rasterizer_cache: Remove reliance on the System singleton

Rather than have a transparent dependency, we can make it explicit in
the interface. This also gets rid of the need to put the core include in
a header.

src/audio_core/stream.cpp

@@ -11,7 +11,6 @@
 #include "audio_core/stream.h"
 #include "common/assert.h"
 #include "common/logging/log.h"
-#include "common/microprofile.h"
 #include "core/core_timing.h"
 #include "core/core_timing_util.h"
 #include "core/settings.h"
@@ -104,10 +103,7 @@ void Stream::PlayNextBuffer() {
     CoreTiming::ScheduleEventThreadsafe(GetBufferReleaseCycles(*active_buffer), release_event, {});
 }
 
-MICROPROFILE_DEFINE(AudioOutput, "Audio", "ReleaseActiveBuffer", MP_RGB(100, 100, 255));
-
 void Stream::ReleaseActiveBuffer() {
-    MICROPROFILE_SCOPE(AudioOutput);
     ASSERT(active_buffer);
     released_buffers.push(std::move(active_buffer));
     release_callback();

src/core/hle/kernel/svc.cpp

@@ -395,16 +395,42 @@ struct BreakReason {
 /// Break program execution
 static void Break(u32 reason, u64 info1, u64 info2) {
     BreakReason break_reason{reason};
+    bool has_dumped_buffer{};
 
+    const auto handle_debug_buffer = [&](VAddr addr, u64 sz) {
+        if (sz == 0 || addr == 0 || has_dumped_buffer) {
+            return;
+        }
+
+        // This typically is an error code so we're going to assume this is the case
+        if (sz == sizeof(u32)) {
+            LOG_CRITICAL(Debug_Emulated, "debug_buffer_err_code={:X}", Memory::Read32(addr));
+        } else {
+            // We don't know what's in here so we'll hexdump it
+            std::vector<u8> debug_buffer(sz);
+            Memory::ReadBlock(addr, debug_buffer.data(), sz);
+            std::string hexdump;
+            for (std::size_t i = 0; i < debug_buffer.size(); i++) {
+                hexdump += fmt::format("{:02X} ", debug_buffer[i]);
+                if (i != 0 && i % 16 == 0) {
+                    hexdump += '\n';
+                }
+            }
+            LOG_CRITICAL(Debug_Emulated, "debug_buffer=\n{}", hexdump);
+        }
+        has_dumped_buffer = true;
+    };
     switch (break_reason.break_type) {
     case BreakType::Panic:
         LOG_CRITICAL(Debug_Emulated, "Signalling debugger, PANIC! info1=0x{:016X}, info2=0x{:016X}",
                      info1, info2);
+        handle_debug_buffer(info1, info2);
         break;
     case BreakType::AssertionFailed:
         LOG_CRITICAL(Debug_Emulated,
                      "Signalling debugger, Assertion failed! info1=0x{:016X}, info2=0x{:016X}",
                      info1, info2);
+        handle_debug_buffer(info1, info2);
         break;
     case BreakType::PreNROLoad:
         LOG_WARNING(
@@ -433,6 +459,7 @@ static void Break(u32 reason, u64 info1, u64 info2) {
             Debug_Emulated,
             "Signalling debugger, Unknown break reason {}, info1=0x{:016X}, info2=0x{:016X}",
             static_cast<u32>(break_reason.break_type.Value()), info1, info2);
+        handle_debug_buffer(info1, info2);
         break;
     }
 
@@ -441,6 +468,7 @@ static void Break(u32 reason, u64 info1, u64 info2) {
             Debug_Emulated,
             "Emulated program broke execution! reason=0x{:016X}, info1=0x{:016X}, info2=0x{:016X}",
             reason, info1, info2);
+        handle_debug_buffer(info1, info2);
         ASSERT(false);
 
         Core::CurrentProcess()->PrepareForTermination();

src/core/hle/service/hid/controllers/npad.cpp

@@ -392,8 +392,10 @@ std::size_t Controller_NPad::GetSupportedNPadIdTypesSize() const {
 }
 
 void Controller_NPad::SetHoldType(NpadHoldType joy_hold_type) {
+    styleset_changed_event->Signal();
     hold_type = joy_hold_type;
 }
+
 Controller_NPad::NpadHoldType Controller_NPad::GetHoldType() const {
     return hold_type;
 }

src/video_core/engines/maxwell_3d.cpp

@@ -37,6 +37,22 @@ 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;
+    }
 }
 
 void Maxwell3D::CallMacroMethod(u32 method, std::vector<u32> parameters) {

src/video_core/engines/maxwell_3d.h

@@ -462,6 +462,16 @@ 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 +581,11 @@ 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 +660,14 @@ 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 +861,9 @@ 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;
@@ -1075,6 +1096,7 @@ 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 +1109,10 @@ 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 +1143,7 @@ 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);

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)};
 
-    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 = 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));

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};

src/video_core/rasterizer_cache.h

@@ -10,10 +10,8 @@
 #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:
@@ -64,6 +62,8 @@ 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)};
@@ -109,14 +109,12 @@ protected:
     void Register(const T& object) {
         object->SetIsRegistered(true);
         object_cache.add({GetInterval(object), ObjectSet{object}});
-        auto& rasterizer = Core::System::GetInstance().Renderer().Rasterizer();
         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
@@ -177,4 +175,5 @@ private:
 
     ObjectCache object_cache; ///< Cache of objects
     u64 modified_ticks{};     ///< Counter of cache state ticks, used for in-order flushing
+    VideoCore::RasterizerInterface& rasterizer;
 };

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) {

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.

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"

src/video_core/renderer_opengl/gl_rasterizer.cpp

@@ -79,7 +79,8 @@ 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();
@@ -511,10 +512,10 @@ void RasterizerOpenGL::Clear() {
 
     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;
+    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.R || regs.clear_buffers.G || regs.clear_buffers.B ||
         regs.clear_buffers.A) {
@@ -573,14 +574,13 @@ void RasterizerOpenGL::DrawArrays() {
     ScopeAcquireGLContext acquire_context{emu_window};
 
     ConfigureFramebuffers();
-
+    SyncColorMask();
     SyncDepthTestState();
     SyncStencilTestState();
     SyncBlendState();
     SyncLogicOpState();
     SyncCullMode();
     SyncPrimitiveRestart();
-    SyncDepthRange();
     SyncScissorTest();
     // Alpha Testing is synced on shaders.
     SyncTransformFeedback();
@@ -899,12 +899,16 @@ u32 RasterizerOpenGL::SetupTextures(Maxwell::ShaderStage stage, Shader& shader,
 
 void RasterizerOpenGL::SyncViewport() {
     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 = state.viewports[i];
+        viewport.x = viewport_rect.left;
+        viewport.y = viewport_rect.bottom;
+        viewport.width = static_cast<GLsizei>(viewport_rect.GetWidth());
+        viewport.height = static_cast<GLsizei>(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 +950,6 @@ 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;
 
@@ -993,26 +990,60 @@ void RasterizerOpenGL::SyncStencilTestState() {
     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;
+    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;
 
-    if (!state.blend.enabled)
+    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 +1062,19 @@ 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() {

src/video_core/renderer_opengl/gl_rasterizer.h

@@ -133,7 +133,7 @@ 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
+    /// Syncs the viewport and depth range to match the guest state
     void SyncViewport();
 
     /// Syncs the clip enabled status to match the guest state
@@ -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();
 

src/video_core/renderer_opengl/gl_rasterizer_cache.cpp

@@ -15,6 +15,7 @@
 #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"
@@ -562,9 +563,11 @@ 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()};
@@ -704,9 +707,11 @@ 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) {
+    MICROPROFILE_SCOPE(OpenGL_CopySurface);
     ASSERT_MSG(dst_attachment == 0, "Unimplemented");
 
     const auto& src_params{src_surface->GetSurfaceParams()};
@@ -975,7 +980,7 @@ 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);
@@ -1157,7 +1162,7 @@ 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;
@@ -1168,7 +1173,8 @@ 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();

src/video_core/renderer_opengl/gl_rasterizer_cache.h

@@ -264,6 +264,8 @@ struct hash<SurfaceReserveKey> {
 
 namespace OpenGL {
 
+class RasterizerOpenGL;
+
 class CachedSurface final : public RasterizerCacheObject {
 public:
     CachedSurface(const SurfaceParams& params);
@@ -311,7 +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,

src/video_core/renderer_opengl/gl_resource_manager.cpp

@@ -5,21 +5,29 @@
 #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;
@@ -28,12 +36,16 @@ void OGLTexture::Release() {
 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;
@@ -44,12 +56,16 @@ void OGLShader::Create(const char* source, GLenum type) {
         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;
 }
@@ -63,12 +79,16 @@ void OGLProgram::CreateFromSource(const char* vert_shader, const char* geo_shade
         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;
@@ -77,12 +97,16 @@ void OGLProgram::Release() {
 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;
@@ -91,12 +115,16 @@ void OGLPipeline::Release() {
 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;
@@ -105,12 +133,16 @@ void OGLBuffer::Release() {
 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;
 }
@@ -118,12 +150,16 @@ void OGLSync::Release() {
 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;
@@ -132,12 +168,16 @@ void OGLVertexArray::Release() {
 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;

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 {
 
@@ -135,6 +135,8 @@ 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)};
 

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;
 
@@ -104,6 +106,8 @@ private:
 
 class ShaderCacheOpenGL final : public RasterizerCache<Shader> {
 public:
+    explicit ShaderCacheOpenGL(RasterizerOpenGL& rasterizer);
+
     /// Gets the current specified shader stage program
     Shader GetStageProgram(Maxwell::ShaderProgram program);
 };

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 +239,141 @@ 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(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) {
+            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(int 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);
@@ -348,12 +460,6 @@ 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 +470,13 @@ 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();

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;
@@ -146,13 +161,6 @@ public:
         GLsizei height;
     } scissor;
 
-    struct {
-        GLint x;
-        GLint y;
-        GLsizei width;
-        GLsizei height;
-    } viewport;
-
     struct {
         float size; // GL_POINT_SIZE
     } point;
@@ -191,14 +199,18 @@ 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(int target, bool force) const;
+    void ApplyGlobalBlending() const;
     void ApplyBlending() const;
     void ApplyLogicOp() const;
     void ApplyTextures() const;
     void ApplySamplers() const;
+    void ApplyScissor() const;
 };
 
 } // namespace OpenGL

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);