mirror of
https://github.com/Ryujinx/Ryujinx.git
synced 2024-10-01 12:30:00 +02:00
6b23a2c125
* Start implementing a new shader translator * Fix shift instructions and a typo * Small refactoring on StructuredProgram, move RemovePhis method to a separate class * Initial geometry shader support * Implement TLD4 * Fix -- There's no negation on FMUL32I * Add constant folding and algebraic simplification optimizations, nits * Some leftovers from constant folding * Avoid cast for constant assignments * Add a branch elimination pass, and misc small fixes * Remove redundant branches, add expression propagation and other improvements on the code * Small leftovers -- add missing break and continue, remove unused properties, other improvements * Add null check to handle empty block cases on block visitor * Add HADD2 and HMUL2 half float shader instructions * Optimize pack/unpack sequences, some fixes related to half float instructions * Add TXQ, TLD, TLDS and TLD4S shader texture instructions, and some support for bindless textures, some refactoring on codegen * Fix copy paste mistake that caused RZ to be ignored on the AST instruction * Add workaround for conditional exit, and fix half float instruction with constant buffer * Add missing 0.0 source for TLDS.LZ variants * Simplify the switch for TLDS.LZ * Texture instructions related fixes * Implement the HFMA instruction, and some misc. fixes * Enable constant folding on UnpackHalf2x16 instructions * Refactor HFMA to use OpCode* for opcode decoding rather than on the helper methods * Remove the old shader translator * Remove ShaderDeclInfo and other unused things * Add dual vertex shader support * Add ShaderConfig, used to pass shader type and maximum cbuffer size * Move and rename some instruction enums * Move texture instructions into a separate file * Move operand GetExpression and locals management to OperandManager * Optimize opcode decoding using a simple list and binary search * Add missing condition for do-while on goto elimination * Misc. fixes on texture instructions * Simplify TLDS switch * Address PR feedback, and a nit
166 lines
No EOL
4.7 KiB
C#
166 lines
No EOL
4.7 KiB
C#
using Ryujinx.Graphics.Gal;
|
|
using Ryujinx.Graphics.Shader.Decoders;
|
|
using System;
|
|
|
|
namespace Ryujinx.Graphics.Shader
|
|
{
|
|
struct OutputMapTarget
|
|
{
|
|
public bool Red { get; }
|
|
public bool Green { get; }
|
|
public bool Blue { get; }
|
|
public bool Alpha { get; }
|
|
|
|
public bool Enabled => Red || Green || Blue || Alpha;
|
|
|
|
public OutputMapTarget(bool red, bool green, bool blue, bool alpha)
|
|
{
|
|
Red = red;
|
|
Green = green;
|
|
Blue = blue;
|
|
Alpha = alpha;
|
|
}
|
|
|
|
public bool ComponentEnabled(int component)
|
|
{
|
|
switch (component)
|
|
{
|
|
case 0: return Red;
|
|
case 1: return Green;
|
|
case 2: return Blue;
|
|
case 3: return Alpha;
|
|
}
|
|
|
|
throw new ArgumentOutOfRangeException(nameof(component));
|
|
}
|
|
}
|
|
|
|
class ShaderHeader
|
|
{
|
|
public int SphType { get; }
|
|
|
|
public int Version { get; }
|
|
|
|
public int ShaderType { get; }
|
|
|
|
public bool MrtEnable { get; }
|
|
|
|
public bool KillsPixels { get; }
|
|
|
|
public bool DoesGlobalStore { get; }
|
|
|
|
public int SassVersion { get; }
|
|
|
|
public bool DoesLoadOrStore { get; }
|
|
|
|
public bool DoesFp64 { get; }
|
|
|
|
public int StreamOutMask{ get; }
|
|
|
|
public int ShaderLocalMemoryLowSize { get; }
|
|
|
|
public int PerPatchAttributeCount { get; }
|
|
|
|
public int ShaderLocalMemoryHighSize { get; }
|
|
|
|
public int ThreadsPerInputPrimitive { get; }
|
|
|
|
public int ShaderLocalMemoryCrsSize { get; }
|
|
|
|
public int OutputTopology { get; }
|
|
|
|
public int MaxOutputVertexCount { get; }
|
|
|
|
public int StoreReqStart { get; }
|
|
public int StoreReqEnd { get; }
|
|
|
|
public OutputMapTarget[] OmapTargets { get; }
|
|
public bool OmapSampleMask { get; }
|
|
public bool OmapDepth { get; }
|
|
|
|
public ShaderHeader(IGalMemory memory, ulong address)
|
|
{
|
|
int commonWord0 = memory.ReadInt32((long)address + 0);
|
|
int commonWord1 = memory.ReadInt32((long)address + 4);
|
|
int commonWord2 = memory.ReadInt32((long)address + 8);
|
|
int commonWord3 = memory.ReadInt32((long)address + 12);
|
|
int commonWord4 = memory.ReadInt32((long)address + 16);
|
|
|
|
SphType = commonWord0.Extract(0, 5);
|
|
|
|
Version = commonWord0.Extract(5, 5);
|
|
|
|
ShaderType = commonWord0.Extract(10, 4);
|
|
|
|
MrtEnable = commonWord0.Extract(14);
|
|
|
|
KillsPixels = commonWord0.Extract(15);
|
|
|
|
DoesGlobalStore = commonWord0.Extract(16);
|
|
|
|
SassVersion = commonWord0.Extract(17, 4);
|
|
|
|
DoesLoadOrStore = commonWord0.Extract(26);
|
|
|
|
DoesFp64 = commonWord0.Extract(27);
|
|
|
|
StreamOutMask = commonWord0.Extract(28, 4);
|
|
|
|
ShaderLocalMemoryLowSize = commonWord1.Extract(0, 24);
|
|
|
|
PerPatchAttributeCount = commonWord1.Extract(24, 8);
|
|
|
|
ShaderLocalMemoryHighSize = commonWord2.Extract(0, 24);
|
|
|
|
ThreadsPerInputPrimitive = commonWord2.Extract(24, 8);
|
|
|
|
ShaderLocalMemoryCrsSize = commonWord3.Extract(0, 24);
|
|
|
|
OutputTopology = commonWord3.Extract(24, 4);
|
|
|
|
MaxOutputVertexCount = commonWord4.Extract(0, 12);
|
|
|
|
StoreReqStart = commonWord4.Extract(12, 8);
|
|
StoreReqEnd = commonWord4.Extract(24, 8);
|
|
|
|
int type2OmapTarget = memory.ReadInt32((long)address + 72);
|
|
int type2Omap = memory.ReadInt32((long)address + 76);
|
|
|
|
OmapTargets = new OutputMapTarget[8];
|
|
|
|
for (int offset = 0; offset < OmapTargets.Length * 4; offset += 4)
|
|
{
|
|
OmapTargets[offset >> 2] = new OutputMapTarget(
|
|
type2OmapTarget.Extract(offset + 0),
|
|
type2OmapTarget.Extract(offset + 1),
|
|
type2OmapTarget.Extract(offset + 2),
|
|
type2OmapTarget.Extract(offset + 3));
|
|
}
|
|
|
|
OmapSampleMask = type2Omap.Extract(0);
|
|
OmapDepth = type2Omap.Extract(1);
|
|
}
|
|
|
|
public int DepthRegister
|
|
{
|
|
get
|
|
{
|
|
int count = 0;
|
|
|
|
for (int index = 0; index < OmapTargets.Length; index++)
|
|
{
|
|
for (int component = 0; component < 4; component++)
|
|
{
|
|
if (OmapTargets[index].ComponentEnabled(component))
|
|
{
|
|
count++;
|
|
}
|
|
}
|
|
}
|
|
|
|
//Depth register is always two registers after the last color output.
|
|
return count + 1;
|
|
}
|
|
}
|
|
}
|
|
} |