R/ARMeilleure/Instructions/InstEmitSystem.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

158 lines
6.3 KiB
C#

using ARMeilleure.Decoders;
using ARMeilleure.IntermediateRepresentation;
using ARMeilleure.State;
using ARMeilleure.Translation;
using System;
using System.Reflection;
using static ARMeilleure.Instructions.InstEmitHelper;
using static ARMeilleure.IntermediateRepresentation.Operand.Factory;
namespace ARMeilleure.Instructions
{
static partial class InstEmit
{
private const int DczSizeLog2 = 4;
public static void Hint(ArmEmitterContext context)
{
// Execute as no-op.
}
public static void Isb(ArmEmitterContext context)
{
// Execute as no-op.
}
public static void Mrs(ArmEmitterContext context)
{
OpCodeSystem op = (OpCodeSystem)context.CurrOp;
MethodInfo info;
switch (GetPackedId(op))
{
case 0b11_011_0000_0000_001: info = typeof(NativeInterface).GetMethod(nameof(NativeInterface.GetCtrEl0)); break;
case 0b11_011_0000_0000_111: info = typeof(NativeInterface).GetMethod(nameof(NativeInterface.GetDczidEl0)); break;
case 0b11_011_0100_0010_000: EmitGetNzcv(context); return;
case 0b11_011_0100_0100_000: info = typeof(NativeInterface).GetMethod(nameof(NativeInterface.GetFpcr)); break;
case 0b11_011_0100_0100_001: info = typeof(NativeInterface).GetMethod(nameof(NativeInterface.GetFpsr)); break;
case 0b11_011_1101_0000_010: info = typeof(NativeInterface).GetMethod(nameof(NativeInterface.GetTpidrEl0)); break;
case 0b11_011_1101_0000_011: info = typeof(NativeInterface).GetMethod(nameof(NativeInterface.GetTpidr)); break;
case 0b11_011_1110_0000_000: info = typeof(NativeInterface).GetMethod(nameof(NativeInterface.GetCntfrqEl0)); break;
case 0b11_011_1110_0000_001: info = typeof(NativeInterface).GetMethod(nameof(NativeInterface.GetCntpctEl0)); break;
case 0b11_011_1110_0000_010: info = typeof(NativeInterface).GetMethod(nameof(NativeInterface.GetCntvctEl0)); break;
default: throw new NotImplementedException($"Unknown MRS 0x{op.RawOpCode:X8} at 0x{op.Address:X16}.");
}
SetIntOrZR(context, op.Rt, context.Call(info));
}
public static void Msr(ArmEmitterContext context)
{
OpCodeSystem op = (OpCodeSystem)context.CurrOp;
MethodInfo info;
switch (GetPackedId(op))
{
case 0b11_011_0100_0010_000: EmitSetNzcv(context); return;
case 0b11_011_0100_0100_000: info = typeof(NativeInterface).GetMethod(nameof(NativeInterface.SetFpcr)); break;
case 0b11_011_0100_0100_001: info = typeof(NativeInterface).GetMethod(nameof(NativeInterface.SetFpsr)); break;
case 0b11_011_1101_0000_010: info = typeof(NativeInterface).GetMethod(nameof(NativeInterface.SetTpidrEl0)); break;
default: throw new NotImplementedException($"Unknown MSR 0x{op.RawOpCode:X8} at 0x{op.Address:X16}.");
}
context.Call(info, GetIntOrZR(context, op.Rt));
}
public static void Nop(ArmEmitterContext context)
{
// Do nothing.
}
public static void Sys(ArmEmitterContext context)
{
// This instruction is used to do some operations on the CPU like cache invalidation,
// address translation and the like.
// We treat it as no-op here since we don't have any cache being emulated anyway.
OpCodeSystem op = (OpCodeSystem)context.CurrOp;
switch (GetPackedId(op))
{
case 0b11_011_0111_0100_001:
{
// DC ZVA
Operand t = GetIntOrZR(context, op.Rt);
for (long offset = 0; offset < (4 << DczSizeLog2); offset += 8)
{
Operand address = context.Add(t, Const(offset));
context.Call(typeof(NativeInterface).GetMethod(nameof(NativeInterface.WriteUInt64)), address, Const(0L));
}
break;
}
// No-op
case 0b11_011_0111_1110_001: //DC CIVAC
break;
}
}
private static int GetPackedId(OpCodeSystem op)
{
int id;
id = op.Op2 << 0;
id |= op.CRm << 3;
id |= op.CRn << 7;
id |= op.Op1 << 11;
id |= op.Op0 << 14;
return id;
}
private static void EmitGetNzcv(ArmEmitterContext context)
{
OpCodeSystem op = (OpCodeSystem)context.CurrOp;
Operand vSh = context.ShiftLeft(GetFlag(PState.VFlag), Const((int)PState.VFlag));
Operand cSh = context.ShiftLeft(GetFlag(PState.CFlag), Const((int)PState.CFlag));
Operand zSh = context.ShiftLeft(GetFlag(PState.ZFlag), Const((int)PState.ZFlag));
Operand nSh = context.ShiftLeft(GetFlag(PState.NFlag), Const((int)PState.NFlag));
Operand nzcvSh = context.BitwiseOr(context.BitwiseOr(nSh, zSh), context.BitwiseOr(cSh, vSh));
SetIntOrZR(context, op.Rt, nzcvSh);
}
private static void EmitSetNzcv(ArmEmitterContext context)
{
OpCodeSystem op = (OpCodeSystem)context.CurrOp;
Operand t = GetIntOrZR(context, op.Rt);
t = context.ConvertI64ToI32(t);
Operand v = context.ShiftRightUI(t, Const((int)PState.VFlag));
v = context.BitwiseAnd (v, Const(1));
Operand c = context.ShiftRightUI(t, Const((int)PState.CFlag));
c = context.BitwiseAnd (c, Const(1));
Operand z = context.ShiftRightUI(t, Const((int)PState.ZFlag));
z = context.BitwiseAnd (z, Const(1));
Operand n = context.ShiftRightUI(t, Const((int)PState.NFlag));
n = context.BitwiseAnd (n, Const(1));
SetFlag(context, PState.VFlag, v);
SetFlag(context, PState.CFlag, c);
SetFlag(context, PState.ZFlag, z);
SetFlag(context, PState.NFlag, n);
}
}
}