0
0
Fork 0
mirror of https://github.com/GreemDev/Ryujinx.git synced 2024-12-27 21:45:46 +00:00
Ryujinx/Ryujinx.Graphics.Shader/CodeGen/Glsl/GlslGenerator.cs

166 lines
5.9 KiB
C#
Raw Normal View History

New shader translator implementation (#654) * 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
2019-04-17 23:57:08 +00:00
using Ryujinx.Graphics.Shader.CodeGen.Glsl.Instructions;
using Ryujinx.Graphics.Shader.IntermediateRepresentation;
using Ryujinx.Graphics.Shader.StructuredIr;
2019-12-29 17:41:50 +00:00
using Ryujinx.Graphics.Shader.Translation;
New shader translator implementation (#654) * 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
2019-04-17 23:57:08 +00:00
using System;
using static Ryujinx.Graphics.Shader.CodeGen.Glsl.TypeConversion;
namespace Ryujinx.Graphics.Shader.CodeGen.Glsl
{
static class GlslGenerator
{
private const string MainFunctionName = "main";
public static string Generate(StructuredProgramInfo info, ShaderConfig config)
New shader translator implementation (#654) * 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
2019-04-17 23:57:08 +00:00
{
CodeGenContext context = new CodeGenContext(info, config);
New shader translator implementation (#654) * 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
2019-04-17 23:57:08 +00:00
Declarations.Declare(context, info);
if (info.Functions.Count != 0)
{
for (int i = 1; i < info.Functions.Count; i++)
{
Vulkan: Add workarounds for MoltenVK (#4202) * Add MVK basics. * Use appropriate output attribute types * 4kb vertex alignment, bunch of fixes * Add reduced shader precision mode for mvk. * Disable ASTC on MVK for now * Only request robustnes2 when it is available. * It's just the one feature actually * Add triangle fan conversion * Allow NullDescriptor on MVK for some reason. * Force safe blit on MoltenVK * Use ASTC only when formats are all available. * Disable multilevel 3d texture views * Filter duplicate render targets (on backend) * Add Automatic MoltenVK Configuration * Do not create color attachment views with formats that are not RT compatible * Make sure that the host format matches the vertex shader input types for invalid/unknown guest formats * FIx rebase for Vertex Attrib State * Fix 4b alignment for vertex * Use asynchronous queue submits for MVK * Ensure color clear shader has correct output type * Update MoltenVK config * Always use MoltenVK workarounds on MacOS * Make MVK supersede all vendors * Fix rebase * Various fixes on rebase * Get portability flags from extension * Fix some minor rebasing issues * Style change * Use LibraryImport for MVKConfiguration * Rename MoltenVK vendor to Apple Intel and AMD GPUs on moltenvk report with the those vendors - only apple silicon reports with vendor 0x106B. * Fix features2 rebase conflict * Rename fragment output type * Add missing check for fragment output types Might have caused the crash in MK8 * Only do fragment output specialization on MoltenVK * Avoid copy when passing capabilities * Self feedback * Address feedback Co-authored-by: gdk <gab.dark.100@gmail.com> Co-authored-by: nastys <nastys@users.noreply.github.com>
2023-01-13 00:31:21 +00:00
context.AppendLine($"{GetFunctionSignature(context, info.Functions[i])};");
}
context.AppendLine();
for (int i = 1; i < info.Functions.Count; i++)
{
PrintFunction(context, info, info.Functions[i]);
context.AppendLine();
}
}
PrintFunction(context, info, info.Functions[0], MainFunctionName);
New shader translator implementation (#654) * 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
2019-04-17 23:57:08 +00:00
return context.GetCode();
New shader translator implementation (#654) * 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
2019-04-17 23:57:08 +00:00
}
private static void PrintFunction(CodeGenContext context, StructuredProgramInfo info, StructuredFunction function, string funcName = null)
New shader translator implementation (#654) * 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
2019-04-17 23:57:08 +00:00
{
context.CurrentFunction = function;
New shader translator implementation (#654) * 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
2019-04-17 23:57:08 +00:00
Vulkan: Add workarounds for MoltenVK (#4202) * Add MVK basics. * Use appropriate output attribute types * 4kb vertex alignment, bunch of fixes * Add reduced shader precision mode for mvk. * Disable ASTC on MVK for now * Only request robustnes2 when it is available. * It's just the one feature actually * Add triangle fan conversion * Allow NullDescriptor on MVK for some reason. * Force safe blit on MoltenVK * Use ASTC only when formats are all available. * Disable multilevel 3d texture views * Filter duplicate render targets (on backend) * Add Automatic MoltenVK Configuration * Do not create color attachment views with formats that are not RT compatible * Make sure that the host format matches the vertex shader input types for invalid/unknown guest formats * FIx rebase for Vertex Attrib State * Fix 4b alignment for vertex * Use asynchronous queue submits for MVK * Ensure color clear shader has correct output type * Update MoltenVK config * Always use MoltenVK workarounds on MacOS * Make MVK supersede all vendors * Fix rebase * Various fixes on rebase * Get portability flags from extension * Fix some minor rebasing issues * Style change * Use LibraryImport for MVKConfiguration * Rename MoltenVK vendor to Apple Intel and AMD GPUs on moltenvk report with the those vendors - only apple silicon reports with vendor 0x106B. * Fix features2 rebase conflict * Rename fragment output type * Add missing check for fragment output types Might have caused the crash in MK8 * Only do fragment output specialization on MoltenVK * Avoid copy when passing capabilities * Self feedback * Address feedback Co-authored-by: gdk <gab.dark.100@gmail.com> Co-authored-by: nastys <nastys@users.noreply.github.com>
2023-01-13 00:31:21 +00:00
context.AppendLine(GetFunctionSignature(context, function, funcName));
New shader translator implementation (#654) * 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
2019-04-17 23:57:08 +00:00
context.EnterScope();
Declarations.DeclareLocals(context, function);
PrintBlock(context, function.MainBlock);
New shader translator implementation (#654) * 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
2019-04-17 23:57:08 +00:00
context.LeaveScope();
}
Vulkan: Add workarounds for MoltenVK (#4202) * Add MVK basics. * Use appropriate output attribute types * 4kb vertex alignment, bunch of fixes * Add reduced shader precision mode for mvk. * Disable ASTC on MVK for now * Only request robustnes2 when it is available. * It's just the one feature actually * Add triangle fan conversion * Allow NullDescriptor on MVK for some reason. * Force safe blit on MoltenVK * Use ASTC only when formats are all available. * Disable multilevel 3d texture views * Filter duplicate render targets (on backend) * Add Automatic MoltenVK Configuration * Do not create color attachment views with formats that are not RT compatible * Make sure that the host format matches the vertex shader input types for invalid/unknown guest formats * FIx rebase for Vertex Attrib State * Fix 4b alignment for vertex * Use asynchronous queue submits for MVK * Ensure color clear shader has correct output type * Update MoltenVK config * Always use MoltenVK workarounds on MacOS * Make MVK supersede all vendors * Fix rebase * Various fixes on rebase * Get portability flags from extension * Fix some minor rebasing issues * Style change * Use LibraryImport for MVKConfiguration * Rename MoltenVK vendor to Apple Intel and AMD GPUs on moltenvk report with the those vendors - only apple silicon reports with vendor 0x106B. * Fix features2 rebase conflict * Rename fragment output type * Add missing check for fragment output types Might have caused the crash in MK8 * Only do fragment output specialization on MoltenVK * Avoid copy when passing capabilities * Self feedback * Address feedback Co-authored-by: gdk <gab.dark.100@gmail.com> Co-authored-by: nastys <nastys@users.noreply.github.com>
2023-01-13 00:31:21 +00:00
private static string GetFunctionSignature(CodeGenContext context, StructuredFunction function, string funcName = null)
{
string[] args = new string[function.InArguments.Length + function.OutArguments.Length];
for (int i = 0; i < function.InArguments.Length; i++)
{
Vulkan: Add workarounds for MoltenVK (#4202) * Add MVK basics. * Use appropriate output attribute types * 4kb vertex alignment, bunch of fixes * Add reduced shader precision mode for mvk. * Disable ASTC on MVK for now * Only request robustnes2 when it is available. * It's just the one feature actually * Add triangle fan conversion * Allow NullDescriptor on MVK for some reason. * Force safe blit on MoltenVK * Use ASTC only when formats are all available. * Disable multilevel 3d texture views * Filter duplicate render targets (on backend) * Add Automatic MoltenVK Configuration * Do not create color attachment views with formats that are not RT compatible * Make sure that the host format matches the vertex shader input types for invalid/unknown guest formats * FIx rebase for Vertex Attrib State * Fix 4b alignment for vertex * Use asynchronous queue submits for MVK * Ensure color clear shader has correct output type * Update MoltenVK config * Always use MoltenVK workarounds on MacOS * Make MVK supersede all vendors * Fix rebase * Various fixes on rebase * Get portability flags from extension * Fix some minor rebasing issues * Style change * Use LibraryImport for MVKConfiguration * Rename MoltenVK vendor to Apple Intel and AMD GPUs on moltenvk report with the those vendors - only apple silicon reports with vendor 0x106B. * Fix features2 rebase conflict * Rename fragment output type * Add missing check for fragment output types Might have caused the crash in MK8 * Only do fragment output specialization on MoltenVK * Avoid copy when passing capabilities * Self feedback * Address feedback Co-authored-by: gdk <gab.dark.100@gmail.com> Co-authored-by: nastys <nastys@users.noreply.github.com>
2023-01-13 00:31:21 +00:00
args[i] = $"{Declarations.GetVarTypeName(context, function.InArguments[i])} {OperandManager.GetArgumentName(i)}";
}
for (int i = 0; i < function.OutArguments.Length; i++)
{
int j = i + function.InArguments.Length;
Vulkan: Add workarounds for MoltenVK (#4202) * Add MVK basics. * Use appropriate output attribute types * 4kb vertex alignment, bunch of fixes * Add reduced shader precision mode for mvk. * Disable ASTC on MVK for now * Only request robustnes2 when it is available. * It's just the one feature actually * Add triangle fan conversion * Allow NullDescriptor on MVK for some reason. * Force safe blit on MoltenVK * Use ASTC only when formats are all available. * Disable multilevel 3d texture views * Filter duplicate render targets (on backend) * Add Automatic MoltenVK Configuration * Do not create color attachment views with formats that are not RT compatible * Make sure that the host format matches the vertex shader input types for invalid/unknown guest formats * FIx rebase for Vertex Attrib State * Fix 4b alignment for vertex * Use asynchronous queue submits for MVK * Ensure color clear shader has correct output type * Update MoltenVK config * Always use MoltenVK workarounds on MacOS * Make MVK supersede all vendors * Fix rebase * Various fixes on rebase * Get portability flags from extension * Fix some minor rebasing issues * Style change * Use LibraryImport for MVKConfiguration * Rename MoltenVK vendor to Apple Intel and AMD GPUs on moltenvk report with the those vendors - only apple silicon reports with vendor 0x106B. * Fix features2 rebase conflict * Rename fragment output type * Add missing check for fragment output types Might have caused the crash in MK8 * Only do fragment output specialization on MoltenVK * Avoid copy when passing capabilities * Self feedback * Address feedback Co-authored-by: gdk <gab.dark.100@gmail.com> Co-authored-by: nastys <nastys@users.noreply.github.com>
2023-01-13 00:31:21 +00:00
args[j] = $"out {Declarations.GetVarTypeName(context, function.OutArguments[i])} {OperandManager.GetArgumentName(j)}";
}
Vulkan: Add workarounds for MoltenVK (#4202) * Add MVK basics. * Use appropriate output attribute types * 4kb vertex alignment, bunch of fixes * Add reduced shader precision mode for mvk. * Disable ASTC on MVK for now * Only request robustnes2 when it is available. * It's just the one feature actually * Add triangle fan conversion * Allow NullDescriptor on MVK for some reason. * Force safe blit on MoltenVK * Use ASTC only when formats are all available. * Disable multilevel 3d texture views * Filter duplicate render targets (on backend) * Add Automatic MoltenVK Configuration * Do not create color attachment views with formats that are not RT compatible * Make sure that the host format matches the vertex shader input types for invalid/unknown guest formats * FIx rebase for Vertex Attrib State * Fix 4b alignment for vertex * Use asynchronous queue submits for MVK * Ensure color clear shader has correct output type * Update MoltenVK config * Always use MoltenVK workarounds on MacOS * Make MVK supersede all vendors * Fix rebase * Various fixes on rebase * Get portability flags from extension * Fix some minor rebasing issues * Style change * Use LibraryImport for MVKConfiguration * Rename MoltenVK vendor to Apple Intel and AMD GPUs on moltenvk report with the those vendors - only apple silicon reports with vendor 0x106B. * Fix features2 rebase conflict * Rename fragment output type * Add missing check for fragment output types Might have caused the crash in MK8 * Only do fragment output specialization on MoltenVK * Avoid copy when passing capabilities * Self feedback * Address feedback Co-authored-by: gdk <gab.dark.100@gmail.com> Co-authored-by: nastys <nastys@users.noreply.github.com>
2023-01-13 00:31:21 +00:00
return $"{Declarations.GetVarTypeName(context, function.ReturnType)} {funcName ?? function.Name}({string.Join(", ", args)})";
}
New shader translator implementation (#654) * 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
2019-04-17 23:57:08 +00:00
private static void PrintBlock(CodeGenContext context, AstBlock block)
{
AstBlockVisitor visitor = new AstBlockVisitor(block);
visitor.BlockEntered += (sender, e) =>
{
switch (e.Block.Type)
{
case AstBlockType.DoWhile:
context.AppendLine("do");
break;
case AstBlockType.Else:
context.AppendLine("else");
break;
case AstBlockType.ElseIf:
context.AppendLine($"else if ({GetCondExpr(context, e.Block.Condition)})");
break;
case AstBlockType.If:
context.AppendLine($"if ({GetCondExpr(context, e.Block.Condition)})");
break;
default: throw new InvalidOperationException($"Found unexpected block type \"{e.Block.Type}\".");
}
context.EnterScope();
};
visitor.BlockLeft += (sender, e) =>
{
context.LeaveScope();
if (e.Block.Type == AstBlockType.DoWhile)
{
context.AppendLine($"while ({GetCondExpr(context, e.Block.Condition)});");
}
};
foreach (IAstNode node in visitor.Visit())
{
if (node is AstOperation operation)
{
string expr = InstGen.GetExpression(context, operation);
if (expr != null)
{
context.AppendLine(expr + ";");
}
New shader translator implementation (#654) * 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
2019-04-17 23:57:08 +00:00
}
else if (node is AstAssignment assignment)
{
AggregateType srcType = OperandManager.GetNodeDestType(context, assignment.Source);
AggregateType dstType = OperandManager.GetNodeDestType(context, assignment.Destination, isAsgDest: true);
New shader translator implementation (#654) * 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
2019-04-17 23:57:08 +00:00
string dest;
if (assignment.Destination is AstOperand operand && operand.Type.IsAttribute())
New shader translator implementation (#654) * 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
2019-04-17 23:57:08 +00:00
{
bool perPatch = operand.Type == OperandType.AttributePerPatch;
dest = OperandManager.GetOutAttributeName(context, operand.Value, perPatch);
New shader translator implementation (#654) * 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
2019-04-17 23:57:08 +00:00
}
else
{
dest = InstGen.GetExpression(context, assignment.Destination);
}
string src = ReinterpretCast(context, assignment.Source, srcType, dstType);
context.AppendLine(dest + " = " + src + ";");
}
2019-10-13 06:02:07 +00:00
else if (node is AstComment comment)
{
context.AppendLine("// " + comment.Comment);
}
New shader translator implementation (#654) * 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
2019-04-17 23:57:08 +00:00
else
{
throw new InvalidOperationException($"Found unexpected node type \"{node?.GetType().Name ?? "null"}\".");
}
}
}
private static string GetCondExpr(CodeGenContext context, IAstNode cond)
{
AggregateType srcType = OperandManager.GetNodeDestType(context, cond);
New shader translator implementation (#654) * 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
2019-04-17 23:57:08 +00:00
return ReinterpretCast(context, cond, srcType, AggregateType.Bool);
New shader translator implementation (#654) * 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
2019-04-17 23:57:08 +00:00
}
}
}