opengl: Sanitize antialiasing config

video_core/vulkan: Fixed loading of Vulkan driver pipeline cache

src/common/settings.cpp

@@ -185,6 +185,7 @@ void RestoreGlobalState(bool is_powered_on) {
     // Renderer
     values.fsr_sharpening_slider.SetGlobal(true);
     values.renderer_backend.SetGlobal(true);
+    values.renderer_force_max_clock.SetGlobal(true);
     values.vulkan_device.SetGlobal(true);
     values.aspect_ratio.SetGlobal(true);
     values.max_anisotropy.SetGlobal(true);
@@ -200,6 +201,7 @@ void RestoreGlobalState(bool is_powered_on) {
     values.use_asynchronous_shaders.SetGlobal(true);
     values.use_fast_gpu_time.SetGlobal(true);
     values.use_pessimistic_flushes.SetGlobal(true);
+    values.use_vulkan_driver_pipeline_cache.SetGlobal(true);
     values.bg_red.SetGlobal(true);
     values.bg_green.SetGlobal(true);
     values.bg_blue.SetGlobal(true);

src/common/settings.h

@@ -415,6 +415,7 @@ struct Values {
     // Renderer
     SwitchableSetting<RendererBackend, true> renderer_backend{
         RendererBackend::Vulkan, RendererBackend::OpenGL, RendererBackend::Null, "backend"};
+    SwitchableSetting<bool> renderer_force_max_clock{true, "force_max_clock"};
     Setting<bool> renderer_debug{false, "debug"};
     Setting<bool> renderer_shader_feedback{false, "shader_feedback"};
     Setting<bool> enable_nsight_aftermath{false, "nsight_aftermath"};
@@ -451,6 +452,8 @@ struct Values {
     SwitchableSetting<bool> use_asynchronous_shaders{false, "use_asynchronous_shaders"};
     SwitchableSetting<bool> use_fast_gpu_time{true, "use_fast_gpu_time"};
     SwitchableSetting<bool> use_pessimistic_flushes{false, "use_pessimistic_flushes"};
+    SwitchableSetting<bool> use_vulkan_driver_pipeline_cache{true,
+                                                             "use_vulkan_driver_pipeline_cache"};
 
     SwitchableSetting<u8> bg_red{0, "bg_red"};
     SwitchableSetting<u8> bg_green{0, "bg_green"};

src/core/arm/dynarmic/arm_dynarmic_32.cpp

@@ -229,7 +229,11 @@ std::shared_ptr<Dynarmic::A32::Jit> ARM_Dynarmic_32::MakeJit(Common::PageTable*
     config.enable_cycle_counting = true;
 
     // Code cache size
+#ifdef ARCHITECTURE_arm64
+    config.code_cache_size = 128_MiB;
+#else
     config.code_cache_size = 512_MiB;
+#endif
 
     // Allow memory fault handling to work
     if (system.DebuggerEnabled()) {

src/core/arm/dynarmic/arm_dynarmic_64.cpp

@@ -288,7 +288,11 @@ std::shared_ptr<Dynarmic::A64::Jit> ARM_Dynarmic_64::MakeJit(Common::PageTable*
     config.enable_cycle_counting = true;
 
     // Code cache size
+#ifdef ARCHITECTURE_arm64
+    config.code_cache_size = 128_MiB;
+#else
     config.code_cache_size = 512_MiB;
+#endif
 
     // Allow memory fault handling to work
     if (system.DebuggerEnabled()) {

src/core/internal_network/network.cpp

@@ -117,6 +117,8 @@ Errno TranslateNativeError(int e) {
         return Errno::NETUNREACH;
     case WSAEMSGSIZE:
         return Errno::MSGSIZE;
+    case WSAETIMEDOUT:
+        return Errno::TIMEDOUT;
     default:
         UNIMPLEMENTED_MSG("Unimplemented errno={}", e);
         return Errno::OTHER;
@@ -211,6 +213,8 @@ Errno TranslateNativeError(int e) {
         return Errno::NETUNREACH;
     case EMSGSIZE:
         return Errno::MSGSIZE;
+    case ETIMEDOUT:
+        return Errno::TIMEDOUT;
     default:
         UNIMPLEMENTED_MSG("Unimplemented errno={}", e);
         return Errno::OTHER;
@@ -226,7 +230,7 @@ Errno GetAndLogLastError() {
     int e = errno;
 #endif
     const Errno err = TranslateNativeError(e);
-    if (err == Errno::AGAIN) {
+    if (err == Errno::AGAIN || err == Errno::TIMEDOUT) {
         return err;
     }
     LOG_ERROR(Network, "Socket operation error: {}", Common::NativeErrorToString(e));

src/core/memory.cpp

@@ -436,7 +436,7 @@ struct Memory::Impl {
         }
 
         if (Settings::IsFastmemEnabled()) {
-            const bool is_read_enable = Settings::IsGPULevelHigh() || !cached;
+            const bool is_read_enable = !Settings::IsGPULevelExtreme() || !cached;
             system.DeviceMemory().buffer.Protect(vaddr, size, is_read_enable, !cached);
         }
 

src/shader_recompiler/backend/spirv/emit_spirv_context_get_set.cpp

@@ -321,8 +321,12 @@ Id EmitGetAttribute(EmitContext& ctx, IR::Attribute attr, Id vertex) {
     case IR::Attribute::PositionY:
     case IR::Attribute::PositionZ:
     case IR::Attribute::PositionW:
-        return ctx.OpLoad(ctx.F32[1], AttrPointer(ctx, ctx.input_f32, vertex, ctx.input_position,
-                                                  ctx.Const(element)));
+        return ctx.OpLoad(
+            ctx.F32[1],
+            ctx.need_input_position_indirect
+                ? AttrPointer(ctx, ctx.input_f32, vertex, ctx.input_position, ctx.u32_zero_value,
+                              ctx.Const(element))
+                : AttrPointer(ctx, ctx.input_f32, vertex, ctx.input_position, ctx.Const(element)));
     case IR::Attribute::InstanceId:
         if (ctx.profile.support_vertex_instance_id) {
             return ctx.OpBitcast(ctx.F32[1], ctx.OpLoad(ctx.U32[1], ctx.instance_id));

src/shader_recompiler/backend/spirv/emit_spirv_warp.cpp

@@ -58,11 +58,10 @@ Id SelectValue(EmitContext& ctx, Id in_range, Id value, Id src_thread_id) {
         ctx.OpGroupNonUniformShuffle(ctx.U32[1], SubgroupScope(ctx), value, src_thread_id), value);
 }
 
-Id GetUpperClamp(EmitContext& ctx, Id invocation_id, Id clamp) {
-    const Id thirty_two{ctx.Const(32u)};
-    const Id is_upper_partition{ctx.OpSGreaterThanEqual(ctx.U1, invocation_id, thirty_two)};
-    const Id upper_clamp{ctx.OpIAdd(ctx.U32[1], thirty_two, clamp)};
-    return ctx.OpSelect(ctx.U32[1], is_upper_partition, upper_clamp, clamp);
+Id AddPartitionBase(EmitContext& ctx, Id thread_id) {
+    const Id partition_idx{ctx.OpShiftRightLogical(ctx.U32[1], GetThreadId(ctx), ctx.Const(5u))};
+    const Id partition_base{ctx.OpShiftLeftLogical(ctx.U32[1], partition_idx, ctx.Const(5u))};
+    return ctx.OpIAdd(ctx.U32[1], thread_id, partition_base);
 }
 } // Anonymous namespace
 
@@ -145,64 +144,63 @@ Id EmitSubgroupGeMask(EmitContext& ctx) {
 Id EmitShuffleIndex(EmitContext& ctx, IR::Inst* inst, Id value, Id index, Id clamp,
                     Id segmentation_mask) {
     const Id not_seg_mask{ctx.OpNot(ctx.U32[1], segmentation_mask)};
-    const Id thread_id{GetThreadId(ctx)};
-    if (ctx.profile.warp_size_potentially_larger_than_guest) {
-        const Id thirty_two{ctx.Const(32u)};
-        const Id is_upper_partition{ctx.OpSGreaterThanEqual(ctx.U1, thread_id, thirty_two)};
-        const Id upper_index{ctx.OpIAdd(ctx.U32[1], thirty_two, index)};
-        const Id upper_clamp{ctx.OpIAdd(ctx.U32[1], thirty_two, clamp)};
-        index = ctx.OpSelect(ctx.U32[1], is_upper_partition, upper_index, index);
-        clamp = ctx.OpSelect(ctx.U32[1], is_upper_partition, upper_clamp, clamp);
-    }
+    const Id thread_id{EmitLaneId(ctx)};
     const Id min_thread_id{ComputeMinThreadId(ctx, thread_id, segmentation_mask)};
     const Id max_thread_id{ComputeMaxThreadId(ctx, min_thread_id, clamp, not_seg_mask)};
 
     const Id lhs{ctx.OpBitwiseAnd(ctx.U32[1], index, not_seg_mask)};
-    const Id src_thread_id{ctx.OpBitwiseOr(ctx.U32[1], lhs, min_thread_id)};
+    Id src_thread_id{ctx.OpBitwiseOr(ctx.U32[1], lhs, min_thread_id)};
     const Id in_range{ctx.OpSLessThanEqual(ctx.U1, src_thread_id, max_thread_id)};
 
+    if (ctx.profile.warp_size_potentially_larger_than_guest) {
+        src_thread_id = AddPartitionBase(ctx, src_thread_id);
+    }
+
     SetInBoundsFlag(inst, in_range);
     return SelectValue(ctx, in_range, value, src_thread_id);
 }
 
 Id EmitShuffleUp(EmitContext& ctx, IR::Inst* inst, Id value, Id index, Id clamp,
                  Id segmentation_mask) {
-    const Id thread_id{GetThreadId(ctx)};
-    if (ctx.profile.warp_size_potentially_larger_than_guest) {
-        clamp = GetUpperClamp(ctx, thread_id, clamp);
-    }
+    const Id thread_id{EmitLaneId(ctx)};
     const Id max_thread_id{GetMaxThreadId(ctx, thread_id, clamp, segmentation_mask)};
-    const Id src_thread_id{ctx.OpISub(ctx.U32[1], thread_id, index)};
+    Id src_thread_id{ctx.OpISub(ctx.U32[1], thread_id, index)};
     const Id in_range{ctx.OpSGreaterThanEqual(ctx.U1, src_thread_id, max_thread_id)};
 
+    if (ctx.profile.warp_size_potentially_larger_than_guest) {
+        src_thread_id = AddPartitionBase(ctx, src_thread_id);
+    }
+
     SetInBoundsFlag(inst, in_range);
     return SelectValue(ctx, in_range, value, src_thread_id);
 }
 
 Id EmitShuffleDown(EmitContext& ctx, IR::Inst* inst, Id value, Id index, Id clamp,
                    Id segmentation_mask) {
-    const Id thread_id{GetThreadId(ctx)};
-    if (ctx.profile.warp_size_potentially_larger_than_guest) {
-        clamp = GetUpperClamp(ctx, thread_id, clamp);
-    }
+    const Id thread_id{EmitLaneId(ctx)};
     const Id max_thread_id{GetMaxThreadId(ctx, thread_id, clamp, segmentation_mask)};
-    const Id src_thread_id{ctx.OpIAdd(ctx.U32[1], thread_id, index)};
+    Id src_thread_id{ctx.OpIAdd(ctx.U32[1], thread_id, index)};
     const Id in_range{ctx.OpSLessThanEqual(ctx.U1, src_thread_id, max_thread_id)};
 
+    if (ctx.profile.warp_size_potentially_larger_than_guest) {
+        src_thread_id = AddPartitionBase(ctx, src_thread_id);
+    }
+
     SetInBoundsFlag(inst, in_range);
     return SelectValue(ctx, in_range, value, src_thread_id);
 }
 
 Id EmitShuffleButterfly(EmitContext& ctx, IR::Inst* inst, Id value, Id index, Id clamp,
                         Id segmentation_mask) {
-    const Id thread_id{GetThreadId(ctx)};
-    if (ctx.profile.warp_size_potentially_larger_than_guest) {
-        clamp = GetUpperClamp(ctx, thread_id, clamp);
-    }
+    const Id thread_id{EmitLaneId(ctx)};
     const Id max_thread_id{GetMaxThreadId(ctx, thread_id, clamp, segmentation_mask)};
-    const Id src_thread_id{ctx.OpBitwiseXor(ctx.U32[1], thread_id, index)};
+    Id src_thread_id{ctx.OpBitwiseXor(ctx.U32[1], thread_id, index)};
     const Id in_range{ctx.OpSLessThanEqual(ctx.U1, src_thread_id, max_thread_id)};
 
+    if (ctx.profile.warp_size_potentially_larger_than_guest) {
+        src_thread_id = AddPartitionBase(ctx, src_thread_id);
+    }
+
     SetInBoundsFlag(inst, in_range);
     return SelectValue(ctx, in_range, value, src_thread_id);
 }

src/shader_recompiler/backend/spirv/spirv_emit_context.cpp

@@ -544,7 +544,7 @@ void EmitContext::DefineCommonTypes(const Info& info) {
         U16 = Name(TypeInt(16, false), "u16");
         S16 = Name(TypeInt(16, true), "s16");
     }
-    if (info.uses_int64) {
+    if (info.uses_int64 && profile.support_int64) {
         AddCapability(spv::Capability::Int64);
         U64 = Name(TypeInt(64, false), "u64");
     }
@@ -721,9 +721,21 @@ void EmitContext::DefineAttributeMemAccess(const Info& info) {
         size_t label_index{0};
         if (info.loads.AnyComponent(IR::Attribute::PositionX)) {
             AddLabel(labels[label_index]);
-            const Id pointer{is_array
-                                 ? OpAccessChain(input_f32, input_position, vertex, masked_index)
-                                 : OpAccessChain(input_f32, input_position, masked_index)};
+            const Id pointer{[&]() {
+                if (need_input_position_indirect) {
+                    if (is_array)
+                        return OpAccessChain(input_f32, input_position, vertex, u32_zero_value,
+                                             masked_index);
+                    else
+                        return OpAccessChain(input_f32, input_position, u32_zero_value,
+                                             masked_index);
+                } else {
+                    if (is_array)
+                        return OpAccessChain(input_f32, input_position, vertex, masked_index);
+                    else
+                        return OpAccessChain(input_f32, input_position, masked_index);
+                }
+            }()};
             const Id result{OpLoad(F32[1], pointer)};
             OpReturnValue(result);
             ++label_index;
@@ -1367,12 +1379,25 @@ void EmitContext::DefineInputs(const IR::Program& program) {
         Decorate(layer, spv::Decoration::Flat);
     }
     if (loads.AnyComponent(IR::Attribute::PositionX)) {
-        const bool is_fragment{stage != Stage::Fragment};
-        const spv::BuiltIn built_in{is_fragment ? spv::BuiltIn::Position : spv::BuiltIn::FragCoord};
-        input_position = DefineInput(*this, F32[4], true, built_in);
-        if (profile.support_geometry_shader_passthrough) {
-            if (info.passthrough.AnyComponent(IR::Attribute::PositionX)) {
-                Decorate(input_position, spv::Decoration::PassthroughNV);
+        const bool is_fragment{stage == Stage::Fragment};
+        if (!is_fragment && profile.has_broken_spirv_position_input) {
+            need_input_position_indirect = true;
+
+            const Id input_position_struct = TypeStruct(F32[4]);
+            input_position = DefineInput(*this, input_position_struct, true);
+
+            MemberDecorate(input_position_struct, 0, spv::Decoration::BuiltIn,
+                           static_cast<unsigned>(spv::BuiltIn::Position));
+            Decorate(input_position_struct, spv::Decoration::Block);
+        } else {
+            const spv::BuiltIn built_in{is_fragment ? spv::BuiltIn::FragCoord
+                                                    : spv::BuiltIn::Position};
+            input_position = DefineInput(*this, F32[4], true, built_in);
+
+            if (profile.support_geometry_shader_passthrough) {
+                if (info.passthrough.AnyComponent(IR::Attribute::PositionX)) {
+                    Decorate(input_position, spv::Decoration::PassthroughNV);
+                }
             }
         }
     }

src/shader_recompiler/backend/spirv/spirv_emit_context.h

@@ -280,6 +280,7 @@ public:
     Id write_global_func_u32x2{};
     Id write_global_func_u32x4{};
 
+    bool need_input_position_indirect{};
     Id input_position{};
     std::array<Id, 32> input_generics{};
 

src/shader_recompiler/frontend/maxwell/translate_program.cpp

@@ -171,6 +171,70 @@ std::map<IR::Attribute, IR::Attribute> GenerateLegacyToGenericMappings(
     }
     return mapping;
 }
+
+void EmitGeometryPassthrough(IR::IREmitter& ir, const IR::Program& program,
+                             const Shader::VaryingState& passthrough_mask,
+                             bool passthrough_position,
+                             std::optional<IR::Attribute> passthrough_layer_attr) {
+    for (u32 i = 0; i < program.output_vertices; i++) {
+        // Assign generics from input
+        for (u32 j = 0; j < 32; j++) {
+            if (!passthrough_mask.Generic(j)) {
+                continue;
+            }
+
+            const IR::Attribute attr = IR::Attribute::Generic0X + (j * 4);
+            ir.SetAttribute(attr + 0, ir.GetAttribute(attr + 0, ir.Imm32(i)), ir.Imm32(0));
+            ir.SetAttribute(attr + 1, ir.GetAttribute(attr + 1, ir.Imm32(i)), ir.Imm32(0));
+            ir.SetAttribute(attr + 2, ir.GetAttribute(attr + 2, ir.Imm32(i)), ir.Imm32(0));
+            ir.SetAttribute(attr + 3, ir.GetAttribute(attr + 3, ir.Imm32(i)), ir.Imm32(0));
+        }
+
+        if (passthrough_position) {
+            // Assign position from input
+            const IR::Attribute attr = IR::Attribute::PositionX;
+            ir.SetAttribute(attr + 0, ir.GetAttribute(attr + 0, ir.Imm32(i)), ir.Imm32(0));
+            ir.SetAttribute(attr + 1, ir.GetAttribute(attr + 1, ir.Imm32(i)), ir.Imm32(0));
+            ir.SetAttribute(attr + 2, ir.GetAttribute(attr + 2, ir.Imm32(i)), ir.Imm32(0));
+            ir.SetAttribute(attr + 3, ir.GetAttribute(attr + 3, ir.Imm32(i)), ir.Imm32(0));
+        }
+
+        if (passthrough_layer_attr) {
+            // Assign layer
+            ir.SetAttribute(IR::Attribute::Layer, ir.GetAttribute(*passthrough_layer_attr),
+                            ir.Imm32(0));
+        }
+
+        // Emit vertex
+        ir.EmitVertex(ir.Imm32(0));
+    }
+    ir.EndPrimitive(ir.Imm32(0));
+}
+
+u32 GetOutputTopologyVertices(OutputTopology output_topology) {
+    switch (output_topology) {
+    case OutputTopology::PointList:
+        return 1;
+    case OutputTopology::LineStrip:
+        return 2;
+    default:
+        return 3;
+    }
+}
+
+void LowerGeometryPassthrough(const IR::Program& program, const HostTranslateInfo& host_info) {
+    for (IR::Block* const block : program.blocks) {
+        for (IR::Inst& inst : block->Instructions()) {
+            if (inst.GetOpcode() == IR::Opcode::Epilogue) {
+                IR::IREmitter ir{*block, IR::Block::InstructionList::s_iterator_to(inst)};
+                EmitGeometryPassthrough(
+                    ir, program, program.info.passthrough,
+                    program.info.passthrough.AnyComponent(IR::Attribute::PositionX), {});
+            }
+        }
+    }
+}
+
 } // Anonymous namespace
 
 IR::Program TranslateProgram(ObjectPool<IR::Inst>& inst_pool, ObjectPool<IR::Block>& block_pool,
@@ -198,6 +262,11 @@ IR::Program TranslateProgram(ObjectPool<IR::Inst>& inst_pool, ObjectPool<IR::Blo
             for (size_t i = 0; i < program.info.passthrough.mask.size(); ++i) {
                 program.info.passthrough.mask[i] = ((mask[i / 32] >> (i % 32)) & 1) == 0;
             }
+
+            if (!host_info.support_geometry_shader_passthrough) {
+                program.output_vertices = GetOutputTopologyVertices(program.output_topology);
+                LowerGeometryPassthrough(program, host_info);
+            }
         }
         break;
     }
@@ -223,7 +292,7 @@ IR::Program TranslateProgram(ObjectPool<IR::Inst>& inst_pool, ObjectPool<IR::Blo
 
     Optimization::PositionPass(env, program);
 
-    Optimization::GlobalMemoryToStorageBufferPass(program);
+    Optimization::GlobalMemoryToStorageBufferPass(program, host_info);
     Optimization::TexturePass(env, program, host_info);
 
     if (Settings::values.resolution_info.active) {
@@ -342,17 +411,7 @@ IR::Program GenerateGeometryPassthrough(ObjectPool<IR::Inst>& inst_pool,
     IR::Program program;
     program.stage = Stage::Geometry;
     program.output_topology = output_topology;
-    switch (output_topology) {
-    case OutputTopology::PointList:
-        program.output_vertices = 1;
-        break;
-    case OutputTopology::LineStrip:
-        program.output_vertices = 2;
-        break;
-    default:
-        program.output_vertices = 3;
-        break;
-    }
+    program.output_vertices = GetOutputTopologyVertices(output_topology);
 
     program.is_geometry_passthrough = false;
     program.info.loads.mask = source_program.info.stores.mask;
@@ -366,35 +425,8 @@ IR::Program GenerateGeometryPassthrough(ObjectPool<IR::Inst>& inst_pool,
     node.data.block = current_block;
 
     IR::IREmitter ir{*current_block};
-    for (u32 i = 0; i < program.output_vertices; i++) {
-        // Assign generics from input
-        for (u32 j = 0; j < 32; j++) {
-            if (!program.info.stores.Generic(j)) {
-                continue;
-            }
-
-            const IR::Attribute attr = IR::Attribute::Generic0X + (j * 4);
-            ir.SetAttribute(attr + 0, ir.GetAttribute(attr + 0, ir.Imm32(i)), ir.Imm32(0));
-            ir.SetAttribute(attr + 1, ir.GetAttribute(attr + 1, ir.Imm32(i)), ir.Imm32(0));
-            ir.SetAttribute(attr + 2, ir.GetAttribute(attr + 2, ir.Imm32(i)), ir.Imm32(0));
-            ir.SetAttribute(attr + 3, ir.GetAttribute(attr + 3, ir.Imm32(i)), ir.Imm32(0));
-        }
-
-        // Assign position from input
-        const IR::Attribute attr = IR::Attribute::PositionX;
-        ir.SetAttribute(attr + 0, ir.GetAttribute(attr + 0, ir.Imm32(i)), ir.Imm32(0));
-        ir.SetAttribute(attr + 1, ir.GetAttribute(attr + 1, ir.Imm32(i)), ir.Imm32(0));
-        ir.SetAttribute(attr + 2, ir.GetAttribute(attr + 2, ir.Imm32(i)), ir.Imm32(0));
-        ir.SetAttribute(attr + 3, ir.GetAttribute(attr + 3, ir.Imm32(i)), ir.Imm32(0));
-
-        // Assign layer
-        ir.SetAttribute(IR::Attribute::Layer, ir.GetAttribute(source_program.info.emulated_layer),
-                        ir.Imm32(0));
-
-        // Emit vertex
-        ir.EmitVertex(ir.Imm32(0));
-    }
-    ir.EndPrimitive(ir.Imm32(0));
+    EmitGeometryPassthrough(ir, program, program.info.stores, true,
+                            source_program.info.emulated_layer);
 
     IR::Block* return_block{block_pool.Create(inst_pool)};
     IR::IREmitter{*return_block}.Epilogue();

src/shader_recompiler/host_translate_info.h

@@ -15,6 +15,9 @@ struct HostTranslateInfo {
     bool needs_demote_reorder{}; ///< True when the device needs DemoteToHelperInvocation reordered
     bool support_snorm_render_buffer{};  ///< True when the device supports SNORM render buffers
     bool support_viewport_index_layer{}; ///< True when the device supports gl_Layer in VS
+    u32 min_ssbo_alignment{};            ///< Minimum alignment supported by the device for SSBOs
+    bool support_geometry_shader_passthrough{}; ///< True when the device supports geometry
+                                                ///< passthrough shaders
 };
 
 } // namespace Shader

src/shader_recompiler/ir_opt/global_memory_to_storage_buffer_pass.cpp

@@ -11,6 +11,7 @@
 #include "shader_recompiler/frontend/ir/breadth_first_search.h"
 #include "shader_recompiler/frontend/ir/ir_emitter.h"
 #include "shader_recompiler/frontend/ir/value.h"
+#include "shader_recompiler/host_translate_info.h"
 #include "shader_recompiler/ir_opt/passes.h"
 
 namespace Shader::Optimization {
@@ -402,7 +403,7 @@ void CollectStorageBuffers(IR::Block& block, IR::Inst& inst, StorageInfo& info)
 }
 
 /// Returns the offset in indices (not bytes) for an equivalent storage instruction
-IR::U32 StorageOffset(IR::Block& block, IR::Inst& inst, StorageBufferAddr buffer) {
+IR::U32 StorageOffset(IR::Block& block, IR::Inst& inst, StorageBufferAddr buffer, u32 alignment) {
     IR::IREmitter ir{block, IR::Block::InstructionList::s_iterator_to(inst)};
     IR::U32 offset;
     if (const std::optional<LowAddrInfo> low_addr{TrackLowAddress(&inst)}) {
@@ -415,7 +416,10 @@ IR::U32 StorageOffset(IR::Block& block, IR::Inst& inst, StorageBufferAddr buffer
     }
     // Subtract the least significant 32 bits from the guest offset. The result is the storage
     // buffer offset in bytes.
-    const IR::U32 low_cbuf{ir.GetCbuf(ir.Imm32(buffer.index), ir.Imm32(buffer.offset))};
+    IR::U32 low_cbuf{ir.GetCbuf(ir.Imm32(buffer.index), ir.Imm32(buffer.offset))};
+
+    // Align the offset base to match the host alignment requirements
+    low_cbuf = ir.BitwiseAnd(low_cbuf, ir.Imm32(~(alignment - 1U)));
     return ir.ISub(offset, low_cbuf);
 }
 
@@ -510,7 +514,7 @@ void Replace(IR::Block& block, IR::Inst& inst, const IR::U32& storage_index,
 }
 } // Anonymous namespace
 
-void GlobalMemoryToStorageBufferPass(IR::Program& program) {
+void GlobalMemoryToStorageBufferPass(IR::Program& program, const HostTranslateInfo& host_info) {
     StorageInfo info;
     for (IR::Block* const block : program.post_order_blocks) {
         for (IR::Inst& inst : block->Instructions()) {
@@ -534,7 +538,8 @@ void GlobalMemoryToStorageBufferPass(IR::Program& program) {
         const IR::U32 index{IR::Value{static_cast<u32>(info.set.index_of(it))}};
         IR::Block* const block{storage_inst.block};
         IR::Inst* const inst{storage_inst.inst};
-        const IR::U32 offset{StorageOffset(*block, *inst, storage_buffer)};
+        const IR::U32 offset{
+            StorageOffset(*block, *inst, storage_buffer, host_info.min_ssbo_alignment)};
         Replace(*block, *inst, index, offset);
     }
 }

src/shader_recompiler/ir_opt/passes.h

@@ -15,7 +15,7 @@ namespace Shader::Optimization {
 void CollectShaderInfoPass(Environment& env, IR::Program& program);
 void ConstantPropagationPass(Environment& env, IR::Program& program);
 void DeadCodeEliminationPass(IR::Program& program);
-void GlobalMemoryToStorageBufferPass(IR::Program& program);
+void GlobalMemoryToStorageBufferPass(IR::Program& program, const HostTranslateInfo& host_info);
 void IdentityRemovalPass(IR::Program& program);
 void LowerFp16ToFp32(IR::Program& program);
 void LowerInt64ToInt32(IR::Program& program);

src/shader_recompiler/profile.h

@@ -55,6 +55,8 @@ struct Profile {
 
     /// OpFClamp is broken and OpFMax + OpFMin should be used instead
     bool has_broken_spirv_clamp{};
+    /// The Position builtin needs to be wrapped in a struct when used as an input
+    bool has_broken_spirv_position_input{};
     /// Offset image operands with an unsigned type do not work
     bool has_broken_unsigned_image_offsets{};
     /// Signed instructions with unsigned data types are misinterpreted

src/shader_recompiler/shader_info.h

@@ -65,6 +65,8 @@ enum class Interpolation {
 struct ConstantBufferDescriptor {
     u32 index;
     u32 count;
+
+    auto operator<=>(const ConstantBufferDescriptor&) const = default;
 };
 
 struct StorageBufferDescriptor {
@@ -72,6 +74,8 @@ struct StorageBufferDescriptor {
     u32 cbuf_offset;
     u32 count;
     bool is_written;
+
+    auto operator<=>(const StorageBufferDescriptor&) const = default;
 };
 
 struct TextureBufferDescriptor {
@@ -84,6 +88,8 @@ struct TextureBufferDescriptor {
     u32 secondary_shift_left;
     u32 count;
     u32 size_shift;
+
+    auto operator<=>(const TextureBufferDescriptor&) const = default;
 };
 using TextureBufferDescriptors = boost::container::small_vector<TextureBufferDescriptor, 6>;
 
@@ -95,6 +101,8 @@ struct ImageBufferDescriptor {
     u32 cbuf_offset;
     u32 count;
     u32 size_shift;
+
+    auto operator<=>(const ImageBufferDescriptor&) const = default;
 };
 using ImageBufferDescriptors = boost::container::small_vector<ImageBufferDescriptor, 2>;
 
@@ -110,6 +118,8 @@ struct TextureDescriptor {
     u32 secondary_shift_left;
     u32 count;
     u32 size_shift;
+
+    auto operator<=>(const TextureDescriptor&) const = default;
 };
 using TextureDescriptors = boost::container::small_vector<TextureDescriptor, 12>;
 
@@ -122,6 +132,8 @@ struct ImageDescriptor {
     u32 cbuf_offset;
     u32 count;
     u32 size_shift;
+
+    auto operator<=>(const ImageDescriptor&) const = default;
 };
 using ImageDescriptors = boost::container::small_vector<ImageDescriptor, 4>;
 

src/tests/video_core/buffer_base.cpp

@@ -538,7 +538,7 @@ TEST_CASE("BufferBase: Cached write downloads") {
     int num = 0;
     buffer.ForEachDownloadRangeAndClear(c, WORD, [&](u64 offset, u64 size) { ++num; });
     buffer.ForEachUploadRange(c, WORD, [&](u64 offset, u64 size) { ++num; });
-    REQUIRE(num == 0);
+    REQUIRE(num == 1);
     REQUIRE(!buffer.IsRegionCpuModified(c + PAGE, PAGE));
     REQUIRE(!buffer.IsRegionGpuModified(c + PAGE, PAGE));
     buffer.FlushCachedWrites();

src/video_core/CMakeLists.txt

@@ -85,6 +85,7 @@ add_library(video_core STATIC
     gpu.h
     gpu_thread.cpp
     gpu_thread.h
+    invalidation_accumulator.h
     memory_manager.cpp
     memory_manager.h
     precompiled_headers.h
@@ -99,8 +100,6 @@ add_library(video_core STATIC
     renderer_null/null_rasterizer.h
     renderer_null/renderer_null.cpp
     renderer_null/renderer_null.h
-    renderer_opengl/blit_image.cpp
-    renderer_opengl/blit_image.h
     renderer_opengl/gl_buffer_cache.cpp
     renderer_opengl/gl_buffer_cache.h
     renderer_opengl/gl_compute_pipeline.cpp
@@ -192,6 +191,8 @@ add_library(video_core STATIC
     renderer_vulkan/vk_texture_cache.cpp
     renderer_vulkan/vk_texture_cache.h
     renderer_vulkan/vk_texture_cache_base.cpp
+    renderer_vulkan/vk_turbo_mode.cpp
+    renderer_vulkan/vk_turbo_mode.h
     renderer_vulkan/vk_update_descriptor.cpp
     renderer_vulkan/vk_update_descriptor.h
     shader_cache.cpp

src/video_core/buffer_cache/buffer_base.h

@@ -430,7 +430,7 @@ private:
         if (query_begin >= SizeBytes() || size < 0) {
             return;
         }
-        u64* const untracked_words = Array<Type::Untracked>();
+        [[maybe_unused]] u64* const untracked_words = Array<Type::Untracked>();
         u64* const state_words = Array<type>();
         const u64 query_end = query_begin + std::min(static_cast<u64>(size), SizeBytes());
         u64* const words_begin = state_words + query_begin / BYTES_PER_WORD;
@@ -483,7 +483,7 @@ private:
                 NotifyRasterizer<true>(word_index, current_bits, ~u64{0});
             }
             // Exclude CPU modified pages when visiting GPU pages
-            const u64 word = current_word & ~(type == Type::GPU ? untracked_words[word_index] : 0);
+            const u64 word = current_word;
             u64 page = page_begin;
             page_begin = 0;
 
@@ -531,7 +531,7 @@ private:
     [[nodiscard]] bool IsRegionModified(u64 offset, u64 size) const noexcept {
         static_assert(type != Type::Untracked);
 
-        const u64* const untracked_words = Array<Type::Untracked>();
+        [[maybe_unused]] const u64* const untracked_words = Array<Type::Untracked>();
         const u64* const state_words = Array<type>();
         const u64 num_query_words = size / BYTES_PER_WORD + 1;
         const u64 word_begin = offset / BYTES_PER_WORD;
@@ -539,8 +539,7 @@ private:
         const u64 page_limit = Common::DivCeil(offset + size, BYTES_PER_PAGE);
         u64 page_index = (offset / BYTES_PER_PAGE) % PAGES_PER_WORD;
         for (u64 word_index = word_begin; word_index < word_end; ++word_index, page_index = 0) {
-            const u64 off_word = type == Type::GPU ? untracked_words[word_index] : 0;
-            const u64 word = state_words[word_index] & ~off_word;
+            const u64 word = state_words[word_index];
             if (word == 0) {
                 continue;
             }
@@ -564,7 +563,7 @@ private:
     [[nodiscard]] std::pair<u64, u64> ModifiedRegion(u64 offset, u64 size) const noexcept {
         static_assert(type != Type::Untracked);
 
-        const u64* const untracked_words = Array<Type::Untracked>();
+        [[maybe_unused]] const u64* const untracked_words = Array<Type::Untracked>();
         const u64* const state_words = Array<type>();
         const u64 num_query_words = size / BYTES_PER_WORD + 1;
         const u64 word_begin = offset / BYTES_PER_WORD;
@@ -574,8 +573,7 @@ private:
         u64 begin = std::numeric_limits<u64>::max();
         u64 end = 0;
         for (u64 word_index = word_begin; word_index < word_end; ++word_index) {
-            const u64 off_word = type == Type::GPU ? untracked_words[word_index] : 0;
-            const u64 word = state_words[word_index] & ~off_word;
+            const u64 word = state_words[word_index];
             if (word == 0) {
                 continue;
             }

src/video_core/buffer_cache/buffer_cache.h

@@ -1938,14 +1938,21 @@ typename BufferCache<P>::Binding BufferCache<P>::StorageBufferBinding(GPUVAddr s
                                                                       bool is_written) const {
     const GPUVAddr gpu_addr = gpu_memory->Read<u64>(ssbo_addr);
     const u32 size = gpu_memory->Read<u32>(ssbo_addr + 8);
-    const std::optional<VAddr> cpu_addr = gpu_memory->GpuToCpuAddress(gpu_addr);
+    const u32 alignment = runtime.GetStorageBufferAlignment();
+
+    const GPUVAddr aligned_gpu_addr = Common::AlignDown(gpu_addr, alignment);
+    const u32 aligned_size =
+        Common::AlignUp(static_cast<u32>(gpu_addr - aligned_gpu_addr) + size, alignment);
+
+    const std::optional<VAddr> cpu_addr = gpu_memory->GpuToCpuAddress(aligned_gpu_addr);
     if (!cpu_addr || size == 0) {
         return NULL_BINDING;
     }
-    const VAddr cpu_end = Common::AlignUp(*cpu_addr + size, Core::Memory::YUZU_PAGESIZE);
+
+    const VAddr cpu_end = Common::AlignUp(*cpu_addr + aligned_size, Core::Memory::YUZU_PAGESIZE);
     const Binding binding{
         .cpu_addr = *cpu_addr,
-        .size = is_written ? size : static_cast<u32>(cpu_end - *cpu_addr),
+        .size = is_written ? aligned_size : static_cast<u32>(cpu_end - *cpu_addr),
         .buffer_id = BufferId{},
     };
     return binding;

src/video_core/engines/draw_manager.cpp

@@ -51,10 +51,6 @@ void DrawManager::ProcessMethodCall(u32 method, u32 argument) {
         LOG_WARNING(HW_GPU, "(STUBBED) called");
         break;
     }
-    case MAXWELL3D_REG_INDEX(draw_texture.src_y0): {
-        DrawTexture();
-        break;
-    }
     default:
         break;
     }
@@ -183,33 +179,6 @@ void DrawManager::DrawIndexSmall(u32 argument) {
     ProcessDraw(true, 1);
 }
 
-void DrawManager::DrawTexture() {
-    const auto& regs{maxwell3d->regs};
-    draw_texture_state.dst_x0 = static_cast<float>(regs.draw_texture.dst_x0) / 4096.f;
-    draw_texture_state.dst_y0 = static_cast<float>(regs.draw_texture.dst_y0) / 4096.f;
-    const auto dst_width = static_cast<float>(regs.draw_texture.dst_width) / 4096.f;
-    const auto dst_height = static_cast<float>(regs.draw_texture.dst_height) / 4096.f;
-    const bool lower_left{regs.window_origin.mode !=
-                          Maxwell3D::Regs::WindowOrigin::Mode::UpperLeft};
-    if (lower_left) {
-        draw_texture_state.dst_y0 -= dst_height;
-    }
-    draw_texture_state.dst_x1 = draw_texture_state.dst_x0 + dst_width;
-    draw_texture_state.dst_y1 = draw_texture_state.dst_y0 + dst_height;
-    draw_texture_state.src_x0 = static_cast<float>(regs.draw_texture.src_x0) / 4096.f;
-    draw_texture_state.src_y0 = static_cast<float>(regs.draw_texture.src_y0) / 4096.f;
-    draw_texture_state.src_x1 =
-        (static_cast<float>(regs.draw_texture.dx_du) / 4294967296.f) * dst_width +
-        draw_texture_state.src_x0;
-    draw_texture_state.src_y1 =
-        (static_cast<float>(regs.draw_texture.dy_dv) / 4294967296.f) * dst_height +
-        draw_texture_state.src_y0;
-    draw_texture_state.src_sampler = regs.draw_texture.src_sampler;
-    draw_texture_state.src_texture = regs.draw_texture.src_texture;
-
-    maxwell3d->rasterizer->DrawTexture();
-}
-
 void DrawManager::UpdateTopology() {
     const auto& regs{maxwell3d->regs};
     switch (regs.primitive_topology_control) {

src/video_core/engines/draw_manager.h

@@ -32,19 +32,6 @@ public:
         std::vector<u8> inline_index_draw_indexes;
     };
 
-    struct DrawTextureState {
-        f32 dst_x0;
-        f32 dst_y0;
-        f32 dst_x1;
-        f32 dst_y1;
-        f32 src_x0;
-        f32 src_y0;
-        f32 src_x1;
-        f32 src_y1;
-        u32 src_sampler;
-        u32 src_texture;
-    };
-
     struct IndirectParams {
         bool is_indexed;
         bool include_count;
@@ -77,10 +64,6 @@ public:
         return draw_state;
     }
 
-    const DrawTextureState& GetDrawTextureState() const {
-        return draw_texture_state;
-    }
-
     IndirectParams& GetIndirectParams() {
         return indirect_state;
     }
@@ -98,8 +81,6 @@ private:
 
     void DrawIndexSmall(u32 argument);
 
-    void DrawTexture();
-
     void UpdateTopology();
 
     void ProcessDraw(bool draw_indexed, u32 instance_count);
@@ -108,7 +89,6 @@ private:
 
     Maxwell3D* maxwell3d{};
     State draw_state{};
-    DrawTextureState draw_texture_state{};
     IndirectParams indirect_state{};
 };
 } // namespace Tegra::Engines

src/video_core/engines/engine_upload.cpp

@@ -76,7 +76,7 @@ void State::ProcessData(std::span<const u8> read_buffer) {
                                        regs.dest.height, regs.dest.depth, x_offset, regs.dest.y,
                                        x_elements, regs.line_count, regs.dest.BlockHeight(),
                                        regs.dest.BlockDepth(), regs.line_length_in);
-        memory_manager.WriteBlock(address, tmp_buffer.data(), dst_size);
+        memory_manager.WriteBlockCached(address, tmp_buffer.data(), dst_size);
     }
 }
 

src/video_core/engines/fermi_2d.cpp

@@ -6,6 +6,7 @@
 #include "common/microprofile.h"
 #include "video_core/engines/fermi_2d.h"
 #include "video_core/engines/sw_blitter/blitter.h"
+#include "video_core/memory_manager.h"
 #include "video_core/rasterizer_interface.h"
 #include "video_core/surface.h"
 #include "video_core/textures/decoders.h"
@@ -20,8 +21,8 @@ namespace Tegra::Engines {
 
 using namespace Texture;
 
-Fermi2D::Fermi2D(MemoryManager& memory_manager_) {
-    sw_blitter = std::make_unique<Blitter::SoftwareBlitEngine>(memory_manager_);
+Fermi2D::Fermi2D(MemoryManager& memory_manager_) : memory_manager{memory_manager_} {
+    sw_blitter = std::make_unique<Blitter::SoftwareBlitEngine>(memory_manager);
     // Nvidia's OpenGL driver seems to assume these values
     regs.src.depth = 1;
     regs.dst.depth = 1;
@@ -104,6 +105,7 @@ void Fermi2D::Blit() {
         config.src_x0 = 0;
     }
 
+    memory_manager.FlushCaching();
     if (!rasterizer->AccelerateSurfaceCopy(src, regs.dst, config)) {
         sw_blitter->Blit(src, regs.dst, config);
     }

src/video_core/engines/fermi_2d.h

@@ -305,6 +305,7 @@ public:
 private:
     VideoCore::RasterizerInterface* rasterizer = nullptr;
     std::unique_ptr<Blitter::SoftwareBlitEngine> sw_blitter;
+    MemoryManager& memory_manager;
 
     /// Performs the copy from the source surface to the destination surface as configured in the
     /// registers.

src/video_core/engines/maxwell_3d.cpp

@@ -149,7 +149,6 @@ bool Maxwell3D::IsMethodExecutable(u32 method) {
     case MAXWELL3D_REG_INDEX(inline_index_4x8.index0):
     case MAXWELL3D_REG_INDEX(vertex_array_instance_first):
     case MAXWELL3D_REG_INDEX(vertex_array_instance_subsequent):
-    case MAXWELL3D_REG_INDEX(draw_texture.src_y0):
     case MAXWELL3D_REG_INDEX(wait_for_idle):
     case MAXWELL3D_REG_INDEX(shadow_ram_control):
     case MAXWELL3D_REG_INDEX(load_mme.instruction_ptr):
@@ -486,11 +485,6 @@ void Maxwell3D::StampQueryResult(u64 payload, bool long_query) {
 }
 
 void Maxwell3D::ProcessQueryGet() {
-    // TODO(Subv): Support the other query units.
-    if (regs.report_semaphore.query.location != Regs::ReportSemaphore::Location::All) {
-        LOG_DEBUG(HW_GPU, "Locations other than ALL are unimplemented");
-    }
-
     switch (regs.report_semaphore.query.operation) {
     case Regs::ReportSemaphore::Operation::Release:
         if (regs.report_semaphore.query.short_query != 0) {
@@ -650,7 +644,7 @@ void Maxwell3D::ProcessCBMultiData(const u32* start_base, u32 amount) {
 
     const GPUVAddr address{buffer_address + regs.const_buffer.offset};
     const size_t copy_size = amount * sizeof(u32);
-    memory_manager.WriteBlock(address, start_base, copy_size);
+    memory_manager.WriteBlockCached(address, start_base, copy_size);
 
     // Increment the current buffer position.
     regs.const_buffer.offset += static_cast<u32>(copy_size);

src/video_core/engines/maxwell_3d.h

@@ -1599,20 +1599,6 @@ public:
         };
         static_assert(sizeof(TIRModulationCoeff) == 0x4);
 
-        struct DrawTexture {
-            s32 dst_x0;
-            s32 dst_y0;
-            s32 dst_width;
-            s32 dst_height;
-            s64 dx_du;
-            s64 dy_dv;
-            u32 src_sampler;
-            u32 src_texture;
-            s32 src_x0;
-            s32 src_y0;
-        };
-        static_assert(sizeof(DrawTexture) == 0x30);
-
         struct ReduceColorThreshold {
             union {
                 BitField<0, 8, u32> all_hit_once;
@@ -2765,7 +2751,7 @@ public:
                 u32 reserved_sw_method2;                                               ///< 0x102C
                 std::array<TIRModulationCoeff, 5> tir_modulation_coeff;                ///< 0x1030
                 std::array<u32, 15> spare_nop;                                         ///< 0x1044
-                DrawTexture draw_texture;                                              ///< 0x1080
+                INSERT_PADDING_BYTES_NOINIT(0x30);
                 std::array<u32, 7> reserved_sw_method3_to_7;                           ///< 0x10B0
                 ReduceColorThreshold reduce_color_thresholds_unorm8;                   ///< 0x10CC
                 std::array<u32, 4> reserved_sw_method10_to_13;                         ///< 0x10D0

src/video_core/engines/maxwell_dma.cpp

@@ -69,7 +69,7 @@ void MaxwellDMA::Launch() {
     if (launch.multi_line_enable) {
         const bool is_src_pitch = launch.src_memory_layout == LaunchDMA::MemoryLayout::PITCH;
         const bool is_dst_pitch = launch.dst_memory_layout == LaunchDMA::MemoryLayout::PITCH;
-
+        memory_manager.FlushCaching();
         if (!is_src_pitch && !is_dst_pitch) {
             // If both the source and the destination are in block layout, assert.
             CopyBlockLinearToBlockLinear();
@@ -104,6 +104,7 @@ void MaxwellDMA::Launch() {
                                             reinterpret_cast<u8*>(tmp_buffer.data()),
                                             regs.line_length_in * sizeof(u32));
         } else {
+            memory_manager.FlushCaching();
             const auto convert_linear_2_blocklinear_addr = [](u64 address) {
                 return (address & ~0x1f0ULL) | ((address & 0x40) >> 2) | ((address & 0x10) << 1) |
                        ((address & 0x180) >> 1) | ((address & 0x20) << 3);
@@ -121,8 +122,8 @@ void MaxwellDMA::Launch() {
                     memory_manager.ReadBlockUnsafe(
                         convert_linear_2_blocklinear_addr(regs.offset_in + offset),
                         tmp_buffer.data(), tmp_buffer.size());
-                    memory_manager.WriteBlock(regs.offset_out + offset, tmp_buffer.data(),
-                                              tmp_buffer.size());
+                    memory_manager.WriteBlockCached(regs.offset_out + offset, tmp_buffer.data(),
+                                                    tmp_buffer.size());
                 }
             } else if (is_src_pitch && !is_dst_pitch) {
                 UNIMPLEMENTED_IF(regs.line_length_in % 16 != 0);
@@ -132,7 +133,7 @@ void MaxwellDMA::Launch() {
                 for (u32 offset = 0; offset < regs.line_length_in; offset += 16) {
                     memory_manager.ReadBlockUnsafe(regs.offset_in + offset, tmp_buffer.data(),
                                                    tmp_buffer.size());
-                    memory_manager.WriteBlock(
+                    memory_manager.WriteBlockCached(
                         convert_linear_2_blocklinear_addr(regs.offset_out + offset),
                         tmp_buffer.data(), tmp_buffer.size());
                 }
@@ -141,8 +142,8 @@ void MaxwellDMA::Launch() {
                     std::vector<u8> tmp_buffer(regs.line_length_in);
                     memory_manager.ReadBlockUnsafe(regs.offset_in, tmp_buffer.data(),
                                                    regs.line_length_in);
-                    memory_manager.WriteBlock(regs.offset_out, tmp_buffer.data(),
-                                              regs.line_length_in);
+                    memory_manager.WriteBlockCached(regs.offset_out, tmp_buffer.data(),
+                                                    regs.line_length_in);
                 }
             }
         }
@@ -204,7 +205,7 @@ void MaxwellDMA::CopyBlockLinearToPitch() {
                      src_params.origin.y, x_elements, regs.line_count, block_height, block_depth,
                      regs.pitch_out);
 
-    memory_manager.WriteBlock(regs.offset_out, write_buffer.data(), dst_size);
+    memory_manager.WriteBlockCached(regs.offset_out, write_buffer.data(), dst_size);
 }
 
 void MaxwellDMA::CopyPitchToBlockLinear() {
@@ -256,7 +257,7 @@ void MaxwellDMA::CopyPitchToBlockLinear() {
                    dst_params.origin.y, x_elements, regs.line_count, block_height, block_depth,
                    regs.pitch_in);
 
-    memory_manager.WriteBlock(regs.offset_out, write_buffer.data(), dst_size);
+    memory_manager.WriteBlockCached(regs.offset_out, write_buffer.data(), dst_size);
 }
 
 void MaxwellDMA::FastCopyBlockLinearToPitch() {
@@ -287,7 +288,7 @@ void MaxwellDMA::FastCopyBlockLinearToPitch() {
                      regs.src_params.block_size.height, regs.src_params.block_size.depth,
                      regs.pitch_out);
 
-    memory_manager.WriteBlock(regs.offset_out, write_buffer.data(), dst_size);
+    memory_manager.WriteBlockCached(regs.offset_out, write_buffer.data(), dst_size);
 }
 
 void MaxwellDMA::CopyBlockLinearToBlockLinear() {
@@ -347,7 +348,7 @@ void MaxwellDMA::CopyBlockLinearToBlockLinear() {
                    dst.depth, dst_x_offset, dst.origin.y, x_elements, regs.line_count,
                    dst.block_size.height, dst.block_size.depth, pitch);
 
-    memory_manager.WriteBlock(regs.offset_out, write_buffer.data(), dst_size);
+    memory_manager.WriteBlockCached(regs.offset_out, write_buffer.data(), dst_size);
 }
 
 void MaxwellDMA::ReleaseSemaphore() {

src/video_core/host_shaders/CMakeLists.txt

@@ -11,7 +11,6 @@ set(GLSL_INCLUDES
 
 set(SHADER_FILES
     astc_decoder.comp
-    blit_color_float.frag
     block_linear_unswizzle_2d.comp
     block_linear_unswizzle_3d.comp
     convert_abgr8_to_d24s8.frag
@@ -37,6 +36,7 @@ set(SHADER_FILES
     smaa_blending_weight_calculation.frag
     smaa_neighborhood_blending.vert
     smaa_neighborhood_blending.frag
+    vulkan_blit_color_float.frag
     vulkan_blit_depth_stencil.frag
     vulkan_fidelityfx_fsr_easu_fp16.comp
     vulkan_fidelityfx_fsr_easu_fp32.comp
@@ -47,6 +47,7 @@ set(SHADER_FILES
     vulkan_present_scaleforce_fp16.frag
     vulkan_present_scaleforce_fp32.frag
     vulkan_quad_indexed.comp
+    vulkan_turbo_mode.comp
     vulkan_uint8.comp
 )
 

src/video_core/host_shaders/full_screen_triangle.vert

@@ -4,20 +4,13 @@
 #version 450
 
 #ifdef VULKAN
-#define VERTEX_ID gl_VertexIndex
 #define BEGIN_PUSH_CONSTANTS layout(push_constant) uniform PushConstants {
 #define END_PUSH_CONSTANTS };
 #define UNIFORM(n)
-#define FLIPY 1
 #else // ^^^ Vulkan ^^^ // vvv OpenGL vvv
-#define VERTEX_ID gl_VertexID
 #define BEGIN_PUSH_CONSTANTS
 #define END_PUSH_CONSTANTS
-#define FLIPY -1
 #define UNIFORM(n) layout (location = n) uniform
-out gl_PerVertex {
-    vec4 gl_Position;
-};
 #endif
 
 BEGIN_PUSH_CONSTANTS
@@ -28,8 +21,8 @@ END_PUSH_CONSTANTS
 layout(location = 0) out vec2 texcoord;
 
 void main() {
-    float x = float((VERTEX_ID & 1) << 2);
-    float y = float((VERTEX_ID & 2) << 1);
-    gl_Position = vec4(x - 1.0, FLIPY * (y - 1.0), 0.0, 1.0);
+    float x = float((gl_VertexIndex & 1) << 2);
+    float y = float((gl_VertexIndex & 2) << 1);
+    gl_Position = vec4(x - 1.0, y - 1.0, 0.0, 1.0);
     texcoord = fma(vec2(x, y) / 2.0, tex_scale, tex_offset);
 }

src/video_core/host_shaders/vulkan_turbo_mode.comp

@@ -0,0 +1,29 @@
+// SPDX-FileCopyrightText: Copyright 2022 yuzu Emulator Project
+// SPDX-License-Identifier: GPL-2.0-or-later
+
+#version 460 core
+
+layout (local_size_x = 16, local_size_y = 8, local_size_z = 1) in;
+
+layout (binding = 0) buffer ThreadData {
+    uint data[];
+};
+
+uint xorshift32(uint x) {
+    x ^= x << 13;
+    x ^= x >> 17;
+    x ^= x << 5;
+    return x;
+}
+
+uint getGlobalIndex() {
+    return gl_GlobalInvocationID.x + gl_GlobalInvocationID.y * gl_WorkGroupSize.y * gl_NumWorkGroups.y;
+}
+
+void main() {
+    uint myIndex = xorshift32(getGlobalIndex());
+    uint otherIndex = xorshift32(myIndex);
+
+    uint otherValue = atomicAdd(data[otherIndex % data.length()], 0) + 1;
+    atomicAdd(data[myIndex % data.length()], otherValue);
+}

src/video_core/invalidation_accumulator.h

@@ -0,0 +1,79 @@
+// SPDX-FileCopyrightText: Copyright 2018 yuzu Emulator Project
+// SPDX-License-Identifier: GPL-2.0-or-later
+
+#pragma once
+
+#include <utility>
+#include <vector>
+
+#include "common/common_types.h"
+
+namespace VideoCommon {
+
+class InvalidationAccumulator {
+public:
+    InvalidationAccumulator() = default;
+    ~InvalidationAccumulator() = default;
+
+    void Add(GPUVAddr address, size_t size) {
+        const auto reset_values = [&]() {
+            if (has_collected) {
+                buffer.emplace_back(start_address, accumulated_size);
+            }
+            start_address = address;
+            accumulated_size = size;
+            last_collection = start_address + size;
+        };
+        if (address >= start_address && address + size <= last_collection) [[likely]] {
+            return;
+        }
+        size = ((address + size + atomicity_size_mask) & atomicity_mask) - address;
+        address = address & atomicity_mask;
+        if (!has_collected) [[unlikely]] {
+            reset_values();
+            has_collected = true;
+            return;
+        }
+        if (address != last_collection) [[unlikely]] {
+            reset_values();
+            return;
+        }
+        accumulated_size += size;
+        last_collection += size;
+    }
+
+    void Clear() {
+        buffer.clear();
+        start_address = 0;
+        last_collection = 0;
+        has_collected = false;
+    }
+
+    bool AnyAccumulated() const {
+        return has_collected;
+    }
+
+    template <typename Func>
+    void Callback(Func&& func) {
+        if (!has_collected) {
+            return;
+        }
+        buffer.emplace_back(start_address, accumulated_size);
+        for (auto& [address, size] : buffer) {
+            func(address, size);
+        }
+    }
+
+private:
+    static constexpr size_t atomicity_bits = 5;
+    static constexpr size_t atomicity_size = 1ULL << atomicity_bits;
+    static constexpr size_t atomicity_size_mask = atomicity_size - 1;
+    static constexpr size_t atomicity_mask = ~atomicity_size_mask;
+    GPUVAddr start_address{};
+    GPUVAddr last_collection{};
+    size_t accumulated_size{};
+    bool has_collected{};
+    std::vector<std::pair<VAddr, size_t>> buffer;
+};
+
+} // namespace VideoCommon

src/video_core/macro/macro_hle.cpp

@@ -50,38 +50,6 @@ protected:
     Maxwell3D& maxwell3d;
 };
 
-class HLE_DrawArrays final : public HLEMacroImpl {
-public:
-    explicit HLE_DrawArrays(Maxwell3D& maxwell3d_) : HLEMacroImpl(maxwell3d_) {}
-
-    void Execute(const std::vector<u32>& parameters, [[maybe_unused]] u32 method) override {
-        maxwell3d.RefreshParameters();
-
-        auto topology = static_cast<Maxwell3D::Regs::PrimitiveTopology>(parameters[0]);
-        maxwell3d.draw_manager->DrawArray(topology, parameters[1], parameters[2],
-                                          maxwell3d.regs.global_base_instance_index, 1);
-    }
-};
-
-class HLE_DrawIndexed final : public HLEMacroImpl {
-public:
-    explicit HLE_DrawIndexed(Maxwell3D& maxwell3d_) : HLEMacroImpl(maxwell3d_) {}
-
-    void Execute(const std::vector<u32>& parameters, [[maybe_unused]] u32 method) override {
-        maxwell3d.RefreshParameters();
-        maxwell3d.regs.index_buffer.start_addr_high = parameters[1];
-        maxwell3d.regs.index_buffer.start_addr_low = parameters[2];
-        maxwell3d.regs.index_buffer.format =
-            static_cast<Engines::Maxwell3D::Regs::IndexFormat>(parameters[3]);
-        maxwell3d.dirty.flags[VideoCommon::Dirty::IndexBuffer] = true;
-
-        auto topology = static_cast<Maxwell3D::Regs::PrimitiveTopology>(parameters[0]);
-        maxwell3d.draw_manager->DrawIndex(topology, 0, parameters[4],
-                                          maxwell3d.regs.global_base_vertex_index,
-                                          maxwell3d.regs.global_base_instance_index, 1);
-    }
-};
-
 /*
  * @note: these macros have two versions, a normal and extended version, with the extended version
  * also assigning the base vertex/instance.
@@ -497,11 +465,6 @@ public:
 } // Anonymous namespace
 
 HLEMacro::HLEMacro(Maxwell3D& maxwell3d_) : maxwell3d{maxwell3d_} {
-    builders.emplace(0xDD6A7FA92A7D2674ULL,
-                     std::function<std::unique_ptr<CachedMacro>(Maxwell3D&)>(
-                         [](Maxwell3D& maxwell3d__) -> std::unique_ptr<CachedMacro> {
-                             return std::make_unique<HLE_DrawArrays>(maxwell3d__);
-                         }));
     builders.emplace(0x0D61FC9FAAC9FCADULL,
                      std::function<std::unique_ptr<CachedMacro>(Maxwell3D&)>(
                          [](Maxwell3D& maxwell3d__) -> std::unique_ptr<CachedMacro> {
@@ -512,11 +475,6 @@ HLEMacro::HLEMacro(Maxwell3D& maxwell3d_) : maxwell3d{maxwell3d_} {
                          [](Maxwell3D& maxwell3d__) -> std::unique_ptr<CachedMacro> {
                              return std::make_unique<HLE_DrawArraysIndirect<true>>(maxwell3d__);
                          }));
-    builders.emplace(0x2DB33AADB741839CULL,
-                     std::function<std::unique_ptr<CachedMacro>(Maxwell3D&)>(
-                         [](Maxwell3D& maxwell3d__) -> std::unique_ptr<CachedMacro> {
-                             return std::make_unique<HLE_DrawIndexed>(maxwell3d__);
-                         }));
     builders.emplace(0x771BB18C62444DA0ULL,
                      std::function<std::unique_ptr<CachedMacro>(Maxwell3D&)>(
                          [](Maxwell3D& maxwell3d__) -> std::unique_ptr<CachedMacro> {

src/video_core/memory_manager.cpp

@@ -6,11 +6,13 @@
 #include "common/alignment.h"
 #include "common/assert.h"
 #include "common/logging/log.h"
+#include "common/settings.h"
 #include "core/core.h"
 #include "core/device_memory.h"
 #include "core/hle/kernel/k_page_table.h"
 #include "core/hle/kernel/k_process.h"
 #include "core/memory.h"
+#include "video_core/invalidation_accumulator.h"
 #include "video_core/memory_manager.h"
 #include "video_core/rasterizer_interface.h"
 #include "video_core/renderer_base.h"
@@ -26,7 +28,8 @@ MemoryManager::MemoryManager(Core::System& system_, u64 address_space_bits_, u64
       entries{}, big_entries{}, page_table{address_space_bits, address_space_bits + page_bits - 38,
                                            page_bits != big_page_bits ? page_bits : 0},
       kind_map{PTEKind::INVALID}, unique_identifier{unique_identifier_generator.fetch_add(
-                                      1, std::memory_order_acq_rel)} {
+                                      1, std::memory_order_acq_rel)},
+      accumulator{std::make_unique<VideoCommon::InvalidationAccumulator>()} {
     address_space_size = 1ULL << address_space_bits;
     page_size = 1ULL << page_bits;
     page_mask = page_size - 1ULL;
@@ -43,6 +46,11 @@ MemoryManager::MemoryManager(Core::System& system_, u64 address_space_bits_, u64
     big_page_table_cpu.resize(big_page_table_size);
     big_page_continous.resize(big_page_table_size / continous_bits, 0);
     entries.resize(page_table_size / 32, 0);
+    if (!Settings::IsGPULevelExtreme() && Settings::IsFastmemEnabled()) {
+        fastmem_arena = system.DeviceMemory().buffer.VirtualBasePointer();
+    } else {
+        fastmem_arena = nullptr;
+    }
 }
 
 MemoryManager::~MemoryManager() = default;
@@ -185,15 +193,12 @@ void MemoryManager::Unmap(GPUVAddr gpu_addr, std::size_t size) {
     if (size == 0) {
         return;
     }
-    const auto submapped_ranges = GetSubmappedRange(gpu_addr, size);
+    GetSubmappedRangeImpl<false>(gpu_addr, size, page_stash);
 
-    for (const auto& [map_addr, map_size] : submapped_ranges) {
-        // Flush and invalidate through the GPU interface, to be asynchronous if possible.
-        const std::optional<VAddr> cpu_addr = GpuToCpuAddress(map_addr);
-        ASSERT(cpu_addr);
-
-        rasterizer->UnmapMemory(*cpu_addr, map_size);
+    for (const auto& [map_addr, map_size] : page_stash) {
+        rasterizer->UnmapMemory(map_addr, map_size);
     }
+    page_stash.clear();
 
     BigPageTableOp<EntryType::Free>(gpu_addr, 0, size, PTEKind::INVALID);
     PageTableOp<EntryType::Free>(gpu_addr, 0, size, PTEKind::INVALID);
@@ -355,7 +360,7 @@ inline void MemoryManager::MemoryOperation(GPUVAddr gpu_src_addr, std::size_t si
     }
 }
 
-template <bool is_safe>
+template <bool is_safe, bool use_fastmem>
 void MemoryManager::ReadBlockImpl(GPUVAddr gpu_src_addr, void* dest_buffer, std::size_t size,
                                   [[maybe_unused]] VideoCommon::CacheType which) const {
     auto set_to_zero = [&]([[maybe_unused]] std::size_t page_index,
@@ -369,8 +374,12 @@ void MemoryManager::ReadBlockImpl(GPUVAddr gpu_src_addr, void* dest_buffer, std:
         if constexpr (is_safe) {
             rasterizer->FlushRegion(cpu_addr_base, copy_amount, which);
         }
-        u8* physical = memory.GetPointer(cpu_addr_base);
-        std::memcpy(dest_buffer, physical, copy_amount);
+        if constexpr (use_fastmem) {
+            std::memcpy(dest_buffer, &fastmem_arena[cpu_addr_base], copy_amount);
+        } else {
+            u8* physical = memory.GetPointer(cpu_addr_base);
+            std::memcpy(dest_buffer, physical, copy_amount);
+        }
         dest_buffer = static_cast<u8*>(dest_buffer) + copy_amount;
     };
     auto mapped_big = [&](std::size_t page_index, std::size_t offset, std::size_t copy_amount) {
@@ -379,11 +388,15 @@ void MemoryManager::ReadBlockImpl(GPUVAddr gpu_src_addr, void* dest_buffer, std:
         if constexpr (is_safe) {
             rasterizer->FlushRegion(cpu_addr_base, copy_amount, which);
         }
-        if (!IsBigPageContinous(page_index)) [[unlikely]] {
-            memory.ReadBlockUnsafe(cpu_addr_base, dest_buffer, copy_amount);
+        if constexpr (use_fastmem) {
+            std::memcpy(dest_buffer, &fastmem_arena[cpu_addr_base], copy_amount);
         } else {
-            u8* physical = memory.GetPointer(cpu_addr_base);
-            std::memcpy(dest_buffer, physical, copy_amount);
+            if (!IsBigPageContinous(page_index)) [[unlikely]] {
+                memory.ReadBlockUnsafe(cpu_addr_base, dest_buffer, copy_amount);
+            } else {
+                u8* physical = memory.GetPointer(cpu_addr_base);
+                std::memcpy(dest_buffer, physical, copy_amount);
+            }
         }
         dest_buffer = static_cast<u8*>(dest_buffer) + copy_amount;
     };
@@ -397,12 +410,20 @@ void MemoryManager::ReadBlockImpl(GPUVAddr gpu_src_addr, void* dest_buffer, std:
 
 void MemoryManager::ReadBlock(GPUVAddr gpu_src_addr, void* dest_buffer, std::size_t size,
                               VideoCommon::CacheType which) const {
-    ReadBlockImpl<true>(gpu_src_addr, dest_buffer, size, which);
+    if (fastmem_arena) [[likely]] {
+        ReadBlockImpl<true, true>(gpu_src_addr, dest_buffer, size, which);
+        return;
+    }
+    ReadBlockImpl<true, false>(gpu_src_addr, dest_buffer, size, which);
 }
 
 void MemoryManager::ReadBlockUnsafe(GPUVAddr gpu_src_addr, void* dest_buffer,
                                     const std::size_t size) const {
-    ReadBlockImpl<false>(gpu_src_addr, dest_buffer, size, VideoCommon::CacheType::None);
+    if (fastmem_arena) [[likely]] {
+        ReadBlockImpl<false, true>(gpu_src_addr, dest_buffer, size, VideoCommon::CacheType::None);
+        return;
+    }
+    ReadBlockImpl<false, false>(gpu_src_addr, dest_buffer, size, VideoCommon::CacheType::None);
 }
 
 template <bool is_safe>
@@ -454,6 +475,12 @@ void MemoryManager::WriteBlockUnsafe(GPUVAddr gpu_dest_addr, const void* src_buf
     WriteBlockImpl<false>(gpu_dest_addr, src_buffer, size, VideoCommon::CacheType::None);
 }
 
+void MemoryManager::WriteBlockCached(GPUVAddr gpu_dest_addr, const void* src_buffer,
+                                     std::size_t size) {
+    WriteBlockImpl<false>(gpu_dest_addr, src_buffer, size, VideoCommon::CacheType::None);
+    accumulator->Add(gpu_dest_addr, size);
+}
+
 void MemoryManager::FlushRegion(GPUVAddr gpu_addr, size_t size,
                                 VideoCommon::CacheType which) const {
     auto do_nothing = [&]([[maybe_unused]] std::size_t page_index,
@@ -663,7 +690,17 @@ bool MemoryManager::IsFullyMappedRange(GPUVAddr gpu_addr, std::size_t size) cons
 std::vector<std::pair<GPUVAddr, std::size_t>> MemoryManager::GetSubmappedRange(
     GPUVAddr gpu_addr, std::size_t size) const {
     std::vector<std::pair<GPUVAddr, std::size_t>> result{};
-    std::optional<std::pair<GPUVAddr, std::size_t>> last_segment{};
+    GetSubmappedRangeImpl<true>(gpu_addr, size, result);
+    return result;
+}
+
+template <bool is_gpu_address>
+void MemoryManager::GetSubmappedRangeImpl(
+    GPUVAddr gpu_addr, std::size_t size,
+    std::vector<std::pair<std::conditional_t<is_gpu_address, GPUVAddr, VAddr>, std::size_t>>&
+        result) const {
+    std::optional<std::pair<std::conditional_t<is_gpu_address, GPUVAddr, VAddr>, std::size_t>>
+        last_segment{};
     std::optional<VAddr> old_page_addr{};
     const auto split = [&last_segment, &result]([[maybe_unused]] std::size_t page_index,
                                                 [[maybe_unused]] std::size_t offset,
@@ -685,8 +722,12 @@ std::vector<std::pair<GPUVAddr, std::size_t>> MemoryManager::GetSubmappedRange(
         }
         old_page_addr = {cpu_addr_base + copy_amount};
         if (!last_segment) {
-            const GPUVAddr new_base_addr = (page_index << big_page_bits) + offset;
-            last_segment = {new_base_addr, copy_amount};
+            if constexpr (is_gpu_address) {
+                const GPUVAddr new_base_addr = (page_index << big_page_bits) + offset;
+                last_segment = {new_base_addr, copy_amount};
+            } else {
+                last_segment = {cpu_addr_base, copy_amount};
+            }
         } else {
             last_segment->second += copy_amount;
         }
@@ -703,8 +744,12 @@ std::vector<std::pair<GPUVAddr, std::size_t>> MemoryManager::GetSubmappedRange(
         }
         old_page_addr = {cpu_addr_base + copy_amount};
         if (!last_segment) {
-            const GPUVAddr new_base_addr = (page_index << page_bits) + offset;
-            last_segment = {new_base_addr, copy_amount};
+            if constexpr (is_gpu_address) {
+                const GPUVAddr new_base_addr = (page_index << page_bits) + offset;
+                last_segment = {new_base_addr, copy_amount};
+            } else {
+                last_segment = {cpu_addr_base, copy_amount};
+            }
         } else {
             last_segment->second += copy_amount;
         }
@@ -715,7 +760,18 @@ std::vector<std::pair<GPUVAddr, std::size_t>> MemoryManager::GetSubmappedRange(
     };
     MemoryOperation<true>(gpu_addr, size, extend_size_big, split, do_short_pages);
     split(0, 0, 0);
-    return result;
+}
+
+void MemoryManager::FlushCaching() {
+    if (!accumulator->AnyAccumulated()) {
+        return;
+    }
+    accumulator->Callback([this](GPUVAddr addr, size_t size) {
+        GetSubmappedRangeImpl<false>(addr, size, page_stash);
+    });
+    rasterizer->InnerInvalidation(page_stash);
+    page_stash.clear();
+    accumulator->Clear();
 }
 
 } // namespace Tegra

src/video_core/memory_manager.h

@@ -19,6 +19,10 @@ namespace VideoCore {
 class RasterizerInterface;
 }
 
+namespace VideoCommon {
+class InvalidationAccumulator;
+}
+
 namespace Core {
 class DeviceMemory;
 namespace Memory {
@@ -80,6 +84,7 @@ public:
      */
     void ReadBlockUnsafe(GPUVAddr gpu_src_addr, void* dest_buffer, std::size_t size) const;
     void WriteBlockUnsafe(GPUVAddr gpu_dest_addr, const void* src_buffer, std::size_t size);
+    void WriteBlockCached(GPUVAddr gpu_dest_addr, const void* src_buffer, std::size_t size);
 
     /**
      * Checks if a gpu region can be simply read with a pointer.
@@ -129,12 +134,14 @@ public:
     size_t GetMemoryLayoutSize(GPUVAddr gpu_addr,
                                size_t max_size = std::numeric_limits<size_t>::max()) const;
 
+    void FlushCaching();
+
 private:
     template <bool is_big_pages, typename FuncMapped, typename FuncReserved, typename FuncUnmapped>
     inline void MemoryOperation(GPUVAddr gpu_src_addr, std::size_t size, FuncMapped&& func_mapped,
                                 FuncReserved&& func_reserved, FuncUnmapped&& func_unmapped) const;
 
-    template <bool is_safe>
+    template <bool is_safe, bool use_fastmem>
     void ReadBlockImpl(GPUVAddr gpu_src_addr, void* dest_buffer, std::size_t size,
                        VideoCommon::CacheType which) const;
 
@@ -154,6 +161,12 @@ private:
     inline bool IsBigPageContinous(size_t big_page_index) const;
     inline void SetBigPageContinous(size_t big_page_index, bool value);
 
+    template <bool is_gpu_address>
+    void GetSubmappedRangeImpl(
+        GPUVAddr gpu_addr, std::size_t size,
+        std::vector<std::pair<std::conditional_t<is_gpu_address, GPUVAddr, VAddr>, std::size_t>>&
+            result) const;
+
     Core::System& system;
     Core::Memory::Memory& memory;
     Core::DeviceMemory& device_memory;
@@ -201,10 +214,13 @@ private:
     Common::VirtualBuffer<u32> big_page_table_cpu;
 
     std::vector<u64> big_page_continous;
+    std::vector<std::pair<VAddr, std::size_t>> page_stash{};
+    u8* fastmem_arena{};
 
     constexpr static size_t continous_bits = 64;
 
     const size_t unique_identifier;
+    std::unique_ptr<VideoCommon::InvalidationAccumulator> accumulator;
 
     static std::atomic<size_t> unique_identifier_generator;
 };

src/video_core/rasterizer_interface.h

@@ -6,6 +6,7 @@
 #include <functional>
 #include <optional>
 #include <span>
+#include <utility>
 #include "common/common_types.h"
 #include "common/polyfill_thread.h"
 #include "video_core/cache_types.h"
@@ -46,9 +47,6 @@ public:
     /// Dispatches an indirect draw invocation
     virtual void DrawIndirect() {}
 
-    /// Dispatches an draw texture invocation
-    virtual void DrawTexture() = 0;
-
     /// Clear the current framebuffer
     virtual void Clear(u32 layer_count) = 0;
 
@@ -98,6 +96,12 @@ public:
     virtual void InvalidateRegion(VAddr addr, u64 size,
                                   VideoCommon::CacheType which = VideoCommon::CacheType::All) = 0;
 
+    virtual void InnerInvalidation(std::span<const std::pair<VAddr, std::size_t>> sequences) {
+        for (const auto& [cpu_addr, size] : sequences) {
+            InvalidateRegion(cpu_addr, size);
+        }
+    }
+
     /// Notify rasterizer that any caches of the specified region are desync with guest
     virtual void OnCPUWrite(VAddr addr, u64 size) = 0;
 

src/video_core/renderer_null/null_rasterizer.cpp

@@ -21,7 +21,6 @@ RasterizerNull::RasterizerNull(Core::Memory::Memory& cpu_memory_, Tegra::GPU& gp
 RasterizerNull::~RasterizerNull() = default;
 
 void RasterizerNull::Draw(bool is_indexed, u32 instance_count) {}
-void RasterizerNull::DrawTexture() {}
 void RasterizerNull::Clear(u32 layer_count) {}
 void RasterizerNull::DispatchCompute() {}
 void RasterizerNull::ResetCounter(VideoCore::QueryType type) {}

src/video_core/renderer_null/null_rasterizer.h

@@ -31,7 +31,6 @@ public:
     ~RasterizerNull() override;
 
     void Draw(bool is_indexed, u32 instance_count) override;
-    void DrawTexture() override;
     void Clear(u32 layer_count) override;
     void DispatchCompute() override;
     void ResetCounter(VideoCore::QueryType type) override;

src/video_core/renderer_opengl/blit_image.cpp

@@ -1,59 +0,0 @@
-// SPDX-FileCopyrightText: Copyright 2023 yuzu Emulator Project
-// SPDX-License-Identifier: GPL-2.0-or-later
-
-#include <algorithm>
-
-#include "video_core/host_shaders/blit_color_float_frag.h"
-#include "video_core/host_shaders/full_screen_triangle_vert.h"
-#include "video_core/renderer_opengl/blit_image.h"
-#include "video_core/renderer_opengl/gl_shader_manager.h"
-#include "video_core/renderer_opengl/gl_shader_util.h"
-
-namespace OpenGL {
-
-BlitImageHelper::BlitImageHelper(ProgramManager& program_manager_)
-    : program_manager(program_manager_),
-      full_screen_vert(CreateProgram(HostShaders::FULL_SCREEN_TRIANGLE_VERT, GL_VERTEX_SHADER)),
-      blit_color_to_color_frag(
-          CreateProgram(HostShaders::BLIT_COLOR_FLOAT_FRAG, GL_FRAGMENT_SHADER)) {}
-
-BlitImageHelper::~BlitImageHelper() = default;
-
-void BlitImageHelper::BlitColor(GLuint dst_framebuffer, GLuint src_image_view, GLuint src_sampler,
-                                const Region2D& dst_region, const Region2D& src_region,
-                                const Extent3D& src_size) {
-    glEnable(GL_CULL_FACE);
-    glDisable(GL_COLOR_LOGIC_OP);
-    glDisable(GL_DEPTH_TEST);
-    glDisable(GL_STENCIL_TEST);
-    glDisable(GL_POLYGON_OFFSET_FILL);
-    glDisable(GL_RASTERIZER_DISCARD);
-    glDisable(GL_ALPHA_TEST);
-    glDisablei(GL_BLEND, 0);
-    glPolygonMode(GL_FRONT_AND_BACK, GL_FILL);
-    glCullFace(GL_BACK);
-    glFrontFace(GL_CW);
-    glColorMaski(0, GL_TRUE, GL_TRUE, GL_TRUE, GL_TRUE);
-    glDepthRangeIndexed(0, 0.0, 0.0);
-
-    program_manager.BindPresentPrograms(full_screen_vert.handle, blit_color_to_color_frag.handle);
-    glProgramUniform2f(full_screen_vert.handle, 0,
-                       static_cast<float>(src_region.end.x - src_region.start.x) /
-                           static_cast<float>(src_size.width),
-                       static_cast<float>(src_region.end.y - src_region.start.y) /
-                           static_cast<float>(src_size.height));
-    glProgramUniform2f(full_screen_vert.handle, 1,
-                       static_cast<float>(src_region.start.x) / static_cast<float>(src_size.width),
-                       static_cast<float>(src_region.start.y) /
-                           static_cast<float>(src_size.height));
-    glViewport(std::min(dst_region.start.x, dst_region.end.x),
-               std::min(dst_region.start.y, dst_region.end.y),
-               std::abs(dst_region.end.x - dst_region.start.x),
-               std::abs(dst_region.end.y - dst_region.start.y));
-    glBindFramebuffer(GL_DRAW_FRAMEBUFFER, dst_framebuffer);
-    glBindSampler(0, src_sampler);
-    glBindTextureUnit(0, src_image_view);
-    glClear(GL_COLOR_BUFFER_BIT);
-    glDrawArrays(GL_TRIANGLES, 0, 3);
-}
-} // namespace OpenGL

src/video_core/renderer_opengl/blit_image.h

@@ -1,38 +0,0 @@
-// SPDX-FileCopyrightText: Copyright 2023 yuzu Emulator Project
-// SPDX-License-Identifier: GPL-2.0-or-later
-
-#pragma once
-
-#include <glad/glad.h>
-
-#include "video_core/engines/fermi_2d.h"
-#include "video_core/renderer_opengl/gl_resource_manager.h"
-#include "video_core/texture_cache/types.h"
-
-namespace OpenGL {
-
-using VideoCommon::Extent3D;
-using VideoCommon::Offset2D;
-using VideoCommon::Region2D;
-
-class ProgramManager;
-class Framebuffer;
-class ImageView;
-
-class BlitImageHelper {
-public:
-    explicit BlitImageHelper(ProgramManager& program_manager);
-    ~BlitImageHelper();
-
-    void BlitColor(GLuint dst_framebuffer, GLuint src_image_view, GLuint src_sampler,
-                   const Region2D& dst_region, const Region2D& src_region,
-                   const Extent3D& src_size);
-
-private:
-    ProgramManager& program_manager;
-
-    OGLProgram full_screen_vert;
-    OGLProgram blit_color_to_color_frag;
-};
-
-} // namespace OpenGL

src/video_core/renderer_opengl/gl_buffer_cache.h

@@ -160,6 +160,10 @@ public:
         return device.CanReportMemoryUsage();
     }
 
+    u32 GetStorageBufferAlignment() const {
+        return static_cast<u32>(device.GetShaderStorageBufferAlignment());
+    }
+
 private:
     static constexpr std::array PABO_LUT{
         GL_VERTEX_PROGRAM_PARAMETER_BUFFER_NV,          GL_TESS_CONTROL_PROGRAM_PARAMETER_BUFFER_NV,

src/video_core/renderer_opengl/gl_device.cpp

@@ -166,7 +166,6 @@ Device::Device(Core::Frontend::EmuWindow& emu_window) {
     has_shader_int64 = HasExtension(extensions, "GL_ARB_gpu_shader_int64");
     has_amd_shader_half_float = GLAD_GL_AMD_gpu_shader_half_float;
     has_sparse_texture_2 = GLAD_GL_ARB_sparse_texture2;
-    has_draw_texture = GLAD_GL_NV_draw_texture;
     warp_size_potentially_larger_than_guest = !is_nvidia && !is_intel;
     need_fastmath_off = is_nvidia;
     can_report_memory = GLAD_GL_NVX_gpu_memory_info;

src/video_core/renderer_opengl/gl_device.h

@@ -4,8 +4,6 @@
 #pragma once
 
 #include <cstddef>
-#include <string>
-
 #include "common/common_types.h"
 #include "core/frontend/emu_window.h"
 #include "shader_recompiler/stage.h"
@@ -148,10 +146,6 @@ public:
         return has_sparse_texture_2;
     }
 
-    bool HasDrawTexture() const {
-        return has_draw_texture;
-    }
-
     bool IsWarpSizePotentiallyLargerThanGuest() const {
         return warp_size_potentially_larger_than_guest;
     }
@@ -222,7 +216,6 @@ private:
     bool has_shader_int64{};
     bool has_amd_shader_half_float{};
     bool has_sparse_texture_2{};
-    bool has_draw_texture{};
     bool warp_size_potentially_larger_than_guest{};
     bool need_fastmath_off{};
     bool has_cbuf_ftou_bug{};

src/video_core/renderer_opengl/gl_rasterizer.cpp

@@ -64,8 +64,7 @@ RasterizerOpenGL::RasterizerOpenGL(Core::Frontend::EmuWindow& emu_window_, Tegra
       shader_cache(*this, emu_window_, device, texture_cache, buffer_cache, program_manager,
                    state_tracker, gpu.ShaderNotify()),
       query_cache(*this), accelerate_dma(buffer_cache),
-      fence_manager(*this, gpu, texture_cache, buffer_cache, query_cache),
-      blit_image(program_manager_) {}
+      fence_manager(*this, gpu, texture_cache, buffer_cache, query_cache) {}
 
 RasterizerOpenGL::~RasterizerOpenGL() = default;
 
@@ -319,47 +318,6 @@ void RasterizerOpenGL::DrawIndirect() {
     buffer_cache.SetDrawIndirect(nullptr);
 }
 
-void RasterizerOpenGL::DrawTexture() {
-    MICROPROFILE_SCOPE(OpenGL_Drawing);
-
-    SCOPE_EXIT({ gpu.TickWork(); });
-    query_cache.UpdateCounters();
-
-    texture_cache.SynchronizeGraphicsDescriptors();
-    texture_cache.UpdateRenderTargets(false);
-
-    SyncState();
-
-    const auto& draw_texture_state = maxwell3d->draw_manager->GetDrawTextureState();
-    const auto& sampler = texture_cache.GetGraphicsSampler(draw_texture_state.src_sampler);
-    const auto& texture = texture_cache.GetImageView(draw_texture_state.src_texture);
-
-    if (device.HasDrawTexture()) {
-        state_tracker.BindFramebuffer(texture_cache.GetFramebuffer()->Handle());
-
-        glDrawTextureNV(texture.DefaultHandle(), sampler->Handle(), draw_texture_state.dst_x0,
-                        draw_texture_state.dst_y0, draw_texture_state.dst_x1,
-                        draw_texture_state.dst_y1, 0,
-                        draw_texture_state.src_x0 / static_cast<float>(texture.size.width),
-                        draw_texture_state.src_y0 / static_cast<float>(texture.size.height),
-                        draw_texture_state.src_x1 / static_cast<float>(texture.size.width),
-                        draw_texture_state.src_y1 / static_cast<float>(texture.size.height));
-    } else {
-        Region2D dst_region = {Offset2D{.x = static_cast<s32>(draw_texture_state.dst_x0),
-                                        .y = static_cast<s32>(draw_texture_state.dst_y0)},
-                               Offset2D{.x = static_cast<s32>(draw_texture_state.dst_x1),
-                                        .y = static_cast<s32>(draw_texture_state.dst_y1)}};
-        Region2D src_region = {Offset2D{.x = static_cast<s32>(draw_texture_state.src_x0),
-                                        .y = static_cast<s32>(draw_texture_state.src_y0)},
-                               Offset2D{.x = static_cast<s32>(draw_texture_state.src_x1),
-                                        .y = static_cast<s32>(draw_texture_state.src_y1)}};
-        blit_image.BlitColor(texture_cache.GetFramebuffer()->Handle(), texture.DefaultHandle(),
-                             sampler->Handle(), dst_region, src_region, texture.size);
-    }
-
-    ++num_queued_commands;
-}
-
 void RasterizerOpenGL::DispatchCompute() {
     ComputePipeline* const pipeline{shader_cache.CurrentComputePipeline()};
     if (!pipeline) {

src/video_core/renderer_opengl/gl_rasterizer.h

@@ -16,7 +16,6 @@
 #include "video_core/engines/maxwell_dma.h"
 #include "video_core/rasterizer_accelerated.h"
 #include "video_core/rasterizer_interface.h"
-#include "video_core/renderer_opengl/blit_image.h"
 #include "video_core/renderer_opengl/gl_buffer_cache.h"
 #include "video_core/renderer_opengl/gl_device.h"
 #include "video_core/renderer_opengl/gl_fence_manager.h"
@@ -71,7 +70,6 @@ public:
 
     void Draw(bool is_indexed, u32 instance_count) override;
     void DrawIndirect() override;
-    void DrawTexture() override;
     void Clear(u32 layer_count) override;
     void DispatchCompute() override;
     void ResetCounter(VideoCore::QueryType type) override;
@@ -226,8 +224,6 @@ private:
     AccelerateDMA accelerate_dma;
     FenceManagerOpenGL fence_manager;
 
-    BlitImageHelper blit_image;
-
     boost::container::static_vector<u32, MAX_IMAGE_VIEWS> image_view_indices;
     std::array<ImageViewId, MAX_IMAGE_VIEWS> image_view_ids;
     boost::container::static_vector<GLuint, MAX_TEXTURES> sampler_handles;

src/video_core/renderer_opengl/gl_shader_cache.cpp

@@ -236,6 +236,8 @@ ShaderCache::ShaderCache(RasterizerOpenGL& rasterizer_, Core::Frontend::EmuWindo
           .needs_demote_reorder = device.IsAmd(),
           .support_snorm_render_buffer = false,
           .support_viewport_index_layer = device.HasVertexViewportLayer(),
+          .min_ssbo_alignment = static_cast<u32>(device.GetShaderStorageBufferAlignment()),
+          .support_geometry_shader_passthrough = device.HasGeometryShaderPassthrough(),
       } {
     if (use_asynchronous_shaders) {
         workers = CreateWorkers();

src/video_core/renderer_opengl/gl_shader_manager.cpp

@@ -1,123 +1,2 @@
 // SPDX-FileCopyrightText: Copyright 2018 yuzu Emulator Project
 // SPDX-License-Identifier: GPL-2.0-or-later
-
-#include <glad/glad.h>
-
-#include "video_core/renderer_opengl/gl_shader_manager.h"
-
-namespace OpenGL {
-
-static constexpr std::array ASSEMBLY_PROGRAM_ENUMS{
-    GL_VERTEX_PROGRAM_NV,   GL_TESS_CONTROL_PROGRAM_NV, GL_TESS_EVALUATION_PROGRAM_NV,
-    GL_GEOMETRY_PROGRAM_NV, GL_FRAGMENT_PROGRAM_NV,
-};
-
-ProgramManager::ProgramManager(const Device& device) {
-    glCreateProgramPipelines(1, &pipeline.handle);
-    if (device.UseAssemblyShaders()) {
-        glEnable(GL_COMPUTE_PROGRAM_NV);
-    }
-}
-
-void ProgramManager::BindComputeProgram(GLuint program) {
-    glUseProgram(program);
-    is_compute_bound = true;
-}
-
-void ProgramManager::BindComputeAssemblyProgram(GLuint program) {
-    if (current_assembly_compute_program != program) {
-        current_assembly_compute_program = program;
-        glBindProgramARB(GL_COMPUTE_PROGRAM_NV, program);
-    }
-    UnbindPipeline();
-}
-
-void ProgramManager::BindSourcePrograms(std::span<const OGLProgram, NUM_STAGES> programs) {
-    static constexpr std::array<GLenum, 5> stage_enums{
-        GL_VERTEX_SHADER_BIT,   GL_TESS_CONTROL_SHADER_BIT, GL_TESS_EVALUATION_SHADER_BIT,
-        GL_GEOMETRY_SHADER_BIT, GL_FRAGMENT_SHADER_BIT,
-    };
-    for (size_t stage = 0; stage < NUM_STAGES; ++stage) {
-        if (current_programs[stage] != programs[stage].handle) {
-            current_programs[stage] = programs[stage].handle;
-            glUseProgramStages(pipeline.handle, stage_enums[stage], programs[stage].handle);
-        }
-    }
-    BindPipeline();
-}
-
-void ProgramManager::BindPresentPrograms(GLuint vertex, GLuint fragment) {
-    if (current_programs[0] != vertex) {
-        current_programs[0] = vertex;
-        glUseProgramStages(pipeline.handle, GL_VERTEX_SHADER_BIT, vertex);
-    }
-    if (current_programs[4] != fragment) {
-        current_programs[4] = fragment;
-        glUseProgramStages(pipeline.handle, GL_FRAGMENT_SHADER_BIT, fragment);
-    }
-    glUseProgramStages(
-        pipeline.handle,
-        GL_TESS_CONTROL_SHADER_BIT | GL_TESS_EVALUATION_SHADER_BIT | GL_GEOMETRY_SHADER_BIT, 0);
-    current_programs[1] = 0;
-    current_programs[2] = 0;
-    current_programs[3] = 0;
-
-    if (current_stage_mask != 0) {
-        current_stage_mask = 0;
-        for (const GLenum program_type : ASSEMBLY_PROGRAM_ENUMS) {
-            glDisable(program_type);
-        }
-    }
-    BindPipeline();
-}
-
-void ProgramManager::BindAssemblyPrograms(std::span<const OGLAssemblyProgram, NUM_STAGES> programs,
-                                          u32 stage_mask) {
-    const u32 changed_mask = current_stage_mask ^ stage_mask;
-    current_stage_mask = stage_mask;
-
-    if (changed_mask != 0) {
-        for (size_t stage = 0; stage < NUM_STAGES; ++stage) {
-            if (((changed_mask >> stage) & 1) != 0) {
-                if (((stage_mask >> stage) & 1) != 0) {
-                    glEnable(ASSEMBLY_PROGRAM_ENUMS[stage]);
-                } else {
-                    glDisable(ASSEMBLY_PROGRAM_ENUMS[stage]);
-                }
-            }
-        }
-    }
-    for (size_t stage = 0; stage < NUM_STAGES; ++stage) {
-        if (current_programs[stage] != programs[stage].handle) {
-            current_programs[stage] = programs[stage].handle;
-            glBindProgramARB(ASSEMBLY_PROGRAM_ENUMS[stage], programs[stage].handle);
-        }
-    }
-    UnbindPipeline();
-}
-
-void ProgramManager::RestoreGuestCompute() {}
-
-void ProgramManager::BindPipeline() {
-    if (!is_pipeline_bound) {
-        is_pipeline_bound = true;
-        glBindProgramPipeline(pipeline.handle);
-    }
-    UnbindCompute();
-}
-
-void ProgramManager::UnbindPipeline() {
-    if (is_pipeline_bound) {
-        is_pipeline_bound = false;
-        glBindProgramPipeline(0);
-    }
-    UnbindCompute();
-}
-
-void ProgramManager::UnbindCompute() {
-    if (is_compute_bound) {
-        is_compute_bound = false;
-        glUseProgram(0);
-    }
-}
-} // namespace OpenGL

src/video_core/renderer_opengl/gl_shader_manager.h

@@ -6,6 +6,8 @@
 #include <array>
 #include <span>
 
+#include <glad/glad.h>
+
 #include "video_core/renderer_opengl/gl_device.h"
 #include "video_core/renderer_opengl/gl_resource_manager.h"
 
@@ -14,28 +16,121 @@ namespace OpenGL {
 class ProgramManager {
     static constexpr size_t NUM_STAGES = 5;
 
+    static constexpr std::array ASSEMBLY_PROGRAM_ENUMS{
+        GL_VERTEX_PROGRAM_NV,   GL_TESS_CONTROL_PROGRAM_NV, GL_TESS_EVALUATION_PROGRAM_NV,
+        GL_GEOMETRY_PROGRAM_NV, GL_FRAGMENT_PROGRAM_NV,
+    };
+
 public:
-    explicit ProgramManager(const Device& device);
+    explicit ProgramManager(const Device& device) {
+        glCreateProgramPipelines(1, &pipeline.handle);
+        if (device.UseAssemblyShaders()) {
+            glEnable(GL_COMPUTE_PROGRAM_NV);
+        }
+    }
 
-    void BindComputeProgram(GLuint program);
+    void BindComputeProgram(GLuint program) {
+        glUseProgram(program);
+        is_compute_bound = true;
+    }
 
-    void BindComputeAssemblyProgram(GLuint program);
+    void BindComputeAssemblyProgram(GLuint program) {
+        if (current_assembly_compute_program != program) {
+            current_assembly_compute_program = program;
+            glBindProgramARB(GL_COMPUTE_PROGRAM_NV, program);
+        }
+        UnbindPipeline();
+    }
 
-    void BindSourcePrograms(std::span<const OGLProgram, NUM_STAGES> programs);
+    void BindSourcePrograms(std::span<const OGLProgram, NUM_STAGES> programs) {
+        static constexpr std::array<GLenum, 5> stage_enums{
+            GL_VERTEX_SHADER_BIT,   GL_TESS_CONTROL_SHADER_BIT, GL_TESS_EVALUATION_SHADER_BIT,
+            GL_GEOMETRY_SHADER_BIT, GL_FRAGMENT_SHADER_BIT,
+        };
+        for (size_t stage = 0; stage < NUM_STAGES; ++stage) {
+            if (current_programs[stage] != programs[stage].handle) {
+                current_programs[stage] = programs[stage].handle;
+                glUseProgramStages(pipeline.handle, stage_enums[stage], programs[stage].handle);
+            }
+        }
+        BindPipeline();
+    }
 
-    void BindPresentPrograms(GLuint vertex, GLuint fragment);
+    void BindPresentPrograms(GLuint vertex, GLuint fragment) {
+        if (current_programs[0] != vertex) {
+            current_programs[0] = vertex;
+            glUseProgramStages(pipeline.handle, GL_VERTEX_SHADER_BIT, vertex);
+        }
+        if (current_programs[4] != fragment) {
+            current_programs[4] = fragment;
+            glUseProgramStages(pipeline.handle, GL_FRAGMENT_SHADER_BIT, fragment);
+        }
+        glUseProgramStages(
+            pipeline.handle,
+            GL_TESS_CONTROL_SHADER_BIT | GL_TESS_EVALUATION_SHADER_BIT | GL_GEOMETRY_SHADER_BIT, 0);
+        current_programs[1] = 0;
+        current_programs[2] = 0;
+        current_programs[3] = 0;
+
+        if (current_stage_mask != 0) {
+            current_stage_mask = 0;
+            for (const GLenum program_type : ASSEMBLY_PROGRAM_ENUMS) {
+                glDisable(program_type);
+            }
+        }
+        BindPipeline();
+    }
 
     void BindAssemblyPrograms(std::span<const OGLAssemblyProgram, NUM_STAGES> programs,
-                              u32 stage_mask);
+                              u32 stage_mask) {
+        const u32 changed_mask = current_stage_mask ^ stage_mask;
+        current_stage_mask = stage_mask;
 
-    void RestoreGuestCompute();
+        if (changed_mask != 0) {
+            for (size_t stage = 0; stage < NUM_STAGES; ++stage) {
+                if (((changed_mask >> stage) & 1) != 0) {
+                    if (((stage_mask >> stage) & 1) != 0) {
+                        glEnable(ASSEMBLY_PROGRAM_ENUMS[stage]);
+                    } else {
+                        glDisable(ASSEMBLY_PROGRAM_ENUMS[stage]);
+                    }
+                }
+            }
+        }
+        for (size_t stage = 0; stage < NUM_STAGES; ++stage) {
+            if (current_programs[stage] != programs[stage].handle) {
+                current_programs[stage] = programs[stage].handle;
+                glBindProgramARB(ASSEMBLY_PROGRAM_ENUMS[stage], programs[stage].handle);
+            }
+        }
+        UnbindPipeline();
+    }
+
+    void RestoreGuestCompute() {}
 
 private:
-    void BindPipeline();
+    void BindPipeline() {
+        if (!is_pipeline_bound) {
+            is_pipeline_bound = true;
+            glBindProgramPipeline(pipeline.handle);
+        }
+        UnbindCompute();
+    }
 
-    void UnbindPipeline();
+    void UnbindPipeline() {
+        if (is_pipeline_bound) {
+            is_pipeline_bound = false;
+            glBindProgramPipeline(0);
+        }
+        UnbindCompute();
+    }
 
-    void UnbindCompute();
+    void UnbindCompute() {
+        if (is_compute_bound) {
+            is_compute_bound = false;
+            glUseProgram(0);
+        }
+    }
 
     OGLPipeline pipeline;
     bool is_pipeline_bound{};

src/video_core/renderer_opengl/renderer_opengl.cpp

@@ -442,7 +442,13 @@ void RendererOpenGL::DrawScreen(const Layout::FramebufferLayout& layout) {
 
     glBindTextureUnit(0, screen_info.display_texture);
 
-    const auto anti_aliasing = Settings::values.anti_aliasing.GetValue();
+    auto anti_aliasing = Settings::values.anti_aliasing.GetValue();
+    if (anti_aliasing > Settings::AntiAliasing::LastAA) {
+        LOG_ERROR(Render_OpenGL, "Invalid antialiasing option selected {}", anti_aliasing);
+        anti_aliasing = Settings::AntiAliasing::None;
+        Settings::values.anti_aliasing.SetValue(anti_aliasing);
+    }
+
     if (anti_aliasing != Settings::AntiAliasing::None) {
         glEnablei(GL_SCISSOR_TEST, 0);
         auto viewport_width = screen_info.texture.width;

src/video_core/renderer_vulkan/blit_image.cpp

@@ -4,13 +4,13 @@
 #include <algorithm>
 
 #include "common/settings.h"
-#include "video_core/host_shaders/blit_color_float_frag_spv.h"
 #include "video_core/host_shaders/convert_abgr8_to_d24s8_frag_spv.h"
 #include "video_core/host_shaders/convert_d24s8_to_abgr8_frag_spv.h"
 #include "video_core/host_shaders/convert_depth_to_float_frag_spv.h"
 #include "video_core/host_shaders/convert_float_to_depth_frag_spv.h"
 #include "video_core/host_shaders/convert_s8d24_to_abgr8_frag_spv.h"
 #include "video_core/host_shaders/full_screen_triangle_vert_spv.h"
+#include "video_core/host_shaders/vulkan_blit_color_float_frag_spv.h"
 #include "video_core/host_shaders/vulkan_blit_depth_stencil_frag_spv.h"
 #include "video_core/renderer_vulkan/blit_image.h"
 #include "video_core/renderer_vulkan/maxwell_to_vk.h"
@@ -303,7 +303,7 @@ void UpdateTwoTexturesDescriptorSet(const Device& device, VkDescriptorSet descri
 }
 
 void BindBlitState(vk::CommandBuffer cmdbuf, VkPipelineLayout layout, const Region2D& dst_region,
-                   const Region2D& src_region, const Extent3D& src_size = {1, 1, 1}) {
+                   const Region2D& src_region) {
     const VkOffset2D offset{
         .x = std::min(dst_region.start.x, dst_region.end.x),
         .y = std::min(dst_region.start.y, dst_region.end.y),
@@ -325,15 +325,12 @@ void BindBlitState(vk::CommandBuffer cmdbuf, VkPipelineLayout layout, const Regi
         .offset = offset,
         .extent = extent,
     };
-    const float scale_x = static_cast<float>(src_region.end.x - src_region.start.x) /
-                          static_cast<float>(src_size.width);
-    const float scale_y = static_cast<float>(src_region.end.y - src_region.start.y) /
-                          static_cast<float>(src_size.height);
+    const float scale_x = static_cast<float>(src_region.end.x - src_region.start.x);
+    const float scale_y = static_cast<float>(src_region.end.y - src_region.start.y);
     const PushConstants push_constants{
         .tex_scale = {scale_x, scale_y},
-        .tex_offset = {static_cast<float>(src_region.start.x) / static_cast<float>(src_size.width),
-                       static_cast<float>(src_region.start.y) /
-                           static_cast<float>(src_size.height)},
+        .tex_offset = {static_cast<float>(src_region.start.x),
+                       static_cast<float>(src_region.start.y)},
     };
     cmdbuf.SetViewport(0, viewport);
     cmdbuf.SetScissor(0, scissor);
@@ -350,51 +347,6 @@ VkExtent2D GetConversionExtent(const ImageView& src_image_view) {
         .height = is_rescaled ? resolution.ScaleUp(height) : height,
     };
 }
-
-void TransitionImageLayout(vk::CommandBuffer& cmdbuf, VkImage image, VkImageLayout target_layout,
-                           VkImageLayout source_layout = VK_IMAGE_LAYOUT_GENERAL) {
-    constexpr VkFlags flags{VK_ACCESS_COLOR_ATTACHMENT_READ_BIT |
-                            VK_ACCESS_COLOR_ATTACHMENT_WRITE_BIT | VK_ACCESS_SHADER_READ_BIT};
-    const VkImageMemoryBarrier barrier{
-        .sType = VK_STRUCTURE_TYPE_IMAGE_MEMORY_BARRIER,
-        .pNext = nullptr,
-        .srcAccessMask = flags,
-        .dstAccessMask = flags,
-        .oldLayout = source_layout,
-        .newLayout = target_layout,
-        .srcQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED,
-        .dstQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED,
-        .image = image,
-        .subresourceRange{
-            .aspectMask = VK_IMAGE_ASPECT_COLOR_BIT,
-            .baseMipLevel = 0,
-            .levelCount = 1,
-            .baseArrayLayer = 0,
-            .layerCount = 1,
-        },
-    };
-    cmdbuf.PipelineBarrier(VK_PIPELINE_STAGE_ALL_COMMANDS_BIT, VK_PIPELINE_STAGE_ALL_COMMANDS_BIT,
-                           0, barrier);
-}
-
-void BeginRenderPass(vk::CommandBuffer& cmdbuf, const Framebuffer* framebuffer) {
-    const VkRenderPass render_pass = framebuffer->RenderPass();
-    const VkFramebuffer framebuffer_handle = framebuffer->Handle();
-    const VkExtent2D render_area = framebuffer->RenderArea();
-    const VkRenderPassBeginInfo renderpass_bi{
-        .sType = VK_STRUCTURE_TYPE_RENDER_PASS_BEGIN_INFO,
-        .pNext = nullptr,
-        .renderPass = render_pass,
-        .framebuffer = framebuffer_handle,
-        .renderArea{
-            .offset{},
-            .extent = render_area,
-        },
-        .clearValueCount = 0,
-        .pClearValues = nullptr,
-    };
-    cmdbuf.BeginRenderPass(renderpass_bi, VK_SUBPASS_CONTENTS_INLINE);
-}
 } // Anonymous namespace
 
 BlitImageHelper::BlitImageHelper(const Device& device_, Scheduler& scheduler_,
@@ -413,7 +365,7 @@ BlitImageHelper::BlitImageHelper(const Device& device_, Scheduler& scheduler_,
       two_textures_pipeline_layout(device.GetLogical().CreatePipelineLayout(
           PipelineLayoutCreateInfo(two_textures_set_layout.address()))),
       full_screen_vert(BuildShader(device, FULL_SCREEN_TRIANGLE_VERT_SPV)),
-      blit_color_to_color_frag(BuildShader(device, BLIT_COLOR_FLOAT_FRAG_SPV)),
+      blit_color_to_color_frag(BuildShader(device, VULKAN_BLIT_COLOR_FLOAT_FRAG_SPV)),
       blit_depth_stencil_frag(BuildShader(device, VULKAN_BLIT_DEPTH_STENCIL_FRAG_SPV)),
       convert_depth_to_float_frag(BuildShader(device, CONVERT_DEPTH_TO_FLOAT_FRAG_SPV)),
       convert_float_to_depth_frag(BuildShader(device, CONVERT_FLOAT_TO_DEPTH_FRAG_SPV)),
@@ -452,32 +404,6 @@ void BlitImageHelper::BlitColor(const Framebuffer* dst_framebuffer, VkImageView
     scheduler.InvalidateState();
 }
 
-void BlitImageHelper::BlitColor(const Framebuffer* dst_framebuffer, VkImageView src_image_view,
-                                VkImage src_image, VkSampler src_sampler,
-                                const Region2D& dst_region, const Region2D& src_region,
-                                const Extent3D& src_size) {
-    const BlitImagePipelineKey key{
-        .renderpass = dst_framebuffer->RenderPass(),
-        .operation = Tegra::Engines::Fermi2D::Operation::SrcCopy,
-    };
-    const VkPipelineLayout layout = *one_texture_pipeline_layout;
-    const VkPipeline pipeline = FindOrEmplaceColorPipeline(key);
-    scheduler.RequestOutsideRenderPassOperationContext();
-    scheduler.Record([this, dst_framebuffer, src_image_view, src_image, src_sampler, dst_region,
-                      src_region, src_size, pipeline, layout](vk::CommandBuffer cmdbuf) {
-        TransitionImageLayout(cmdbuf, src_image, VK_IMAGE_LAYOUT_READ_ONLY_OPTIMAL);
-        BeginRenderPass(cmdbuf, dst_framebuffer);
-        const VkDescriptorSet descriptor_set = one_texture_descriptor_allocator.Commit();
-        UpdateOneTextureDescriptorSet(device, descriptor_set, src_sampler, src_image_view);
-        cmdbuf.BindPipeline(VK_PIPELINE_BIND_POINT_GRAPHICS, pipeline);
-        cmdbuf.BindDescriptorSets(VK_PIPELINE_BIND_POINT_GRAPHICS, layout, 0, descriptor_set,
-                                  nullptr);
-        BindBlitState(cmdbuf, layout, dst_region, src_region, src_size);
-        cmdbuf.Draw(3, 1, 0, 0);
-        cmdbuf.EndRenderPass();
-    });
-}
-
 void BlitImageHelper::BlitDepthStencil(const Framebuffer* dst_framebuffer,
                                        VkImageView src_depth_view, VkImageView src_stencil_view,
                                        const Region2D& dst_region, const Region2D& src_region,

src/video_core/renderer_vulkan/blit_image.h

@@ -10,8 +10,6 @@
 
 namespace Vulkan {
 
-using VideoCommon::Extent3D;
-using VideoCommon::Offset2D;
 using VideoCommon::Region2D;
 
 class Device;
@@ -38,10 +36,6 @@ public:
                    Tegra::Engines::Fermi2D::Filter filter,
                    Tegra::Engines::Fermi2D::Operation operation);
 
-    void BlitColor(const Framebuffer* dst_framebuffer, VkImageView src_image_view,
-                   VkImage src_image, VkSampler src_sampler, const Region2D& dst_region,
-                   const Region2D& src_region, const Extent3D& src_size);
-
     void BlitDepthStencil(const Framebuffer* dst_framebuffer, VkImageView src_depth_view,
                           VkImageView src_stencil_view, const Region2D& dst_region,
                           const Region2D& src_region, Tegra::Engines::Fermi2D::Filter filter,

src/video_core/renderer_vulkan/fixed_pipeline_state.cpp

@@ -148,7 +148,7 @@ void FixedPipelineState::Refresh(Tegra::Engines::Maxwell3D& maxwell3d, DynamicFe
         });
     }
     if (!extended_dynamic_state_2_extra) {
-        dynamic_state.Refresh2(regs, topology, extended_dynamic_state_2);
+        dynamic_state.Refresh2(regs, topology_, extended_dynamic_state_2);
     }
     if (!extended_dynamic_state_3_blend) {
         if (maxwell3d.dirty.flags[Dirty::Blending]) {

src/video_core/renderer_vulkan/renderer_vulkan.cpp

@@ -78,6 +78,8 @@ std::string BuildCommaSeparatedExtensions(std::vector<std::string> available_ext
     return separated_extensions;
 }
 
+} // Anonymous namespace
+
 Device CreateDevice(const vk::Instance& instance, const vk::InstanceDispatch& dld,
                     VkSurfaceKHR surface) {
     const std::vector<VkPhysicalDevice> devices = instance.EnumeratePhysicalDevices();
@@ -89,7 +91,6 @@ Device CreateDevice(const vk::Instance& instance, const vk::InstanceDispatch& dl
     const vk::PhysicalDevice physical_device(devices[device_index], dld);
     return Device(*instance, physical_device, surface, dld);
 }
-} // Anonymous namespace
 
 RendererVulkan::RendererVulkan(Core::TelemetrySession& telemetry_session_,
                                Core::Frontend::EmuWindow& emu_window,
@@ -98,7 +99,7 @@ RendererVulkan::RendererVulkan(Core::TelemetrySession& telemetry_session_,
     : RendererBase(emu_window, std::move(context_)), telemetry_session(telemetry_session_),
       cpu_memory(cpu_memory_), gpu(gpu_), library(OpenLibrary()),
       instance(CreateInstance(library, dld, VK_API_VERSION_1_1, render_window.GetWindowInfo().type,
-                              true, Settings::values.renderer_debug.GetValue())),
+                              Settings::values.renderer_debug.GetValue())),
       debug_callback(Settings::values.renderer_debug ? CreateDebugCallback(instance) : nullptr),
       surface(CreateSurface(instance, render_window)),
       device(CreateDevice(instance, dld, *surface)), memory_allocator(device, false),
@@ -109,6 +110,9 @@ RendererVulkan::RendererVulkan(Core::TelemetrySession& telemetry_session_,
                   screen_info),
       rasterizer(render_window, gpu, cpu_memory, screen_info, device, memory_allocator,
                  state_tracker, scheduler) {
+    if (Settings::values.renderer_force_max_clock.GetValue()) {
+        turbo_mode.emplace(instance, dld);
+    }
     Report();
 } catch (const vk::Exception& exception) {
     LOG_ERROR(Render_Vulkan, "Vulkan initialization failed with error: {}", exception.what());

src/video_core/renderer_vulkan/renderer_vulkan.h

@@ -13,6 +13,7 @@
 #include "video_core/renderer_vulkan/vk_scheduler.h"
 #include "video_core/renderer_vulkan/vk_state_tracker.h"
 #include "video_core/renderer_vulkan/vk_swapchain.h"
+#include "video_core/renderer_vulkan/vk_turbo_mode.h"
 #include "video_core/vulkan_common/vulkan_device.h"
 #include "video_core/vulkan_common/vulkan_memory_allocator.h"
 #include "video_core/vulkan_common/vulkan_wrapper.h"
@@ -31,6 +32,9 @@ class GPU;
 
 namespace Vulkan {
 
+Device CreateDevice(const vk::Instance& instance, const vk::InstanceDispatch& dld,
+                    VkSurfaceKHR surface);
+
 class RendererVulkan final : public VideoCore::RendererBase {
 public:
     explicit RendererVulkan(Core::TelemetrySession& telemtry_session,
@@ -74,6 +78,7 @@ private:
     Swapchain swapchain;
     BlitScreen blit_screen;
     RasterizerVulkan rasterizer;
+    std::optional<TurboMode> turbo_mode;
 };
 
 } // namespace Vulkan

src/video_core/renderer_vulkan/vk_buffer_cache.cpp

@@ -330,12 +330,19 @@ bool BufferCacheRuntime::CanReportMemoryUsage() const {
     return device.CanReportMemoryUsage();
 }
 
+u32 BufferCacheRuntime::GetStorageBufferAlignment() const {
+    return static_cast<u32>(device.GetStorageBufferAlignment());
+}
+
 void BufferCacheRuntime::Finish() {
     scheduler.Finish();
 }
 
 void BufferCacheRuntime::CopyBuffer(VkBuffer dst_buffer, VkBuffer src_buffer,
                                     std::span<const VideoCommon::BufferCopy> copies, bool barrier) {
+    if (dst_buffer == VK_NULL_HANDLE || src_buffer == VK_NULL_HANDLE) {
+        return;
+    }
     static constexpr VkMemoryBarrier READ_BARRIER{
         .sType = VK_STRUCTURE_TYPE_MEMORY_BARRIER,
         .pNext = nullptr,
@@ -394,6 +401,9 @@ void BufferCacheRuntime::PostCopyBarrier() {
 }
 
 void BufferCacheRuntime::ClearBuffer(VkBuffer dest_buffer, u32 offset, size_t size, u32 value) {
+    if (dest_buffer == VK_NULL_HANDLE) {
+        return;
+    }
     static constexpr VkMemoryBarrier READ_BARRIER{
         .sType = VK_STRUCTURE_TYPE_MEMORY_BARRIER,
         .pNext = nullptr,
@@ -473,6 +483,11 @@ void BufferCacheRuntime::BindVertexBuffer(u32 index, VkBuffer buffer, u32 offset
             cmdbuf.BindVertexBuffers2EXT(index, 1, &buffer, &vk_offset, &vk_size, &vk_stride);
         });
     } else {
+        if (!device.HasNullDescriptor() && buffer == VK_NULL_HANDLE) {
+            ReserveNullBuffer();
+            buffer = *null_buffer;
+            offset = 0;
+        }
         scheduler.Record([index, buffer, offset](vk::CommandBuffer cmdbuf) {
             cmdbuf.BindVertexBuffer(index, buffer, offset);
         });

src/video_core/renderer_vulkan/vk_buffer_cache.h

@@ -73,6 +73,8 @@ public:
 
     bool CanReportMemoryUsage() const;
 
+    u32 GetStorageBufferAlignment() const;
+
     [[nodiscard]] StagingBufferRef UploadStagingBuffer(size_t size);
 
     [[nodiscard]] StagingBufferRef DownloadStagingBuffer(size_t size);

src/video_core/renderer_vulkan/vk_compute_pipeline.cpp

@@ -24,13 +24,15 @@ using Shader::ImageBufferDescriptor;
 using Shader::Backend::SPIRV::RESCALING_LAYOUT_WORDS_OFFSET;
 using Tegra::Texture::TexturePair;
 
-ComputePipeline::ComputePipeline(const Device& device_, DescriptorPool& descriptor_pool,
+ComputePipeline::ComputePipeline(const Device& device_, vk::PipelineCache& pipeline_cache_,
+                                 DescriptorPool& descriptor_pool,
                                  UpdateDescriptorQueue& update_descriptor_queue_,
                                  Common::ThreadWorker* thread_worker,
                                  PipelineStatistics* pipeline_statistics,
                                  VideoCore::ShaderNotify* shader_notify, const Shader::Info& info_,
                                  vk::ShaderModule spv_module_)
-    : device{device_}, update_descriptor_queue{update_descriptor_queue_}, info{info_},
+    : device{device_}, pipeline_cache(pipeline_cache_),
+      update_descriptor_queue{update_descriptor_queue_}, info{info_},
       spv_module(std::move(spv_module_)) {
     if (shader_notify) {
         shader_notify->MarkShaderBuilding();
@@ -56,23 +58,27 @@ ComputePipeline::ComputePipeline(const Device& device_, DescriptorPool& descript
         if (device.IsKhrPipelineExecutablePropertiesEnabled()) {
             flags |= VK_PIPELINE_CREATE_CAPTURE_STATISTICS_BIT_KHR;
         }
-        pipeline = device.GetLogical().CreateComputePipeline({
-            .sType = VK_STRUCTURE_TYPE_COMPUTE_PIPELINE_CREATE_INFO,
-            .pNext = nullptr,
-            .flags = flags,
-            .stage{
-                .sType = VK_STRUCTURE_TYPE_PIPELINE_SHADER_STAGE_CREATE_INFO,
-                .pNext = device.IsExtSubgroupSizeControlSupported() ? &subgroup_size_ci : nullptr,
-                .flags = 0,
-                .stage = VK_SHADER_STAGE_COMPUTE_BIT,
-                .module = *spv_module,
-                .pName = "main",
-                .pSpecializationInfo = nullptr,
+        pipeline = device.GetLogical().CreateComputePipeline(
+            {
+                .sType = VK_STRUCTURE_TYPE_COMPUTE_PIPELINE_CREATE_INFO,
+                .pNext = nullptr,
+                .flags = flags,
+                .stage{
+                    .sType = VK_STRUCTURE_TYPE_PIPELINE_SHADER_STAGE_CREATE_INFO,
+                    .pNext =
+                        device.IsExtSubgroupSizeControlSupported() ? &subgroup_size_ci : nullptr,
+                    .flags = 0,
+                    .stage = VK_SHADER_STAGE_COMPUTE_BIT,
+                    .module = *spv_module,
+                    .pName = "main",
+                    .pSpecializationInfo = nullptr,
+                },
+                .layout = *pipeline_layout,
+                .basePipelineHandle = 0,
+                .basePipelineIndex = 0,
             },
-            .layout = *pipeline_layout,
-            .basePipelineHandle = 0,
-            .basePipelineIndex = 0,
-        });
+            *pipeline_cache);
+
         if (pipeline_statistics) {
             pipeline_statistics->Collect(*pipeline);
         }

src/video_core/renderer_vulkan/vk_compute_pipeline.h

@@ -28,7 +28,8 @@ class Scheduler;
 
 class ComputePipeline {
 public:
-    explicit ComputePipeline(const Device& device, DescriptorPool& descriptor_pool,
+    explicit ComputePipeline(const Device& device, vk::PipelineCache& pipeline_cache,
+                             DescriptorPool& descriptor_pool,
                              UpdateDescriptorQueue& update_descriptor_queue,
                              Common::ThreadWorker* thread_worker,
                              PipelineStatistics* pipeline_statistics,
@@ -46,6 +47,7 @@ public:
 
 private:
     const Device& device;
+    vk::PipelineCache& pipeline_cache;
     UpdateDescriptorQueue& update_descriptor_queue;
     Shader::Info info;
 

src/video_core/renderer_vulkan/vk_graphics_pipeline.cpp

@@ -234,13 +234,14 @@ ConfigureFuncPtr ConfigureFunc(const std::array<vk::ShaderModule, NUM_STAGES>& m
 
 GraphicsPipeline::GraphicsPipeline(
     Scheduler& scheduler_, BufferCache& buffer_cache_, TextureCache& texture_cache_,
-    VideoCore::ShaderNotify* shader_notify, const Device& device_, DescriptorPool& descriptor_pool,
+    vk::PipelineCache& pipeline_cache_, VideoCore::ShaderNotify* shader_notify,
+    const Device& device_, DescriptorPool& descriptor_pool,
     UpdateDescriptorQueue& update_descriptor_queue_, Common::ThreadWorker* worker_thread,
     PipelineStatistics* pipeline_statistics, RenderPassCache& render_pass_cache,
     const GraphicsPipelineCacheKey& key_, std::array<vk::ShaderModule, NUM_STAGES> stages,
     const std::array<const Shader::Info*, NUM_STAGES>& infos)
-    : key{key_}, device{device_}, texture_cache{texture_cache_},
-      buffer_cache{buffer_cache_}, scheduler{scheduler_},
+    : key{key_}, device{device_}, texture_cache{texture_cache_}, buffer_cache{buffer_cache_},
+      pipeline_cache(pipeline_cache_), scheduler{scheduler_},
       update_descriptor_queue{update_descriptor_queue_}, spv_modules{std::move(stages)} {
     if (shader_notify) {
         shader_notify->MarkShaderBuilding();
@@ -644,12 +645,15 @@ void GraphicsPipeline::MakePipeline(VkRenderPass render_pass) {
         .pNext = nullptr,
         .flags = 0,
         .topology = input_assembly_topology,
-        .primitiveRestartEnable = dynamic.primitive_restart_enable != 0 &&
-                                  ((input_assembly_topology != VK_PRIMITIVE_TOPOLOGY_PATCH_LIST &&
-                                    device.IsTopologyListPrimitiveRestartSupported()) ||
-                                   SupportsPrimitiveRestart(input_assembly_topology) ||
-                                   (input_assembly_topology == VK_PRIMITIVE_TOPOLOGY_PATCH_LIST &&
-                                    device.IsPatchListPrimitiveRestartSupported())),
+        .primitiveRestartEnable =
+            dynamic.primitive_restart_enable != 0 &&
+                    ((input_assembly_topology != VK_PRIMITIVE_TOPOLOGY_PATCH_LIST &&
+                      device.IsTopologyListPrimitiveRestartSupported()) ||
+                     SupportsPrimitiveRestart(input_assembly_topology) ||
+                     (input_assembly_topology == VK_PRIMITIVE_TOPOLOGY_PATCH_LIST &&
+                      device.IsPatchListPrimitiveRestartSupported()))
+                ? VK_TRUE
+                : VK_FALSE,
     };
     const VkPipelineTessellationStateCreateInfo tessellation_ci{
         .sType = VK_STRUCTURE_TYPE_PIPELINE_TESSELLATION_STATE_CREATE_INFO,
@@ -699,7 +703,7 @@ void GraphicsPipeline::MakePipeline(VkRenderPass render_pass) {
         .cullMode = static_cast<VkCullModeFlags>(
             dynamic.cull_enable ? MaxwellToVK::CullFace(dynamic.CullFace()) : VK_CULL_MODE_NONE),
         .frontFace = MaxwellToVK::FrontFace(dynamic.FrontFace()),
-        .depthBiasEnable = (dynamic.depth_bias_enable == 0 ? VK_TRUE : VK_FALSE),
+        .depthBiasEnable = (dynamic.depth_bias_enable != 0 ? VK_TRUE : VK_FALSE),
         .depthBiasConstantFactor = 0.0f,
         .depthBiasClamp = 0.0f,
         .depthBiasSlopeFactor = 0.0f,
@@ -894,27 +898,29 @@ void GraphicsPipeline::MakePipeline(VkRenderPass render_pass) {
     if (device.IsKhrPipelineExecutablePropertiesEnabled()) {
         flags |= VK_PIPELINE_CREATE_CAPTURE_STATISTICS_BIT_KHR;
     }
-    pipeline = device.GetLogical().CreateGraphicsPipeline({
-        .sType = VK_STRUCTURE_TYPE_GRAPHICS_PIPELINE_CREATE_INFO,
-        .pNext = nullptr,
-        .flags = flags,
-        .stageCount = static_cast<u32>(shader_stages.size()),
-        .pStages = shader_stages.data(),
-        .pVertexInputState = &vertex_input_ci,
-        .pInputAssemblyState = &input_assembly_ci,
-        .pTessellationState = &tessellation_ci,
-        .pViewportState = &viewport_ci,
-        .pRasterizationState = &rasterization_ci,
-        .pMultisampleState = &multisample_ci,
-        .pDepthStencilState = &depth_stencil_ci,
-        .pColorBlendState = &color_blend_ci,
-        .pDynamicState = &dynamic_state_ci,
-        .layout = *pipeline_layout,
-        .renderPass = render_pass,
-        .subpass = 0,
-        .basePipelineHandle = nullptr,
-        .basePipelineIndex = 0,
-    });
+    pipeline = device.GetLogical().CreateGraphicsPipeline(
+        {
+            .sType = VK_STRUCTURE_TYPE_GRAPHICS_PIPELINE_CREATE_INFO,
+            .pNext = nullptr,
+            .flags = flags,
+            .stageCount = static_cast<u32>(shader_stages.size()),
+            .pStages = shader_stages.data(),
+            .pVertexInputState = &vertex_input_ci,
+            .pInputAssemblyState = &input_assembly_ci,
+            .pTessellationState = &tessellation_ci,
+            .pViewportState = &viewport_ci,
+            .pRasterizationState = &rasterization_ci,
+            .pMultisampleState = &multisample_ci,
+            .pDepthStencilState = &depth_stencil_ci,
+            .pColorBlendState = &color_blend_ci,
+            .pDynamicState = &dynamic_state_ci,
+            .layout = *pipeline_layout,
+            .renderPass = render_pass,
+            .subpass = 0,
+            .basePipelineHandle = nullptr,
+            .basePipelineIndex = 0,
+        },
+        *pipeline_cache);
 }
 
 void GraphicsPipeline::Validate() {

src/video_core/renderer_vulkan/vk_graphics_pipeline.h

@@ -70,16 +70,14 @@ class GraphicsPipeline {
     static constexpr size_t NUM_STAGES = Tegra::Engines::Maxwell3D::Regs::MaxShaderStage;
 
 public:
-    explicit GraphicsPipeline(Scheduler& scheduler, BufferCache& buffer_cache,
-                              TextureCache& texture_cache, VideoCore::ShaderNotify* shader_notify,
-                              const Device& device, DescriptorPool& descriptor_pool,
-                              UpdateDescriptorQueue& update_descriptor_queue,
-                              Common::ThreadWorker* worker_thread,
-                              PipelineStatistics* pipeline_statistics,
-                              RenderPassCache& render_pass_cache,
-                              const GraphicsPipelineCacheKey& key,
-                              std::array<vk::ShaderModule, NUM_STAGES> stages,
-                              const std::array<const Shader::Info*, NUM_STAGES>& infos);
+    explicit GraphicsPipeline(
+        Scheduler& scheduler, BufferCache& buffer_cache, TextureCache& texture_cache,
+        vk::PipelineCache& pipeline_cache, VideoCore::ShaderNotify* shader_notify,
+        const Device& device, DescriptorPool& descriptor_pool,
+        UpdateDescriptorQueue& update_descriptor_queue, Common::ThreadWorker* worker_thread,
+        PipelineStatistics* pipeline_statistics, RenderPassCache& render_pass_cache,
+        const GraphicsPipelineCacheKey& key, std::array<vk::ShaderModule, NUM_STAGES> stages,
+        const std::array<const Shader::Info*, NUM_STAGES>& infos);
 
     GraphicsPipeline& operator=(GraphicsPipeline&&) noexcept = delete;
     GraphicsPipeline(GraphicsPipeline&&) noexcept = delete;
@@ -133,6 +131,7 @@ private:
     const Device& device;
     TextureCache& texture_cache;
     BufferCache& buffer_cache;
+    vk::PipelineCache& pipeline_cache;
     Scheduler& scheduler;
     UpdateDescriptorQueue& update_descriptor_queue;
 

src/video_core/renderer_vulkan/vk_pipeline_cache.cpp

@@ -55,6 +55,7 @@ using VideoCommon::GenericEnvironment;
 using VideoCommon::GraphicsEnvironment;
 
 constexpr u32 CACHE_VERSION = 10;
+constexpr std::array<char, 8> VULKAN_CACHE_MAGIC_NUMBER{'y', 'u', 'z', 'u', 'v', 'k', 'c', 'h'};
 
 template <typename Container>
 auto MakeSpan(Container& container) {
@@ -284,6 +285,7 @@ PipelineCache::PipelineCache(RasterizerVulkan& rasterizer_, const Device& device
       render_pass_cache{render_pass_cache_}, buffer_cache{buffer_cache_},
       texture_cache{texture_cache_}, shader_notify{shader_notify_},
       use_asynchronous_shaders{Settings::values.use_asynchronous_shaders.GetValue()},
+      use_vulkan_pipeline_cache{Settings::values.use_vulkan_driver_pipeline_cache.GetValue()},
       workers(std::max(std::thread::hardware_concurrency(), 2U) - 1, "VkPipelineBuilder"),
       serialization_thread(1, "VkPipelineSerialization") {
     const auto& float_control{device.FloatControlProperties()};
@@ -329,6 +331,7 @@ PipelineCache::PipelineCache(RasterizerVulkan& rasterizer_, const Device& device
         .need_declared_frag_colors = false,
 
         .has_broken_spirv_clamp = driver_id == VK_DRIVER_ID_INTEL_PROPRIETARY_WINDOWS,
+        .has_broken_spirv_position_input = driver_id == VK_DRIVER_ID_QUALCOMM_PROPRIETARY,
         .has_broken_unsigned_image_offsets = false,
         .has_broken_signed_operations = false,
         .has_broken_fp16_float_controls = driver_id == VK_DRIVER_ID_NVIDIA_PROPRIETARY,
@@ -341,6 +344,8 @@ PipelineCache::PipelineCache(RasterizerVulkan& rasterizer_, const Device& device
             driver_id == VK_DRIVER_ID_AMD_PROPRIETARY || driver_id == VK_DRIVER_ID_AMD_OPEN_SOURCE,
         .support_snorm_render_buffer = true,
         .support_viewport_index_layer = device.IsExtShaderViewportIndexLayerSupported(),
+        .min_ssbo_alignment = static_cast<u32>(device.GetStorageBufferAlignment()),
+        .support_geometry_shader_passthrough = device.IsNvGeometryShaderPassthroughSupported(),
     };
 
     if (device.GetMaxVertexInputAttributes() < Maxwell::NumVertexAttributes) {
@@ -362,7 +367,12 @@ PipelineCache::PipelineCache(RasterizerVulkan& rasterizer_, const Device& device
     };
 }
 
-PipelineCache::~PipelineCache() = default;
+PipelineCache::~PipelineCache() {
+    if (use_vulkan_pipeline_cache && !vulkan_pipeline_cache_filename.empty()) {
+        SerializeVulkanPipelineCache(vulkan_pipeline_cache_filename, vulkan_pipeline_cache,
+                                     CACHE_VERSION);
+    }
+}
 
 GraphicsPipeline* PipelineCache::CurrentGraphicsPipeline() {
     MICROPROFILE_SCOPE(Vulkan_PipelineCache);
@@ -418,6 +428,12 @@ void PipelineCache::LoadDiskResources(u64 title_id, std::stop_token stop_loading
     }
     pipeline_cache_filename = base_dir / "vulkan.bin";
 
+    if (use_vulkan_pipeline_cache) {
+        vulkan_pipeline_cache_filename = base_dir / "vulkan_pipelines.bin";
+        vulkan_pipeline_cache =
+            LoadVulkanPipelineCache(vulkan_pipeline_cache_filename, CACHE_VERSION);
+    }
+
     struct {
         std::mutex mutex;
         size_t total{};
@@ -496,6 +512,11 @@ void PipelineCache::LoadDiskResources(u64 title_id, std::stop_token stop_loading
 
     workers.WaitForRequests(stop_loading);
 
+    if (use_vulkan_pipeline_cache) {
+        SerializeVulkanPipelineCache(vulkan_pipeline_cache_filename, vulkan_pipeline_cache,
+                                     CACHE_VERSION);
+    }
+
     if (state.statistics) {
         state.statistics->Report();
     }
@@ -616,10 +637,10 @@ std::unique_ptr<GraphicsPipeline> PipelineCache::CreateGraphicsPipeline(
         previous_stage = &program;
     }
     Common::ThreadWorker* const thread_worker{build_in_parallel ? &workers : nullptr};
-    return std::make_unique<GraphicsPipeline>(scheduler, buffer_cache, texture_cache,
-                                              &shader_notify, device, descriptor_pool,
-                                              update_descriptor_queue, thread_worker, statistics,
-                                              render_pass_cache, key, std::move(modules), infos);
+    return std::make_unique<GraphicsPipeline>(
+        scheduler, buffer_cache, texture_cache, vulkan_pipeline_cache, &shader_notify, device,
+        descriptor_pool, update_descriptor_queue, thread_worker, statistics, render_pass_cache, key,
+        std::move(modules), infos);
 
 } catch (const Shader::Exception& exception) {
     LOG_ERROR(Render_Vulkan, "{}", exception.what());
@@ -689,13 +710,108 @@ std::unique_ptr<ComputePipeline> PipelineCache::CreateComputePipeline(
         spv_module.SetObjectNameEXT(name.c_str());
     }
     Common::ThreadWorker* const thread_worker{build_in_parallel ? &workers : nullptr};
-    return std::make_unique<ComputePipeline>(device, descriptor_pool, update_descriptor_queue,
-                                             thread_worker, statistics, &shader_notify,
-                                             program.info, std::move(spv_module));
+    return std::make_unique<ComputePipeline>(device, vulkan_pipeline_cache, descriptor_pool,
+                                             update_descriptor_queue, thread_worker, statistics,
+                                             &shader_notify, program.info, std::move(spv_module));
 
 } catch (const Shader::Exception& exception) {
     LOG_ERROR(Render_Vulkan, "{}", exception.what());
     return nullptr;
 }
 
+void PipelineCache::SerializeVulkanPipelineCache(const std::filesystem::path& filename,
+                                                 const vk::PipelineCache& pipeline_cache,
+                                                 u32 cache_version) try {
+    std::ofstream file(filename, std::ios::binary);
+    file.exceptions(std::ifstream::failbit);
+    if (!file.is_open()) {
+        LOG_ERROR(Common_Filesystem, "Failed to open Vulkan driver pipeline cache file {}",
+                  Common::FS::PathToUTF8String(filename));
+        return;
+    }
+    file.write(VULKAN_CACHE_MAGIC_NUMBER.data(), VULKAN_CACHE_MAGIC_NUMBER.size())
+        .write(reinterpret_cast<const char*>(&cache_version), sizeof(cache_version));
+
+    size_t cache_size = 0;
+    std::vector<char> cache_data;
+    if (pipeline_cache) {
+        pipeline_cache.Read(&cache_size, nullptr);
+        cache_data.resize(cache_size);
+        pipeline_cache.Read(&cache_size, cache_data.data());
+    }
+    file.write(cache_data.data(), cache_size);
+
+    LOG_INFO(Render_Vulkan, "Vulkan driver pipelines cached at: {}",
+             Common::FS::PathToUTF8String(filename));
+
+} catch (const std::ios_base::failure& e) {
+    LOG_ERROR(Common_Filesystem, "{}", e.what());
+    if (!Common::FS::RemoveFile(filename)) {
+        LOG_ERROR(Common_Filesystem, "Failed to delete Vulkan driver pipeline cache file {}",
+                  Common::FS::PathToUTF8String(filename));
+    }
+}
+
+vk::PipelineCache PipelineCache::LoadVulkanPipelineCache(const std::filesystem::path& filename,
+                                                         u32 expected_cache_version) {
+    const auto create_pipeline_cache = [this](size_t data_size, const void* data) {
+        VkPipelineCacheCreateInfo pipeline_cache_ci = {
+            .sType = VK_STRUCTURE_TYPE_PIPELINE_CACHE_CREATE_INFO,
+            .pNext = nullptr,
+            .flags = 0,
+            .initialDataSize = data_size,
+            .pInitialData = data};
+        return device.GetLogical().CreatePipelineCache(pipeline_cache_ci);
+    };
+    try {
+        std::ifstream file(filename, std::ios::binary | std::ios::ate);
+        if (!file.is_open()) {
+            return create_pipeline_cache(0, nullptr);
+        }
+        file.exceptions(std::ifstream::failbit);
+        const auto end{file.tellg()};
+        file.seekg(0, std::ios::beg);
+
+        std::array<char, 8> magic_number;
+        u32 cache_version;
+        file.read(magic_number.data(), magic_number.size())
+            .read(reinterpret_cast<char*>(&cache_version), sizeof(cache_version));
+        if (magic_number != VULKAN_CACHE_MAGIC_NUMBER || cache_version != expected_cache_version) {
+            file.close();
+            if (Common::FS::RemoveFile(filename)) {
+                if (magic_number != VULKAN_CACHE_MAGIC_NUMBER) {
+                    LOG_ERROR(Common_Filesystem, "Invalid Vulkan driver pipeline cache file");
+                }
+                if (cache_version != expected_cache_version) {
+                    LOG_INFO(Common_Filesystem, "Deleting old Vulkan driver pipeline cache");
+                }
+            } else {
+                LOG_ERROR(Common_Filesystem,
+                          "Invalid Vulkan pipeline cache file and failed to delete it in \"{}\"",
+                          Common::FS::PathToUTF8String(filename));
+            }
+            return create_pipeline_cache(0, nullptr);
+        }
+
+        static constexpr size_t header_size = magic_number.size() + sizeof(cache_version);
+        const size_t cache_size = static_cast<size_t>(end) - header_size;
+        std::vector<char> cache_data(cache_size);
+        file.read(cache_data.data(), cache_size);
+
+        LOG_INFO(Render_Vulkan,
+                 "Loaded Vulkan driver pipeline cache: ", Common::FS::PathToUTF8String(filename));
+
+        return create_pipeline_cache(cache_size, cache_data.data());
+
+    } catch (const std::ios_base::failure& e) {
+        LOG_ERROR(Common_Filesystem, "{}", e.what());
+        if (!Common::FS::RemoveFile(filename)) {
+            LOG_ERROR(Common_Filesystem, "Failed to delete Vulkan driver pipeline cache file {}",
+                      Common::FS::PathToUTF8String(filename));
+        }
+
+        return create_pipeline_cache(0, nullptr);
+    }
+}
+
 } // namespace Vulkan

src/video_core/renderer_vulkan/vk_pipeline_cache.h

@@ -135,6 +135,12 @@ private:
                                                            PipelineStatistics* statistics,
                                                            bool build_in_parallel);
 
+    void SerializeVulkanPipelineCache(const std::filesystem::path& filename,
+                                      const vk::PipelineCache& pipeline_cache, u32 cache_version);
+
+    vk::PipelineCache LoadVulkanPipelineCache(const std::filesystem::path& filename,
+                                              u32 expected_cache_version);
+
     const Device& device;
     Scheduler& scheduler;
     DescriptorPool& descriptor_pool;
@@ -144,6 +150,7 @@ private:
     TextureCache& texture_cache;
     VideoCore::ShaderNotify& shader_notify;
     bool use_asynchronous_shaders{};
+    bool use_vulkan_pipeline_cache{};
 
     GraphicsPipelineCacheKey graphics_key{};
     GraphicsPipeline* current_pipeline{};
@@ -158,6 +165,9 @@ private:
 
     std::filesystem::path pipeline_cache_filename;
 
+    std::filesystem::path vulkan_pipeline_cache_filename;
+    vk::PipelineCache vulkan_pipeline_cache;
+
     Common::ThreadWorker workers;
     Common::ThreadWorker serialization_thread;
     DynamicFeatures dynamic_features;

src/video_core/renderer_vulkan/vk_rasterizer.cpp

@@ -186,6 +186,7 @@ void RasterizerVulkan::PrepareDraw(bool is_indexed, Func&& draw_func) {
 
     SCOPE_EXIT({ gpu.TickWork(); });
     FlushWork();
+    gpu_memory->FlushCaching();
 
     query_cache.UpdateCounters();
 
@@ -265,35 +266,6 @@ void RasterizerVulkan::DrawIndirect() {
     buffer_cache.SetDrawIndirect(nullptr);
 }
 
-void RasterizerVulkan::DrawTexture() {
-    MICROPROFILE_SCOPE(Vulkan_Drawing);
-
-    SCOPE_EXIT({ gpu.TickWork(); });
-    FlushWork();
-
-    query_cache.UpdateCounters();
-
-    texture_cache.SynchronizeGraphicsDescriptors();
-    texture_cache.UpdateRenderTargets(false);
-
-    UpdateDynamicStates();
-
-    const auto& draw_texture_state = maxwell3d->draw_manager->GetDrawTextureState();
-    const auto& sampler = texture_cache.GetGraphicsSampler(draw_texture_state.src_sampler);
-    const auto& texture = texture_cache.GetImageView(draw_texture_state.src_texture);
-    Region2D dst_region = {Offset2D{.x = static_cast<s32>(draw_texture_state.dst_x0),
-                                    .y = static_cast<s32>(draw_texture_state.dst_y0)},
-                           Offset2D{.x = static_cast<s32>(draw_texture_state.dst_x1),
-                                    .y = static_cast<s32>(draw_texture_state.dst_y1)}};
-    Region2D src_region = {Offset2D{.x = static_cast<s32>(draw_texture_state.src_x0),
-                                    .y = static_cast<s32>(draw_texture_state.src_y0)},
-                           Offset2D{.x = static_cast<s32>(draw_texture_state.src_x1),
-                                    .y = static_cast<s32>(draw_texture_state.src_y1)}};
-    blit_image.BlitColor(texture_cache.GetFramebuffer(), texture.RenderTarget(),
-                         texture.ImageHandle(), sampler->Handle(), dst_region, src_region,
-                         texture.size);
-}
-
 void RasterizerVulkan::Clear(u32 layer_count) {
     MICROPROFILE_SCOPE(Vulkan_Clearing);
 
@@ -422,6 +394,7 @@ void RasterizerVulkan::Clear(u32 layer_count) {
 
 void RasterizerVulkan::DispatchCompute() {
     FlushWork();
+    gpu_memory->FlushCaching();
 
     ComputePipeline* const pipeline{pipeline_cache.CurrentComputePipeline()};
     if (!pipeline) {
@@ -510,6 +483,27 @@ void RasterizerVulkan::InvalidateRegion(VAddr addr, u64 size, VideoCommon::Cache
     }
 }
 
+void RasterizerVulkan::InnerInvalidation(std::span<const std::pair<VAddr, std::size_t>> sequences) {
+    {
+        std::scoped_lock lock{texture_cache.mutex};
+        for (const auto& [addr, size] : sequences) {
+            texture_cache.WriteMemory(addr, size);
+        }
+    }
+    {
+        std::scoped_lock lock{buffer_cache.mutex};
+        for (const auto& [addr, size] : sequences) {
+            buffer_cache.WriteMemory(addr, size);
+        }
+    }
+    {
+        for (const auto& [addr, size] : sequences) {
+            query_cache.InvalidateRegion(addr, size);
+            pipeline_cache.InvalidateRegion(addr, size);
+        }
+    }
+}
+
 void RasterizerVulkan::OnCPUWrite(VAddr addr, u64 size) {
     if (addr == 0 || size == 0) {
         return;

src/video_core/renderer_vulkan/vk_rasterizer.h

@@ -66,7 +66,6 @@ public:
 
     void Draw(bool is_indexed, u32 instance_count) override;
     void DrawIndirect() override;
-    void DrawTexture() override;
     void Clear(u32 layer_count) override;
     void DispatchCompute() override;
     void ResetCounter(VideoCore::QueryType type) override;
@@ -80,6 +79,7 @@ public:
                          VideoCommon::CacheType which = VideoCommon::CacheType::All) override;
     void InvalidateRegion(VAddr addr, u64 size,
                           VideoCommon::CacheType which = VideoCommon::CacheType::All) override;
+    void InnerInvalidation(std::span<const std::pair<VAddr, std::size_t>> sequences) override;
     void OnCPUWrite(VAddr addr, u64 size) override;
     void InvalidateGPUCache() override;
     void UnmapMemory(VAddr addr, u64 size) override;

src/video_core/renderer_vulkan/vk_turbo_mode.cpp

@@ -0,0 +1,205 @@
+// SPDX-FileCopyrightText: Copyright 2022 yuzu Emulator Project
+// SPDX-License-Identifier: GPL-2.0-or-later
+
+#include "common/literals.h"
+#include "video_core/host_shaders/vulkan_turbo_mode_comp_spv.h"
+#include "video_core/renderer_vulkan/renderer_vulkan.h"
+#include "video_core/renderer_vulkan/vk_shader_util.h"
+#include "video_core/renderer_vulkan/vk_turbo_mode.h"
+#include "video_core/vulkan_common/vulkan_device.h"
+
+namespace Vulkan {
+
+using namespace Common::Literals;
+
+TurboMode::TurboMode(const vk::Instance& instance, const vk::InstanceDispatch& dld)
+    : m_device{CreateDevice(instance, dld, VK_NULL_HANDLE)}, m_allocator{m_device, false} {
+    m_thread = std::jthread([&](auto stop_token) { Run(stop_token); });
+}
+
+TurboMode::~TurboMode() = default;
+
+void TurboMode::Run(std::stop_token stop_token) {
+    auto& dld = m_device.GetLogical();
+
+    // Allocate buffer. 2MiB should be sufficient.
+    auto buffer = dld.CreateBuffer(VkBufferCreateInfo{
+        .sType = VK_STRUCTURE_TYPE_BUFFER_CREATE_INFO,
+        .pNext = nullptr,
+        .flags = 0,
+        .size = 2_MiB,
+        .usage = VK_BUFFER_USAGE_STORAGE_BUFFER_BIT | VK_BUFFER_USAGE_TRANSFER_DST_BIT,
+        .sharingMode = VK_SHARING_MODE_EXCLUSIVE,
+        .queueFamilyIndexCount = 0,
+        .pQueueFamilyIndices = nullptr,
+    });
+
+    // Commit some device local memory for the buffer.
+    auto commit = m_allocator.Commit(buffer, MemoryUsage::DeviceLocal);
+
+    // Create the descriptor pool to contain our descriptor.
+    constexpr VkDescriptorPoolSize pool_size{
+        .type = VK_DESCRIPTOR_TYPE_STORAGE_BUFFER,
+        .descriptorCount = 1,
+    };
+
+    auto descriptor_pool = dld.CreateDescriptorPool(VkDescriptorPoolCreateInfo{
+        .sType = VK_STRUCTURE_TYPE_DESCRIPTOR_POOL_CREATE_INFO,
+        .pNext = nullptr,
+        .flags = VK_DESCRIPTOR_POOL_CREATE_FREE_DESCRIPTOR_SET_BIT,
+        .maxSets = 1,
+        .poolSizeCount = 1,
+        .pPoolSizes = &pool_size,
+    });
+
+    // Create the descriptor set layout from the pool.
+    constexpr VkDescriptorSetLayoutBinding layout_binding{
+        .binding = 0,
+        .descriptorType = VK_DESCRIPTOR_TYPE_STORAGE_BUFFER,
+        .descriptorCount = 1,
+        .stageFlags = VK_SHADER_STAGE_COMPUTE_BIT,
+        .pImmutableSamplers = nullptr,
+    };
+
+    auto descriptor_set_layout = dld.CreateDescriptorSetLayout(VkDescriptorSetLayoutCreateInfo{
+        .sType = VK_STRUCTURE_TYPE_DESCRIPTOR_SET_LAYOUT_CREATE_INFO,
+        .pNext = nullptr,
+        .flags = 0,
+        .bindingCount = 1,
+        .pBindings = &layout_binding,
+    });
+
+    // Actually create the descriptor set.
+    auto descriptor_set = descriptor_pool.Allocate(VkDescriptorSetAllocateInfo{
+        .sType = VK_STRUCTURE_TYPE_DESCRIPTOR_SET_ALLOCATE_INFO,
+        .pNext = nullptr,
+        .descriptorPool = *descriptor_pool,
+        .descriptorSetCount = 1,
+        .pSetLayouts = descriptor_set_layout.address(),
+    });
+
+    // Create the shader.
+    auto shader = BuildShader(m_device, VULKAN_TURBO_MODE_COMP_SPV);
+
+    // Create the pipeline layout.
+    auto pipeline_layout = dld.CreatePipelineLayout(VkPipelineLayoutCreateInfo{
+        .sType = VK_STRUCTURE_TYPE_PIPELINE_LAYOUT_CREATE_INFO,
+        .pNext = nullptr,
+        .flags = 0,
+        .setLayoutCount = 1,
+        .pSetLayouts = descriptor_set_layout.address(),
+        .pushConstantRangeCount = 0,
+        .pPushConstantRanges = nullptr,
+    });
+
+    // Actually create the pipeline.
+    const VkPipelineShaderStageCreateInfo shader_stage{
+        .sType = VK_STRUCTURE_TYPE_PIPELINE_SHADER_STAGE_CREATE_INFO,
+        .pNext = nullptr,
+        .flags = 0,
+        .stage = VK_SHADER_STAGE_COMPUTE_BIT,
+        .module = *shader,
+        .pName = "main",
+        .pSpecializationInfo = nullptr,
+    };
+
+    auto pipeline = dld.CreateComputePipeline(VkComputePipelineCreateInfo{
+        .sType = VK_STRUCTURE_TYPE_COMPUTE_PIPELINE_CREATE_INFO,
+        .pNext = nullptr,
+        .flags = 0,
+        .stage = shader_stage,
+        .layout = *pipeline_layout,
+        .basePipelineHandle = VK_NULL_HANDLE,
+        .basePipelineIndex = 0,
+    });
+
+    // Create a fence to wait on.
+    auto fence = dld.CreateFence(VkFenceCreateInfo{
+        .sType = VK_STRUCTURE_TYPE_FENCE_CREATE_INFO,
+        .pNext = nullptr,
+        .flags = 0,
+    });
+
+    // Create a command pool to allocate a command buffer from.
+    auto command_pool = dld.CreateCommandPool(VkCommandPoolCreateInfo{
+        .sType = VK_STRUCTURE_TYPE_COMMAND_POOL_CREATE_INFO,
+        .pNext = nullptr,
+        .flags =
+            VK_COMMAND_POOL_CREATE_TRANSIENT_BIT | VK_COMMAND_POOL_CREATE_RESET_COMMAND_BUFFER_BIT,
+        .queueFamilyIndex = m_device.GetGraphicsFamily(),
+    });
+
+    // Create a single command buffer.
+    auto cmdbufs = command_pool.Allocate(1, VK_COMMAND_BUFFER_LEVEL_PRIMARY);
+    auto cmdbuf = vk::CommandBuffer{cmdbufs[0], m_device.GetDispatchLoader()};
+
+    while (!stop_token.stop_requested()) {
+        // Reset the fence.
+        fence.Reset();
+
+        // Update descriptor set.
+        const VkDescriptorBufferInfo buffer_info{
+            .buffer = *buffer,
+            .offset = 0,
+            .range = VK_WHOLE_SIZE,
+        };
+
+        const VkWriteDescriptorSet buffer_write{
+            .sType = VK_STRUCTURE_TYPE_WRITE_DESCRIPTOR_SET,
+            .pNext = nullptr,
+            .dstSet = descriptor_set[0],
+            .dstBinding = 0,
+            .dstArrayElement = 0,
+            .descriptorCount = 1,
+            .descriptorType = VK_DESCRIPTOR_TYPE_STORAGE_BUFFER,
+            .pImageInfo = nullptr,
+            .pBufferInfo = &buffer_info,
+            .pTexelBufferView = nullptr,
+        };
+
+        dld.UpdateDescriptorSets(std::array{buffer_write}, {});
+
+        // Set up the command buffer.
+        cmdbuf.Begin(VkCommandBufferBeginInfo{
+            .sType = VK_STRUCTURE_TYPE_COMMAND_BUFFER_BEGIN_INFO,
+            .pNext = nullptr,
+            .flags = VK_COMMAND_BUFFER_USAGE_ONE_TIME_SUBMIT_BIT,
+            .pInheritanceInfo = nullptr,
+        });
+
+        // Clear the buffer.
+        cmdbuf.FillBuffer(*buffer, 0, VK_WHOLE_SIZE, 0);
+
+        // Bind descriptor set.
+        cmdbuf.BindDescriptorSets(VK_PIPELINE_BIND_POINT_COMPUTE, *pipeline_layout, 0,
+                                  descriptor_set, {});
+
+        // Bind the pipeline.
+        cmdbuf.BindPipeline(VK_PIPELINE_BIND_POINT_COMPUTE, *pipeline);
+
+        // Dispatch.
+        cmdbuf.Dispatch(64, 64, 1);
+
+        // Finish.
+        cmdbuf.End();
+
+        const VkSubmitInfo submit_info{
+            .sType = VK_STRUCTURE_TYPE_SUBMIT_INFO,
+            .pNext = nullptr,
+            .waitSemaphoreCount = 0,
+            .pWaitSemaphores = nullptr,
+            .pWaitDstStageMask = nullptr,
+            .commandBufferCount = 1,
+            .pCommandBuffers = cmdbuf.address(),
+            .signalSemaphoreCount = 0,
+            .pSignalSemaphores = nullptr,
+        };
+
+        m_device.GetGraphicsQueue().Submit(std::array{submit_info}, *fence);
+
+        // Wait for completion.
+        fence.Wait();
+    }
+}
+
+} // namespace Vulkan

src/video_core/renderer_vulkan/vk_turbo_mode.h

@@ -0,0 +1,26 @@
+// SPDX-FileCopyrightText: Copyright 2022 yuzu Emulator Project
+// SPDX-License-Identifier: GPL-2.0-or-later
+
+#pragma once
+
+#include "common/polyfill_thread.h"
+#include "video_core/vulkan_common/vulkan_device.h"
+#include "video_core/vulkan_common/vulkan_memory_allocator.h"
+#include "video_core/vulkan_common/vulkan_wrapper.h"
+
+namespace Vulkan {
+
+class TurboMode {
+public:
+    explicit TurboMode(const vk::Instance& instance, const vk::InstanceDispatch& dld);
+    ~TurboMode();
+
+private:
+    void Run(std::stop_token stop_token);
+
+    Device m_device;
+    MemoryAllocator m_allocator;
+    std::jthread m_thread;
+};
+
+} // namespace Vulkan

src/video_core/texture_cache/texture_cache.h

@@ -148,13 +148,6 @@ typename P::ImageView& TextureCache<P>::GetImageView(ImageViewId id) noexcept {
     return slot_image_views[id];
 }
 
-template <class P>
-typename P::ImageView& TextureCache<P>::GetImageView(u32 index) noexcept {
-    const auto image_view_id = VisitImageView(channel_state->graphics_image_table,
-                                              channel_state->graphics_image_view_ids, index);
-    return slot_image_views[image_view_id];
-}
-
 template <class P>
 void TextureCache<P>::MarkModification(ImageId id) noexcept {
     MarkModification(slot_images[id]);

src/video_core/texture_cache/texture_cache_base.h

@@ -129,9 +129,6 @@ public:
     /// Return a reference to the given image view id
     [[nodiscard]] ImageView& GetImageView(ImageViewId id) noexcept;
 
-    /// Get the imageview from the graphics descriptor table in the specified index
-    [[nodiscard]] ImageView& GetImageView(u32 index) noexcept;
-
     /// Mark an image as modified from the GPU
     void MarkModification(ImageId id) noexcept;
 

src/video_core/vulkan_common/vulkan_device.cpp

@@ -1472,7 +1472,7 @@ std::vector<const char*> Device::LoadExtensions(bool requires_surface) {
         is_patch_list_restart_supported =
             primitive_topology_list_restart.primitiveTopologyPatchListRestart;
     }
-    if (has_khr_image_format_list && has_khr_swapchain_mutable_format) {
+    if (requires_surface && has_khr_image_format_list && has_khr_swapchain_mutable_format) {
         extensions.push_back(VK_KHR_IMAGE_FORMAT_LIST_EXTENSION_NAME);
         extensions.push_back(VK_KHR_SWAPCHAIN_MUTABLE_FORMAT_EXTENSION_NAME);
         khr_swapchain_mutable_format = true;
@@ -1487,6 +1487,9 @@ std::vector<const char*> Device::LoadExtensions(bool requires_surface) {
 
         max_push_descriptors = push_descriptor.maxPushDescriptors;
     }
+
+    has_null_descriptor = true;
+
     return extensions;
 }
 

src/video_core/vulkan_common/vulkan_device.h

@@ -397,6 +397,10 @@ public:
         return must_emulate_bgr565;
     }
 
+    bool HasNullDescriptor() const {
+        return has_null_descriptor;
+    }
+
     u32 GetMaxVertexInputAttributes() const {
         return max_vertex_input_attributes;
     }
@@ -511,6 +515,7 @@ private:
     bool supports_d24_depth{};                   ///< Supports D24 depth buffers.
     bool cant_blit_msaa{};                       ///< Does not support MSAA<->MSAA blitting.
     bool must_emulate_bgr565{};                  ///< Emulates BGR565 by swizzling RGB565 format.
+    bool has_null_descriptor{};                  ///< Has support for null descriptors.
     u32 max_vertex_input_attributes{};           ///< Max vertex input attributes in pipeline
     u32 max_vertex_input_bindings{};             ///< Max vertex input buffers in pipeline
 

src/video_core/vulkan_common/vulkan_instance.cpp

@@ -32,7 +32,7 @@
 namespace Vulkan {
 namespace {
 [[nodiscard]] std::vector<const char*> RequiredExtensions(
-    Core::Frontend::WindowSystemType window_type, bool enable_debug_utils) {
+    Core::Frontend::WindowSystemType window_type, bool enable_validation) {
     std::vector<const char*> extensions;
     extensions.reserve(6);
     switch (window_type) {
@@ -65,7 +65,7 @@ namespace {
     if (window_type != Core::Frontend::WindowSystemType::Headless) {
         extensions.push_back(VK_KHR_SURFACE_EXTENSION_NAME);
     }
-    if (enable_debug_utils) {
+    if (enable_validation) {
         extensions.push_back(VK_EXT_DEBUG_UTILS_EXTENSION_NAME);
     }
     extensions.push_back(VK_KHR_GET_PHYSICAL_DEVICE_PROPERTIES_2_EXTENSION_NAME);
@@ -95,9 +95,9 @@ namespace {
     return true;
 }
 
-[[nodiscard]] std::vector<const char*> Layers(bool enable_layers) {
+[[nodiscard]] std::vector<const char*> Layers(bool enable_validation) {
     std::vector<const char*> layers;
-    if (enable_layers) {
+    if (enable_validation) {
         layers.push_back("VK_LAYER_KHRONOS_validation");
     }
     return layers;
@@ -125,7 +125,7 @@ void RemoveUnavailableLayers(const vk::InstanceDispatch& dld, std::vector<const
 
 vk::Instance CreateInstance(const Common::DynamicLibrary& library, vk::InstanceDispatch& dld,
                             u32 required_version, Core::Frontend::WindowSystemType window_type,
-                            bool enable_debug_utils, bool enable_layers) {
+                            bool enable_validation) {
     if (!library.IsOpen()) {
         LOG_ERROR(Render_Vulkan, "Vulkan library not available");
         throw vk::Exception(VK_ERROR_INITIALIZATION_FAILED);
@@ -138,11 +138,11 @@ vk::Instance CreateInstance(const Common::DynamicLibrary& library, vk::InstanceD
         LOG_ERROR(Render_Vulkan, "Failed to load Vulkan function pointers");
         throw vk::Exception(VK_ERROR_INITIALIZATION_FAILED);
     }
-    const std::vector<const char*> extensions = RequiredExtensions(window_type, enable_debug_utils);
+    const std::vector<const char*> extensions = RequiredExtensions(window_type, enable_validation);
     if (!AreExtensionsSupported(dld, extensions)) {
         throw vk::Exception(VK_ERROR_EXTENSION_NOT_PRESENT);
     }
-    std::vector<const char*> layers = Layers(enable_layers);
+    std::vector<const char*> layers = Layers(enable_validation);
     RemoveUnavailableLayers(dld, layers);
 
     const u32 available_version = vk::AvailableVersion(dld);

src/video_core/vulkan_common/vulkan_instance.h

@@ -17,8 +17,7 @@ namespace Vulkan {
  * @param dld                Dispatch table to load function pointers into
  * @param required_version   Required Vulkan version (for example, VK_API_VERSION_1_1)
  * @param window_type        Window system type's enabled extension
- * @param enable_debug_utils Whether to enable VK_EXT_debug_utils_extension_name or not
- * @param enable_layers      Whether to enable Vulkan validation layers or not
+ * @param enable_validation  Whether to enable Vulkan validation layers or not
  *
  * @return A new Vulkan instance
  * @throw vk::Exception on failure
@@ -26,6 +25,6 @@ namespace Vulkan {
 [[nodiscard]] vk::Instance CreateInstance(
     const Common::DynamicLibrary& library, vk::InstanceDispatch& dld, u32 required_version,
     Core::Frontend::WindowSystemType window_type = Core::Frontend::WindowSystemType::Headless,
-    bool enable_debug_utils = false, bool enable_layers = false);
+    bool enable_validation = false);
 
 } // namespace Vulkan

src/video_core/vulkan_common/vulkan_wrapper.cpp

@@ -152,6 +152,7 @@ void Load(VkDevice device, DeviceDispatch& dld) noexcept {
     X(vkCreateGraphicsPipelines);
     X(vkCreateImage);
     X(vkCreateImageView);
+    X(vkCreatePipelineCache);
     X(vkCreatePipelineLayout);
     X(vkCreateQueryPool);
     X(vkCreateRenderPass);
@@ -171,6 +172,7 @@ void Load(VkDevice device, DeviceDispatch& dld) noexcept {
     X(vkDestroyImage);
     X(vkDestroyImageView);
     X(vkDestroyPipeline);
+    X(vkDestroyPipelineCache);
     X(vkDestroyPipelineLayout);
     X(vkDestroyQueryPool);
     X(vkDestroyRenderPass);
@@ -188,6 +190,7 @@ void Load(VkDevice device, DeviceDispatch& dld) noexcept {
     X(vkGetEventStatus);
     X(vkGetFenceStatus);
     X(vkGetImageMemoryRequirements);
+    X(vkGetPipelineCacheData);
     X(vkGetMemoryFdKHR);
 #ifdef _WIN32
     X(vkGetMemoryWin32HandleKHR);
@@ -431,6 +434,10 @@ void Destroy(VkDevice device, VkPipeline handle, const DeviceDispatch& dld) noex
     dld.vkDestroyPipeline(device, handle, nullptr);
 }
 
+void Destroy(VkDevice device, VkPipelineCache handle, const DeviceDispatch& dld) noexcept {
+    dld.vkDestroyPipelineCache(device, handle, nullptr);
+}
+
 void Destroy(VkDevice device, VkPipelineLayout handle, const DeviceDispatch& dld) noexcept {
     dld.vkDestroyPipelineLayout(device, handle, nullptr);
 }
@@ -651,6 +658,10 @@ void ShaderModule::SetObjectNameEXT(const char* name) const {
     SetObjectName(dld, owner, handle, VK_OBJECT_TYPE_SHADER_MODULE, name);
 }
 
+void PipelineCache::SetObjectNameEXT(const char* name) const {
+    SetObjectName(dld, owner, handle, VK_OBJECT_TYPE_PIPELINE_CACHE, name);
+}
+
 void Semaphore::SetObjectNameEXT(const char* name) const {
     SetObjectName(dld, owner, handle, VK_OBJECT_TYPE_SEMAPHORE, name);
 }
@@ -746,21 +757,29 @@ DescriptorSetLayout Device::CreateDescriptorSetLayout(
     return DescriptorSetLayout(object, handle, *dld);
 }
 
+PipelineCache Device::CreatePipelineCache(const VkPipelineCacheCreateInfo& ci) const {
+    VkPipelineCache cache;
+    Check(dld->vkCreatePipelineCache(handle, &ci, nullptr, &cache));
+    return PipelineCache(cache, handle, *dld);
+}
+
 PipelineLayout Device::CreatePipelineLayout(const VkPipelineLayoutCreateInfo& ci) const {
     VkPipelineLayout object;
     Check(dld->vkCreatePipelineLayout(handle, &ci, nullptr, &object));
     return PipelineLayout(object, handle, *dld);
 }
 
-Pipeline Device::CreateGraphicsPipeline(const VkGraphicsPipelineCreateInfo& ci) const {
+Pipeline Device::CreateGraphicsPipeline(const VkGraphicsPipelineCreateInfo& ci,
+                                        VkPipelineCache cache) const {
     VkPipeline object;
-    Check(dld->vkCreateGraphicsPipelines(handle, nullptr, 1, &ci, nullptr, &object));
+    Check(dld->vkCreateGraphicsPipelines(handle, cache, 1, &ci, nullptr, &object));
     return Pipeline(object, handle, *dld);
 }
 
-Pipeline Device::CreateComputePipeline(const VkComputePipelineCreateInfo& ci) const {
+Pipeline Device::CreateComputePipeline(const VkComputePipelineCreateInfo& ci,
+                                       VkPipelineCache cache) const {
     VkPipeline object;
-    Check(dld->vkCreateComputePipelines(handle, nullptr, 1, &ci, nullptr, &object));
+    Check(dld->vkCreateComputePipelines(handle, cache, 1, &ci, nullptr, &object));
     return Pipeline(object, handle, *dld);
 }
 

src/video_core/vulkan_common/vulkan_wrapper.h

@@ -270,6 +270,7 @@ struct DeviceDispatch : InstanceDispatch {
     PFN_vkCreateGraphicsPipelines vkCreateGraphicsPipelines{};
     PFN_vkCreateImage vkCreateImage{};
     PFN_vkCreateImageView vkCreateImageView{};
+    PFN_vkCreatePipelineCache vkCreatePipelineCache{};
     PFN_vkCreatePipelineLayout vkCreatePipelineLayout{};
     PFN_vkCreateQueryPool vkCreateQueryPool{};
     PFN_vkCreateRenderPass vkCreateRenderPass{};
@@ -289,6 +290,7 @@ struct DeviceDispatch : InstanceDispatch {
     PFN_vkDestroyImage vkDestroyImage{};
     PFN_vkDestroyImageView vkDestroyImageView{};
     PFN_vkDestroyPipeline vkDestroyPipeline{};
+    PFN_vkDestroyPipelineCache vkDestroyPipelineCache{};
     PFN_vkDestroyPipelineLayout vkDestroyPipelineLayout{};
     PFN_vkDestroyQueryPool vkDestroyQueryPool{};
     PFN_vkDestroyRenderPass vkDestroyRenderPass{};
@@ -306,6 +308,7 @@ struct DeviceDispatch : InstanceDispatch {
     PFN_vkGetEventStatus vkGetEventStatus{};
     PFN_vkGetFenceStatus vkGetFenceStatus{};
     PFN_vkGetImageMemoryRequirements vkGetImageMemoryRequirements{};
+    PFN_vkGetPipelineCacheData vkGetPipelineCacheData{};
     PFN_vkGetMemoryFdKHR vkGetMemoryFdKHR{};
 #ifdef _WIN32
     PFN_vkGetMemoryWin32HandleKHR vkGetMemoryWin32HandleKHR{};
@@ -351,6 +354,7 @@ void Destroy(VkDevice, VkFramebuffer, const DeviceDispatch&) noexcept;
 void Destroy(VkDevice, VkImage, const DeviceDispatch&) noexcept;
 void Destroy(VkDevice, VkImageView, const DeviceDispatch&) noexcept;
 void Destroy(VkDevice, VkPipeline, const DeviceDispatch&) noexcept;
+void Destroy(VkDevice, VkPipelineCache, const DeviceDispatch&) noexcept;
 void Destroy(VkDevice, VkPipelineLayout, const DeviceDispatch&) noexcept;
 void Destroy(VkDevice, VkQueryPool, const DeviceDispatch&) noexcept;
 void Destroy(VkDevice, VkRenderPass, const DeviceDispatch&) noexcept;
@@ -773,6 +777,18 @@ public:
     void SetObjectNameEXT(const char* name) const;
 };
 
+class PipelineCache : public Handle<VkPipelineCache, VkDevice, DeviceDispatch> {
+    using Handle<VkPipelineCache, VkDevice, DeviceDispatch>::Handle;
+
+public:
+    /// Set object name.
+    void SetObjectNameEXT(const char* name) const;
+
+    VkResult Read(size_t* size, void* data) const noexcept {
+        return dld->vkGetPipelineCacheData(owner, handle, size, data);
+    }
+};
+
 class Semaphore : public Handle<VkSemaphore, VkDevice, DeviceDispatch> {
     using Handle<VkSemaphore, VkDevice, DeviceDispatch>::Handle;
 
@@ -844,11 +860,15 @@ public:
 
     DescriptorSetLayout CreateDescriptorSetLayout(const VkDescriptorSetLayoutCreateInfo& ci) const;
 
+    PipelineCache CreatePipelineCache(const VkPipelineCacheCreateInfo& ci) const;
+
     PipelineLayout CreatePipelineLayout(const VkPipelineLayoutCreateInfo& ci) const;
 
-    Pipeline CreateGraphicsPipeline(const VkGraphicsPipelineCreateInfo& ci) const;
+    Pipeline CreateGraphicsPipeline(const VkGraphicsPipelineCreateInfo& ci,
+                                    VkPipelineCache cache = nullptr) const;
 
-    Pipeline CreateComputePipeline(const VkComputePipelineCreateInfo& ci) const;
+    Pipeline CreateComputePipeline(const VkComputePipelineCreateInfo& ci,
+                                   VkPipelineCache cache = nullptr) const;
 
     Sampler CreateSampler(const VkSamplerCreateInfo& ci) const;
 

src/yuzu/configuration/config.cpp

@@ -690,6 +690,7 @@ void Config::ReadRendererValues() {
     qt_config->beginGroup(QStringLiteral("Renderer"));
 
     ReadGlobalSetting(Settings::values.renderer_backend);
+    ReadGlobalSetting(Settings::values.renderer_force_max_clock);
     ReadGlobalSetting(Settings::values.vulkan_device);
     ReadGlobalSetting(Settings::values.fullscreen_mode);
     ReadGlobalSetting(Settings::values.aspect_ratio);
@@ -709,6 +710,7 @@ void Config::ReadRendererValues() {
     ReadGlobalSetting(Settings::values.use_asynchronous_shaders);
     ReadGlobalSetting(Settings::values.use_fast_gpu_time);
     ReadGlobalSetting(Settings::values.use_pessimistic_flushes);
+    ReadGlobalSetting(Settings::values.use_vulkan_driver_pipeline_cache);
     ReadGlobalSetting(Settings::values.bg_red);
     ReadGlobalSetting(Settings::values.bg_green);
     ReadGlobalSetting(Settings::values.bg_blue);
@@ -1305,6 +1307,9 @@ void Config::SaveRendererValues() {
                  static_cast<u32>(Settings::values.renderer_backend.GetValue(global)),
                  static_cast<u32>(Settings::values.renderer_backend.GetDefault()),
                  Settings::values.renderer_backend.UsingGlobal());
+    WriteSetting(QString::fromStdString(Settings::values.renderer_force_max_clock.GetLabel()),
+                 static_cast<u32>(Settings::values.renderer_force_max_clock.GetValue(global)),
+                 static_cast<u32>(Settings::values.renderer_force_max_clock.GetDefault()));
     WriteGlobalSetting(Settings::values.vulkan_device);
     WriteSetting(QString::fromStdString(Settings::values.fullscreen_mode.GetLabel()),
                  static_cast<u32>(Settings::values.fullscreen_mode.GetValue(global)),
@@ -1348,6 +1353,7 @@ void Config::SaveRendererValues() {
     WriteGlobalSetting(Settings::values.use_asynchronous_shaders);
     WriteGlobalSetting(Settings::values.use_fast_gpu_time);
     WriteGlobalSetting(Settings::values.use_pessimistic_flushes);
+    WriteGlobalSetting(Settings::values.use_vulkan_driver_pipeline_cache);
     WriteGlobalSetting(Settings::values.bg_red);
     WriteGlobalSetting(Settings::values.bg_green);
     WriteGlobalSetting(Settings::values.bg_blue);

src/yuzu/configuration/configure_graphics_advanced.cpp

@@ -25,10 +25,13 @@ void ConfigureGraphicsAdvanced::SetConfiguration() {
     ui->use_asynchronous_shaders->setEnabled(runtime_lock);
     ui->anisotropic_filtering_combobox->setEnabled(runtime_lock);
 
+    ui->renderer_force_max_clock->setChecked(Settings::values.renderer_force_max_clock.GetValue());
     ui->use_vsync->setChecked(Settings::values.use_vsync.GetValue());
     ui->use_asynchronous_shaders->setChecked(Settings::values.use_asynchronous_shaders.GetValue());
     ui->use_fast_gpu_time->setChecked(Settings::values.use_fast_gpu_time.GetValue());
     ui->use_pessimistic_flushes->setChecked(Settings::values.use_pessimistic_flushes.GetValue());
+    ui->use_vulkan_driver_pipeline_cache->setChecked(
+        Settings::values.use_vulkan_driver_pipeline_cache.GetValue());
 
     if (Settings::IsConfiguringGlobal()) {
         ui->gpu_accuracy->setCurrentIndex(
@@ -37,6 +40,8 @@ void ConfigureGraphicsAdvanced::SetConfiguration() {
             Settings::values.max_anisotropy.GetValue());
     } else {
         ConfigurationShared::SetPerGameSetting(ui->gpu_accuracy, &Settings::values.gpu_accuracy);
+        ConfigurationShared::SetPerGameSetting(ui->renderer_force_max_clock,
+                                               &Settings::values.renderer_force_max_clock);
         ConfigurationShared::SetPerGameSetting(ui->anisotropic_filtering_combobox,
                                                &Settings::values.max_anisotropy);
         ConfigurationShared::SetHighlight(ui->label_gpu_accuracy,
@@ -48,6 +53,9 @@ void ConfigureGraphicsAdvanced::SetConfiguration() {
 
 void ConfigureGraphicsAdvanced::ApplyConfiguration() {
     ConfigurationShared::ApplyPerGameSetting(&Settings::values.gpu_accuracy, ui->gpu_accuracy);
+    ConfigurationShared::ApplyPerGameSetting(&Settings::values.renderer_force_max_clock,
+                                             ui->renderer_force_max_clock,
+                                             renderer_force_max_clock);
     ConfigurationShared::ApplyPerGameSetting(&Settings::values.max_anisotropy,
                                              ui->anisotropic_filtering_combobox);
     ConfigurationShared::ApplyPerGameSetting(&Settings::values.use_vsync, ui->use_vsync, use_vsync);
@@ -58,6 +66,9 @@ void ConfigureGraphicsAdvanced::ApplyConfiguration() {
                                              ui->use_fast_gpu_time, use_fast_gpu_time);
     ConfigurationShared::ApplyPerGameSetting(&Settings::values.use_pessimistic_flushes,
                                              ui->use_pessimistic_flushes, use_pessimistic_flushes);
+    ConfigurationShared::ApplyPerGameSetting(&Settings::values.use_vulkan_driver_pipeline_cache,
+                                             ui->use_vulkan_driver_pipeline_cache,
+                                             use_vulkan_driver_pipeline_cache);
 }
 
 void ConfigureGraphicsAdvanced::changeEvent(QEvent* event) {
@@ -76,18 +87,25 @@ void ConfigureGraphicsAdvanced::SetupPerGameUI() {
     // Disable if not global (only happens during game)
     if (Settings::IsConfiguringGlobal()) {
         ui->gpu_accuracy->setEnabled(Settings::values.gpu_accuracy.UsingGlobal());
+        ui->renderer_force_max_clock->setEnabled(
+            Settings::values.renderer_force_max_clock.UsingGlobal());
         ui->use_vsync->setEnabled(Settings::values.use_vsync.UsingGlobal());
         ui->use_asynchronous_shaders->setEnabled(
             Settings::values.use_asynchronous_shaders.UsingGlobal());
         ui->use_fast_gpu_time->setEnabled(Settings::values.use_fast_gpu_time.UsingGlobal());
         ui->use_pessimistic_flushes->setEnabled(
             Settings::values.use_pessimistic_flushes.UsingGlobal());
+        ui->use_vulkan_driver_pipeline_cache->setEnabled(
+            Settings::values.use_vulkan_driver_pipeline_cache.UsingGlobal());
         ui->anisotropic_filtering_combobox->setEnabled(
             Settings::values.max_anisotropy.UsingGlobal());
 
         return;
     }
 
+    ConfigurationShared::SetColoredTristate(ui->renderer_force_max_clock,
+                                            Settings::values.renderer_force_max_clock,
+                                            renderer_force_max_clock);
     ConfigurationShared::SetColoredTristate(ui->use_vsync, Settings::values.use_vsync, use_vsync);
     ConfigurationShared::SetColoredTristate(ui->use_asynchronous_shaders,
                                             Settings::values.use_asynchronous_shaders,
@@ -97,6 +115,9 @@ void ConfigureGraphicsAdvanced::SetupPerGameUI() {
     ConfigurationShared::SetColoredTristate(ui->use_pessimistic_flushes,
                                             Settings::values.use_pessimistic_flushes,
                                             use_pessimistic_flushes);
+    ConfigurationShared::SetColoredTristate(ui->use_vulkan_driver_pipeline_cache,
+                                            Settings::values.use_vulkan_driver_pipeline_cache,
+                                            use_vulkan_driver_pipeline_cache);
     ConfigurationShared::SetColoredComboBox(
         ui->gpu_accuracy, ui->label_gpu_accuracy,
         static_cast<int>(Settings::values.gpu_accuracy.GetValue(true)));

src/yuzu/configuration/configure_graphics_advanced.h

@@ -36,10 +36,12 @@ private:
 
     std::unique_ptr<Ui::ConfigureGraphicsAdvanced> ui;
 
+    ConfigurationShared::CheckState renderer_force_max_clock;
     ConfigurationShared::CheckState use_vsync;
     ConfigurationShared::CheckState use_asynchronous_shaders;
     ConfigurationShared::CheckState use_fast_gpu_time;
     ConfigurationShared::CheckState use_pessimistic_flushes;
+    ConfigurationShared::CheckState use_vulkan_driver_pipeline_cache;
 
     const Core::System& system;
 };

src/yuzu/configuration/configure_graphics_advanced.ui

@@ -69,6 +69,16 @@
           </layout>
          </widget>
         </item>
+        <item>
+         <widget class="QCheckBox" name="renderer_force_max_clock">
+          <property name="toolTip">
+           <string>Runs work in the background while waiting for graphics commands to keep the GPU from lowering its clock speed.</string>
+          </property>
+          <property name="text">
+           <string>Force maximum clocks (Vulkan only)</string>
+          </property>
+         </widget>
+        </item>
         <item>
          <widget class="QCheckBox" name="use_vsync">
           <property name="toolTip">
@@ -109,6 +119,16 @@
           </property>
          </widget>
         </item>
+        <item>
+         <widget class="QCheckBox" name="use_vulkan_driver_pipeline_cache">
+          <property name="toolTip">
+            <string>Enables GPU vendor-specific pipeline cache. This option can improve shader loading time significantly in cases where the Vulkan driver does not store pipeline cache files internally.</string>
+          </property>
+          <property name="text">
+           <string>Use Vulkan pipeline cache</string>
+          </property>
+         </widget>
+        </item>
         <item>
          <widget class="QWidget" name="af_layout" native="true">
           <layout class="QHBoxLayout" name="horizontalLayout_1">

src/yuzu/main.cpp

@@ -2229,8 +2229,10 @@ void GMainWindow::OnGameListRemoveFile(u64 program_id, GameListRemoveTarget targ
     }
 
     switch (target) {
-    case GameListRemoveTarget::GlShaderCache:
     case GameListRemoveTarget::VkShaderCache:
+        RemoveVulkanDriverPipelineCache(program_id);
+        [[fallthrough]];
+    case GameListRemoveTarget::GlShaderCache:
         RemoveTransferableShaderCache(program_id, target);
         break;
     case GameListRemoveTarget::AllShaderCache:
@@ -2271,6 +2273,22 @@ void GMainWindow::RemoveTransferableShaderCache(u64 program_id, GameListRemoveTa
     }
 }
 
+void GMainWindow::RemoveVulkanDriverPipelineCache(u64 program_id) {
+    static constexpr std::string_view target_file_name = "vulkan_pipelines.bin";
+
+    const auto shader_cache_dir = Common::FS::GetYuzuPath(Common::FS::YuzuPath::ShaderDir);
+    const auto shader_cache_folder_path = shader_cache_dir / fmt::format("{:016x}", program_id);
+    const auto target_file = shader_cache_folder_path / target_file_name;
+
+    if (!Common::FS::Exists(target_file)) {
+        return;
+    }
+    if (!Common::FS::RemoveFile(target_file)) {
+        QMessageBox::warning(this, tr("Error Removing Vulkan Driver Pipeline Cache"),
+                             tr("Failed to remove the driver pipeline cache."));
+    }
+}
+
 void GMainWindow::RemoveAllTransferableShaderCaches(u64 program_id) {
     const auto shader_cache_dir = Common::FS::GetYuzuPath(Common::FS::YuzuPath::ShaderDir);
     const auto program_shader_cache_dir = shader_cache_dir / fmt::format("{:016x}", program_id);

src/yuzu/main.h

@@ -347,6 +347,7 @@ private:
     void RemoveUpdateContent(u64 program_id, InstalledEntryType type);
     void RemoveAddOnContent(u64 program_id, InstalledEntryType type);
     void RemoveTransferableShaderCache(u64 program_id, GameListRemoveTarget target);
+    void RemoveVulkanDriverPipelineCache(u64 program_id);
     void RemoveAllTransferableShaderCaches(u64 program_id);
     void RemoveCustomConfiguration(u64 program_id, const std::string& game_path);
     std::optional<u64> SelectRomFSDumpTarget(const FileSys::ContentProvider&, u64 program_id);

src/yuzu_cmd/config.cpp

@@ -296,6 +296,7 @@ void Config::ReadValues() {
 
     // Renderer
     ReadSetting("Renderer", Settings::values.renderer_backend);
+    ReadSetting("Renderer", Settings::values.renderer_force_max_clock);
     ReadSetting("Renderer", Settings::values.renderer_debug);
     ReadSetting("Renderer", Settings::values.renderer_shader_feedback);
     ReadSetting("Renderer", Settings::values.enable_nsight_aftermath);
@@ -321,6 +322,7 @@ void Config::ReadValues() {
     ReadSetting("Renderer", Settings::values.accelerate_astc);
     ReadSetting("Renderer", Settings::values.use_fast_gpu_time);
     ReadSetting("Renderer", Settings::values.use_pessimistic_flushes);
+    ReadSetting("Renderer", Settings::values.use_vulkan_driver_pipeline_cache);
 
     ReadSetting("Renderer", Settings::values.bg_red);
     ReadSetting("Renderer", Settings::values.bg_green);