video_core: compact the macro memo arena; download profiling counters

Calls stored again with another shape leave their old record in the memo's arena;
title 1 filled its 16 MiB every ~2 s and clearing it made the next ~20,000 calls
interpret again. The live records are now copied to a new arena (a clear only when
they alone fill three quarters of it). The memo statistics count hits on the last
call's slot and compactions. With EDEN_NX2_GPU_PROFILE the GPU thread also reports
how much downloaded buffer data lands on pages the native MME has DMA-read.

1 files changed, 13 insertions, 0 deletions

diff --git a/src/video_core/buffer_cache/buffer_cache.h b/src/video_core/buffer_cache/buffer_cache.h
index e2de0eede1..b909971e42 100644
--- a/src/video_core/buffer_cache/buffer_cache.h
+++ b/src/video_core/buffer_cache/buffer_cache.h
@@ -679,6 +679,19 @@ void BufferCache<P>::CommitAsyncFlushesHigh() {
         Buffer& buffer = slot_buffers[buffer_id];
         second_copy.src_offset = static_cast<size_t>(buffer.CpuAddr()) + copy.src_offset;
         const DAddr orig_device_addr = static_cast<DAddr>(second_copy.src_offset);
+        if (Nx2Profile::Enabled()) {
+            u64 on_dma_pages = 0;
+            const auto& pages = Nx2Profile::DmaReadPages();
+            for (DAddr page = orig_device_addr >> 12; page <= (orig_device_addr + copy.size - 1) >> 12;
+                 ++page) {
+                if (pages.contains(page)) {
+                    const DAddr begin = (std::max)(orig_device_addr, page << 12);
+                    const DAddr end = (std::min)(orig_device_addr + copy.size, (page + 1) << 12);
+                    on_dma_pages += end - begin;
+                }
+            }
+            Nx2Profile::Add(Nx2Profile::Counter::DownloadBytesOnDmaPages, on_dma_pages);
+        }
         async_downloads.Add(orig_device_addr, copy.size);
         buffer.MarkUsage(copy.src_offset, copy.size);
         runtime.CopyBuffer(download_staging.buffer, buffer, copies, false);