that we should introduce a temporary for any branch argument that "involves" a block parameter (whether directly using it or using it as a sub-expression). Unfortunately, the ad hoc approach we use for folding sub-expressions today means that testing if an operand "involves" something would be both expensive and unwieldy.
A more expedient fix is to disallow *all* folding of sub-expressions into unconditional branch instructions (the ones that can pass arguments to the target block), which is what I ended up implementing in this change. Making that defensive change alters the GLSL we output for some of our cross-compilation tests, in a way that required me to update the baseline/gold GLSL.
A better long-term fix for this whole space of issues would be to have the "de-SSA" operation be something we do explicitly on the IR. Such an IR pass would still need to be careful about the first issue addressed in this change, but the second one should (in principle) be a non-issue given that our emit/folding logic already handles code with explicit mutable local variables correctly.
|
