diff options
Diffstat (limited to 'src/video_core')
33 files changed, 1093 insertions, 545 deletions
diff --git a/src/video_core/CMakeLists.txt b/src/video_core/CMakeLists.txt index ddb1a1d69..a780215c1 100644 --- a/src/video_core/CMakeLists.txt +++ b/src/video_core/CMakeLists.txt @@ -21,6 +21,7 @@ add_library(video_core STATIC macro_interpreter.h memory_manager.cpp memory_manager.h + rasterizer_cache.cpp rasterizer_cache.h rasterizer_interface.h renderer_base.cpp @@ -33,6 +34,7 @@ add_library(video_core STATIC renderer_opengl/gl_rasterizer.h renderer_opengl/gl_rasterizer_cache.cpp renderer_opengl/gl_rasterizer_cache.h + renderer_opengl/gl_resource_manager.cpp renderer_opengl/gl_resource_manager.h renderer_opengl/gl_shader_cache.cpp renderer_opengl/gl_shader_cache.h diff --git a/src/video_core/engines/maxwell_3d.cpp b/src/video_core/engines/maxwell_3d.cpp index d79c50919..6de07ea56 100644 --- a/src/video_core/engines/maxwell_3d.cpp +++ b/src/video_core/engines/maxwell_3d.cpp @@ -37,6 +37,35 @@ void Maxwell3D::InitializeRegisterDefaults() { regs.viewport[viewport].depth_range_near = 0.0f; regs.viewport[viewport].depth_range_far = 1.0f; } + // Doom and Bomberman seems to use the uninitialized registers and just enable blend + // so initialize blend registers with sane values + regs.blend.equation_rgb = Regs::Blend::Equation::Add; + regs.blend.factor_source_rgb = Regs::Blend::Factor::One; + regs.blend.factor_dest_rgb = Regs::Blend::Factor::Zero; + regs.blend.equation_a = Regs::Blend::Equation::Add; + regs.blend.factor_source_a = Regs::Blend::Factor::One; + regs.blend.factor_dest_a = Regs::Blend::Factor::Zero; + for (std::size_t blend_index = 0; blend_index < Regs::NumRenderTargets; blend_index++) { + regs.independent_blend[blend_index].equation_rgb = Regs::Blend::Equation::Add; + regs.independent_blend[blend_index].factor_source_rgb = Regs::Blend::Factor::One; + regs.independent_blend[blend_index].factor_dest_rgb = Regs::Blend::Factor::Zero; + regs.independent_blend[blend_index].equation_a = Regs::Blend::Equation::Add; + regs.independent_blend[blend_index].factor_source_a = Regs::Blend::Factor::One; + regs.independent_blend[blend_index].factor_dest_a = Regs::Blend::Factor::Zero; + } + regs.stencil_front_op_fail = Regs::StencilOp::Keep; + regs.stencil_front_op_zfail = Regs::StencilOp::Keep; + regs.stencil_front_op_zpass = Regs::StencilOp::Keep; + regs.stencil_front_func_func = Regs::ComparisonOp::Always; + regs.stencil_front_func_mask = 0xFFFFFFFF; + regs.stencil_front_mask = 0xFFFFFFFF; + regs.stencil_two_side_enable = 1; + regs.stencil_back_op_fail = Regs::StencilOp::Keep; + regs.stencil_back_op_zfail = Regs::StencilOp::Keep; + regs.stencil_back_op_zpass = Regs::StencilOp::Keep; + regs.stencil_back_func_func = Regs::ComparisonOp::Always; + regs.stencil_back_func_mask = 0xFFFFFFFF; + regs.stencil_back_mask = 0xFFFFFFFF; } void Maxwell3D::CallMacroMethod(u32 method, std::vector<u32> parameters) { @@ -92,7 +121,13 @@ void Maxwell3D::WriteReg(u32 method, u32 value, u32 remaining_params) { debug_context->OnEvent(Tegra::DebugContext::Event::MaxwellCommandLoaded, nullptr); } - regs.reg_array[method] = value; + if (regs.reg_array[method] != value) { + regs.reg_array[method] = value; + if (method >= MAXWELL3D_REG_INDEX(vertex_attrib_format) && + method < MAXWELL3D_REG_INDEX(vertex_attrib_format) + regs.vertex_attrib_format.size()) { + dirty_flags.vertex_attrib_format = true; + } + } switch (method) { case MAXWELL3D_REG_INDEX(macros.data): { diff --git a/src/video_core/engines/maxwell_3d.h b/src/video_core/engines/maxwell_3d.h index 50873813e..91ca57883 100644 --- a/src/video_core/engines/maxwell_3d.h +++ b/src/video_core/engines/maxwell_3d.h @@ -345,6 +345,14 @@ public: Invert = 6, IncrWrap = 7, DecrWrap = 8, + KeepOGL = 0x1E00, + ZeroOGL = 0, + ReplaceOGL = 0x1E01, + IncrOGL = 0x1E02, + DecrOGL = 0x1E03, + InvertOGL = 0x150A, + IncrWrapOGL = 0x8507, + DecrWrapOGL = 0x8508, }; enum class MemoryLayout : u32 { @@ -462,6 +470,16 @@ public: } }; + struct ColorMask { + union { + u32 raw; + BitField<0, 4, u32> R; + BitField<4, 4, u32> G; + BitField<8, 4, u32> B; + BitField<12, 4, u32> A; + }; + }; + bool IsShaderConfigEnabled(std::size_t index) const { // The VertexB is always enabled. if (index == static_cast<std::size_t>(Regs::ShaderProgram::VertexB)) { @@ -571,7 +589,11 @@ public: u32 stencil_back_mask; u32 stencil_back_func_mask; - INSERT_PADDING_WORDS(0x13); + INSERT_PADDING_WORDS(0xC); + + u32 color_mask_common; + + INSERT_PADDING_WORDS(0x6); u32 rt_separate_frag_data; @@ -646,8 +668,14 @@ public: ComparisonOp depth_test_func; float alpha_test_ref; ComparisonOp alpha_test_func; - - INSERT_PADDING_WORDS(0x9); + u32 draw_tfb_stride; + struct { + float r; + float g; + float b; + float a; + } blend_color; + INSERT_PADDING_WORDS(0x4); struct { u32 separate_alpha; @@ -841,8 +869,9 @@ public: BitField<6, 4, u32> RT; BitField<10, 11, u32> layer; } clear_buffers; - - INSERT_PADDING_WORDS(0x4B); + INSERT_PADDING_WORDS(0xB); + std::array<ColorMask, NumRenderTargets> color_mask; + INSERT_PADDING_WORDS(0x38); struct { u32 query_address_high; @@ -983,6 +1012,12 @@ public: State state{}; MemoryManager& memory_manager; + struct DirtyFlags { + bool vertex_attrib_format = true; + }; + + DirtyFlags dirty_flags; + /// Reads a register value located at the input method address u32 GetRegisterValue(u32 method) const; @@ -1075,6 +1110,7 @@ ASSERT_REG_POSITION(scissor_test, 0x380); ASSERT_REG_POSITION(stencil_back_func_ref, 0x3D5); ASSERT_REG_POSITION(stencil_back_mask, 0x3D6); ASSERT_REG_POSITION(stencil_back_func_mask, 0x3D7); +ASSERT_REG_POSITION(color_mask_common, 0x3E4); ASSERT_REG_POSITION(rt_separate_frag_data, 0x3EB); ASSERT_REG_POSITION(zeta, 0x3F8); ASSERT_REG_POSITION(vertex_attrib_format, 0x458); @@ -1087,6 +1123,10 @@ ASSERT_REG_POSITION(depth_write_enabled, 0x4BA); ASSERT_REG_POSITION(alpha_test_enabled, 0x4BB); ASSERT_REG_POSITION(d3d_cull_mode, 0x4C2); ASSERT_REG_POSITION(depth_test_func, 0x4C3); +ASSERT_REG_POSITION(alpha_test_ref, 0x4C4); +ASSERT_REG_POSITION(alpha_test_func, 0x4C5); +ASSERT_REG_POSITION(draw_tfb_stride, 0x4C6); +ASSERT_REG_POSITION(blend_color, 0x4C7); ASSERT_REG_POSITION(blend, 0x4CF); ASSERT_REG_POSITION(stencil_enable, 0x4E0); ASSERT_REG_POSITION(stencil_front_op_fail, 0x4E1); @@ -1117,6 +1157,7 @@ ASSERT_REG_POSITION(instanced_arrays, 0x620); ASSERT_REG_POSITION(cull, 0x646); ASSERT_REG_POSITION(logic_op, 0x671); ASSERT_REG_POSITION(clear_buffers, 0x674); +ASSERT_REG_POSITION(color_mask, 0x680); ASSERT_REG_POSITION(query, 0x6C0); ASSERT_REG_POSITION(vertex_array[0], 0x700); ASSERT_REG_POSITION(independent_blend, 0x780); diff --git a/src/video_core/memory_manager.cpp b/src/video_core/memory_manager.cpp index 90a8e825d..77a20bb84 100644 --- a/src/video_core/memory_manager.cpp +++ b/src/video_core/memory_manager.cpp @@ -4,18 +4,21 @@ #include "common/alignment.h" #include "common/assert.h" +#include "common/logging/log.h" #include "video_core/memory_manager.h" namespace Tegra { GPUVAddr MemoryManager::AllocateSpace(u64 size, u64 align) { - std::optional<GPUVAddr> gpu_addr = FindFreeBlock(size, align); - ASSERT(gpu_addr); + const std::optional<GPUVAddr> gpu_addr{FindFreeBlock(0, size, align, PageStatus::Unmapped)}; - for (u64 offset = 0; offset < size; offset += PAGE_SIZE) { - VAddr& slot = PageSlot(*gpu_addr + offset); + ASSERT_MSG(gpu_addr, "unable to find available GPU memory"); + + for (u64 offset{}; offset < size; offset += PAGE_SIZE) { + VAddr& slot{PageSlot(*gpu_addr + offset)}; ASSERT(slot == static_cast<u64>(PageStatus::Unmapped)); + slot = static_cast<u64>(PageStatus::Allocated); } @@ -23,10 +26,11 @@ GPUVAddr MemoryManager::AllocateSpace(u64 size, u64 align) { } GPUVAddr MemoryManager::AllocateSpace(GPUVAddr gpu_addr, u64 size, u64 align) { - for (u64 offset = 0; offset < size; offset += PAGE_SIZE) { - VAddr& slot = PageSlot(gpu_addr + offset); + for (u64 offset{}; offset < size; offset += PAGE_SIZE) { + VAddr& slot{PageSlot(gpu_addr + offset)}; ASSERT(slot == static_cast<u64>(PageStatus::Unmapped)); + slot = static_cast<u64>(PageStatus::Allocated); } @@ -34,17 +38,19 @@ GPUVAddr MemoryManager::AllocateSpace(GPUVAddr gpu_addr, u64 size, u64 align) { } GPUVAddr MemoryManager::MapBufferEx(VAddr cpu_addr, u64 size) { - std::optional<GPUVAddr> gpu_addr = FindFreeBlock(size, PAGE_SIZE); - ASSERT(gpu_addr); + const std::optional<GPUVAddr> gpu_addr{FindFreeBlock(0, size, PAGE_SIZE, PageStatus::Unmapped)}; + + ASSERT_MSG(gpu_addr, "unable to find available GPU memory"); - for (u64 offset = 0; offset < size; offset += PAGE_SIZE) { - VAddr& slot = PageSlot(*gpu_addr + offset); + for (u64 offset{}; offset < size; offset += PAGE_SIZE) { + VAddr& slot{PageSlot(*gpu_addr + offset)}; ASSERT(slot == static_cast<u64>(PageStatus::Unmapped)); + slot = cpu_addr + offset; } - MappedRegion region{cpu_addr, *gpu_addr, size}; + const MappedRegion region{cpu_addr, *gpu_addr, size}; mapped_regions.push_back(region); return *gpu_addr; @@ -53,14 +59,31 @@ GPUVAddr MemoryManager::MapBufferEx(VAddr cpu_addr, u64 size) { GPUVAddr MemoryManager::MapBufferEx(VAddr cpu_addr, GPUVAddr gpu_addr, u64 size) { ASSERT((gpu_addr & PAGE_MASK) == 0); - for (u64 offset = 0; offset < size; offset += PAGE_SIZE) { - VAddr& slot = PageSlot(gpu_addr + offset); + if (PageSlot(gpu_addr) != static_cast<u64>(PageStatus::Allocated)) { + // Page has been already mapped. In this case, we must find a new area of memory to use that + // is different than the specified one. Super Mario Odyssey hits this scenario when changing + // areas, but we do not want to overwrite the old pages. + // TODO(bunnei): We need to write a hardware test to confirm this behavior. + + LOG_ERROR(HW_GPU, "attempting to map addr 0x{:016X}, which is not available!", gpu_addr); + + const std::optional<GPUVAddr> new_gpu_addr{ + FindFreeBlock(gpu_addr, size, PAGE_SIZE, PageStatus::Allocated)}; + + ASSERT_MSG(new_gpu_addr, "unable to find available GPU memory"); + + gpu_addr = *new_gpu_addr; + } + + for (u64 offset{}; offset < size; offset += PAGE_SIZE) { + VAddr& slot{PageSlot(gpu_addr + offset)}; ASSERT(slot == static_cast<u64>(PageStatus::Allocated)); + slot = cpu_addr + offset; } - MappedRegion region{cpu_addr, gpu_addr, size}; + const MappedRegion region{cpu_addr, gpu_addr, size}; mapped_regions.push_back(region); return gpu_addr; @@ -69,11 +92,12 @@ GPUVAddr MemoryManager::MapBufferEx(VAddr cpu_addr, GPUVAddr gpu_addr, u64 size) GPUVAddr MemoryManager::UnmapBuffer(GPUVAddr gpu_addr, u64 size) { ASSERT((gpu_addr & PAGE_MASK) == 0); - for (u64 offset = 0; offset < size; offset += PAGE_SIZE) { - VAddr& slot = PageSlot(gpu_addr + offset); + for (u64 offset{}; offset < size; offset += PAGE_SIZE) { + VAddr& slot{PageSlot(gpu_addr + offset)}; ASSERT(slot != static_cast<u64>(PageStatus::Allocated) && slot != static_cast<u64>(PageStatus::Unmapped)); + slot = static_cast<u64>(PageStatus::Unmapped); } @@ -97,13 +121,14 @@ GPUVAddr MemoryManager::GetRegionEnd(GPUVAddr region_start) const { return {}; } -std::optional<GPUVAddr> MemoryManager::FindFreeBlock(u64 size, u64 align) { - GPUVAddr gpu_addr = 0; - u64 free_space = 0; +std::optional<GPUVAddr> MemoryManager::FindFreeBlock(GPUVAddr region_start, u64 size, u64 align, + PageStatus status) { + GPUVAddr gpu_addr{region_start}; + u64 free_space{}; align = (align + PAGE_MASK) & ~PAGE_MASK; while (gpu_addr + free_space < MAX_ADDRESS) { - if (!IsPageMapped(gpu_addr + free_space)) { + if (PageSlot(gpu_addr + free_space) == static_cast<u64>(status)) { free_space += PAGE_SIZE; if (free_space >= size) { return gpu_addr; @@ -119,7 +144,7 @@ std::optional<GPUVAddr> MemoryManager::FindFreeBlock(u64 size, u64 align) { } std::optional<VAddr> MemoryManager::GpuToCpuAddress(GPUVAddr gpu_addr) { - VAddr base_addr = PageSlot(gpu_addr); + const VAddr base_addr{PageSlot(gpu_addr)}; if (base_addr == static_cast<u64>(PageStatus::Allocated) || base_addr == static_cast<u64>(PageStatus::Unmapped)) { @@ -133,19 +158,15 @@ std::vector<GPUVAddr> MemoryManager::CpuToGpuAddress(VAddr cpu_addr) const { std::vector<GPUVAddr> results; for (const auto& region : mapped_regions) { if (cpu_addr >= region.cpu_addr && cpu_addr < (region.cpu_addr + region.size)) { - u64 offset = cpu_addr - region.cpu_addr; + const u64 offset{cpu_addr - region.cpu_addr}; results.push_back(region.gpu_addr + offset); } } return results; } -bool MemoryManager::IsPageMapped(GPUVAddr gpu_addr) { - return PageSlot(gpu_addr) != static_cast<u64>(PageStatus::Unmapped); -} - VAddr& MemoryManager::PageSlot(GPUVAddr gpu_addr) { - auto& block = page_table[(gpu_addr >> (PAGE_BITS + PAGE_TABLE_BITS)) & PAGE_TABLE_MASK]; + auto& block{page_table[(gpu_addr >> (PAGE_BITS + PAGE_TABLE_BITS)) & PAGE_TABLE_MASK]}; if (!block) { block = std::make_unique<PageBlock>(); block->fill(static_cast<VAddr>(PageStatus::Unmapped)); diff --git a/src/video_core/memory_manager.h b/src/video_core/memory_manager.h index b1255fd56..4eb338aa2 100644 --- a/src/video_core/memory_manager.h +++ b/src/video_core/memory_manager.h @@ -34,15 +34,15 @@ public: static constexpr u64 PAGE_MASK = PAGE_SIZE - 1; private: - std::optional<GPUVAddr> FindFreeBlock(u64 size, u64 align = 1); - bool IsPageMapped(GPUVAddr gpu_addr); - VAddr& PageSlot(GPUVAddr gpu_addr); - enum class PageStatus : u64 { Unmapped = 0xFFFFFFFFFFFFFFFFULL, Allocated = 0xFFFFFFFFFFFFFFFEULL, }; + std::optional<GPUVAddr> FindFreeBlock(GPUVAddr region_start, u64 size, u64 align, + PageStatus status); + VAddr& PageSlot(GPUVAddr gpu_addr); + static constexpr u64 MAX_ADDRESS{0x10000000000ULL}; static constexpr u64 PAGE_TABLE_BITS{10}; static constexpr u64 PAGE_TABLE_SIZE{1 << PAGE_TABLE_BITS}; diff --git a/src/video_core/rasterizer_cache.cpp b/src/video_core/rasterizer_cache.cpp new file mode 100644 index 000000000..093b2cdf4 --- /dev/null +++ b/src/video_core/rasterizer_cache.cpp @@ -0,0 +1,7 @@ +// Copyright 2018 yuzu Emulator Project +// Licensed under GPLv2 or any later version +// Refer to the license.txt file included. + +#include "video_core/rasterizer_cache.h" + +RasterizerCacheObject::~RasterizerCacheObject() = default; diff --git a/src/video_core/rasterizer_cache.h b/src/video_core/rasterizer_cache.h index 0a3b3951e..bcf0c15a4 100644 --- a/src/video_core/rasterizer_cache.h +++ b/src/video_core/rasterizer_cache.h @@ -5,18 +5,19 @@ #pragma once #include <set> +#include <unordered_map> #include <boost/icl/interval_map.hpp> #include <boost/range/iterator_range_core.hpp> #include "common/common_types.h" -#include "core/core.h" #include "core/settings.h" #include "video_core/rasterizer_interface.h" -#include "video_core/renderer_base.h" class RasterizerCacheObject { public: + virtual ~RasterizerCacheObject(); + /// Gets the address of the shader in guest memory, required for cache management virtual VAddr GetAddr() const = 0; @@ -64,6 +65,8 @@ class RasterizerCache : NonCopyable { friend class RasterizerCacheObject; public: + explicit RasterizerCache(VideoCore::RasterizerInterface& rasterizer) : rasterizer{rasterizer} {} + /// Write any cached resources overlapping the specified region back to memory void FlushRegion(Tegra::GPUVAddr addr, size_t size) { const auto& objects{GetSortedObjectsFromRegion(addr, size)}; @@ -86,45 +89,39 @@ public: /// Invalidates everything in the cache void InvalidateAll() { - while (object_cache.begin() != object_cache.end()) { - Unregister(*object_cache.begin()->second.begin()); + while (interval_cache.begin() != interval_cache.end()) { + Unregister(*interval_cache.begin()->second.begin()); } } protected: /// Tries to get an object from the cache with the specified address T TryGet(VAddr addr) const { - const ObjectInterval interval{addr}; - for (auto& pair : boost::make_iterator_range(object_cache.equal_range(interval))) { - for (auto& cached_object : pair.second) { - if (cached_object->GetAddr() == addr) { - return cached_object; - } - } - } + const auto iter = map_cache.find(addr); + if (iter != map_cache.end()) + return iter->second; return nullptr; } /// Register an object into the cache void Register(const T& object) { object->SetIsRegistered(true); - object_cache.add({GetInterval(object), ObjectSet{object}}); - auto& rasterizer = Core::System::GetInstance().Renderer().Rasterizer(); + interval_cache.add({GetInterval(object), ObjectSet{object}}); + map_cache.insert({object->GetAddr(), object}); rasterizer.UpdatePagesCachedCount(object->GetAddr(), object->GetSizeInBytes(), 1); } /// Unregisters an object from the cache void Unregister(const T& object) { object->SetIsRegistered(false); - auto& rasterizer = Core::System::GetInstance().Renderer().Rasterizer(); rasterizer.UpdatePagesCachedCount(object->GetAddr(), object->GetSizeInBytes(), -1); - // Only flush if use_accurate_gpu_emulation is enabled, as it incurs a performance hit if (Settings::values.use_accurate_gpu_emulation) { FlushObject(object); } - object_cache.subtract({GetInterval(object), ObjectSet{object}}); + interval_cache.subtract({GetInterval(object), ObjectSet{object}}); + map_cache.erase(object->GetAddr()); } /// Returns a ticks counter used for tracking when cached objects were last modified @@ -141,7 +138,7 @@ private: std::vector<T> objects; const ObjectInterval interval{addr, addr + size}; - for (auto& pair : boost::make_iterator_range(object_cache.equal_range(interval))) { + for (auto& pair : boost::make_iterator_range(interval_cache.equal_range(interval))) { for (auto& cached_object : pair.second) { if (!cached_object) { continue; @@ -167,14 +164,17 @@ private: } using ObjectSet = std::set<T>; - using ObjectCache = boost::icl::interval_map<VAddr, ObjectSet>; - using ObjectInterval = typename ObjectCache::interval_type; + using ObjectCache = std::unordered_map<VAddr, T>; + using IntervalCache = boost::icl::interval_map<VAddr, ObjectSet>; + using ObjectInterval = typename IntervalCache::interval_type; static auto GetInterval(const T& object) { return ObjectInterval::right_open(object->GetAddr(), object->GetAddr() + object->GetSizeInBytes()); } - ObjectCache object_cache; ///< Cache of objects - u64 modified_ticks{}; ///< Counter of cache state ticks, used for in-order flushing + ObjectCache map_cache; + IntervalCache interval_cache; ///< Cache of objects + u64 modified_ticks{}; ///< Counter of cache state ticks, used for in-order flushing + VideoCore::RasterizerInterface& rasterizer; }; diff --git a/src/video_core/renderer_base.cpp b/src/video_core/renderer_base.cpp index 0df3725c2..1482cdb40 100644 --- a/src/video_core/renderer_base.cpp +++ b/src/video_core/renderer_base.cpp @@ -5,7 +5,6 @@ #include "core/frontend/emu_window.h" #include "core/settings.h" #include "video_core/renderer_base.h" -#include "video_core/renderer_opengl/gl_rasterizer.h" namespace VideoCore { diff --git a/src/video_core/renderer_opengl/gl_buffer_cache.cpp b/src/video_core/renderer_opengl/gl_buffer_cache.cpp index 41a54b3e7..075192c3f 100644 --- a/src/video_core/renderer_opengl/gl_buffer_cache.cpp +++ b/src/video_core/renderer_opengl/gl_buffer_cache.cpp @@ -9,10 +9,12 @@ #include "core/core.h" #include "core/memory.h" #include "video_core/renderer_opengl/gl_buffer_cache.h" +#include "video_core/renderer_opengl/gl_rasterizer.h" namespace OpenGL { -OGLBufferCache::OGLBufferCache(std::size_t size) : stream_buffer(GL_ARRAY_BUFFER, size) {} +OGLBufferCache::OGLBufferCache(RasterizerOpenGL& rasterizer, std::size_t size) + : RasterizerCache{rasterizer}, stream_buffer(GL_ARRAY_BUFFER, size) {} GLintptr OGLBufferCache::UploadMemory(Tegra::GPUVAddr gpu_addr, std::size_t size, std::size_t alignment, bool cache) { diff --git a/src/video_core/renderer_opengl/gl_buffer_cache.h b/src/video_core/renderer_opengl/gl_buffer_cache.h index be29dc8be..91fca3f6c 100644 --- a/src/video_core/renderer_opengl/gl_buffer_cache.h +++ b/src/video_core/renderer_opengl/gl_buffer_cache.h @@ -15,6 +15,8 @@ namespace OpenGL { +class RasterizerOpenGL; + struct CachedBufferEntry final : public RasterizerCacheObject { VAddr GetAddr() const override { return addr; @@ -35,7 +37,7 @@ struct CachedBufferEntry final : public RasterizerCacheObject { class OGLBufferCache final : public RasterizerCache<std::shared_ptr<CachedBufferEntry>> { public: - explicit OGLBufferCache(std::size_t size); + explicit OGLBufferCache(RasterizerOpenGL& rasterizer, std::size_t size); /// Uploads data from a guest GPU address. Returns host's buffer offset where it's been /// allocated. diff --git a/src/video_core/renderer_opengl/gl_primitive_assembler.cpp b/src/video_core/renderer_opengl/gl_primitive_assembler.cpp index 741f14bc3..d9ed08437 100644 --- a/src/video_core/renderer_opengl/gl_primitive_assembler.cpp +++ b/src/video_core/renderer_opengl/gl_primitive_assembler.cpp @@ -6,6 +6,7 @@ #include <array> #include "common/assert.h" #include "common/common_types.h" +#include "core/core.h" #include "core/memory.h" #include "video_core/renderer_opengl/gl_buffer_cache.h" #include "video_core/renderer_opengl/gl_primitive_assembler.h" diff --git a/src/video_core/renderer_opengl/gl_rasterizer.cpp b/src/video_core/renderer_opengl/gl_rasterizer.cpp index a0527fe57..84bd91eed 100644 --- a/src/video_core/renderer_opengl/gl_rasterizer.cpp +++ b/src/video_core/renderer_opengl/gl_rasterizer.cpp @@ -33,7 +33,8 @@ using Maxwell = Tegra::Engines::Maxwell3D::Regs; using PixelFormat = VideoCore::Surface::PixelFormat; using SurfaceType = VideoCore::Surface::SurfaceType; -MICROPROFILE_DEFINE(OpenGL_VAO, "OpenGL", "Vertex Array Setup", MP_RGB(128, 128, 192)); +MICROPROFILE_DEFINE(OpenGL_VAO, "OpenGL", "Vertex Format Setup", MP_RGB(128, 128, 192)); +MICROPROFILE_DEFINE(OpenGL_VB, "OpenGL", "Vertex Buffer Setup", MP_RGB(128, 128, 192)); MICROPROFILE_DEFINE(OpenGL_Shader, "OpenGL", "Shader Setup", MP_RGB(128, 128, 192)); MICROPROFILE_DEFINE(OpenGL_UBO, "OpenGL", "Const Buffer Setup", MP_RGB(128, 128, 192)); MICROPROFILE_DEFINE(OpenGL_Index, "OpenGL", "Index Buffer Setup", MP_RGB(128, 128, 192)); @@ -79,7 +80,8 @@ struct DrawParameters { }; RasterizerOpenGL::RasterizerOpenGL(Core::Frontend::EmuWindow& window, ScreenInfo& info) - : emu_window{window}, screen_info{info}, buffer_cache(STREAM_BUFFER_SIZE) { + : res_cache{*this}, shader_cache{*this}, emu_window{window}, screen_info{info}, + buffer_cache(*this, STREAM_BUFFER_SIZE) { // Create sampler objects for (std::size_t i = 0; i < texture_samplers.size(); ++i) { texture_samplers[i].Create(); @@ -105,8 +107,6 @@ RasterizerOpenGL::RasterizerOpenGL(Core::Frontend::EmuWindow& window, ScreenInfo ASSERT_MSG(has_ARB_separate_shader_objects, "has_ARB_separate_shader_objects is unsupported"); OpenGLState::ApplyDefaultState(); - // Clipping plane 0 is always enabled for PICA fixed clip plane z <= 0 - state.clip_distance[0] = true; // Create render framebuffer framebuffer.Create(); @@ -122,18 +122,23 @@ RasterizerOpenGL::RasterizerOpenGL(Core::Frontend::EmuWindow& window, ScreenInfo RasterizerOpenGL::~RasterizerOpenGL() {} -void RasterizerOpenGL::SetupVertexArrays() { - MICROPROFILE_SCOPE(OpenGL_VAO); - const auto& gpu = Core::System::GetInstance().GPU().Maxwell3D(); +void RasterizerOpenGL::SetupVertexFormat() { + auto& gpu = Core::System::GetInstance().GPU().Maxwell3D(); const auto& regs = gpu.regs; + if (!gpu.dirty_flags.vertex_attrib_format) + return; + gpu.dirty_flags.vertex_attrib_format = false; + + MICROPROFILE_SCOPE(OpenGL_VAO); + auto [iter, is_cache_miss] = vertex_array_cache.try_emplace(regs.vertex_attrib_format); auto& VAO = iter->second; if (is_cache_miss) { VAO.Create(); state.draw.vertex_array = VAO.handle; - state.Apply(); + state.ApplyVertexBufferState(); // The index buffer binding is stored within the VAO. Stupid OpenGL, but easy to work // around. @@ -175,8 +180,13 @@ void RasterizerOpenGL::SetupVertexArrays() { } } state.draw.vertex_array = VAO.handle; - state.draw.vertex_buffer = buffer_cache.GetHandle(); - state.Apply(); + state.ApplyVertexBufferState(); +} + +void RasterizerOpenGL::SetupVertexBuffer() { + MICROPROFILE_SCOPE(OpenGL_VB); + const auto& gpu = Core::System::GetInstance().GPU().Maxwell3D(); + const auto& regs = gpu.regs; // Upload all guest vertex arrays sequentially to our buffer for (u32 index = 0; index < Maxwell::NumVertexArrays; ++index) { @@ -203,6 +213,9 @@ void RasterizerOpenGL::SetupVertexArrays() { glVertexBindingDivisor(index, 0); } } + + // Implicit set by glBindVertexBuffer. Stupid glstate handling... + state.draw.vertex_buffer = buffer_cache.GetHandle(); } DrawParameters RasterizerOpenGL::SetupDraw() { @@ -327,8 +340,6 @@ void RasterizerOpenGL::SetupShaders(GLenum primitive_mode) { index++; } } - - state.Apply(); } std::size_t RasterizerOpenGL::CalculateVertexArraysSize() const { @@ -397,8 +408,8 @@ void RasterizerOpenGL::UpdatePagesCachedCount(VAddr addr, u64 size, int delta) { cached_pages.add({pages_interval, delta}); } -void RasterizerOpenGL::ConfigureFramebuffers(bool using_color_fb, bool using_depth_fb, - bool preserve_contents, +void RasterizerOpenGL::ConfigureFramebuffers(OpenGLState& current_state, bool using_color_fb, + bool using_depth_fb, bool preserve_contents, std::optional<std::size_t> single_color_target) { MICROPROFILE_SCOPE(OpenGL_Framebuffer); const auto& regs = Core::System::GetInstance().GPU().Maxwell3D().regs; @@ -414,9 +425,9 @@ void RasterizerOpenGL::ConfigureFramebuffers(bool using_color_fb, bool using_dep ASSERT_MSG(regs.rt_separate_frag_data == 0, "Unimplemented"); // Bind the framebuffer surfaces - state.draw.draw_framebuffer = framebuffer.handle; - state.Apply(); - state.framebuffer_srgb.enabled = regs.framebuffer_srgb != 0; + current_state.draw.draw_framebuffer = framebuffer.handle; + current_state.ApplyFramebufferState(); + current_state.framebuffer_srgb.enabled = regs.framebuffer_srgb != 0; if (using_color_fb) { if (single_color_target) { @@ -494,10 +505,7 @@ void RasterizerOpenGL::ConfigureFramebuffers(bool using_color_fb, bool using_dep glFramebufferTexture2D(GL_DRAW_FRAMEBUFFER, GL_DEPTH_STENCIL_ATTACHMENT, GL_TEXTURE_2D, 0, 0); } - - SyncViewport(); - - state.Apply(); + SyncViewport(current_state); } void RasterizerOpenGL::Clear() { @@ -510,22 +518,23 @@ void RasterizerOpenGL::Clear() { bool use_stencil{}; OpenGLState clear_state; - clear_state.draw.draw_framebuffer = framebuffer.handle; - clear_state.color_mask.red_enabled = regs.clear_buffers.R ? GL_TRUE : GL_FALSE; - clear_state.color_mask.green_enabled = regs.clear_buffers.G ? GL_TRUE : GL_FALSE; - clear_state.color_mask.blue_enabled = regs.clear_buffers.B ? GL_TRUE : GL_FALSE; - clear_state.color_mask.alpha_enabled = regs.clear_buffers.A ? GL_TRUE : GL_FALSE; - if (regs.clear_buffers.R || regs.clear_buffers.G || regs.clear_buffers.B || regs.clear_buffers.A) { use_color = true; } + if (use_color) { + clear_state.color_mask[0].red_enabled = regs.clear_buffers.R ? GL_TRUE : GL_FALSE; + clear_state.color_mask[0].green_enabled = regs.clear_buffers.G ? GL_TRUE : GL_FALSE; + clear_state.color_mask[0].blue_enabled = regs.clear_buffers.B ? GL_TRUE : GL_FALSE; + clear_state.color_mask[0].alpha_enabled = regs.clear_buffers.A ? GL_TRUE : GL_FALSE; + } if (regs.clear_buffers.Z) { ASSERT_MSG(regs.zeta_enable != 0, "Tried to clear Z but buffer is not enabled!"); use_depth = true; // Always enable the depth write when clearing the depth buffer. The depth write mask is - // ignored when clearing the buffer in the Switch, but OpenGL obeys it so we set it to true. + // ignored when clearing the buffer in the Switch, but OpenGL obeys it so we set it to + // true. clear_state.depth.test_enabled = true; clear_state.depth.test_func = GL_ALWAYS; } @@ -542,11 +551,8 @@ void RasterizerOpenGL::Clear() { ScopeAcquireGLContext acquire_context{emu_window}; - ConfigureFramebuffers(use_color, use_depth || use_stencil, false, + ConfigureFramebuffers(clear_state, use_color, use_depth || use_stencil, false, regs.clear_buffers.RT.Value()); - // Copy the sRGB setting to the clear state to avoid problem with - // specific driver implementations - clear_state.framebuffer_srgb.enabled = state.framebuffer_srgb.enabled; clear_state.Apply(); if (use_color) { @@ -572,15 +578,14 @@ void RasterizerOpenGL::DrawArrays() { ScopeAcquireGLContext acquire_context{emu_window}; - ConfigureFramebuffers(); - + ConfigureFramebuffers(state); + SyncColorMask(); SyncDepthTestState(); SyncStencilTestState(); SyncBlendState(); SyncLogicOpState(); SyncCullMode(); SyncPrimitiveRestart(); - SyncDepthRange(); SyncScissorTest(); // Alpha Testing is synced on shaders. SyncTransformFeedback(); @@ -594,7 +599,7 @@ void RasterizerOpenGL::DrawArrays() { const bool is_indexed = accelerate_draw == AccelDraw::Indexed; state.draw.vertex_buffer = buffer_cache.GetHandle(); - state.Apply(); + state.ApplyVertexBufferState(); std::size_t buffer_size = CalculateVertexArraysSize(); @@ -621,7 +626,8 @@ void RasterizerOpenGL::DrawArrays() { buffer_cache.Map(buffer_size); - SetupVertexArrays(); + SetupVertexFormat(); + SetupVertexBuffer(); DrawParameters params = SetupDraw(); SetupShaders(params.primitive_mode); @@ -725,9 +731,9 @@ void RasterizerOpenGL::SamplerInfo::Create() { glSamplerParameteri(sampler.handle, GL_TEXTURE_COMPARE_FUNC, GL_NEVER); } -void RasterizerOpenGL::SamplerInfo::SyncWithConfig(const Tegra::Texture::TSCEntry& config) { +void RasterizerOpenGL::SamplerInfo::SyncWithConfig(const Tegra::Texture::FullTextureInfo& info) { const GLuint s = sampler.handle; - + const Tegra::Texture::TSCEntry& config = info.tsc; if (mag_filter != config.mag_filter) { mag_filter = config.mag_filter; glSamplerParameteri( @@ -778,6 +784,22 @@ void RasterizerOpenGL::SamplerInfo::SyncWithConfig(const Tegra::Texture::TSCEntr glSamplerParameterfv(s, GL_TEXTURE_BORDER_COLOR, border_color.data()); } } + if (info.tic.use_header_opt_control == 0) { + if (GLAD_GL_ARB_texture_filter_anisotropic) { + glSamplerParameterf(s, GL_TEXTURE_MAX_ANISOTROPY, + static_cast<float>(1 << info.tic.max_anisotropy.Value())); + } else if (GLAD_GL_EXT_texture_filter_anisotropic) { + glSamplerParameterf(s, GL_TEXTURE_MAX_ANISOTROPY_EXT, + static_cast<float>(1 << info.tic.max_anisotropy.Value())); + } + glSamplerParameterf(s, GL_TEXTURE_MIN_LOD, + static_cast<float>(info.tic.res_min_mip_level.Value())); + glSamplerParameterf(s, GL_TEXTURE_MAX_LOD, + static_cast<float>(info.tic.res_max_mip_level.Value() == 0 + ? 16 + : info.tic.res_max_mip_level.Value())); + glSamplerParameterf(s, GL_TEXTURE_LOD_BIAS, info.tic.mip_lod_bias.Value() / 256.f); + } } u32 RasterizerOpenGL::SetupConstBuffers(Maxwell::ShaderStage stage, Shader& shader, @@ -875,7 +897,7 @@ u32 RasterizerOpenGL::SetupTextures(Maxwell::ShaderStage stage, Shader& shader, continue; } - texture_samplers[current_bindpoint].SyncWithConfig(texture.tsc); + texture_samplers[current_bindpoint].SyncWithConfig(texture); Surface surface = res_cache.GetTextureSurface(texture, entry); if (surface != nullptr) { state.texture_units[current_bindpoint].texture = surface->Texture().handle; @@ -897,14 +919,18 @@ u32 RasterizerOpenGL::SetupTextures(Maxwell::ShaderStage stage, Shader& shader, return current_unit + static_cast<u32>(entries.size()); } -void RasterizerOpenGL::SyncViewport() { +void RasterizerOpenGL::SyncViewport(OpenGLState& current_state) { const auto& regs = Core::System::GetInstance().GPU().Maxwell3D().regs; - const MathUtil::Rectangle<s32> viewport_rect{regs.viewport_transform[0].GetRect()}; - - state.viewport.x = viewport_rect.left; - state.viewport.y = viewport_rect.bottom; - state.viewport.width = static_cast<GLsizei>(viewport_rect.GetWidth()); - state.viewport.height = static_cast<GLsizei>(viewport_rect.GetHeight()); + for (size_t i = 0; i < Tegra::Engines::Maxwell3D::Regs::NumRenderTargets; i++) { + const MathUtil::Rectangle<s32> viewport_rect{regs.viewport_transform[i].GetRect()}; + auto& viewport = current_state.viewports[i]; + viewport.x = viewport_rect.left; + viewport.y = viewport_rect.bottom; + viewport.width = static_cast<GLfloat>(viewport_rect.GetWidth()); + viewport.height = static_cast<GLfloat>(viewport_rect.GetHeight()); + viewport.depth_range_far = regs.viewport[i].depth_range_far; + viewport.depth_range_near = regs.viewport[i].depth_range_near; + } } void RasterizerOpenGL::SyncClipEnabled() { @@ -946,13 +972,6 @@ void RasterizerOpenGL::SyncPrimitiveRestart() { state.primitive_restart.index = regs.primitive_restart.index; } -void RasterizerOpenGL::SyncDepthRange() { - const auto& regs = Core::System::GetInstance().GPU().Maxwell3D().regs; - - state.depth.depth_range_near = regs.viewport->depth_range_near; - state.depth.depth_range_far = regs.viewport->depth_range_far; -} - void RasterizerOpenGL::SyncDepthTestState() { const auto& regs = Core::System::GetInstance().GPU().Maxwell3D().regs; @@ -973,9 +992,6 @@ void RasterizerOpenGL::SyncStencilTestState() { return; } - // TODO(bunnei): Verify behavior when this is not set - ASSERT(regs.stencil_two_side_enable); - state.stencil.front.test_func = MaxwellToGL::ComparisonOp(regs.stencil_front_func_func); state.stencil.front.test_ref = regs.stencil_front_func_ref; state.stencil.front.test_mask = regs.stencil_front_func_mask; @@ -983,36 +999,79 @@ void RasterizerOpenGL::SyncStencilTestState() { state.stencil.front.action_depth_fail = MaxwellToGL::StencilOp(regs.stencil_front_op_zfail); state.stencil.front.action_depth_pass = MaxwellToGL::StencilOp(regs.stencil_front_op_zpass); state.stencil.front.write_mask = regs.stencil_front_mask; + if (regs.stencil_two_side_enable) { + state.stencil.back.test_func = MaxwellToGL::ComparisonOp(regs.stencil_back_func_func); + state.stencil.back.test_ref = regs.stencil_back_func_ref; + state.stencil.back.test_mask = regs.stencil_back_func_mask; + state.stencil.back.action_stencil_fail = MaxwellToGL::StencilOp(regs.stencil_back_op_fail); + state.stencil.back.action_depth_fail = MaxwellToGL::StencilOp(regs.stencil_back_op_zfail); + state.stencil.back.action_depth_pass = MaxwellToGL::StencilOp(regs.stencil_back_op_zpass); + state.stencil.back.write_mask = regs.stencil_back_mask; + } else { + state.stencil.back.test_func = GL_ALWAYS; + state.stencil.back.test_ref = 0; + state.stencil.back.test_mask = 0xFFFFFFFF; + state.stencil.back.write_mask = 0xFFFFFFFF; + state.stencil.back.action_stencil_fail = GL_KEEP; + state.stencil.back.action_depth_fail = GL_KEEP; + state.stencil.back.action_depth_pass = GL_KEEP; + } +} - state.stencil.back.test_func = MaxwellToGL::ComparisonOp(regs.stencil_back_func_func); - state.stencil.back.test_ref = regs.stencil_back_func_ref; - state.stencil.back.test_mask = regs.stencil_back_func_mask; - state.stencil.back.action_stencil_fail = MaxwellToGL::StencilOp(regs.stencil_back_op_fail); - state.stencil.back.action_depth_fail = MaxwellToGL::StencilOp(regs.stencil_back_op_zfail); - state.stencil.back.action_depth_pass = MaxwellToGL::StencilOp(regs.stencil_back_op_zpass); - state.stencil.back.write_mask = regs.stencil_back_mask; +void RasterizerOpenGL::SyncColorMask() { + const auto& regs = Core::System::GetInstance().GPU().Maxwell3D().regs; + for (size_t i = 0; i < Tegra::Engines::Maxwell3D::Regs::NumRenderTargets; i++) { + const auto& source = regs.color_mask[regs.color_mask_common ? 0 : i]; + auto& dest = state.color_mask[i]; + dest.red_enabled = (source.R == 0) ? GL_FALSE : GL_TRUE; + dest.green_enabled = (source.G == 0) ? GL_FALSE : GL_TRUE; + dest.blue_enabled = (source.B == 0) ? GL_FALSE : GL_TRUE; + dest.alpha_enabled = (source.A == 0) ? GL_FALSE : GL_TRUE; + } } void RasterizerOpenGL::SyncBlendState() { const auto& regs = Core::System::GetInstance().GPU().Maxwell3D().regs; - // TODO(Subv): Support more than just render target 0. - state.blend.enabled = regs.blend.enable[0] != 0; - - if (!state.blend.enabled) + state.blend_color.red = regs.blend_color.r; + state.blend_color.green = regs.blend_color.g; + state.blend_color.blue = regs.blend_color.b; + state.blend_color.alpha = regs.blend_color.a; + + state.independant_blend.enabled = regs.independent_blend_enable; + if (!state.independant_blend.enabled) { + auto& blend = state.blend[0]; + blend.enabled = regs.blend.enable[0] != 0; + blend.separate_alpha = regs.blend.separate_alpha; + blend.rgb_equation = MaxwellToGL::BlendEquation(regs.blend.equation_rgb); + blend.src_rgb_func = MaxwellToGL::BlendFunc(regs.blend.factor_source_rgb); + blend.dst_rgb_func = MaxwellToGL::BlendFunc(regs.blend.factor_dest_rgb); + if (blend.separate_alpha) { + blend.a_equation = MaxwellToGL::BlendEquation(regs.blend.equation_a); + blend.src_a_func = MaxwellToGL::BlendFunc(regs.blend.factor_source_a); + blend.dst_a_func = MaxwellToGL::BlendFunc(regs.blend.factor_dest_a); + } + for (size_t i = 1; i < Tegra::Engines::Maxwell3D::Regs::NumRenderTargets; i++) { + state.blend[i].enabled = false; + } return; + } - ASSERT_MSG(regs.logic_op.enable == 0, - "Blending and logic op can't be enabled at the same time."); - - ASSERT_MSG(regs.independent_blend_enable == 1, "Only independent blending is implemented"); - ASSERT_MSG(!regs.independent_blend[0].separate_alpha, "Unimplemented"); - state.blend.rgb_equation = MaxwellToGL::BlendEquation(regs.independent_blend[0].equation_rgb); - state.blend.src_rgb_func = MaxwellToGL::BlendFunc(regs.independent_blend[0].factor_source_rgb); - state.blend.dst_rgb_func = MaxwellToGL::BlendFunc(regs.independent_blend[0].factor_dest_rgb); - state.blend.a_equation = MaxwellToGL::BlendEquation(regs.independent_blend[0].equation_a); - state.blend.src_a_func = MaxwellToGL::BlendFunc(regs.independent_blend[0].factor_source_a); - state.blend.dst_a_func = MaxwellToGL::BlendFunc(regs.independent_blend[0].factor_dest_a); + for (size_t i = 0; i < Tegra::Engines::Maxwell3D::Regs::NumRenderTargets; i++) { + auto& blend = state.blend[i]; + blend.enabled = regs.blend.enable[i] != 0; + if (!blend.enabled) + continue; + blend.separate_alpha = regs.independent_blend[i].separate_alpha; + blend.rgb_equation = MaxwellToGL::BlendEquation(regs.independent_blend[i].equation_rgb); + blend.src_rgb_func = MaxwellToGL::BlendFunc(regs.independent_blend[i].factor_source_rgb); + blend.dst_rgb_func = MaxwellToGL::BlendFunc(regs.independent_blend[i].factor_dest_rgb); + if (blend.separate_alpha) { + blend.a_equation = MaxwellToGL::BlendEquation(regs.independent_blend[i].equation_a); + blend.src_a_func = MaxwellToGL::BlendFunc(regs.independent_blend[i].factor_source_a); + blend.dst_a_func = MaxwellToGL::BlendFunc(regs.independent_blend[i].factor_dest_a); + } + } } void RasterizerOpenGL::SyncLogicOpState() { @@ -1031,19 +1090,19 @@ void RasterizerOpenGL::SyncLogicOpState() { } void RasterizerOpenGL::SyncScissorTest() { + // TODO: what is the correct behavior here, a single scissor for all targets + // or scissor disabled for the rest of the targets? const auto& regs = Core::System::GetInstance().GPU().Maxwell3D().regs; - state.scissor.enabled = (regs.scissor_test.enable != 0); - // TODO(Blinkhawk): Figure if the hardware supports scissor testing per viewport and how it's - // implemented. - if (regs.scissor_test.enable != 0) { - const u32 width = regs.scissor_test.max_x - regs.scissor_test.min_x; - const u32 height = regs.scissor_test.max_y - regs.scissor_test.min_y; - state.scissor.x = regs.scissor_test.min_x; - state.scissor.y = regs.scissor_test.min_y; - state.scissor.width = width; - state.scissor.height = height; + if (regs.scissor_test.enable == 0) { + return; } + const u32 width = regs.scissor_test.max_x - regs.scissor_test.min_x; + const u32 height = regs.scissor_test.max_y - regs.scissor_test.min_y; + state.scissor.x = regs.scissor_test.min_x; + state.scissor.y = regs.scissor_test.min_y; + state.scissor.width = width; + state.scissor.height = height; } void RasterizerOpenGL::SyncTransformFeedback() { @@ -1068,9 +1127,8 @@ void RasterizerOpenGL::CheckAlphaTests() { const auto& regs = Core::System::GetInstance().GPU().Maxwell3D().regs; if (regs.alpha_test_enabled != 0 && regs.rt_control.count > 1) { - LOG_CRITICAL( - Render_OpenGL, - "Alpha Testing is enabled with Multiple Render Targets, this behavior is undefined."); + LOG_CRITICAL(Render_OpenGL, "Alpha Testing is enabled with Multiple Render Targets, " + "this behavior is undefined."); UNREACHABLE(); } } diff --git a/src/video_core/renderer_opengl/gl_rasterizer.h b/src/video_core/renderer_opengl/gl_rasterizer.h index 47097c569..8ef0f6c12 100644 --- a/src/video_core/renderer_opengl/gl_rasterizer.h +++ b/src/video_core/renderer_opengl/gl_rasterizer.h @@ -88,7 +88,7 @@ private: /// SamplerInfo struct. void Create(); /// Syncs the sampler object with the config, updating any necessary state. - void SyncWithConfig(const Tegra::Texture::TSCEntry& config); + void SyncWithConfig(const Tegra::Texture::FullTextureInfo& info); private: Tegra::Texture::TextureFilter mag_filter; @@ -109,8 +109,8 @@ private: * @param preserve_contents If true, tries to preserve data from a previously used framebuffer. * @param single_color_target Specifies if a single color buffer target should be used. */ - void ConfigureFramebuffers(bool use_color_fb = true, bool using_depth_fb = true, - bool preserve_contents = true, + void ConfigureFramebuffers(OpenGLState& current_state, bool use_color_fb = true, + bool using_depth_fb = true, bool preserve_contents = true, std::optional<std::size_t> single_color_target = {}); /* @@ -133,8 +133,8 @@ private: u32 SetupTextures(Tegra::Engines::Maxwell3D::Regs::ShaderStage stage, Shader& shader, GLenum primitive_mode, u32 current_unit); - /// Syncs the viewport to match the guest state - void SyncViewport(); + /// Syncs the viewport and depth range to match the guest state + void SyncViewport(OpenGLState& current_state); /// Syncs the clip enabled status to match the guest state void SyncClipEnabled(); @@ -148,9 +148,6 @@ private: /// Syncs the primitve restart to match the guest state void SyncPrimitiveRestart(); - /// Syncs the depth range to match the guest state - void SyncDepthRange(); - /// Syncs the depth test state to match the guest state void SyncDepthTestState(); @@ -172,6 +169,9 @@ private: /// Syncs the point state to match the guest state void SyncPointState(); + /// Syncs Color Mask + void SyncColorMask(); + /// Check asserts for alpha testing. void CheckAlphaTests(); @@ -207,7 +207,8 @@ private: std::size_t CalculateIndexBufferSize() const; - void SetupVertexArrays(); + void SetupVertexFormat(); + void SetupVertexBuffer(); DrawParameters SetupDraw(); diff --git a/src/video_core/renderer_opengl/gl_rasterizer_cache.cpp b/src/video_core/renderer_opengl/gl_rasterizer_cache.cpp index f194a7687..b44ecfa1c 100644 --- a/src/video_core/renderer_opengl/gl_rasterizer_cache.cpp +++ b/src/video_core/renderer_opengl/gl_rasterizer_cache.cpp @@ -15,7 +15,9 @@ #include "core/memory.h" #include "core/settings.h" #include "video_core/engines/maxwell_3d.h" +#include "video_core/renderer_opengl/gl_rasterizer.h" #include "video_core/renderer_opengl/gl_rasterizer_cache.h" +#include "video_core/renderer_opengl/gl_state.h" #include "video_core/renderer_opengl/utils.h" #include "video_core/surface.h" #include "video_core/textures/astc.h" @@ -58,16 +60,14 @@ void SurfaceParams::InitCacheParameters(Tegra::GPUVAddr gpu_addr_) { std::size_t SurfaceParams::InnerMipmapMemorySize(u32 mip_level, bool force_gl, bool layer_only, bool uncompressed) const { - const u32 compression_factor{GetCompressionFactor(pixel_format)}; + const u32 tile_x{GetDefaultBlockWidth(pixel_format)}; + const u32 tile_y{GetDefaultBlockHeight(pixel_format)}; const u32 bytes_per_pixel{GetBytesPerPixel(pixel_format)}; u32 m_depth = (layer_only ? 1U : depth); u32 m_width = MipWidth(mip_level); u32 m_height = MipHeight(mip_level); - m_width = uncompressed ? m_width - : std::max(1U, (m_width + compression_factor - 1) / compression_factor); - m_height = uncompressed - ? m_height - : std::max(1U, (m_height + compression_factor - 1) / compression_factor); + m_width = uncompressed ? m_width : std::max(1U, (m_width + tile_x - 1) / tile_x); + m_height = uncompressed ? m_height : std::max(1U, (m_height + tile_y - 1) / tile_y); m_depth = std::max(1U, m_depth >> mip_level); u32 m_block_height = MipBlockHeight(mip_level); u32 m_block_depth = MipBlockDepth(mip_level); @@ -128,6 +128,13 @@ std::size_t SurfaceParams::InnerMemorySize(bool force_gl, bool layer_only, params.target = SurfaceTarget::Texture2D; } break; + case SurfaceTarget::TextureCubeArray: + params.depth = config.tic.Depth() * 6; + if (!entry.IsArray()) { + ASSERT(params.depth == 6); + params.target = SurfaceTarget::TextureCubemap; + } + break; default: LOG_CRITICAL(HW_GPU, "Unknown depth for target={}", static_cast<u32>(params.target)); UNREACHABLE(); @@ -305,6 +312,10 @@ static constexpr std::array<FormatTuple, VideoCore::Surface::MaxPixelFormat> tex {GL_SRGB8_ALPHA8, GL_RGBA, GL_UNSIGNED_BYTE, ComponentType::UNorm, false}, // ASTC_2D_8X8_SRGB {GL_SRGB8_ALPHA8, GL_RGBA, GL_UNSIGNED_BYTE, ComponentType::UNorm, false}, // ASTC_2D_8X5_SRGB {GL_SRGB8_ALPHA8, GL_RGBA, GL_UNSIGNED_BYTE, ComponentType::UNorm, false}, // ASTC_2D_5X4_SRGB + {GL_RGBA8, GL_RGBA, GL_UNSIGNED_BYTE, ComponentType::UNorm, false}, // ASTC_2D_5X5 + {GL_SRGB8_ALPHA8, GL_RGBA, GL_UNSIGNED_BYTE, ComponentType::UNorm, false}, // ASTC_2D_5X5_SRGB + {GL_RGBA8, GL_RGBA, GL_UNSIGNED_BYTE, ComponentType::UNorm, false}, // ASTC_2D_10X8 + {GL_SRGB8_ALPHA8, GL_RGBA, GL_UNSIGNED_BYTE, ComponentType::UNorm, false}, // ASTC_2D_10X8_SRGB // Depth formats {GL_DEPTH_COMPONENT32F, GL_DEPTH_COMPONENT, GL_FLOAT, ComponentType::Float, false}, // Z32F @@ -334,6 +345,8 @@ static GLenum SurfaceTargetToGL(SurfaceTarget target) { return GL_TEXTURE_2D_ARRAY; case SurfaceTarget::TextureCubemap: return GL_TEXTURE_CUBE_MAP; + case SurfaceTarget::TextureCubeArray: + return GL_TEXTURE_CUBE_MAP_ARRAY_ARB; } LOG_CRITICAL(Render_OpenGL, "Unimplemented texture target={}", static_cast<u32>(target)); UNREACHABLE(); @@ -364,15 +377,18 @@ void MortonCopy(u32 stride, u32 block_height, u32 height, u32 block_depth, u32 d // With the BCn formats (DXT and DXN), each 4x4 tile is swizzled instead of just individual // pixel values. - const u32 tile_size{IsFormatBCn(format) ? 4U : 1U}; + const u32 tile_size_x{GetDefaultBlockWidth(format)}; + const u32 tile_size_y{GetDefaultBlockHeight(format)}; if (morton_to_gl) { - const std::vector<u8> data = Tegra::Texture::UnswizzleTexture( - addr, tile_size, bytes_per_pixel, stride, height, depth, block_height, block_depth); + const std::vector<u8> data = + Tegra::Texture::UnswizzleTexture(addr, tile_size_x, tile_size_y, bytes_per_pixel, + stride, height, depth, block_height, block_depth); const std::size_t size_to_copy{std::min(gl_buffer_size, data.size())}; memcpy(gl_buffer, data.data(), size_to_copy); } else { - Tegra::Texture::CopySwizzledData(stride / tile_size, height / tile_size, depth, + Tegra::Texture::CopySwizzledData((stride + tile_size_x - 1) / tile_size_x, + (height + tile_size_y - 1) / tile_size_y, depth, bytes_per_pixel, bytes_per_pixel, Memory::GetPointer(addr), gl_buffer, false, block_height, block_depth); } @@ -440,6 +456,10 @@ static constexpr GLConversionArray morton_to_gl_fns = { MortonCopy<true, PixelFormat::ASTC_2D_8X8_SRGB>, MortonCopy<true, PixelFormat::ASTC_2D_8X5_SRGB>, MortonCopy<true, PixelFormat::ASTC_2D_5X4_SRGB>, + MortonCopy<true, PixelFormat::ASTC_2D_5X5>, + MortonCopy<true, PixelFormat::ASTC_2D_5X5_SRGB>, + MortonCopy<true, PixelFormat::ASTC_2D_10X8>, + MortonCopy<true, PixelFormat::ASTC_2D_10X8_SRGB>, MortonCopy<true, PixelFormat::Z32F>, MortonCopy<true, PixelFormat::Z16>, MortonCopy<true, PixelFormat::Z24S8>, @@ -508,6 +528,10 @@ static constexpr GLConversionArray gl_to_morton_fns = { nullptr, nullptr, nullptr, + nullptr, + nullptr, + nullptr, + nullptr, MortonCopy<false, PixelFormat::Z32F>, MortonCopy<false, PixelFormat::Z16>, MortonCopy<false, PixelFormat::Z24S8>, @@ -526,8 +550,8 @@ void SwizzleFunc(const GLConversionArray& functions, const SurfaceParams& params if (params.is_layered) { u64 offset = params.GetMipmapLevelOffset(mip_level); u64 offset_gl = 0; - u64 layer_size = params.LayerMemorySize(); - u64 gl_size = params.LayerSizeGL(mip_level); + const u64 layer_size = params.LayerMemorySize(); + const u64 gl_size = params.LayerSizeGL(mip_level); for (u32 i = 0; i < params.depth; i++) { functions[static_cast<std::size_t>(params.pixel_format)]( params.MipWidth(mip_level), params.MipBlockHeight(mip_level), @@ -537,7 +561,7 @@ void SwizzleFunc(const GLConversionArray& functions, const SurfaceParams& params offset_gl += gl_size; } } else { - u64 offset = params.GetMipmapLevelOffset(mip_level); + const u64 offset = params.GetMipmapLevelOffset(mip_level); functions[static_cast<std::size_t>(params.pixel_format)]( params.MipWidth(mip_level), params.MipBlockHeight(mip_level), params.MipHeight(mip_level), params.MipBlockDepth(mip_level), depth, gl_buffer.data(), @@ -545,9 +569,11 @@ void SwizzleFunc(const GLConversionArray& functions, const SurfaceParams& params } } +MICROPROFILE_DEFINE(OpenGL_BlitSurface, "OpenGL", "BlitSurface", MP_RGB(128, 192, 64)); static bool BlitSurface(const Surface& src_surface, const Surface& dst_surface, GLuint read_fb_handle, GLuint draw_fb_handle, GLenum src_attachment = 0, GLenum dst_attachment = 0, std::size_t cubemap_face = 0) { + MICROPROFILE_SCOPE(OpenGL_BlitSurface); const auto& src_params{src_surface->GetSurfaceParams()}; const auto& dst_params{dst_surface->GetSurfaceParams()}; @@ -560,7 +586,7 @@ static bool BlitSurface(const Surface& src_surface, const Surface& dst_surface, state.draw.draw_framebuffer = draw_fb_handle; // Set sRGB enabled if the destination surfaces need it state.framebuffer_srgb.enabled = dst_params.srgb_conversion; - state.Apply(); + state.ApplyFramebufferState(); u32 buffers{}; @@ -687,18 +713,20 @@ static void FastCopySurface(const Surface& src_surface, const Surface& dst_surfa 0, 0, width, height, 1); } +MICROPROFILE_DEFINE(OpenGL_CopySurface, "OpenGL", "CopySurface", MP_RGB(128, 192, 64)); static void CopySurface(const Surface& src_surface, const Surface& dst_surface, - GLuint copy_pbo_handle, GLenum src_attachment = 0, - GLenum dst_attachment = 0, std::size_t cubemap_face = 0) { + const GLuint copy_pbo_handle, const GLenum src_attachment = 0, + const GLenum dst_attachment = 0, const std::size_t cubemap_face = 0) { + MICROPROFILE_SCOPE(OpenGL_CopySurface); ASSERT_MSG(dst_attachment == 0, "Unimplemented"); const auto& src_params{src_surface->GetSurfaceParams()}; const auto& dst_params{dst_surface->GetSurfaceParams()}; - auto source_format = GetFormatTuple(src_params.pixel_format, src_params.component_type); - auto dest_format = GetFormatTuple(dst_params.pixel_format, dst_params.component_type); + const auto source_format = GetFormatTuple(src_params.pixel_format, src_params.component_type); + const auto dest_format = GetFormatTuple(dst_params.pixel_format, dst_params.component_type); - std::size_t buffer_size = std::max(src_params.size_in_bytes, dst_params.size_in_bytes); + const std::size_t buffer_size = std::max(src_params.size_in_bytes, dst_params.size_in_bytes); glBindBuffer(GL_PIXEL_PACK_BUFFER, copy_pbo_handle); glBufferData(GL_PIXEL_PACK_BUFFER, buffer_size, nullptr, GL_STREAM_DRAW_ARB); @@ -722,13 +750,10 @@ static void CopySurface(const Surface& src_surface, const Surface& dst_surface, LOG_DEBUG(HW_GPU, "Trying to upload extra texture data from the CPU during " "reinterpretation but the texture is tiled."); } - std::size_t remaining_size = dst_params.size_in_bytes - src_params.size_in_bytes; - std::vector<u8> data(remaining_size); - std::memcpy(data.data(), Memory::GetPointer(dst_params.addr + src_params.size_in_bytes), - data.size()); + const std::size_t remaining_size = dst_params.size_in_bytes - src_params.size_in_bytes; glBufferSubData(GL_PIXEL_PACK_BUFFER, src_params.size_in_bytes, remaining_size, - data.data()); + Memory::GetPointer(dst_params.addr + src_params.size_in_bytes)); } glBindBuffer(GL_PIXEL_PACK_BUFFER, 0); @@ -754,6 +779,7 @@ static void CopySurface(const Surface& src_surface, const Surface& dst_surface, break; case SurfaceTarget::Texture3D: case SurfaceTarget::Texture2DArray: + case SurfaceTarget::TextureCubeArray: glTextureSubImage3D(dst_surface->Texture().handle, 0, 0, 0, 0, width, height, static_cast<GLsizei>(dst_params.depth), dest_format.format, dest_format.type, nullptr); @@ -806,6 +832,7 @@ CachedSurface::CachedSurface(const SurfaceParams& params) break; case SurfaceTarget::Texture3D: case SurfaceTarget::Texture2DArray: + case SurfaceTarget::TextureCubeArray: glTexStorage3D(SurfaceTargetToGL(params.target), params.max_mip_level, format_tuple.internal_format, rect.GetWidth(), rect.GetHeight(), params.depth); @@ -897,21 +924,26 @@ static void ConvertG8R8ToR8G8(std::vector<u8>& data, u32 width, u32 height) { * typical desktop GPUs. */ static void ConvertFormatAsNeeded_LoadGLBuffer(std::vector<u8>& data, PixelFormat pixel_format, - u32 width, u32 height) { + u32 width, u32 height, u32 depth) { switch (pixel_format) { case PixelFormat::ASTC_2D_4X4: case PixelFormat::ASTC_2D_8X8: case PixelFormat::ASTC_2D_8X5: case PixelFormat::ASTC_2D_5X4: + case PixelFormat::ASTC_2D_5X5: case PixelFormat::ASTC_2D_4X4_SRGB: case PixelFormat::ASTC_2D_8X8_SRGB: case PixelFormat::ASTC_2D_8X5_SRGB: - case PixelFormat::ASTC_2D_5X4_SRGB: { + case PixelFormat::ASTC_2D_5X4_SRGB: + case PixelFormat::ASTC_2D_5X5_SRGB: + case PixelFormat::ASTC_2D_10X8: + case PixelFormat::ASTC_2D_10X8_SRGB: { // Convert ASTC pixel formats to RGBA8, as most desktop GPUs do not support ASTC. u32 block_width{}; u32 block_height{}; std::tie(block_width, block_height) = GetASTCBlockSize(pixel_format); - data = Tegra::Texture::ASTC::Decompress(data, width, height, block_width, block_height); + data = + Tegra::Texture::ASTC::Decompress(data, width, height, depth, block_width, block_height); break; } case PixelFormat::S8Z24: @@ -940,7 +972,11 @@ static void ConvertFormatAsNeeded_FlushGLBuffer(std::vector<u8>& data, PixelForm case PixelFormat::ASTC_2D_4X4: case PixelFormat::ASTC_2D_8X8: case PixelFormat::ASTC_2D_4X4_SRGB: - case PixelFormat::ASTC_2D_8X8_SRGB: { + case PixelFormat::ASTC_2D_8X8_SRGB: + case PixelFormat::ASTC_2D_5X5: + case PixelFormat::ASTC_2D_5X5_SRGB: + case PixelFormat::ASTC_2D_10X8: + case PixelFormat::ASTC_2D_10X8_SRGB: { LOG_CRITICAL(HW_GPU, "Conversion of format {} after texture flushing is not implemented", static_cast<u32>(pixel_format)); UNREACHABLE(); @@ -953,7 +989,7 @@ static void ConvertFormatAsNeeded_FlushGLBuffer(std::vector<u8>& data, PixelForm } } -MICROPROFILE_DEFINE(OpenGL_SurfaceLoad, "OpenGL", "Surface Load", MP_RGB(128, 64, 192)); +MICROPROFILE_DEFINE(OpenGL_SurfaceLoad, "OpenGL", "Surface Load", MP_RGB(128, 192, 64)); void CachedSurface::LoadGLBuffer() { MICROPROFILE_SCOPE(OpenGL_SurfaceLoad); gl_buffer.resize(params.max_mip_level); @@ -971,7 +1007,7 @@ void CachedSurface::LoadGLBuffer() { } for (u32 i = 0; i < params.max_mip_level; i++) ConvertFormatAsNeeded_LoadGLBuffer(gl_buffer[i], params.pixel_format, params.MipWidth(i), - params.MipHeight(i)); + params.MipHeight(i), params.MipDepth(i)); } MICROPROFILE_DEFINE(OpenGL_SurfaceFlush, "OpenGL", "Surface Flush", MP_RGB(128, 192, 64)); @@ -1055,6 +1091,7 @@ void CachedSurface::UploadGLMipmapTexture(u32 mip_map, GLuint read_fb_handle, &gl_buffer[mip_map][buffer_offset]); break; case SurfaceTarget::Texture2DArray: + case SurfaceTarget::TextureCubeArray: glCompressedTexImage3D(SurfaceTargetToGL(params.target), mip_map, tuple.internal_format, static_cast<GLsizei>(params.MipWidth(mip_map)), static_cast<GLsizei>(params.MipHeight(mip_map)), @@ -1104,6 +1141,7 @@ void CachedSurface::UploadGLMipmapTexture(u32 mip_map, GLuint read_fb_handle, tuple.format, tuple.type, &gl_buffer[mip_map][buffer_offset]); break; case SurfaceTarget::Texture2DArray: + case SurfaceTarget::TextureCubeArray: glTexSubImage3D(SurfaceTargetToGL(params.target), mip_map, x0, y0, 0, static_cast<GLsizei>(rect.GetWidth()), static_cast<GLsizei>(rect.GetHeight()), params.depth, tuple.format, @@ -1133,7 +1171,7 @@ void CachedSurface::UploadGLMipmapTexture(u32 mip_map, GLuint read_fb_handle, glPixelStorei(GL_UNPACK_ROW_LENGTH, 0); } -MICROPROFILE_DEFINE(OpenGL_TextureUL, "OpenGL", "Texture Upload", MP_RGB(128, 64, 192)); +MICROPROFILE_DEFINE(OpenGL_TextureUL, "OpenGL", "Texture Upload", MP_RGB(128, 192, 64)); void CachedSurface::UploadGLTexture(GLuint read_fb_handle, GLuint draw_fb_handle) { if (params.type == SurfaceType::Fill) return; @@ -1144,7 +1182,8 @@ void CachedSurface::UploadGLTexture(GLuint read_fb_handle, GLuint draw_fb_handle UploadGLMipmapTexture(i, read_fb_handle, draw_fb_handle); } -RasterizerCacheOpenGL::RasterizerCacheOpenGL() { +RasterizerCacheOpenGL::RasterizerCacheOpenGL(RasterizerOpenGL& rasterizer) + : RasterizerCache{rasterizer} { read_framebuffer.Create(); draw_framebuffer.Create(); copy_pbo.Create(); @@ -1306,6 +1345,8 @@ Surface RasterizerCacheOpenGL::RecreateSurface(const Surface& old_surface, break; case SurfaceTarget::TextureCubemap: case SurfaceTarget::Texture3D: + case SurfaceTarget::Texture2DArray: + case SurfaceTarget::TextureCubeArray: AccurateCopySurface(old_surface, new_surface); break; default: diff --git a/src/video_core/renderer_opengl/gl_rasterizer_cache.h b/src/video_core/renderer_opengl/gl_rasterizer_cache.h index f255f4419..494f6b903 100644 --- a/src/video_core/renderer_opengl/gl_rasterizer_cache.h +++ b/src/video_core/renderer_opengl/gl_rasterizer_cache.h @@ -49,6 +49,8 @@ struct SurfaceParams { return "Texture2DArray"; case SurfaceTarget::TextureCubemap: return "TextureCubemap"; + case SurfaceTarget::TextureCubeArray: + return "TextureCubeArray"; default: LOG_CRITICAL(HW_GPU, "Unimplemented surface_target={}", static_cast<u32>(target)); UNREACHABLE(); @@ -139,7 +141,7 @@ struct SurfaceParams { } u32 MipDepth(u32 mip_level) const { - return std::max(1U, depth >> mip_level); + return is_layered ? depth : std::max(1U, depth >> mip_level); } // Auto block resizing algorithm from: @@ -262,6 +264,8 @@ struct hash<SurfaceReserveKey> { namespace OpenGL { +class RasterizerOpenGL; + class CachedSurface final : public RasterizerCacheObject { public: CachedSurface(const SurfaceParams& params); @@ -309,7 +313,7 @@ private: class RasterizerCacheOpenGL final : public RasterizerCache<Surface> { public: - RasterizerCacheOpenGL(); + explicit RasterizerCacheOpenGL(RasterizerOpenGL& rasterizer); /// Get a surface based on the texture configuration Surface GetTextureSurface(const Tegra::Texture::FullTextureInfo& config, diff --git a/src/video_core/renderer_opengl/gl_resource_manager.cpp b/src/video_core/renderer_opengl/gl_resource_manager.cpp new file mode 100644 index 000000000..c17d5ac00 --- /dev/null +++ b/src/video_core/renderer_opengl/gl_resource_manager.cpp @@ -0,0 +1,186 @@ +// Copyright 2015 Citra Emulator Project +// Licensed under GPLv2 or any later version +// Refer to the license.txt file included. + +#include <utility> +#include <glad/glad.h> +#include "common/common_types.h" +#include "common/microprofile.h" +#include "video_core/renderer_opengl/gl_resource_manager.h" +#include "video_core/renderer_opengl/gl_shader_util.h" +#include "video_core/renderer_opengl/gl_state.h" + +MICROPROFILE_DEFINE(OpenGL_ResourceCreation, "OpenGL", "Resource Creation", MP_RGB(128, 128, 192)); +MICROPROFILE_DEFINE(OpenGL_ResourceDeletion, "OpenGL", "Resource Deletion", MP_RGB(128, 128, 192)); + +namespace OpenGL { + +void OGLTexture::Create() { + if (handle != 0) + return; + + MICROPROFILE_SCOPE(OpenGL_ResourceCreation); + glGenTextures(1, &handle); +} + +void OGLTexture::Release() { + if (handle == 0) + return; + + MICROPROFILE_SCOPE(OpenGL_ResourceDeletion); + glDeleteTextures(1, &handle); + OpenGLState::GetCurState().UnbindTexture(handle).Apply(); + handle = 0; +} + +void OGLSampler::Create() { + if (handle != 0) + return; + + MICROPROFILE_SCOPE(OpenGL_ResourceCreation); + glGenSamplers(1, &handle); +} + +void OGLSampler::Release() { + if (handle == 0) + return; + + MICROPROFILE_SCOPE(OpenGL_ResourceDeletion); + glDeleteSamplers(1, &handle); + OpenGLState::GetCurState().ResetSampler(handle).Apply(); + handle = 0; +} + +void OGLShader::Create(const char* source, GLenum type) { + if (handle != 0) + return; + if (source == nullptr) + return; + + MICROPROFILE_SCOPE(OpenGL_ResourceCreation); + handle = GLShader::LoadShader(source, type); +} + +void OGLShader::Release() { + if (handle == 0) + return; + + MICROPROFILE_SCOPE(OpenGL_ResourceDeletion); + glDeleteShader(handle); + handle = 0; +} + +void OGLProgram::CreateFromSource(const char* vert_shader, const char* geo_shader, + const char* frag_shader, bool separable_program) { + OGLShader vert, geo, frag; + if (vert_shader) + vert.Create(vert_shader, GL_VERTEX_SHADER); + if (geo_shader) + geo.Create(geo_shader, GL_GEOMETRY_SHADER); + if (frag_shader) + frag.Create(frag_shader, GL_FRAGMENT_SHADER); + + MICROPROFILE_SCOPE(OpenGL_ResourceCreation); + Create(separable_program, vert.handle, geo.handle, frag.handle); +} + +void OGLProgram::Release() { + if (handle == 0) + return; + + MICROPROFILE_SCOPE(OpenGL_ResourceDeletion); + glDeleteProgram(handle); + OpenGLState::GetCurState().ResetProgram(handle).Apply(); + handle = 0; +} + +void OGLPipeline::Create() { + if (handle != 0) + return; + + MICROPROFILE_SCOPE(OpenGL_ResourceCreation); + glGenProgramPipelines(1, &handle); +} + +void OGLPipeline::Release() { + if (handle == 0) + return; + + MICROPROFILE_SCOPE(OpenGL_ResourceDeletion); + glDeleteProgramPipelines(1, &handle); + OpenGLState::GetCurState().ResetPipeline(handle).Apply(); + handle = 0; +} + +void OGLBuffer::Create() { + if (handle != 0) + return; + + MICROPROFILE_SCOPE(OpenGL_ResourceCreation); + glGenBuffers(1, &handle); +} + +void OGLBuffer::Release() { + if (handle == 0) + return; + + MICROPROFILE_SCOPE(OpenGL_ResourceDeletion); + glDeleteBuffers(1, &handle); + OpenGLState::GetCurState().ResetBuffer(handle).Apply(); + handle = 0; +} + +void OGLSync::Create() { + if (handle != 0) + return; + + // Don't profile here, this one is expected to happen ingame. + handle = glFenceSync(GL_SYNC_GPU_COMMANDS_COMPLETE, 0); +} + +void OGLSync::Release() { + if (handle == 0) + return; + + // Don't profile here, this one is expected to happen ingame. + glDeleteSync(handle); + handle = 0; +} + +void OGLVertexArray::Create() { + if (handle != 0) + return; + + MICROPROFILE_SCOPE(OpenGL_ResourceCreation); + glGenVertexArrays(1, &handle); +} + +void OGLVertexArray::Release() { + if (handle == 0) + return; + + MICROPROFILE_SCOPE(OpenGL_ResourceDeletion); + glDeleteVertexArrays(1, &handle); + OpenGLState::GetCurState().ResetVertexArray(handle).Apply(); + handle = 0; +} + +void OGLFramebuffer::Create() { + if (handle != 0) + return; + + MICROPROFILE_SCOPE(OpenGL_ResourceCreation); + glGenFramebuffers(1, &handle); +} + +void OGLFramebuffer::Release() { + if (handle == 0) + return; + + MICROPROFILE_SCOPE(OpenGL_ResourceDeletion); + glDeleteFramebuffers(1, &handle); + OpenGLState::GetCurState().ResetFramebuffer(handle).Apply(); + handle = 0; +} + +} // namespace OpenGL diff --git a/src/video_core/renderer_opengl/gl_resource_manager.h b/src/video_core/renderer_opengl/gl_resource_manager.h index 3bc1b83b5..e33f1e973 100644 --- a/src/video_core/renderer_opengl/gl_resource_manager.h +++ b/src/video_core/renderer_opengl/gl_resource_manager.h @@ -8,7 +8,6 @@ #include <glad/glad.h> #include "common/common_types.h" #include "video_core/renderer_opengl/gl_shader_util.h" -#include "video_core/renderer_opengl/gl_state.h" namespace OpenGL { @@ -29,20 +28,10 @@ public: } /// Creates a new internal OpenGL resource and stores the handle - void Create() { - if (handle != 0) - return; - glGenTextures(1, &handle); - } + void Create(); /// Deletes the internal OpenGL resource - void Release() { - if (handle == 0) - return; - glDeleteTextures(1, &handle); - OpenGLState::GetCurState().UnbindTexture(handle).Apply(); - handle = 0; - } + void Release(); GLuint handle = 0; }; @@ -64,20 +53,10 @@ public: } /// Creates a new internal OpenGL resource and stores the handle - void Create() { - if (handle != 0) - return; - glGenSamplers(1, &handle); - } + void Create(); /// Deletes the internal OpenGL resource - void Release() { - if (handle == 0) - return; - glDeleteSamplers(1, &handle); - OpenGLState::GetCurState().ResetSampler(handle).Apply(); - handle = 0; - } + void Release(); GLuint handle = 0; }; @@ -98,20 +77,9 @@ public: return *this; } - void Create(const char* source, GLenum type) { - if (handle != 0) - return; - if (source == nullptr) - return; - handle = GLShader::LoadShader(source, type); - } + void Create(const char* source, GLenum type); - void Release() { - if (handle == 0) - return; - glDeleteShader(handle); - handle = 0; - } + void Release(); GLuint handle = 0; }; @@ -141,25 +109,10 @@ public: /// Creates a new internal OpenGL resource and stores the handle void CreateFromSource(const char* vert_shader, const char* geo_shader, const char* frag_shader, - bool separable_program = false) { - OGLShader vert, geo, frag; - if (vert_shader) - vert.Create(vert_shader, GL_VERTEX_SHADER); - if (geo_shader) - geo.Create(geo_shader, GL_GEOMETRY_SHADER); - if (frag_shader) - frag.Create(frag_shader, GL_FRAGMENT_SHADER); - Create(separable_program, vert.handle, geo.handle, frag.handle); - } + bool separable_program = false); /// Deletes the internal OpenGL resource - void Release() { - if (handle == 0) - return; - glDeleteProgram(handle); - OpenGLState::GetCurState().ResetProgram(handle).Apply(); - handle = 0; - } + void Release(); GLuint handle = 0; }; @@ -178,20 +131,10 @@ public: } /// Creates a new internal OpenGL resource and stores the handle - void Create() { - if (handle != 0) - return; - glGenProgramPipelines(1, &handle); - } + void Create(); /// Deletes the internal OpenGL resource - void Release() { - if (handle == 0) - return; - glDeleteProgramPipelines(1, &handle); - OpenGLState::GetCurState().ResetPipeline(handle).Apply(); - handle = 0; - } + void Release(); GLuint handle = 0; }; @@ -213,20 +156,10 @@ public: } /// Creates a new internal OpenGL resource and stores the handle - void Create() { - if (handle != 0) - return; - glGenBuffers(1, &handle); - } + void Create(); /// Deletes the internal OpenGL resource - void Release() { - if (handle == 0) - return; - glDeleteBuffers(1, &handle); - OpenGLState::GetCurState().ResetBuffer(handle).Apply(); - handle = 0; - } + void Release(); GLuint handle = 0; }; @@ -247,19 +180,10 @@ public: } /// Creates a new internal OpenGL resource and stores the handle - void Create() { - if (handle != 0) - return; - handle = glFenceSync(GL_SYNC_GPU_COMMANDS_COMPLETE, 0); - } + void Create(); /// Deletes the internal OpenGL resource - void Release() { - if (handle == 0) - return; - glDeleteSync(handle); - handle = 0; - } + void Release(); GLsync handle = 0; }; @@ -281,20 +205,10 @@ public: } /// Creates a new internal OpenGL resource and stores the handle - void Create() { - if (handle != 0) - return; - glGenVertexArrays(1, &handle); - } + void Create(); /// Deletes the internal OpenGL resource - void Release() { - if (handle == 0) - return; - glDeleteVertexArrays(1, &handle); - OpenGLState::GetCurState().ResetVertexArray(handle).Apply(); - handle = 0; - } + void Release(); GLuint handle = 0; }; @@ -316,20 +230,10 @@ public: } /// Creates a new internal OpenGL resource and stores the handle - void Create() { - if (handle != 0) - return; - glGenFramebuffers(1, &handle); - } + void Create(); /// Deletes the internal OpenGL resource - void Release() { - if (handle == 0) - return; - glDeleteFramebuffers(1, &handle); - OpenGLState::GetCurState().ResetFramebuffer(handle).Apply(); - handle = 0; - } + void Release(); GLuint handle = 0; }; diff --git a/src/video_core/renderer_opengl/gl_shader_cache.cpp b/src/video_core/renderer_opengl/gl_shader_cache.cpp index 9522fd344..a85a7c0c5 100644 --- a/src/video_core/renderer_opengl/gl_shader_cache.cpp +++ b/src/video_core/renderer_opengl/gl_shader_cache.cpp @@ -6,10 +6,10 @@ #include "core/core.h" #include "core/memory.h" #include "video_core/engines/maxwell_3d.h" +#include "video_core/renderer_opengl/gl_rasterizer.h" #include "video_core/renderer_opengl/gl_shader_cache.h" #include "video_core/renderer_opengl/gl_shader_manager.h" #include "video_core/renderer_opengl/utils.h" -#include "video_core/utils.h" namespace OpenGL { @@ -121,12 +121,16 @@ GLint CachedShader::GetUniformLocation(const GLShader::SamplerEntry& sampler) { } GLuint CachedShader::LazyGeometryProgram(OGLProgram& target_program, - const std::string& glsl_topology, + const std::string& glsl_topology, u32 max_vertices, const std::string& debug_name) { if (target_program.handle != 0) { return target_program.handle; } - const std::string source{geometry_programs.code + "layout (" + glsl_topology + ") in;\n"}; + std::string source = "#version 430 core\n"; + source += "layout (" + glsl_topology + ") in;\n"; + source += "#define MAX_VERTEX_INPUT " + std::to_string(max_vertices) + '\n'; + source += geometry_programs.code; + OGLShader shader; shader.Create(source.c_str(), GL_GEOMETRY_SHADER); target_program.Create(true, shader.handle); @@ -135,6 +139,8 @@ GLuint CachedShader::LazyGeometryProgram(OGLProgram& target_program, return target_program.handle; }; +ShaderCacheOpenGL::ShaderCacheOpenGL(RasterizerOpenGL& rasterizer) : RasterizerCache{rasterizer} {} + Shader ShaderCacheOpenGL::GetStageProgram(Maxwell::ShaderProgram program) { const VAddr program_addr{GetShaderAddress(program)}; diff --git a/src/video_core/renderer_opengl/gl_shader_cache.h b/src/video_core/renderer_opengl/gl_shader_cache.h index a210f1731..ffbf21831 100644 --- a/src/video_core/renderer_opengl/gl_shader_cache.h +++ b/src/video_core/renderer_opengl/gl_shader_cache.h @@ -16,6 +16,8 @@ namespace OpenGL { class CachedShader; +class RasterizerOpenGL; + using Shader = std::shared_ptr<CachedShader>; using Maxwell = Tegra::Engines::Maxwell3D::Regs; @@ -46,22 +48,23 @@ public: } switch (primitive_mode) { case GL_POINTS: - return LazyGeometryProgram(geometry_programs.points, "points", "ShaderPoints"); + return LazyGeometryProgram(geometry_programs.points, "points", 1, "ShaderPoints"); case GL_LINES: case GL_LINE_STRIP: - return LazyGeometryProgram(geometry_programs.lines, "lines", "ShaderLines"); + return LazyGeometryProgram(geometry_programs.lines, "lines", 2, "ShaderLines"); case GL_LINES_ADJACENCY: case GL_LINE_STRIP_ADJACENCY: - return LazyGeometryProgram(geometry_programs.lines_adjacency, "lines_adjacency", + return LazyGeometryProgram(geometry_programs.lines_adjacency, "lines_adjacency", 4, "ShaderLinesAdjacency"); case GL_TRIANGLES: case GL_TRIANGLE_STRIP: case GL_TRIANGLE_FAN: - return LazyGeometryProgram(geometry_programs.triangles, "triangles", "ShaderTriangles"); + return LazyGeometryProgram(geometry_programs.triangles, "triangles", 3, + "ShaderTriangles"); case GL_TRIANGLES_ADJACENCY: case GL_TRIANGLE_STRIP_ADJACENCY: return LazyGeometryProgram(geometry_programs.triangles_adjacency, "triangles_adjacency", - "ShaderLines"); + 6, "ShaderTrianglesAdjacency"); default: UNREACHABLE_MSG("Unknown primitive mode."); } @@ -76,7 +79,7 @@ public: private: /// Generates a geometry shader or returns one that already exists. GLuint LazyGeometryProgram(OGLProgram& target_program, const std::string& glsl_topology, - const std::string& debug_name); + u32 max_vertices, const std::string& debug_name); VAddr addr; Maxwell::ShaderProgram program_type; @@ -104,6 +107,8 @@ private: class ShaderCacheOpenGL final : public RasterizerCache<Shader> { public: + explicit ShaderCacheOpenGL(RasterizerOpenGL& rasterizer); + /// Gets the current specified shader stage program Shader GetStageProgram(Maxwell::ShaderProgram program); }; diff --git a/src/video_core/renderer_opengl/gl_shader_decompiler.cpp b/src/video_core/renderer_opengl/gl_shader_decompiler.cpp index 09b003c59..5fde22ad4 100644 --- a/src/video_core/renderer_opengl/gl_shader_decompiler.cpp +++ b/src/video_core/renderer_opengl/gl_shader_decompiler.cpp @@ -494,10 +494,10 @@ public: // instruction for now. if (stage == Maxwell3D::Regs::ShaderStage::Geometry) { // TODO(Rodrigo): nouveau sets some attributes after setting emitting a geometry - // shader. These instructions use a dirty register as buffer index. To avoid some - // drivers from complaining for the out of boundary writes, guard them. - const std::string buf_index{"min(" + GetRegisterAsInteger(buf_reg) + ", " + - std::to_string(MAX_GEOMETRY_BUFFERS - 1) + ')'}; + // shader. These instructions use a dirty register as buffer index, to avoid some + // drivers from complaining about out of boundary writes, guard them. + const std::string buf_index{"((" + GetRegisterAsInteger(buf_reg) + ") % " + + std::to_string(MAX_GEOMETRY_BUFFERS) + ')'}; shader.AddLine("amem[" + buf_index + "][" + std::to_string(static_cast<u32>(attribute)) + ']' + GetSwizzle(elem) + " = " + src + ';'); @@ -811,7 +811,11 @@ private: std::optional<Register> vertex = {}) { auto GeometryPass = [&](const std::string& name) { if (stage == Maxwell3D::Regs::ShaderStage::Geometry && vertex) { - return "gs_" + name + '[' + GetRegisterAsInteger(*vertex, 0, false) + ']'; + // TODO(Rodrigo): Guard geometry inputs against out of bound reads. Some games set + // an 0x80000000 index for those and the shader fails to build. Find out why this + // happens and what's its intent. + return "gs_" + name + '[' + GetRegisterAsInteger(*vertex, 0, false) + + " % MAX_VERTEX_INPUT]"; } return name; }; @@ -2742,12 +2746,12 @@ private: } case 3: { if (is_array) { - UNIMPLEMENTED_MSG("3-coordinate arrays not fully implemented"); - const std::string x = regs.GetRegisterAsFloat(instr.gpr8); - const std::string y = regs.GetRegisterAsFloat(instr.gpr20); - coord = "vec2 coords = vec2(" + x + ", " + y + ");"; - texture_type = Tegra::Shader::TextureType::Texture2D; - is_array = false; + const std::string index = regs.GetRegisterAsInteger(instr.gpr8); + const std::string x = regs.GetRegisterAsFloat(instr.gpr8.Value() + 1); + const std::string y = regs.GetRegisterAsFloat(instr.gpr8.Value() + 2); + const std::string z = regs.GetRegisterAsFloat(instr.gpr20); + coord = + "vec4 coords = vec4(" + x + ", " + y + ", " + z + ", " + index + ");"; } else { const std::string x = regs.GetRegisterAsFloat(instr.gpr8); const std::string y = regs.GetRegisterAsFloat(instr.gpr8.Value() + 1); @@ -2777,7 +2781,11 @@ private: break; } case Tegra::Shader::TextureProcessMode::LZ: { - texture = "textureLod(" + sampler + ", coords, 0.0)"; + if (depth_compare && is_array) { + texture = "texture(" + sampler + ", coords)"; + } else { + texture = "textureLod(" + sampler + ", coords, 0.0)"; + } break; } case Tegra::Shader::TextureProcessMode::LL: { diff --git a/src/video_core/renderer_opengl/gl_shader_gen.cpp b/src/video_core/renderer_opengl/gl_shader_gen.cpp index 9d17edd63..eea090e52 100644 --- a/src/video_core/renderer_opengl/gl_shader_gen.cpp +++ b/src/video_core/renderer_opengl/gl_shader_gen.cpp @@ -82,8 +82,8 @@ void main() { } ProgramResult GenerateGeometryShader(const ShaderSetup& setup) { - std::string out = "#version 430 core\n"; - out += "#extension GL_ARB_separate_shader_objects : enable\n\n"; + // Version is intentionally skipped in shader generation, it's added by the lazy compilation. + std::string out = "#extension GL_ARB_separate_shader_objects : enable\n\n"; out += Decompiler::GetCommonDeclarations(); out += "bool exec_geometry();\n"; diff --git a/src/video_core/renderer_opengl/gl_shader_manager.h b/src/video_core/renderer_opengl/gl_shader_manager.h index 36fe1f04c..2a069cdd8 100644 --- a/src/video_core/renderer_opengl/gl_shader_manager.h +++ b/src/video_core/renderer_opengl/gl_shader_manager.h @@ -7,6 +7,7 @@ #include <glad/glad.h> #include "video_core/renderer_opengl/gl_resource_manager.h" +#include "video_core/renderer_opengl/gl_state.h" #include "video_core/renderer_opengl/maxwell_to_gl.h" namespace OpenGL::GLShader { diff --git a/src/video_core/renderer_opengl/gl_state.cpp b/src/video_core/renderer_opengl/gl_state.cpp index b6b426f34..98622a058 100644 --- a/src/video_core/renderer_opengl/gl_state.cpp +++ b/src/video_core/renderer_opengl/gl_state.cpp @@ -22,17 +22,15 @@ OpenGLState::OpenGLState() { depth.test_enabled = false; depth.test_func = GL_LESS; depth.write_mask = GL_TRUE; - depth.depth_range_near = 0.0f; - depth.depth_range_far = 1.0f; primitive_restart.enabled = false; primitive_restart.index = 0; - - color_mask.red_enabled = GL_TRUE; - color_mask.green_enabled = GL_TRUE; - color_mask.blue_enabled = GL_TRUE; - color_mask.alpha_enabled = GL_TRUE; - + for (auto& item : color_mask) { + item.red_enabled = GL_TRUE; + item.green_enabled = GL_TRUE; + item.blue_enabled = GL_TRUE; + item.alpha_enabled = GL_TRUE; + } stencil.test_enabled = false; auto reset_stencil = [](auto& config) { config.test_func = GL_ALWAYS; @@ -45,19 +43,33 @@ OpenGLState::OpenGLState() { }; reset_stencil(stencil.front); reset_stencil(stencil.back); - - blend.enabled = true; - blend.rgb_equation = GL_FUNC_ADD; - blend.a_equation = GL_FUNC_ADD; - blend.src_rgb_func = GL_ONE; - blend.dst_rgb_func = GL_ZERO; - blend.src_a_func = GL_ONE; - blend.dst_a_func = GL_ZERO; - blend.color.red = 0.0f; - blend.color.green = 0.0f; - blend.color.blue = 0.0f; - blend.color.alpha = 0.0f; - + for (auto& item : viewports) { + item.x = 0; + item.y = 0; + item.width = 0; + item.height = 0; + item.depth_range_near = 0.0f; + item.depth_range_far = 1.0f; + } + scissor.enabled = false; + scissor.x = 0; + scissor.y = 0; + scissor.width = 0; + scissor.height = 0; + for (auto& item : blend) { + item.enabled = true; + item.rgb_equation = GL_FUNC_ADD; + item.a_equation = GL_FUNC_ADD; + item.src_rgb_func = GL_ONE; + item.dst_rgb_func = GL_ZERO; + item.src_a_func = GL_ONE; + item.dst_a_func = GL_ZERO; + } + independant_blend.enabled = false; + blend_color.red = 0.0f; + blend_color.green = 0.0f; + blend_color.blue = 0.0f; + blend_color.alpha = 0.0f; logic_op.enabled = false; logic_op.operation = GL_COPY; @@ -73,17 +85,6 @@ OpenGLState::OpenGLState() { draw.shader_program = 0; draw.program_pipeline = 0; - scissor.enabled = false; - scissor.x = 0; - scissor.y = 0; - scissor.width = 0; - scissor.height = 0; - - viewport.x = 0; - viewport.y = 0; - viewport.width = 0; - viewport.height = 0; - clip_distance = {}; point.size = 1; @@ -134,6 +135,32 @@ void OpenGLState::ApplyCulling() const { } } +void OpenGLState::ApplyColorMask() const { + if (GLAD_GL_ARB_viewport_array) { + for (size_t i = 0; i < Tegra::Engines::Maxwell3D::Regs::NumRenderTargets; i++) { + const auto& updated = color_mask[i]; + const auto& current = cur_state.color_mask[i]; + if (updated.red_enabled != current.red_enabled || + updated.green_enabled != current.green_enabled || + updated.blue_enabled != current.blue_enabled || + updated.alpha_enabled != current.alpha_enabled) { + glColorMaski(static_cast<GLuint>(i), updated.red_enabled, updated.green_enabled, + updated.blue_enabled, updated.alpha_enabled); + } + } + } else { + const auto& updated = color_mask[0]; + const auto& current = cur_state.color_mask[0]; + if (updated.red_enabled != current.red_enabled || + updated.green_enabled != current.green_enabled || + updated.blue_enabled != current.blue_enabled || + updated.alpha_enabled != current.alpha_enabled) { + glColorMask(updated.red_enabled, updated.green_enabled, updated.blue_enabled, + updated.alpha_enabled); + } + } +} + void OpenGLState::ApplyDepth() const { // Depth test const bool depth_test_changed = depth.test_enabled != cur_state.depth.test_enabled; @@ -152,11 +179,6 @@ void OpenGLState::ApplyDepth() const { if (depth.write_mask != cur_state.depth.write_mask) { glDepthMask(depth.write_mask); } - // Depth range - if (depth.depth_range_near != cur_state.depth.depth_range_near || - depth.depth_range_far != cur_state.depth.depth_range_far) { - glDepthRange(depth.depth_range_near, depth.depth_range_far); - } } void OpenGLState::ApplyPrimitiveRestart() const { @@ -208,7 +230,7 @@ void OpenGLState::ApplyStencilTest() const { } } -void OpenGLState::ApplyScissorTest() const { +void OpenGLState::ApplyScissor() const { const bool scissor_changed = scissor.enabled != cur_state.scissor.enabled; if (scissor_changed) { if (scissor.enabled) { @@ -217,51 +239,142 @@ void OpenGLState::ApplyScissorTest() const { glDisable(GL_SCISSOR_TEST); } } - if (scissor_changed || scissor_changed || scissor.x != cur_state.scissor.x || - scissor.y != cur_state.scissor.y || scissor.width != cur_state.scissor.width || - scissor.height != cur_state.scissor.height) { + if (scissor.enabled && + (scissor_changed || scissor.x != cur_state.scissor.x || scissor.y != cur_state.scissor.y || + scissor.width != cur_state.scissor.width || scissor.height != cur_state.scissor.height)) { glScissor(scissor.x, scissor.y, scissor.width, scissor.height); } } -void OpenGLState::ApplyBlending() const { - const bool blend_changed = blend.enabled != cur_state.blend.enabled; +void OpenGLState::ApplyViewport() const { + if (GLAD_GL_ARB_viewport_array) { + for (GLuint i = 0; + i < static_cast<GLuint>(Tegra::Engines::Maxwell3D::Regs::NumRenderTargets); i++) { + const auto& current = cur_state.viewports[i]; + const auto& updated = viewports[i]; + if (updated.x != current.x || updated.y != current.y || + updated.width != current.width || updated.height != current.height) { + glViewportIndexedf(i, updated.x, updated.y, updated.width, updated.height); + } + if (updated.depth_range_near != current.depth_range_near || + updated.depth_range_far != current.depth_range_far) { + glDepthRangeIndexed(i, updated.depth_range_near, updated.depth_range_far); + } + } + } else { + const auto& current = cur_state.viewports[0]; + const auto& updated = viewports[0]; + if (updated.x != current.x || updated.y != current.y || updated.width != current.width || + updated.height != current.height) { + glViewport(static_cast<GLint>(updated.x), static_cast<GLint>(updated.y), + static_cast<GLsizei>(updated.width), static_cast<GLsizei>(updated.height)); + } + if (updated.depth_range_near != current.depth_range_near || + updated.depth_range_far != current.depth_range_far) { + glDepthRange(updated.depth_range_near, updated.depth_range_far); + } + } +} + +void OpenGLState::ApplyGlobalBlending() const { + const Blend& current = cur_state.blend[0]; + const Blend& updated = blend[0]; + const bool blend_changed = updated.enabled != current.enabled; if (blend_changed) { - if (blend.enabled) { - ASSERT(!logic_op.enabled); + if (updated.enabled) { glEnable(GL_BLEND); } else { glDisable(GL_BLEND); } } - if (blend.enabled) { - if (blend_changed || blend.color.red != cur_state.blend.color.red || - blend.color.green != cur_state.blend.color.green || - blend.color.blue != cur_state.blend.color.blue || - blend.color.alpha != cur_state.blend.color.alpha) { - glBlendColor(blend.color.red, blend.color.green, blend.color.blue, blend.color.alpha); + if (!updated.enabled) { + return; + } + if (updated.separate_alpha) { + if (blend_changed || updated.src_rgb_func != current.src_rgb_func || + updated.dst_rgb_func != current.dst_rgb_func || + updated.src_a_func != current.src_a_func || updated.dst_a_func != current.dst_a_func) { + glBlendFuncSeparate(updated.src_rgb_func, updated.dst_rgb_func, updated.src_a_func, + updated.dst_a_func); } - if (blend_changed || blend.src_rgb_func != cur_state.blend.src_rgb_func || - blend.dst_rgb_func != cur_state.blend.dst_rgb_func || - blend.src_a_func != cur_state.blend.src_a_func || - blend.dst_a_func != cur_state.blend.dst_a_func) { - glBlendFuncSeparate(blend.src_rgb_func, blend.dst_rgb_func, blend.src_a_func, - blend.dst_a_func); + if (blend_changed || updated.rgb_equation != current.rgb_equation || + updated.a_equation != current.a_equation) { + glBlendEquationSeparate(updated.rgb_equation, updated.a_equation); + } + } else { + if (blend_changed || updated.src_rgb_func != current.src_rgb_func || + updated.dst_rgb_func != current.dst_rgb_func) { + glBlendFunc(updated.src_rgb_func, updated.dst_rgb_func); } - if (blend_changed || blend.rgb_equation != cur_state.blend.rgb_equation || - blend.a_equation != cur_state.blend.a_equation) { - glBlendEquationSeparate(blend.rgb_equation, blend.a_equation); + if (blend_changed || updated.rgb_equation != current.rgb_equation) { + glBlendEquation(updated.rgb_equation); } } } +void OpenGLState::ApplyTargetBlending(std::size_t target, bool force) const { + const Blend& updated = blend[target]; + const Blend& current = cur_state.blend[target]; + const bool blend_changed = updated.enabled != current.enabled || force; + if (blend_changed) { + if (updated.enabled) { + glEnablei(GL_BLEND, static_cast<GLuint>(target)); + } else { + glDisablei(GL_BLEND, static_cast<GLuint>(target)); + } + } + if (!updated.enabled) { + return; + } + if (updated.separate_alpha) { + if (blend_changed || updated.src_rgb_func != current.src_rgb_func || + updated.dst_rgb_func != current.dst_rgb_func || + updated.src_a_func != current.src_a_func || updated.dst_a_func != current.dst_a_func) { + glBlendFuncSeparateiARB(static_cast<GLuint>(target), updated.src_rgb_func, + updated.dst_rgb_func, updated.src_a_func, updated.dst_a_func); + } + + if (blend_changed || updated.rgb_equation != current.rgb_equation || + updated.a_equation != current.a_equation) { + glBlendEquationSeparateiARB(static_cast<GLuint>(target), updated.rgb_equation, + updated.a_equation); + } + } else { + if (blend_changed || updated.src_rgb_func != current.src_rgb_func || + updated.dst_rgb_func != current.dst_rgb_func) { + glBlendFunciARB(static_cast<GLuint>(target), updated.src_rgb_func, + updated.dst_rgb_func); + } + + if (blend_changed || updated.rgb_equation != current.rgb_equation) { + glBlendEquationiARB(static_cast<GLuint>(target), updated.rgb_equation); + } + } +} + +void OpenGLState::ApplyBlending() const { + if (independant_blend.enabled) { + for (size_t i = 0; i < Tegra::Engines::Maxwell3D::Regs::NumRenderTargets; i++) { + ApplyTargetBlending(i, + independant_blend.enabled != cur_state.independant_blend.enabled); + } + } else { + ApplyGlobalBlending(); + } + if (blend_color.red != cur_state.blend_color.red || + blend_color.green != cur_state.blend_color.green || + blend_color.blue != cur_state.blend_color.blue || + blend_color.alpha != cur_state.blend_color.alpha) { + glBlendColor(blend_color.red, blend_color.green, blend_color.blue, blend_color.alpha); + } +} + void OpenGLState::ApplyLogicOp() const { const bool logic_op_changed = logic_op.enabled != cur_state.logic_op.enabled; if (logic_op_changed) { if (logic_op.enabled) { - ASSERT(!blend.enabled); glEnable(GL_COLOR_LOGIC_OP); } else { glDisable(GL_COLOR_LOGIC_OP); @@ -315,7 +428,7 @@ void OpenGLState::ApplySamplers() const { } } -void OpenGLState::Apply() const { +void OpenGLState::ApplyFramebufferState() const { // Framebuffer if (draw.read_framebuffer != cur_state.draw.read_framebuffer) { glBindFramebuffer(GL_READ_FRAMEBUFFER, draw.read_framebuffer); @@ -323,7 +436,9 @@ void OpenGLState::Apply() const { if (draw.draw_framebuffer != cur_state.draw.draw_framebuffer) { glBindFramebuffer(GL_DRAW_FRAMEBUFFER, draw.draw_framebuffer); } +} +void OpenGLState::ApplyVertexBufferState() const { // Vertex array if (draw.vertex_array != cur_state.draw.vertex_array) { glBindVertexArray(draw.vertex_array); @@ -333,7 +448,11 @@ void OpenGLState::Apply() const { if (draw.vertex_buffer != cur_state.draw.vertex_buffer) { glBindBuffer(GL_ARRAY_BUFFER, draw.vertex_buffer); } +} +void OpenGLState::Apply() const { + ApplyFramebufferState(); + ApplyVertexBufferState(); // Uniform buffer if (draw.uniform_buffer != cur_state.draw.uniform_buffer) { glBindBuffer(GL_UNIFORM_BUFFER, draw.uniform_buffer); @@ -348,12 +467,6 @@ void OpenGLState::Apply() const { if (draw.program_pipeline != cur_state.draw.program_pipeline) { glBindProgramPipeline(draw.program_pipeline); } - // Viewport - if (viewport.x != cur_state.viewport.x || viewport.y != cur_state.viewport.y || - viewport.width != cur_state.viewport.width || - viewport.height != cur_state.viewport.height) { - glViewport(viewport.x, viewport.y, viewport.width, viewport.height); - } // Clip distance for (std::size_t i = 0; i < clip_distance.size(); ++i) { if (clip_distance[i] != cur_state.clip_distance[i]) { @@ -364,19 +477,13 @@ void OpenGLState::Apply() const { } } } - // Color mask - if (color_mask.red_enabled != cur_state.color_mask.red_enabled || - color_mask.green_enabled != cur_state.color_mask.green_enabled || - color_mask.blue_enabled != cur_state.color_mask.blue_enabled || - color_mask.alpha_enabled != cur_state.color_mask.alpha_enabled) { - glColorMask(color_mask.red_enabled, color_mask.green_enabled, color_mask.blue_enabled, - color_mask.alpha_enabled); - } // Point if (point.size != cur_state.point.size) { glPointSize(point.size); } - ApplyScissorTest(); + ApplyColorMask(); + ApplyViewport(); + ApplyScissor(); ApplyStencilTest(); ApplySRgb(); ApplyCulling(); diff --git a/src/video_core/renderer_opengl/gl_state.h b/src/video_core/renderer_opengl/gl_state.h index fe648aff6..e5d1baae6 100644 --- a/src/video_core/renderer_opengl/gl_state.h +++ b/src/video_core/renderer_opengl/gl_state.h @@ -46,11 +46,9 @@ public: } cull; struct { - bool test_enabled; // GL_DEPTH_TEST - GLenum test_func; // GL_DEPTH_FUNC - GLboolean write_mask; // GL_DEPTH_WRITEMASK - GLfloat depth_range_near; // GL_DEPTH_RANGE - GLfloat depth_range_far; // GL_DEPTH_RANGE + bool test_enabled; // GL_DEPTH_TEST + GLenum test_func; // GL_DEPTH_FUNC + GLboolean write_mask; // GL_DEPTH_WRITEMASK } depth; struct { @@ -58,13 +56,14 @@ public: GLuint index; } primitive_restart; // GL_PRIMITIVE_RESTART - struct { + struct ColorMask { GLboolean red_enabled; GLboolean green_enabled; GLboolean blue_enabled; GLboolean alpha_enabled; - } color_mask; // GL_COLOR_WRITEMASK - + }; + std::array<ColorMask, Tegra::Engines::Maxwell3D::Regs::NumRenderTargets> + color_mask; // GL_COLOR_WRITEMASK struct { bool test_enabled; // GL_STENCIL_TEST struct { @@ -78,22 +77,28 @@ public: } front, back; } stencil; - struct { + struct Blend { bool enabled; // GL_BLEND + bool separate_alpha; // Independent blend enabled GLenum rgb_equation; // GL_BLEND_EQUATION_RGB GLenum a_equation; // GL_BLEND_EQUATION_ALPHA GLenum src_rgb_func; // GL_BLEND_SRC_RGB GLenum dst_rgb_func; // GL_BLEND_DST_RGB GLenum src_a_func; // GL_BLEND_SRC_ALPHA GLenum dst_a_func; // GL_BLEND_DST_ALPHA + }; + std::array<Blend, Tegra::Engines::Maxwell3D::Regs::NumRenderTargets> blend; - struct { - GLclampf red; - GLclampf green; - GLclampf blue; - GLclampf alpha; - } color; // GL_BLEND_COLOR - } blend; + struct { + bool enabled; + } independant_blend; + + struct { + GLclampf red; + GLclampf green; + GLclampf blue; + GLclampf alpha; + } blend_color; // GL_BLEND_COLOR struct { bool enabled; // GL_LOGIC_OP_MODE @@ -138,6 +143,16 @@ public: GLuint program_pipeline; // GL_PROGRAM_PIPELINE_BINDING } draw; + struct viewport { + GLfloat x; + GLfloat y; + GLfloat width; + GLfloat height; + GLfloat depth_range_near; // GL_DEPTH_RANGE + GLfloat depth_range_far; // GL_DEPTH_RANGE + }; + std::array<viewport, Tegra::Engines::Maxwell3D::Regs::NumRenderTargets> viewports; + struct { bool enabled; // GL_SCISSOR_TEST GLint x; @@ -147,13 +162,6 @@ public: } scissor; struct { - GLint x; - GLint y; - GLsizei width; - GLsizei height; - } viewport; - - struct { float size; // GL_POINT_SIZE } point; @@ -173,6 +181,10 @@ public: } /// Apply this state as the current OpenGL state void Apply() const; + /// Apply only the state afecting the framebuffer + void ApplyFramebufferState() const; + /// Apply only the state afecting the vertex buffer + void ApplyVertexBufferState() const; /// Set the initial OpenGL state static void ApplyDefaultState(); /// Resets any references to the given resource @@ -191,14 +203,18 @@ private: static bool s_rgb_used; void ApplySRgb() const; void ApplyCulling() const; + void ApplyColorMask() const; void ApplyDepth() const; void ApplyPrimitiveRestart() const; void ApplyStencilTest() const; - void ApplyScissorTest() const; + void ApplyViewport() const; + void ApplyTargetBlending(std::size_t target, bool force) const; + void ApplyGlobalBlending() const; void ApplyBlending() const; void ApplyLogicOp() const; void ApplyTextures() const; void ApplySamplers() const; + void ApplyScissor() const; }; } // namespace OpenGL diff --git a/src/video_core/renderer_opengl/gl_stream_buffer.cpp b/src/video_core/renderer_opengl/gl_stream_buffer.cpp index e409228cc..b97b895a4 100644 --- a/src/video_core/renderer_opengl/gl_stream_buffer.cpp +++ b/src/video_core/renderer_opengl/gl_stream_buffer.cpp @@ -6,9 +6,13 @@ #include <vector> #include "common/alignment.h" #include "common/assert.h" +#include "common/microprofile.h" #include "video_core/renderer_opengl/gl_state.h" #include "video_core/renderer_opengl/gl_stream_buffer.h" +MICROPROFILE_DEFINE(OpenGL_StreamBuffer, "OpenGL", "Stream Buffer Orphaning", + MP_RGB(128, 128, 192)); + namespace OpenGL { OGLStreamBuffer::OGLStreamBuffer(GLenum target, GLsizeiptr size, bool prefer_coherent) @@ -75,6 +79,7 @@ std::tuple<u8*, GLintptr, bool> OGLStreamBuffer::Map(GLsizeiptr size, GLintptr a } if (invalidate || !persistent) { + MICROPROFILE_SCOPE(OpenGL_StreamBuffer); GLbitfield flags = GL_MAP_WRITE_BIT | (persistent ? GL_MAP_PERSISTENT_BIT : 0) | (coherent ? GL_MAP_COHERENT_BIT : GL_MAP_FLUSH_EXPLICIT_BIT) | (invalidate ? GL_MAP_INVALIDATE_BUFFER_BIT : GL_MAP_UNSYNCHRONIZED_BIT); diff --git a/src/video_core/renderer_opengl/maxwell_to_gl.h b/src/video_core/renderer_opengl/maxwell_to_gl.h index 87d511c38..3ce2cc6d2 100644 --- a/src/video_core/renderer_opengl/maxwell_to_gl.h +++ b/src/video_core/renderer_opengl/maxwell_to_gl.h @@ -159,10 +159,8 @@ inline GLenum TextureFilterMode(Tegra::Texture::TextureFilter filter_mode, } } } - LOG_CRITICAL(Render_OpenGL, "Unimplemented texture filter mode={}", - static_cast<u32>(filter_mode)); - UNREACHABLE(); - return {}; + LOG_ERROR(Render_OpenGL, "Unimplemented texture filter mode={}", static_cast<u32>(filter_mode)); + return GL_LINEAR; } inline GLenum WrapMode(Tegra::Texture::WrapMode wrap_mode) { @@ -183,9 +181,8 @@ inline GLenum WrapMode(Tegra::Texture::WrapMode wrap_mode) { case Tegra::Texture::WrapMode::MirrorOnceClampToEdge: return GL_MIRROR_CLAMP_TO_EDGE; } - LOG_CRITICAL(Render_OpenGL, "Unimplemented texture wrap mode={}", static_cast<u32>(wrap_mode)); - UNREACHABLE(); - return {}; + LOG_ERROR(Render_OpenGL, "Unimplemented texture wrap mode={}", static_cast<u32>(wrap_mode)); + return GL_REPEAT; } inline GLenum DepthCompareFunc(Tegra::Texture::DepthCompareFunc func) { @@ -207,10 +204,9 @@ inline GLenum DepthCompareFunc(Tegra::Texture::DepthCompareFunc func) { case Tegra::Texture::DepthCompareFunc::Always: return GL_ALWAYS; } - LOG_CRITICAL(Render_OpenGL, "Unimplemented texture depth compare function ={}", - static_cast<u32>(func)); - UNREACHABLE(); - return {}; + LOG_ERROR(Render_OpenGL, "Unimplemented texture depth compare function ={}", + static_cast<u32>(func)); + return GL_GREATER; } inline GLenum BlendEquation(Maxwell::Blend::Equation equation) { @@ -226,9 +222,8 @@ inline GLenum BlendEquation(Maxwell::Blend::Equation equation) { case Maxwell::Blend::Equation::Max: return GL_MAX; } - LOG_CRITICAL(Render_OpenGL, "Unimplemented blend equation={}", static_cast<u32>(equation)); - UNREACHABLE(); - return {}; + LOG_ERROR(Render_OpenGL, "Unimplemented blend equation={}", static_cast<u32>(equation)); + return GL_FUNC_ADD; } inline GLenum BlendFunc(Maxwell::Blend::Factor factor) { @@ -291,9 +286,8 @@ inline GLenum BlendFunc(Maxwell::Blend::Factor factor) { case Maxwell::Blend::Factor::OneMinusConstantAlphaGL: return GL_ONE_MINUS_CONSTANT_ALPHA; } - LOG_CRITICAL(Render_OpenGL, "Unimplemented blend factor={}", static_cast<u32>(factor)); - UNREACHABLE(); - return {}; + LOG_ERROR(Render_OpenGL, "Unimplemented blend factor={}", static_cast<u32>(factor)); + return GL_ZERO; } inline GLenum SwizzleSource(Tegra::Texture::SwizzleSource source) { @@ -312,9 +306,8 @@ inline GLenum SwizzleSource(Tegra::Texture::SwizzleSource source) { case Tegra::Texture::SwizzleSource::OneFloat: return GL_ONE; } - LOG_CRITICAL(Render_OpenGL, "Unimplemented swizzle source={}", static_cast<u32>(source)); - UNREACHABLE(); - return {}; + LOG_ERROR(Render_OpenGL, "Unimplemented swizzle source={}", static_cast<u32>(source)); + return GL_ZERO; } inline GLenum ComparisonOp(Maxwell::ComparisonOp comparison) { @@ -344,33 +337,39 @@ inline GLenum ComparisonOp(Maxwell::ComparisonOp comparison) { case Maxwell::ComparisonOp::AlwaysOld: return GL_ALWAYS; } - LOG_CRITICAL(Render_OpenGL, "Unimplemented comparison op={}", static_cast<u32>(comparison)); - UNREACHABLE(); - return {}; + LOG_ERROR(Render_OpenGL, "Unimplemented comparison op={}", static_cast<u32>(comparison)); + return GL_ALWAYS; } inline GLenum StencilOp(Maxwell::StencilOp stencil) { switch (stencil) { case Maxwell::StencilOp::Keep: + case Maxwell::StencilOp::KeepOGL: return GL_KEEP; case Maxwell::StencilOp::Zero: + case Maxwell::StencilOp::ZeroOGL: return GL_ZERO; case Maxwell::StencilOp::Replace: + case Maxwell::StencilOp::ReplaceOGL: return GL_REPLACE; case Maxwell::StencilOp::Incr: + case Maxwell::StencilOp::IncrOGL: return GL_INCR; case Maxwell::StencilOp::Decr: + case Maxwell::StencilOp::DecrOGL: return GL_DECR; case Maxwell::StencilOp::Invert: + case Maxwell::StencilOp::InvertOGL: return GL_INVERT; case Maxwell::StencilOp::IncrWrap: + case Maxwell::StencilOp::IncrWrapOGL: return GL_INCR_WRAP; case Maxwell::StencilOp::DecrWrap: + case Maxwell::StencilOp::DecrWrapOGL: return GL_DECR_WRAP; } - LOG_CRITICAL(Render_OpenGL, "Unimplemented stencil op={}", static_cast<u32>(stencil)); - UNREACHABLE(); - return {}; + LOG_ERROR(Render_OpenGL, "Unimplemented stencil op={}", static_cast<u32>(stencil)); + return GL_KEEP; } inline GLenum FrontFace(Maxwell::Cull::FrontFace front_face) { @@ -380,9 +379,8 @@ inline GLenum FrontFace(Maxwell::Cull::FrontFace front_face) { case Maxwell::Cull::FrontFace::CounterClockWise: return GL_CCW; } - LOG_CRITICAL(Render_OpenGL, "Unimplemented front face cull={}", static_cast<u32>(front_face)); - UNREACHABLE(); - return {}; + LOG_ERROR(Render_OpenGL, "Unimplemented front face cull={}", static_cast<u32>(front_face)); + return GL_CCW; } inline GLenum CullFace(Maxwell::Cull::CullFace cull_face) { @@ -394,9 +392,8 @@ inline GLenum CullFace(Maxwell::Cull::CullFace cull_face) { case Maxwell::Cull::CullFace::FrontAndBack: return GL_FRONT_AND_BACK; } - LOG_CRITICAL(Render_OpenGL, "Unimplemented cull face={}", static_cast<u32>(cull_face)); - UNREACHABLE(); - return {}; + LOG_ERROR(Render_OpenGL, "Unimplemented cull face={}", static_cast<u32>(cull_face)); + return GL_BACK; } inline GLenum LogicOp(Maxwell::LogicOperation operation) { @@ -434,9 +431,8 @@ inline GLenum LogicOp(Maxwell::LogicOperation operation) { case Maxwell::LogicOperation::Set: return GL_SET; } - LOG_CRITICAL(Render_OpenGL, "Unimplemented logic operation={}", static_cast<u32>(operation)); - UNREACHABLE(); - return {}; + LOG_ERROR(Render_OpenGL, "Unimplemented logic operation={}", static_cast<u32>(operation)); + return GL_COPY; } } // namespace MaxwellToGL diff --git a/src/video_core/surface.cpp b/src/video_core/surface.cpp index d9a97e30b..9582dd2ca 100644 --- a/src/video_core/surface.cpp +++ b/src/video_core/surface.cpp @@ -19,6 +19,8 @@ SurfaceTarget SurfaceTargetFromTextureType(Tegra::Texture::TextureType texture_t return SurfaceTarget::Texture3D; case Tegra::Texture::TextureType::TextureCubemap: return SurfaceTarget::TextureCubemap; + case Tegra::Texture::TextureType::TextureCubeArray: + return SurfaceTarget::TextureCubeArray; case Tegra::Texture::TextureType::Texture1DArray: return SurfaceTarget::Texture1DArray; case Tegra::Texture::TextureType::Texture2DArray: @@ -39,6 +41,7 @@ bool SurfaceTargetIsLayered(SurfaceTarget target) { case SurfaceTarget::Texture1DArray: case SurfaceTarget::Texture2DArray: case SurfaceTarget::TextureCubemap: + case SurfaceTarget::TextureCubeArray: return true; default: LOG_CRITICAL(HW_GPU, "Unimplemented surface_target={}", static_cast<u32>(target)); @@ -297,10 +300,14 @@ PixelFormat PixelFormatFromTextureFormat(Tegra::Texture::TextureFormat format, return is_srgb ? PixelFormat::ASTC_2D_4X4_SRGB : PixelFormat::ASTC_2D_4X4; case Tegra::Texture::TextureFormat::ASTC_2D_5X4: return is_srgb ? PixelFormat::ASTC_2D_5X4_SRGB : PixelFormat::ASTC_2D_5X4; + case Tegra::Texture::TextureFormat::ASTC_2D_5X5: + return is_srgb ? PixelFormat::ASTC_2D_5X5_SRGB : PixelFormat::ASTC_2D_5X5; case Tegra::Texture::TextureFormat::ASTC_2D_8X8: return is_srgb ? PixelFormat::ASTC_2D_8X8_SRGB : PixelFormat::ASTC_2D_8X8; case Tegra::Texture::TextureFormat::ASTC_2D_8X5: return is_srgb ? PixelFormat::ASTC_2D_8X5_SRGB : PixelFormat::ASTC_2D_8X5; + case Tegra::Texture::TextureFormat::ASTC_2D_10X8: + return is_srgb ? PixelFormat::ASTC_2D_10X8_SRGB : PixelFormat::ASTC_2D_10X8; case Tegra::Texture::TextureFormat::R16_G16: switch (component_type) { case Tegra::Texture::ComponentType::FLOAT: @@ -440,12 +447,16 @@ bool IsPixelFormatASTC(PixelFormat format) { switch (format) { case PixelFormat::ASTC_2D_4X4: case PixelFormat::ASTC_2D_5X4: + case PixelFormat::ASTC_2D_5X5: case PixelFormat::ASTC_2D_8X8: case PixelFormat::ASTC_2D_8X5: case PixelFormat::ASTC_2D_4X4_SRGB: case PixelFormat::ASTC_2D_5X4_SRGB: + case PixelFormat::ASTC_2D_5X5_SRGB: case PixelFormat::ASTC_2D_8X8_SRGB: case PixelFormat::ASTC_2D_8X5_SRGB: + case PixelFormat::ASTC_2D_10X8: + case PixelFormat::ASTC_2D_10X8_SRGB: return true; default: return false; @@ -453,27 +464,7 @@ bool IsPixelFormatASTC(PixelFormat format) { } std::pair<u32, u32> GetASTCBlockSize(PixelFormat format) { - switch (format) { - case PixelFormat::ASTC_2D_4X4: - return {4, 4}; - case PixelFormat::ASTC_2D_5X4: - return {5, 4}; - case PixelFormat::ASTC_2D_8X8: - return {8, 8}; - case PixelFormat::ASTC_2D_8X5: - return {8, 5}; - case PixelFormat::ASTC_2D_4X4_SRGB: - return {4, 4}; - case PixelFormat::ASTC_2D_5X4_SRGB: - return {5, 4}; - case PixelFormat::ASTC_2D_8X8_SRGB: - return {8, 8}; - case PixelFormat::ASTC_2D_8X5_SRGB: - return {8, 5}; - default: - LOG_CRITICAL(HW_GPU, "Unhandled format: {}", static_cast<u32>(format)); - UNREACHABLE(); - } + return {GetDefaultBlockWidth(format), GetDefaultBlockHeight(format)}; } bool IsFormatBCn(PixelFormat format) { diff --git a/src/video_core/surface.h b/src/video_core/surface.h index 3232e437f..0dd3eb2e4 100644 --- a/src/video_core/surface.h +++ b/src/video_core/surface.h @@ -72,19 +72,23 @@ enum class PixelFormat { ASTC_2D_8X8_SRGB = 54, ASTC_2D_8X5_SRGB = 55, ASTC_2D_5X4_SRGB = 56, + ASTC_2D_5X5 = 57, + ASTC_2D_5X5_SRGB = 58, + ASTC_2D_10X8 = 59, + ASTC_2D_10X8_SRGB = 60, MaxColorFormat, // Depth formats - Z32F = 57, - Z16 = 58, + Z32F = 61, + Z16 = 62, MaxDepthFormat, // DepthStencil formats - Z24S8 = 59, - S8Z24 = 60, - Z32FS8 = 61, + Z24S8 = 63, + S8Z24 = 64, + Z32FS8 = 65, MaxDepthStencilFormat, @@ -118,6 +122,7 @@ enum class SurfaceTarget { Texture1DArray, Texture2DArray, TextureCubemap, + TextureCubeArray, }; /** @@ -188,6 +193,10 @@ static constexpr u32 GetCompressionFactor(PixelFormat format) { 4, // ASTC_2D_8X8_SRGB 4, // ASTC_2D_8X5_SRGB 4, // ASTC_2D_5X4_SRGB + 4, // ASTC_2D_5X5 + 4, // ASTC_2D_5X5_SRGB + 4, // ASTC_2D_10X8 + 4, // ASTC_2D_10X8_SRGB 1, // Z32F 1, // Z16 1, // Z24S8 @@ -199,6 +208,81 @@ static constexpr u32 GetCompressionFactor(PixelFormat format) { return compression_factor_table[static_cast<std::size_t>(format)]; } +static constexpr u32 GetDefaultBlockWidth(PixelFormat format) { + if (format == PixelFormat::Invalid) + return 0; + constexpr std::array<u32, MaxPixelFormat> block_width_table = {{ + 1, // ABGR8U + 1, // ABGR8S + 1, // ABGR8UI + 1, // B5G6R5U + 1, // A2B10G10R10U + 1, // A1B5G5R5U + 1, // R8U + 1, // R8UI + 1, // RGBA16F + 1, // RGBA16U + 1, // RGBA16UI + 1, // R11FG11FB10F + 1, // RGBA32UI + 4, // DXT1 + 4, // DXT23 + 4, // DXT45 + 4, // DXN1 + 4, // DXN2UNORM + 4, // DXN2SNORM + 4, // BC7U + 4, // BC6H_UF16 + 4, // BC6H_SF16 + 4, // ASTC_2D_4X4 + 1, // G8R8U + 1, // G8R8S + 1, // BGRA8 + 1, // RGBA32F + 1, // RG32F + 1, // R32F + 1, // R16F + 1, // R16U + 1, // R16S + 1, // R16UI + 1, // R16I + 1, // RG16 + 1, // RG16F + 1, // RG16UI + 1, // RG16I + 1, // RG16S + 1, // RGB32F + 1, // RGBA8_SRGB + 1, // RG8U + 1, // RG8S + 1, // RG32UI + 1, // R32UI + 8, // ASTC_2D_8X8 + 8, // ASTC_2D_8X5 + 5, // ASTC_2D_5X4 + 1, // BGRA8_SRGB + 4, // DXT1_SRGB + 4, // DXT23_SRGB + 4, // DXT45_SRGB + 4, // BC7U_SRGB + 4, // ASTC_2D_4X4_SRGB + 8, // ASTC_2D_8X8_SRGB + 8, // ASTC_2D_8X5_SRGB + 5, // ASTC_2D_5X4_SRGB + 5, // ASTC_2D_5X5 + 5, // ASTC_2D_5X5_SRGB + 10, // ASTC_2D_10X8 + 10, // ASTC_2D_10X8_SRGB + 1, // Z32F + 1, // Z16 + 1, // Z24S8 + 1, // S8Z24 + 1, // Z32FS8 + }}; + ASSERT(static_cast<std::size_t>(format) < block_width_table.size()); + return block_width_table[static_cast<std::size_t>(format)]; +} + static constexpr u32 GetDefaultBlockHeight(PixelFormat format) { if (format == PixelFormat::Invalid) return 0; @@ -261,6 +345,10 @@ static constexpr u32 GetDefaultBlockHeight(PixelFormat format) { 8, // ASTC_2D_8X8_SRGB 5, // ASTC_2D_8X5_SRGB 4, // ASTC_2D_5X4_SRGB + 5, // ASTC_2D_5X5 + 5, // ASTC_2D_5X5_SRGB + 8, // ASTC_2D_10X8 + 8, // ASTC_2D_10X8_SRGB 1, // Z32F 1, // Z16 1, // Z24S8 @@ -299,7 +387,7 @@ static constexpr u32 GetFormatBpp(PixelFormat format) { 128, // BC7U 128, // BC6H_UF16 128, // BC6H_SF16 - 32, // ASTC_2D_4X4 + 128, // ASTC_2D_4X4 16, // G8R8U 16, // G8R8S 32, // BGRA8 @@ -322,18 +410,22 @@ static constexpr u32 GetFormatBpp(PixelFormat format) { 16, // RG8S 64, // RG32UI 32, // R32UI - 16, // ASTC_2D_8X8 - 16, // ASTC_2D_8X5 - 32, // ASTC_2D_5X4 + 128, // ASTC_2D_8X8 + 128, // ASTC_2D_8X5 + 128, // ASTC_2D_5X4 32, // BGRA8_SRGB 64, // DXT1_SRGB 128, // DXT23_SRGB 128, // DXT45_SRGB 128, // BC7U - 32, // ASTC_2D_4X4_SRGB - 16, // ASTC_2D_8X8_SRGB - 16, // ASTC_2D_8X5_SRGB - 32, // ASTC_2D_5X4_SRGB + 128, // ASTC_2D_4X4_SRGB + 128, // ASTC_2D_8X8_SRGB + 128, // ASTC_2D_8X5_SRGB + 128, // ASTC_2D_5X4_SRGB + 128, // ASTC_2D_5X5 + 128, // ASTC_2D_5X5_SRGB + 128, // ASTC_2D_10X8 + 128, // ASTC_2D_10X8_SRGB 32, // Z32F 16, // Z16 32, // Z24S8 diff --git a/src/video_core/textures/astc.cpp b/src/video_core/textures/astc.cpp index b1feacae9..bc50a4876 100644 --- a/src/video_core/textures/astc.cpp +++ b/src/video_core/textures/astc.cpp @@ -1598,27 +1598,29 @@ static void DecompressBlock(uint8_t inBuf[16], const uint32_t blockWidth, namespace Tegra::Texture::ASTC { std::vector<uint8_t> Decompress(std::vector<uint8_t>& data, uint32_t width, uint32_t height, - uint32_t block_width, uint32_t block_height) { + uint32_t depth, uint32_t block_width, uint32_t block_height) { uint32_t blockIdx = 0; - std::vector<uint8_t> outData(height * width * 4); - for (uint32_t j = 0; j < height; j += block_height) { - for (uint32_t i = 0; i < width; i += block_width) { + std::vector<uint8_t> outData(height * width * depth * 4); + for (uint32_t k = 0; k < depth; k++) { + for (uint32_t j = 0; j < height; j += block_height) { + for (uint32_t i = 0; i < width; i += block_width) { - uint8_t* blockPtr = data.data() + blockIdx * 16; + uint8_t* blockPtr = data.data() + blockIdx * 16; - // Blocks can be at most 12x12 - uint32_t uncompData[144]; - ASTCC::DecompressBlock(blockPtr, block_width, block_height, uncompData); + // Blocks can be at most 12x12 + uint32_t uncompData[144]; + ASTCC::DecompressBlock(blockPtr, block_width, block_height, uncompData); - uint32_t decompWidth = std::min(block_width, width - i); - uint32_t decompHeight = std::min(block_height, height - j); + uint32_t decompWidth = std::min(block_width, width - i); + uint32_t decompHeight = std::min(block_height, height - j); - uint8_t* outRow = outData.data() + (j * width + i) * 4; - for (uint32_t jj = 0; jj < decompHeight; jj++) { - memcpy(outRow + jj * width * 4, uncompData + jj * block_width, decompWidth * 4); - } + uint8_t* outRow = outData.data() + (j * width + i) * 4; + for (uint32_t jj = 0; jj < decompHeight; jj++) { + memcpy(outRow + jj * width * 4, uncompData + jj * block_width, decompWidth * 4); + } - blockIdx++; + blockIdx++; + } } } diff --git a/src/video_core/textures/astc.h b/src/video_core/textures/astc.h index f0d7c0e56..d419dd025 100644 --- a/src/video_core/textures/astc.h +++ b/src/video_core/textures/astc.h @@ -10,6 +10,6 @@ namespace Tegra::Texture::ASTC { std::vector<uint8_t> Decompress(std::vector<uint8_t>& data, uint32_t width, uint32_t height, - uint32_t block_width, uint32_t block_height); + uint32_t depth, uint32_t block_width, uint32_t block_height); } // namespace Tegra::Texture::ASTC diff --git a/src/video_core/textures/decoders.cpp b/src/video_core/textures/decoders.cpp index 550ca856c..19f30b1b5 100644 --- a/src/video_core/textures/decoders.cpp +++ b/src/video_core/textures/decoders.cpp @@ -45,7 +45,7 @@ constexpr auto fast_swizzle_table = SwizzleTable<8, 4, 16>(); * Instead of going gob by gob, we map the coordinates inside a block and manage from * those. Block_Width is assumed to be 1. */ -void PreciseProcessBlock(u8* swizzled_data, u8* unswizzled_data, const bool unswizzle, +void PreciseProcessBlock(u8* const swizzled_data, u8* const unswizzled_data, const bool unswizzle, const u32 x_start, const u32 y_start, const u32 z_start, const u32 x_end, const u32 y_end, const u32 z_end, const u32 tile_offset, const u32 xy_block_size, const u32 layer_z, const u32 stride_x, @@ -81,7 +81,7 @@ void PreciseProcessBlock(u8* swizzled_data, u8* unswizzled_data, const bool unsw * Instead of going gob by gob, we map the coordinates inside a block and manage from * those. Block_Width is assumed to be 1. */ -void FastProcessBlock(u8* swizzled_data, u8* unswizzled_data, const bool unswizzle, +void FastProcessBlock(u8* const swizzled_data, u8* const unswizzled_data, const bool unswizzle, const u32 x_start, const u32 y_start, const u32 z_start, const u32 x_end, const u32 y_end, const u32 z_end, const u32 tile_offset, const u32 xy_block_size, const u32 layer_z, const u32 stride_x, @@ -90,10 +90,10 @@ void FastProcessBlock(u8* swizzled_data, u8* unswizzled_data, const bool unswizz u32 z_address = tile_offset; const u32 x_startb = x_start * bytes_per_pixel; const u32 x_endb = x_end * bytes_per_pixel; - const u32 copy_size = 16; - const u32 gob_size_x = 64; - const u32 gob_size_y = 8; - const u32 gob_size_z = 1; + constexpr u32 copy_size = 16; + constexpr u32 gob_size_x = 64; + constexpr u32 gob_size_y = 8; + constexpr u32 gob_size_z = 1; const u32 gob_size = gob_size_x * gob_size_y * gob_size_z; for (u32 z = z_start; z < z_end; z++) { u32 y_address = z_address; @@ -126,23 +126,23 @@ void FastProcessBlock(u8* swizzled_data, u8* unswizzled_data, const bool unswizz * https://envytools.readthedocs.io/en/latest/hw/memory/g80-surface.html#blocklinear-surfaces */ template <bool fast> -void SwizzledData(u8* swizzled_data, u8* unswizzled_data, const bool unswizzle, const u32 width, - const u32 height, const u32 depth, const u32 bytes_per_pixel, +void SwizzledData(u8* const swizzled_data, u8* const unswizzled_data, const bool unswizzle, + const u32 width, const u32 height, const u32 depth, const u32 bytes_per_pixel, const u32 out_bytes_per_pixel, const u32 block_height, const u32 block_depth) { auto div_ceil = [](const u32 x, const u32 y) { return ((x + y - 1) / y); }; const u32 stride_x = width * out_bytes_per_pixel; const u32 layer_z = height * stride_x; - const u32 gob_x_bytes = 64; + constexpr u32 gob_x_bytes = 64; const u32 gob_elements_x = gob_x_bytes / bytes_per_pixel; - const u32 gob_elements_y = 8; - const u32 gob_elements_z = 1; + constexpr u32 gob_elements_y = 8; + constexpr u32 gob_elements_z = 1; const u32 block_x_elements = gob_elements_x; const u32 block_y_elements = gob_elements_y * block_height; const u32 block_z_elements = gob_elements_z * block_depth; const u32 blocks_on_x = div_ceil(width, block_x_elements); const u32 blocks_on_y = div_ceil(height, block_y_elements); const u32 blocks_on_z = div_ceil(depth, block_z_elements); - const u32 gob_size = gob_x_bytes * gob_elements_y * gob_elements_z; + constexpr u32 gob_size = gob_x_bytes * gob_elements_y * gob_elements_z; const u32 xy_block_size = gob_size * block_height; const u32 block_size = xy_block_size * block_depth; u32 tile_offset = 0; @@ -171,7 +171,7 @@ void SwizzledData(u8* swizzled_data, u8* unswizzled_data, const bool unswizzle, } void CopySwizzledData(u32 width, u32 height, u32 depth, u32 bytes_per_pixel, - u32 out_bytes_per_pixel, u8* swizzled_data, u8* unswizzled_data, + u32 out_bytes_per_pixel, u8* const swizzled_data, u8* const unswizzled_data, bool unswizzle, u32 block_height, u32 block_depth) { if (bytes_per_pixel % 3 != 0 && (width * bytes_per_pixel) % 16 == 0) { SwizzledData<true>(swizzled_data, unswizzled_data, unswizzle, width, height, depth, @@ -202,6 +202,8 @@ u32 BytesPerPixel(TextureFormat format) { case TextureFormat::ASTC_2D_5X4: case TextureFormat::ASTC_2D_8X8: case TextureFormat::ASTC_2D_8X5: + case TextureFormat::ASTC_2D_10X8: + case TextureFormat::ASTC_2D_5X5: case TextureFormat::A8R8G8B8: case TextureFormat::A2B10G10R10: case TextureFormat::BF10GF11RF11: @@ -227,12 +229,14 @@ u32 BytesPerPixel(TextureFormat format) { } } -std::vector<u8> UnswizzleTexture(VAddr address, u32 tile_size, u32 bytes_per_pixel, u32 width, - u32 height, u32 depth, u32 block_height, u32 block_depth) { +std::vector<u8> UnswizzleTexture(VAddr address, u32 tile_size_x, u32 tile_size_y, + u32 bytes_per_pixel, u32 width, u32 height, u32 depth, + u32 block_height, u32 block_depth) { std::vector<u8> unswizzled_data(width * height * depth * bytes_per_pixel); - CopySwizzledData(width / tile_size, height / tile_size, depth, bytes_per_pixel, bytes_per_pixel, - Memory::GetPointer(address), unswizzled_data.data(), true, block_height, - block_depth); + CopySwizzledData((width + tile_size_x - 1) / tile_size_x, + (height + tile_size_y - 1) / tile_size_y, depth, bytes_per_pixel, + bytes_per_pixel, Memory::GetPointer(address), unswizzled_data.data(), true, + block_height, block_depth); return unswizzled_data; } @@ -292,6 +296,8 @@ std::vector<u8> DecodeTexture(const std::vector<u8>& texture_data, TextureFormat case TextureFormat::BC6H_SF16: case TextureFormat::ASTC_2D_4X4: case TextureFormat::ASTC_2D_8X8: + case TextureFormat::ASTC_2D_5X5: + case TextureFormat::ASTC_2D_10X8: case TextureFormat::A8R8G8B8: case TextureFormat::A2B10G10R10: case TextureFormat::A1B5G5R5: @@ -319,9 +325,9 @@ std::vector<u8> DecodeTexture(const std::vector<u8>& texture_data, TextureFormat std::size_t CalculateSize(bool tiled, u32 bytes_per_pixel, u32 width, u32 height, u32 depth, u32 block_height, u32 block_depth) { if (tiled) { - const u32 gobs_in_x = 64; - const u32 gobs_in_y = 8; - const u32 gobs_in_z = 1; + constexpr u32 gobs_in_x = 64; + constexpr u32 gobs_in_y = 8; + constexpr u32 gobs_in_z = 1; const u32 aligned_width = Common::AlignUp(width * bytes_per_pixel, gobs_in_x); const u32 aligned_height = Common::AlignUp(height, gobs_in_y * block_height); const u32 aligned_depth = Common::AlignUp(depth, gobs_in_z * block_depth); diff --git a/src/video_core/textures/decoders.h b/src/video_core/textures/decoders.h index b390219e4..ba065510b 100644 --- a/src/video_core/textures/decoders.h +++ b/src/video_core/textures/decoders.h @@ -19,8 +19,8 @@ inline std::size_t GetGOBSize() { /** * Unswizzles a swizzled texture without changing its format. */ -std::vector<u8> UnswizzleTexture(VAddr address, u32 tile_size, u32 bytes_per_pixel, u32 width, - u32 height, u32 depth, +std::vector<u8> UnswizzleTexture(VAddr address, u32 tile_size_x, u32 tile_size_y, + u32 bytes_per_pixel, u32 width, u32 height, u32 depth, u32 block_height = TICEntry::DefaultBlockHeight, u32 block_depth = TICEntry::DefaultBlockHeight); diff --git a/src/video_core/textures/texture.h b/src/video_core/textures/texture.h index d12d2ecb8..e199d019a 100644 --- a/src/video_core/textures/texture.h +++ b/src/video_core/textures/texture.h @@ -168,20 +168,29 @@ struct TICEntry { // High 16 bits of the pitch value BitField<0, 16, u32> pitch_high; - + BitField<26, 1, u32> use_header_opt_control; + BitField<27, 1, u32> depth_texture; BitField<28, 4, u32> max_mip_level; }; union { BitField<0, 16, u32> width_minus_1; BitField<22, 1, u32> srgb_conversion; BitField<23, 4, TextureType> texture_type; + BitField<29, 3, u32> border_size; }; union { BitField<0, 16, u32> height_minus_1; BitField<16, 15, u32> depth_minus_1; }; + union { + BitField<6, 13, u32> mip_lod_bias; + BitField<27, 3, u32> max_anisotropy; + }; - INSERT_PADDING_BYTES(8); + union { + BitField<0, 4, u32> res_min_mip_level; + BitField<4, 4, u32> res_max_mip_level; + }; GPUVAddr Address() const { return static_cast<GPUVAddr>((static_cast<GPUVAddr>(address_high) << 32) | address_low); |