Working towards a full e-graph implementation as described by the [egg
paper](https://arxiv.org/pdf/2004.03082).
The type system has a couple of places where e-graphs would've been
useful and solved some classes of problems trivially. For example:
1. Normalization and simplification cannot handle cyclic types due to
the nature of their implementation.
2. Normalization can't tell when two tables or functions are equivalent,
but simplification theoretically can albeit not implemented.
3. Normalization requires deep normalization for inhabitance check,
whereas simplification would've returned the `never` type itself
indicating uninhabited.
4. Simplification requires constraint ordering to have perfect timing to
simplify.
5. Adding a rewrite rule requires implementing it twice, once in
simplification and once again in normalization with completely different
code design making it hard to verify that their behavior is materially
equivalent.
6. In cases where we must cache for performance, two different types
that are isomorphic have different cache entries resulting in cache
misses.
7. Type family reduction can handle cyclic type families, but only if
the cycle is not obscured by a different type family instance. (`t1
where t1 = union<number, add<t1, number>>` is irreducible)
I think we're getting the point!
---
Currently the implementation is missing a few features that makes
e-graphs actually useful. Those will be coming in a future PR.
1. Pattern matching,
6. Applying rewrites,
7. Rewrite until saturation, and
8. Extracting the best e-node according to some cost function.
# What's Changed?
- Performance improvement in the old solver
- Bugfixes in the new solver
## Old Solver
- Mark types that do not need instantiation when being exported to
prevent unnecessary work from being done
## New Solver
- Refactored instances of "type family" with "type function"
- Index-out-of-bounds bug fix in the resolution resolver
- Subtyping reasonings are merged only if all failed
---
### Internal Contributors
Co-authored-by: Aaron Weiss <aaronweiss@roblox.com>
Co-authored-by: Vighnesh Vijay <vvijay@roblox.com>
Co-authored-by: Vyacheslav Egorov <vegorov@roblox.com>
---------
Co-authored-by: Aaron Weiss <aaronweiss@roblox.com>
Co-authored-by: Alexander McCord <amccord@roblox.com>
Co-authored-by: Andy Friesen <afriesen@roblox.com>
Co-authored-by: Vighnesh <vvijay@roblox.com>
Co-authored-by: Aviral Goel <agoel@roblox.com>
Co-authored-by: David Cope <dcope@roblox.com>
Co-authored-by: Lily Brown <lbrown@roblox.com>
Co-authored-by: Vyacheslav Egorov <vegorov@roblox.com>
The description of SUBRK/DIVRK is out of sync with the VM code.
Comment:
```cpp
// SUBRK, DIVRK: compute arithmetic operation between the constant and a source register and put the result into target register
// A: target register
// B: source register
// C: constant table index (0..255); must refer to a number
LOP_SUBRK,
LOP_DIVRK,
```
VM snippet:
```cpp
VM_CASE(LOP_DIVRK)
{
Instruction insn = *pc++;
StkId ra = VM_REG(LUAU_INSN_A(insn));
TValue* kv = VM_KV(LUAU_INSN_B(insn));
StkId rc = VM_REG(LUAU_INSN_C(insn));
...
```
Right now, the commentLocations are not transferred over to the
SourceModule for definition files, like they are for normal source
modules. This means we lose out on finding documentation comments. We
add that in here
https://github.com/luau-lang/luau/issues/1137#issuecomment-2212413633
# What's Changed?
- Mostly stability and bugfixes with the new solver.
## New Solver
- Typechecking with the new solver should respect the no-check hot
comment.
- Record type alias locations and property locations of table
assignments
- Maintain location information for exported table types
- Stability fixes for normalization
- Report internal constraint solver errors.
---
### Internal Contributors
Co-authored-by: Aaron Weiss <aaronweiss@roblox.com>
Co-authored-by: Alexander McCord <amccord@roblox.com>
Co-authored-by: Andy Friesen <afriesen@roblox.com>
Co-authored-by: Vighnesh Vijay <vvijay@roblox.com>
Co-authored-by: Vyacheslav Egorov <vegorov@roblox.com>
---------
Co-authored-by: Aaron Weiss <aaronweiss@roblox.com>
Co-authored-by: Alexander McCord <amccord@roblox.com>
Co-authored-by: Andy Friesen <afriesen@roblox.com>
Co-authored-by: Aviral Goel <agoel@roblox.com>
Co-authored-by: David Cope <dcope@roblox.com>
Co-authored-by: Lily Brown <lbrown@roblox.com>
Co-authored-by: Vyacheslav Egorov <vegorov@roblox.com>
# What's Changed?
- Fix#1137 by appropriately retaining additional metadata from
definition files throughout the type system.
- Improve Frontend for LSPs by appropriately allowing the cancellation
of typechecking while running its destructor.
## New Solver
- Added support for the `rawget` type function.
- Reduced overall static memory usage of builtin type functions.
- Fixed a crash where visitors could mutate a union or intersection type
and fail to invalidate iteration over them in doing so.
- Revised autocomplete functionality to not rely on a separate run of
the type solver when using the new solver.
- Implemented a more relaxed semantic rule for casting.
- Fixed some smaller crashes in the new solver.
## Native Code Generation
- Add additional codegen specialization for `math.sign`
- Cleaned up a large number of outstanding fflags in the code.
### Internal Contributors
Co-authored-by: Aaron Weiss <aaronweiss@roblox.com>
Co-authored-by: Alexander McCord <amccord@roblox.com>
Co-authored-by: Andy Friesen <afriesen@roblox.com>
Co-authored-by: James McNellis <jmcnellis@roblox.com>
Co-authored-by: Jeremy Yoo <jyoo@roblox.com>
Co-authored-by: Vighnesh Vijay <vvijay@roblox.com>
Co-authored-by: Vyacheslav Egorov <vegorov@roblox.com>
---------
Co-authored-by: Alexander McCord <amccord@roblox.com>
Co-authored-by: Andy Friesen <afriesen@roblox.com>
Co-authored-by: Vighnesh <vvijay@roblox.com>
Co-authored-by: Aviral Goel <agoel@roblox.com>
Co-authored-by: David Cope <dcope@roblox.com>
Co-authored-by: Lily Brown <lbrown@roblox.com>
Co-authored-by: Vyacheslav Egorov <vegorov@roblox.com>
### What's new
* Added lint warning for using redundant `@native` attributes on
functions inside a `--!native` module
* Improved typechecking speed in old solver for modules with large types
### New Solver
* Fixed the length type function sealing the table prematurely
* Fixed crashes caused by general table indexing expressions
### VM
* Added support for a specialized 3-argument fast-call instruction to
improve performance of `vector` constructor, `buffer` writes and a few
`bit32` methods
---
### Internal Contributors
Co-authored-by: Aaron Weiss <aaronweiss@roblox.com>
Co-authored-by: Andy Friesen <afriesen@roblox.com>
Co-authored-by: Aviral Goel <agoel@roblox.com>
Co-authored-by: Vighnesh Vijay <vvijay@roblox.com>
Co-authored-by: Vyacheslav Egorov <vegorov@roblox.com>
---------
Co-authored-by: Aaron Weiss <aaronweiss@roblox.com>
Co-authored-by: Alexander McCord <amccord@roblox.com>
Co-authored-by: Andy Friesen <afriesen@roblox.com>
Co-authored-by: Vighnesh <vvijay@roblox.com>
Co-authored-by: Aviral Goel <agoel@roblox.com>
Co-authored-by: David Cope <dcope@roblox.com>
Co-authored-by: Lily Brown <lbrown@roblox.com>
### What's new
* A bug in exception handling in GCC(11/12/13) on MacOS prevents our
test suite from running.
* Parser now supports leading `|` or `&` when declaring `Union` and
`Intersection` types (#1286)
* We now support parsing of attributes on functions as described in the
[rfc](https://github.com/luau-lang/rfcs/pull/30)
* With this change, expressions such as `local x = @native function(x)
return x+1 end` and `f(@native function(x) return x+1 end)` are now
valid.
* Added support for `@native` attribute - we can now force native
compilation of individual functions if the `@native` attribute is
specified before the `function` keyword (works for lambdas too).
### New Solver
* Many fixes in the new solver for crashes and instability
* Refinements now use simplification and not normalization in a specific
case of two tables
* Assume that compound assignments do not change the type of the
left-side operand
* Fix error that prevented Class Methods from being overloaded
### VM
* Updated description of Garbage Collector invariant
---
### Internal Contributors
Co-authored-by: Aaron Weiss <aaronweiss@roblox.com>
Co-authored-by: Alexander McCord <amccord@roblox.com>
Co-authored-by: Andy Friesen <afriesen@roblox.com>
Co-authored-by: Aviral Goel <agoel@roblox.com>
Co-authored-by: Vighnesh Vijay <vvijay@roblox.com>
Co-authored-by: Vyacheslav Egorov <vegorov@roblox.com>
---------
Co-authored-by: Aaron Weiss <aaronweiss@roblox.com>
Co-authored-by: Alexander McCord <amccord@roblox.com>
Co-authored-by: Andy Friesen <afriesen@roblox.com>
Co-authored-by: Aviral Goel <agoel@roblox.com>
Co-authored-by: David Cope <dcope@roblox.com>
Co-authored-by: Lily Brown <lbrown@roblox.com>
Co-authored-by: Vyacheslav Egorov <vegorov@roblox.com>
### What's new
* Implemented parsing logic for attributes
* Added `lua_setuserdatametatable` and `lua_getuserdatametatable` C API
methods for a faster userdata metatable fetch compared to
`luaL_getmetatable`. Note that metatable reference has to still be
pinned in memory!
### New Solver
* Further improvement to the assignment inference logic
* Fix many bugs surrounding constraint dispatch order
### Native Codegen
* Add IR lowering hooks for custom host userdata types
* Add IR to create new tagged userdata objects
* Remove outdated NativeState
---
### Internal Contributors
Co-authored-by: Aaron Weiss <aaronweiss@roblox.com>
Co-authored-by: Alexander McCord <amccord@roblox.com>
Co-authored-by: Andy Friesen <afriesen@roblox.com>
Co-authored-by: Aviral Goel <agoel@roblox.com>
Co-authored-by: Vighnesh Vijay <vvijay@roblox.com>
Co-authored-by: Vyacheslav Egorov <vegorov@roblox.com>
These were written before compiler optimizations and array invariant. It
is now impossible for t[1] to be stored in the hash part, as this would
violate the array invariant that says that elements 1..#t are stored in
the array.
For ipairs, it doesn't traverse the hash part anymore now, so we adjust
the code to make sure no elements outside of the 1..#t slice are
covered. For table.find, we can use find-with-offset to still access the
hash part.
Fixes#1283.
The comment gave an incorrect (reversed) version of the invariant, which
could be confusing for people who haven't read the full description in
lgc.cpp.
Unfortunately this change is difficult to flag.
Fixes#1282.
Implements the [Leading `|` and `&` in types](https://rfcs.luau-lang.org/syntax-leading-bar-and-ampersand.html) RFC.
The changes to the parser are exactly as described in the RFC.
---------
Co-authored-by: Alexander McCord <11488393+alexmccord@users.noreply.github.com>
The 'findBindingAtPosition' AstQuery function can be used to lookup a
local or global binding.
Inside of this function is a check to "Ignore this binding if we're
inside its definition. e.g. local abc = abc -- Will take the definition
of abc from outer scope".
However, this check is incorrect when we are looking up a global binding
at the start of a file.
Consider a complete file with the contents:
```lua
local x = stri|ng.char(1)
```
and we pass the location of the marker `|` as the position to the find
binding position. We will pick up the global binding of the definition
`string` coming from a builtin source (either defined via C++ code or a
definitions file and loaded into the global scope).
The global binding `string` will have a zero position: `0,0,0,0`.
However, the `findBindingLocalStatement` check works by looking up the
AstAncestry at the binding's defined begin position *in the current
source module*. This will then incorrectly return the local statement
for `local x`, as that is at the start of the source code. Then in turn,
we assume we are in the `local abc = abc` case, and end up skipping over
the correct binding.
We fix this by checking if the binding is at the global position. If so,
we early exit because it is impossible for a global binding to be
defined in a local statement.
### What's new?
* Remove a case of unsound `table.move` optimization
* Add Luau stack slot reservations that were missing in REPL (fixes
#1273)
### New Type Solver
* Assignments have been completely reworked to fix a case of cyclic
constraint dependency
* When indexing, if the fresh type's upper bound already contains a
compatible indexer, do not add another upper bound
* Distribute type arguments over all type families sans `eq`, `keyof`,
`rawkeyof`, and other internal type families
* Fix a case where `buffers` component weren't read in two places (fixes
#1267)
* Fix a case where things that constitutes a strong ref were slightly
incorrect
* Fix a case where constraint dependencies weren't setup wrt `for ...
in` statement
### Native Codegen
* Fix an optimization that splits TValue store only when its value and
its tag are compatible
* Implement a system to plug additional type information for custom host
userdata types
---
### Internal Contributors
Co-authored-by: Aaron Weiss <aaronweiss@roblox.com>
Co-authored-by: Alexander McCord <amccord@roblox.com>
Co-authored-by: Andy Friesen <afriesen@roblox.com>
Co-authored-by: Vighnesh Vijay <vvijay@roblox.com>
Co-authored-by: Vyacheslav Egorov <vegorov@roblox.com>
---------
Co-authored-by: Aaron Weiss <aaronweiss@roblox.com>
Co-authored-by: Andy Friesen <afriesen@roblox.com>
Co-authored-by: Vighnesh <vvijay@roblox.com>
Co-authored-by: Aviral Goel <agoel@roblox.com>
Co-authored-by: David Cope <dcope@roblox.com>
Co-authored-by: Lily Brown <lbrown@roblox.com>
Co-authored-by: Vyacheslav Egorov <vegorov@roblox.com>
### What's new?
* Removed new `table.move` optimization because of correctness problems.
### New Type Solver
* Improved error messages for type families to describe what's wrong in
more detail, and ideally without using the term `type family` at all.
* Change `boolean` and `string` singletons in type checking to report
errors to the user when they've gotten an impossible type (indicating a
type error from their context).
* Split debugging flags for type family reduction
(`DebugLuauLogTypeFamilies`) from general solver logging
(`DebugLuauLogSolver`).
* Improve type simplification to support patterns like `(number |
string) | (string | number)` becoming `number | string`.
### Native Code Generation
* Use templated `luaV_doarith` to speedup vector operation fallbacks.
* Various small changes to better support arm64 on Windows.
### Internal Contributors
Co-authored-by: Aaron Weiss <aaronweiss@roblox.com>
Co-authored-by: Andy Friesen <afriesen@roblox.com>
Co-authored-by: James McNellis <jmcnellis@roblox.com>
Co-authored-by: Vighnesh Vijay <vvijay@roblox.com>
Co-authored-by: Vyacheslav Egorov <vegorov@roblox.com>
---------
Co-authored-by: Alexander McCord <amccord@roblox.com>
Co-authored-by: Andy Friesen <afriesen@roblox.com>
Co-authored-by: Vighnesh <vvijay@roblox.com>
Co-authored-by: Aviral Goel <agoel@roblox.com>
Co-authored-by: David Cope <dcope@roblox.com>
Co-authored-by: Lily Brown <lbrown@roblox.com>
Co-authored-by: Vyacheslav Egorov <vegorov@roblox.com>
In the readme file under the building section, it specifies that you
should run the command `mkdir cmake && cd cmake`; however, the folder is
not currently being ignored.
### New Type Solver
* Fixed crash in numeric binary operation type families
* Results of an indexing operation are now comparable to `nil` without a
false positive error
* Fixed a crash when a type that failed normalization was accessed
* Iterating on a free value now implies that it is iterable
---
### Internal Contributors
Co-authored-by: Aaron Weiss <aaronweiss@roblox.com>
Co-authored-by: Alexander McCord <amccord@roblox.com>
Co-authored-by: James McNellis <jmcnellis@roblox.com>
Co-authored-by: Vighnesh Vijay <vvijay@roblox.com>
# What's changed?
* Optimize table.maxn. This function is now 5-14x faster
* Reserve Luau stack space for error message.
## New Solver
* Globals can be type-stated, but only if they are already in scope
* Fix a stack overflow that could occur when normalizing certain kinds
of recursive unions of intersections (of unions of intersections...)
* Fix an assertion failure that would trigger when the __iter metamethod
has a bad signature
## Native Codegen
* Type propagation and temporary register type hints
* Direct vector property access should only happen for names of right
length
* BytecodeAnalysis will only predict that some of the vector value
fields are numbers
---
## Internal Contributors
Co-authored-by: Alexander McCord <amccord@roblox.com>
Co-authored-by: Andy Friesen <afriesen@roblox.com>
Co-authored-by: Aviral Goel <agoel@roblox.com>
Co-authored-by: Vyacheslav Egorov <vegorov@roblox.com>
table.concat is idiomatic and should be the fastest way to concatenate
all table array elements together, but apparently you can beat it by
using `string.format`, `string.rep` and `table.unpack`:
```lua
string.format(string.rep("%*", #t), table.unpack(t))
```
... this just won't do, so we should fix table.concat performance.
The deficit comes from two places:
- rawgeti overhead followed by other stack accesses, all to extract a
string from what is almost always an in-bounds array lookup
- addlstring overhead in case separator is empty (extra function calls)
This change fixes this by using a fast path for in-bounds array lookup
for a string. Note that `table.concat` also supports numbers (these need
to be converted to strings which is a little cumbersome and has innate
overhead), and out-of-bounds accesses*. In these cases we fall back to
the old implementation.
To trigger out-of-bounds accesses, you need to skip the past-array-end
element (which is nil per array invariant), but this is achievable
because table.concat supports offset+length arguments. This should
almost never come up in practice but the per-element branches et al are
fairly cheap compared to the eventual string copy/alloc anyway.
This change makes table.concat ~2x faster when the separator is empty;
the table.concat benchmark shows +40% gains but it uses a variety of
string separators of different lengths so it doesn't get the full
benefit from this change.
---------
Co-authored-by: vegorov-rbx <75688451+vegorov-rbx@users.noreply.github.com>
# What's changed?
### New Type Solver
- Unification of two fresh types no longer binds them together.
- Replaced uses of raw `emplace` with `emplaceType` to catch cyclic
bound types when they are created.
- `SetIndexerConstraint` is blocked until the indexer result type is not
blocked.
- Fix a case where a blocked type got past the constraint solver.
- Searching for free types should no longer traverse into `ClassType`s.
- Fix a corner case that could result in the non-testable type `~{}`.
- Fix incorrect flagging when `any` was a parameter of some checked
function in nonstrict type checker.
- `IterableConstraint` now consider tables without `__iter` to be
iterables.
### Native Code Generation
- Improve register type info lookup by program counter.
- Generate type information for locals and upvalues
---
### Internal Contributors
Co-authored-by: Aaron Weiss <aaronweiss@roblox.com>
Co-authored-by: Alexander McCord <amccord@roblox.com>
Co-authored-by: Andy Friesen <afriesen@roblox.com>
Co-authored-by: James McNellis <jmcnellis@roblox.com>
Co-authored-by: Vighnesh Vijay <vvijay@roblox.com>
Co-authored-by: Vyacheslav Egorov <vegorov@roblox.com>
---------
Co-authored-by: Aaron Weiss <aaronweiss@roblox.com>
Co-authored-by: Andy Friesen <afriesen@roblox.com>
Co-authored-by: Vighnesh <vvijay@roblox.com>
Co-authored-by: Aviral Goel <agoel@roblox.com>
Co-authored-by: David Cope <dcope@roblox.com>
Co-authored-by: Lily Brown <lbrown@roblox.com>
Co-authored-by: Vyacheslav Egorov <vegorov@roblox.com>
# What's changed?
* Improved the actual message for the type errors for `cannot call
non-function` when attempting to call a union of functions/callable
tables. The error now correctly explains the issue is an inability to
determine the return type of the call in this situation.
* Resolve an issue where tables and metatables were not correctly being
cloned during instantiation (fixes#1176).
* Refactor `luaM_getnextgcopage` to `luaM_getnextpage` (generally
removing `gco` prefix where appropriate).
* Optimize `table.move` between tables for large ranges of elements.
* Reformat a bunch of code automatically using `clang-format`.
### New Type Solver
* Clean up minimally-used or unused constraints in the constraint solver
(`InstantiationConstraint`, `SetOpConstraint`,
`SingletonOrTopTypeConstraint`).
* Add a builtin `singleton` type family to replace
`SingletonOrTopTypeConstraint` when inferring refinements.
* Fixed a crash involving type path reasoning by recording when type
family reduction has taken place in the path.
* Improved constraint ordering by blocking on unreduced types families
that are not yet proven uninhabitable.
* Improved the handling of `SetIndexerConstraints` for both better
inference quality and to resolve crashes.
* Fix a crash when normalizing cyclic unions of intersections.
* Fix a crash when normalizing an intersection with the negation of
`unknown`.
* Fix a number of crashes caused by missing `follow` calls on `TypeId`s.
* Changed type family reduction to correctly use a semantic notion of
uninhabited types, rather than checking for `never` types specifically.
* Refactor the `union` and `intersect` type families to be variadic.
### Native Code Generation
* Improve translation for userdata key get/set and userdata/vector
namecall.
* Provide `[top level]` and `[anonymous]` as function names to
`FunctionStats` as appropriate when no function name is available.
* Disable unwind support on Android platforms since it is unsupported.
*
---
### Internal Contributors
Co-authored-by: Aaron Weiss <aaronweiss@roblox.com>
Co-authored-by: Alexander McCord <amccord@roblox.com>
Co-authored-by: Andy Friesen <afriesen@roblox.com>
Co-authored-by: Aviral Goel <agoel@roblox.com>
Co-authored-by: Vighnesh Vijay <vvijay@roblox.com>
Co-authored-by: Vyacheslav Egorov <vegorov@roblox.com>
---------
Co-authored-by: Alexander McCord <amccord@roblox.com>
Co-authored-by: Andy Friesen <afriesen@roblox.com>
Co-authored-by: Vighnesh <vvijay@roblox.com>
Co-authored-by: Aviral Goel <agoel@roblox.com>
Co-authored-by: David Cope <dcope@roblox.com>
Co-authored-by: Lily Brown <lbrown@roblox.com>
Co-authored-by: Vyacheslav Egorov <vegorov@roblox.com>
# What's changed?
* Support for new 'require by string' RFC with relative paths and
aliases in now enabled in Luau REPL application
### New Type Solver
* Fixed assertion failure on generic table keys (`[expr] = value`)
* Fixed an issue with type substitution traversing into the substituted
parts during type instantiation
* Fixed crash in union simplification when that union contained
uninhabited unions and other types inside
* Union types in binary type families like `add<a | b, c>` are expanded
into `add<a, c> | add<b, c>` to handle
* Added handling for type family solving creating new type families
* Fixed a bug with normalization operation caching types with unsolved
parts
* Tables with uninhabited properties are now simplified to `never`
* Fixed failures found by fuzzer
### Native Code Generation
* Added support for shared code generation between multiple Luau VM
instances
* Fixed issue in load-store propagation and new tagged LOAD_TVALUE
instructions
* Fixed issues with partial register dead store elimination causing
failures in GC assists
---
### Internal Contributors
Co-authored-by: Aaron Weiss <aaronweiss@roblox.com>
Co-authored-by: Alexander McCord <amccord@roblox.com>
Co-authored-by: Andy Friesen <afriesen@roblox.com>
Co-authored-by: James McNellis <jmcnellis@roblox.com>
Co-authored-by: Vighnesh Vijay <vvijay@roblox.com>
Co-authored-by: Vyacheslav Egorov <vegorov@roblox.com>
# What's Changed
## New Type Solver
- Many more fixes to crashes, assertions, and hangs
- Annotated locals now countermand the inferred types of locals, meaning
that for a type `type MyType = number | string`, `local foo : MyType =
5` behaves the same as `local foo = 5 :: MyType`, where before, foo
would be assigned the type of the value on the rhs.
- Type Normalization now respects resource limits.
- Subtyping between classes and cyclic tables now supported
## Native Code Generation
- Work on the Native Code Generation(NCG) allocator continues
---
# Internal Contributors
Co-authored-by: Aaron Weiss <aaronweiss@roblox.com>
Co-authored-by: Alexander McCord <amccord@roblox.com>
Co-authored-by: Andy Friesen <afriesen@roblox.com>
Co-authored-by: James McNellis <jmcnellis@roblox.com>
Co-authored-by: Vighnesh Vijay <vvijay@roblox.com>
Co-authored-by: Vyacheslav Egorov <vegorov@roblox.com>
---------
Co-authored-by: Aaron Weiss <aaronweiss@roblox.com>
Co-authored-by: Alexander McCord <amccord@roblox.com>
Co-authored-by: Andy Friesen <afriesen@roblox.com>
Co-authored-by: Aviral Goel <agoel@roblox.com>
Co-authored-by: David Cope <dcope@roblox.com>
Co-authored-by: Lily Brown <lbrown@roblox.com>
Co-authored-by: Vyacheslav Egorov <vegorov@roblox.com>
# What's Changed
## New Type Solver
- Many fixes to crashes, assertions, and hangs
- Binary type family aliases now have a default parameter
- Added a debug check for unsolved types escaping the constraint solver
- Overloaded functions are no longer inferred
- Unification creates additional subtyping constraints for blocked types
- Attempt to guess the result type for type families that are too large
to resolve timely
## Native Code Generation
- Fixed `IrCmd::CHECK_TRUTHY` lowering in a specific case
- Detailed compilation errors are now supported
- More work on the new allocator
---
# Internal Contributors
Co-authored-by: Aaron Weiss <aaronweiss@roblox.com>
Co-authored-by: Alexander McCord <amccord@roblox.com>
Co-authored-by: Andy Friesen <afriesen@roblox.com>
Co-authored-by: James McNellis <jmcnellis@roblox.com>
Co-authored-by: Lily Brown <lbrown@roblox.com>
Co-authored-by: Vyacheslav Egorov <vegorov@roblox.com>
Co-authored-by: Vighnesh Vijay <vvijay@roblox.com>
# What's changed
### Debugger
* Values after a 'continue' statement should not be accessible by
debugger in the 'until' condition
### New Type Solver
* Many fixes to crashes and hangs
* Better bidirectional inference of table literal expressions
### Native Code Generation
* Initial steps toward a shared code allocator
---
### Internal Contributors
Co-authored-by: Aaron Weiss <aaronweiss@roblox.com>
Co-authored-by: Lily Brown <lbrown@roblox.com>
Co-authored-by: Vyacheslav Egorov <vegorov@roblox.com>
# What's Changed
* Fix a case where the stack wasn't completely cleaned up where
`debug.info` errored when passed `"f"` option and a thread.
* Fix a case of uninitialized field in `luaF_newproto`.
### New Type Solver
* When a local is captured in a function, don't add a new entry to the
`DfgScope::bindings` if the capture occurs within a loop.
* Fix a poor performance characteristic during unification by not trying
to simplify an intersection.
* Fix a case of multiple constraints mutating the same blocked type
causing incorrect inferences.
* Fix a case of assertion failure when overload resolution encounters a
return typepack mismatch.
* When refining a property of the top `table` type, we no longer signal
an unknown property error.
* Fix a misuse of free types when trying to infer the type of a
subscript expression.
* Fix a case of assertion failure when trying to resolve an overload
from `never`.
### Native Code Generation
* Fix dead store optimization issues caused by partial stores.
---
### Internal Contributors
Co-authored-by: Aaron Weiss <aaronweiss@roblox.com>
Co-authored-by: Andy Friesen <afriesen@roblox.com>
Co-authored-by: Aviral Goel <agoel@roblox.com>
Co-authored-by: David Cope <dcope@roblox.com>
Co-authored-by: Lily Brown <lbrown@roblox.com>
Co-authored-by: Vyacheslav Egorov <vegorov@roblox.com>
---------
Co-authored-by: Aaron Weiss <aaronweiss@roblox.com>
Co-authored-by: Andy Friesen <afriesen@roblox.com>
Co-authored-by: Vighnesh <vvijay@roblox.com>
Co-authored-by: Aviral Goel <agoel@roblox.com>
Co-authored-by: David Cope <dcope@roblox.com>
Co-authored-by: Lily Brown <lbrown@roblox.com>
Co-authored-by: Vyacheslav Egorov <vegorov@roblox.com>
When lowering LOADK for booleans/numbers/nils, we deconstruct the
operation using STORE_TAG which informs the rest of the optimization
pipeline about the tag of the value. This is helpful to remove various
tag checks.
When the constant is a string or a vector, we just use
LOAD_TVALUE/STORE_TVALUE. For strings, this could be replaced by pointer
load/store, but for vectors there's no great alternative using current
IR ops; in either case, the optimization needs to be carefully examined
for profitability as simply copying constants into registers for
function calls could become more expensive.
However, there are cases where it's still valuable to preserve the tag.
For vectors, doing any math with vector constants contains tag checks
that could be removed. For both strings and vectors, storing them into a
table has a barrier that for vectors could be elided, and for strings
could be simplified as there's no need to confirm the tag.
With this change we now carry the optional tag of the value with
LOAD_TVALUE. This has no performance effect on existing benchmarks but
does reduce the generated code for benchmarks by ~0.1%, and it makes
vector code more efficient (~5% lift on X64 log1p approximation).
vm.mmap_rnd_bits has been recently changed to 32 on GHA, which triggers
issues in ASAN builds that spuriously fail on startup. The fix requires
a more recent clang/gcc than the agents have available (clang 17, not
sure what GCC version), so for now we need to work around this by
restricting the ASLR randomness.
See https://github.com/google/sanitizers/issues/1614
When the input is a constant, we use a fairly inefficient sequence of
fmov+fcvt+dup or, when the double isn't encodable in fmov,
adr+ldr+fcvt+dup.
Instead, we can use the same lowering as X64 when the input is a
constant, and load the vector from memory. However, if the constant is
encodable via fmov, we can use a vector fmov instead (which is just one
instruction and doesn't need constant space).
Fortunately the bit encoding of fmov for 32-bit floating point numbers
matches that of 64-bit: the decoding algorithm is a little different
because it expands into a larger exponent, but the values are
compatible, so if a double can be encoded into a scalar fmov with a
given abcdefgh pattern, the same pattern should encode the same float;
due to the very limited number of mantissa and exponent bits, all values
that are encodable are also exact in both 32-bit and 64-bit floats.
This strategy is ~same as what gcc uses. For complex vectors, we
previously used 4 instructions and 8 bytes of constant storage, and now
we use 2 instructions and 16 bytes of constant storage, so the memory
footprint is the same; for simple vectors we just need 1 instruction (4
bytes).
clang lowers vector constants a little differently, opting to synthesize
a 64-bit integer using 4 instructions (mov/movk) and then move it to the
vector register - this requires 5 instructions and 20 bytes, vs ours/gcc
2 instructions and 8+16=24 bytes. I tried a simpler version of this that
would be more compact - synthesize a 32-bit integer constant with
mov+movk, and move it to vector register via dup.4s - but this was a
little slower on M2, so for now we prefer the slightly larger version as
it's not a regression vs current implementation.
On the vector approximation benchmark we get:
- Before this PR (flag=false): ~7.85 ns/op
- After this PR (flag=true): ~7.74 ns/op
- After this PR, with 0.125 instead of 0.123 in the benchmark code (to
use fmov): ~7.52 ns/op
- Not part of this PR, but the mov/dup strategy described above: ~8.00
ns/op
The last line of the help message was missing a newline character. I
feel a little silly creating a pull request for a 2 character change but
it was bothering me. Fixes#1185
# What's Changed
* Add a compiler hint to improve Luau memory allocation inlining
### New Type Solver
* Added a system for recommending explicit type annotations to users in
cases where we've inferred complex generic types with type families.
* Marked string library functions as `@checked` for use in new
non-strict mode.
* Fixed a bug with new non-strict mode where we would incorrectly report
arity mismatches when missing optional arguments.
* Implement an occurs check for unifications that would produce
self-recursive types.
* Fix bug where overload resolution would fail when applied to
non-overloaded functions.
* Fix bug that caused the subtyping to report an error whenever a
generic was instantiated in an invariant context.
* Fix crash caused by `SetPropConstraint` not blocking properly.
### Native Code Generation
* Implement optimization to eliminate dead stores
* Optimize vector ops for X64 when the source is computed (thanks,
@zeux!)
* Use more efficient lowering for UNM_* (thanks, @zeux!)
---
### Internal Contributors
Co-authored-by: Aaron Weiss <aaronweiss@roblox.com>
Co-authored-by: Alexander McCord <amccord@roblox.com>
Co-authored-by: Andy Friesen <afriesen@roblox.com>
Co-authored-by: David Cope <dcope@roblox.com>
Co-authored-by: Lily Brown <lbrown@roblox.com>
Co-authored-by: Vyacheslav Egorov <vegorov@roblox.com>
---------
Co-authored-by: Alexander McCord <amccord@roblox.com>
Co-authored-by: Andy Friesen <afriesen@roblox.com>
Co-authored-by: Vighnesh <vvijay@roblox.com>
Co-authored-by: Aviral Goel <agoel@roblox.com>
Co-authored-by: David Cope <dcope@roblox.com>
Co-authored-by: Lily Brown <lbrown@roblox.com>
Co-authored-by: Vyacheslav Egorov <vegorov@roblox.com>
# What's changed?
* Luau allocation scheme was changed to handle allocations in 513-1024
byte range internally without falling back to global allocator
* coroutine/thread creation no longer requires any global allocations,
making it up to 15% faster (vs libc malloc)
* table construction for 17-32 keys or 33-64 array elements is up to 30%
faster (vs libc malloc)
### New Type Solver
* Cyclic unary negation type families are reduced to `number` when
possible
* Class types are skipped when searching for free types in unifier to
improve performance
* Fixed issues with table type inference when metatables are present
* Improved inference of iteration loop types
* Fixed an issue with bidirectional inference of method calls
* Type simplification will now preserve error suppression markers
### Native Code Generation
* Fixed TAG_VECTOR skip optimization to not break instruction use counts
(broken optimization wasn't included in 614)
* Fixed missing side-effect when optimizing generic loop preparation
instruction
---
### Internal Contributors
Co-authored-by: Aaron Weiss <aaronweiss@roblox.com>
Co-authored-by: Andy Friesen <afriesen@roblox.com>
Co-authored-by: Lily Brown <lbrown@roblox.com>
Co-authored-by: Vyacheslav Egorov <vegorov@roblox.com>
---------
Co-authored-by: Aaron Weiss <aaronweiss@roblox.com>
Co-authored-by: Alexander McCord <amccord@roblox.com>
Co-authored-by: Andy Friesen <afriesen@roblox.com>
Co-authored-by: Vighnesh <vvijay@roblox.com>
Co-authored-by: Aviral Goel <agoel@roblox.com>
Co-authored-by: David Cope <dcope@roblox.com>
Co-authored-by: Lily Brown <lbrown@roblox.com>
With the TAG_VECTOR change, we can now confidently distinguish cases
when the .w component
contains TVECTOR tag from cases where it doesn't: loads and tag ops
produce the tag, whereas
other instructions don't.
We now take advantage of this fact and only apply vandps with a mask
when we need to.
It would be possible to use a positive filter (explicitly checking for
source coming from ADD_VEC
et al), but there are more instructions to check this way and this is
purely an optimization so
it is allowed to be conservative (as in, the cost of a mistake here is a
potential slowdown,
not a correctness issue).
Additionally, this change only performs vandps once when the arguments
are the same instead
of doing it twice.
On the function that computes a polynomial approximation this change
makes it ~20% faster on Zen4.
When --vector-lib is not specified, CompileOptions::vectorLib was set to
an empty string. This resulted in the builtin matching not working,
since vectorLib must either be a null pointer or a pointer to a valid
global identifier.
---------
Co-authored-by: vegorov-rbx <75688451+vegorov-rbx@users.noreply.github.com>
# What's changed?
Add program argument passing to scripts run using the Luau REPL! You can
now pass `--program-args` (or shorthand `-a`) to the REPL which will
treat all remaining arguments as arguments to pass to executed scripts.
These values can be accessed through variadic argument expansion. You
can read these values like so:
```
local args = {...} -- gets you an array of all the arguments
```
For example if we run the following script like `luau test.lua -a test1
test2 test3`:
```
-- test.lua
print(...)
```
you should get the output:
```
test1 test2 test3
```
### Native Code Generation
* Improve A64 lowering for vector operations by using vector
instructions
* Fix lowering issue in IR value location tracking!
- A developer reported a divergence between code run in the VM and
Native Code Generation which we have now fixed
### New Type Solver
* Apply substitution to type families, and emit new constraints to
reduce those further
* More progress on reducing comparison (`lt/le`)type families
* Resolve two major sources of cyclic types in the new solver
### Miscellaneous
* Turned internal compiler errors (ICE's) into warnings and errors
-------
Co-authored-by: Aaron Weiss <aaronweiss@roblox.com>
Co-authored-by: Alexander McCord <amccord@roblox.com>
Co-authored-by: Andy Friesen <afriesen@roblox.com>
Co-authored-by: Aviral Goel <agoel@roblox.com>
Co-authored-by: Vyacheslav Egorov <vegorov@roblox.com>
---------
Co-authored-by: Aaron Weiss <aaronweiss@roblox.com>
Co-authored-by: Alexander McCord <amccord@roblox.com>
Co-authored-by: Andy Friesen <afriesen@roblox.com>
Co-authored-by: Aviral Goel <agoel@roblox.com>
Co-authored-by: David Cope <dcope@roblox.com>
Co-authored-by: Lily Brown <lbrown@roblox.com>
Co-authored-by: Vyacheslav Egorov <vegorov@roblox.com>
Instead of patching the tag component with TVECTOR in every instruction
that produces a vector value, we now use a separate IR instruction to do
this. This reduces implementation redundancy, but more importantly
allows for a class of optimizations:
- NUM_TO_VECTOR previously patched the component unconditionally but the
result was used only in MUL/DIV_VEC instructions that ignore it anyway;
we can now remove this.
- ADD_VEC et al can now forward the source of TAG_VECTOR instruction of
either input; this shortens the latency chain and in the future could
allow us to generate optimal vector instruction sequence once the
temporary stores are marked as dead.
- In the future on X64, ADD_VEC et al will be able to analyze the input
instruction and remove tag masking conditionally. This is not part of
this PR as it requires a decision around expected FP environment and/or
the necessity of the existing masking to begin with.
I've also renamed NUM_TO_VECTOR to NUM_TO_VEC so that "VEC" always
refers to "3 float values" and for consistency with ADD/etc.
Note: ADD_VEC input forwarding is currently performed unconditionally;
it may or may not increase the spills that can't be reloaded from the
stack.
On A64 this makes the Taylor series computation a tiny bit faster
(11.3ns => 11.0ns) as it removes the redundant ins instructions along
the NUM_TO_VEC path. Curiously, the optimization of forwarding
TAG_VECTOR input to arithmetic instructions actually has a small penalty
as without it this PR runs at 10.9 ns. I don't know if this is a
property of the benchmark though, as I just noticed that in this
benchmark type inference actually fails to infer parts of the
computation as a vector op. If desired I will happily omit this part of
the change and we can explore that separately.
This change replaces scalar versions of vector opcodes for A64 with
actual vector instructions.
We take the approach similar to X64: patch last component with zero,
perform the math, patch last component with type tag. I'm hoping that in
the future the type tag will be placed separately (separate IR opcode?)
because right now chains of math operations result in excessive type tag
operations.
To patch the type tag without always keeping a mask in a register,
ins.4s instructions can be used; unfortunately it's only capable of
patching a register in-place, so we need an extra register copy in case
it's not last-use. Usually it's last-use so the patch is free; probably
with IR rework mentioned above all of this can be improved (e.g.
load-with-patch will never need to copy).
~It's not 100% clear if we *have* to patch type tag: Apple does preserve
denormals but we'd need to benchmark this to see if there's an actual
performance impact. But for now we're playing it safe.~
This was tested by running the conformance tests, and new opcode
implementations were checked by comparing the result with
https://armconverter.com/.
Performance testing is complicated by the fact that OSS Luau doesn't
support vector constructor out of the box, and other limitations of
codegen. I've hacked vector constructor/type into REPL and confirmed
that on a test that calls this function in a loop (not inlined):
```
function fma(a: vector, b: vector, c: vector)
return a * b + c
end
```
... this PR improves performance by ~6% (note that probably most of the
overhead here is the call dispatch; I didn't want to brave testing a
more complex expression). The assembly for an individual operation
changes as follows:
Before:
```
# %14 = MUL_VEC %12, %13 ; useCount: 2, lastUse: %22
dup s29,v31.s[0]
dup s28,v30.s[0]
fmul s29,s29,s28
ins v31.s[0],v29.s[0]
dup s29,v31.s[1]
dup s28,v30.s[1]
fmul s29,s29,s28
ins v31.s[1],v29.s[0]
dup s29,v31.s[2]
dup s28,v30.s[2]
fmul s29,s29,s28
ins v31.s[2],v29.s[0]
```
After:
```
# %14 = MUL_VEC %12, %13 ; useCount: 2, lastUse: %22
ins v31.s[3],w31
ins v30.s[3],w31
fmul v31.4s,v31.4s,v30.4s
movz w17,#4
ins v31.s[3],w17
```
**edit** final form (see comments):
```
# %14 = MUL_VEC %12, %13 ; useCount: 2, lastUse: %22
fmul v31.4s,v31.4s,v30.4s
movz w17,#4
ins v31.s[3],w17
```
# What's changed?
* Compiler now targets bytecode version 5 by default, this includes
support for vector type literals and sub/div opcodes with a constant on
lhs
### New Type Solver
* Normalizer type inhabitance check has been optimized
* Added ability to reduce cyclic `and`/`or` type families
### Native Code Generation
* `CodeGen::compile` now returns more specific causes of a code
generation failure
* Fixed linking issues on platforms that don't support unwind frame data
registration
---
### Internal Contributors
Co-authored-by: Andy Friesen <afriesen@roblox.com>
Co-authored-by: Aviral Goel <agoel@roblox.com>
Co-authored-by: Vyacheslav Egorov <vegorov@roblox.com>
---------
Co-authored-by: Aaron Weiss <aaronweiss@roblox.com>
Co-authored-by: Alexander McCord <amccord@roblox.com>
Co-authored-by: Andy Friesen <afriesen@roblox.com>
Co-authored-by: Vighnesh <vvijay@roblox.com>
Co-authored-by: Aviral Goel <agoel@roblox.com>
Co-authored-by: David Cope <dcope@roblox.com>
Co-authored-by: Lily Brown <lbrown@roblox.com>
New solver
* Fix bugs where bidirectional type inference would fail to take effect
at the proper stage.
* Improve inference of mutually recursive functions
* Fix crashes
---------
Co-authored-by: Aaron Weiss <aaronweiss@roblox.com>
Co-authored-by: Andy Friesen <afriesen@roblox.com>
Co-authored-by: Vyacheslav Egorov <vegorov@roblox.com>
