Merge remote-tracking branch 'origin/nx2' into fe-fsr31

# Conflicts:
#	src/video_core/renderer_vulkan/blit_image.cpp
#	src/video_core/renderer_vulkan/blit_image.h

3 files changed, 15 insertions, 7 deletions

diff --git a/src/video_core/texture_cache/descriptor_table.h b/src/video_core/texture_cache/descriptor_table.h
index 75383809c5..a441e572de 100644
--- a/src/video_core/texture_cache/descriptor_table.h
+++ b/src/video_core/texture_cache/descriptor_table.h
@@ -35,10 +35,11 @@ public:
         invalidated = true;
     }
 
-    [[nodiscard]] std::pair<T, bool> Read(Tegra::MemoryManager const& gpu_memory, u32 index) noexcept {
+    template <class MemoryManager>
+    [[nodiscard]] std::pair<T, bool> Read(const MemoryManager& gpu_memory, u32 index) noexcept {
         DEBUG_ASSERT(index <= current_limit);
-        if (invalidated) [[unlikely]] {
-            std::ranges::fill(read_descriptors, 0);
+        if (invalidated) {
+            ClearReadDescriptors();
             invalidated = false;
         }
         const GPUVAddr gpu_addr = current_gpu_addr + index * sizeof(T);
@@ -62,10 +63,8 @@ public:
         // Mario Brothership reallocates a lot of times, so use aggressive pre-alloc sizes
         // std::vector<T> by default uses quadratic growth, but that isn't even enough to satisfy brothership
         const size_t num_descriptors = ((limit + 0x80000) & (~0x7ffff)) + 1;
-        size_t old_size = read_descriptors.size();
         read_descriptors.resize(Common::DivCeil(num_descriptors, 64U));
-        old_size = (std::min)(old_size, read_descriptors.size());
-        std::fill(read_descriptors.begin(), read_descriptors.begin() + old_size, 0);
+        ClearReadDescriptors();
         invalidated = false;
         //
         descriptors.resize(num_descriptors);
@@ -76,6 +75,13 @@ public:
     GPUVAddr current_gpu_addr{};
     u32 current_limit{};
     bool invalidated{}; ///< read_descriptors is to be cleared before the next read
+
+private:
+    void ClearReadDescriptors() noexcept {
+        // Reserved capacity is not addressable. A later limit increase clears its new range.
+        const size_t words = (std::min)(read_descriptors.size(), size_t{current_limit} / 64 + 1);
+        std::fill_n(read_descriptors.begin(), words, u64{0});
+    }
 };
 
 } // namespace VideoCommon
diff --git a/src/video_core/texture_cache/format_lookup_table.cpp b/src/video_core/texture_cache/format_lookup_table.cpp
index 20bb96aae3..18be2dcbf3 100644
--- a/src/video_core/texture_cache/format_lookup_table.cpp
+++ b/src/video_core/texture_cache/format_lookup_table.cpp
@@ -150,6 +150,7 @@ PixelFormat PixelFormatFromTextureInfo(TextureFormat format, ComponentType red,
     case Hash(TextureFormat::X8Z24, UNORM):
         return PixelFormat::X8_D24_UNORM;
     case Hash(TextureFormat::X8Z24, UNORM, UINT, UINT, UINT, LINEAR):
+    case Hash(TextureFormat::X8Z24, UNORM, UINT, UNORM, UNORM, LINEAR):
         return PixelFormat::X8_D24_UNORM;
     case Hash(TextureFormat::Z24S8, UINT, UNORM, UNORM, UNORM, LINEAR):
         return PixelFormat::S8_UINT_D24_UNORM;
diff --git a/src/video_core/texture_cache/texture_cache.h b/src/video_core/texture_cache/texture_cache.h
index ecd8429fa2..abb22b9bfb 100644
--- a/src/video_core/texture_cache/texture_cache.h
+++ b/src/video_core/texture_cache/texture_cache.h
@@ -809,7 +809,8 @@ typename P::Framebuffer* TextureCache<P>::GetFramebuffer() {
 template <class P>
 ImageViewId TextureCache<P>::VisitImageView(u32 index, bool compute) {
     auto& table = compute ? channel_state->compute_image_table : channel_state->graphics_image_table;
-    index = Tegra::Texture::TextureHeaderIndex(index, table.current_limit);
+    index = Tegra::Texture::TextureHeaderIndex(index, table.current_limit,
+                                               gpu_memory->UsesNativeTiling());
     if (index > table.current_limit) {
         LOG_DEBUG(HW_GPU, "Invalid image view index={}", index);
         return NULL_IMAGE_VIEW_ID;