luau/Analysis/src/TypeFamily.cpp

1025 lines
39 KiB
C++
Raw Normal View History

Sync to upstream/release/576 (#928) * `ClassType` can now have an indexer defined on it. This allows custom types to be used in `t[x]` expressions. * Fixed search for closest executable breakpoint line. Previously, breakpoints might have been skipped in `else` blocks at the end of a function * Fixed how unification is performed for two optional types `a? <: b?`, previously it might have unified either 'a' or 'b' with 'nil'. Note that this fix is not enabled by default yet (see the list in `ExperimentalFlags.h`) In the new type solver, a concept of 'Type Families' has been introduced. Type families can be thought of as type aliases with custom type inference/reduction logic included with them. For example, we can have an `Add<T, U>` type family that will resolve the type that is the result of adding two values together. This will help type inference to figure out what 'T' and 'U' might be when explicit type annotations are not provided. In this update we don't define any type families, but they will be added in the near future. It is also possible for Luau embedders to define their own type families in the global/environment scope. Other changes include: * Fixed scope used to find out which generic types should be included in the function generic type list * Fixed a crash after cyclic bound types were created during unification And in native code generation (jit): * Use of arm64 target on M1 now requires macOS 13 * Entry into native code has been optimized. This is especially important for coroutine call/pcall performance as they involve going through a C call frame * LOP_LOADK(X) translation into IR has been improved to enable type tag/constant propagation * arm64 can use integer immediate values to synthesize floating-point values * x64 assembler removes duplicate 64bit numbers from the data section to save space * Linux `perf` can now be used to profile native Luau code (when running with --codegen-perf CLI argument)
2023-05-12 18:50:47 +01:00
// This file is part of the Luau programming language and is licensed under MIT License; see LICENSE.txt for details
#include "Luau/TypeFamily.h"
#include "Luau/ConstraintSolver.h"
Sync to upstream/release/576 (#928) * `ClassType` can now have an indexer defined on it. This allows custom types to be used in `t[x]` expressions. * Fixed search for closest executable breakpoint line. Previously, breakpoints might have been skipped in `else` blocks at the end of a function * Fixed how unification is performed for two optional types `a? <: b?`, previously it might have unified either 'a' or 'b' with 'nil'. Note that this fix is not enabled by default yet (see the list in `ExperimentalFlags.h`) In the new type solver, a concept of 'Type Families' has been introduced. Type families can be thought of as type aliases with custom type inference/reduction logic included with them. For example, we can have an `Add<T, U>` type family that will resolve the type that is the result of adding two values together. This will help type inference to figure out what 'T' and 'U' might be when explicit type annotations are not provided. In this update we don't define any type families, but they will be added in the near future. It is also possible for Luau embedders to define their own type families in the global/environment scope. Other changes include: * Fixed scope used to find out which generic types should be included in the function generic type list * Fixed a crash after cyclic bound types were created during unification And in native code generation (jit): * Use of arm64 target on M1 now requires macOS 13 * Entry into native code has been optimized. This is especially important for coroutine call/pcall performance as they involve going through a C call frame * LOP_LOADK(X) translation into IR has been improved to enable type tag/constant propagation * arm64 can use integer immediate values to synthesize floating-point values * x64 assembler removes duplicate 64bit numbers from the data section to save space * Linux `perf` can now be used to profile native Luau code (when running with --codegen-perf CLI argument)
2023-05-12 18:50:47 +01:00
#include "Luau/DenseHash.h"
#include "Luau/Instantiation.h"
#include "Luau/Normalize.h"
#include "Luau/Simplify.h"
Sync to upstream/release/576 (#928) * `ClassType` can now have an indexer defined on it. This allows custom types to be used in `t[x]` expressions. * Fixed search for closest executable breakpoint line. Previously, breakpoints might have been skipped in `else` blocks at the end of a function * Fixed how unification is performed for two optional types `a? <: b?`, previously it might have unified either 'a' or 'b' with 'nil'. Note that this fix is not enabled by default yet (see the list in `ExperimentalFlags.h`) In the new type solver, a concept of 'Type Families' has been introduced. Type families can be thought of as type aliases with custom type inference/reduction logic included with them. For example, we can have an `Add<T, U>` type family that will resolve the type that is the result of adding two values together. This will help type inference to figure out what 'T' and 'U' might be when explicit type annotations are not provided. In this update we don't define any type families, but they will be added in the near future. It is also possible for Luau embedders to define their own type families in the global/environment scope. Other changes include: * Fixed scope used to find out which generic types should be included in the function generic type list * Fixed a crash after cyclic bound types were created during unification And in native code generation (jit): * Use of arm64 target on M1 now requires macOS 13 * Entry into native code has been optimized. This is especially important for coroutine call/pcall performance as they involve going through a C call frame * LOP_LOADK(X) translation into IR has been improved to enable type tag/constant propagation * arm64 can use integer immediate values to synthesize floating-point values * x64 assembler removes duplicate 64bit numbers from the data section to save space * Linux `perf` can now be used to profile native Luau code (when running with --codegen-perf CLI argument)
2023-05-12 18:50:47 +01:00
#include "Luau/Substitution.h"
#include "Luau/Subtyping.h"
Sync to upstream/release/576 (#928) * `ClassType` can now have an indexer defined on it. This allows custom types to be used in `t[x]` expressions. * Fixed search for closest executable breakpoint line. Previously, breakpoints might have been skipped in `else` blocks at the end of a function * Fixed how unification is performed for two optional types `a? <: b?`, previously it might have unified either 'a' or 'b' with 'nil'. Note that this fix is not enabled by default yet (see the list in `ExperimentalFlags.h`) In the new type solver, a concept of 'Type Families' has been introduced. Type families can be thought of as type aliases with custom type inference/reduction logic included with them. For example, we can have an `Add<T, U>` type family that will resolve the type that is the result of adding two values together. This will help type inference to figure out what 'T' and 'U' might be when explicit type annotations are not provided. In this update we don't define any type families, but they will be added in the near future. It is also possible for Luau embedders to define their own type families in the global/environment scope. Other changes include: * Fixed scope used to find out which generic types should be included in the function generic type list * Fixed a crash after cyclic bound types were created during unification And in native code generation (jit): * Use of arm64 target on M1 now requires macOS 13 * Entry into native code has been optimized. This is especially important for coroutine call/pcall performance as they involve going through a C call frame * LOP_LOADK(X) translation into IR has been improved to enable type tag/constant propagation * arm64 can use integer immediate values to synthesize floating-point values * x64 assembler removes duplicate 64bit numbers from the data section to save space * Linux `perf` can now be used to profile native Luau code (when running with --codegen-perf CLI argument)
2023-05-12 18:50:47 +01:00
#include "Luau/ToString.h"
#include "Luau/TxnLog.h"
#include "Luau/TypeCheckLimits.h"
Sync to upstream/release/577 (#934) Lots of things going on this week: * Fix a crash that could occur in the presence of a cyclic union. We shouldn't be creating cyclic unions, but we shouldn't be crashing when they arise either. * Minor cleanup of `luau_precall` * Internal change to make L->top handling slightly more uniform * Optimize SETGLOBAL & GETGLOBAL fallback C functions. * https://github.com/Roblox/luau/pull/929 * The syntax to the `luau-reduce` commandline tool has changed. It now accepts a script, a command to execute, and an error to search for. It no longer automatically passes the script to the command which makes it a lot more flexible. Also be warned that it edits the script it is passed **in place**. Do not point it at something that is not in source control! New solver * Switch to a greedier but more fallible algorithm for simplifying union and intersection types that are created as part of refinement calculation. This has much better and more predictable performance. * Fix a constraint cycle in recursive function calls. * Much improved inference of binary addition. Functions like `function add(x, y) return x + y end` can now be inferred without annotations. We also accurately typecheck calls to functions like this. * Many small bugfixes surrounding things like table indexers * Add support for indexers on class types. This was previously added to the old solver; we now add it to the new one for feature parity. JIT * https://github.com/Roblox/luau/pull/931 * Fuse key.value and key.tt loads for CEHCK_SLOT_MATCH in A64 * Implement remaining aliases of BFM for A64 * Implement new callinfo flag for A64 * Add instruction simplification for int->num->int conversion chains * Don't even load execdata for X64 calls * Treat opcode fallbacks the same as manually written fallbacks --------- Co-authored-by: Arseny Kapoulkine <arseny.kapoulkine@gmail.com> Co-authored-by: Vyacheslav Egorov <vegorov@roblox.com>
2023-05-19 20:37:30 +01:00
#include "Luau/TypeUtils.h"
#include "Luau/Unifier2.h"
#include "Luau/VisitType.h"
Sync to upstream/release/576 (#928) * `ClassType` can now have an indexer defined on it. This allows custom types to be used in `t[x]` expressions. * Fixed search for closest executable breakpoint line. Previously, breakpoints might have been skipped in `else` blocks at the end of a function * Fixed how unification is performed for two optional types `a? <: b?`, previously it might have unified either 'a' or 'b' with 'nil'. Note that this fix is not enabled by default yet (see the list in `ExperimentalFlags.h`) In the new type solver, a concept of 'Type Families' has been introduced. Type families can be thought of as type aliases with custom type inference/reduction logic included with them. For example, we can have an `Add<T, U>` type family that will resolve the type that is the result of adding two values together. This will help type inference to figure out what 'T' and 'U' might be when explicit type annotations are not provided. In this update we don't define any type families, but they will be added in the near future. It is also possible for Luau embedders to define their own type families in the global/environment scope. Other changes include: * Fixed scope used to find out which generic types should be included in the function generic type list * Fixed a crash after cyclic bound types were created during unification And in native code generation (jit): * Use of arm64 target on M1 now requires macOS 13 * Entry into native code has been optimized. This is especially important for coroutine call/pcall performance as they involve going through a C call frame * LOP_LOADK(X) translation into IR has been improved to enable type tag/constant propagation * arm64 can use integer immediate values to synthesize floating-point values * x64 assembler removes duplicate 64bit numbers from the data section to save space * Linux `perf` can now be used to profile native Luau code (when running with --codegen-perf CLI argument)
2023-05-12 18:50:47 +01:00
LUAU_DYNAMIC_FASTINTVARIABLE(LuauTypeFamilyGraphReductionMaximumSteps, 1'000'000);
namespace Luau
{
struct InstanceCollector : TypeOnceVisitor
{
std::deque<TypeId> tys;
std::deque<TypePackId> tps;
bool visit(TypeId ty, const TypeFamilyInstanceType&) override
{
// TypeOnceVisitor performs a depth-first traversal in the absence of
// cycles. This means that by pushing to the front of the queue, we will
// try to reduce deeper instances first if we start with the first thing
// in the queue. Consider Add<Add<Add<number, number>, number>, number>:
// we want to reduce the innermost Add<number, number> instantiation
// first.
tys.push_front(ty);
return true;
}
Sync to upstream/release/577 (#934) Lots of things going on this week: * Fix a crash that could occur in the presence of a cyclic union. We shouldn't be creating cyclic unions, but we shouldn't be crashing when they arise either. * Minor cleanup of `luau_precall` * Internal change to make L->top handling slightly more uniform * Optimize SETGLOBAL & GETGLOBAL fallback C functions. * https://github.com/Roblox/luau/pull/929 * The syntax to the `luau-reduce` commandline tool has changed. It now accepts a script, a command to execute, and an error to search for. It no longer automatically passes the script to the command which makes it a lot more flexible. Also be warned that it edits the script it is passed **in place**. Do not point it at something that is not in source control! New solver * Switch to a greedier but more fallible algorithm for simplifying union and intersection types that are created as part of refinement calculation. This has much better and more predictable performance. * Fix a constraint cycle in recursive function calls. * Much improved inference of binary addition. Functions like `function add(x, y) return x + y end` can now be inferred without annotations. We also accurately typecheck calls to functions like this. * Many small bugfixes surrounding things like table indexers * Add support for indexers on class types. This was previously added to the old solver; we now add it to the new one for feature parity. JIT * https://github.com/Roblox/luau/pull/931 * Fuse key.value and key.tt loads for CEHCK_SLOT_MATCH in A64 * Implement remaining aliases of BFM for A64 * Implement new callinfo flag for A64 * Add instruction simplification for int->num->int conversion chains * Don't even load execdata for X64 calls * Treat opcode fallbacks the same as manually written fallbacks --------- Co-authored-by: Arseny Kapoulkine <arseny.kapoulkine@gmail.com> Co-authored-by: Vyacheslav Egorov <vegorov@roblox.com>
2023-05-19 20:37:30 +01:00
bool visit(TypeId ty, const ClassType&) override
{
return false;
}
Sync to upstream/release/576 (#928) * `ClassType` can now have an indexer defined on it. This allows custom types to be used in `t[x]` expressions. * Fixed search for closest executable breakpoint line. Previously, breakpoints might have been skipped in `else` blocks at the end of a function * Fixed how unification is performed for two optional types `a? <: b?`, previously it might have unified either 'a' or 'b' with 'nil'. Note that this fix is not enabled by default yet (see the list in `ExperimentalFlags.h`) In the new type solver, a concept of 'Type Families' has been introduced. Type families can be thought of as type aliases with custom type inference/reduction logic included with them. For example, we can have an `Add<T, U>` type family that will resolve the type that is the result of adding two values together. This will help type inference to figure out what 'T' and 'U' might be when explicit type annotations are not provided. In this update we don't define any type families, but they will be added in the near future. It is also possible for Luau embedders to define their own type families in the global/environment scope. Other changes include: * Fixed scope used to find out which generic types should be included in the function generic type list * Fixed a crash after cyclic bound types were created during unification And in native code generation (jit): * Use of arm64 target on M1 now requires macOS 13 * Entry into native code has been optimized. This is especially important for coroutine call/pcall performance as they involve going through a C call frame * LOP_LOADK(X) translation into IR has been improved to enable type tag/constant propagation * arm64 can use integer immediate values to synthesize floating-point values * x64 assembler removes duplicate 64bit numbers from the data section to save space * Linux `perf` can now be used to profile native Luau code (when running with --codegen-perf CLI argument)
2023-05-12 18:50:47 +01:00
bool visit(TypePackId tp, const TypeFamilyInstanceTypePack&) override
{
// TypeOnceVisitor performs a depth-first traversal in the absence of
// cycles. This means that by pushing to the front of the queue, we will
// try to reduce deeper instances first if we start with the first thing
// in the queue. Consider Add<Add<Add<number, number>, number>, number>:
// we want to reduce the innermost Add<number, number> instantiation
// first.
tps.push_front(tp);
return true;
}
};
struct FamilyReducer
{
TypeFamilyContext ctx;
Sync to upstream/release/576 (#928) * `ClassType` can now have an indexer defined on it. This allows custom types to be used in `t[x]` expressions. * Fixed search for closest executable breakpoint line. Previously, breakpoints might have been skipped in `else` blocks at the end of a function * Fixed how unification is performed for two optional types `a? <: b?`, previously it might have unified either 'a' or 'b' with 'nil'. Note that this fix is not enabled by default yet (see the list in `ExperimentalFlags.h`) In the new type solver, a concept of 'Type Families' has been introduced. Type families can be thought of as type aliases with custom type inference/reduction logic included with them. For example, we can have an `Add<T, U>` type family that will resolve the type that is the result of adding two values together. This will help type inference to figure out what 'T' and 'U' might be when explicit type annotations are not provided. In this update we don't define any type families, but they will be added in the near future. It is also possible for Luau embedders to define their own type families in the global/environment scope. Other changes include: * Fixed scope used to find out which generic types should be included in the function generic type list * Fixed a crash after cyclic bound types were created during unification And in native code generation (jit): * Use of arm64 target on M1 now requires macOS 13 * Entry into native code has been optimized. This is especially important for coroutine call/pcall performance as they involve going through a C call frame * LOP_LOADK(X) translation into IR has been improved to enable type tag/constant propagation * arm64 can use integer immediate values to synthesize floating-point values * x64 assembler removes duplicate 64bit numbers from the data section to save space * Linux `perf` can now be used to profile native Luau code (when running with --codegen-perf CLI argument)
2023-05-12 18:50:47 +01:00
std::deque<TypeId> queuedTys;
std::deque<TypePackId> queuedTps;
DenseHashSet<const void*> irreducible{nullptr};
FamilyGraphReductionResult result;
bool force = false;
// Local to the constraint being reduced.
Location location;
FamilyReducer(std::deque<TypeId> queuedTys, std::deque<TypePackId> queuedTps, Location location, TypeFamilyContext ctx, bool force = false)
: ctx(ctx)
, queuedTys(std::move(queuedTys))
, queuedTps(std::move(queuedTps))
Sync to upstream/release/576 (#928) * `ClassType` can now have an indexer defined on it. This allows custom types to be used in `t[x]` expressions. * Fixed search for closest executable breakpoint line. Previously, breakpoints might have been skipped in `else` blocks at the end of a function * Fixed how unification is performed for two optional types `a? <: b?`, previously it might have unified either 'a' or 'b' with 'nil'. Note that this fix is not enabled by default yet (see the list in `ExperimentalFlags.h`) In the new type solver, a concept of 'Type Families' has been introduced. Type families can be thought of as type aliases with custom type inference/reduction logic included with them. For example, we can have an `Add<T, U>` type family that will resolve the type that is the result of adding two values together. This will help type inference to figure out what 'T' and 'U' might be when explicit type annotations are not provided. In this update we don't define any type families, but they will be added in the near future. It is also possible for Luau embedders to define their own type families in the global/environment scope. Other changes include: * Fixed scope used to find out which generic types should be included in the function generic type list * Fixed a crash after cyclic bound types were created during unification And in native code generation (jit): * Use of arm64 target on M1 now requires macOS 13 * Entry into native code has been optimized. This is especially important for coroutine call/pcall performance as they involve going through a C call frame * LOP_LOADK(X) translation into IR has been improved to enable type tag/constant propagation * arm64 can use integer immediate values to synthesize floating-point values * x64 assembler removes duplicate 64bit numbers from the data section to save space * Linux `perf` can now be used to profile native Luau code (when running with --codegen-perf CLI argument)
2023-05-12 18:50:47 +01:00
, force(force)
, location(location)
Sync to upstream/release/576 (#928) * `ClassType` can now have an indexer defined on it. This allows custom types to be used in `t[x]` expressions. * Fixed search for closest executable breakpoint line. Previously, breakpoints might have been skipped in `else` blocks at the end of a function * Fixed how unification is performed for two optional types `a? <: b?`, previously it might have unified either 'a' or 'b' with 'nil'. Note that this fix is not enabled by default yet (see the list in `ExperimentalFlags.h`) In the new type solver, a concept of 'Type Families' has been introduced. Type families can be thought of as type aliases with custom type inference/reduction logic included with them. For example, we can have an `Add<T, U>` type family that will resolve the type that is the result of adding two values together. This will help type inference to figure out what 'T' and 'U' might be when explicit type annotations are not provided. In this update we don't define any type families, but they will be added in the near future. It is also possible for Luau embedders to define their own type families in the global/environment scope. Other changes include: * Fixed scope used to find out which generic types should be included in the function generic type list * Fixed a crash after cyclic bound types were created during unification And in native code generation (jit): * Use of arm64 target on M1 now requires macOS 13 * Entry into native code has been optimized. This is especially important for coroutine call/pcall performance as they involve going through a C call frame * LOP_LOADK(X) translation into IR has been improved to enable type tag/constant propagation * arm64 can use integer immediate values to synthesize floating-point values * x64 assembler removes duplicate 64bit numbers from the data section to save space * Linux `perf` can now be used to profile native Luau code (when running with --codegen-perf CLI argument)
2023-05-12 18:50:47 +01:00
{
}
enum class SkipTestResult
{
Irreducible,
Defer,
Okay,
};
SkipTestResult testForSkippability(TypeId ty)
{
ty = follow(ty);
Sync to upstream/release/576 (#928) * `ClassType` can now have an indexer defined on it. This allows custom types to be used in `t[x]` expressions. * Fixed search for closest executable breakpoint line. Previously, breakpoints might have been skipped in `else` blocks at the end of a function * Fixed how unification is performed for two optional types `a? <: b?`, previously it might have unified either 'a' or 'b' with 'nil'. Note that this fix is not enabled by default yet (see the list in `ExperimentalFlags.h`) In the new type solver, a concept of 'Type Families' has been introduced. Type families can be thought of as type aliases with custom type inference/reduction logic included with them. For example, we can have an `Add<T, U>` type family that will resolve the type that is the result of adding two values together. This will help type inference to figure out what 'T' and 'U' might be when explicit type annotations are not provided. In this update we don't define any type families, but they will be added in the near future. It is also possible for Luau embedders to define their own type families in the global/environment scope. Other changes include: * Fixed scope used to find out which generic types should be included in the function generic type list * Fixed a crash after cyclic bound types were created during unification And in native code generation (jit): * Use of arm64 target on M1 now requires macOS 13 * Entry into native code has been optimized. This is especially important for coroutine call/pcall performance as they involve going through a C call frame * LOP_LOADK(X) translation into IR has been improved to enable type tag/constant propagation * arm64 can use integer immediate values to synthesize floating-point values * x64 assembler removes duplicate 64bit numbers from the data section to save space * Linux `perf` can now be used to profile native Luau code (when running with --codegen-perf CLI argument)
2023-05-12 18:50:47 +01:00
if (is<TypeFamilyInstanceType>(ty))
Sync to upstream/release/576 (#928) * `ClassType` can now have an indexer defined on it. This allows custom types to be used in `t[x]` expressions. * Fixed search for closest executable breakpoint line. Previously, breakpoints might have been skipped in `else` blocks at the end of a function * Fixed how unification is performed for two optional types `a? <: b?`, previously it might have unified either 'a' or 'b' with 'nil'. Note that this fix is not enabled by default yet (see the list in `ExperimentalFlags.h`) In the new type solver, a concept of 'Type Families' has been introduced. Type families can be thought of as type aliases with custom type inference/reduction logic included with them. For example, we can have an `Add<T, U>` type family that will resolve the type that is the result of adding two values together. This will help type inference to figure out what 'T' and 'U' might be when explicit type annotations are not provided. In this update we don't define any type families, but they will be added in the near future. It is also possible for Luau embedders to define their own type families in the global/environment scope. Other changes include: * Fixed scope used to find out which generic types should be included in the function generic type list * Fixed a crash after cyclic bound types were created during unification And in native code generation (jit): * Use of arm64 target on M1 now requires macOS 13 * Entry into native code has been optimized. This is especially important for coroutine call/pcall performance as they involve going through a C call frame * LOP_LOADK(X) translation into IR has been improved to enable type tag/constant propagation * arm64 can use integer immediate values to synthesize floating-point values * x64 assembler removes duplicate 64bit numbers from the data section to save space * Linux `perf` can now be used to profile native Luau code (when running with --codegen-perf CLI argument)
2023-05-12 18:50:47 +01:00
{
if (!irreducible.contains(ty))
return SkipTestResult::Defer;
else
return SkipTestResult::Irreducible;
}
else if (is<GenericType>(ty))
Sync to upstream/release/576 (#928) * `ClassType` can now have an indexer defined on it. This allows custom types to be used in `t[x]` expressions. * Fixed search for closest executable breakpoint line. Previously, breakpoints might have been skipped in `else` blocks at the end of a function * Fixed how unification is performed for two optional types `a? <: b?`, previously it might have unified either 'a' or 'b' with 'nil'. Note that this fix is not enabled by default yet (see the list in `ExperimentalFlags.h`) In the new type solver, a concept of 'Type Families' has been introduced. Type families can be thought of as type aliases with custom type inference/reduction logic included with them. For example, we can have an `Add<T, U>` type family that will resolve the type that is the result of adding two values together. This will help type inference to figure out what 'T' and 'U' might be when explicit type annotations are not provided. In this update we don't define any type families, but they will be added in the near future. It is also possible for Luau embedders to define their own type families in the global/environment scope. Other changes include: * Fixed scope used to find out which generic types should be included in the function generic type list * Fixed a crash after cyclic bound types were created during unification And in native code generation (jit): * Use of arm64 target on M1 now requires macOS 13 * Entry into native code has been optimized. This is especially important for coroutine call/pcall performance as they involve going through a C call frame * LOP_LOADK(X) translation into IR has been improved to enable type tag/constant propagation * arm64 can use integer immediate values to synthesize floating-point values * x64 assembler removes duplicate 64bit numbers from the data section to save space * Linux `perf` can now be used to profile native Luau code (when running with --codegen-perf CLI argument)
2023-05-12 18:50:47 +01:00
{
return SkipTestResult::Irreducible;
}
return SkipTestResult::Okay;
}
SkipTestResult testForSkippability(TypePackId ty)
{
ty = follow(ty);
Sync to upstream/release/576 (#928) * `ClassType` can now have an indexer defined on it. This allows custom types to be used in `t[x]` expressions. * Fixed search for closest executable breakpoint line. Previously, breakpoints might have been skipped in `else` blocks at the end of a function * Fixed how unification is performed for two optional types `a? <: b?`, previously it might have unified either 'a' or 'b' with 'nil'. Note that this fix is not enabled by default yet (see the list in `ExperimentalFlags.h`) In the new type solver, a concept of 'Type Families' has been introduced. Type families can be thought of as type aliases with custom type inference/reduction logic included with them. For example, we can have an `Add<T, U>` type family that will resolve the type that is the result of adding two values together. This will help type inference to figure out what 'T' and 'U' might be when explicit type annotations are not provided. In this update we don't define any type families, but they will be added in the near future. It is also possible for Luau embedders to define their own type families in the global/environment scope. Other changes include: * Fixed scope used to find out which generic types should be included in the function generic type list * Fixed a crash after cyclic bound types were created during unification And in native code generation (jit): * Use of arm64 target on M1 now requires macOS 13 * Entry into native code has been optimized. This is especially important for coroutine call/pcall performance as they involve going through a C call frame * LOP_LOADK(X) translation into IR has been improved to enable type tag/constant propagation * arm64 can use integer immediate values to synthesize floating-point values * x64 assembler removes duplicate 64bit numbers from the data section to save space * Linux `perf` can now be used to profile native Luau code (when running with --codegen-perf CLI argument)
2023-05-12 18:50:47 +01:00
if (is<TypeFamilyInstanceTypePack>(ty))
Sync to upstream/release/576 (#928) * `ClassType` can now have an indexer defined on it. This allows custom types to be used in `t[x]` expressions. * Fixed search for closest executable breakpoint line. Previously, breakpoints might have been skipped in `else` blocks at the end of a function * Fixed how unification is performed for two optional types `a? <: b?`, previously it might have unified either 'a' or 'b' with 'nil'. Note that this fix is not enabled by default yet (see the list in `ExperimentalFlags.h`) In the new type solver, a concept of 'Type Families' has been introduced. Type families can be thought of as type aliases with custom type inference/reduction logic included with them. For example, we can have an `Add<T, U>` type family that will resolve the type that is the result of adding two values together. This will help type inference to figure out what 'T' and 'U' might be when explicit type annotations are not provided. In this update we don't define any type families, but they will be added in the near future. It is also possible for Luau embedders to define their own type families in the global/environment scope. Other changes include: * Fixed scope used to find out which generic types should be included in the function generic type list * Fixed a crash after cyclic bound types were created during unification And in native code generation (jit): * Use of arm64 target on M1 now requires macOS 13 * Entry into native code has been optimized. This is especially important for coroutine call/pcall performance as they involve going through a C call frame * LOP_LOADK(X) translation into IR has been improved to enable type tag/constant propagation * arm64 can use integer immediate values to synthesize floating-point values * x64 assembler removes duplicate 64bit numbers from the data section to save space * Linux `perf` can now be used to profile native Luau code (when running with --codegen-perf CLI argument)
2023-05-12 18:50:47 +01:00
{
if (!irreducible.contains(ty))
return SkipTestResult::Defer;
else
return SkipTestResult::Irreducible;
}
else if (is<GenericTypePack>(ty))
Sync to upstream/release/576 (#928) * `ClassType` can now have an indexer defined on it. This allows custom types to be used in `t[x]` expressions. * Fixed search for closest executable breakpoint line. Previously, breakpoints might have been skipped in `else` blocks at the end of a function * Fixed how unification is performed for two optional types `a? <: b?`, previously it might have unified either 'a' or 'b' with 'nil'. Note that this fix is not enabled by default yet (see the list in `ExperimentalFlags.h`) In the new type solver, a concept of 'Type Families' has been introduced. Type families can be thought of as type aliases with custom type inference/reduction logic included with them. For example, we can have an `Add<T, U>` type family that will resolve the type that is the result of adding two values together. This will help type inference to figure out what 'T' and 'U' might be when explicit type annotations are not provided. In this update we don't define any type families, but they will be added in the near future. It is also possible for Luau embedders to define their own type families in the global/environment scope. Other changes include: * Fixed scope used to find out which generic types should be included in the function generic type list * Fixed a crash after cyclic bound types were created during unification And in native code generation (jit): * Use of arm64 target on M1 now requires macOS 13 * Entry into native code has been optimized. This is especially important for coroutine call/pcall performance as they involve going through a C call frame * LOP_LOADK(X) translation into IR has been improved to enable type tag/constant propagation * arm64 can use integer immediate values to synthesize floating-point values * x64 assembler removes duplicate 64bit numbers from the data section to save space * Linux `perf` can now be used to profile native Luau code (when running with --codegen-perf CLI argument)
2023-05-12 18:50:47 +01:00
{
return SkipTestResult::Irreducible;
}
return SkipTestResult::Okay;
}
template<typename T>
void replace(T subject, T replacement)
{
asMutable(subject)->ty.template emplace<Unifiable::Bound<T>>(replacement);
Sync to upstream/release/576 (#928) * `ClassType` can now have an indexer defined on it. This allows custom types to be used in `t[x]` expressions. * Fixed search for closest executable breakpoint line. Previously, breakpoints might have been skipped in `else` blocks at the end of a function * Fixed how unification is performed for two optional types `a? <: b?`, previously it might have unified either 'a' or 'b' with 'nil'. Note that this fix is not enabled by default yet (see the list in `ExperimentalFlags.h`) In the new type solver, a concept of 'Type Families' has been introduced. Type families can be thought of as type aliases with custom type inference/reduction logic included with them. For example, we can have an `Add<T, U>` type family that will resolve the type that is the result of adding two values together. This will help type inference to figure out what 'T' and 'U' might be when explicit type annotations are not provided. In this update we don't define any type families, but they will be added in the near future. It is also possible for Luau embedders to define their own type families in the global/environment scope. Other changes include: * Fixed scope used to find out which generic types should be included in the function generic type list * Fixed a crash after cyclic bound types were created during unification And in native code generation (jit): * Use of arm64 target on M1 now requires macOS 13 * Entry into native code has been optimized. This is especially important for coroutine call/pcall performance as they involve going through a C call frame * LOP_LOADK(X) translation into IR has been improved to enable type tag/constant propagation * arm64 can use integer immediate values to synthesize floating-point values * x64 assembler removes duplicate 64bit numbers from the data section to save space * Linux `perf` can now be used to profile native Luau code (when running with --codegen-perf CLI argument)
2023-05-12 18:50:47 +01:00
if constexpr (std::is_same_v<T, TypeId>)
result.reducedTypes.insert(subject);
else if constexpr (std::is_same_v<T, TypePackId>)
result.reducedPacks.insert(subject);
}
template<typename T>
void handleFamilyReduction(T subject, TypeFamilyReductionResult<T> reduction)
{
if (reduction.result)
replace(subject, *reduction.result);
else
{
irreducible.insert(subject);
if (reduction.uninhabited || force)
{
if constexpr (std::is_same_v<T, TypeId>)
result.errors.push_back(TypeError{location, UninhabitedTypeFamily{subject}});
else if constexpr (std::is_same_v<T, TypePackId>)
result.errors.push_back(TypeError{location, UninhabitedTypePackFamily{subject}});
}
else if (!reduction.uninhabited && !force)
{
for (TypeId b : reduction.blockedTypes)
result.blockedTypes.insert(b);
for (TypePackId b : reduction.blockedPacks)
result.blockedPacks.insert(b);
}
}
}
bool done()
{
return queuedTys.empty() && queuedTps.empty();
}
template<typename T, typename I>
bool testParameters(T subject, const I* tfit)
{
for (TypeId p : tfit->typeArguments)
{
SkipTestResult skip = testForSkippability(p);
if (skip == SkipTestResult::Irreducible)
{
irreducible.insert(subject);
return false;
}
else if (skip == SkipTestResult::Defer)
{
if constexpr (std::is_same_v<T, TypeId>)
queuedTys.push_back(subject);
else if constexpr (std::is_same_v<T, TypePackId>)
queuedTps.push_back(subject);
return false;
}
}
for (TypePackId p : tfit->packArguments)
{
SkipTestResult skip = testForSkippability(p);
if (skip == SkipTestResult::Irreducible)
{
irreducible.insert(subject);
return false;
}
else if (skip == SkipTestResult::Defer)
{
if constexpr (std::is_same_v<T, TypeId>)
queuedTys.push_back(subject);
else if constexpr (std::is_same_v<T, TypePackId>)
queuedTps.push_back(subject);
return false;
}
}
return true;
}
void stepType()
{
TypeId subject = follow(queuedTys.front());
Sync to upstream/release/576 (#928) * `ClassType` can now have an indexer defined on it. This allows custom types to be used in `t[x]` expressions. * Fixed search for closest executable breakpoint line. Previously, breakpoints might have been skipped in `else` blocks at the end of a function * Fixed how unification is performed for two optional types `a? <: b?`, previously it might have unified either 'a' or 'b' with 'nil'. Note that this fix is not enabled by default yet (see the list in `ExperimentalFlags.h`) In the new type solver, a concept of 'Type Families' has been introduced. Type families can be thought of as type aliases with custom type inference/reduction logic included with them. For example, we can have an `Add<T, U>` type family that will resolve the type that is the result of adding two values together. This will help type inference to figure out what 'T' and 'U' might be when explicit type annotations are not provided. In this update we don't define any type families, but they will be added in the near future. It is also possible for Luau embedders to define their own type families in the global/environment scope. Other changes include: * Fixed scope used to find out which generic types should be included in the function generic type list * Fixed a crash after cyclic bound types were created during unification And in native code generation (jit): * Use of arm64 target on M1 now requires macOS 13 * Entry into native code has been optimized. This is especially important for coroutine call/pcall performance as they involve going through a C call frame * LOP_LOADK(X) translation into IR has been improved to enable type tag/constant propagation * arm64 can use integer immediate values to synthesize floating-point values * x64 assembler removes duplicate 64bit numbers from the data section to save space * Linux `perf` can now be used to profile native Luau code (when running with --codegen-perf CLI argument)
2023-05-12 18:50:47 +01:00
queuedTys.pop_front();
if (irreducible.contains(subject))
return;
if (const TypeFamilyInstanceType* tfit = get<TypeFamilyInstanceType>(subject))
Sync to upstream/release/576 (#928) * `ClassType` can now have an indexer defined on it. This allows custom types to be used in `t[x]` expressions. * Fixed search for closest executable breakpoint line. Previously, breakpoints might have been skipped in `else` blocks at the end of a function * Fixed how unification is performed for two optional types `a? <: b?`, previously it might have unified either 'a' or 'b' with 'nil'. Note that this fix is not enabled by default yet (see the list in `ExperimentalFlags.h`) In the new type solver, a concept of 'Type Families' has been introduced. Type families can be thought of as type aliases with custom type inference/reduction logic included with them. For example, we can have an `Add<T, U>` type family that will resolve the type that is the result of adding two values together. This will help type inference to figure out what 'T' and 'U' might be when explicit type annotations are not provided. In this update we don't define any type families, but they will be added in the near future. It is also possible for Luau embedders to define their own type families in the global/environment scope. Other changes include: * Fixed scope used to find out which generic types should be included in the function generic type list * Fixed a crash after cyclic bound types were created during unification And in native code generation (jit): * Use of arm64 target on M1 now requires macOS 13 * Entry into native code has been optimized. This is especially important for coroutine call/pcall performance as they involve going through a C call frame * LOP_LOADK(X) translation into IR has been improved to enable type tag/constant propagation * arm64 can use integer immediate values to synthesize floating-point values * x64 assembler removes duplicate 64bit numbers from the data section to save space * Linux `perf` can now be used to profile native Luau code (when running with --codegen-perf CLI argument)
2023-05-12 18:50:47 +01:00
{
if (!testParameters(subject, tfit))
return;
TypeFamilyReductionResult<TypeId> result = tfit->family->reducer(tfit->typeArguments, tfit->packArguments, NotNull{&ctx});
Sync to upstream/release/576 (#928) * `ClassType` can now have an indexer defined on it. This allows custom types to be used in `t[x]` expressions. * Fixed search for closest executable breakpoint line. Previously, breakpoints might have been skipped in `else` blocks at the end of a function * Fixed how unification is performed for two optional types `a? <: b?`, previously it might have unified either 'a' or 'b' with 'nil'. Note that this fix is not enabled by default yet (see the list in `ExperimentalFlags.h`) In the new type solver, a concept of 'Type Families' has been introduced. Type families can be thought of as type aliases with custom type inference/reduction logic included with them. For example, we can have an `Add<T, U>` type family that will resolve the type that is the result of adding two values together. This will help type inference to figure out what 'T' and 'U' might be when explicit type annotations are not provided. In this update we don't define any type families, but they will be added in the near future. It is also possible for Luau embedders to define their own type families in the global/environment scope. Other changes include: * Fixed scope used to find out which generic types should be included in the function generic type list * Fixed a crash after cyclic bound types were created during unification And in native code generation (jit): * Use of arm64 target on M1 now requires macOS 13 * Entry into native code has been optimized. This is especially important for coroutine call/pcall performance as they involve going through a C call frame * LOP_LOADK(X) translation into IR has been improved to enable type tag/constant propagation * arm64 can use integer immediate values to synthesize floating-point values * x64 assembler removes duplicate 64bit numbers from the data section to save space * Linux `perf` can now be used to profile native Luau code (when running with --codegen-perf CLI argument)
2023-05-12 18:50:47 +01:00
handleFamilyReduction(subject, result);
}
}
void stepPack()
{
TypePackId subject = follow(queuedTps.front());
Sync to upstream/release/576 (#928) * `ClassType` can now have an indexer defined on it. This allows custom types to be used in `t[x]` expressions. * Fixed search for closest executable breakpoint line. Previously, breakpoints might have been skipped in `else` blocks at the end of a function * Fixed how unification is performed for two optional types `a? <: b?`, previously it might have unified either 'a' or 'b' with 'nil'. Note that this fix is not enabled by default yet (see the list in `ExperimentalFlags.h`) In the new type solver, a concept of 'Type Families' has been introduced. Type families can be thought of as type aliases with custom type inference/reduction logic included with them. For example, we can have an `Add<T, U>` type family that will resolve the type that is the result of adding two values together. This will help type inference to figure out what 'T' and 'U' might be when explicit type annotations are not provided. In this update we don't define any type families, but they will be added in the near future. It is also possible for Luau embedders to define their own type families in the global/environment scope. Other changes include: * Fixed scope used to find out which generic types should be included in the function generic type list * Fixed a crash after cyclic bound types were created during unification And in native code generation (jit): * Use of arm64 target on M1 now requires macOS 13 * Entry into native code has been optimized. This is especially important for coroutine call/pcall performance as they involve going through a C call frame * LOP_LOADK(X) translation into IR has been improved to enable type tag/constant propagation * arm64 can use integer immediate values to synthesize floating-point values * x64 assembler removes duplicate 64bit numbers from the data section to save space * Linux `perf` can now be used to profile native Luau code (when running with --codegen-perf CLI argument)
2023-05-12 18:50:47 +01:00
queuedTps.pop_front();
if (irreducible.contains(subject))
return;
if (const TypeFamilyInstanceTypePack* tfit = get<TypeFamilyInstanceTypePack>(subject))
Sync to upstream/release/576 (#928) * `ClassType` can now have an indexer defined on it. This allows custom types to be used in `t[x]` expressions. * Fixed search for closest executable breakpoint line. Previously, breakpoints might have been skipped in `else` blocks at the end of a function * Fixed how unification is performed for two optional types `a? <: b?`, previously it might have unified either 'a' or 'b' with 'nil'. Note that this fix is not enabled by default yet (see the list in `ExperimentalFlags.h`) In the new type solver, a concept of 'Type Families' has been introduced. Type families can be thought of as type aliases with custom type inference/reduction logic included with them. For example, we can have an `Add<T, U>` type family that will resolve the type that is the result of adding two values together. This will help type inference to figure out what 'T' and 'U' might be when explicit type annotations are not provided. In this update we don't define any type families, but they will be added in the near future. It is also possible for Luau embedders to define their own type families in the global/environment scope. Other changes include: * Fixed scope used to find out which generic types should be included in the function generic type list * Fixed a crash after cyclic bound types were created during unification And in native code generation (jit): * Use of arm64 target on M1 now requires macOS 13 * Entry into native code has been optimized. This is especially important for coroutine call/pcall performance as they involve going through a C call frame * LOP_LOADK(X) translation into IR has been improved to enable type tag/constant propagation * arm64 can use integer immediate values to synthesize floating-point values * x64 assembler removes duplicate 64bit numbers from the data section to save space * Linux `perf` can now be used to profile native Luau code (when running with --codegen-perf CLI argument)
2023-05-12 18:50:47 +01:00
{
if (!testParameters(subject, tfit))
return;
TypeFamilyReductionResult<TypePackId> result = tfit->family->reducer(tfit->typeArguments, tfit->packArguments, NotNull{&ctx});
Sync to upstream/release/576 (#928) * `ClassType` can now have an indexer defined on it. This allows custom types to be used in `t[x]` expressions. * Fixed search for closest executable breakpoint line. Previously, breakpoints might have been skipped in `else` blocks at the end of a function * Fixed how unification is performed for two optional types `a? <: b?`, previously it might have unified either 'a' or 'b' with 'nil'. Note that this fix is not enabled by default yet (see the list in `ExperimentalFlags.h`) In the new type solver, a concept of 'Type Families' has been introduced. Type families can be thought of as type aliases with custom type inference/reduction logic included with them. For example, we can have an `Add<T, U>` type family that will resolve the type that is the result of adding two values together. This will help type inference to figure out what 'T' and 'U' might be when explicit type annotations are not provided. In this update we don't define any type families, but they will be added in the near future. It is also possible for Luau embedders to define their own type families in the global/environment scope. Other changes include: * Fixed scope used to find out which generic types should be included in the function generic type list * Fixed a crash after cyclic bound types were created during unification And in native code generation (jit): * Use of arm64 target on M1 now requires macOS 13 * Entry into native code has been optimized. This is especially important for coroutine call/pcall performance as they involve going through a C call frame * LOP_LOADK(X) translation into IR has been improved to enable type tag/constant propagation * arm64 can use integer immediate values to synthesize floating-point values * x64 assembler removes duplicate 64bit numbers from the data section to save space * Linux `perf` can now be used to profile native Luau code (when running with --codegen-perf CLI argument)
2023-05-12 18:50:47 +01:00
handleFamilyReduction(subject, result);
}
}
void step()
{
if (!queuedTys.empty())
stepType();
else if (!queuedTps.empty())
stepPack();
}
};
static FamilyGraphReductionResult reduceFamiliesInternal(
std::deque<TypeId> queuedTys, std::deque<TypePackId> queuedTps, Location location, TypeFamilyContext ctx, bool force)
Sync to upstream/release/576 (#928) * `ClassType` can now have an indexer defined on it. This allows custom types to be used in `t[x]` expressions. * Fixed search for closest executable breakpoint line. Previously, breakpoints might have been skipped in `else` blocks at the end of a function * Fixed how unification is performed for two optional types `a? <: b?`, previously it might have unified either 'a' or 'b' with 'nil'. Note that this fix is not enabled by default yet (see the list in `ExperimentalFlags.h`) In the new type solver, a concept of 'Type Families' has been introduced. Type families can be thought of as type aliases with custom type inference/reduction logic included with them. For example, we can have an `Add<T, U>` type family that will resolve the type that is the result of adding two values together. This will help type inference to figure out what 'T' and 'U' might be when explicit type annotations are not provided. In this update we don't define any type families, but they will be added in the near future. It is also possible for Luau embedders to define their own type families in the global/environment scope. Other changes include: * Fixed scope used to find out which generic types should be included in the function generic type list * Fixed a crash after cyclic bound types were created during unification And in native code generation (jit): * Use of arm64 target on M1 now requires macOS 13 * Entry into native code has been optimized. This is especially important for coroutine call/pcall performance as they involve going through a C call frame * LOP_LOADK(X) translation into IR has been improved to enable type tag/constant propagation * arm64 can use integer immediate values to synthesize floating-point values * x64 assembler removes duplicate 64bit numbers from the data section to save space * Linux `perf` can now be used to profile native Luau code (when running with --codegen-perf CLI argument)
2023-05-12 18:50:47 +01:00
{
FamilyReducer reducer{std::move(queuedTys), std::move(queuedTps), location, ctx, force};
Sync to upstream/release/576 (#928) * `ClassType` can now have an indexer defined on it. This allows custom types to be used in `t[x]` expressions. * Fixed search for closest executable breakpoint line. Previously, breakpoints might have been skipped in `else` blocks at the end of a function * Fixed how unification is performed for two optional types `a? <: b?`, previously it might have unified either 'a' or 'b' with 'nil'. Note that this fix is not enabled by default yet (see the list in `ExperimentalFlags.h`) In the new type solver, a concept of 'Type Families' has been introduced. Type families can be thought of as type aliases with custom type inference/reduction logic included with them. For example, we can have an `Add<T, U>` type family that will resolve the type that is the result of adding two values together. This will help type inference to figure out what 'T' and 'U' might be when explicit type annotations are not provided. In this update we don't define any type families, but they will be added in the near future. It is also possible for Luau embedders to define their own type families in the global/environment scope. Other changes include: * Fixed scope used to find out which generic types should be included in the function generic type list * Fixed a crash after cyclic bound types were created during unification And in native code generation (jit): * Use of arm64 target on M1 now requires macOS 13 * Entry into native code has been optimized. This is especially important for coroutine call/pcall performance as they involve going through a C call frame * LOP_LOADK(X) translation into IR has been improved to enable type tag/constant propagation * arm64 can use integer immediate values to synthesize floating-point values * x64 assembler removes duplicate 64bit numbers from the data section to save space * Linux `perf` can now be used to profile native Luau code (when running with --codegen-perf CLI argument)
2023-05-12 18:50:47 +01:00
int iterationCount = 0;
while (!reducer.done())
{
reducer.step();
++iterationCount;
if (iterationCount > DFInt::LuauTypeFamilyGraphReductionMaximumSteps)
{
reducer.result.errors.push_back(TypeError{location, CodeTooComplex{}});
break;
}
}
return std::move(reducer.result);
}
FamilyGraphReductionResult reduceFamilies(TypeId entrypoint, Location location, TypeFamilyContext ctx, bool force)
{
InstanceCollector collector;
try
{
collector.traverse(entrypoint);
}
catch (RecursionLimitException&)
{
return FamilyGraphReductionResult{};
}
if (collector.tys.empty() && collector.tps.empty())
return {};
return reduceFamiliesInternal(std::move(collector.tys), std::move(collector.tps), location, ctx, force);
}
FamilyGraphReductionResult reduceFamilies(TypePackId entrypoint, Location location, TypeFamilyContext ctx, bool force)
{
InstanceCollector collector;
try
{
collector.traverse(entrypoint);
}
catch (RecursionLimitException&)
{
return FamilyGraphReductionResult{};
}
if (collector.tys.empty() && collector.tps.empty())
return {};
return reduceFamiliesInternal(std::move(collector.tys), std::move(collector.tps), location, ctx, force);
Sync to upstream/release/577 (#934) Lots of things going on this week: * Fix a crash that could occur in the presence of a cyclic union. We shouldn't be creating cyclic unions, but we shouldn't be crashing when they arise either. * Minor cleanup of `luau_precall` * Internal change to make L->top handling slightly more uniform * Optimize SETGLOBAL & GETGLOBAL fallback C functions. * https://github.com/Roblox/luau/pull/929 * The syntax to the `luau-reduce` commandline tool has changed. It now accepts a script, a command to execute, and an error to search for. It no longer automatically passes the script to the command which makes it a lot more flexible. Also be warned that it edits the script it is passed **in place**. Do not point it at something that is not in source control! New solver * Switch to a greedier but more fallible algorithm for simplifying union and intersection types that are created as part of refinement calculation. This has much better and more predictable performance. * Fix a constraint cycle in recursive function calls. * Much improved inference of binary addition. Functions like `function add(x, y) return x + y end` can now be inferred without annotations. We also accurately typecheck calls to functions like this. * Many small bugfixes surrounding things like table indexers * Add support for indexers on class types. This was previously added to the old solver; we now add it to the new one for feature parity. JIT * https://github.com/Roblox/luau/pull/931 * Fuse key.value and key.tt loads for CEHCK_SLOT_MATCH in A64 * Implement remaining aliases of BFM for A64 * Implement new callinfo flag for A64 * Add instruction simplification for int->num->int conversion chains * Don't even load execdata for X64 calls * Treat opcode fallbacks the same as manually written fallbacks --------- Co-authored-by: Arseny Kapoulkine <arseny.kapoulkine@gmail.com> Co-authored-by: Vyacheslav Egorov <vegorov@roblox.com>
2023-05-19 20:37:30 +01:00
}
bool isPending(TypeId ty, ConstraintSolver* solver)
Sync to upstream/release/577 (#934) Lots of things going on this week: * Fix a crash that could occur in the presence of a cyclic union. We shouldn't be creating cyclic unions, but we shouldn't be crashing when they arise either. * Minor cleanup of `luau_precall` * Internal change to make L->top handling slightly more uniform * Optimize SETGLOBAL & GETGLOBAL fallback C functions. * https://github.com/Roblox/luau/pull/929 * The syntax to the `luau-reduce` commandline tool has changed. It now accepts a script, a command to execute, and an error to search for. It no longer automatically passes the script to the command which makes it a lot more flexible. Also be warned that it edits the script it is passed **in place**. Do not point it at something that is not in source control! New solver * Switch to a greedier but more fallible algorithm for simplifying union and intersection types that are created as part of refinement calculation. This has much better and more predictable performance. * Fix a constraint cycle in recursive function calls. * Much improved inference of binary addition. Functions like `function add(x, y) return x + y end` can now be inferred without annotations. We also accurately typecheck calls to functions like this. * Many small bugfixes surrounding things like table indexers * Add support for indexers on class types. This was previously added to the old solver; we now add it to the new one for feature parity. JIT * https://github.com/Roblox/luau/pull/931 * Fuse key.value and key.tt loads for CEHCK_SLOT_MATCH in A64 * Implement remaining aliases of BFM for A64 * Implement new callinfo flag for A64 * Add instruction simplification for int->num->int conversion chains * Don't even load execdata for X64 calls * Treat opcode fallbacks the same as manually written fallbacks --------- Co-authored-by: Arseny Kapoulkine <arseny.kapoulkine@gmail.com> Co-authored-by: Vyacheslav Egorov <vegorov@roblox.com>
2023-05-19 20:37:30 +01:00
{
return is<BlockedType>(ty) || is<PendingExpansionType>(ty) || is<TypeFamilyInstanceType>(ty) || (solver && solver->hasUnresolvedConstraints(ty));
Sync to upstream/release/577 (#934) Lots of things going on this week: * Fix a crash that could occur in the presence of a cyclic union. We shouldn't be creating cyclic unions, but we shouldn't be crashing when they arise either. * Minor cleanup of `luau_precall` * Internal change to make L->top handling slightly more uniform * Optimize SETGLOBAL & GETGLOBAL fallback C functions. * https://github.com/Roblox/luau/pull/929 * The syntax to the `luau-reduce` commandline tool has changed. It now accepts a script, a command to execute, and an error to search for. It no longer automatically passes the script to the command which makes it a lot more flexible. Also be warned that it edits the script it is passed **in place**. Do not point it at something that is not in source control! New solver * Switch to a greedier but more fallible algorithm for simplifying union and intersection types that are created as part of refinement calculation. This has much better and more predictable performance. * Fix a constraint cycle in recursive function calls. * Much improved inference of binary addition. Functions like `function add(x, y) return x + y end` can now be inferred without annotations. We also accurately typecheck calls to functions like this. * Many small bugfixes surrounding things like table indexers * Add support for indexers on class types. This was previously added to the old solver; we now add it to the new one for feature parity. JIT * https://github.com/Roblox/luau/pull/931 * Fuse key.value and key.tt loads for CEHCK_SLOT_MATCH in A64 * Implement remaining aliases of BFM for A64 * Implement new callinfo flag for A64 * Add instruction simplification for int->num->int conversion chains * Don't even load execdata for X64 calls * Treat opcode fallbacks the same as manually written fallbacks --------- Co-authored-by: Arseny Kapoulkine <arseny.kapoulkine@gmail.com> Co-authored-by: Vyacheslav Egorov <vegorov@roblox.com>
2023-05-19 20:37:30 +01:00
}
TypeFamilyReductionResult<TypeId> notFamilyFn(const std::vector<TypeId>& typeParams, const std::vector<TypePackId>& packParams, NotNull<TypeFamilyContext> ctx)
{
if (typeParams.size() != 1 || !packParams.empty())
{
ctx->ice->ice("not type family: encountered a type family instance without the required argument structure");
LUAU_ASSERT(false);
}
TypeId ty = follow(typeParams.at(0));
if (isPending(ty, ctx->solver))
return {std::nullopt, false, {ty}, {}};
// `not` operates on anything and returns a `boolean` always.
return {ctx->builtins->booleanType, false, {}, {}};
}
Sync to upstream/release/600 (#1076) ### 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>
2023-10-21 02:10:30 +01:00
TypeFamilyReductionResult<TypeId> lenFamilyFn(const std::vector<TypeId>& typeParams, const std::vector<TypePackId>& packParams, NotNull<TypeFamilyContext> ctx)
{
if (typeParams.size() != 1 || !packParams.empty())
{
ctx->ice->ice("len type family: encountered a type family instance without the required argument structure");
LUAU_ASSERT(false);
}
TypeId operandTy = follow(typeParams.at(0));
Sync to upstream/release/601 (#1084) ## What's changed - `bit32.byteswap` added ([RFC](https://github.com/luau-lang/rfcs/blob/4f543ec23b6a1b53396e0803dd253c83041bae62/docs/function-bit32-byteswap.md)) - Buffer library implementation ([RFC](https://github.com/luau-lang/rfcs/blob/4f543ec23b6a1b53396e0803dd253c83041bae62/docs/type-byte-buffer.md)) - Fixed a missing `stdint.h` include - Fixed parser limiter for recursive type annotations being kind of weird (fixes #645) ### Native Codegen - Fixed a pair of issues when lowering `bit32.extract` - Fixed a narrow edge case that could result in an infinite loop without an interruption - Fixed a negative array out-of-bounds access issue - Temporarily reverted linear block predecessor value propagation ### New type solver - We now type check assignments to annotated variables - Fixed some test cases under local type inference - Moved `isPending` checks for type families to improve performance - Optimized our process for testing if a free type is sufficiently solved - Removed "none ptr" from lea instruction disassembly logging ### Build system & tooling - CMake configuration now validates dependencies to maintain separation between components - Improvements to the fuzzer coverage - Deduplicator for fuzzed callstacks --------- 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: Vighnesh Vijay <vvijay@roblox.com>
2023-10-27 22:18:41 +01:00
// check to see if the operand type is resolved enough, and wait to reduce if not
if (isPending(operandTy, ctx->solver))
return {std::nullopt, false, {operandTy}, {}};
Sync to upstream/release/600 (#1076) ### 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>
2023-10-21 02:10:30 +01:00
const NormalizedType* normTy = ctx->normalizer->normalize(operandTy);
// if the type failed to normalize, we can't reduce, but know nothing about inhabitance.
if (!normTy)
return {std::nullopt, false, {}, {}};
// if the operand type is error suppressing, we can immediately reduce to `number`.
if (normTy->shouldSuppressErrors())
return {ctx->builtins->numberType, false, {}, {}};
// if we have a `never`, we can never observe that the operator didn't work.
if (is<NeverType>(operandTy))
return {ctx->builtins->neverType, false, {}, {}};
// if we're checking the length of a string, that works!
if (normTy->isSubtypeOfString())
return {ctx->builtins->numberType, false, {}, {}};
// we use the normalized operand here in case there was an intersection or union.
TypeId normalizedOperand = ctx->normalizer->typeFromNormal(*normTy);
if (normTy->hasTopTable() || get<TableType>(normalizedOperand))
return {ctx->builtins->numberType, false, {}, {}};
// findMetatableEntry demands the ability to emit errors, so we must give it
// the necessary state to do that, even if we intend to just eat the errors.
ErrorVec dummy;
std::optional<TypeId> mmType = findMetatableEntry(ctx->builtins, dummy, operandTy, "__len", Location{});
if (!mmType)
return {std::nullopt, true, {}, {}};
mmType = follow(*mmType);
if (isPending(*mmType, ctx->solver))
return {std::nullopt, false, {*mmType}, {}};
const FunctionType* mmFtv = get<FunctionType>(*mmType);
if (!mmFtv)
return {std::nullopt, true, {}, {}};
std::optional<TypeId> instantiatedMmType = instantiate(ctx->builtins, ctx->arena, ctx->limits, ctx->scope, *mmType);
if (!instantiatedMmType)
return {std::nullopt, true, {}, {}};
const FunctionType* instantiatedMmFtv = get<FunctionType>(*instantiatedMmType);
if (!instantiatedMmFtv)
return {ctx->builtins->errorRecoveryType(), false, {}, {}};
TypePackId inferredArgPack = ctx->arena->addTypePack({operandTy});
Unifier2 u2{ctx->arena, ctx->builtins, ctx->scope, ctx->ice};
if (!u2.unify(inferredArgPack, instantiatedMmFtv->argTypes))
return {std::nullopt, true, {}, {}}; // occurs check failed
Subtyping subtyping{ctx->builtins, ctx->arena, ctx->normalizer, ctx->ice, ctx->scope};
if (!subtyping.isSubtype(inferredArgPack, instantiatedMmFtv->argTypes).isSubtype) // TODO: is this the right variance?
return {std::nullopt, true, {}, {}};
// `len` must return a `number`.
return {ctx->builtins->numberType, false, {}, {}};
}
TypeFamilyReductionResult<TypeId> unmFamilyFn(
const std::vector<TypeId>& typeParams, const std::vector<TypePackId>& packParams, NotNull<TypeFamilyContext> ctx)
{
if (typeParams.size() != 1 || !packParams.empty())
{
ctx->ice->ice("unm type family: encountered a type family instance without the required argument structure");
LUAU_ASSERT(false);
}
TypeId operandTy = follow(typeParams.at(0));
Sync to upstream/release/601 (#1084) ## What's changed - `bit32.byteswap` added ([RFC](https://github.com/luau-lang/rfcs/blob/4f543ec23b6a1b53396e0803dd253c83041bae62/docs/function-bit32-byteswap.md)) - Buffer library implementation ([RFC](https://github.com/luau-lang/rfcs/blob/4f543ec23b6a1b53396e0803dd253c83041bae62/docs/type-byte-buffer.md)) - Fixed a missing `stdint.h` include - Fixed parser limiter for recursive type annotations being kind of weird (fixes #645) ### Native Codegen - Fixed a pair of issues when lowering `bit32.extract` - Fixed a narrow edge case that could result in an infinite loop without an interruption - Fixed a negative array out-of-bounds access issue - Temporarily reverted linear block predecessor value propagation ### New type solver - We now type check assignments to annotated variables - Fixed some test cases under local type inference - Moved `isPending` checks for type families to improve performance - Optimized our process for testing if a free type is sufficiently solved - Removed "none ptr" from lea instruction disassembly logging ### Build system & tooling - CMake configuration now validates dependencies to maintain separation between components - Improvements to the fuzzer coverage - Deduplicator for fuzzed callstacks --------- 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: Vighnesh Vijay <vvijay@roblox.com>
2023-10-27 22:18:41 +01:00
// check to see if the operand type is resolved enough, and wait to reduce if not
if (isPending(operandTy, ctx->solver))
return {std::nullopt, false, {operandTy}, {}};
Sync to upstream/release/600 (#1076) ### 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>
2023-10-21 02:10:30 +01:00
const NormalizedType* normTy = ctx->normalizer->normalize(operandTy);
// if the operand failed to normalize, we can't reduce, but know nothing about inhabitance.
if (!normTy)
return {std::nullopt, false, {}, {}};
// if the operand is error suppressing, we can just go ahead and reduce.
if (normTy->shouldSuppressErrors())
return {operandTy, false, {}, {}};
// if we have a `never`, we can never observe that the operation didn't work.
if (is<NeverType>(operandTy))
return {ctx->builtins->neverType, false, {}, {}};
// If the type is exactly `number`, we can reduce now.
if (normTy->isExactlyNumber())
return {ctx->builtins->numberType, false, {}, {}};
// findMetatableEntry demands the ability to emit errors, so we must give it
// the necessary state to do that, even if we intend to just eat the errors.
ErrorVec dummy;
std::optional<TypeId> mmType = findMetatableEntry(ctx->builtins, dummy, operandTy, "__unm", Location{});
if (!mmType)
return {std::nullopt, true, {}, {}};
mmType = follow(*mmType);
if (isPending(*mmType, ctx->solver))
return {std::nullopt, false, {*mmType}, {}};
const FunctionType* mmFtv = get<FunctionType>(*mmType);
if (!mmFtv)
return {std::nullopt, true, {}, {}};
std::optional<TypeId> instantiatedMmType = instantiate(ctx->builtins, ctx->arena, ctx->limits, ctx->scope, *mmType);
if (!instantiatedMmType)
return {std::nullopt, true, {}, {}};
const FunctionType* instantiatedMmFtv = get<FunctionType>(*instantiatedMmType);
if (!instantiatedMmFtv)
return {ctx->builtins->errorRecoveryType(), false, {}, {}};
TypePackId inferredArgPack = ctx->arena->addTypePack({operandTy});
Unifier2 u2{ctx->arena, ctx->builtins, ctx->scope, ctx->ice};
if (!u2.unify(inferredArgPack, instantiatedMmFtv->argTypes))
return {std::nullopt, true, {}, {}}; // occurs check failed
Subtyping subtyping{ctx->builtins, ctx->arena, ctx->normalizer, ctx->ice, ctx->scope};
if (!subtyping.isSubtype(inferredArgPack, instantiatedMmFtv->argTypes).isSubtype) // TODO: is this the right variance?
return {std::nullopt, true, {}, {}};
if (std::optional<TypeId> ret = first(instantiatedMmFtv->retTypes))
return {*ret, false, {}, {}};
else
return {std::nullopt, true, {}, {}};
}
TypeFamilyReductionResult<TypeId> numericBinopFamilyFn(
const std::vector<TypeId>& typeParams, const std::vector<TypePackId>& packParams, NotNull<TypeFamilyContext> ctx, const std::string metamethod)
Sync to upstream/release/577 (#934) Lots of things going on this week: * Fix a crash that could occur in the presence of a cyclic union. We shouldn't be creating cyclic unions, but we shouldn't be crashing when they arise either. * Minor cleanup of `luau_precall` * Internal change to make L->top handling slightly more uniform * Optimize SETGLOBAL & GETGLOBAL fallback C functions. * https://github.com/Roblox/luau/pull/929 * The syntax to the `luau-reduce` commandline tool has changed. It now accepts a script, a command to execute, and an error to search for. It no longer automatically passes the script to the command which makes it a lot more flexible. Also be warned that it edits the script it is passed **in place**. Do not point it at something that is not in source control! New solver * Switch to a greedier but more fallible algorithm for simplifying union and intersection types that are created as part of refinement calculation. This has much better and more predictable performance. * Fix a constraint cycle in recursive function calls. * Much improved inference of binary addition. Functions like `function add(x, y) return x + y end` can now be inferred without annotations. We also accurately typecheck calls to functions like this. * Many small bugfixes surrounding things like table indexers * Add support for indexers on class types. This was previously added to the old solver; we now add it to the new one for feature parity. JIT * https://github.com/Roblox/luau/pull/931 * Fuse key.value and key.tt loads for CEHCK_SLOT_MATCH in A64 * Implement remaining aliases of BFM for A64 * Implement new callinfo flag for A64 * Add instruction simplification for int->num->int conversion chains * Don't even load execdata for X64 calls * Treat opcode fallbacks the same as manually written fallbacks --------- Co-authored-by: Arseny Kapoulkine <arseny.kapoulkine@gmail.com> Co-authored-by: Vyacheslav Egorov <vegorov@roblox.com>
2023-05-19 20:37:30 +01:00
{
if (typeParams.size() != 2 || !packParams.empty())
{
ctx->ice->ice("encountered a type family instance without the required argument structure");
Sync to upstream/release/577 (#934) Lots of things going on this week: * Fix a crash that could occur in the presence of a cyclic union. We shouldn't be creating cyclic unions, but we shouldn't be crashing when they arise either. * Minor cleanup of `luau_precall` * Internal change to make L->top handling slightly more uniform * Optimize SETGLOBAL & GETGLOBAL fallback C functions. * https://github.com/Roblox/luau/pull/929 * The syntax to the `luau-reduce` commandline tool has changed. It now accepts a script, a command to execute, and an error to search for. It no longer automatically passes the script to the command which makes it a lot more flexible. Also be warned that it edits the script it is passed **in place**. Do not point it at something that is not in source control! New solver * Switch to a greedier but more fallible algorithm for simplifying union and intersection types that are created as part of refinement calculation. This has much better and more predictable performance. * Fix a constraint cycle in recursive function calls. * Much improved inference of binary addition. Functions like `function add(x, y) return x + y end` can now be inferred without annotations. We also accurately typecheck calls to functions like this. * Many small bugfixes surrounding things like table indexers * Add support for indexers on class types. This was previously added to the old solver; we now add it to the new one for feature parity. JIT * https://github.com/Roblox/luau/pull/931 * Fuse key.value and key.tt loads for CEHCK_SLOT_MATCH in A64 * Implement remaining aliases of BFM for A64 * Implement new callinfo flag for A64 * Add instruction simplification for int->num->int conversion chains * Don't even load execdata for X64 calls * Treat opcode fallbacks the same as manually written fallbacks --------- Co-authored-by: Arseny Kapoulkine <arseny.kapoulkine@gmail.com> Co-authored-by: Vyacheslav Egorov <vegorov@roblox.com>
2023-05-19 20:37:30 +01:00
LUAU_ASSERT(false);
}
TypeId lhsTy = follow(typeParams.at(0));
TypeId rhsTy = follow(typeParams.at(1));
Sync to upstream/release/601 (#1084) ## What's changed - `bit32.byteswap` added ([RFC](https://github.com/luau-lang/rfcs/blob/4f543ec23b6a1b53396e0803dd253c83041bae62/docs/function-bit32-byteswap.md)) - Buffer library implementation ([RFC](https://github.com/luau-lang/rfcs/blob/4f543ec23b6a1b53396e0803dd253c83041bae62/docs/type-byte-buffer.md)) - Fixed a missing `stdint.h` include - Fixed parser limiter for recursive type annotations being kind of weird (fixes #645) ### Native Codegen - Fixed a pair of issues when lowering `bit32.extract` - Fixed a narrow edge case that could result in an infinite loop without an interruption - Fixed a negative array out-of-bounds access issue - Temporarily reverted linear block predecessor value propagation ### New type solver - We now type check assignments to annotated variables - Fixed some test cases under local type inference - Moved `isPending` checks for type families to improve performance - Optimized our process for testing if a free type is sufficiently solved - Removed "none ptr" from lea instruction disassembly logging ### Build system & tooling - CMake configuration now validates dependencies to maintain separation between components - Improvements to the fuzzer coverage - Deduplicator for fuzzed callstacks --------- 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: Vighnesh Vijay <vvijay@roblox.com>
2023-10-27 22:18:41 +01:00
// check to see if both operand types are resolved enough, and wait to reduce if not
if (isPending(lhsTy, ctx->solver))
return {std::nullopt, false, {lhsTy}, {}};
else if (isPending(rhsTy, ctx->solver))
return {std::nullopt, false, {rhsTy}, {}};
const NormalizedType* normLhsTy = ctx->normalizer->normalize(lhsTy);
const NormalizedType* normRhsTy = ctx->normalizer->normalize(rhsTy);
// if either failed to normalize, we can't reduce, but know nothing about inhabitance.
if (!normLhsTy || !normRhsTy)
return {std::nullopt, false, {}, {}};
// if one of the types is error suppressing, we can reduce to `any` since we should suppress errors in the result of the usage.
if (normLhsTy->shouldSuppressErrors() || normRhsTy->shouldSuppressErrors())
return {ctx->builtins->anyType, false, {}, {}};
// if we have a `never`, we can never observe that the numeric operator didn't work.
if (is<NeverType>(lhsTy) || is<NeverType>(rhsTy))
return {ctx->builtins->neverType, false, {}, {}};
// if we're adding two `number` types, the result is `number`.
if (normLhsTy->isExactlyNumber() && normRhsTy->isExactlyNumber())
return {ctx->builtins->numberType, false, {}, {}};
Sync to upstream/release/577 (#934) Lots of things going on this week: * Fix a crash that could occur in the presence of a cyclic union. We shouldn't be creating cyclic unions, but we shouldn't be crashing when they arise either. * Minor cleanup of `luau_precall` * Internal change to make L->top handling slightly more uniform * Optimize SETGLOBAL & GETGLOBAL fallback C functions. * https://github.com/Roblox/luau/pull/929 * The syntax to the `luau-reduce` commandline tool has changed. It now accepts a script, a command to execute, and an error to search for. It no longer automatically passes the script to the command which makes it a lot more flexible. Also be warned that it edits the script it is passed **in place**. Do not point it at something that is not in source control! New solver * Switch to a greedier but more fallible algorithm for simplifying union and intersection types that are created as part of refinement calculation. This has much better and more predictable performance. * Fix a constraint cycle in recursive function calls. * Much improved inference of binary addition. Functions like `function add(x, y) return x + y end` can now be inferred without annotations. We also accurately typecheck calls to functions like this. * Many small bugfixes surrounding things like table indexers * Add support for indexers on class types. This was previously added to the old solver; we now add it to the new one for feature parity. JIT * https://github.com/Roblox/luau/pull/931 * Fuse key.value and key.tt loads for CEHCK_SLOT_MATCH in A64 * Implement remaining aliases of BFM for A64 * Implement new callinfo flag for A64 * Add instruction simplification for int->num->int conversion chains * Don't even load execdata for X64 calls * Treat opcode fallbacks the same as manually written fallbacks --------- Co-authored-by: Arseny Kapoulkine <arseny.kapoulkine@gmail.com> Co-authored-by: Vyacheslav Egorov <vegorov@roblox.com>
2023-05-19 20:37:30 +01:00
// findMetatableEntry demands the ability to emit errors, so we must give it
// the necessary state to do that, even if we intend to just eat the errors.
ErrorVec dummy;
std::optional<TypeId> mmType = findMetatableEntry(ctx->builtins, dummy, lhsTy, metamethod, Location{});
Sync to upstream/release/577 (#934) Lots of things going on this week: * Fix a crash that could occur in the presence of a cyclic union. We shouldn't be creating cyclic unions, but we shouldn't be crashing when they arise either. * Minor cleanup of `luau_precall` * Internal change to make L->top handling slightly more uniform * Optimize SETGLOBAL & GETGLOBAL fallback C functions. * https://github.com/Roblox/luau/pull/929 * The syntax to the `luau-reduce` commandline tool has changed. It now accepts a script, a command to execute, and an error to search for. It no longer automatically passes the script to the command which makes it a lot more flexible. Also be warned that it edits the script it is passed **in place**. Do not point it at something that is not in source control! New solver * Switch to a greedier but more fallible algorithm for simplifying union and intersection types that are created as part of refinement calculation. This has much better and more predictable performance. * Fix a constraint cycle in recursive function calls. * Much improved inference of binary addition. Functions like `function add(x, y) return x + y end` can now be inferred without annotations. We also accurately typecheck calls to functions like this. * Many small bugfixes surrounding things like table indexers * Add support for indexers on class types. This was previously added to the old solver; we now add it to the new one for feature parity. JIT * https://github.com/Roblox/luau/pull/931 * Fuse key.value and key.tt loads for CEHCK_SLOT_MATCH in A64 * Implement remaining aliases of BFM for A64 * Implement new callinfo flag for A64 * Add instruction simplification for int->num->int conversion chains * Don't even load execdata for X64 calls * Treat opcode fallbacks the same as manually written fallbacks --------- Co-authored-by: Arseny Kapoulkine <arseny.kapoulkine@gmail.com> Co-authored-by: Vyacheslav Egorov <vegorov@roblox.com>
2023-05-19 20:37:30 +01:00
bool reversed = false;
if (!mmType)
Sync to upstream/release/577 (#934) Lots of things going on this week: * Fix a crash that could occur in the presence of a cyclic union. We shouldn't be creating cyclic unions, but we shouldn't be crashing when they arise either. * Minor cleanup of `luau_precall` * Internal change to make L->top handling slightly more uniform * Optimize SETGLOBAL & GETGLOBAL fallback C functions. * https://github.com/Roblox/luau/pull/929 * The syntax to the `luau-reduce` commandline tool has changed. It now accepts a script, a command to execute, and an error to search for. It no longer automatically passes the script to the command which makes it a lot more flexible. Also be warned that it edits the script it is passed **in place**. Do not point it at something that is not in source control! New solver * Switch to a greedier but more fallible algorithm for simplifying union and intersection types that are created as part of refinement calculation. This has much better and more predictable performance. * Fix a constraint cycle in recursive function calls. * Much improved inference of binary addition. Functions like `function add(x, y) return x + y end` can now be inferred without annotations. We also accurately typecheck calls to functions like this. * Many small bugfixes surrounding things like table indexers * Add support for indexers on class types. This was previously added to the old solver; we now add it to the new one for feature parity. JIT * https://github.com/Roblox/luau/pull/931 * Fuse key.value and key.tt loads for CEHCK_SLOT_MATCH in A64 * Implement remaining aliases of BFM for A64 * Implement new callinfo flag for A64 * Add instruction simplification for int->num->int conversion chains * Don't even load execdata for X64 calls * Treat opcode fallbacks the same as manually written fallbacks --------- Co-authored-by: Arseny Kapoulkine <arseny.kapoulkine@gmail.com> Co-authored-by: Vyacheslav Egorov <vegorov@roblox.com>
2023-05-19 20:37:30 +01:00
{
mmType = findMetatableEntry(ctx->builtins, dummy, rhsTy, metamethod, Location{});
Sync to upstream/release/577 (#934) Lots of things going on this week: * Fix a crash that could occur in the presence of a cyclic union. We shouldn't be creating cyclic unions, but we shouldn't be crashing when they arise either. * Minor cleanup of `luau_precall` * Internal change to make L->top handling slightly more uniform * Optimize SETGLOBAL & GETGLOBAL fallback C functions. * https://github.com/Roblox/luau/pull/929 * The syntax to the `luau-reduce` commandline tool has changed. It now accepts a script, a command to execute, and an error to search for. It no longer automatically passes the script to the command which makes it a lot more flexible. Also be warned that it edits the script it is passed **in place**. Do not point it at something that is not in source control! New solver * Switch to a greedier but more fallible algorithm for simplifying union and intersection types that are created as part of refinement calculation. This has much better and more predictable performance. * Fix a constraint cycle in recursive function calls. * Much improved inference of binary addition. Functions like `function add(x, y) return x + y end` can now be inferred without annotations. We also accurately typecheck calls to functions like this. * Many small bugfixes surrounding things like table indexers * Add support for indexers on class types. This was previously added to the old solver; we now add it to the new one for feature parity. JIT * https://github.com/Roblox/luau/pull/931 * Fuse key.value and key.tt loads for CEHCK_SLOT_MATCH in A64 * Implement remaining aliases of BFM for A64 * Implement new callinfo flag for A64 * Add instruction simplification for int->num->int conversion chains * Don't even load execdata for X64 calls * Treat opcode fallbacks the same as manually written fallbacks --------- Co-authored-by: Arseny Kapoulkine <arseny.kapoulkine@gmail.com> Co-authored-by: Vyacheslav Egorov <vegorov@roblox.com>
2023-05-19 20:37:30 +01:00
reversed = true;
}
if (!mmType)
Sync to upstream/release/577 (#934) Lots of things going on this week: * Fix a crash that could occur in the presence of a cyclic union. We shouldn't be creating cyclic unions, but we shouldn't be crashing when they arise either. * Minor cleanup of `luau_precall` * Internal change to make L->top handling slightly more uniform * Optimize SETGLOBAL & GETGLOBAL fallback C functions. * https://github.com/Roblox/luau/pull/929 * The syntax to the `luau-reduce` commandline tool has changed. It now accepts a script, a command to execute, and an error to search for. It no longer automatically passes the script to the command which makes it a lot more flexible. Also be warned that it edits the script it is passed **in place**. Do not point it at something that is not in source control! New solver * Switch to a greedier but more fallible algorithm for simplifying union and intersection types that are created as part of refinement calculation. This has much better and more predictable performance. * Fix a constraint cycle in recursive function calls. * Much improved inference of binary addition. Functions like `function add(x, y) return x + y end` can now be inferred without annotations. We also accurately typecheck calls to functions like this. * Many small bugfixes surrounding things like table indexers * Add support for indexers on class types. This was previously added to the old solver; we now add it to the new one for feature parity. JIT * https://github.com/Roblox/luau/pull/931 * Fuse key.value and key.tt loads for CEHCK_SLOT_MATCH in A64 * Implement remaining aliases of BFM for A64 * Implement new callinfo flag for A64 * Add instruction simplification for int->num->int conversion chains * Don't even load execdata for X64 calls * Treat opcode fallbacks the same as manually written fallbacks --------- Co-authored-by: Arseny Kapoulkine <arseny.kapoulkine@gmail.com> Co-authored-by: Vyacheslav Egorov <vegorov@roblox.com>
2023-05-19 20:37:30 +01:00
return {std::nullopt, true, {}, {}};
mmType = follow(*mmType);
if (isPending(*mmType, ctx->solver))
return {std::nullopt, false, {*mmType}, {}};
Sync to upstream/release/577 (#934) Lots of things going on this week: * Fix a crash that could occur in the presence of a cyclic union. We shouldn't be creating cyclic unions, but we shouldn't be crashing when they arise either. * Minor cleanup of `luau_precall` * Internal change to make L->top handling slightly more uniform * Optimize SETGLOBAL & GETGLOBAL fallback C functions. * https://github.com/Roblox/luau/pull/929 * The syntax to the `luau-reduce` commandline tool has changed. It now accepts a script, a command to execute, and an error to search for. It no longer automatically passes the script to the command which makes it a lot more flexible. Also be warned that it edits the script it is passed **in place**. Do not point it at something that is not in source control! New solver * Switch to a greedier but more fallible algorithm for simplifying union and intersection types that are created as part of refinement calculation. This has much better and more predictable performance. * Fix a constraint cycle in recursive function calls. * Much improved inference of binary addition. Functions like `function add(x, y) return x + y end` can now be inferred without annotations. We also accurately typecheck calls to functions like this. * Many small bugfixes surrounding things like table indexers * Add support for indexers on class types. This was previously added to the old solver; we now add it to the new one for feature parity. JIT * https://github.com/Roblox/luau/pull/931 * Fuse key.value and key.tt loads for CEHCK_SLOT_MATCH in A64 * Implement remaining aliases of BFM for A64 * Implement new callinfo flag for A64 * Add instruction simplification for int->num->int conversion chains * Don't even load execdata for X64 calls * Treat opcode fallbacks the same as manually written fallbacks --------- Co-authored-by: Arseny Kapoulkine <arseny.kapoulkine@gmail.com> Co-authored-by: Vyacheslav Egorov <vegorov@roblox.com>
2023-05-19 20:37:30 +01:00
const FunctionType* mmFtv = get<FunctionType>(*mmType);
Sync to upstream/release/577 (#934) Lots of things going on this week: * Fix a crash that could occur in the presence of a cyclic union. We shouldn't be creating cyclic unions, but we shouldn't be crashing when they arise either. * Minor cleanup of `luau_precall` * Internal change to make L->top handling slightly more uniform * Optimize SETGLOBAL & GETGLOBAL fallback C functions. * https://github.com/Roblox/luau/pull/929 * The syntax to the `luau-reduce` commandline tool has changed. It now accepts a script, a command to execute, and an error to search for. It no longer automatically passes the script to the command which makes it a lot more flexible. Also be warned that it edits the script it is passed **in place**. Do not point it at something that is not in source control! New solver * Switch to a greedier but more fallible algorithm for simplifying union and intersection types that are created as part of refinement calculation. This has much better and more predictable performance. * Fix a constraint cycle in recursive function calls. * Much improved inference of binary addition. Functions like `function add(x, y) return x + y end` can now be inferred without annotations. We also accurately typecheck calls to functions like this. * Many small bugfixes surrounding things like table indexers * Add support for indexers on class types. This was previously added to the old solver; we now add it to the new one for feature parity. JIT * https://github.com/Roblox/luau/pull/931 * Fuse key.value and key.tt loads for CEHCK_SLOT_MATCH in A64 * Implement remaining aliases of BFM for A64 * Implement new callinfo flag for A64 * Add instruction simplification for int->num->int conversion chains * Don't even load execdata for X64 calls * Treat opcode fallbacks the same as manually written fallbacks --------- Co-authored-by: Arseny Kapoulkine <arseny.kapoulkine@gmail.com> Co-authored-by: Vyacheslav Egorov <vegorov@roblox.com>
2023-05-19 20:37:30 +01:00
if (!mmFtv)
return {std::nullopt, true, {}, {}};
std::optional<TypeId> instantiatedMmType = instantiate(ctx->builtins, ctx->arena, ctx->limits, ctx->scope, *mmType);
if (!instantiatedMmType)
return {std::nullopt, true, {}, {}};
Sync to upstream/release/577 (#934) Lots of things going on this week: * Fix a crash that could occur in the presence of a cyclic union. We shouldn't be creating cyclic unions, but we shouldn't be crashing when they arise either. * Minor cleanup of `luau_precall` * Internal change to make L->top handling slightly more uniform * Optimize SETGLOBAL & GETGLOBAL fallback C functions. * https://github.com/Roblox/luau/pull/929 * The syntax to the `luau-reduce` commandline tool has changed. It now accepts a script, a command to execute, and an error to search for. It no longer automatically passes the script to the command which makes it a lot more flexible. Also be warned that it edits the script it is passed **in place**. Do not point it at something that is not in source control! New solver * Switch to a greedier but more fallible algorithm for simplifying union and intersection types that are created as part of refinement calculation. This has much better and more predictable performance. * Fix a constraint cycle in recursive function calls. * Much improved inference of binary addition. Functions like `function add(x, y) return x + y end` can now be inferred without annotations. We also accurately typecheck calls to functions like this. * Many small bugfixes surrounding things like table indexers * Add support for indexers on class types. This was previously added to the old solver; we now add it to the new one for feature parity. JIT * https://github.com/Roblox/luau/pull/931 * Fuse key.value and key.tt loads for CEHCK_SLOT_MATCH in A64 * Implement remaining aliases of BFM for A64 * Implement new callinfo flag for A64 * Add instruction simplification for int->num->int conversion chains * Don't even load execdata for X64 calls * Treat opcode fallbacks the same as manually written fallbacks --------- Co-authored-by: Arseny Kapoulkine <arseny.kapoulkine@gmail.com> Co-authored-by: Vyacheslav Egorov <vegorov@roblox.com>
2023-05-19 20:37:30 +01:00
const FunctionType* instantiatedMmFtv = get<FunctionType>(*instantiatedMmType);
if (!instantiatedMmFtv)
return {ctx->builtins->errorRecoveryType(), false, {}, {}};
Sync to upstream/release/577 (#934) Lots of things going on this week: * Fix a crash that could occur in the presence of a cyclic union. We shouldn't be creating cyclic unions, but we shouldn't be crashing when they arise either. * Minor cleanup of `luau_precall` * Internal change to make L->top handling slightly more uniform * Optimize SETGLOBAL & GETGLOBAL fallback C functions. * https://github.com/Roblox/luau/pull/929 * The syntax to the `luau-reduce` commandline tool has changed. It now accepts a script, a command to execute, and an error to search for. It no longer automatically passes the script to the command which makes it a lot more flexible. Also be warned that it edits the script it is passed **in place**. Do not point it at something that is not in source control! New solver * Switch to a greedier but more fallible algorithm for simplifying union and intersection types that are created as part of refinement calculation. This has much better and more predictable performance. * Fix a constraint cycle in recursive function calls. * Much improved inference of binary addition. Functions like `function add(x, y) return x + y end` can now be inferred without annotations. We also accurately typecheck calls to functions like this. * Many small bugfixes surrounding things like table indexers * Add support for indexers on class types. This was previously added to the old solver; we now add it to the new one for feature parity. JIT * https://github.com/Roblox/luau/pull/931 * Fuse key.value and key.tt loads for CEHCK_SLOT_MATCH in A64 * Implement remaining aliases of BFM for A64 * Implement new callinfo flag for A64 * Add instruction simplification for int->num->int conversion chains * Don't even load execdata for X64 calls * Treat opcode fallbacks the same as manually written fallbacks --------- Co-authored-by: Arseny Kapoulkine <arseny.kapoulkine@gmail.com> Co-authored-by: Vyacheslav Egorov <vegorov@roblox.com>
2023-05-19 20:37:30 +01:00
std::vector<TypeId> inferredArgs;
if (!reversed)
inferredArgs = {lhsTy, rhsTy};
else
inferredArgs = {rhsTy, lhsTy};
TypePackId inferredArgPack = ctx->arena->addTypePack(std::move(inferredArgs));
Unifier2 u2{ctx->arena, ctx->builtins, ctx->scope, ctx->ice};
if (!u2.unify(inferredArgPack, instantiatedMmFtv->argTypes))
return {std::nullopt, true, {}, {}}; // occurs check failed
if (std::optional<TypeId> ret = first(instantiatedMmFtv->retTypes))
return {*ret, false, {}, {}};
Sync to upstream/release/577 (#934) Lots of things going on this week: * Fix a crash that could occur in the presence of a cyclic union. We shouldn't be creating cyclic unions, but we shouldn't be crashing when they arise either. * Minor cleanup of `luau_precall` * Internal change to make L->top handling slightly more uniform * Optimize SETGLOBAL & GETGLOBAL fallback C functions. * https://github.com/Roblox/luau/pull/929 * The syntax to the `luau-reduce` commandline tool has changed. It now accepts a script, a command to execute, and an error to search for. It no longer automatically passes the script to the command which makes it a lot more flexible. Also be warned that it edits the script it is passed **in place**. Do not point it at something that is not in source control! New solver * Switch to a greedier but more fallible algorithm for simplifying union and intersection types that are created as part of refinement calculation. This has much better and more predictable performance. * Fix a constraint cycle in recursive function calls. * Much improved inference of binary addition. Functions like `function add(x, y) return x + y end` can now be inferred without annotations. We also accurately typecheck calls to functions like this. * Many small bugfixes surrounding things like table indexers * Add support for indexers on class types. This was previously added to the old solver; we now add it to the new one for feature parity. JIT * https://github.com/Roblox/luau/pull/931 * Fuse key.value and key.tt loads for CEHCK_SLOT_MATCH in A64 * Implement remaining aliases of BFM for A64 * Implement new callinfo flag for A64 * Add instruction simplification for int->num->int conversion chains * Don't even load execdata for X64 calls * Treat opcode fallbacks the same as manually written fallbacks --------- Co-authored-by: Arseny Kapoulkine <arseny.kapoulkine@gmail.com> Co-authored-by: Vyacheslav Egorov <vegorov@roblox.com>
2023-05-19 20:37:30 +01:00
else
return {std::nullopt, true, {}, {}};
}
TypeFamilyReductionResult<TypeId> addFamilyFn(const std::vector<TypeId>& typeParams, const std::vector<TypePackId>& packParams, NotNull<TypeFamilyContext> ctx)
{
if (typeParams.size() != 2 || !packParams.empty())
{
ctx->ice->ice("add type family: encountered a type family instance without the required argument structure");
LUAU_ASSERT(false);
}
return numericBinopFamilyFn(typeParams, packParams, ctx, "__add");
}
TypeFamilyReductionResult<TypeId> subFamilyFn(const std::vector<TypeId>& typeParams, const std::vector<TypePackId>& packParams, NotNull<TypeFamilyContext> ctx)
{
if (typeParams.size() != 2 || !packParams.empty())
{
ctx->ice->ice("sub type family: encountered a type family instance without the required argument structure");
LUAU_ASSERT(false);
}
return numericBinopFamilyFn(typeParams, packParams, ctx, "__sub");
}
TypeFamilyReductionResult<TypeId> mulFamilyFn(const std::vector<TypeId>& typeParams, const std::vector<TypePackId>& packParams, NotNull<TypeFamilyContext> ctx)
{
if (typeParams.size() != 2 || !packParams.empty())
{
ctx->ice->ice("mul type family: encountered a type family instance without the required argument structure");
LUAU_ASSERT(false);
}
return numericBinopFamilyFn(typeParams, packParams, ctx, "__mul");
}
TypeFamilyReductionResult<TypeId> divFamilyFn(const std::vector<TypeId>& typeParams, const std::vector<TypePackId>& packParams, NotNull<TypeFamilyContext> ctx)
{
if (typeParams.size() != 2 || !packParams.empty())
{
ctx->ice->ice("div type family: encountered a type family instance without the required argument structure");
LUAU_ASSERT(false);
}
return numericBinopFamilyFn(typeParams, packParams, ctx, "__div");
}
TypeFamilyReductionResult<TypeId> idivFamilyFn(const std::vector<TypeId>& typeParams, const std::vector<TypePackId>& packParams, NotNull<TypeFamilyContext> ctx)
{
if (typeParams.size() != 2 || !packParams.empty())
{
ctx->ice->ice("integer div type family: encountered a type family instance without the required argument structure");
LUAU_ASSERT(false);
}
return numericBinopFamilyFn(typeParams, packParams, ctx, "__idiv");
}
TypeFamilyReductionResult<TypeId> powFamilyFn(const std::vector<TypeId>& typeParams, const std::vector<TypePackId>& packParams, NotNull<TypeFamilyContext> ctx)
{
if (typeParams.size() != 2 || !packParams.empty())
{
ctx->ice->ice("pow type family: encountered a type family instance without the required argument structure");
LUAU_ASSERT(false);
}
return numericBinopFamilyFn(typeParams, packParams, ctx, "__pow");
}
TypeFamilyReductionResult<TypeId> modFamilyFn(const std::vector<TypeId>& typeParams, const std::vector<TypePackId>& packParams, NotNull<TypeFamilyContext> ctx)
{
if (typeParams.size() != 2 || !packParams.empty())
{
ctx->ice->ice("modulo type family: encountered a type family instance without the required argument structure");
LUAU_ASSERT(false);
}
return numericBinopFamilyFn(typeParams, packParams, ctx, "__mod");
}
TypeFamilyReductionResult<TypeId> concatFamilyFn(const std::vector<TypeId>& typeParams, const std::vector<TypePackId>& packParams, NotNull<TypeFamilyContext> ctx)
{
if (typeParams.size() != 2 || !packParams.empty())
{
ctx->ice->ice("concat type family: encountered a type family instance without the required argument structure");
LUAU_ASSERT(false);
}
TypeId lhsTy = follow(typeParams.at(0));
TypeId rhsTy = follow(typeParams.at(1));
Sync to upstream/release/601 (#1084) ## What's changed - `bit32.byteswap` added ([RFC](https://github.com/luau-lang/rfcs/blob/4f543ec23b6a1b53396e0803dd253c83041bae62/docs/function-bit32-byteswap.md)) - Buffer library implementation ([RFC](https://github.com/luau-lang/rfcs/blob/4f543ec23b6a1b53396e0803dd253c83041bae62/docs/type-byte-buffer.md)) - Fixed a missing `stdint.h` include - Fixed parser limiter for recursive type annotations being kind of weird (fixes #645) ### Native Codegen - Fixed a pair of issues when lowering `bit32.extract` - Fixed a narrow edge case that could result in an infinite loop without an interruption - Fixed a negative array out-of-bounds access issue - Temporarily reverted linear block predecessor value propagation ### New type solver - We now type check assignments to annotated variables - Fixed some test cases under local type inference - Moved `isPending` checks for type families to improve performance - Optimized our process for testing if a free type is sufficiently solved - Removed "none ptr" from lea instruction disassembly logging ### Build system & tooling - CMake configuration now validates dependencies to maintain separation between components - Improvements to the fuzzer coverage - Deduplicator for fuzzed callstacks --------- 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: Vighnesh Vijay <vvijay@roblox.com>
2023-10-27 22:18:41 +01:00
// check to see if both operand types are resolved enough, and wait to reduce if not
if (isPending(lhsTy, ctx->solver))
return {std::nullopt, false, {lhsTy}, {}};
else if (isPending(rhsTy, ctx->solver))
return {std::nullopt, false, {rhsTy}, {}};
const NormalizedType* normLhsTy = ctx->normalizer->normalize(lhsTy);
const NormalizedType* normRhsTy = ctx->normalizer->normalize(rhsTy);
// if either failed to normalize, we can't reduce, but know nothing about inhabitance.
if (!normLhsTy || !normRhsTy)
return {std::nullopt, false, {}, {}};
// if one of the types is error suppressing, we can reduce to `any` since we should suppress errors in the result of the usage.
if (normLhsTy->shouldSuppressErrors() || normRhsTy->shouldSuppressErrors())
return {ctx->builtins->anyType, false, {}, {}};
// if we have a `never`, we can never observe that the numeric operator didn't work.
if (is<NeverType>(lhsTy) || is<NeverType>(rhsTy))
return {ctx->builtins->neverType, false, {}, {}};
// if we're concatenating two elements that are either strings or numbers, the result is `string`.
if ((normLhsTy->isSubtypeOfString() || normLhsTy->isExactlyNumber()) && (normRhsTy->isSubtypeOfString() || normRhsTy->isExactlyNumber()))
return {ctx->builtins->stringType, false, {}, {}};
// findMetatableEntry demands the ability to emit errors, so we must give it
// the necessary state to do that, even if we intend to just eat the errors.
ErrorVec dummy;
std::optional<TypeId> mmType = findMetatableEntry(ctx->builtins, dummy, lhsTy, "__concat", Location{});
bool reversed = false;
if (!mmType)
{
mmType = findMetatableEntry(ctx->builtins, dummy, rhsTy, "__concat", Location{});
reversed = true;
}
if (!mmType)
return {std::nullopt, true, {}, {}};
mmType = follow(*mmType);
if (isPending(*mmType, ctx->solver))
return {std::nullopt, false, {*mmType}, {}};
const FunctionType* mmFtv = get<FunctionType>(*mmType);
if (!mmFtv)
return {std::nullopt, true, {}, {}};
std::optional<TypeId> instantiatedMmType = instantiate(ctx->builtins, ctx->arena, ctx->limits, ctx->scope, *mmType);
if (!instantiatedMmType)
return {std::nullopt, true, {}, {}};
const FunctionType* instantiatedMmFtv = get<FunctionType>(*instantiatedMmType);
if (!instantiatedMmFtv)
return {ctx->builtins->errorRecoveryType(), false, {}, {}};
std::vector<TypeId> inferredArgs;
if (!reversed)
inferredArgs = {lhsTy, rhsTy};
else
inferredArgs = {rhsTy, lhsTy};
TypePackId inferredArgPack = ctx->arena->addTypePack(std::move(inferredArgs));
Unifier2 u2{ctx->arena, ctx->builtins, ctx->scope, ctx->ice};
if (!u2.unify(inferredArgPack, instantiatedMmFtv->argTypes))
return {std::nullopt, true, {}, {}}; // occurs check failed
Subtyping subtyping{ctx->builtins, ctx->arena, ctx->normalizer, ctx->ice, ctx->scope};
if (!subtyping.isSubtype(inferredArgPack, instantiatedMmFtv->argTypes).isSubtype) // TODO: is this the right variance?
return {std::nullopt, true, {}, {}};
return {ctx->builtins->stringType, false, {}, {}};
}
TypeFamilyReductionResult<TypeId> andFamilyFn(const std::vector<TypeId>& typeParams, const std::vector<TypePackId>& packParams, NotNull<TypeFamilyContext> ctx)
{
if (typeParams.size() != 2 || !packParams.empty())
{
ctx->ice->ice("and type family: encountered a type family instance without the required argument structure");
LUAU_ASSERT(false);
}
TypeId lhsTy = follow(typeParams.at(0));
TypeId rhsTy = follow(typeParams.at(1));
Sync to upstream/release/601 (#1084) ## What's changed - `bit32.byteswap` added ([RFC](https://github.com/luau-lang/rfcs/blob/4f543ec23b6a1b53396e0803dd253c83041bae62/docs/function-bit32-byteswap.md)) - Buffer library implementation ([RFC](https://github.com/luau-lang/rfcs/blob/4f543ec23b6a1b53396e0803dd253c83041bae62/docs/type-byte-buffer.md)) - Fixed a missing `stdint.h` include - Fixed parser limiter for recursive type annotations being kind of weird (fixes #645) ### Native Codegen - Fixed a pair of issues when lowering `bit32.extract` - Fixed a narrow edge case that could result in an infinite loop without an interruption - Fixed a negative array out-of-bounds access issue - Temporarily reverted linear block predecessor value propagation ### New type solver - We now type check assignments to annotated variables - Fixed some test cases under local type inference - Moved `isPending` checks for type families to improve performance - Optimized our process for testing if a free type is sufficiently solved - Removed "none ptr" from lea instruction disassembly logging ### Build system & tooling - CMake configuration now validates dependencies to maintain separation between components - Improvements to the fuzzer coverage - Deduplicator for fuzzed callstacks --------- 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: Vighnesh Vijay <vvijay@roblox.com>
2023-10-27 22:18:41 +01:00
// check to see if both operand types are resolved enough, and wait to reduce if not
if (isPending(lhsTy, ctx->solver))
return {std::nullopt, false, {lhsTy}, {}};
else if (isPending(rhsTy, ctx->solver))
return {std::nullopt, false, {rhsTy}, {}};
// And evalutes to a boolean if the LHS is falsey, and the RHS type if LHS is truthy.
SimplifyResult filteredLhs = simplifyIntersection(ctx->builtins, ctx->arena, lhsTy, ctx->builtins->falsyType);
SimplifyResult overallResult = simplifyUnion(ctx->builtins, ctx->arena, rhsTy, filteredLhs.result);
Sync to upstream/release/603 (#1097) # What's changed? - Record the location of properties for table types (closes #802) - Implement stricter UTF-8 validations as per the RFC (https://github.com/luau-lang/rfcs/pull/1) - Implement `buffer` as a new type in both the old and new solvers. - Changed errors produced by some `buffer` builtins to be a bit more generic to avoid platform-dependent error messages. - Fixed a bug where `Unifier` would copy some persistent types, tripping some internal assertions. - Type checking rules on relational operators is now a little bit more lax. - Improve dead code elimination for some `if` statements with complex always-false conditions ## New type solver - Dataflow analysis now generates phi nodes on exit of branches. - Dataflow analysis avoids producing a new definition for locals or properties that are not owned by that loop. - If a function parameter has been constrained to `never`, report errors at all uses of that parameter within that function. - Switch to using the new `Luau::Set` to replace `std::unordered_set` to alleviate some poor allocation characteristics which was negatively affecting overall performance. - Subtyping can now report many failing reasons instead of just the first one that we happened to find during the test. - Subtyping now also report reasons for type pack mismatches. - When visiting `if` statements or expressions, the resulting context are the common terms in both branches. ## Native codegen - Implement support for `buffer` builtins to its IR for x64 and A64. - Optimized `table.insert` by not inserting a table barrier if it is fastcalled with a constant. ## 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: Arseny Kapoulkine <arseny@roblox.com> Co-authored-by: Aviral Goel <agoel@roblox.com> Co-authored-by: Lily Brown <lbrown@roblox.com> Co-authored-by: Vyacheslav Egorov <vegorov@roblox.com>
2023-11-10 21:10:07 +00:00
std::vector<TypeId> blockedTypes{};
for (auto ty : filteredLhs.blockedTypes)
blockedTypes.push_back(ty);
for (auto ty : overallResult.blockedTypes)
blockedTypes.push_back(ty);
return {overallResult.result, false, std::move(blockedTypes), {}};
}
TypeFamilyReductionResult<TypeId> orFamilyFn(const std::vector<TypeId>& typeParams, const std::vector<TypePackId>& packParams, NotNull<TypeFamilyContext> ctx)
{
if (typeParams.size() != 2 || !packParams.empty())
{
ctx->ice->ice("or type family: encountered a type family instance without the required argument structure");
LUAU_ASSERT(false);
}
TypeId lhsTy = follow(typeParams.at(0));
TypeId rhsTy = follow(typeParams.at(1));
Sync to upstream/release/601 (#1084) ## What's changed - `bit32.byteswap` added ([RFC](https://github.com/luau-lang/rfcs/blob/4f543ec23b6a1b53396e0803dd253c83041bae62/docs/function-bit32-byteswap.md)) - Buffer library implementation ([RFC](https://github.com/luau-lang/rfcs/blob/4f543ec23b6a1b53396e0803dd253c83041bae62/docs/type-byte-buffer.md)) - Fixed a missing `stdint.h` include - Fixed parser limiter for recursive type annotations being kind of weird (fixes #645) ### Native Codegen - Fixed a pair of issues when lowering `bit32.extract` - Fixed a narrow edge case that could result in an infinite loop without an interruption - Fixed a negative array out-of-bounds access issue - Temporarily reverted linear block predecessor value propagation ### New type solver - We now type check assignments to annotated variables - Fixed some test cases under local type inference - Moved `isPending` checks for type families to improve performance - Optimized our process for testing if a free type is sufficiently solved - Removed "none ptr" from lea instruction disassembly logging ### Build system & tooling - CMake configuration now validates dependencies to maintain separation between components - Improvements to the fuzzer coverage - Deduplicator for fuzzed callstacks --------- 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: Vighnesh Vijay <vvijay@roblox.com>
2023-10-27 22:18:41 +01:00
// check to see if both operand types are resolved enough, and wait to reduce if not
if (isPending(lhsTy, ctx->solver))
return {std::nullopt, false, {lhsTy}, {}};
else if (isPending(rhsTy, ctx->solver))
return {std::nullopt, false, {rhsTy}, {}};
// Or evalutes to the LHS type if the LHS is truthy, and the RHS type if LHS is falsy.
SimplifyResult filteredLhs = simplifyIntersection(ctx->builtins, ctx->arena, lhsTy, ctx->builtins->truthyType);
SimplifyResult overallResult = simplifyUnion(ctx->builtins, ctx->arena, rhsTy, filteredLhs.result);
Sync to upstream/release/603 (#1097) # What's changed? - Record the location of properties for table types (closes #802) - Implement stricter UTF-8 validations as per the RFC (https://github.com/luau-lang/rfcs/pull/1) - Implement `buffer` as a new type in both the old and new solvers. - Changed errors produced by some `buffer` builtins to be a bit more generic to avoid platform-dependent error messages. - Fixed a bug where `Unifier` would copy some persistent types, tripping some internal assertions. - Type checking rules on relational operators is now a little bit more lax. - Improve dead code elimination for some `if` statements with complex always-false conditions ## New type solver - Dataflow analysis now generates phi nodes on exit of branches. - Dataflow analysis avoids producing a new definition for locals or properties that are not owned by that loop. - If a function parameter has been constrained to `never`, report errors at all uses of that parameter within that function. - Switch to using the new `Luau::Set` to replace `std::unordered_set` to alleviate some poor allocation characteristics which was negatively affecting overall performance. - Subtyping can now report many failing reasons instead of just the first one that we happened to find during the test. - Subtyping now also report reasons for type pack mismatches. - When visiting `if` statements or expressions, the resulting context are the common terms in both branches. ## Native codegen - Implement support for `buffer` builtins to its IR for x64 and A64. - Optimized `table.insert` by not inserting a table barrier if it is fastcalled with a constant. ## 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: Arseny Kapoulkine <arseny@roblox.com> Co-authored-by: Aviral Goel <agoel@roblox.com> Co-authored-by: Lily Brown <lbrown@roblox.com> Co-authored-by: Vyacheslav Egorov <vegorov@roblox.com>
2023-11-10 21:10:07 +00:00
std::vector<TypeId> blockedTypes{};
for (auto ty : filteredLhs.blockedTypes)
blockedTypes.push_back(ty);
for (auto ty : overallResult.blockedTypes)
blockedTypes.push_back(ty);
return {overallResult.result, false, std::move(blockedTypes), {}};
}
static TypeFamilyReductionResult<TypeId> comparisonFamilyFn(
const std::vector<TypeId>& typeParams, const std::vector<TypePackId>& packParams, NotNull<TypeFamilyContext> ctx, const std::string metamethod)
{
if (typeParams.size() != 2 || !packParams.empty())
{
ctx->ice->ice("encountered a type family instance without the required argument structure");
LUAU_ASSERT(false);
}
TypeId lhsTy = follow(typeParams.at(0));
TypeId rhsTy = follow(typeParams.at(1));
Sync to upstream/release/601 (#1084) ## What's changed - `bit32.byteswap` added ([RFC](https://github.com/luau-lang/rfcs/blob/4f543ec23b6a1b53396e0803dd253c83041bae62/docs/function-bit32-byteswap.md)) - Buffer library implementation ([RFC](https://github.com/luau-lang/rfcs/blob/4f543ec23b6a1b53396e0803dd253c83041bae62/docs/type-byte-buffer.md)) - Fixed a missing `stdint.h` include - Fixed parser limiter for recursive type annotations being kind of weird (fixes #645) ### Native Codegen - Fixed a pair of issues when lowering `bit32.extract` - Fixed a narrow edge case that could result in an infinite loop without an interruption - Fixed a negative array out-of-bounds access issue - Temporarily reverted linear block predecessor value propagation ### New type solver - We now type check assignments to annotated variables - Fixed some test cases under local type inference - Moved `isPending` checks for type families to improve performance - Optimized our process for testing if a free type is sufficiently solved - Removed "none ptr" from lea instruction disassembly logging ### Build system & tooling - CMake configuration now validates dependencies to maintain separation between components - Improvements to the fuzzer coverage - Deduplicator for fuzzed callstacks --------- 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: Vighnesh Vijay <vvijay@roblox.com>
2023-10-27 22:18:41 +01:00
// check to see if both operand types are resolved enough, and wait to reduce if not
if (isPending(lhsTy, ctx->solver))
return {std::nullopt, false, {lhsTy}, {}};
else if (isPending(rhsTy, ctx->solver))
return {std::nullopt, false, {rhsTy}, {}};
const NormalizedType* normLhsTy = ctx->normalizer->normalize(lhsTy);
const NormalizedType* normRhsTy = ctx->normalizer->normalize(rhsTy);
// if either failed to normalize, we can't reduce, but know nothing about inhabitance.
if (!normLhsTy || !normRhsTy)
return {std::nullopt, false, {}, {}};
// if one of the types is error suppressing, we can just go ahead and reduce.
if (normLhsTy->shouldSuppressErrors() || normRhsTy->shouldSuppressErrors())
return {ctx->builtins->booleanType, false, {}, {}};
// if we have a `never`, we can never observe that the comparison didn't work.
if (is<NeverType>(lhsTy) || is<NeverType>(rhsTy))
return {ctx->builtins->booleanType, false, {}, {}};
// If both types are some strict subset of `string`, we can reduce now.
if (normLhsTy->isSubtypeOfString() && normRhsTy->isSubtypeOfString())
return {ctx->builtins->booleanType, false, {}, {}};
// If both types are exactly `number`, we can reduce now.
if (normLhsTy->isExactlyNumber() && normRhsTy->isExactlyNumber())
return {ctx->builtins->booleanType, false, {}, {}};
// findMetatableEntry demands the ability to emit errors, so we must give it
// the necessary state to do that, even if we intend to just eat the errors.
ErrorVec dummy;
std::optional<TypeId> mmType = findMetatableEntry(ctx->builtins, dummy, lhsTy, metamethod, Location{});
if (!mmType)
mmType = findMetatableEntry(ctx->builtins, dummy, rhsTy, metamethod, Location{});
if (!mmType)
return {std::nullopt, true, {}, {}};
mmType = follow(*mmType);
if (isPending(*mmType, ctx->solver))
return {std::nullopt, false, {*mmType}, {}};
const FunctionType* mmFtv = get<FunctionType>(*mmType);
if (!mmFtv)
return {std::nullopt, true, {}, {}};
std::optional<TypeId> instantiatedMmType = instantiate(ctx->builtins, ctx->arena, ctx->limits, ctx->scope, *mmType);
if (!instantiatedMmType)
return {std::nullopt, true, {}, {}};
const FunctionType* instantiatedMmFtv = get<FunctionType>(*instantiatedMmType);
if (!instantiatedMmFtv)
return {ctx->builtins->errorRecoveryType(), false, {}, {}};
TypePackId inferredArgPack = ctx->arena->addTypePack({lhsTy, rhsTy});
Unifier2 u2{ctx->arena, ctx->builtins, ctx->scope, ctx->ice};
if (!u2.unify(inferredArgPack, instantiatedMmFtv->argTypes))
return {std::nullopt, true, {}, {}}; // occurs check failed
Subtyping subtyping{ctx->builtins, ctx->arena, ctx->normalizer, ctx->ice, ctx->scope};
if (!subtyping.isSubtype(inferredArgPack, instantiatedMmFtv->argTypes).isSubtype) // TODO: is this the right variance?
return {std::nullopt, true, {}, {}};
return {ctx->builtins->booleanType, false, {}, {}};
}
TypeFamilyReductionResult<TypeId> ltFamilyFn(const std::vector<TypeId>& typeParams, const std::vector<TypePackId>& packParams, NotNull<TypeFamilyContext> ctx)
{
if (typeParams.size() != 2 || !packParams.empty())
{
ctx->ice->ice("lt type family: encountered a type family instance without the required argument structure");
LUAU_ASSERT(false);
}
return comparisonFamilyFn(typeParams, packParams, ctx, "__lt");
}
TypeFamilyReductionResult<TypeId> leFamilyFn(const std::vector<TypeId>& typeParams, const std::vector<TypePackId>& packParams, NotNull<TypeFamilyContext> ctx)
{
if (typeParams.size() != 2 || !packParams.empty())
{
ctx->ice->ice("le type family: encountered a type family instance without the required argument structure");
LUAU_ASSERT(false);
}
return comparisonFamilyFn(typeParams, packParams, ctx, "__le");
}
TypeFamilyReductionResult<TypeId> eqFamilyFn(const std::vector<TypeId>& typeParams, const std::vector<TypePackId>& packParams, NotNull<TypeFamilyContext> ctx)
{
if (typeParams.size() != 2 || !packParams.empty())
{
ctx->ice->ice("eq type family: encountered a type family instance without the required argument structure");
LUAU_ASSERT(false);
}
TypeId lhsTy = follow(typeParams.at(0));
TypeId rhsTy = follow(typeParams.at(1));
Sync to upstream/release/601 (#1084) ## What's changed - `bit32.byteswap` added ([RFC](https://github.com/luau-lang/rfcs/blob/4f543ec23b6a1b53396e0803dd253c83041bae62/docs/function-bit32-byteswap.md)) - Buffer library implementation ([RFC](https://github.com/luau-lang/rfcs/blob/4f543ec23b6a1b53396e0803dd253c83041bae62/docs/type-byte-buffer.md)) - Fixed a missing `stdint.h` include - Fixed parser limiter for recursive type annotations being kind of weird (fixes #645) ### Native Codegen - Fixed a pair of issues when lowering `bit32.extract` - Fixed a narrow edge case that could result in an infinite loop without an interruption - Fixed a negative array out-of-bounds access issue - Temporarily reverted linear block predecessor value propagation ### New type solver - We now type check assignments to annotated variables - Fixed some test cases under local type inference - Moved `isPending` checks for type families to improve performance - Optimized our process for testing if a free type is sufficiently solved - Removed "none ptr" from lea instruction disassembly logging ### Build system & tooling - CMake configuration now validates dependencies to maintain separation between components - Improvements to the fuzzer coverage - Deduplicator for fuzzed callstacks --------- 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: Vighnesh Vijay <vvijay@roblox.com>
2023-10-27 22:18:41 +01:00
// check to see if both operand types are resolved enough, and wait to reduce if not
if (isPending(lhsTy, ctx->solver))
return {std::nullopt, false, {lhsTy}, {}};
else if (isPending(rhsTy, ctx->solver))
return {std::nullopt, false, {rhsTy}, {}};
const NormalizedType* normLhsTy = ctx->normalizer->normalize(lhsTy);
const NormalizedType* normRhsTy = ctx->normalizer->normalize(rhsTy);
// if either failed to normalize, we can't reduce, but know nothing about inhabitance.
if (!normLhsTy || !normRhsTy)
return {std::nullopt, false, {}, {}};
// if one of the types is error suppressing, we can just go ahead and reduce.
if (normLhsTy->shouldSuppressErrors() || normRhsTy->shouldSuppressErrors())
return {ctx->builtins->booleanType, false, {}, {}};
// if we have a `never`, we can never observe that the comparison didn't work.
if (is<NeverType>(lhsTy) || is<NeverType>(rhsTy))
return {ctx->builtins->booleanType, false, {}, {}};
// findMetatableEntry demands the ability to emit errors, so we must give it
// the necessary state to do that, even if we intend to just eat the errors.
ErrorVec dummy;
std::optional<TypeId> mmType = findMetatableEntry(ctx->builtins, dummy, lhsTy, "__eq", Location{});
if (!mmType)
mmType = findMetatableEntry(ctx->builtins, dummy, rhsTy, "__eq", Location{});
// if neither type has a metatable entry for `__eq`, then we'll check for inhabitance of the intersection!
if (!mmType && ctx->normalizer->isIntersectionInhabited(lhsTy, rhsTy))
return {ctx->builtins->booleanType, false, {}, {}}; // if it's inhabited, everything is okay!
else if (!mmType)
return {std::nullopt, true, {}, {}}; // if it's not, then this family is irreducible!
mmType = follow(*mmType);
if (isPending(*mmType, ctx->solver))
return {std::nullopt, false, {*mmType}, {}};
const FunctionType* mmFtv = get<FunctionType>(*mmType);
if (!mmFtv)
return {std::nullopt, true, {}, {}};
std::optional<TypeId> instantiatedMmType = instantiate(ctx->builtins, ctx->arena, ctx->limits, ctx->scope, *mmType);
if (!instantiatedMmType)
return {std::nullopt, true, {}, {}};
const FunctionType* instantiatedMmFtv = get<FunctionType>(*instantiatedMmType);
if (!instantiatedMmFtv)
return {ctx->builtins->errorRecoveryType(), false, {}, {}};
TypePackId inferredArgPack = ctx->arena->addTypePack({lhsTy, rhsTy});
Unifier2 u2{ctx->arena, ctx->builtins, ctx->scope, ctx->ice};
if (!u2.unify(inferredArgPack, instantiatedMmFtv->argTypes))
return {std::nullopt, true, {}, {}}; // occurs check failed
Subtyping subtyping{ctx->builtins, ctx->arena, ctx->normalizer, ctx->ice, ctx->scope};
if (!subtyping.isSubtype(inferredArgPack, instantiatedMmFtv->argTypes).isSubtype) // TODO: is this the right variance?
return {std::nullopt, true, {}, {}};
return {ctx->builtins->booleanType, false, {}, {}};
}
Sync to upstream/release/577 (#934) Lots of things going on this week: * Fix a crash that could occur in the presence of a cyclic union. We shouldn't be creating cyclic unions, but we shouldn't be crashing when they arise either. * Minor cleanup of `luau_precall` * Internal change to make L->top handling slightly more uniform * Optimize SETGLOBAL & GETGLOBAL fallback C functions. * https://github.com/Roblox/luau/pull/929 * The syntax to the `luau-reduce` commandline tool has changed. It now accepts a script, a command to execute, and an error to search for. It no longer automatically passes the script to the command which makes it a lot more flexible. Also be warned that it edits the script it is passed **in place**. Do not point it at something that is not in source control! New solver * Switch to a greedier but more fallible algorithm for simplifying union and intersection types that are created as part of refinement calculation. This has much better and more predictable performance. * Fix a constraint cycle in recursive function calls. * Much improved inference of binary addition. Functions like `function add(x, y) return x + y end` can now be inferred without annotations. We also accurately typecheck calls to functions like this. * Many small bugfixes surrounding things like table indexers * Add support for indexers on class types. This was previously added to the old solver; we now add it to the new one for feature parity. JIT * https://github.com/Roblox/luau/pull/931 * Fuse key.value and key.tt loads for CEHCK_SLOT_MATCH in A64 * Implement remaining aliases of BFM for A64 * Implement new callinfo flag for A64 * Add instruction simplification for int->num->int conversion chains * Don't even load execdata for X64 calls * Treat opcode fallbacks the same as manually written fallbacks --------- Co-authored-by: Arseny Kapoulkine <arseny.kapoulkine@gmail.com> Co-authored-by: Vyacheslav Egorov <vegorov@roblox.com>
2023-05-19 20:37:30 +01:00
BuiltinTypeFamilies::BuiltinTypeFamilies()
: notFamily{"not", notFamilyFn}
Sync to upstream/release/600 (#1076) ### 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>
2023-10-21 02:10:30 +01:00
, lenFamily{"len", lenFamilyFn}
, unmFamily{"unm", unmFamilyFn}
, addFamily{"add", addFamilyFn}
, subFamily{"sub", subFamilyFn}
, mulFamily{"mul", mulFamilyFn}
, divFamily{"div", divFamilyFn}
, idivFamily{"idiv", idivFamilyFn}
, powFamily{"pow", powFamilyFn}
, modFamily{"mod", modFamilyFn}
, concatFamily{"concat", concatFamilyFn}
, andFamily{"and", andFamilyFn}
, orFamily{"or", orFamilyFn}
, ltFamily{"lt", ltFamilyFn}
, leFamily{"le", leFamilyFn}
, eqFamily{"eq", eqFamilyFn}
{
}
void BuiltinTypeFamilies::addToScope(NotNull<TypeArena> arena, NotNull<Scope> scope) const
Sync to upstream/release/577 (#934) Lots of things going on this week: * Fix a crash that could occur in the presence of a cyclic union. We shouldn't be creating cyclic unions, but we shouldn't be crashing when they arise either. * Minor cleanup of `luau_precall` * Internal change to make L->top handling slightly more uniform * Optimize SETGLOBAL & GETGLOBAL fallback C functions. * https://github.com/Roblox/luau/pull/929 * The syntax to the `luau-reduce` commandline tool has changed. It now accepts a script, a command to execute, and an error to search for. It no longer automatically passes the script to the command which makes it a lot more flexible. Also be warned that it edits the script it is passed **in place**. Do not point it at something that is not in source control! New solver * Switch to a greedier but more fallible algorithm for simplifying union and intersection types that are created as part of refinement calculation. This has much better and more predictable performance. * Fix a constraint cycle in recursive function calls. * Much improved inference of binary addition. Functions like `function add(x, y) return x + y end` can now be inferred without annotations. We also accurately typecheck calls to functions like this. * Many small bugfixes surrounding things like table indexers * Add support for indexers on class types. This was previously added to the old solver; we now add it to the new one for feature parity. JIT * https://github.com/Roblox/luau/pull/931 * Fuse key.value and key.tt loads for CEHCK_SLOT_MATCH in A64 * Implement remaining aliases of BFM for A64 * Implement new callinfo flag for A64 * Add instruction simplification for int->num->int conversion chains * Don't even load execdata for X64 calls * Treat opcode fallbacks the same as manually written fallbacks --------- Co-authored-by: Arseny Kapoulkine <arseny.kapoulkine@gmail.com> Co-authored-by: Vyacheslav Egorov <vegorov@roblox.com>
2023-05-19 20:37:30 +01:00
{
Sync to upstream/release/600 (#1076) ### 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>
2023-10-21 02:10:30 +01:00
// make a type function for a one-argument type family
auto mkUnaryTypeFamily = [&](const TypeFamily* family) {
TypeId t = arena->addType(GenericType{"T"});
GenericTypeDefinition genericT{t};
return TypeFun{{genericT}, arena->addType(TypeFamilyInstanceType{NotNull{family}, {t}, {}})};
};
// make a type function for a two-argument type family
auto mkBinaryTypeFamily = [&](const TypeFamily* family) {
TypeId t = arena->addType(GenericType{"T"});
TypeId u = arena->addType(GenericType{"U"});
GenericTypeDefinition genericT{t};
GenericTypeDefinition genericU{u};
return TypeFun{{genericT, genericU}, arena->addType(TypeFamilyInstanceType{NotNull{family}, {t, u}, {}})};
};
Sync to upstream/release/600 (#1076) ### 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>
2023-10-21 02:10:30 +01:00
scope->exportedTypeBindings[lenFamily.name] = mkUnaryTypeFamily(&lenFamily);
scope->exportedTypeBindings[unmFamily.name] = mkUnaryTypeFamily(&unmFamily);
scope->exportedTypeBindings[addFamily.name] = mkBinaryTypeFamily(&addFamily);
scope->exportedTypeBindings[subFamily.name] = mkBinaryTypeFamily(&subFamily);
scope->exportedTypeBindings[mulFamily.name] = mkBinaryTypeFamily(&mulFamily);
scope->exportedTypeBindings[divFamily.name] = mkBinaryTypeFamily(&divFamily);
scope->exportedTypeBindings[idivFamily.name] = mkBinaryTypeFamily(&idivFamily);
scope->exportedTypeBindings[powFamily.name] = mkBinaryTypeFamily(&powFamily);
scope->exportedTypeBindings[modFamily.name] = mkBinaryTypeFamily(&modFamily);
scope->exportedTypeBindings[concatFamily.name] = mkBinaryTypeFamily(&concatFamily);
scope->exportedTypeBindings[ltFamily.name] = mkBinaryTypeFamily(&ltFamily);
scope->exportedTypeBindings[leFamily.name] = mkBinaryTypeFamily(&leFamily);
scope->exportedTypeBindings[eqFamily.name] = mkBinaryTypeFamily(&eqFamily);
Sync to upstream/release/576 (#928) * `ClassType` can now have an indexer defined on it. This allows custom types to be used in `t[x]` expressions. * Fixed search for closest executable breakpoint line. Previously, breakpoints might have been skipped in `else` blocks at the end of a function * Fixed how unification is performed for two optional types `a? <: b?`, previously it might have unified either 'a' or 'b' with 'nil'. Note that this fix is not enabled by default yet (see the list in `ExperimentalFlags.h`) In the new type solver, a concept of 'Type Families' has been introduced. Type families can be thought of as type aliases with custom type inference/reduction logic included with them. For example, we can have an `Add<T, U>` type family that will resolve the type that is the result of adding two values together. This will help type inference to figure out what 'T' and 'U' might be when explicit type annotations are not provided. In this update we don't define any type families, but they will be added in the near future. It is also possible for Luau embedders to define their own type families in the global/environment scope. Other changes include: * Fixed scope used to find out which generic types should be included in the function generic type list * Fixed a crash after cyclic bound types were created during unification And in native code generation (jit): * Use of arm64 target on M1 now requires macOS 13 * Entry into native code has been optimized. This is especially important for coroutine call/pcall performance as they involve going through a C call frame * LOP_LOADK(X) translation into IR has been improved to enable type tag/constant propagation * arm64 can use integer immediate values to synthesize floating-point values * x64 assembler removes duplicate 64bit numbers from the data section to save space * Linux `perf` can now be used to profile native Luau code (when running with --codegen-perf CLI argument)
2023-05-12 18:50:47 +01:00
}
} // namespace Luau