Implement CAL and RET shader instructions (#1618)

* Add support for CAL and RET shader instructions

* Remove unused stuff

* Fix a bug that could cause the wrong values to be passed to a function

* Avoid repopulating function id dictionary every time

* PR feedback

* Fix vertex shader A/B merge
This commit is contained in:
gdkchan 2020-10-25 17:00:44 -03:00 committed by GitHub
parent 973a615d40
commit 49f970d5bd
No known key found for this signature in database
GPG key ID: 4AEE18F83AFDEB23
33 changed files with 1337 additions and 401 deletions

View file

@ -14,6 +14,16 @@ namespace Ryujinx.Graphics.Shader.Translation
{
private const int HeaderSize = 0x50;
private struct FunctionCode
{
public Operation[] Code { get; }
public FunctionCode(Operation[] code)
{
Code = code;
}
}
public static ShaderProgram Translate(ulong address, IGpuAccessor gpuAccessor, TranslationFlags flags)
{
return Translate(DecodeShader(address, gpuAccessor, flags, out ShaderConfig config), config);
@ -21,32 +31,65 @@ namespace Ryujinx.Graphics.Shader.Translation
public static ShaderProgram Translate(ulong addressA, ulong addressB, IGpuAccessor gpuAccessor, TranslationFlags flags)
{
Operation[] opsA = DecodeShader(addressA, gpuAccessor, flags | TranslationFlags.VertexA, out ShaderConfig configA);
Operation[] opsB = DecodeShader(addressB, gpuAccessor, flags, out ShaderConfig config);
FunctionCode[] funcA = DecodeShader(addressA, gpuAccessor, flags | TranslationFlags.VertexA, out ShaderConfig configA);
FunctionCode[] funcB = DecodeShader(addressB, gpuAccessor, flags, out ShaderConfig config);
config.SetUsedFeature(configA.UsedFeatures);
return Translate(Combine(opsA, opsB), config, configA.Size);
return Translate(Combine(funcA, funcB), config, configA.Size);
}
private static ShaderProgram Translate(Operation[] ops, ShaderConfig config, int sizeA = 0)
private static ShaderProgram Translate(FunctionCode[] functions, ShaderConfig config, int sizeA = 0)
{
BasicBlock[] blocks = ControlFlowGraph.MakeCfg(ops);
var cfgs = new ControlFlowGraph[functions.Length];
var frus = new RegisterUsage.FunctionRegisterUsage[functions.Length];
if (blocks.Length > 0)
for (int i = 0; i < functions.Length; i++)
{
Dominance.FindDominators(blocks[0], blocks.Length);
cfgs[i] = ControlFlowGraph.Create(functions[i].Code);
Dominance.FindDominanceFrontiers(blocks);
Ssa.Rename(blocks);
Optimizer.RunPass(blocks, config);
Lowering.RunPass(blocks, config);
if (i != 0)
{
frus[i] = RegisterUsage.RunPass(cfgs[i]);
}
}
StructuredProgramInfo sInfo = StructuredProgram.MakeStructuredProgram(blocks, config);
Function[] funcs = new Function[functions.Length];
for (int i = 0; i < functions.Length; i++)
{
var cfg = cfgs[i];
int inArgumentsCount = 0;
int outArgumentsCount = 0;
if (i != 0)
{
var fru = frus[i];
inArgumentsCount = fru.InArguments.Length;
outArgumentsCount = fru.OutArguments.Length;
}
if (cfg.Blocks.Length != 0)
{
RegisterUsage.FixupCalls(cfg.Blocks, frus);
Dominance.FindDominators(cfg);
Dominance.FindDominanceFrontiers(cfg.Blocks);
Ssa.Rename(cfg.Blocks);
Optimizer.RunPass(cfg.Blocks, config);
Lowering.RunPass(cfg.Blocks, config);
}
funcs[i] = new Function(cfg.Blocks, $"fun{i}", false, inArgumentsCount, outArgumentsCount);
}
StructuredProgramInfo sInfo = StructuredProgram.MakeStructuredProgram(funcs, config);
GlslProgram program = GlslGenerator.Generate(sInfo, config);
@ -62,9 +105,9 @@ namespace Ryujinx.Graphics.Shader.Translation
return new ShaderProgram(spInfo, config.Stage, glslCode, config.Size, sizeA);
}
private static Operation[] DecodeShader(ulong address, IGpuAccessor gpuAccessor, TranslationFlags flags, out ShaderConfig config)
private static FunctionCode[] DecodeShader(ulong address, IGpuAccessor gpuAccessor, TranslationFlags flags, out ShaderConfig config)
{
Block[] cfg;
Block[][] cfg;
if ((flags & TranslationFlags.Compute) != 0)
{
@ -83,112 +126,131 @@ namespace Ryujinx.Graphics.Shader.Translation
{
gpuAccessor.Log("Invalid branch detected, failed to build CFG.");
return Array.Empty<Operation>();
return Array.Empty<FunctionCode>();
}
EmitterContext context = new EmitterContext(config);
Dictionary<ulong, int> funcIds = new Dictionary<ulong, int>();
for (int funcIndex = 0; funcIndex < cfg.Length; funcIndex++)
{
funcIds.Add(cfg[funcIndex][0].Address, funcIndex);
}
List<FunctionCode> funcs = new List<FunctionCode>();
ulong maxEndAddress = 0;
for (int blkIndex = 0; blkIndex < cfg.Length; blkIndex++)
for (int funcIndex = 0; funcIndex < cfg.Length; funcIndex++)
{
Block block = cfg[blkIndex];
EmitterContext context = new EmitterContext(config, funcIndex != 0, funcIds);
if (maxEndAddress < block.EndAddress)
for (int blkIndex = 0; blkIndex < cfg[funcIndex].Length; blkIndex++)
{
maxEndAddress = block.EndAddress;
Block block = cfg[funcIndex][blkIndex];
if (maxEndAddress < block.EndAddress)
{
maxEndAddress = block.EndAddress;
}
context.CurrBlock = block;
context.MarkLabel(context.GetLabel(block.Address));
EmitOps(context, block);
}
context.CurrBlock = block;
context.MarkLabel(context.GetLabel(block.Address));
for (int opIndex = 0; opIndex < block.OpCodes.Count; opIndex++)
{
OpCode op = block.OpCodes[opIndex];
if ((flags & TranslationFlags.DebugMode) != 0)
{
string instName;
if (op.Emitter != null)
{
instName = op.Emitter.Method.Name;
}
else
{
instName = "???";
gpuAccessor.Log($"Invalid instruction at 0x{op.Address:X6} (0x{op.RawOpCode:X16}).");
}
string dbgComment = $"0x{op.Address:X6}: 0x{op.RawOpCode:X16} {instName}";
context.Add(new CommentNode(dbgComment));
}
if (op.NeverExecute)
{
continue;
}
Operand predSkipLbl = null;
bool skipPredicateCheck = op is OpCodeBranch opBranch && !opBranch.PushTarget;
if (op is OpCodeBranchPop opBranchPop)
{
// If the instruction is a SYNC or BRK instruction with only one
// possible target address, then the instruction is basically
// just a simple branch, we can generate code similar to branch
// instructions, with the condition check on the branch itself.
skipPredicateCheck = opBranchPop.Targets.Count < 2;
}
if (!(op.Predicate.IsPT || skipPredicateCheck))
{
Operand label;
if (opIndex == block.OpCodes.Count - 1 && block.Next != null)
{
label = context.GetLabel(block.Next.Address);
}
else
{
label = Label();
predSkipLbl = label;
}
Operand pred = Register(op.Predicate);
if (op.InvertPredicate)
{
context.BranchIfTrue(label, pred);
}
else
{
context.BranchIfFalse(label, pred);
}
}
context.CurrOp = op;
op.Emitter?.Invoke(context);
if (predSkipLbl != null)
{
context.MarkLabel(predSkipLbl);
}
}
funcs.Add(new FunctionCode(context.GetOperations()));
}
config.SizeAdd((int)maxEndAddress + (flags.HasFlag(TranslationFlags.Compute) ? 0 : HeaderSize));
return context.GetOperations();
return funcs.ToArray();
}
private static Operation[] Combine(Operation[] a, Operation[] b)
internal static void EmitOps(EmitterContext context, Block block)
{
for (int opIndex = 0; opIndex < block.OpCodes.Count; opIndex++)
{
OpCode op = block.OpCodes[opIndex];
if ((context.Config.Flags & TranslationFlags.DebugMode) != 0)
{
string instName;
if (op.Emitter != null)
{
instName = op.Emitter.Method.Name;
}
else
{
instName = "???";
context.Config.GpuAccessor.Log($"Invalid instruction at 0x{op.Address:X6} (0x{op.RawOpCode:X16}).");
}
string dbgComment = $"0x{op.Address:X6}: 0x{op.RawOpCode:X16} {instName}";
context.Add(new CommentNode(dbgComment));
}
if (op.NeverExecute)
{
continue;
}
Operand predSkipLbl = null;
bool skipPredicateCheck = op is OpCodeBranch opBranch && !opBranch.PushTarget;
if (op is OpCodeBranchPop opBranchPop)
{
// If the instruction is a SYNC or BRK instruction with only one
// possible target address, then the instruction is basically
// just a simple branch, we can generate code similar to branch
// instructions, with the condition check on the branch itself.
skipPredicateCheck = opBranchPop.Targets.Count < 2;
}
if (!(op.Predicate.IsPT || skipPredicateCheck))
{
Operand label;
if (opIndex == block.OpCodes.Count - 1 && block.Next != null)
{
label = context.GetLabel(block.Next.Address);
}
else
{
label = Label();
predSkipLbl = label;
}
Operand pred = Register(op.Predicate);
if (op.InvertPredicate)
{
context.BranchIfTrue(label, pred);
}
else
{
context.BranchIfFalse(label, pred);
}
}
context.CurrOp = op;
op.Emitter?.Invoke(context);
if (predSkipLbl != null)
{
context.MarkLabel(predSkipLbl);
}
}
}
private static FunctionCode[] Combine(FunctionCode[] a, FunctionCode[] b)
{
// Here we combine two shaders.
// For shader A:
@ -199,15 +261,17 @@ namespace Ryujinx.Graphics.Shader.Translation
// For shader B:
// - All user attribute loads on shader B are turned into copies from a
// temporary variable, as long that attribute is written by shader A.
List<Operation> output = new List<Operation>(a.Length + b.Length);
FunctionCode[] output = new FunctionCode[a.Length + b.Length - 1];
List<Operation> ops = new List<Operation>(a.Length + b.Length);
Operand[] temps = new Operand[AttributeConsts.UserAttributesCount * 4];
Operand lblB = Label();
for (int index = 0; index < a.Length; index++)
for (int index = 0; index < a[0].Code.Length; index++)
{
Operation operation = a[index];
Operation operation = a[0].Code[index];
if (IsUserAttribute(operation.Dest))
{
@ -227,19 +291,19 @@ namespace Ryujinx.Graphics.Shader.Translation
if (operation.Inst == Instruction.Return)
{
output.Add(new Operation(Instruction.Branch, lblB));
ops.Add(new Operation(Instruction.Branch, lblB));
}
else
{
output.Add(operation);
ops.Add(operation);
}
}
output.Add(new Operation(Instruction.MarkLabel, lblB));
ops.Add(new Operation(Instruction.MarkLabel, lblB));
for (int index = 0; index < b.Length; index++)
for (int index = 0; index < b[0].Code.Length; index++)
{
Operation operation = b[index];
Operation operation = b[0].Code[index];
for (int srcIndex = 0; srcIndex < operation.SourcesCount; srcIndex++)
{
@ -256,10 +320,22 @@ namespace Ryujinx.Graphics.Shader.Translation
}
}
output.Add(operation);
ops.Add(operation);
}
return output.ToArray();
output[0] = new FunctionCode(ops.ToArray());
for (int i = 1; i < a.Length; i++)
{
output[i] = a[i];
}
for (int i = 1; i < b.Length; i++)
{
output[a.Length + i - 1] = b[i];
}
return output;
}
private static bool IsUserAttribute(Operand operand)