emit_spirv_image: Fix depth image implicit lod sample in compute

Ensures all drivers behave the same way in this case.

src/shader_recompiler/backend/spirv/emit_spirv_image.cpp

@@ -355,11 +355,22 @@ Id EmitImageSampleExplicitLod(EmitContext& ctx, IR::Inst* inst, const IR::Value&
 Id EmitImageSampleDrefImplicitLod(EmitContext& ctx, IR::Inst* inst, const IR::Value& index,
                                   Id coords, Id dref, Id bias_lc, const IR::Value& offset) {
     const auto info{inst->Flags<IR::TextureInstInfo>()};
-    const ImageOperands operands(ctx, info.has_bias != 0, false, info.has_lod_clamp != 0, bias_lc,
-                                 offset);
-    return Emit(&EmitContext::OpImageSparseSampleDrefImplicitLod,
-                &EmitContext::OpImageSampleDrefImplicitLod, ctx, inst, ctx.F32[1],
-                Texture(ctx, info, index), coords, dref, operands.MaskOptional(), operands.Span());
+    if (ctx.stage == Stage::Fragment) {
+        const ImageOperands operands(ctx, info.has_bias != 0, false, info.has_lod_clamp != 0,
+                                     bias_lc, offset);
+        return Emit(&EmitContext::OpImageSparseSampleDrefImplicitLod,
+                    &EmitContext::OpImageSampleDrefImplicitLod, ctx, inst, ctx.F32[1],
+                    Texture(ctx, info, index), coords, dref, operands.MaskOptional(),
+                    operands.Span());
+    } else {
+        // Implicit lods in compute behave on hardware as if sampling from LOD 0.
+        // This check is to ensure all drivers behave this way.
+        const Id lod{ctx.Const(0.0f)};
+        const ImageOperands operands(ctx, false, true, false, lod, offset);
+        return Emit(&EmitContext::OpImageSparseSampleDrefExplicitLod,
+                    &EmitContext::OpImageSampleDrefExplicitLod, ctx, inst, ctx.F32[1],
+                    Texture(ctx, info, index), coords, dref, operands.Mask(), operands.Span());
+    }
 }
 
 Id EmitImageSampleDrefExplicitLod(EmitContext& ctx, IR::Inst* inst, const IR::Value& index,

src/shader_recompiler/ir_opt/texture_pass.cpp

@@ -492,7 +492,8 @@ void TexturePass(Environment& env, IR::Program& program) {
             const auto insert_point{IR::Block::InstructionList::s_iterator_to(*inst)};
             IR::IREmitter ir{*texture_inst.block, insert_point};
             const IR::U32 shift{ir.Imm32(std::countr_zero(DESCRIPTOR_SIZE))};
-            inst->SetArg(0, ir.ShiftRightArithmetic(cbuf.dynamic_offset, shift));
+            inst->SetArg(0, ir.SMin(ir.ShiftRightArithmetic(cbuf.dynamic_offset, shift),
+                                    ir.Imm32(DESCRIPTOR_SIZE - 1)));
         } else {
             inst->SetArg(0, IR::Value{});
         }

src/video_core/query_cache.h

@@ -258,9 +258,9 @@ private:
 
     void AsyncFlushQuery(VAddr addr) {
         if (!uncommitted_flushes) {
-            uncommitted_flushes = std::make_shared<std::unordered_set<VAddr>>();
+            uncommitted_flushes = std::make_shared<std::vector<VAddr>>();
         }
-        uncommitted_flushes->insert(addr);
+        uncommitted_flushes->push_back(addr);
     }
 
     static constexpr std::uintptr_t PAGE_SIZE = 4096;
@@ -276,8 +276,8 @@ private:
 
     std::array<CounterStream, VideoCore::NumQueryTypes> streams;
 
-    std::shared_ptr<std::unordered_set<VAddr>> uncommitted_flushes{};
-    std::list<std::shared_ptr<std::unordered_set<VAddr>>> committed_flushes;
+    std::shared_ptr<std::vector<VAddr>> uncommitted_flushes{};
+    std::list<std::shared_ptr<std::vector<VAddr>>> committed_flushes;
 };
 
 template <class QueryCache, class HostCounter>

src/video_core/rasterizer_accelerated.h

@@ -42,7 +42,7 @@ private:
     };
     static_assert(sizeof(CacheEntry) == 8, "CacheEntry should be 8 bytes!");
 
-    std::array<CacheEntry, 0x1000000> cached_pages;
+    std::array<CacheEntry, 0x2000000> cached_pages;
     Core::Memory::Memory& cpu_memory;
 };
 

src/video_core/renderer_vulkan/vk_master_semaphore.h

@@ -21,12 +21,12 @@ public:
 
     /// Returns the current logical tick.
     [[nodiscard]] u64 CurrentTick() const noexcept {
-        return current_tick.load(std::memory_order_relaxed);
+        return current_tick.load(std::memory_order_acquire);
     }
 
     /// Returns the last known GPU tick.
     [[nodiscard]] u64 KnownGpuTick() const noexcept {
-        return gpu_tick.load(std::memory_order_relaxed);
+        return gpu_tick.load(std::memory_order_acquire);
     }
 
     /// Returns the timeline semaphore handle.
@@ -41,12 +41,21 @@ public:
 
     /// Advance to the logical tick and return the old one
     [[nodiscard]] u64 NextTick() noexcept {
-        return current_tick.fetch_add(1, std::memory_order::relaxed);
+        return current_tick.fetch_add(1, std::memory_order_release);
     }
 
     /// Refresh the known GPU tick
     void Refresh() {
-        gpu_tick.store(semaphore.GetCounter(), std::memory_order_relaxed);
+        u64 this_tick{};
+        u64 counter{};
+        do {
+            this_tick = gpu_tick.load(std::memory_order_acquire);
+            counter = semaphore.GetCounter();
+            if (counter < this_tick) {
+                return;
+            }
+        } while (!gpu_tick.compare_exchange_weak(this_tick, counter, std::memory_order_release,
+                                                 std::memory_order_relaxed));
     }
 
     /// Waits for a tick to be hit on the GPU

src/video_core/renderer_vulkan/vk_query_cache.cpp

@@ -117,7 +117,8 @@ u64 HostCounter::BlockingQuery() const {
     cache.GetScheduler().Wait(tick);
     u64 data;
     const VkResult query_result = cache.GetDevice().GetLogical().GetQueryResults(
-        query.first, query.second, 1, sizeof(data), &data, sizeof(data), VK_QUERY_RESULT_64_BIT);
+        query.first, query.second, 1, sizeof(data), &data, sizeof(data),
+        VK_QUERY_RESULT_64_BIT | VK_QUERY_RESULT_WAIT_BIT);
 
     switch (query_result) {
     case VK_SUCCESS:

src/yuzu/main.cpp

@@ -1666,7 +1666,7 @@ void GMainWindow::OnTransferableShaderCacheOpenFile(u64 program_id) {
     const auto shader_cache_folder_path{shader_cache_dir / fmt::format("{:016x}", program_id)};
     if (!Common::FS::CreateDirs(shader_cache_folder_path)) {
         QMessageBox::warning(this, tr("Error Opening Transferable Shader Cache"),
-                             tr("Filed to create the shader cache directory for this title."));
+                             tr("Failed to create the shader cache directory for this title."));
         return;
     }
     const auto shader_path_string{Common::FS::PathToUTF8String(shader_cache_folder_path)};