From 31eb658fea6ecae8fb4f0dde5abdeb1d8eb44ae0 Mon Sep 17 00:00:00 2001 From: ReinUsesLisp Date: Tue, 19 May 2020 22:01:25 -0300 Subject: maxwell_3d: Initialize polygon modes NVN expects this to be initialized as Fill, otherwise games that never bind a rasterizer state will log an invalid polygon mode. --- src/video_core/engines/maxwell_3d.cpp | 2 ++ 1 file changed, 2 insertions(+) (limited to 'src/video_core/engines') diff --git a/src/video_core/engines/maxwell_3d.cpp b/src/video_core/engines/maxwell_3d.cpp index 024c9e43b..6113da7f8 100644 --- a/src/video_core/engines/maxwell_3d.cpp +++ b/src/video_core/engines/maxwell_3d.cpp @@ -107,6 +107,8 @@ void Maxwell3D::InitializeRegisterDefaults() { regs.rt_separate_frag_data = 1; regs.framebuffer_srgb = 1; regs.front_face = Maxwell3D::Regs::FrontFace::ClockWise; + regs.polygon_mode_back = Maxwell3D::Regs::PolygonMode::Fill; + regs.polygon_mode_front = Maxwell3D::Regs::PolygonMode::Fill; shadow_state = regs; -- cgit v1.2.3 From f3f056c3b60fda242ae08f3866832dd307d537ba Mon Sep 17 00:00:00 2001 From: ReinUsesLisp Date: Wed, 20 May 2020 01:13:40 -0300 Subject: maxwell_3d: Initialize line widths Initialize line widths to avoid setting a line width of zero. --- src/video_core/engines/maxwell_3d.cpp | 2 ++ 1 file changed, 2 insertions(+) (limited to 'src/video_core/engines') diff --git a/src/video_core/engines/maxwell_3d.cpp b/src/video_core/engines/maxwell_3d.cpp index 6113da7f8..b5a70b9fc 100644 --- a/src/video_core/engines/maxwell_3d.cpp +++ b/src/video_core/engines/maxwell_3d.cpp @@ -106,6 +106,8 @@ void Maxwell3D::InitializeRegisterDefaults() { regs.rasterize_enable = 1; regs.rt_separate_frag_data = 1; regs.framebuffer_srgb = 1; + regs.line_width_aliased = 1.0f; + regs.line_width_smooth = 1.0f; regs.front_face = Maxwell3D::Regs::FrontFace::ClockWise; regs.polygon_mode_back = Maxwell3D::Regs::PolygonMode::Fill; regs.polygon_mode_front = Maxwell3D::Regs::PolygonMode::Fill; -- cgit v1.2.3 From b032ebdfee1928c4458eaf15faa0cff299371e65 Mon Sep 17 00:00:00 2001 From: David Marcec Date: Fri, 29 May 2020 14:53:27 +1000 Subject: Implement macro JIT --- src/video_core/engines/maxwell_3d.cpp | 19 +++++++------------ src/video_core/engines/maxwell_3d.h | 19 ++++--------------- 2 files changed, 11 insertions(+), 27 deletions(-) (limited to 'src/video_core/engines') diff --git a/src/video_core/engines/maxwell_3d.cpp b/src/video_core/engines/maxwell_3d.cpp index 004f6b261..934a1d6db 100644 --- a/src/video_core/engines/maxwell_3d.cpp +++ b/src/video_core/engines/maxwell_3d.cpp @@ -25,9 +25,8 @@ constexpr u32 MacroRegistersStart = 0xE00; Maxwell3D::Maxwell3D(Core::System& system, VideoCore::RasterizerInterface& rasterizer, MemoryManager& memory_manager) : system{system}, rasterizer{rasterizer}, memory_manager{memory_manager}, - macro_interpreter{*this}, upload_state{memory_manager, regs.upload} { + macro_engine(GetMacroEngine(*this)), upload_state{memory_manager, regs.upload} { dirty.flags.flip(); - InitializeRegisterDefaults(); } @@ -116,7 +115,7 @@ void Maxwell3D::InitializeRegisterDefaults() { mme_inline[MAXWELL3D_REG_INDEX(index_array.count)] = true; } -void Maxwell3D::CallMacroMethod(u32 method, std::size_t num_parameters, const u32* parameters) { +void Maxwell3D::CallMacroMethod(u32 method, std::vector&& parameters) { // Reset the current macro. executing_macro = 0; @@ -125,7 +124,7 @@ void Maxwell3D::CallMacroMethod(u32 method, std::size_t num_parameters, const u3 ((method - MacroRegistersStart) >> 1) % static_cast(macro_positions.size()); // Execute the current macro. - macro_interpreter.Execute(macro_positions[entry], num_parameters, parameters); + macro_engine->Execute(macro_positions[entry], std::move(parameters)); if (mme_draw.current_mode != MMEDrawMode::Undefined) { FlushMMEInlineDraw(); } @@ -161,8 +160,7 @@ void Maxwell3D::CallMethod(u32 method, u32 method_argument, bool is_last_call) { // Call the macro when there are no more parameters in the command buffer if (is_last_call) { - CallMacroMethod(executing_macro, macro_params.size(), macro_params.data()); - macro_params.clear(); + CallMacroMethod(executing_macro, std::move(macro_params)); } return; } @@ -197,7 +195,7 @@ void Maxwell3D::CallMethod(u32 method, u32 method_argument, bool is_last_call) { break; } case MAXWELL3D_REG_INDEX(macros.data): { - ProcessMacroUpload(arg); + macro_engine->AddCode(regs.macros.upload_address, arg); break; } case MAXWELL3D_REG_INDEX(macros.bind): { @@ -306,8 +304,7 @@ void Maxwell3D::CallMultiMethod(u32 method, const u32* base_start, u32 amount, // Call the macro when there are no more parameters in the command buffer if (amount == methods_pending) { - CallMacroMethod(executing_macro, macro_params.size(), macro_params.data()); - macro_params.clear(); + CallMacroMethod(executing_macro, std::move(macro_params)); } return; } @@ -420,9 +417,7 @@ void Maxwell3D::FlushMMEInlineDraw() { } void Maxwell3D::ProcessMacroUpload(u32 data) { - ASSERT_MSG(regs.macros.upload_address < macro_memory.size(), - "upload_address exceeded macro_memory size!"); - macro_memory[regs.macros.upload_address++] = data; + macro_engine->AddCode(regs.macros.upload_address++, data); } void Maxwell3D::ProcessMacroBind(u32 data) { diff --git a/src/video_core/engines/maxwell_3d.h b/src/video_core/engines/maxwell_3d.h index 05dd6b39b..077bc9841 100644 --- a/src/video_core/engines/maxwell_3d.h +++ b/src/video_core/engines/maxwell_3d.h @@ -23,7 +23,7 @@ #include "video_core/engines/engine_upload.h" #include "video_core/engines/shader_type.h" #include "video_core/gpu.h" -#include "video_core/macro_interpreter.h" +#include "video_core/macro/macro.h" #include "video_core/textures/texture.h" namespace Core { @@ -1411,15 +1411,6 @@ public: const VideoCore::GuestDriverProfile& AccessGuestDriverProfile() const override; - /// Memory for macro code - it's undetermined how big this is, however 1MB is much larger than - /// we've seen used. - using MacroMemory = std::array; - - /// Gets a reference to macro memory. - const MacroMemory& GetMacroMemory() const { - return macro_memory; - } - bool ShouldExecute() const { return execute_on; } @@ -1468,16 +1459,14 @@ private: std::array mme_inline{}; - /// Memory for macro code - MacroMemory macro_memory; - /// Macro method that is currently being executed / being fed parameters. u32 executing_macro = 0; /// Parameters that have been submitted to the macro call so far. std::vector macro_params; /// Interpreter for the macro codes uploaded to the GPU. - MacroInterpreter macro_interpreter; + std::unique_ptr macro_engine; + // MacroInterpreter macro_interpreter; static constexpr u32 null_cb_data = 0xFFFFFFFF; struct { @@ -1506,7 +1495,7 @@ private: * @param num_parameters Number of arguments * @param parameters Arguments to the method call */ - void CallMacroMethod(u32 method, std::size_t num_parameters, const u32* parameters); + void CallMacroMethod(u32 method, std::vector&& parameters); /// Handles writes to the macro uploading register. void ProcessMacroUpload(u32 data); -- cgit v1.2.3 From 3a20e74f4056a03253e563510048e99ea548d5b4 Mon Sep 17 00:00:00 2001 From: David Marcec Date: Tue, 2 Jun 2020 16:37:06 +1000 Subject: Pass by reference instead of copying parameters --- src/video_core/engines/maxwell_3d.cpp | 10 ++++++---- src/video_core/engines/maxwell_3d.h | 2 +- 2 files changed, 7 insertions(+), 5 deletions(-) (limited to 'src/video_core/engines') diff --git a/src/video_core/engines/maxwell_3d.cpp b/src/video_core/engines/maxwell_3d.cpp index 934a1d6db..0c0a7c1ed 100644 --- a/src/video_core/engines/maxwell_3d.cpp +++ b/src/video_core/engines/maxwell_3d.cpp @@ -115,7 +115,7 @@ void Maxwell3D::InitializeRegisterDefaults() { mme_inline[MAXWELL3D_REG_INDEX(index_array.count)] = true; } -void Maxwell3D::CallMacroMethod(u32 method, std::vector&& parameters) { +void Maxwell3D::CallMacroMethod(u32 method, std::vector& parameters) { // Reset the current macro. executing_macro = 0; @@ -124,7 +124,7 @@ void Maxwell3D::CallMacroMethod(u32 method, std::vector&& parameters) { ((method - MacroRegistersStart) >> 1) % static_cast(macro_positions.size()); // Execute the current macro. - macro_engine->Execute(macro_positions[entry], std::move(parameters)); + macro_engine->Execute(macro_positions[entry], parameters); if (mme_draw.current_mode != MMEDrawMode::Undefined) { FlushMMEInlineDraw(); } @@ -160,7 +160,8 @@ void Maxwell3D::CallMethod(u32 method, u32 method_argument, bool is_last_call) { // Call the macro when there are no more parameters in the command buffer if (is_last_call) { - CallMacroMethod(executing_macro, std::move(macro_params)); + CallMacroMethod(executing_macro, macro_params); + macro_params.clear(); } return; } @@ -304,7 +305,8 @@ void Maxwell3D::CallMultiMethod(u32 method, const u32* base_start, u32 amount, // Call the macro when there are no more parameters in the command buffer if (amount == methods_pending) { - CallMacroMethod(executing_macro, std::move(macro_params)); + CallMacroMethod(executing_macro, macro_params); + macro_params.clear(); } return; } diff --git a/src/video_core/engines/maxwell_3d.h b/src/video_core/engines/maxwell_3d.h index 077bc9841..651f37b4d 100644 --- a/src/video_core/engines/maxwell_3d.h +++ b/src/video_core/engines/maxwell_3d.h @@ -1495,7 +1495,7 @@ private: * @param num_parameters Number of arguments * @param parameters Arguments to the method call */ - void CallMacroMethod(u32 method, std::vector&& parameters); + void CallMacroMethod(u32 method, std::vector& parameters); /// Handles writes to the macro uploading register. void ProcessMacroUpload(u32 data); -- cgit v1.2.3 From 411f5527d41ba5c4f09b914b4fb4df0c6493f744 Mon Sep 17 00:00:00 2001 From: David Marcec Date: Wed, 3 Jun 2020 16:33:38 +1000 Subject: Mark parameters as const --- src/video_core/engines/maxwell_3d.cpp | 2 +- src/video_core/engines/maxwell_3d.h | 3 +-- 2 files changed, 2 insertions(+), 3 deletions(-) (limited to 'src/video_core/engines') diff --git a/src/video_core/engines/maxwell_3d.cpp b/src/video_core/engines/maxwell_3d.cpp index 0c0a7c1ed..14ad40250 100644 --- a/src/video_core/engines/maxwell_3d.cpp +++ b/src/video_core/engines/maxwell_3d.cpp @@ -115,7 +115,7 @@ void Maxwell3D::InitializeRegisterDefaults() { mme_inline[MAXWELL3D_REG_INDEX(index_array.count)] = true; } -void Maxwell3D::CallMacroMethod(u32 method, std::vector& parameters) { +void Maxwell3D::CallMacroMethod(u32 method, const std::vector& parameters) { // Reset the current macro. executing_macro = 0; diff --git a/src/video_core/engines/maxwell_3d.h b/src/video_core/engines/maxwell_3d.h index 651f37b4d..b827b112f 100644 --- a/src/video_core/engines/maxwell_3d.h +++ b/src/video_core/engines/maxwell_3d.h @@ -1466,7 +1466,6 @@ private: /// Interpreter for the macro codes uploaded to the GPU. std::unique_ptr macro_engine; - // MacroInterpreter macro_interpreter; static constexpr u32 null_cb_data = 0xFFFFFFFF; struct { @@ -1495,7 +1494,7 @@ private: * @param num_parameters Number of arguments * @param parameters Arguments to the method call */ - void CallMacroMethod(u32 method, std::vector& parameters); + void CallMacroMethod(u32 method, const std::vector& parameters); /// Handles writes to the macro uploading register. void ProcessMacroUpload(u32 data); -- cgit v1.2.3 From eca3d16e548451f403cc4cc2e9fb38ed96693102 Mon Sep 17 00:00:00 2001 From: David Marcec Date: Thu, 4 Jun 2020 22:23:07 +1000 Subject: Default init labels and use initializer list for macro engine --- src/video_core/engines/maxwell_3d.cpp | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'src/video_core/engines') diff --git a/src/video_core/engines/maxwell_3d.cpp b/src/video_core/engines/maxwell_3d.cpp index 14ad40250..951016c3e 100644 --- a/src/video_core/engines/maxwell_3d.cpp +++ b/src/video_core/engines/maxwell_3d.cpp @@ -25,7 +25,7 @@ constexpr u32 MacroRegistersStart = 0xE00; Maxwell3D::Maxwell3D(Core::System& system, VideoCore::RasterizerInterface& rasterizer, MemoryManager& memory_manager) : system{system}, rasterizer{rasterizer}, memory_manager{memory_manager}, - macro_engine(GetMacroEngine(*this)), upload_state{memory_manager, regs.upload} { + macro_engine{GetMacroEngine(*this)}, upload_state{memory_manager, regs.upload} { dirty.flags.flip(); InitializeRegisterDefaults(); } -- cgit v1.2.3 From 5b2b6d594c6cfa77c3fb92faee63ad524bfe7204 Mon Sep 17 00:00:00 2001 From: ReinUsesLisp Date: Thu, 4 Jun 2020 23:03:49 -0300 Subject: shader/texture: Join separate image and sampler pairs offline Games using D3D idioms can join images and samplers when a shader executes, instead of baking them into a combined sampler image. This is also possible on Vulkan. One approach to this solution would be to use separate samplers on Vulkan and leave this unimplemented on OpenGL, but we can't do this because there's no consistent way of determining which constant buffer holds a sampler and which one an image. We could in theory find the first bit and if it's in the TIC area, it's an image; but this falls apart when an image or sampler handle use an index of zero. The used approach is to track for a LOP.OR operation (this is done at an IR level, not at an ISA level), track again the constant buffers used as source and store this pair. Then, outside of shader execution, join the sample and image pair with a bitwise or operation. This approach won't work on games that truly use separate samplers in a meaningful way. For example, pooling textures in a 2D array and determining at runtime what sampler to use. This invalidates OpenGL's disk shader cache :) - Used mostly by D3D ports to Switch --- src/video_core/engines/const_buffer_engine_interface.h | 1 + src/video_core/engines/kepler_compute.cpp | 5 ++++- src/video_core/engines/kepler_compute.h | 2 ++ src/video_core/engines/maxwell_3d.cpp | 5 ++++- src/video_core/engines/maxwell_3d.h | 2 ++ 5 files changed, 13 insertions(+), 2 deletions(-) (limited to 'src/video_core/engines') diff --git a/src/video_core/engines/const_buffer_engine_interface.h b/src/video_core/engines/const_buffer_engine_interface.h index ebe139504..f46e81bb7 100644 --- a/src/video_core/engines/const_buffer_engine_interface.h +++ b/src/video_core/engines/const_buffer_engine_interface.h @@ -93,6 +93,7 @@ public: virtual SamplerDescriptor AccessBoundSampler(ShaderType stage, u64 offset) const = 0; virtual SamplerDescriptor AccessBindlessSampler(ShaderType stage, u64 const_buffer, u64 offset) const = 0; + virtual SamplerDescriptor AccessSampler(u32 handle) const = 0; virtual u32 GetBoundBuffer() const = 0; virtual VideoCore::GuestDriverProfile& AccessGuestDriverProfile() = 0; diff --git a/src/video_core/engines/kepler_compute.cpp b/src/video_core/engines/kepler_compute.cpp index f6237fc6a..a82b06a38 100644 --- a/src/video_core/engines/kepler_compute.cpp +++ b/src/video_core/engines/kepler_compute.cpp @@ -92,8 +92,11 @@ SamplerDescriptor KeplerCompute::AccessBindlessSampler(ShaderType stage, u64 con ASSERT(stage == ShaderType::Compute); const auto& tex_info_buffer = launch_description.const_buffer_config[const_buffer]; const GPUVAddr tex_info_address = tex_info_buffer.Address() + offset; + return AccessSampler(memory_manager.Read(tex_info_address)); +} - const Texture::TextureHandle tex_handle{memory_manager.Read(tex_info_address)}; +SamplerDescriptor KeplerCompute::AccessSampler(u32 handle) const { + const Texture::TextureHandle tex_handle{handle}; const Texture::FullTextureInfo tex_info = GetTextureInfo(tex_handle); SamplerDescriptor result = SamplerDescriptor::FromTIC(tex_info.tic); result.is_shadow.Assign(tex_info.tsc.depth_compare_enabled.Value()); diff --git a/src/video_core/engines/kepler_compute.h b/src/video_core/engines/kepler_compute.h index 18ceedfaf..b7f668d88 100644 --- a/src/video_core/engines/kepler_compute.h +++ b/src/video_core/engines/kepler_compute.h @@ -219,6 +219,8 @@ public: SamplerDescriptor AccessBindlessSampler(ShaderType stage, u64 const_buffer, u64 offset) const override; + SamplerDescriptor AccessSampler(u32 handle) const override; + u32 GetBoundBuffer() const override { return regs.tex_cb_index; } diff --git a/src/video_core/engines/maxwell_3d.cpp b/src/video_core/engines/maxwell_3d.cpp index 13ef2e42d..d539ae7b1 100644 --- a/src/video_core/engines/maxwell_3d.cpp +++ b/src/video_core/engines/maxwell_3d.cpp @@ -743,8 +743,11 @@ SamplerDescriptor Maxwell3D::AccessBindlessSampler(ShaderType stage, u64 const_b const auto& shader = state.shader_stages[static_cast(stage)]; const auto& tex_info_buffer = shader.const_buffers[const_buffer]; const GPUVAddr tex_info_address = tex_info_buffer.address + offset; + return AccessSampler(memory_manager.Read(tex_info_address)); +} - const Texture::TextureHandle tex_handle{memory_manager.Read(tex_info_address)}; +SamplerDescriptor Maxwell3D::AccessSampler(u32 handle) const { + const Texture::TextureHandle tex_handle{handle}; const Texture::FullTextureInfo tex_info = GetTextureInfo(tex_handle); SamplerDescriptor result = SamplerDescriptor::FromTIC(tex_info.tic); result.is_shadow.Assign(tex_info.tsc.depth_compare_enabled.Value()); diff --git a/src/video_core/engines/maxwell_3d.h b/src/video_core/engines/maxwell_3d.h index 05dd6b39b..c5f1b0499 100644 --- a/src/video_core/engines/maxwell_3d.h +++ b/src/video_core/engines/maxwell_3d.h @@ -1403,6 +1403,8 @@ public: SamplerDescriptor AccessBindlessSampler(ShaderType stage, u64 const_buffer, u64 offset) const override; + SamplerDescriptor AccessSampler(u32 handle) const override; + u32 GetBoundBuffer() const override { return regs.tex_cb_index; } -- cgit v1.2.3 From c95c254f3eda75476ad221a4828033f4140a3470 Mon Sep 17 00:00:00 2001 From: ReinUsesLisp Date: Fri, 29 May 2020 23:32:41 -0300 Subject: texture_cache: Implement rendering to 3D textures This allows rendering to 3D textures with more than one slice. Applications are allowed to render to more than one slice of a texture using gl_Layer from a VTG shader. This also requires reworking how 3D texture collisions are handled, for now, this commit allows rendering to slices but not to miplevels. When a render target attempts to write to a mipmap, we fallback to the previous implementation (copying or flushing as needed). - Fixes color correction 3D textures on UE4 games (rainbow effects). - Allows Xenoblade games to render to 3D textures directly. --- src/video_core/engines/maxwell_3d.h | 1 + 1 file changed, 1 insertion(+) (limited to 'src/video_core/engines') diff --git a/src/video_core/engines/maxwell_3d.h b/src/video_core/engines/maxwell_3d.h index b827b112f..79fc9bbea 100644 --- a/src/video_core/engines/maxwell_3d.h +++ b/src/video_core/engines/maxwell_3d.h @@ -598,6 +598,7 @@ public: BitField<4, 3, u32> block_height; BitField<8, 3, u32> block_depth; BitField<12, 1, InvMemoryLayout> type; + BitField<16, 1, u32> is_3d; } memory_layout; union { BitField<0, 16, u32> layers; -- cgit v1.2.3 From 6ce5f3120be6a65a798d3abc6fda0fe6171d0296 Mon Sep 17 00:00:00 2001 From: David Marcec Date: Fri, 5 Jun 2020 01:42:19 +1000 Subject: Macro HLE support --- src/video_core/engines/maxwell_3d.cpp | 2 +- src/video_core/engines/maxwell_3d.h | 4 ++++ 2 files changed, 5 insertions(+), 1 deletion(-) (limited to 'src/video_core/engines') diff --git a/src/video_core/engines/maxwell_3d.cpp b/src/video_core/engines/maxwell_3d.cpp index ea3c8a963..c01436295 100644 --- a/src/video_core/engines/maxwell_3d.cpp +++ b/src/video_core/engines/maxwell_3d.cpp @@ -128,7 +128,7 @@ void Maxwell3D::CallMacroMethod(u32 method, const std::vector& parameters) ((method - MacroRegistersStart) >> 1) % static_cast(macro_positions.size()); // Execute the current macro. - macro_engine->Execute(macro_positions[entry], parameters); + macro_engine->Execute(*this, macro_positions[entry], parameters); if (mme_draw.current_mode != MMEDrawMode::Undefined) { FlushMMEInlineDraw(); } diff --git a/src/video_core/engines/maxwell_3d.h b/src/video_core/engines/maxwell_3d.h index d5fe25065..5926c4d2d 100644 --- a/src/video_core/engines/maxwell_3d.h +++ b/src/video_core/engines/maxwell_3d.h @@ -1418,6 +1418,10 @@ public: return execute_on; } + VideoCore::RasterizerInterface& GetRasterizer() { + return rasterizer; + } + /// Notify a memory write has happened. void OnMemoryWrite() { dirty.flags |= dirty.on_write_stores; -- cgit v1.2.3 From fabdf5d3850c078d173653f259845c26a2ce6e7d Mon Sep 17 00:00:00 2001 From: David Marcec Date: Fri, 5 Jun 2020 13:09:52 +1000 Subject: Addressed issues --- src/video_core/engines/maxwell_3d.h | 4 ++++ 1 file changed, 4 insertions(+) (limited to 'src/video_core/engines') diff --git a/src/video_core/engines/maxwell_3d.h b/src/video_core/engines/maxwell_3d.h index 5926c4d2d..ef1618990 100644 --- a/src/video_core/engines/maxwell_3d.h +++ b/src/video_core/engines/maxwell_3d.h @@ -1422,6 +1422,10 @@ public: return rasterizer; } + const VideoCore::RasterizerInterface& GetRasterizer() const { + return rasterizer; + } + /// Notify a memory write has happened. void OnMemoryWrite() { dirty.flags |= dirty.on_write_stores; -- cgit v1.2.3