early-access version 1358

src/video_core/buffer_cache/buffer_cache.h

@@ -1243,9 +1243,15 @@ typename BufferCache<P>::Binding BufferCache<P>::StorageBufferBinding(GPUVAddr s
     if (!cpu_addr || size == 0) {
         return NULL_BINDING;
     }
+    // HACK(Rodrigo): This is the number of bytes bound in host beyond the guest API's range.
+    // It exists due to some games like Astral Chain operate out of bounds.
+    // Binding the whole map range would be technically correct, but games have large maps that make
+    // this approach unaffordable for now.
+    static constexpr u32 arbitrary_extra_bytes = 0xc000;
+    const u32 bytes_to_map_end = static_cast<u32>(gpu_memory.BytesToMapEnd(gpu_addr));
     const Binding binding{
         .cpu_addr = *cpu_addr,
-        .size = size,
+        .size = std::min(size + arbitrary_extra_bytes, bytes_to_map_end),
         .buffer_id = BufferId{},
     };
     return binding;

src/video_core/memory_manager.cpp

@@ -4,6 +4,7 @@
 
 #include "common/alignment.h"
 #include "common/assert.h"
+#include "common/logging/log.h"
 #include "core/core.h"
 #include "core/hle/kernel/memory/page_table.h"
 #include "core/hle/kernel/process.h"
@@ -38,6 +39,12 @@ GPUVAddr MemoryManager::UpdateRange(GPUVAddr gpu_addr, PageEntry page_entry, std
 }
 
 GPUVAddr MemoryManager::Map(VAddr cpu_addr, GPUVAddr gpu_addr, std::size_t size) {
+    const auto it = std::ranges::lower_bound(map_ranges, gpu_addr, {}, &MapRange::first);
+    if (it != map_ranges.end() && it->first == gpu_addr) {
+        it->second = size;
+    } else {
+        map_ranges.insert(it, MapRange{gpu_addr, size});
+    }
     return UpdateRange(gpu_addr, cpu_addr, size);
 }
 
@@ -52,10 +59,16 @@ GPUVAddr MemoryManager::MapAllocate32(VAddr cpu_addr, std::size_t size) {
 }
 
 void MemoryManager::Unmap(GPUVAddr gpu_addr, std::size_t size) {
-    if (!size) {
+    if (size == 0) {
         return;
     }
-
+    const auto it = std::ranges::lower_bound(map_ranges, gpu_addr, {}, &MapRange::first);
+    if (it != map_ranges.end()) {
+        ASSERT(it->first == gpu_addr);
+        map_ranges.erase(it);
+    } else {
+        UNREACHABLE_MSG("Unmapping non-existent GPU address=0x{:x}", gpu_addr);
+    }
     // Flush and invalidate through the GPU interface, to be asynchronous if possible.
     const std::optional<VAddr> cpu_addr = GpuToCpuAddress(gpu_addr);
     ASSERT(cpu_addr);
@@ -237,6 +250,12 @@ const u8* MemoryManager::GetPointer(GPUVAddr gpu_addr) const {
     return system.Memory().GetPointer(*address);
 }
 
+size_t MemoryManager::BytesToMapEnd(GPUVAddr gpu_addr) const noexcept {
+    auto it = std::ranges::upper_bound(map_ranges, gpu_addr, {}, &MapRange::first);
+    --it;
+    return it->second - (gpu_addr - it->first);
+}
+
 void MemoryManager::ReadBlock(GPUVAddr gpu_src_addr, void* dest_buffer, std::size_t size) const {
     std::size_t remaining_size{size};
     std::size_t page_index{gpu_src_addr >> page_bits};

src/video_core/memory_manager.h

@@ -85,6 +85,9 @@ public:
     [[nodiscard]] u8* GetPointer(GPUVAddr addr);
     [[nodiscard]] const u8* GetPointer(GPUVAddr addr) const;
 
+    /// Returns the number of bytes until the end of the memory map containing the given GPU address
+    [[nodiscard]] size_t BytesToMapEnd(GPUVAddr gpu_addr) const noexcept;
+
     /**
      * ReadBlock and WriteBlock are full read and write operations over virtual
      * GPU Memory. It's important to use these when GPU memory may not be continuous
@@ -160,6 +163,9 @@ private:
 
     std::vector<PageEntry> page_table;
     std::vector<std::pair<VAddr, std::size_t>> cache_invalidate_queue;
+
+    using MapRange = std::pair<GPUVAddr, size_t>;
+    std::vector<MapRange> map_ranges;
 };
 
 } // namespace Tegra

src/video_core/shader/node.h

@@ -465,6 +465,14 @@ public:
         return operands.size();
     }
 
+    NodeBlock& GetOperands() {
+        return operands;
+    }
+
+    const NodeBlock& GetOperands() const {
+        return operands;
+    }
+
     [[nodiscard]] const Node& operator[](std::size_t operand_index) const {
         return operands.at(operand_index);
     }

src/video_core/shader/shader_ir.cpp

@@ -388,9 +388,54 @@ void ShaderIR::SetInternalFlagsFromInteger(NodeBlock& bb, Node value, bool sets_
     if (!sets_cc) {
         return;
     }
-    Node zerop = Operation(OperationCode::LogicalIEqual, std::move(value), Immediate(0));
-    SetInternalFlag(bb, InternalFlag::Zero, std::move(zerop));
-    LOG_WARNING(HW_GPU, "Condition codes implementation is incomplete");
+    switch (value->index()) {
+    case 0: // Operation Node
+        SearchOperands(bb, value);
+        break;
+    case 2: // Genral Purpose Node
+        if (const auto* gpr = std::get_if<GprNode>(value.get())) {
+            LOG_DEBUG(HW_GPU, "GprNode: index={}", gpr->GetIndex());
+            Node zerop = Operation(OperationCode::LogicalIEqual, std::move(value),
+                                   Immediate(gpr->GetIndex()));
+            SetInternalFlag(bb, InternalFlag::Zero, std::move(zerop));
+        }
+        break;
+
+    default:
+        Node zerop = Operation(OperationCode::LogicalIEqual, std::move(value), Immediate(0));
+        SetInternalFlag(bb, InternalFlag::Zero, std::move(zerop));
+        LOG_WARNING(HW_GPU, "Node Type: {}", value->index());
+        break;
+    }
+}
+
+void ShaderIR::SearchOperands(NodeBlock& nb, Node var) {
+    const auto* op = std::get_if<OperationNode>(var.get());
+    if (op == nullptr) {
+        return;
+    }
+
+    if (op->GetOperandsCount() == 0) {
+        return;
+    }
+
+    for (auto& operand : op->GetOperands()) {
+        switch (operand->index()) {
+        case 0: // Operation Node
+            return SearchOperands(nb, operand);
+        case 2: // General Purpose Node
+            if (const auto* gpr = std::get_if<GprNode>(operand.get())) {
+                LOG_DEBUG(HW_GPU, "Child GprNode: index={}", gpr->GetIndex());
+                Node zerop = Operation(OperationCode::LogicalIEqual, std::move(operand),
+                                       Immediate(gpr->GetIndex()));
+                SetInternalFlag(nb, InternalFlag::Zero, std::move(zerop));
+            }
+            break;
+        default:
+            LOG_WARNING(HW_GPU, "Child Node Type: {}", operand->index());
+            break;
+        }
+    }
 }
 
 Node ShaderIR::BitfieldExtract(Node value, u32 offset, u32 bits) {

src/video_core/shader/shader_ir.h

@@ -350,6 +350,9 @@ private:
     /// Access a bindless image sampler.
     ImageEntry& GetBindlessImage(Tegra::Shader::Register reg, Tegra::Shader::ImageType type);
 
+    /// Recursive Iteration over the OperationNode operands, searching for GprNodes.
+    void SearchOperands(NodeBlock& nb, Node var);
+
     /// Extracts a sequence of bits from a node
     Node BitfieldExtract(Node value, u32 offset, u32 bits);