Revert "[android] Add Enable GPU Buffer Readback option to Android (#4132)"

This reverts commit d142b5dd6a.

src/android/app/src/main/java/org/yuzu/yuzu_emu/features/settings/model/BooleanSetting.kt

@@ -31,7 +31,6 @@ enum class BooleanSetting(override val key: String) : AbstractBooleanSetting {
     RENDERER_REACTIVE_FLUSHING("use_reactive_flushing"),
     ENABLE_BUFFER_HISTORY("enable_buffer_history"),
     USE_OPTIMIZED_VERTEX_BUFFERS("use_optimized_vertex_buffers"),
-    ENABLE_GPU_BUFFER_READBACK("enable_gpu_buffer_readback"),
     SYNC_MEMORY_OPERATIONS("sync_memory_operations"),
     BUFFER_REORDER_DISABLE("disable_buffer_reorder"),
     RENDERER_DEBUG("debug"),

src/android/app/src/main/java/org/yuzu/yuzu_emu/features/settings/model/view/SettingsItem.kt

@@ -806,13 +806,6 @@ abstract class SettingsItem(
                     descriptionId = R.string.enable_buffer_history_description
                 )
             )
-            put(
-                SwitchSetting(
-                    BooleanSetting.ENABLE_GPU_BUFFER_READBACK,
-                    titleId = R.string.enable_gpu_buffer_readback,
-                    descriptionId = R.string.enable_gpu_buffer_readback_description
-                )
-            )
             put(
                 SwitchSetting(
                     BooleanSetting.USE_OPTIMIZED_VERTEX_BUFFERS,

src/android/app/src/main/java/org/yuzu/yuzu_emu/features/settings/ui/SettingsFragmentPresenter.kt

@@ -292,7 +292,6 @@ class SettingsFragmentPresenter(
             add(BooleanSetting.RENDERER_FORCE_MAX_CLOCK.key)
             add(BooleanSetting.RENDERER_REACTIVE_FLUSHING.key)
             add(BooleanSetting.ENABLE_BUFFER_HISTORY.key)
-            add(BooleanSetting.ENABLE_GPU_BUFFER_READBACK.key)
             add(BooleanSetting.USE_OPTIMIZED_VERTEX_BUFFERS.key)
 
             add(HeaderSetting(R.string.hacks))

src/android/app/src/main/res/values/strings.xml

@@ -503,8 +503,6 @@
     <string name="renderer_reactive_flushing_description">Improves rendering accuracy in some games at the cost of performance.</string>
     <string name="enable_buffer_history">Enable buffer history</string>
     <string name="enable_buffer_history_description">Enables access to previous buffer states. This option may improve rendering quality and performance consistency in some games.</string>
-    <string name="enable_gpu_buffer_readback">Enable GPU Buffer Readback</string>
-    <string name="enable_gpu_buffer_readback_description">Preserves GPU-modified buffer data by reading it back before uploads. Some games require this to render certain effects properly. May cause issues if the hardware cannot handle the additional workload.</string>
     <string name="use_optimized_vertex_buffers">Optimized Vertex Buffers</string>
     <string name="use_optimized_vertex_buffers_description">Enables optimized vertex buffer binding for improved performance. Requires Mesa 26.0+ Turnip drivers/ QCOM drivers. Will crash on older Turnip drivers (25.3 and below).</string>
 

src/video_core/buffer_cache/buffer_cache.h

@@ -1619,35 +1619,36 @@ void BufferCache<P>::TouchBuffer(Buffer& buffer, BufferId buffer_id) noexcept {
 template <class P>
 bool BufferCache<P>::SynchronizeBuffer(Buffer& buffer, DAddr device_addr, u32 size) {
     upload_copies.clear();
-    u64 total_size_bytes = 0;
+    u64 staging_offset = 0;
     u64 largest_copy = 0;
-    const DAddr buffer_start = buffer.cpu_addr_cached;
-    memory_tracker.ForEachUploadRange(device_addr, size, [&](u64 device_addr_out, u64 range_size) {
-        upload_copies.push_back(BufferCopy{
-            .src_offset = total_size_bytes,
-            .dst_offset = device_addr_out - buffer_start,
-            .size = range_size,
+    DAddr buffer_start = buffer.CpuAddr();
+    auto push = [&](u64 start, u64 end) {
+        if (start >= end) {
+            return;
+        }
+        u64 sz = end - start;
+        upload_copies.push_back({
+            .src_offset = staging_offset,
+            .dst_offset = start - buffer_start,
+            .size = sz
         });
-        total_size_bytes += range_size;
-        largest_copy = (std::max)(largest_copy, range_size);
+        staging_offset += sz;
+        largest_copy = (std::max)(largest_copy, sz);
+    };
+    memory_tracker.ForEachUploadRange(device_addr, size, [&](u64 addr, u64 range_size) {
+        u64 start = addr;
+        u64 end = addr + range_size;
+        gpu_modified_ranges.ForEachInRange(start, range_size, [&](u64 gstart, u64 gsize) {
+            u64 gend = gstart + gsize;
+            push(start, gstart);
+            start = (std::max)(start, gend);
+        });
+        push(start, end);
     });
-    if (total_size_bytes == 0) {
+    if (upload_copies.empty()) {
         return true;
     }
-    if (Settings::values.enable_gpu_buffer_readback.GetValue()) {
-        u64 min_offset = (std::numeric_limits<u64>::max)();
-        u64 max_offset = 0;
-        for (const auto& copy : upload_copies) {
-            min_offset = (std::min)(min_offset, copy.dst_offset);
-            max_offset = (std::max)(max_offset, copy.dst_offset + copy.size);
-        }
-        const DAddr sync_addr = buffer.CpuAddr() + min_offset;
-        const u64 sync_size = max_offset - min_offset;
-        DownloadBufferMemory(buffer, sync_addr, sync_size);
-    }
-    const std::span<BufferCopy> copies_span(upload_copies.data(), upload_copies.size());
-    UploadMemory(buffer, total_size_bytes, largest_copy, copies_span);
-    any_buffer_uploaded = true;
+    UploadMemory(buffer, staging_offset, largest_copy, std::span(upload_copies));
     return false;
 }
 
@@ -1698,6 +1699,7 @@ void BufferCache<P>::MappedUploadMemory([[maybe_unused]] Buffer& buffer,
             u8* const src_pointer = staging_pointer.data() + copy.src_offset;
             const DAddr device_addr = buffer.CpuAddr() + copy.dst_offset;
             device_memory.ReadBlockUnsafe(device_addr, src_pointer, copy.size);
+
             // Apply the staging offset
             copy.src_offset += upload_staging.offset;
         }