shader_recompiler: Implement most integer image atomics, workgroup barriers and shared memory load/store (#231)

* shader_recompiler: Add LDEXP

* shader_recompiler: Add most image integer atomic ops

* shader_recompiler: Implement shared memory load/store

* shader_recompiler: More image atomics

* externals: Update sirit

* clang format

* cmake: Add missing files

* shader_recompiler: Fix some atomic bugs

* shader_recompiler: Vs outputs

* shader_recompiler: Shared mem has side-effects, fix format component order

* shader_recompiler: Inline constant buffer impl

* video_core: Fix regressions

* Work

* Fixup a few things

src/video_core/texture_cache/image.cpp

@@ -321,12 +321,15 @@ void Image::Upload(vk::Buffer buffer, u64 offset) {
     Transit(vk::ImageLayout::eTransferDstOptimal, vk::AccessFlagBits::eTransferWrite);
 
     // Copy to the image.
+    const auto aspect = aspect_mask & vk::ImageAspectFlagBits::eStencil
+                            ? vk::ImageAspectFlagBits::eDepth
+                            : aspect_mask;
     const vk::BufferImageCopy image_copy = {
         .bufferOffset = offset,
         .bufferRowLength = info.pitch,
         .bufferImageHeight = info.size.height,
         .imageSubresource{
-            .aspectMask = aspect_mask,
+            .aspectMask = aspect,
             .mipLevel = 0,
             .baseArrayLayer = 0,
             .layerCount = 1,

src/video_core/texture_cache/image_view.cpp

@@ -77,7 +77,6 @@ ImageView::ImageView(const Vulkan::Instance& instance, const ImageViewInfo& info
     if (usage_override) {
         usage_ci.usage = usage_override.value();
     }
-
     // When sampling D32 texture from shader, the T# specifies R32 Float format so adjust it.
     vk::Format format = info.format;
     vk::ImageAspectFlags aspect = image.aspect_mask;

src/video_core/texture_cache/texture_cache.cpp

@@ -134,13 +134,13 @@ ImageId TextureCache::FindImage(const ImageInfo& info, VAddr cpu_address, bool r
         image_id = slot_images.insert(instance, scheduler, info, cpu_address);
         RegisterImage(image_id);
     } else {
-        image_id = image_ids.size() > 1 ? image_ids[1] : image_ids[0];
+        image_id = image_ids[0];
     }
 
     RegisterMeta(info, image_id);
 
     Image& image = slot_images[image_id];
-    if (True(image.flags & ImageFlagBits::CpuModified)) {
+    if (True(image.flags & ImageFlagBits::CpuModified) && refresh_on_create) {
         RefreshImage(image);
         TrackImage(image, image_id);
     }
@@ -193,7 +193,7 @@ ImageView& TextureCache::FindImageView(const AmdGpu::Image& desc, bool is_storag
 ImageView& TextureCache::RenderTarget(const AmdGpu::Liverpool::ColorBuffer& buffer,
                                       const AmdGpu::Liverpool::CbDbExtent& hint) {
     const ImageInfo info{buffer, hint};
-    const ImageId image_id = FindImage(info, buffer.Address(), false);
+    const ImageId image_id = FindImage(info, buffer.Address());
     Image& image = slot_images[image_id];
     image.flags &= ~ImageFlagBits::CpuModified;
 

src/video_core/texture_cache/tile_manager.cpp

@@ -179,19 +179,19 @@ vk::Format DemoteImageFormatForDetiling(vk::Format format) {
     case vk::Format::eR8Unorm:
         return vk::Format::eR8Uint;
     case vk::Format::eR8G8Unorm:
+    case vk::Format::eR16Sfloat:
         return vk::Format::eR8G8Uint;
     case vk::Format::eR8G8B8A8Srgb:
-        [[fallthrough]];
     case vk::Format::eB8G8R8A8Srgb:
-        [[fallthrough]];
     case vk::Format::eB8G8R8A8Unorm:
-        [[fallthrough]];
     case vk::Format::eR8G8B8A8Unorm:
+    case vk::Format::eR32Sfloat:
+    case vk::Format::eR32Uint:
         return vk::Format::eR32Uint;
     case vk::Format::eBc1RgbaUnormBlock:
+    case vk::Format::eR32G32Sfloat:
         return vk::Format::eR32G32Uint;
     case vk::Format::eBc3SrgbBlock:
-        [[fallthrough]];
     case vk::Format::eBc3UnormBlock:
     case vk::Format::eBc7SrgbBlock:
     case vk::Format::eBc7UnormBlock: