Merge pull request #10996 from Kelebek1/readblock_optimisation

Use spans over guest memory where possible instead of copying data

3 files changed, 21 insertions, 32 deletions

diff --git a/src/video_core/texture_cache/texture_cache.h b/src/video_core/texture_cache/texture_cache.h
index 3a859139c..4457b366f 100644
--- a/src/video_core/texture_cache/texture_cache.h
+++ b/src/video_core/texture_cache/texture_cache.h
@@ -8,6 +8,7 @@
 
 #include "common/alignment.h"
 #include "common/settings.h"
+#include "core/memory.h"
 #include "video_core/control/channel_state.h"
 #include "video_core/dirty_flags.h"
 #include "video_core/engines/kepler_compute.h"
@@ -1026,19 +1027,19 @@ void TextureCache<P>::UploadImageContents(Image& image, StagingBuffer& staging)
         runtime.AccelerateImageUpload(image, staging, uploads);
         return;
     }
-    const size_t guest_size_bytes = image.guest_size_bytes;
-    swizzle_data_buffer.resize_destructive(guest_size_bytes);
-    gpu_memory->ReadBlockUnsafe(gpu_addr, swizzle_data_buffer.data(), guest_size_bytes);
+
+    Core::Memory::GpuGuestMemory<u8, Core::Memory::GuestMemoryFlags::UnsafeRead> swizzle_data(
+        *gpu_memory, gpu_addr, image.guest_size_bytes, &swizzle_data_buffer);
 
     if (True(image.flags & ImageFlagBits::Converted)) {
         unswizzle_data_buffer.resize_destructive(image.unswizzled_size_bytes);
-        auto copies = UnswizzleImage(*gpu_memory, gpu_addr, image.info, swizzle_data_buffer,
-                                     unswizzle_data_buffer);
+        auto copies =
+            UnswizzleImage(*gpu_memory, gpu_addr, image.info, swizzle_data, unswizzle_data_buffer);
         ConvertImage(unswizzle_data_buffer, image.info, mapped_span, copies);
         image.UploadMemory(staging, copies);
     } else {
         const auto copies =
-            UnswizzleImage(*gpu_memory, gpu_addr, image.info, swizzle_data_buffer, mapped_span);
+            UnswizzleImage(*gpu_memory, gpu_addr, image.info, swizzle_data, mapped_span);
         image.UploadMemory(staging, copies);
     }
 }
@@ -1231,11 +1232,12 @@ void TextureCache<P>::QueueAsyncDecode(Image& image, ImageId image_id) {
     decode->image_id = image_id;
     async_decodes.push_back(std::move(decode));
 
-    Common::ScratchBuffer<u8> local_unswizzle_data_buffer(image.unswizzled_size_bytes);
-    const size_t guest_size_bytes = image.guest_size_bytes;
-    swizzle_data_buffer.resize_destructive(guest_size_bytes);
-    gpu_memory->ReadBlockUnsafe(image.gpu_addr, swizzle_data_buffer.data(), guest_size_bytes);
-    auto copies = UnswizzleImage(*gpu_memory, image.gpu_addr, image.info, swizzle_data_buffer,
+    static Common::ScratchBuffer<u8> local_unswizzle_data_buffer;
+    local_unswizzle_data_buffer.resize_destructive(image.unswizzled_size_bytes);
+    Core::Memory::GpuGuestMemory<u8, Core::Memory::GuestMemoryFlags::UnsafeRead> swizzle_data(
+        *gpu_memory, image.gpu_addr, image.guest_size_bytes, &swizzle_data_buffer);
+
+    auto copies = UnswizzleImage(*gpu_memory, image.gpu_addr, image.info, swizzle_data,
                                  local_unswizzle_data_buffer);
     const size_t out_size = MapSizeBytes(image);
 
diff --git a/src/video_core/texture_cache/util.cpp b/src/video_core/texture_cache/util.cpp
index 0de6ed09d..a83f5d41c 100644
--- a/src/video_core/texture_cache/util.cpp
+++ b/src/video_core/texture_cache/util.cpp
@@ -20,6 +20,7 @@
 #include "common/div_ceil.h"
 #include "common/scratch_buffer.h"
 #include "common/settings.h"
+#include "core/memory.h"
 #include "video_core/compatible_formats.h"
 #include "video_core/engines/maxwell_3d.h"
 #include "video_core/memory_manager.h"
@@ -544,17 +545,15 @@ void SwizzleBlockLinearImage(Tegra::MemoryManager& gpu_memory, GPUVAddr gpu_addr
                        tile_size.height, info.tile_width_spacing);
     const size_t subresource_size = sizes[level];
 
-    tmp_buffer.resize_destructive(subresource_size);
-    const std::span<u8> dst(tmp_buffer);
-
     for (s32 layer = 0; layer < info.resources.layers; ++layer) {
         const std::span<const u8> src = input.subspan(host_offset);
-        gpu_memory.ReadBlockUnsafe(gpu_addr + guest_offset, dst.data(), dst.size_bytes());
-
-        SwizzleTexture(dst, src, bytes_per_block, num_tiles.width, num_tiles.height,
-                       num_tiles.depth, block.height, block.depth);
+        {
+            Core::Memory::GpuGuestMemoryScoped<u8, Core::Memory::GuestMemoryFlags::UnsafeReadWrite>
+                dst(gpu_memory, gpu_addr + guest_offset, subresource_size, &tmp_buffer);
 
-        gpu_memory.WriteBlockUnsafe(gpu_addr + guest_offset, dst.data(), dst.size_bytes());
+            SwizzleTexture(dst, src, bytes_per_block, num_tiles.width, num_tiles.height,
+                           num_tiles.depth, block.height, block.depth);
+        }
 
         host_offset += host_bytes_per_layer;
         guest_offset += layer_stride;
@@ -837,6 +836,7 @@ boost::container::small_vector<BufferImageCopy, 16> UnswizzleImage(Tegra::Memory
     const Extent3D size = info.size;
 
     if (info.type == ImageType::Linear) {
+        ASSERT(output.size_bytes() >= guest_size_bytes);
         gpu_memory.ReadBlockUnsafe(gpu_addr, output.data(), guest_size_bytes);
 
         ASSERT((info.pitch >> bpp_log2) << bpp_log2 == info.pitch);
@@ -904,16 +904,6 @@ boost::container::small_vector<BufferImageCopy, 16> UnswizzleImage(Tegra::Memory
     return copies;
 }
 
-BufferCopy UploadBufferCopy(Tegra::MemoryManager& gpu_memory, GPUVAddr gpu_addr,
-                            const ImageBase& image, std::span<u8> output) {
-    gpu_memory.ReadBlockUnsafe(gpu_addr, output.data(), image.guest_size_bytes);
-    return BufferCopy{
-        .src_offset = 0,
-        .dst_offset = 0,
-        .size = image.guest_size_bytes,
-    };
-}
-
 void ConvertImage(std::span<const u8> input, const ImageInfo& info, std::span<u8> output,
                   std::span<BufferImageCopy> copies) {
     u32 output_offset = 0;
diff --git a/src/video_core/texture_cache/util.h b/src/video_core/texture_cache/util.h
index ab45a43c4..5a0649d24 100644
--- a/src/video_core/texture_cache/util.h
+++ b/src/video_core/texture_cache/util.h
@@ -66,9 +66,6 @@ struct OverlapResult {
     Tegra::MemoryManager& gpu_memory, GPUVAddr gpu_addr, const ImageInfo& info,
     std::span<const u8> input, std::span<u8> output);
 
-[[nodiscard]] BufferCopy UploadBufferCopy(Tegra::MemoryManager& gpu_memory, GPUVAddr gpu_addr,
-                                          const ImageBase& image, std::span<u8> output);
-
 void ConvertImage(std::span<const u8> input, const ImageInfo& info, std::span<u8> output,
                   std::span<BufferImageCopy> copies);