Ryujinx/src/Ryujinx.Graphics.Shader/Instructions/InstEmitAttribute.cs

383 lines
15 KiB
C#
Raw Normal View History

using Ryujinx.Graphics.Shader.Decoders;
using Ryujinx.Graphics.Shader.IntermediateRepresentation;
using Ryujinx.Graphics.Shader.Translation;
using static Ryujinx.Graphics.Shader.Instructions.InstEmitHelper;
using static Ryujinx.Graphics.Shader.IntermediateRepresentation.OperandHelper;
namespace Ryujinx.Graphics.Shader.Instructions
{
static partial class InstEmit
{
public static void Al2p(EmitterContext context)
{
InstAl2p op = context.GetOp<InstAl2p>();
context.Copy(GetDest(op.Dest), context.IAdd(GetSrcReg(context, op.SrcA), Const(op.Imm11)));
}
public static void Ald(EmitterContext context)
{
InstAld op = context.GetOp<InstAld>();
// Some of those attributes are per invocation,
// so we should ignore any primitive vertex indexing for those.
bool hasPrimitiveVertex = AttributeMap.HasPrimitiveVertex(context.Config.Stage, op.O) && !op.P;
if (!op.Phys)
{
hasPrimitiveVertex &= HasPrimitiveVertex(op.Imm11);
}
Operand primVertex = hasPrimitiveVertex ? context.Copy(GetSrcReg(context, op.SrcB)) : null;
for (int index = 0; index < (int)op.AlSize + 1; index++)
{
Register rd = new Register(op.Dest + index, RegisterType.Gpr);
if (rd.IsRZ)
{
break;
}
if (op.Phys)
{
Operand offset = context.ISubtract(GetSrcReg(context, op.SrcA), Const(AttributeConsts.UserAttributeBase));
Operand vecIndex = context.ShiftRightU32(offset, Const(4));
Operand elemIndex = context.BitwiseAnd(context.ShiftRightU32(offset, Const(2)), Const(3));
StorageKind storageKind = op.O ? StorageKind.Output : StorageKind.Input;
context.Copy(Register(rd), context.Load(storageKind, IoVariable.UserDefined, primVertex, vecIndex, elemIndex));
}
else if (op.SrcB == RegisterConsts.RegisterZeroIndex || op.P)
{
int offset = FixedFuncToUserAttribute(context.Config, op.Imm11 + index * 4, op.O);
context.FlagAttributeRead(offset);
bool isOutput = op.O && CanLoadOutput(offset);
if (!op.P && !isOutput && TryConvertIdToIndexForVulkan(context, offset, out Operand value))
{
context.Copy(Register(rd), value);
}
else
{
context.Copy(Register(rd), AttributeMap.GenerateAttributeLoad(context, primVertex, offset, isOutput, op.P));
}
}
else
{
int offset = FixedFuncToUserAttribute(context.Config, op.Imm11 + index * 4, op.O);
context.FlagAttributeRead(offset);
bool isOutput = op.O && CanLoadOutput(offset);
context.Copy(Register(rd), AttributeMap.GenerateAttributeLoad(context, primVertex, offset, isOutput, false));
}
}
}
public static void Ast(EmitterContext context)
{
InstAst op = context.GetOp<InstAst>();
for (int index = 0; index < (int)op.AlSize + 1; index++)
{
if (op.SrcB + index > RegisterConsts.RegisterZeroIndex)
{
break;
}
Register rd = new Register(op.SrcB + index, RegisterType.Gpr);
if (op.Phys)
{
Operand offset = context.ISubtract(GetSrcReg(context, op.SrcA), Const(AttributeConsts.UserAttributeBase));
Operand vecIndex = context.ShiftRightU32(offset, Const(4));
Operand elemIndex = context.BitwiseAnd(context.ShiftRightU32(offset, Const(2)), Const(3));
Operand invocationId = AttributeMap.HasInvocationId(context.Config.Stage, isOutput: true)
? context.Load(StorageKind.Input, IoVariable.InvocationId)
: null;
context.Store(StorageKind.Output, IoVariable.UserDefined, invocationId, vecIndex, elemIndex, Register(rd));
}
else
{
// TODO: Support indirect stores using Ra.
int offset = op.Imm11 + index * 4;
if (!context.Config.IsUsedOutputAttribute(offset))
{
return;
}
offset = FixedFuncToUserAttribute(context.Config, offset, isOutput: true);
context.FlagAttributeWritten(offset);
AttributeMap.GenerateAttributeStore(context, offset, op.P, Register(rd));
}
}
}
public static void Ipa(EmitterContext context)
{
InstIpa op = context.GetOp<InstIpa>();
context.FlagAttributeRead(op.Imm10);
Operand res;
bool isFixedFunc = false;
if (op.Idx)
{
Operand offset = context.ISubtract(GetSrcReg(context, op.SrcA), Const(AttributeConsts.UserAttributeBase));
Operand vecIndex = context.ShiftRightU32(offset, Const(4));
Operand elemIndex = context.BitwiseAnd(context.ShiftRightU32(offset, Const(2)), Const(3));
res = context.Load(StorageKind.Input, IoVariable.UserDefined, null, vecIndex, elemIndex);
res = context.FPMultiply(res, context.Load(StorageKind.Input, IoVariable.FragmentCoord, null, Const(3)));
}
else
{
isFixedFunc = TryFixedFuncToUserAttributeIpa(context, op.Imm10, out res);
if (op.Imm10 >= AttributeConsts.UserAttributeBase && op.Imm10 < AttributeConsts.UserAttributeEnd)
{
int index = (op.Imm10 - AttributeConsts.UserAttributeBase) >> 4;
if (context.Config.ImapTypes[index].GetFirstUsedType() == PixelImap.Perspective)
{
res = context.FPMultiply(res, context.Load(StorageKind.Input, IoVariable.FragmentCoord, null, Const(3)));
}
}
else if (op.Imm10 == AttributeConsts.PositionX || op.Imm10 == AttributeConsts.PositionY)
{
// FragCoord X/Y must be divided by the render target scale, if resolution scaling is active,
// because the shader code is not expecting scaled values.
res = context.FPDivide(res, context.Load(StorageKind.Input, IoVariable.SupportBlockRenderScale, null, Const(0)));
}
else if (op.Imm10 == AttributeConsts.FrontFacing && context.Config.GpuAccessor.QueryHostHasFrontFacingBug())
{
// gl_FrontFacing sometimes has incorrect (flipped) values depending how it is accessed on Intel GPUs.
// This weird trick makes it behave.
res = context.ICompareLess(context.INegate(context.IConvertS32ToFP32(res)), Const(0));
}
}
if (op.IpaOp == IpaOp.Multiply && !isFixedFunc)
{
Operand srcB = GetSrcReg(context, op.SrcB);
res = context.FPMultiply(res, srcB);
}
res = context.FPSaturate(res, op.Sat);
context.Copy(GetDest(op.Dest), res);
}
public static void Isberd(EmitterContext context)
{
InstIsberd op = context.GetOp<InstIsberd>();
// This instruction performs a load from ISBE (Internal Stage Buffer Entry) memory.
// Here, we just propagate the offset, as the result from this instruction is usually
// used with ALD to perform vertex load on geometry or tessellation shaders.
// The offset is calculated as (PrimitiveIndex * VerticesPerPrimitive) + VertexIndex.
// Since we hardcode PrimitiveIndex to zero, then the offset will be just VertexIndex.
context.Copy(GetDest(op.Dest), GetSrcReg(context, op.SrcA));
}
public static void OutR(EmitterContext context)
{
InstOutR op = context.GetOp<InstOutR>();
EmitOut(context, op.OutType.HasFlag(OutType.Emit), op.OutType.HasFlag(OutType.Cut));
}
public static void OutI(EmitterContext context)
{
InstOutI op = context.GetOp<InstOutI>();
EmitOut(context, op.OutType.HasFlag(OutType.Emit), op.OutType.HasFlag(OutType.Cut));
}
public static void OutC(EmitterContext context)
{
InstOutC op = context.GetOp<InstOutC>();
EmitOut(context, op.OutType.HasFlag(OutType.Emit), op.OutType.HasFlag(OutType.Cut));
}
private static void EmitOut(EmitterContext context, bool emit, bool cut)
{
if (!(emit || cut))
{
context.Config.GpuAccessor.Log("Invalid OUT encoding.");
}
if (emit)
{
if (context.Config.LastInVertexPipeline)
{
context.PrepareForVertexReturn(out var tempXLocal, out var tempYLocal, out var tempZLocal);
context.EmitVertex();
// Restore output position value before transformation.
if (tempXLocal != null)
{
context.Copy(context.Load(StorageKind.Input, IoVariable.Position, null, Const(0)), tempXLocal);
}
if (tempYLocal != null)
{
context.Copy(context.Load(StorageKind.Input, IoVariable.Position, null, Const(1)), tempYLocal);
}
if (tempZLocal != null)
{
context.Copy(context.Load(StorageKind.Input, IoVariable.Position, null, Const(2)), tempZLocal);
}
}
else
{
context.EmitVertex();
}
}
if (cut)
{
context.EndPrimitive();
}
}
private static bool HasPrimitiveVertex(int attr)
{
return attr != AttributeConsts.PrimitiveId &&
attr != AttributeConsts.TessCoordX &&
attr != AttributeConsts.TessCoordY;
}
private static bool CanLoadOutput(int attr)
{
return attr != AttributeConsts.TessCoordX && attr != AttributeConsts.TessCoordY;
}
private static bool TryFixedFuncToUserAttributeIpa(EmitterContext context, int attr, out Operand selectedAttr)
{
if (attr >= AttributeConsts.FrontColorDiffuseR && attr < AttributeConsts.BackColorDiffuseR)
{
// TODO: If two sided rendering is enabled, then this should return
// FrontColor if the fragment is front facing, and back color otherwise.
selectedAttr = GenerateIpaLoad(context, FixedFuncToUserAttribute(context.Config, attr, isOutput: false));
return true;
}
else if (attr == AttributeConsts.FogCoord)
{
// TODO: We likely need to emulate the fixed-function functionality for FogCoord here.
selectedAttr = GenerateIpaLoad(context, FixedFuncToUserAttribute(context.Config, attr, isOutput: false));
return true;
}
else if (attr >= AttributeConsts.BackColorDiffuseR && attr < AttributeConsts.ClipDistance0)
{
selectedAttr = ConstF(((attr >> 2) & 3) == 3 ? 1f : 0f);
return true;
}
else if (attr >= AttributeConsts.TexCoordBase && attr < AttributeConsts.TexCoordEnd)
{
selectedAttr = GenerateIpaLoad(context, FixedFuncToUserAttribute(context.Config, attr, isOutput: false));
return true;
}
selectedAttr = GenerateIpaLoad(context, attr);
return false;
}
private static Operand GenerateIpaLoad(EmitterContext context, int offset)
{
return AttributeMap.GenerateAttributeLoad(context, null, offset, isOutput: false, isPerPatch: false);
}
private static int FixedFuncToUserAttribute(ShaderConfig config, int attr, bool isOutput)
{
bool supportsLayerFromVertexOrTess = config.GpuAccessor.QueryHostSupportsLayerVertexTessellation();
int fixedStartAttr = supportsLayerFromVertexOrTess ? 0 : 1;
if (attr == AttributeConsts.Layer && config.Stage != ShaderStage.Geometry && !supportsLayerFromVertexOrTess)
{
attr = FixedFuncToUserAttribute(config, attr, AttributeConsts.Layer, 0, isOutput);
config.SetLayerOutputAttribute(attr);
}
else if (attr == AttributeConsts.FogCoord)
{
attr = FixedFuncToUserAttribute(config, attr, AttributeConsts.FogCoord, fixedStartAttr, isOutput);
}
else if (attr >= AttributeConsts.FrontColorDiffuseR && attr < AttributeConsts.ClipDistance0)
{
attr = FixedFuncToUserAttribute(config, attr, AttributeConsts.FrontColorDiffuseR, fixedStartAttr + 1, isOutput);
}
else if (attr >= AttributeConsts.TexCoordBase && attr < AttributeConsts.TexCoordEnd)
{
attr = FixedFuncToUserAttribute(config, attr, AttributeConsts.TexCoordBase, fixedStartAttr + 5, isOutput);
}
return attr;
}
private static int FixedFuncToUserAttribute(ShaderConfig config, int attr, int baseAttr, int baseIndex, bool isOutput)
{
int index = (attr - baseAttr) >> 4;
int userAttrIndex = config.GetFreeUserAttribute(isOutput, baseIndex + index);
if ((uint)userAttrIndex < Constants.MaxAttributes)
{
attr = AttributeConsts.UserAttributeBase + userAttrIndex * 16 + (attr & 0xf);
if (isOutput)
{
config.SetOutputUserAttributeFixedFunc(userAttrIndex);
}
else
{
config.SetInputUserAttributeFixedFunc(userAttrIndex);
}
}
else
{
config.GpuAccessor.Log($"No enough user attributes for fixed attribute offset 0x{attr:X}.");
}
return attr;
}
private static bool TryConvertIdToIndexForVulkan(EmitterContext context, int attr, out Operand value)
{
if (context.Config.Options.TargetApi == TargetApi.Vulkan)
{
if (attr == AttributeConsts.InstanceId)
{
value = context.ISubtract(
context.Load(StorageKind.Input, IoVariable.InstanceIndex),
context.Load(StorageKind.Input, IoVariable.BaseInstance));
return true;
}
else if (attr == AttributeConsts.VertexId)
{
value = context.Load(StorageKind.Input, IoVariable.VertexIndex);
return true;
}
}
value = null;
return false;
}
}
}