texture_cache: free deleted images once the GPU is past them

Deleted images kept their device memory for TICKS_TO_DESTROY (8)
frames. While the collector deletes dozens a frame that held 1-1.5 GB
of VRAM on top of its threshold, which the collector does not count
(Mario Kart World on an 8 GB card: 6.6-7.0 GB in use against 5 GiB). On
runtimes that report GPU progress (Vulkan) an image is destroyed as
soon as the GPU is past it. Title screen, 10 minutes: VRAM in use 6.75
-> 6.34 GB on average, frames over 100 ms 22 -> 15.

(cherry picked from commit 8c899541baf398deb892a857a932590f32d2aa4f)

2 files changed, 27 insertions, 2 deletions

diff --git a/src/video_core/texture_cache/texture_cache.h b/src/video_core/texture_cache/texture_cache.h
index 71f89db367..46ebf6a25f 100644
--- a/src/video_core/texture_cache/texture_cache.h
+++ b/src/video_core/texture_cache/texture_cache.h
@@ -247,6 +247,13 @@ void TextureCache<P>::TickFrame() {
             }
         }
     }
+    if constexpr (requires { runtime.CurrentTick(); }) {
+        while (!gpu_sentenced_images.empty() &&
+               runtime.IsFree(gpu_sentenced_images.front().tick)) {
+            gpu_sentenced_bytes -= gpu_sentenced_images.front().bytes;
+            gpu_sentenced_images.pop_front();
+        }
+    }
     // If we can obtain the memory info, use it instead of the estimate.
     if (runtime.CanReportMemoryUsage()) {
         total_used_memory = runtime.GetDeviceMemoryUsage();
@@ -256,6 +263,7 @@ void TextureCache<P>::TickFrame() {
         for (const u64 bytes : sentenced_image_bytes) {
             total_used_memory -= (std::min)(total_used_memory, bytes);
         }
+        total_used_memory -= (std::min)(total_used_memory, gpu_sentenced_bytes);
     }
     // Every 600 frames: device memory in use against the collector's thresholds, and what the
     // images themselves account for (VRAM pressure makes the collector thrash).
@@ -2906,8 +2914,14 @@ void TextureCache<P>::DeleteImage(ImageId image_id, bool immediate_delete) {
     if (retained) {
         LOG_INFO(Render_Vulkan, "Retained image for queued download: id={}", image_id.Value());
     } else if (!immediate_delete) {
-        sentenced_images.Push(std::move(slot_images[image_id]));
-        sentenced_image_bytes[sentenced_image_index] += freed_bytes;
+        if constexpr (requires { runtime.CurrentTick(); }) {
+            gpu_sentenced_images.push_back(
+                {runtime.CurrentTick(), freed_bytes, std::move(slot_images[image_id])});
+            gpu_sentenced_bytes += freed_bytes;
+        } else {
+            sentenced_images.Push(std::move(slot_images[image_id]));
+            sentenced_image_bytes[sentenced_image_index] += freed_bytes;
+        }
     }
     slot_images.erase(image_id);
 
diff --git a/src/video_core/texture_cache/texture_cache_base.h b/src/video_core/texture_cache/texture_cache_base.h
index d51eefeee8..151dd9bfa1 100644
--- a/src/video_core/texture_cache/texture_cache_base.h
+++ b/src/video_core/texture_cache/texture_cache_base.h
@@ -556,6 +556,17 @@ private:
     /// counted in the driver's usage until the ring destroys them.
     std::array<u64, TICKS_TO_DESTROY> sentenced_image_bytes{};
     size_t sentenced_image_index = 0;
+    /// Deleted images on runtimes that report GPU progress (CurrentTick, IsFree): destroyed, with
+    /// their device memory, as soon as the GPU is past them rather than TICKS_TO_DESTROY frames
+    /// later. While the collector deletes dozens a frame, eight frames of them held 1-1.5 GB of
+    /// VRAM above its threshold (title 1 on an 8 GB card: 6.6-7.0 GB in use against 5 GiB).
+    struct GpuSentencedImage {
+        u64 tick;
+        u64 bytes;
+        Image image;
+    };
+    std::deque<GpuSentencedImage> gpu_sentenced_images;
+    u64 gpu_sentenced_bytes{};
     DelayedDestructionRing<ImageView, TICKS_TO_DESTROY> sentenced_image_view;
     DelayedDestructionRing<Framebuffer, TICKS_TO_DESTROY> sentenced_framebuffers;