nx2: memo for compute macros; one lock per inline upload; cheaper shader invalidation

- Compute engine macros replay repeated calls like 3D ones (title 1's
  compute macro path fell from ~27% to ~14% of the GPU thread).
- AccelerateInlineToMemory takes the buffer and texture cache mutexes
  together instead of one after the other.
- ShaderCache::InvalidatePagesInRegion looks at the rejected programs
  only when the write can reach one (bounds kept as they are added);
  it ran a locked scan for every inline upload and pre-draw
  invalidation.

Inline uploads went from ~10% to ~4% of the GPU thread.

1 files changed, 2 insertions, 4 deletions

diff --git a/src/video_core/renderer_vulkan/vk_rasterizer.cpp b/src/video_core/renderer_vulkan/vk_rasterizer.cpp
index decfb0d136..6bd267e566 100644
--- a/src/video_core/renderer_vulkan/vk_rasterizer.cpp
+++ b/src/video_core/renderer_vulkan/vk_rasterizer.cpp
@@ -1251,13 +1251,11 @@ void RasterizerVulkan::AccelerateInlineToMemory(GPUVAddr address, size_t copy_si
     }
     gpu_memory->WriteBlockUnsafe(address, memory.data(), copy_size);
     {
-        std::unique_lock<std::recursive_mutex> lock{buffer_cache.mutex};
+        // Both caches under one lock (title 1 makes thousands of inline uploads a second).
+        std::scoped_lock lock{buffer_cache.mutex, texture_cache.mutex};
         if (!buffer_cache.InlineMemory(*cpu_addr, copy_size, memory)) {
             buffer_cache.WriteMemory(*cpu_addr, copy_size);
         }
-    }
-    {
-        std::scoped_lock lock_texture{texture_cache.mutex};
         texture_cache.WriteMemory(*cpu_addr, copy_size);
     }
     pipeline_cache.InvalidateRegion(*cpu_addr, copy_size);