From b5bcd8c71b2d5fd0528191990b4e11bc916b5d7a Mon Sep 17 00:00:00 2001
From: ameerj <52414509+ameerj@users.noreply.github.com>
Date: Wed, 22 Feb 2023 00:48:12 -0500
Subject: configuration: Add async ASTC decode setting

---
 src/video_core/renderer_opengl/gl_texture_cache.cpp | 17 ++++++++++++++---
 src/video_core/renderer_vulkan/vk_texture_cache.cpp |  7 ++++---
 src/video_core/texture_cache/texture_cache.h        |  1 +
 src/video_core/textures/astc.cpp                    |  4 ++--
 4 files changed, 21 insertions(+), 8 deletions(-)

(limited to 'src/video_core')

diff --git a/src/video_core/renderer_opengl/gl_texture_cache.cpp b/src/video_core/renderer_opengl/gl_texture_cache.cpp
index eb6e43a08..b047e7b3d 100644
--- a/src/video_core/renderer_opengl/gl_texture_cache.cpp
+++ b/src/video_core/renderer_opengl/gl_texture_cache.cpp
@@ -228,8 +228,9 @@ void ApplySwizzle(GLuint handle, PixelFormat format, std::array<SwizzleSource, 4
 
 [[nodiscard]] bool CanBeAccelerated(const TextureCacheRuntime& runtime,
                                     const VideoCommon::ImageInfo& info) {
-    if (IsPixelFormatASTC(info.format)) {
-        return !runtime.HasNativeASTC() && Settings::values.accelerate_astc.GetValue();
+    if (IsPixelFormatASTC(info.format) && !runtime.HasNativeASTC()) {
+        return Settings::values.accelerate_astc.GetValue() &&
+               !Settings::values.async_astc.GetValue();
     }
     // Disable other accelerated uploads for now as they don't implement swizzled uploads
     return false;
@@ -258,6 +259,14 @@ void ApplySwizzle(GLuint handle, PixelFormat format, std::array<SwizzleSource, 4
     return format_info.compatibility_class == store_class;
 }
 
+[[nodiscard]] bool CanBeDecodedAsync(const TextureCacheRuntime& runtime,
+                                     const VideoCommon::ImageInfo& info) {
+    if (IsPixelFormatASTC(info.format) && !runtime.HasNativeASTC()) {
+        return Settings::values.async_astc.GetValue();
+    }
+    return false;
+}
+
 [[nodiscard]] CopyOrigin MakeCopyOrigin(VideoCommon::Offset3D offset,
                                         VideoCommon::SubresourceLayers subresource, GLenum target) {
     switch (target) {
@@ -721,7 +730,9 @@ std::optional<size_t> TextureCacheRuntime::StagingBuffers::FindBuffer(size_t req
 Image::Image(TextureCacheRuntime& runtime_, const VideoCommon::ImageInfo& info_, GPUVAddr gpu_addr_,
              VAddr cpu_addr_)
     : VideoCommon::ImageBase(info_, gpu_addr_, cpu_addr_), runtime{&runtime_} {
-    if (CanBeAccelerated(*runtime, info)) {
+    if (CanBeDecodedAsync(*runtime, info)) {
+        flags |= ImageFlagBits::AsynchronousDecode;
+    } else if (CanBeAccelerated(*runtime, info)) {
         flags |= ImageFlagBits::AcceleratedUpload;
     }
     if (IsConverted(runtime->device, info.format, info.type)) {
diff --git a/src/video_core/renderer_vulkan/vk_texture_cache.cpp b/src/video_core/renderer_vulkan/vk_texture_cache.cpp
index 9b85dfb5e..80adb70eb 100644
--- a/src/video_core/renderer_vulkan/vk_texture_cache.cpp
+++ b/src/video_core/renderer_vulkan/vk_texture_cache.cpp
@@ -1256,11 +1256,12 @@ Image::Image(TextureCacheRuntime& runtime_, const ImageInfo& info_, GPUVAddr gpu
       commit(runtime_.memory_allocator.Commit(original_image, MemoryUsage::DeviceLocal)),
       aspect_mask(ImageAspectMask(info.format)) {
     if (IsPixelFormatASTC(info.format) && !runtime->device.IsOptimalAstcSupported()) {
-        if (Settings::values.accelerate_astc.GetValue()) {
+        if (Settings::values.async_astc.GetValue()) {
+            flags |= VideoCommon::ImageFlagBits::AsynchronousDecode;
+        } else if (Settings::values.accelerate_astc.GetValue()) {
             flags |= VideoCommon::ImageFlagBits::AcceleratedUpload;
-        } else {
-            flags |= VideoCommon::ImageFlagBits::Converted;
         }
+        flags |= VideoCommon::ImageFlagBits::Converted;
         flags |= VideoCommon::ImageFlagBits::CostlyLoad;
     }
     if (runtime->device.HasDebuggingToolAttached()) {
diff --git a/src/video_core/texture_cache/texture_cache.h b/src/video_core/texture_cache/texture_cache.h
index 4159bc796..9dd152fbe 100644
--- a/src/video_core/texture_cache/texture_cache.h
+++ b/src/video_core/texture_cache/texture_cache.h
@@ -1003,6 +1003,7 @@ u64 TextureCache<P>::GetScaledImageSizeBytes(const ImageBase& image) {
 template <class P>
 void TextureCache<P>::QueueAsyncDecode(Image& image, ImageId image_id) {
     UNIMPLEMENTED_IF(False(image.flags & ImageFlagBits::Converted));
+    LOG_INFO(HW_GPU, "Queuing async texture decode");
 
     image.flags |= ImageFlagBits::IsDecoding;
     auto decode = std::make_unique<AsyncDecodeContext>();
diff --git a/src/video_core/textures/astc.cpp b/src/video_core/textures/astc.cpp
index e8d7c7863..4381eed1d 100644
--- a/src/video_core/textures/astc.cpp
+++ b/src/video_core/textures/astc.cpp
@@ -1656,8 +1656,8 @@ void Decompress(std::span<const uint8_t> data, uint32_t width, uint32_t height,
     const u32 rows = Common::DivideUp(height, block_height);
     const u32 cols = Common::DivideUp(width, block_width);
 
-    Common::ThreadWorker workers{std::max(std::thread::hardware_concurrency(), 2U) / 2,
-                                 "ASTCDecompress"};
+    static Common::ThreadWorker workers{std::max(std::thread::hardware_concurrency(), 2U) / 2,
+                                        "ASTCDecompress"};
 
     for (u32 z = 0; z < depth; ++z) {
         const u32 depth_offset = z * height * width * 4;
-- 
cgit v1.2.3