Ryujinx/ARMeilleure/Diagnostics/IRDumper.cs
FICTURE7 22b2cb39af
Reduce JIT GC allocations (#2515)
* Turn `MemoryOperand` into a struct

* Remove `IntrinsicOperation`

* Remove `PhiNode`

* Remove `Node`

* Turn `Operand` into a struct

* Turn `Operation` into a struct

* Clean up pool management methods

* Add `Arena` allocator

* Move `OperationHelper` to `Operation.Factory`

* Move `OperandHelper` to `Operand.Factory`

* Optimize `Operation` a bit

* Fix `Arena` initialization

* Rename `NativeList<T>` to `ArenaList<T>`

* Reduce `Operand` size from 88 to 56 bytes

* Reduce `Operation` size from 56 to 40 bytes

* Add optimistic interning of Register & Constant operands

* Optimize `RegisterUsage` pass a bit

* Optimize `RemoveUnusedNodes` pass a bit

Iterating in reverse-order allows killing dependency chains in a single
pass.

* Fix PPTC symbols

* Optimize `BasicBlock` a bit

Reduce allocations from `_successor` & `DominanceFrontiers`

* Fix `Operation` resize

* Make `Arena` expandable

Change the arena allocator to be expandable by allocating in pages, with
some of them being pooled. Currently 32 pages are pooled. An LRU removal
mechanism should probably be added to it.

Apparently MHR can allocate bitmaps large enough to exceed the 16MB
limit for the type.

* Move `Arena` & `ArenaList` to `Common`

* Remove `ThreadStaticPool` & co

* Add `PhiOperation`

* Reduce `Operand` size from 56 from 48 bytes

* Add linear-probing to `Operand` intern table

* Optimize `HybridAllocator` a bit

* Add `Allocators` class

* Tune `ArenaAllocator` sizes

* Add page removal mechanism to `ArenaAllocator`

Remove pages which have not been used for more than 5s after each reset.

I am on fence if this would be better using a Gen2 callback object like
the one in System.Buffers.ArrayPool<T>, to trim the pool. Because right
now if a large translation happens, the pages will be freed only after a
reset. This reset may not happen for a while because no new translation
is hit, but the arena base sizes are rather small.

* Fix `OOM` when allocating larger than page size in `ArenaAllocator`

Tweak resizing mechanism for Operand.Uses and Assignemnts.

* Optimize `Optimizer` a bit

* Optimize `Operand.Add<T>/Remove<T>` a bit

* Clean up `PreAllocator`

* Fix phi insertion order

Reduce codegen diffs.

* Fix code alignment

* Use new heuristics for degree of parallelism

* Suppress warnings

* Address gdkchan's feedback

Renamed `GetValue()` to `GetValueUnsafe()` to make it more clear that
`Operand.Value` should usually not be modified directly.

* Add fast path to `ArenaAllocator`

* Assembly for `ArenaAllocator.Allocate(ulong)`:

  .L0:
    mov rax, [rcx+0x18]
    lea r8, [rax+rdx]
    cmp r8, [rcx+0x10]
    ja short .L2
  .L1:
    mov rdx, [rcx+8]
    add rax, [rdx+8]
    mov [rcx+0x18], r8
    ret
  .L2:
    jmp ArenaAllocator.AllocateSlow(UInt64)

  A few variable/field had to be changed to ulong so that RyuJIT avoids
  emitting zero-extends.

* Implement a new heuristic to free pooled pages.

  If an arena is used often, it is more likely that its pages will be
  needed, so the pages are kept for longer (e.g: during PPTC rebuild or
  burst sof compilations). If is not used often, then it is more likely
  that its pages will not be needed (e.g: after PPTC rebuild or bursts
  of compilations).

* Address riperiperi's feedback

* Use `EqualityComparer<T>` in `IntrusiveList<T>`

Avoids a potential GC hole in `Equals(T, T)`.
2021-08-17 15:08:34 -03:00

311 lines
9.3 KiB
C#

using ARMeilleure.IntermediateRepresentation;
using ARMeilleure.Translation;
using System;
using System.Collections.Generic;
using System.Linq;
using System.Text;
namespace ARMeilleure.Diagnostics
{
class IRDumper
{
private const string Indentation = " ";
private int _indentLevel;
private readonly StringBuilder _builder;
private readonly Dictionary<Operand, string> _localNames;
private readonly Dictionary<ulong, string> _symbolNames;
public IRDumper(int indent)
{
_indentLevel = indent;
_builder = new StringBuilder();
_localNames = new Dictionary<Operand, string>();
_symbolNames = new Dictionary<ulong, string>();
}
private void Indent()
{
_builder.EnsureCapacity(_builder.Capacity + _indentLevel * Indentation.Length);
for (int index = 0; index < _indentLevel; index++)
{
_builder.Append(Indentation);
}
}
private void IncreaseIndentation()
{
_indentLevel++;
}
private void DecreaseIndentation()
{
_indentLevel--;
}
private void DumpBlockName(BasicBlock block)
{
_builder.Append("block").Append(block.Index);
}
private void DumpBlockHeader(BasicBlock block)
{
DumpBlockName(block);
if (block.Frequency == BasicBlockFrequency.Cold)
{
_builder.Append(" cold");
}
if (block.SuccessorsCount > 0)
{
_builder.Append(" (");
for (int i = 0; i < block.SuccessorsCount; i++)
{
DumpBlockName(block.GetSuccessor(i));
if (i < block.SuccessorsCount - 1)
{
_builder.Append(", ");
}
}
_builder.Append(')');
}
_builder.Append(':');
}
private void DumpOperand(Operand operand)
{
if (operand == default)
{
_builder.Append("<NULL>");
return;
}
_builder.Append(GetTypeName(operand.Type)).Append(' ');
switch (operand.Kind)
{
case OperandKind.LocalVariable:
if (!_localNames.TryGetValue(operand, out string localName))
{
localName = $"%{_localNames.Count}";
_localNames.Add(operand, localName);
}
_builder.Append(localName);
break;
case OperandKind.Register:
Register reg = operand.GetRegister();
switch (reg.Type)
{
case RegisterType.Flag: _builder.Append('b'); break;
case RegisterType.FpFlag: _builder.Append('f'); break;
case RegisterType.Integer: _builder.Append('r'); break;
case RegisterType.Vector: _builder.Append('v'); break;
}
_builder.Append(reg.Index);
break;
case OperandKind.Constant:
string symbolName = Symbols.Get(operand.Value);
if (symbolName != null && !_symbolNames.ContainsKey(operand.Value))
{
_symbolNames.Add(operand.Value, symbolName);
}
_builder.Append("0x").Append(operand.Value.ToString("X"));
break;
case OperandKind.Memory:
var memOp = operand.GetMemory();
_builder.Append('[');
DumpOperand(memOp.BaseAddress);
if (memOp.Index != default)
{
_builder.Append(" + ");
DumpOperand(memOp.Index);
switch (memOp.Scale)
{
case Multiplier.x2: _builder.Append("*2"); break;
case Multiplier.x4: _builder.Append("*4"); break;
case Multiplier.x8: _builder.Append("*8"); break;
}
}
if (memOp.Displacement != 0)
{
_builder.Append(" + 0x").Append(memOp.Displacement.ToString("X"));
}
_builder.Append(']');
break;
default:
_builder.Append(operand.Type);
break;
}
}
private void DumpNode(ControlFlowGraph cfg, Operation node)
{
for (int index = 0; index < node.DestinationsCount; index++)
{
DumpOperand(node.GetDestination(index));
if (index == node.DestinationsCount - 1)
{
_builder.Append(" = ");
}
else
{
_builder.Append(", ");
}
}
switch (node)
{
case Operation operation:
if (operation.Instruction == Instruction.Phi)
{
PhiOperation phi = operation.AsPhi();
_builder.Append("Phi ");
for (int index = 0; index < phi.SourcesCount; index++)
{
_builder.Append('(');
DumpBlockName(phi.GetBlock(cfg, index));
_builder.Append(": ");
DumpOperand(phi.GetSource(index));
_builder.Append(')');
if (index < phi.SourcesCount - 1)
{
_builder.Append(", ");
}
}
break;
}
bool comparison = false;
_builder.Append(operation.Instruction);
if (operation.Instruction == Instruction.Extended)
{
_builder.Append('.').Append(operation.Intrinsic);
}
else if (operation.Instruction == Instruction.BranchIf ||
operation.Instruction == Instruction.Compare)
{
comparison = true;
}
_builder.Append(' ');
for (int index = 0; index < operation.SourcesCount; index++)
{
Operand source = operation.GetSource(index);
if (index < operation.SourcesCount - 1)
{
DumpOperand(source);
_builder.Append(", ");
}
else if (comparison)
{
_builder.Append((Comparison)source.AsInt32());
}
else
{
DumpOperand(source);
}
}
break;
}
if (_symbolNames.Count == 1)
{
_builder.Append(" ;; ").Append(_symbolNames.First().Value);
}
else if (_symbolNames.Count > 1)
{
_builder.Append(" ;;");
foreach ((ulong value, string name) in _symbolNames)
{
_builder.Append(" 0x").Append(value.ToString("X")).Append(" = ").Append(name);
}
}
// Reset the set of symbols for the next Node we're going to dump.
_symbolNames.Clear();
}
public static string GetDump(ControlFlowGraph cfg)
{
var dumper = new IRDumper(1);
for (BasicBlock block = cfg.Blocks.First; block != null; block = block.ListNext)
{
dumper.Indent();
dumper.DumpBlockHeader(block);
dumper._builder.AppendLine();
dumper.IncreaseIndentation();
for (Operation node = block.Operations.First; node != default; node = node.ListNext)
{
dumper.Indent();
dumper.DumpNode(cfg, node);
dumper._builder.AppendLine();
}
dumper.DecreaseIndentation();
}
return dumper._builder.ToString();
}
private static string GetTypeName(OperandType type)
{
return type switch
{
OperandType.None => "none",
OperandType.I32 => "i32",
OperandType.I64 => "i64",
OperandType.FP32 => "f32",
OperandType.FP64 => "f64",
OperandType.V128 => "v128",
_ => throw new ArgumentException($"Invalid operand type \"{type}\"."),
};
}
}
}