I've decided to take a stab at implementing `bit32.byteswap` from the
[recently merged
RFC](https://github.com/Roblox/luau/blob/master/rfcs/function-bit32-byteswap.md).
I asked on Discord for some guidance, but for the sake of posterity:
this is my first time doing this and I am likely to have made some
mistakes.
The biggest gaps in this implementation are the lack of tests and the
lack of native codegen support. I'd appreciate help with those since I'm
not sure what's relevant for me to touch for tests, and I'm told that
relevant assembler instructions don't exist publicly yet. Intuition
tells me that Luau-side tests would go into
`tests/conformance/bitwise.luau` but this is not well documented and I'm
not sure how I'm meant to test built-in implementations.
The current implementation compiles down to `bswap` and `rev` on x86 and
ARM respectively when optimized.
---------
Co-authored-by: Arseny Kapoulkine <arseny.kapoulkine@gmail.com>
### What's Changed
- Improve readability of unions and intersections by limiting the number
of elements of those types that can be presented on a single line (gated
under `FFlag::LuauToStringSimpleCompositeTypesSingleLine`)
- Adds a new option to the compiler `--record-stats` to record and
output compilation statistics
- `if...then...else` expressions are now optimized into `AND/OR` form
when possible.
### VM
- Add a new `buffer` type to Luau based on the [buffer
RFC](https://github.com/Roblox/luau/pull/739) and additional C API
functions to work with it; this release does not include the library.
- Internal C API to work with string buffers has been updated to align
with Lua version more closely
### Native Codegen
- Added support for new X64 instruction (rev) and new A64 instruction
(bswap) in the assembler
- Simplified the way numerical loop condition is translated to IR
### New Type Solver
- Operator inference now handled by type families
- Created a new system called `Type Paths` to explain why subtyping
tests fail in order to improve the quality of error messages.
- Systematic changes to implement Data Flow analysis in the new solver
(`Breadcrumb` removed and replaced with `RefinementKey`)
---
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: Lily Brown <lbrown@roblox.com>
Co-authored-by: Vighnesh Vijay <vvijay@roblox.com>
Co-authored-by: Vyacheslav Egorov <vegorov@roblox.com>
---------
Co-authored-by: Arseny Kapoulkine <arseny.kapoulkine@gmail.com>
Co-authored-by: Vyacheslav Egorov <vegorov@roblox.com>
Co-authored-by: Andy Friesen <afriesen@roblox.com>
Co-authored-by: Lily Brown <lbrown@roblox.com>
Co-authored-by: Aaron Weiss <aaronweiss@roblox.com>
Co-authored-by: Alexander McCord <amccord@roblox.com>
Co-authored-by: Aviral Goel <agoel@roblox.com>
When the original version was written, bidirectional type inference
didn't exist, so this new kind of pattern was simply not possible.
Now that it does, we can stop telling users to use the original typeof
workaround.
The only thing that's a bit unfortunate is the cast of `{}` into
`AccountImpl`, but it's a necessary evil for now until we get a few more
big ticket items. When we have that, we can likely kill the cast and
call it redundant.
This function might be of particular use if #739 is accepted, but as it
stands, it still brings a measurable reduction in complexity for
swapping byte order and is measurably faster (around a 50% reduction in
time per operation on my machine), albeit on the order of magnitude of
nanoseconds.
## What's Changed
- Improve POSIX compliance in `CLI/FileUtils.cpp` by @SamuraiCrow #1064
- `AstStat*::hasEnd` is deprecated; use `AstStatBlock::hasEnd` instead
- Added a lint for common misuses of the `#` operator
- Luau now issues deprecated diagnostics for some uses of `getfenv` and
`setfenv`
- Fixed a case where we included a trailing space in some error
stringifications
### Compiler
- Do not do further analysis in O2 on self functions
- Improve detection of invalid repeat..until expressions vs continue
### New Type Solver
- We now cache subtype test results to improve performance
- Improved operator inference mechanics (aka type families)
- Further work towards type states
- Work towards [new non-strict
mode](https://github.com/Roblox/luau/blob/master/rfcs/new-nonstrict.md)
continues
### Native Codegen
- Instruction last use locations should follow the order in which blocks
are lowered
- Add a bonus assertion to IrLoweringA64::tempAddr
---
Co-authored-by: Arseny Kapoulkine <arseny.kapoulkine@gmail.com>
Co-authored-by: Vyacheslav Egorov <vegorov@roblox.com>
Co-authored-by: Andy Friesen <afriesen@roblox.com>
Co-authored-by: Aaron Weiss <aaronweiss@roblox.com>
Co-authored-by: Alexander McCord <amccord@roblox.com>
Co-authored-by: Vighnesh Vijay <vvijay@roblox.com>
Adds an option `compositeTypesSingleLineLimit` to control when
unions/intersections are broken up onto multiple lines when printed.
Simple unions/intersections will remain on one line. Default value is
set to `5`.
I took the liberty to always expand overloaded function intersections on
multiple lines, as I believe they are best viewed that way, even if
theres only 2 functions.
Closes#963
Some POSIX platforms, such as Haiku and some BSDs, don't supply DT_*
identifiers and the corresponding d_type field in stat. This fixes that
and has been tested to still work on 64-bit Linux as well as 64-bit
Haiku.
String interpolation and table.getn/et al deprecation has been live for
a while.
getfenv/setfenv linter diagnostics is going to ship in the next release,
so preemptively marking this as Implemented as well.
* Include `windows.h` rather than `Windows.h` to make things compile on
MinGW.
* Custom implementation of timegm/os.time for all platforms
* Disable builtin constant folding when getfenv/setfenv are used
* Fixes https://github.com/Roblox/luau/issues/1042
* Fixes https://github.com/Roblox/luau/issues/1043
New Type Checker
* Initial work toward type states.
* Rework most overloadable operators to use type families.
* Initial work toward our new nonstrict mode.
Native Codegen
* Fix native code generation for dead loops
* Annotate top-level functions as cold
* Slightly smaller/faster x64 Luau calls
* emitInstCall used to not set savedpc itself, but now it does for
consistency with all other implementations
* Implement cmov support for X64
* Fix assertion in luau-compile when module is empty
* Optimize A64 calls at some code size cost
* Inline constant array index offset into the load/store instruction
* Increase x64 spill slots from 5 to 13
---------
Co-authored-by: Arseny Kapoulkine <arseny.kapoulkine@gmail.com>
Co-authored-by: Vyacheslav Egorov <vegorov@roblox.com>
Co-authored-by: Lily Brown <lbrown@roblox.com>
Co-authored-by: Aaron Weiss <aaronweiss@roblox.com>
Co-authored-by: Alexander McCord <amccord@roblox.com>
Add a note about Lua 5.3 change to semantics of __eq metamethod.
We currently do not plan to implement this as this breaks backwards
compatibility and makes these operators harder to reason about from the
type checking perspective.
Fixes#1051.
# New Type Solver
- Implement bidirectional type inference for higher order functions so
that we can provide a more precise type improving the autocomplete's
human factors.
- We seal all tables, so we changed the stringification to make it a
little lighter on users.
- Fixed a case of array-out-of-bound access.
- Type families no longer depends on `TxnLog` and `Unifier`.
- Type refinements now waits until the free types are sufficiently
solved.
# Native Code Generation
- Remove cached slot lookup for `executeSETTABLEKS` function because it
is a fallback in the event of a cache miss, making the cached slot
lookup redundant.
- Optimized repeated array lookups, e.g. `a[3]` in `a[3] = a[3] / 2` is
done once.
# Misc
- On some platforms, it is necessary to use `gmtime_s` with the
arguments reversed to get the current time. You can now define
`DOCTEST_CONFIG_USE_GMTIME_S` to build and run unit tests on those
platforms.
---------
Co-authored-by: Arseny Kapoulkine <arseny.kapoulkine@gmail.com>
Co-authored-by: Vyacheslav Egorov <vegorov@roblox.com>
Co-authored-by: Andy Friesen <afriesen@roblox.com>
Co-authored-by: Lily Brown <lbrown@roblox.com>
Co-authored-by: Aaron Weiss <aaronweiss@roblox.com>
- Cleaned up `FFlag::FixFindBindingAtFunctionName`,
`FFlag::LuauNormalizeBlockedTypes`, `FFlag::LuauPCallDebuggerFix`
- Added support for break and continue into control flow analysis
- The old type unification engine will now report a more fine-grained
error at times, indicating that type normalization in particular failed
# New Type Solver
- Refactor of Unifier2, the new unification implementation for Luau
- Completed MVP of new unification implementation
- Dramatically simplified overload selection logic
- Type family reduction can now apply sooner to free types that have
been solved
- Subtyping now supports table indexers
- Generalization now replaces bad generics with unknown
# Native Code Generation
- Reduce stack spills caused by FINDUPVAL and STORE_TAG
- Improve Generate SHL/SHR/SAR/rotates with immediate operands in X64
- Removed redundant case re-check in table lookup fallback
---------
Co-authored-by: Arseny Kapoulkine <arseny.kapoulkine@gmail.com>
Co-authored-by: Vyacheslav Egorov <vegorov@roblox.com>
Co-authored-by: Andy Friesen <afriesen@roblox.com>
Co-authored-by: Lily Brown <lbrown@roblox.com>
Fixes: https://github.com/Roblox/luau/issues/913
This PR adds support for type refinements around guard clauses that use
`break` and `continue` statements inside a loop, similar to how guard
clauses with `return` is supported.
I had some free time today, so I figure I'd give a shot at a naïve fix
for this at the very least.
---
## Resulting Change:
Luau now supports type refinements within loops where a `continue` or
`break` guard clause was used.
For example:
```lua
for _, object in objects :: {{value: string?}} do
if not object.value then
continue
end
local x: string = object.value -- OK; Used to emit "Type 'string?' could not be converted into 'string'"
end
```
---------
Co-authored-by: Alexander McCord <amccord@roblox.com>
* Rerun clang-format on the code
* Fix the variance on indexer result subtyping. This fixes some issues
with inconsistent error reporting.
* Fix a bug in the normalization logic for intersections of strings
New Type Solver
* New overload selection logic
* Subtype tests now correctly treat a generic as its upper bound within
that generic's scope
* Semantic subtyping for negation types
* Semantic subtyping between strings and compatible table types like
`{lower: (string) -> string}`
* Further work toward finalizing our new subtype test
* Correctly generalize module-scope symbols
Native Codegen
* Lowering statistics for assembly
* Make executable allocation size/limit configurable without a rebuild.
Use `FInt::LuauCodeGenBlockSize` and `FInt::LuauCodeGenMaxTotalSize`.
---------
Co-authored-by: Arseny Kapoulkine <arseny.kapoulkine@gmail.com>
Co-authored-by: Vyacheslav Egorov <vegorov@roblox.com>
Co-authored-by: Lily Brown <lbrown@roblox.com>
* Fixed `Frontend::markDirty` not working on modules that were not
typechecked yet
* Fixed generic variadic function unification succeeding when it should
have reported an error
New Type Solver:
* Implemented semantic subtyping check for function types
Native Code Generation:
* Improved performance of numerical loops with a constant step
* Simplified IR for `bit32.extract` calls extracting first/last bits
* Improved performance of NaN checks
CLI/Reduce.cpp:
Never worked under MinGW as-is. Add check for MinGW as they're defined.
VM/src/lmem.cpp:
MinGW layout follows more closely MSVC. Checks before were only correct
for 32-bit gcc / clang in non-Windows.
NOTES:
__MINGW32__ is defined on both 32-bit and 64-bit, __MINGW64__ is 64-bit
only. 32-bit MinGW compilers (both Clang & GCC) have a floating point
test error on math.noise, specifically math.lua:258. All other test
cases / unit tests pass modulo stack size issues (so requires optimized build).
---------
Co-authored-by: jdp_ <42700985+jdpatdiscord@users.noreply.github.com>
- Updated Roblox copyright to 2023
- Floor division operator `//` (implements #832)
- Autocomplete now shows `end` within `do` blocks
- Restore BraceType when using `Lexer::lookahead` (fixes#1019)
# New typechecker
- Subtyping tests between metatables and tables
- Subtyping tests between string singletons and tables
- Subtyping tests for class types
# Native codegen
- Fixed macOS test failure (wrong spill restore offset)
- Fixed clobbering of non-volatile xmm registers on Windows
- Fixed wrong storage location of SSA reg spills
- Implemented A64 support for add/sub extended
- Eliminated zextReg from A64 lowering
- Remove identical table slot lookups
- Propagate values from predecessor into the linear block
- Disabled reuse slot optimization
- Keep `LuaNode::val` check for nil when optimizing `CHECK_SLOT_MATCH`
- Implemented IR translation of `table.insert` builtin
- Fixed mmap error handling on macOS/Linux
# Tooling
- Used `|` as a column separator instead of `+` in `bench.py`
- Added a `table.sort` micro-benchmark
- Switched `libprotobuf-mutator` to a less problematic version
* AST queries at position where function name is will now return
AstExprLocal
* Lexer performance has been slightly improved
* Fixed incorrect string singleton autocomplete suggestions (fixes#858)
* Improved parsing error messages
* Fixed crash on null pointer access in unification (fixes#1017)
* Native code support is enabled by default and `native=1`
(make)/`LUAU_NATIVE` (CMake)/`-DLUA_CUSTOM_EXECUTION` configuration is
no longer required
New typechecker:
* New subtyping check can now handle generic functions and tables
(including those that contain cycles)
Native code generation:
* Loops with non-numeric parameters are now handled by VM to streamline
native code
* Array size check can be optimized away in SETLIST
* On failure, CodeGen::compile returns a reason
* Fixed clobbering of non-volatile xmm registers on Windows
* Fix a use-after-free bug in the new type cloning algorithm
* Tighten up the type of `coroutine.wrap`. It is now `<A..., R...>(f:
(A...) -> R...) -> ((A...) -> R...)`
* Break `.luaurc` out into a separate library target `Luau.Config`. This
makes it easier for applications to reason about config files without
also depending on the type inference engine.
* Move typechecking limits into `FrontendOptions`. This allows embedders
more finely-grained control over autocomplete's internal time limits.
* Fix stability issue with debugger onprotectederror callback allowing
break in non-yieldable contexts
New solver:
* Initial work toward [Local Type
Inference](0e1082108f/rfcs/local-type-inference.md)
* Introduce a new subtyping test. This will be much nicer than the old
test because it is completely separate both from actual type inference
and from error reporting.
Native code generation:
* Added function to compute iterated dominance frontier
* Optimize barriers in SET_UPVALUE when tag is known
* Cache lua_State::global in a register on A64
* Optimize constant stores in A64 lowering
* Track table array size state to optimize array size checks
* Add split tag/value store into a VM register
* Check that spills can outlive the block only in specific conditions
---------
Co-authored-by: Arseny Kapoulkine <arseny.kapoulkine@gmail.com>
Co-authored-by: Vyacheslav Egorov <vegorov@roblox.com>
* Better indentation in multi-line type mismatch error messages
* Error message clone can no longer cause a stack overflow (when
typechecking with retainFullTypeGraphs set to false); fixes
https://github.com/Roblox/luau/issues/975
* `string.format` with %s is now ~2x faster on strings smaller than 100
characters
Native code generation:
* All VM side exits will block return to the native execution of the
current function to preserve correctness
* Fixed executable page allocation on Apple platforms when using
hardened runtime
* Added statistics for code generation (no. of functions compiler,
memory used for different areas)
* Fixed issue with function entry type checks performed more that once
in some functions
Nobody is maintaining this and we haven't really used it and are
unlikely to start due to a high degree of noise and lack of dedicated
machines for this setup. Callgrind has worked for us well enough, with
additional profiling performed locally by engineers - this is not
perfect but it doesn't look like we have a path to making this better.
* Progress toward a diffing algorithm for types. We hope that this will
be useful for writing clearer error messages.
* Add a missing recursion limiter in `Unifier::tryUnifyTables`. This was
causing a crash in certain situations.
* Luau heap graph enumeration improvements:
* Weak references are not reported
* Added tag as a fallback name of non-string table links
* Included top Luau function information in thread name to understand
where thread might be suspended
* Constant folding for `math.pi` and `math.huge` at -O2
* Optimize `string.format` and `%*`
* This change makes string interpolation 1.5x-2x faster depending on the
number and type of formatted components, assuming a few are using
primitive types, and reduces associated GC pressure.
New type checker:
* Initial work toward tracking the upper and lower bounds of types
accurately.
Native code generation (JIT):
* Add IrCmd::CHECK_TRUTHY for improved assert fast-calls
* Do not compute type map for modules without types
* Capture metatable+readonly state for NEW_TABLE IR instructions
* Replace JUMP_CMP_ANY with CMP_ANY and existing JUMP_EQ_INT
* Add support for exits to VM with reentry lock in VmExit
---------
Co-authored-by: Arseny Kapoulkine <arseny.kapoulkine@gmail.com>
Co-authored-by: Vyacheslav Egorov <vegorov@roblox.com>
Type checker/autocomplete:
* `Luau::autocomplete` no longer performs typechecking internally, make
sure to run `Frontend::check` before performing autocomplete requests
* Autocomplete string suggestions without "" are now only suggested
inside the ""
* Autocomplete suggestions now include `function (anonymous autofilled)`
key with a full suggestion for the function expression (with arguments
included) stored in `AutocompleteEntry::insertText`
* `AutocompleteEntry::indexedWithSelf` is provided for function call
suggestions made with `:`
* Cyclic modules now see each other type exports as `any` to prevent
memory use-after-free (similar to module return type)
Runtime:
* Updated inline/loop unroll cost model to better handle assignments
(Fixes https://github.com/Roblox/luau/issues/978)
* `math.noise` speed was improved by ~30%
* `table.concat` speed was improved by ~5-7%
* `tonumber` and `tostring` now have fastcall paths that execute ~1.5x
and ~2.5x faster respectively (fixes#777)
* Fixed crash in `luaL_typename` when index refers to a non-existing
value
* Fixed potential out of memory scenario when using `string.sub` or
`string.char` in a loop
* Fixed behavior of some fastcall builtins when called without arguments
under -O2 to match original functions
* Support for native code execution in VM is now enabled by default
(note: native code still has to be generated explicitly)
* `Codegen::compile` now accepts `CodeGen_OnlyNativeModules` flag. When
set, only modules that have a `--!native` hot-comment at the top will be
compiled to native code
In our new typechecker:
* Generic type packs are no longer considered to be variadic during
unification
* Timeout and cancellation now works in new solver
* Fixed false positive errors around 'table' and 'function' type
refinements
* Table literals now use covariant unification rules. This is sound
since literal has no type specified and has no aliases
* Fixed issues with blocked types escaping the constraint solver
* Fixed more places where error messages that should've been suppressed
were still reported
* Fixed errors when iterating over a top table type
In our native code generation (jit):
* 'DebugLuauAbortingChecks' flag is now supported on A64
* LOP_NEWCLOSURE has been translated to IR
