luau/CodeGen/include/Luau/IrRegAllocX64.h

119 lines
3.2 KiB
C
Raw Normal View History

// This file is part of the Luau programming language and is licensed under MIT License; see LICENSE.txt for details
#pragma once
#include "Luau/AssemblyBuilderX64.h"
#include "Luau/IrData.h"
#include "Luau/RegisterX64.h"
#include <array>
#include <initializer_list>
namespace Luau
{
namespace CodeGen
{
Sync to upstream/release/566 (#853) * Fixed incorrect lexeme generated for string parts in the middle of an interpolated string (Fixes https://github.com/Roblox/luau/issues/744) * DeprecatedApi lint can report some issues without type inference information * Fixed performance of autocomplete requests when suggestions have large intersection types (Solves https://github.com/Roblox/luau/discussions/847) * Marked `table.getn`/`foreach`/`foreachi` as deprecated ([RFC: Deprecate table.getn/foreach/foreachi](https://github.com/Roblox/luau/blob/master/rfcs/deprecate-table-getn-foreach.md)) * With -O2 optimization level, we now optimize builtin calls based on known argument/return count. Note that this change can be observable if `getfenv/setfenv` is used to substitute a builtin, especially if arity is different. Fastcall heavy tests show a 1-2% improvement. * Luau can now be built with clang-cl (Fixes https://github.com/Roblox/luau/issues/736) We also made many improvements to our experimental components. For our new type solver: * Overhauled data flow analysis system, fixed issues with 'repeat' loops, global variables and type annotations * Type refinements now work on generic table indexing with a string literal * Type refinements will properly track potentially 'nil' values (like t[x] for a missing key) and their further refinements * Internal top table type is now isomorphic to `{}` which fixes issues when `typeof(v) == 'table'` type refinement is handled * References to non-existent types in type annotations no longer resolve to 'error' type like in old solver * Improved handling of class unions in property access expressions * Fixed default type packs * Unsealed tables can now have metatables * Restored expected types for function arguments And for native code generation: * Added min and max IR instructions mapping to vminsd/vmaxsd on x64 * We now speculatively extract direct execution fast-paths based on expected types of expressions which provides better optimization opportunities inside a single basic block * Translated existing math fastcalls to IR form to improve tag guard removal and constant propagation
2023-03-03 15:21:14 -05:00
namespace X64
{
constexpr uint8_t kNoStackSlot = 0xff;
struct IrSpillX64
{
uint32_t instIdx = 0;
bool useDoubleSlot = 0;
// Spill location can be a stack location or be empty
// When it's empty, it means that instruction value can be rematerialized
uint8_t stackSlot = kNoStackSlot;
RegisterX64 originalLoc = noreg;
};
struct IrRegAllocX64
{
IrRegAllocX64(AssemblyBuilderX64& build, IrFunction& function);
RegisterX64 allocGprReg(SizeX64 preferredSize, uint32_t instIdx);
RegisterX64 allocXmmReg(uint32_t instIdx);
RegisterX64 allocGprRegOrReuse(SizeX64 preferredSize, uint32_t instIdx, std::initializer_list<IrOp> oprefs);
RegisterX64 allocXmmRegOrReuse(uint32_t instIdx, std::initializer_list<IrOp> oprefs);
RegisterX64 takeReg(RegisterX64 reg, uint32_t instIdx);
Sync to upstream/release/566 (#853) * Fixed incorrect lexeme generated for string parts in the middle of an interpolated string (Fixes https://github.com/Roblox/luau/issues/744) * DeprecatedApi lint can report some issues without type inference information * Fixed performance of autocomplete requests when suggestions have large intersection types (Solves https://github.com/Roblox/luau/discussions/847) * Marked `table.getn`/`foreach`/`foreachi` as deprecated ([RFC: Deprecate table.getn/foreach/foreachi](https://github.com/Roblox/luau/blob/master/rfcs/deprecate-table-getn-foreach.md)) * With -O2 optimization level, we now optimize builtin calls based on known argument/return count. Note that this change can be observable if `getfenv/setfenv` is used to substitute a builtin, especially if arity is different. Fastcall heavy tests show a 1-2% improvement. * Luau can now be built with clang-cl (Fixes https://github.com/Roblox/luau/issues/736) We also made many improvements to our experimental components. For our new type solver: * Overhauled data flow analysis system, fixed issues with 'repeat' loops, global variables and type annotations * Type refinements now work on generic table indexing with a string literal * Type refinements will properly track potentially 'nil' values (like t[x] for a missing key) and their further refinements * Internal top table type is now isomorphic to `{}` which fixes issues when `typeof(v) == 'table'` type refinement is handled * References to non-existent types in type annotations no longer resolve to 'error' type like in old solver * Improved handling of class unions in property access expressions * Fixed default type packs * Unsealed tables can now have metatables * Restored expected types for function arguments And for native code generation: * Added min and max IR instructions mapping to vminsd/vmaxsd on x64 * We now speculatively extract direct execution fast-paths based on expected types of expressions which provides better optimization opportunities inside a single basic block * Translated existing math fastcalls to IR form to improve tag guard removal and constant propagation
2023-03-03 15:21:14 -05:00
void freeReg(RegisterX64 reg);
void freeLastUseReg(IrInst& target, uint32_t instIdx);
void freeLastUseRegs(const IrInst& inst, uint32_t instIdx);
bool isLastUseReg(const IrInst& target, uint32_t instIdx) const;
bool shouldFreeGpr(RegisterX64 reg) const;
// Register used by instruction is about to be freed, have to find a way to restore value later
void preserve(IrInst& inst);
void restore(IrInst& inst, bool intoOriginalLocation);
void preserveAndFreeInstValues();
uint32_t findInstructionWithFurthestNextUse(const std::array<uint32_t, 16>& regInstUsers) const;
void assertFree(RegisterX64 reg) const;
Sync to upstream/release/566 (#853) * Fixed incorrect lexeme generated for string parts in the middle of an interpolated string (Fixes https://github.com/Roblox/luau/issues/744) * DeprecatedApi lint can report some issues without type inference information * Fixed performance of autocomplete requests when suggestions have large intersection types (Solves https://github.com/Roblox/luau/discussions/847) * Marked `table.getn`/`foreach`/`foreachi` as deprecated ([RFC: Deprecate table.getn/foreach/foreachi](https://github.com/Roblox/luau/blob/master/rfcs/deprecate-table-getn-foreach.md)) * With -O2 optimization level, we now optimize builtin calls based on known argument/return count. Note that this change can be observable if `getfenv/setfenv` is used to substitute a builtin, especially if arity is different. Fastcall heavy tests show a 1-2% improvement. * Luau can now be built with clang-cl (Fixes https://github.com/Roblox/luau/issues/736) We also made many improvements to our experimental components. For our new type solver: * Overhauled data flow analysis system, fixed issues with 'repeat' loops, global variables and type annotations * Type refinements now work on generic table indexing with a string literal * Type refinements will properly track potentially 'nil' values (like t[x] for a missing key) and their further refinements * Internal top table type is now isomorphic to `{}` which fixes issues when `typeof(v) == 'table'` type refinement is handled * References to non-existent types in type annotations no longer resolve to 'error' type like in old solver * Improved handling of class unions in property access expressions * Fixed default type packs * Unsealed tables can now have metatables * Restored expected types for function arguments And for native code generation: * Added min and max IR instructions mapping to vminsd/vmaxsd on x64 * We now speculatively extract direct execution fast-paths based on expected types of expressions which provides better optimization opportunities inside a single basic block * Translated existing math fastcalls to IR form to improve tag guard removal and constant propagation
2023-03-03 15:21:14 -05:00
void assertAllFree() const;
void assertNoSpills() const;
Sync to upstream/release/566 (#853) * Fixed incorrect lexeme generated for string parts in the middle of an interpolated string (Fixes https://github.com/Roblox/luau/issues/744) * DeprecatedApi lint can report some issues without type inference information * Fixed performance of autocomplete requests when suggestions have large intersection types (Solves https://github.com/Roblox/luau/discussions/847) * Marked `table.getn`/`foreach`/`foreachi` as deprecated ([RFC: Deprecate table.getn/foreach/foreachi](https://github.com/Roblox/luau/blob/master/rfcs/deprecate-table-getn-foreach.md)) * With -O2 optimization level, we now optimize builtin calls based on known argument/return count. Note that this change can be observable if `getfenv/setfenv` is used to substitute a builtin, especially if arity is different. Fastcall heavy tests show a 1-2% improvement. * Luau can now be built with clang-cl (Fixes https://github.com/Roblox/luau/issues/736) We also made many improvements to our experimental components. For our new type solver: * Overhauled data flow analysis system, fixed issues with 'repeat' loops, global variables and type annotations * Type refinements now work on generic table indexing with a string literal * Type refinements will properly track potentially 'nil' values (like t[x] for a missing key) and their further refinements * Internal top table type is now isomorphic to `{}` which fixes issues when `typeof(v) == 'table'` type refinement is handled * References to non-existent types in type annotations no longer resolve to 'error' type like in old solver * Improved handling of class unions in property access expressions * Fixed default type packs * Unsealed tables can now have metatables * Restored expected types for function arguments And for native code generation: * Added min and max IR instructions mapping to vminsd/vmaxsd on x64 * We now speculatively extract direct execution fast-paths based on expected types of expressions which provides better optimization opportunities inside a single basic block * Translated existing math fastcalls to IR form to improve tag guard removal and constant propagation
2023-03-03 15:21:14 -05:00
AssemblyBuilderX64& build;
IrFunction& function;
uint32_t currInstIdx = ~0u;
std::array<bool, 16> freeGprMap;
std::array<uint32_t, 16> gprInstUsers;
std::array<bool, 16> freeXmmMap;
std::array<uint32_t, 16> xmmInstUsers;
std::bitset<256> usedSpillSlots;
unsigned maxUsedSlot = 0;
std::vector<IrSpillX64> spills;
};
struct ScopedRegX64
{
explicit ScopedRegX64(IrRegAllocX64& owner);
ScopedRegX64(IrRegAllocX64& owner, SizeX64 size);
ScopedRegX64(IrRegAllocX64& owner, RegisterX64 reg);
~ScopedRegX64();
ScopedRegX64(const ScopedRegX64&) = delete;
ScopedRegX64& operator=(const ScopedRegX64&) = delete;
void alloc(SizeX64 size);
void free();
RegisterX64 release();
IrRegAllocX64& owner;
RegisterX64 reg;
};
// When IR instruction makes a call under a condition that's not reflected as a real branch in IR,
// spilled values have to be restored to their exact original locations, so that both after a call
// and after the skip, values are found in the same place
struct ScopedSpills
{
explicit ScopedSpills(IrRegAllocX64& owner);
~ScopedSpills();
ScopedSpills(const ScopedSpills&) = delete;
ScopedSpills& operator=(const ScopedSpills&) = delete;
bool wasSpilledBefore(const IrSpillX64& spill) const;
IrRegAllocX64& owner;
std::vector<IrSpillX64> snapshot;
};
Sync to upstream/release/566 (#853) * Fixed incorrect lexeme generated for string parts in the middle of an interpolated string (Fixes https://github.com/Roblox/luau/issues/744) * DeprecatedApi lint can report some issues without type inference information * Fixed performance of autocomplete requests when suggestions have large intersection types (Solves https://github.com/Roblox/luau/discussions/847) * Marked `table.getn`/`foreach`/`foreachi` as deprecated ([RFC: Deprecate table.getn/foreach/foreachi](https://github.com/Roblox/luau/blob/master/rfcs/deprecate-table-getn-foreach.md)) * With -O2 optimization level, we now optimize builtin calls based on known argument/return count. Note that this change can be observable if `getfenv/setfenv` is used to substitute a builtin, especially if arity is different. Fastcall heavy tests show a 1-2% improvement. * Luau can now be built with clang-cl (Fixes https://github.com/Roblox/luau/issues/736) We also made many improvements to our experimental components. For our new type solver: * Overhauled data flow analysis system, fixed issues with 'repeat' loops, global variables and type annotations * Type refinements now work on generic table indexing with a string literal * Type refinements will properly track potentially 'nil' values (like t[x] for a missing key) and their further refinements * Internal top table type is now isomorphic to `{}` which fixes issues when `typeof(v) == 'table'` type refinement is handled * References to non-existent types in type annotations no longer resolve to 'error' type like in old solver * Improved handling of class unions in property access expressions * Fixed default type packs * Unsealed tables can now have metatables * Restored expected types for function arguments And for native code generation: * Added min and max IR instructions mapping to vminsd/vmaxsd on x64 * We now speculatively extract direct execution fast-paths based on expected types of expressions which provides better optimization opportunities inside a single basic block * Translated existing math fastcalls to IR form to improve tag guard removal and constant propagation
2023-03-03 15:21:14 -05:00
} // namespace X64
} // namespace CodeGen
} // namespace Luau