/*
*
* if (last.Output._PacketType.Type == GPUPacketType.PM4_DRAW_INDX_2)
* {
* var dword0 = last.Raw.Words[0];
* var indexCount = dword0 >> 16;
* var primitiveType = (GPUPrimitiveType)(dword0 & 0x3F);
*
* ret += "Draw2 " + Enum.GetName(typeof(GPUPrimitiveType), primitiveType) + " (Count: " + indexCount.ToString() + ")";
* }
* else if (last._PacketType.Type == GPUPacketType.PM4_DRAW_INDX)
* {
* var dword0 = last.Raw.Words[0];
* var dword1 = last.Raw.Words[1];
*
* var indexCount = dword1 >> 16;
* var primitiveType = (GPUPrimitiveType)(dword1 & 0x3F);
* var source = (dword1 >> 6) & 0x3;
*
* if (source == 0x0)
* {
* bool is32Bit = 0 != ((dword1 >> 11) & 0x1);
* ret += (is32Bit ? "DrawIndexed32 " : "DrawIndexed16 ") + Enum.GetName(typeof(GPUPrimitiveType), primitiveType) + " (Count: " + indexCount.ToString() + ")";
* }
* else if (source == 0x2)
* {
* ret += "Draw " + Enum.GetName(typeof(GPUPrimitiveType), primitiveType) + " (Count: " + indexCount.ToString() + ")";
* }
* }
* else if (last.Type == GPUPacketType.PM4_HACK_SWAP)
* {
* return "SWAP";
* }
*/
private GPUExecutorResult ExecutePacketType0(RawPacket _packet, GPUCommandOutput _out)
{
var packetData = _packet.Data;
var regCount = ((packetData >> 16) & 0x3FFF) + 1;
var baseIndex = (packetData & 0x7FFF);
var writeOneReg = 0 != ((packetData >> 15) & 0x1);
for (UInt32 i = 0; i < regCount; i++)
{
UInt32 regData = _packet.Words[i];
UInt32 regIndex = writeOneReg ? baseIndex : baseIndex + i;
WriteRegister(regIndex, regData, _out);
}
if (_out != null)
{
if (writeOneReg)
{
_out._Desc = String.Format("SetRegs [Start: {0}, Count: {1}]", baseIndex, regCount);
}
else
{
_out._Desc = String.Format("SetRegs [At: {0}, Times: {1}]", baseIndex, regCount);
}
_out._PacketType = GPUPacketType.PM1_REGS;
}
return(GPUExecutorResult.Executed);
}
private GPUExecutorResult ExecutePacketType2(RawPacket _packet, GPUCommandOutput _out)
{
if (_out != null)
{
_out._PacketType = GPUPacketType.PM3_NOP;
}
return(GPUExecutorResult.Skipped);
}
private void WriteRegister(UInt32 _regIndex, UInt32 _value, GPUCommandOutput _out)
{
_State.RegValues[_regIndex] = _value;
if (_out != null)
{
_out.Log(String.Format("Register {0} ({1}) written with {2} ({3})",
Enum.GetName(typeof(GPURegisterName), _regIndex), _regIndex,
_value, ToFloat(_value)));
}
}
// advance the GPU state by executing the packet - NOTE, some packets don't change the regs
public GPUExecutorResult ExecutePacket(RawPacket _packet, GPUCommandOutput _out)
{
var packetType = _packet.Data >> 30;
switch (packetType)
{
case 0x00: return(ExecutePacketType0(_packet, _out));
case 0x01: return(ExecutePacketType1(_packet, _out));
case 0x02: return(ExecutePacketType2(_packet, _out));
case 0x03: return(ExecutePacketType3(_packet, _out));
}
return(GPUExecutorResult.Invalid);
}
private GPUExecutorResult ExecutePacketType1(RawPacket _packet, GPUCommandOutput _out)
{
var packetData = _packet.Data;
var regIndexA = packetData & 0x7FF;
var regIndexB = (packetData >> 11) & 0x7FF;
var regDataA = _packet.Words[0];
var regDataB = _packet.Words[1];
WriteRegister(regIndexA, regDataA, _out);
WriteRegister(regIndexB, regDataB, _out);
if (_out != null)
{
_out._Desc = String.Format("SetReg2 [RegA: {0}, RegB: {1}]", regIndexA, regIndexB);
_out._PacketType = GPUPacketType.PM2_REGS;
}
return(GPUExecutorResult.Executed);
}
private GPUExecutorResult ExecutePacketType3_EVENT_WRITE_EXT(RawPacket _packet, UInt32 packetData, UInt32 count, GPUCommandOutput _out)
{
return(GPUExecutorResult.Executed);
}
private GPUExecutorResult ExecutePacketType3_DRAW_INDX_2(RawPacket _packet, UInt32 packetData, UInt32 count, GPUCommandOutput _out)
{
// draw using supplied indices in packet
var dword0 = _packet.Words[0];
var indexCount = dword0 >> 16;
var primitiveType = (GPUPrimitiveType)(dword0 & 0x3F);
if (_out != null)
{
_out._Desc = String.Format("Draw: Indexed2 {0} {1}", Enum.GetName(typeof(GPUPrimitiveType), primitiveType), indexCount);
_out._DumpInDrawCall = true;
_out.Log("Draw: Indexed2");
}
var source = (dword0 >> 6) & 0x3;
if (source == 2)
{
DrawInfo ds = new DrawInfo();
ds.IndexCount = indexCount;
ds.PrimitiveType = primitiveType;
ds.IndexData = 0;
ds.IndexFormat = GPUIndexFormat.Index16;
ds.IndexEndianess = GPUEndianFormat.FormatUnspecified;
ds.BaseVertexIndex = 0;
IssueDraw(_packet, ds, _out);
}
return(GPUExecutorResult.Executed);
}
private GPUExecutorResult ExecutePacketType3_EVENT_WRITE(RawPacket _packet, UInt32 packetData, UInt32 count, GPUCommandOutput _out)
{
var initiator = _packet.Words[0];
WriteRegister(GPURegisterName.VGT_EVENT_INITIATOR, initiator & 0x3F, _out);
return(GPUExecutorResult.Executed);
}
private GPUExecutorResult ExecutePacketType3_REG_RMW(RawPacket _packet, UInt32 packetData, UInt32 count, GPUCommandOutput _out)
{
var rmwSetup = _packet.Words[0];
var andMask = _packet.Words[0];
var orMask = _packet.Words[0];
// read value
var regIndex = (rmwSetup & 0x1FFF);
if (_out != null)
{
_out.Log(String.Format("Register: {0} ({1})", Enum.GetName(typeof(GPURegisterName), regIndex), regIndex));
}
var value = _State.RegValues[regIndex];
if (_out != null)
{
_out.Log(String.Format("Incoming value: 0x{0:X}", value));
}
var oldValue = value;
// OR value (with reg or immediate value)
if (0 != ((rmwSetup >> 30) & 0x1))
{
// | reg
if (_out != null)
{
_out.Log(String.Format("Or Mask Register: {0} ({1})", Enum.GetName(typeof(GPURegisterName), orMask & 0x1FFF), orMask & 0x1FFF));
}
var orValue = _State.RegValues[orMask & 0x1FFF];
if (_out != null)
{
_out.Log(String.Format("Or Mask Value: 0x{0:X}", orValue));
}
value |= orValue;
}
else
{
// | imm
if (_out != null)
{
_out.Log(String.Format("Or Mask Value: 0x{0:X}", orMask));
}
value |= orMask;
}
// AND value (with reg or immediate value)
if (0 != ((rmwSetup >> 31) & 0x1))
{
// & reg
if (_out != null)
{
_out.Log(String.Format("And Mask Register: {0} ({1})", Enum.GetName(typeof(GPURegisterName), andMask & 0x1FFF), andMask & 0x1FFF));
}
var andValue = _State.RegValues[andMask & 0x1FFF];
if (_out != null)
{
_out.Log(String.Format("And Mask Value: 0x{0:X}", andValue));
}
value &= andValue;
}
else
{
// & imm
value &= andMask;
if (_out != null)
{
_out.Log(String.Format("And Mask Value: 0x{0:X}", andMask));
}
}
// write new value
if (_out != null)
{
_out.Log(String.Format("Final Value: 0x{0:X}", value));
}
WriteRegister(regIndex, value, _out);
// executed
return(GPUExecutorResult.Executed);
}
private GPUExecutorResult ExecutePacketType3_COND_WRITE(RawPacket _packet, UInt32 packetData, UInt32 count, GPUCommandOutput _out)
{
// conditional write to memory or register
var waitInfo = _packet.Words[0];
var pollRegAddr = _packet.Words[1];
var refValue = _packet.Words[2];
var mask = _packet.Words[3];
var writeRegAddr = _packet.Words[4];
var writeData = _packet.Words[5];
// read input value
UInt32 value = 0;
if (0 != (waitInfo & 0x10))
{
// get swap format (swap)
var readFormat = (pollRegAddr & 3);
var readAddress = (pollRegAddr & ~3);
if (_out != null)
{
_out.Log(String.Format("Read Address: 0x{0:X}", readAddress));
}
if (_out != null)
{
_out.Log(String.Format("Read Mode: {0}", readFormat));
}
// load the value
var memBlock = _packet.FindMemoryReadBlock((UInt32)readAddress);
if (memBlock != null)
{
value = GPUSwap32((UInt32)memBlock.LoadUint32At(0), readFormat);
}
}
else
{
if (_out != null)
{
_out.Log(String.Format("Read Register: {0} ({1})", Enum.GetName(typeof(GPURegisterName), pollRegAddr), pollRegAddr));
}
value = _State.RegValues[pollRegAddr];
}
if (_out != null)
{
_out.Log(String.Format("Incoming Value: 0x{0:X}", value));
_out.Log(String.Format("Incoming Mask: 0x{0:X}", mask));
_out.Log(String.Format("Masked Value: 0x{0:X}", mask & value));
if ((waitInfo & 7) > 0 && ((waitInfo & 7) < 7))
{
_out.Log(String.Format("Reference Value: 0x{0:X}", refValue));
}
}
// compare
bool matched = false;
switch (waitInfo & 0x7)
{
case 0x0:
matched = false;
if (_out != null)
{
_out.Log("Condition: Never");
}
break;
case 0x1: // Less than reference.
matched = (value & mask) < refValue;
if (_out != null)
{
_out.Log(String.Format("Condition: 0x{0:X} < 0x{1:X}", (value & mask), refValue));
}
break;
case 0x2: // Less than or equal to reference.
matched = (value & mask) <= refValue;
if (_out != null)
{
_out.Log(String.Format("Condition: 0x{0:X} <= 0x{1:X}", (value & mask), refValue));
}
break;
case 0x3: // Equal to reference.
matched = (value & mask) == refValue;
if (_out != null)
{
_out.Log(String.Format("Condition: 0x{0:X} == 0x{1:X}", (value & mask), refValue));
}
break;
case 0x4: // Not equal to reference.
matched = (value & mask) != refValue;
if (_out != null)
{
_out.Log(String.Format("Condition: 0x{0:X} != 0x{1:X}", (value & mask), refValue));
}
break;
case 0x5: // Greater than or equal to reference.
matched = (value & mask) >= refValue;
if (_out != null)
{
_out.Log(String.Format("Condition: 0x{0:X} >= 0x{1:X}", (value & mask), refValue));
}
break;
case 0x6: // Greater than reference.
matched = (value & mask) > refValue;
if (_out != null)
{
_out.Log(String.Format("Condition: 0x{0:X} > 0x{1:X}", (value & mask), refValue));
}
break;
case 0x7: // Always
matched = true;
if (_out != null)
{
_out.Log("Condition: Always");
}
break;
}
// write result
_out.Log(matched ? "Matched: YES" : "Matched: NO");
if (matched)
{
if (_out != null)
{
_out.Log(String.Format("Write Data: 0x{0:X}", writeData));
}
// Write.
if (0 != (waitInfo & 0x100))
{
var writeFormat = (writeRegAddr & 3);
var writeAddress = (writeRegAddr & ~3);
if (_out != null)
{
_out.Log(String.Format("Write Address: 0x{0:X}", writeAddress));
}
if (_out != null)
{
_out.Log(String.Format("Write Mode: {0}", writeFormat));
}
}
else
{
WriteRegister(writeRegAddr, writeData, _out);
}
}
return(GPUExecutorResult.Executed);
}
private GPUExecutorResult ExecutePacketType3_HACK_SWAP(RawPacket _packet, UInt32 packetData, UInt32 count, GPUCommandOutput _out)
{
// nothing happens in the replay
return(GPUExecutorResult.Executed);
}
private GPUExecutorResult ExecutePacketType3_WAIT_REG_MEM(RawPacket _packet, UInt32 packetData, UInt32 count, GPUCommandOutput _out)
{
if (_out == null)
{
return(GPUExecutorResult.Executed);
}
var waitInfo = _packet.Words[0];
var pollRegAddr = _packet.Words[1];
var refValue = _packet.Words[2];
var mask = _packet.Words[3];
var wait = _packet.Words[4];
UInt32 incomingValue = 0;
if (0 != (waitInfo & 0x10))
{
var memoryAddress = pollRegAddr & ~0x3;
var memoryMode = pollRegAddr & 3;
_out.Log(String.Format("Memory address: 0x{0:X}", memoryAddress));
_out.Log(String.Format("Memory mode: {0}", memoryMode));
var memBlock = _packet.FindMemoryReadBlock((UInt32)memoryAddress);
if (memBlock != null)
{
incomingValue = memBlock.LoadUint32AtBE(0);
}
}
else
{
_out.Log(String.Format("Read register: {0}", Enum.GetName(typeof(GPURegisterName), pollRegAddr)));
incomingValue = _State.RegValues[pollRegAddr];
}
_out.Log(String.Format("Incoming value: 0x{0:X}", incomingValue));
_out.Log(String.Format("Incoming mask: 0x{0:X}", mask));
_out.Log(String.Format("Masked value: 0x{0:X}", incomingValue & mask));
switch (waitInfo & 0x7)
{
case 0x0: _out.Log("Condition: Never"); break;
case 0x1: _out.Log(String.Format("Condition: 0x{0:X} < 0x{1:X}", incomingValue & mask, refValue)); break;
case 0x2: _out.Log(String.Format("Condition: 0x{0:X} <= 0x{1:X}", incomingValue & mask, refValue)); break;
case 0x3: _out.Log(String.Format("Condition: 0x{0:X} == 0x{1:X}", incomingValue & mask, refValue)); break;
case 0x4: _out.Log(String.Format("Condition: 0x{0:X} != 0x{1:X}", incomingValue & mask, refValue)); break;
case 0x5: _out.Log(String.Format("Condition: 0x{0:X} >= 0x{1:X}", incomingValue & mask, refValue)); break;
case 0x6: _out.Log(String.Format("Condition: 0x{0:X} > 0x{1:X}", incomingValue & mask, refValue)); break;
case 0x7: _out.Log("Condition: Always"); break;
}
return(GPUExecutorResult.Executed);
}
private GPUExecutorResult ExecutePacketType3_SET_CONSTANT(RawPacket _packet, UInt32 packetData, UInt32 count, GPUCommandOutput _out)
{
var offsetType = _packet.Words[0];
var index = offsetType & 0x7FF;
var type = (offsetType >> 16) & 0xFF;
var baseIndex = index;
switch (type)
{
case 0: // ALU
if (_out != null)
{
_out.Log("Block: ALU");
}
baseIndex += 0x4000;
break;
case 1: // FETCH
if (_out != null)
{
_out.Log("Block: FETCH");
}
baseIndex += 0x4800;
break;
case 2: // BOOL
if (_out != null)
{
_out.Log("Block: BOOL");
}
baseIndex += 0x4900;
break;
case 3: // LOOP
if (_out != null)
{
_out.Log("Block: LOOP");
}
baseIndex += 0x4908;
break;
case 4: // REGISTERS
if (_out != null)
{
_out.Log("Block: REGISTERS");
}
baseIndex += 0x2000;
break;
default:
return(GPUExecutorResult.Invalid);
}
// set data
for (UInt32 n = 0; n < (count - 1); n++)
{
UInt32 data = _packet.Words[n + 1];
WriteRegister(index + n, data, _out);
}
return(GPUExecutorResult.Executed);
}
private GPUExecutorResult ExecutePacketType3(RawPacket _packet, GPUCommandOutput _out)
{
var packetData = _packet.Data;
var opcode = (packetData >> 8) & 0x7F;
var count = ((packetData >> 16) & 0x3FFF) + 1;
if (_out != null)
{
_out._PacketType = (GPUPacketType)opcode;
_out._Desc = Enum.GetName(typeof(GPUPacketType), opcode);
}
// in predicated mode we may need to skip this command if the tile mask does not match tile selector
if ((packetData & 1) != 0)
{
// skip if we are NOT the target
bool shouldSkip = (_TileMask & _TileSelector) == 0;
if (shouldSkip)
{
return(GPUExecutorResult.Skipped);
}
}
// process opcode
switch ((GPUPacketType)opcode)
{
case GPUPacketType.PM4_ME_INIT:
return(ExecutePacketType3_ME_INIT(_packet, packetData, count, _out));
case GPUPacketType.PM4_NOP:
return(ExecutePacketType3_NOP(_packet, packetData, count, _out));
case GPUPacketType.PM4_INTERRUPT:
return(ExecutePacketType3_INTERRUPT(_packet, packetData, count, _out));
case GPUPacketType.PM4_HACK_SWAP:
return(ExecutePacketType3_HACK_SWAP(_packet, packetData, count, _out));
case GPUPacketType.PM4_WAIT_REG_MEM:
return(ExecutePacketType3_WAIT_REG_MEM(_packet, packetData, count, _out));
case GPUPacketType.PM4_REG_RMW:
return(ExecutePacketType3_REG_RMW(_packet, packetData, count, _out));
case GPUPacketType.PM4_COND_WRITE:
return(ExecutePacketType3_COND_WRITE(_packet, packetData, count, _out));
case GPUPacketType.PM4_EVENT_WRITE:
return(ExecutePacketType3_EVENT_WRITE(_packet, packetData, count, _out));
case GPUPacketType.PM4_EVENT_WRITE_SHD:
return(ExecutePacketType3_EVENT_WRITE_SHD(_packet, packetData, count, _out));
case GPUPacketType.PM4_EVENT_WRITE_EXT:
return(ExecutePacketType3_EVENT_WRITE_EXT(_packet, packetData, count, _out));
case GPUPacketType.PM4_DRAW_INDX:
return(ExecutePacketType3_DRAW_INDX(_packet, packetData, count, _out));
case GPUPacketType.PM4_DRAW_INDX_2:
return(ExecutePacketType3_DRAW_INDX_2(_packet, packetData, count, _out));
case GPUPacketType.PM4_SET_CONSTANT:
return(ExecutePacketType3_SET_CONSTANT(_packet, packetData, count, _out));
case GPUPacketType.PM4_SET_CONSTANT2:
return(ExecutePacketType3_SET_CONSTANT2(_packet, packetData, count, _out));
case GPUPacketType.PM4_LOAD_ALU_CONSTANT:
return(ExecutePacketType3_LOAD_ALU_CONSTANT(_packet, packetData, count, _out));
case GPUPacketType.PM4_SET_SHADER_CONSTANTS:
return(ExecutePacketType3_SET_SHADER_CONSTANTS(_packet, packetData, count, _out));
case GPUPacketType.PM4_IM_LOAD:
return(ExecutePacketType3_IM_LOAD(_packet, packetData, count, _out));
case GPUPacketType.PM4_IM_LOAD_IMMEDIATE:
return(ExecutePacketType3_IM_LOAD_IMMEDIATE(_packet, packetData, count, _out));
case GPUPacketType.PM4_INVALIDATE_STATE:
return(ExecutePacketType3_INVALIDATE_STATE(_packet, packetData, count, _out));
// tiled rendering bit mask
case GPUPacketType.PM4_SET_BIN_MASK_LO:
{
var loMask = _packet.Words[0];
_TileMask = (_TileMask & 0xFFFFFFFF00000000UL) | ((UInt64)loMask);
_out.Log(String.Format("LoMask = {0:X}", loMask));
return(GPUExecutorResult.Executed);
}
case GPUPacketType.PM4_SET_BIN_MASK_HI:
{
var hiMask = _packet.Words[0];
_TileMask = (_TileMask & 0xFFFFFFFFUL) | (((UInt64)hiMask) << 32);
_out.Log(String.Format("HiMask = {0:X}", hiMask));
return(GPUExecutorResult.Executed);
}
case GPUPacketType.PM4_SET_BIN_SELECT_LO:
{
var loSelect = _packet.Words[0];
_TileSelector = (_TileSelector & 0xFFFFFFFF00000000UL) | ((UInt64)loSelect);
_out.Log(String.Format("LoSelector = {0:X}", loSelect));
return(GPUExecutorResult.Executed);
}
case GPUPacketType.PM4_SET_BIN_SELECT_HI:
{
var hiSelect = _packet.Words[0];
_TileSelector = (_TileSelector & 0xFFFFFFFFUL) | (((UInt64)hiSelect) << 32);
_out.Log(String.Format("HiSelector = {0:X}", hiSelect));
return(GPUExecutorResult.Executed);
}
// Ignored packets - useful if breaking on the default handler below.
case (GPUPacketType)0x50: // 0xC0015000 usually 2 words, 0xFFFFFFFF / 0x00000000
case (GPUPacketType)0x51: // 0xC0015100 usually 2 words, 0xFFFFFFFF / 0xFFFFFFFF
{
return(GPUExecutorResult.Skipped);
}
}
// invalid packet encountered
return(GPUExecutorResult.Invalid);
}
private void IssueDraw(RawPacket _packet, DrawInfo ds, GPUCommandOutput _out)
{
if (_out == null)
{
return;
}
if (ds.IndexData != 0)
{
var indexSize = ((ds.IndexFormat == GPUIndexFormat.Index16) ? 2U : 4U);
_out.Log(String.Format("Index Data Endianess: {0}", Enum.GetName(typeof(GPUEndianFormat), ds.IndexEndianess)));
_out.Log(String.Format("Index Data Address: {0}", MakeMemoryAddr(_packet, ds.IndexData, indexSize, ds.IndexEndianess, ds.IndexFormat)));
}
_out.Log(String.Format("Index Count: {0}", ds.IndexCount));
_out.Log(String.Format("Base Vertex: {0}", ds.BaseVertexIndex));
_out.Log(String.Format("Primitive Type: {0}", Enum.GetName(typeof(GPUPrimitiveType), ds.PrimitiveType)));
// raw shaders
foreach (var memoryRef in _packet.Memory)
{
if (memoryRef.Tag == "PSSourceCode")
{
_out.Log(String.Format("Pixel shader HLSL: <a href=\"hlsl,{0}\">0x{1:X6}</a>", memoryRef.Block.LocalIndex, memoryRef.Block.Adress));
}
else if (memoryRef.Tag == "VSSourceCode")
{
_out.Log(String.Format("Vertex shader HLSL: <a href=\"hlsl,{0}\">0x{1:X6}</a>", memoryRef.Block.LocalIndex, memoryRef.Block.Adress));
}
}
if (_State.VertexShader != null)
{
VFetchMerger vfetchMerger = new VFetchMerger();
// output fetch parameters, merge streams
var fetches = _State.VertexShader.VertexFetches;
foreach (var vfetch in fetches)
{
UInt32 regIndex = ((UInt32)GPURegisterName.SHADER_CONSTANT_FETCH_00_0) + (vfetch.slot * 2);
GPUVertexFetchData fetchReg = new GPUVertexFetchData(_State.RegValues[regIndex + 0], _State.RegValues[regIndex + 1]);
var fetchAddress = (fetchReg.address << 2);
_out.Log(String.Format("VFetch: Address=0x{0:X6} Slot={1} Offset={2} Stride={3} Format={4} DataType={5}",
fetchAddress, vfetch.slot, vfetch.offset, vfetch.stride,
Enum.GetName(typeof(GPUFetchFormat), vfetch.format),
Enum.GetName(typeof(GPUFetchDataType), vfetch.dataType)));
var block = _packet.FindMemoryReadBlock(fetchAddress);
if (block != null)
{
vfetchMerger.AddVFetch(block, vfetch);
}
}
// create vertex data references
foreach (var stream in vfetchMerger.Streams)
{
_out.Log(String.Format("Stream: <a href=\"{0}\">0x{1:X6}, {2} element(s)</a>", stream.MakeLink(), stream.Block.Adress, stream.Elements.Count));
}
}
if (_State.PixelShader != null)
{
var tfetches = _State.PixelShader.TextureFetches;
foreach (var tfetch in tfetches)
{
_out.Log(String.Format("<h2>Pixel Shader Texture {0}</h2>", tfetch.slot));
{
string tfetchText = "";
tfetch.Descibe(ref tfetchText);
_out.Log("<h4>Shader texture:</h4>" + tfetchText);
}
{
var fetchReg = (UInt32)GPURegisterName.SHADER_CONSTANT_FETCH_00_0 + (tfetch.slot * 6);
var data0 = _State.RegValues[fetchReg + 0];
var data1 = _State.RegValues[fetchReg + 1];
var data2 = _State.RegValues[fetchReg + 2];
var data3 = _State.RegValues[fetchReg + 3];
var data4 = _State.RegValues[fetchReg + 4];
var data5 = _State.RegValues[fetchReg + 5];
var boundedTex = new GPUTextureFetchData(data0, data1, data2, data3, data4, data5);
_out._Textures.Add(boundedTex);
string boundexText = "";
boundedTex.Describe(ref boundexText);
_out.Log("<h4>Bounded texture:</h4>" + boundexText);
}
_out.Log("<hr/>");
}
}
}
public static ParsedPacket Parse(RawPacket raw, int packetIndex, ParsedDrawCall drawcall, GPUCommandOutput commandOutput)
{
ParsedPacket ret = new ParsedPacket();
ret._Raw = raw;
ret._Index = packetIndex;
ret._DrawCall = drawcall;
ret._CmdOutput = commandOutput;
if (drawcall != null)
{
drawcall.Linkup(ret);
}
return(ret);
}
private GPUExecutorResult ExecutePacketType3_LOAD_ALU_CONSTANT(RawPacket _packet, UInt32 packetData, UInt32 count, GPUCommandOutput _out)
{
var address = _packet.Words[0] & 0x3FFFFFFF;
var offset_type = _packet.Words[1];
var index = offset_type & 0x7FF;
var baseIndex = index;
var size_dwords = _packet.Words[2] & 0xFFF;
var type = (offset_type >> 16) & 0xFF;
switch (type)
{
case 0: // ALU
index += 0x4000;
break;
case 1: // FETCH
index += 0x4800;
break;
case 2: // BOOL
index += 0x4900;
break;
case 3: // LOOP
index += 0x4908;
break;
case 4: // REGISTERS
index += 0x2000;
break;
}
if (_out != null)
{
_out.Log(String.Format("ALU First Register: {0}", baseIndex));
_out.Log(String.Format("ALU Actual Register: {0}", Enum.GetName(typeof(GPURegisterName), index)));
_out.Log(String.Format("ALU Num Registers: {0}", size_dwords / 4));
_out.Log(String.Format("ALU Data Address: {0}", MakeMemoryAddrALUConst(_packet, address)));
}
var block = _packet.FindMemoryReadBlock(address);
if (block != null)
{
var data = block.LoadAllDataAs32BE();
for (int i = 0; i < (int)size_dwords; ++i)
{
WriteRegister((UInt32)(index + i), data[i], _out);
}
}
return(GPUExecutorResult.Executed);
}
private GPUExecutorResult ExecutePacketType3_INVALIDATE_STATE(RawPacket _packet, UInt32 packetData, UInt32 count, GPUCommandOutput _out)
{
return(GPUExecutorResult.Executed);
}
private GPUExecutorResult ExecutePacketType3_IM_LOAD_IMMEDIATE(RawPacket _packet, UInt32 packetData, UInt32 count, GPUCommandOutput _out)
{
// load sequencer instruction memory (code embedded in packet)
var dword0 = _packet.Words[0];
var dword1 = _packet.Words[1];
var shaderType = (GPUShaderType)(dword0 & 0x3);
var startSize = dword1;
var start = startSize >> 16;
var sizeDwords = startSize & 0xFFFF; // dwords
if (_packet.Memory.Length == 1)
{
var block = _packet.Memory[0].Block;
var pixel = (shaderType == GPUShaderType.ShaderPixel);
if (_out != null)
{
_out.Log("Shader type: " + ((shaderType == GPUShaderType.ShaderPixel) ? "Pixel" : "Vertex"));
_out.Log("Shader words: " + sizeDwords.ToString());
_out.Log(String.Format("Shader data: {0}", MakeShaderAddr(block, pixel)));
_out._DumpInDrawCall = true;
}
if (block != null)
{
if (pixel)
{
_out._Desc = "LoadPixelShaderIM";
_State.SetPixelShader(block);
}
else
{
_out._Desc = "LoadVertexShaderIM";
_State.SetVertexShader(block);
}
}
}
return(GPUExecutorResult.Executed);
}
private GPUExecutorResult ExecutePacketType3_IM_LOAD(RawPacket _packet, UInt32 packetData, UInt32 count, GPUCommandOutput _out)
{
// load sequencer instruction memory (pointer-based)
var addrType = _packet.Words[0];
var shaderType = (GPUShaderType)(addrType & 0x3);
var addr = addrType & ~0x3;
var startSize = _packet.Words[1];
var start = startSize >> 16;
var sizeDwords = startSize & 0xFFFF; // dwords
var block = _packet.FindMemoryReadBlock((UInt32)addr);
var pixel = (shaderType == GPUShaderType.ShaderPixel);
if (_out != null)
{
_out.Log("Shader type: " + ((shaderType == GPUShaderType.ShaderPixel) ? "Pixel" : "Vertex"));
_out.Log("Shader words: " + sizeDwords.ToString());
_out.Log(String.Format("Shader data: {0}", MakeShaderAddr(block, pixel)));
_out._DumpInDrawCall = true;
}
if (block != null)
{
if (pixel)
{
_out._Desc = "LoadPixelShader";
_State.SetPixelShader(block);
}
else
{
_out._Desc = "LoadVertexShader";
_State.SetVertexShader(block);
}
}
return(GPUExecutorResult.Executed);
}
private GPUExecutorResult ExecutePacketType3_SET_SHADER_CONSTANTS(RawPacket _packet, UInt32 packetData, UInt32 count, GPUCommandOutput _out)
{
var offsetType = _packet.Words[0];
var index = offsetType & 0xFFFF;
for (UInt32 n = 0; n < count - 1; n++)
{
var data = _packet.Words[1 + n];
WriteRegister(index + n, data, _out);
}
return(GPUExecutorResult.Executed);
}
private GPUExecutorResult ExecutePacketType3_DRAW_INDX(RawPacket _packet, UInt32 packetData, UInt32 count, GPUCommandOutput _out)
{
var dword0 = _packet.Words[0];
var dword1 = _packet.Words[1];
var indexCount = dword1 >> 16;
var primitiveType = (GPUPrimitiveType)(dword1 & 0x3F);
var source = (dword1 >> 6) & 0x3;
if (source == 0x0)
{
if (_out != null)
{
_out._Desc = String.Format("Draw: Indexed {0} {1}", Enum.GetName(typeof(GPUPrimitiveType), primitiveType), indexCount);
_out._DumpInDrawCall = true;
_out.Log("Draw: Indexed");
}
// extract information
bool is32Bit = 0 != ((dword1 >> 11) & 0x1);
var indexAddr = _packet.Words[2];
var indexSize = _packet.Words[3];
// setup index state
var indexFormat = is32Bit ? GPUIndexFormat.Index32 : GPUIndexFormat.Index16;
var indexEndianess = (GPUEndianFormat)(indexSize >> 30);
// setup draw state
DrawInfo ds = new DrawInfo();
ds.IndexData = indexAddr;
ds.IndexFormat = indexFormat;
ds.IndexEndianess = indexEndianess;
ds.IndexCount = indexCount;
ds.BaseVertexIndex = _State.RegValues[(int)GPURegisterName.VGT_INDX_OFFSET];
ds.PrimitiveType = primitiveType;
IssueDraw(_packet, ds, _out);
}
else if (source == 0x2)
{
if (_out != null)
{
_out._Desc = String.Format("Draw: {0} {1}", Enum.GetName(typeof(GPUPrimitiveType), primitiveType), indexCount);
_out._DumpInDrawCall = true;
_out.Log("Draw: Not Indexed");
}
DrawInfo ds = new DrawInfo();
ds.IndexData = 0;
ds.IndexFormat = GPUIndexFormat.Index16;
ds.IndexEndianess = GPUEndianFormat.FormatUnspecified;
ds.IndexCount = indexCount;
ds.BaseVertexIndex = 0; //_State.RegValues[ GPURegisterName.VGT_INDX_OFFSET ];
ds.PrimitiveType = primitiveType;
IssueDraw(_packet, ds, _out);
}
return(GPUExecutorResult.Executed);
}
static public ParsedData Parse(RawDumpData raw)
{
ParsedData ret = new ParsedData();
var packetsMap = new Dictionary <RawPacket, ParsedPacket>();
// create drawcall list
ret._ShaderCache = new GPUShaderCache();
ret._DrawCalls = new List <ParsedDrawCall>();
ret._DrawGroups = new List <ParsedDrawGroup>();
// prepare GPU state for the duration of parsing
GPUState parseState = new GPUState(ret._ShaderCache);
GPUExecutor parseExec = new GPUExecutor(parseState);
// parse all packets
int packedIndex = 1;
int drawCallIndex = 1;
int drawGroupIndex = 1;
ParsedDrawCall drawCall = null;
ParsedDrawGroup drawGroup = null;
foreach (var rawPacket in raw.Packets)
{
// start new drawcall
if (drawCall == null)
{
drawCall = new ParsedDrawCall(drawCallIndex);
ret._DrawCalls.Add(drawCall);
drawCallIndex += 1;
}
// execute packet
GPUCommandOutput executeOutput = new GPUCommandOutput();
var executionResult = parseExec.ExecutePacket(rawPacket, executeOutput);
if (executionResult != GPUExecutorResult.Invalid)
{
// add data
packetsMap[rawPacket] = ParsedPacket.Parse(rawPacket, packedIndex, drawCall, executeOutput);
++packedIndex;
}
// restart after each drawcall
if (packetsMap[rawPacket].Output.IsDraw())
{
// capture final state at each drawcall
if (drawCall != null)
{
drawCall.CapturedState = new GPUStateCapture(parseState);
// extract the viewport/render target settings for the draw call - required to match it to the draw group
var stateRenderTargets = new GPUStateRenderTargets(drawCall.CapturedState);
var stateViewport = new GPUStateCaptureViewport(drawCall.CapturedState);
// determine if we should add this draw call to current draw group
if ((drawGroup != null) && drawGroup.Compatible(stateViewport, stateRenderTargets))
{
drawGroup.AddDrawCall(drawCall);
}
else
{
drawGroup = new ParsedDrawGroup(drawGroupIndex++, stateViewport, stateRenderTargets);
drawGroup.AddDrawCall(drawCall);
ret._DrawGroups.Add(drawGroup);
}
}
// reset
drawCall = null;
}
}
return(ret);
}
