diff options
| author | gdkchan <gab.dark.100@gmail.com> | 2019-08-08 15:56:22 -0300 |
|---|---|---|
| committer | emmauss <emmausssss@gmail.com> | 2019-08-08 21:56:22 +0300 |
| commit | a731ab3a2aad56e6ceb8b4e2444a61353246295c (patch) | |
| tree | c7f13f51bfec6b19431e62167811ae31e9d2fea9 /ARMeilleure/CodeGen/RegisterAllocators/HybridAllocator.cs | |
| parent | 1ba58e9942e54175e3f3a0e1d57a48537f4888b1 (diff) | |
Add a new JIT compiler for CPU code (#693)
* Start of the ARMeilleure project
* Refactoring around the old IRAdapter, now renamed to PreAllocator
* Optimize the LowestBitSet method
* Add CLZ support and fix CLS implementation
* Add missing Equals and GetHashCode overrides on some structs, misc small tweaks
* Implement the ByteSwap IR instruction, and some refactoring on the assembler
* Implement the DivideUI IR instruction and fix 64-bits IDIV
* Correct constant operand type on CSINC
* Move division instructions implementation to InstEmitDiv
* Fix destination type for the ConditionalSelect IR instruction
* Implement UMULH and SMULH, with new IR instructions
* Fix some issues with shift instructions
* Fix constant types for BFM instructions
* Fix up new tests using the new V128 struct
* Update tests
* Move DIV tests to a separate file
* Add support for calls, and some instructions that depends on them
* Start adding support for SIMD & FP types, along with some of the related ARM instructions
* Fix some typos and the divide instruction with FP operands
* Fix wrong method call on Clz_V
* Implement ARM FP & SIMD move instructions, Saddlv_V, and misc. fixes
* Implement SIMD logical instructions and more misc. fixes
* Fix PSRAD x86 instruction encoding, TRN, UABD and UABDL implementations
* Implement float conversion instruction, merge in LDj3SNuD fixes, and some other misc. fixes
* Implement SIMD shift instruction and fix Dup_V
* Add SCVTF and UCVTF (vector, fixed-point) variants to the opcode table
* Fix check with tolerance on tester
* Implement FP & SIMD comparison instructions, and some fixes
* Update FCVT (Scalar) encoding on the table to support the Half-float variants
* Support passing V128 structs, some cleanup on the register allocator, merge LDj3SNuD fixes
* Use old memory access methods, made a start on SIMD memory insts support, some fixes
* Fix float constant passed to functions, save and restore non-volatile XMM registers, other fixes
* Fix arguments count with struct return values, other fixes
* More instructions
* Misc. fixes and integrate LDj3SNuD fixes
* Update tests
* Add a faster linear scan allocator, unwinding support on windows, and other changes
* Update Ryujinx.HLE
* Update Ryujinx.Graphics
* Fix V128 return pointer passing, RCX is clobbered
* Update Ryujinx.Tests
* Update ITimeZoneService
* Stop using GetFunctionPointer as that can't be called from native code, misc. fixes and tweaks
* Use generic GetFunctionPointerForDelegate method and other tweaks
* Some refactoring on the code generator, assert on invalid operations and use a separate enum for intrinsics
* Remove some unused code on the assembler
* Fix REX.W prefix regression on float conversion instructions, add some sort of profiler
* Add hardware capability detection
* Fix regression on Sha1h and revert Fcm** changes
* Add SSE2-only paths on vector extract and insert, some refactoring on the pre-allocator
* Fix silly mistake introduced on last commit on CpuId
* Generate inline stack probes when the stack allocation is too large
* Initial support for the System-V ABI
* Support multiple destination operands
* Fix SSE2 VectorInsert8 path, and other fixes
* Change placement of XMM callee save and restore code to match other compilers
* Rename Dest to Destination and Inst to Instruction
* Fix a regression related to calls and the V128 type
* Add an extra space on comments to match code style
* Some refactoring
* Fix vector insert FP32 SSE2 path
* Port over the ARM32 instructions
* Avoid memory protection races on JIT Cache
* Another fix on VectorInsert FP32 (thanks to LDj3SNuD
* Float operands don't need to use the same register when VEX is supported
* Add a new register allocator, higher quality code for hot code (tier up), and other tweaks
* Some nits, small improvements on the pre allocator
* CpuThreadState is gone
* Allow changing CPU emulators with a config entry
* Add runtime identifiers on the ARMeilleure project
* Allow switching between CPUs through a config entry (pt. 2)
* Change win10-x64 to win-x64 on projects
* Update the Ryujinx project to use ARMeilleure
* Ensure that the selected register is valid on the hybrid allocator
* Allow exiting on returns to 0 (should fix test regression)
* Remove register assignments for most used variables on the hybrid allocator
* Do not use fixed registers as spill temp
* Add missing namespace and remove unneeded using
* Address PR feedback
* Fix types, etc
* Enable AssumeStrictAbiCompliance by default
* Ensure that Spill and Fill don't load or store any more than necessary
Diffstat (limited to 'ARMeilleure/CodeGen/RegisterAllocators/HybridAllocator.cs')
| -rw-r--r-- | ARMeilleure/CodeGen/RegisterAllocators/HybridAllocator.cs | 382 |
1 files changed, 382 insertions, 0 deletions
diff --git a/ARMeilleure/CodeGen/RegisterAllocators/HybridAllocator.cs b/ARMeilleure/CodeGen/RegisterAllocators/HybridAllocator.cs new file mode 100644 index 00000000..9a827420 --- /dev/null +++ b/ARMeilleure/CodeGen/RegisterAllocators/HybridAllocator.cs @@ -0,0 +1,382 @@ +using ARMeilleure.Common; +using ARMeilleure.IntermediateRepresentation; +using ARMeilleure.Translation; +using System.Collections.Generic; +using System.Diagnostics; + +using static ARMeilleure.IntermediateRepresentation.OperandHelper; + +namespace ARMeilleure.CodeGen.RegisterAllocators +{ + class HybridAllocator : IRegisterAllocator + { + private const int RegistersCount = 16; + private const int MaxIROperands = 4; + + private struct BlockInfo + { + public bool HasCall { get; } + + public int IntFixedRegisters { get; } + public int VecFixedRegisters { get; } + + public BlockInfo(bool hasCall, int intFixedRegisters, int vecFixedRegisters) + { + HasCall = hasCall; + IntFixedRegisters = intFixedRegisters; + VecFixedRegisters = vecFixedRegisters; + } + } + + private class LocalInfo + { + public int Uses { get; set; } + public int UseCount { get; set; } + + public bool PreAllocated { get; set; } + public int Register { get; set; } + public int SpillOffset { get; set; } + + public int Sequence { get; set; } + + public Operand Temp { get; set; } + + public OperandType Type { get; } + + private int _first; + private int _last; + + public bool IsBlockLocal => _first == _last; + + public LocalInfo(OperandType type, int uses) + { + Uses = uses; + Type = type; + + _first = -1; + _last = -1; + } + + public void SetBlockIndex(int blkIndex) + { + if (_first == -1 || blkIndex < _first) + { + _first = blkIndex; + } + + if (_last == -1 || blkIndex > _last) + { + _last = blkIndex; + } + } + } + + public AllocationResult RunPass( + ControlFlowGraph cfg, + StackAllocator stackAlloc, + RegisterMasks regMasks) + { + int intUsedRegisters = 0; + int vecUsedRegisters = 0; + + int intFreeRegisters = regMasks.IntAvailableRegisters; + int vecFreeRegisters = regMasks.VecAvailableRegisters; + + BlockInfo[] blockInfo = new BlockInfo[cfg.Blocks.Count]; + + List<LocalInfo> locInfo = new List<LocalInfo>(); + + for (int index = cfg.PostOrderBlocks.Length - 1; index >= 0; index--) + { + BasicBlock block = cfg.PostOrderBlocks[index]; + + int intFixedRegisters = 0; + int vecFixedRegisters = 0; + + bool hasCall = false; + + foreach (Node node in block.Operations) + { + if (node is Operation operation && operation.Instruction == Instruction.Call) + { + hasCall = true; + } + + for (int srcIndex = 0; srcIndex < node.SourcesCount; srcIndex++) + { + Operand source = node.GetSource(srcIndex); + + if (source.Kind == OperandKind.LocalVariable) + { + locInfo[source.AsInt32() - 1].SetBlockIndex(block.Index); + } + } + + for (int dstIndex = 0; dstIndex < node.DestinationsCount; dstIndex++) + { + Operand dest = node.GetDestination(dstIndex); + + if (dest.Kind == OperandKind.LocalVariable) + { + LocalInfo info; + + if (dest.Value != 0) + { + info = locInfo[dest.AsInt32() - 1]; + } + else + { + dest.NumberLocal(locInfo.Count + 1); + + info = new LocalInfo(dest.Type, UsesCount(dest)); + + locInfo.Add(info); + } + + info.SetBlockIndex(block.Index); + } + else if (dest.Kind == OperandKind.Register) + { + if (dest.Type.IsInteger()) + { + intFixedRegisters |= 1 << dest.GetRegister().Index; + } + else + { + vecFixedRegisters |= 1 << dest.GetRegister().Index; + } + } + } + } + + blockInfo[block.Index] = new BlockInfo(hasCall, intFixedRegisters, vecFixedRegisters); + } + + int sequence = 0; + + for (int index = cfg.PostOrderBlocks.Length - 1; index >= 0; index--) + { + BasicBlock block = cfg.PostOrderBlocks[index]; + + BlockInfo blkInfo = blockInfo[block.Index]; + + int intLocalFreeRegisters = intFreeRegisters & ~blkInfo.IntFixedRegisters; + int vecLocalFreeRegisters = vecFreeRegisters & ~blkInfo.VecFixedRegisters; + + int intCallerSavedRegisters = blkInfo.HasCall ? regMasks.IntCallerSavedRegisters : 0; + int vecCallerSavedRegisters = blkInfo.HasCall ? regMasks.VecCallerSavedRegisters : 0; + + int intSpillTempRegisters = SelectSpillTemps( + intCallerSavedRegisters & ~blkInfo.IntFixedRegisters, + intLocalFreeRegisters); + int vecSpillTempRegisters = SelectSpillTemps( + vecCallerSavedRegisters & ~blkInfo.VecFixedRegisters, + vecLocalFreeRegisters); + + intLocalFreeRegisters &= ~(intSpillTempRegisters | intCallerSavedRegisters); + vecLocalFreeRegisters &= ~(vecSpillTempRegisters | vecCallerSavedRegisters); + + for (LinkedListNode<Node> llNode = block.Operations.First; llNode != null; llNode = llNode.Next) + { + Node node = llNode.Value; + + int intLocalUse = 0; + int vecLocalUse = 0; + + for (int srcIndex = 0; srcIndex < node.SourcesCount; srcIndex++) + { + Operand source = node.GetSource(srcIndex); + + if (source.Kind != OperandKind.LocalVariable) + { + continue; + } + + LocalInfo info = locInfo[source.AsInt32() - 1]; + + info.UseCount++; + + Debug.Assert(info.UseCount <= info.Uses); + + if (info.Register != -1) + { + node.SetSource(srcIndex, Register(info.Register, source.Type.ToRegisterType(), source.Type)); + + if (info.UseCount == info.Uses && !info.PreAllocated) + { + if (source.Type.IsInteger()) + { + intLocalFreeRegisters |= 1 << info.Register; + } + else + { + vecLocalFreeRegisters |= 1 << info.Register; + } + } + } + else + { + Operand temp = info.Temp; + + if (temp == null || info.Sequence != sequence) + { + temp = source.Type.IsInteger() + ? GetSpillTemp(source, intSpillTempRegisters, ref intLocalUse) + : GetSpillTemp(source, vecSpillTempRegisters, ref vecLocalUse); + + info.Sequence = sequence; + info.Temp = temp; + } + + node.SetSource(srcIndex, temp); + + Operation fillOp = new Operation(Instruction.Fill, temp, Const(info.SpillOffset)); + + block.Operations.AddBefore(llNode, fillOp); + } + } + + int intLocalAsg = 0; + int vecLocalAsg = 0; + + for (int dstIndex = 0; dstIndex < node.DestinationsCount; dstIndex++) + { + Operand dest = node.GetDestination(dstIndex); + + if (dest.Kind != OperandKind.LocalVariable) + { + continue; + } + + LocalInfo info = locInfo[dest.AsInt32() - 1]; + + if (info.UseCount == 0 && !info.PreAllocated) + { + int mask = dest.Type.IsInteger() + ? intLocalFreeRegisters + : vecLocalFreeRegisters; + + if (info.IsBlockLocal && mask != 0) + { + int selectedReg = BitUtils.LowestBitSet(mask); + + info.Register = selectedReg; + + if (dest.Type.IsInteger()) + { + intLocalFreeRegisters &= ~(1 << selectedReg); + intUsedRegisters |= 1 << selectedReg; + } + else + { + vecLocalFreeRegisters &= ~(1 << selectedReg); + vecUsedRegisters |= 1 << selectedReg; + } + } + else + { + info.Register = -1; + info.SpillOffset = stackAlloc.Allocate(dest.Type.GetSizeInBytes()); + } + } + + info.UseCount++; + + Debug.Assert(info.UseCount <= info.Uses); + + if (info.Register != -1) + { + node.SetDestination(dstIndex, Register(info.Register, dest.Type.ToRegisterType(), dest.Type)); + } + else + { + Operand temp = info.Temp; + + if (temp == null || info.Sequence != sequence) + { + temp = dest.Type.IsInteger() + ? GetSpillTemp(dest, intSpillTempRegisters, ref intLocalAsg) + : GetSpillTemp(dest, vecSpillTempRegisters, ref vecLocalAsg); + + info.Sequence = sequence; + info.Temp = temp; + } + + node.SetDestination(dstIndex, temp); + + Operation spillOp = new Operation(Instruction.Spill, null, Const(info.SpillOffset), temp); + + llNode = block.Operations.AddAfter(llNode, spillOp); + } + } + + sequence++; + + intUsedRegisters |= intLocalAsg | intLocalUse; + vecUsedRegisters |= vecLocalAsg | vecLocalUse; + } + } + + return new AllocationResult(intUsedRegisters, vecUsedRegisters, stackAlloc.TotalSize); + } + + private static int SelectSpillTemps(int mask0, int mask1) + { + int selection = 0; + int count = 0; + + while (count < MaxIROperands && mask0 != 0) + { + int mask = mask0 & -mask0; + + selection |= mask; + + mask0 &= ~mask; + + count++; + } + + while (count < MaxIROperands && mask1 != 0) + { + int mask = mask1 & -mask1; + + selection |= mask; + + mask1 &= ~mask; + + count++; + } + + Debug.Assert(count == MaxIROperands, "No enough registers for spill temps."); + + return selection; + } + + private static Operand GetSpillTemp(Operand local, int freeMask, ref int useMask) + { + int selectedReg = BitUtils.LowestBitSet(freeMask & ~useMask); + + useMask |= 1 << selectedReg; + + return Register(selectedReg, local.Type.ToRegisterType(), local.Type); + } + + private static int UsesCount(Operand local) + { + return local.Assignments.Count + local.Uses.Count; + } + + private static IEnumerable<BasicBlock> Successors(BasicBlock block) + { + if (block.Next != null) + { + yield return block.Next; + } + + if (block.Branch != null) + { + yield return block.Branch; + } + } + } +}
\ No newline at end of file |
