Fix the P6 scratch save in R2UR/MATCH/HFMA2.RELU; GPU memory dump tool

The colour cast on title 1 (magenta/cyan, then red/green, changing per
course) came from the sky/atmosphere volume shader (a pixel shader drawing
a 32^3 RGBA16F table slice by slice), which starts with an unguarded R2UR.
R2UR, MATCH and HFMA2.RELU use P6 as scratch and saved/restored it with the
uniform-guard mask, 1 << (6 - 7) for an unguarded instruction: undefined, in
practice bit 31, which shifted into the opcode bits and turned the save into
another instruction. They now always save and restore P6.

Debugging aids used to find it:
- EDEN_NX2_DUMP=<present>:<addr>:<size>,...: flush GPU memory ranges at a
  present (on the GPU thread) and save them next to the frame captures;
  nx2work/vol3d.py untiles Switch 2 3D textures from such dumps.
- GPU::RunOnGpuThread.
- Logs of Switch 2 3D images and descriptors.

4 files changed, 27 insertions, 1 deletions

diff --git a/src/video_core/gpu.cpp b/src/video_core/gpu.cpp
index 2a61a66d82..6441b9511a 100644
--- a/src/video_core/gpu.cpp
+++ b/src/video_core/gpu.cpp
@@ -411,6 +411,10 @@ u64 GPU::RequestFlush(DAddr addr, std::size_t size) {
     });
 }
 
+u64 GPU::RunOnGpuThread(std::function<void()>&& action) {
+    return impl->RequestSyncOperation(std::move(action));
+}
+
 u64 GPU::CurrentSyncRequestFence() const {
     return impl->CurrentSyncRequestFence();
 }
diff --git a/src/video_core/gpu.h b/src/video_core/gpu.h
index 41ba186cd3..3a6546b41b 100644
--- a/src/video_core/gpu.h
+++ b/src/video_core/gpu.h
@@ -164,6 +164,9 @@ public:
     /// Request a host GPU memory flush from the CPU.
     [[nodiscard]] u64 RequestFlush(DAddr addr, std::size_t size);
 
+    /// Runs an action on the GPU thread between command lists (returns its sync fence).
+    u64 RunOnGpuThread(std::function<void()>&& action);
+
     /// Obtains current flush request fence id.
     [[nodiscard]] u64 CurrentSyncRequestFence() const;
 
diff --git a/src/video_core/texture_cache/image_info.cpp b/src/video_core/texture_cache/image_info.cpp
index 8b59eb6841..7e86bf00cf 100644
--- a/src/video_core/texture_cache/image_info.cpp
+++ b/src/video_core/texture_cache/image_info.cpp
@@ -4,9 +4,13 @@
 // SPDX-FileCopyrightText: Copyright 2020 yuzu Emulator Project
 // SPDX-License-Identifier: GPL-2.0-or-later
 
+#include <algorithm>
+#include <array>
+#include <vector>
 #include <fmt/ranges.h>
 
 #include "common/assert.h"
+#include "common/logging.h"
 #include "common/settings.h"
 #include "video_core/surface.h"
 #include "video_core/texture_cache/format_lookup_table.h"
@@ -109,14 +113,23 @@ ImageInfo::ImageInfo(const TICEntry& config) noexcept {
         size.height = config.Height();
         resources.layers = config.BaseLayer() + config.Depth() * 6;
         break;
-    case TextureType::Texture3D:
+    case TextureType::Texture3D: {
         ASSERT(config.BaseLayer() == 0);
         type = ImageType::e3D;
         size.width = config.Width();
         size.height = config.Height();
         size.depth = config.Depth();
         resources.layers = 1;
+        // A few distinct 3D descriptors, raw (to check Switch 2 volume fields).
+        static std::vector<std::array<u64, 4>> logged;
+        if (logged.size() < 16 && std::find(logged.begin(), logged.end(), config.raw) == logged.end()) {
+            logged.push_back(config.raw);
+            LOG_INFO(HW_GPU, "3D TIC {:016x} {:016x} {:016x} {:016x}: {}x{}x{} block_depth {} levels {}",
+                     config.raw[0], config.raw[1], config.raw[2], config.raw[3], size.width, size.height,
+                     size.depth, config.block_depth.Value(), config.max_mip_level.Value() + 1);
+        }
         break;
+    }
     case TextureType::Texture1DBuffer:
         type = ImageType::Buffer;
         size.width = config.Width();
diff --git a/src/video_core/texture_cache/texture_cache.h b/src/video_core/texture_cache/texture_cache.h
index f3da5684a5..75005f0731 100644
--- a/src/video_core/texture_cache/texture_cache.h
+++ b/src/video_core/texture_cache/texture_cache.h
@@ -1569,6 +1569,12 @@ ImageId TextureCache<P>::InsertImage(const ImageInfo& info, GPUVAddr gpu_addr,
     ASSERT_MSG(cpu_addr, "Tried to insert an image to an invalid gpu_addr={:#x}", gpu_addr);
     ImageInfo storage_info = info;
     storage_info.native_tiling = gpu_memory->UsesNativeTiling();
+    if (storage_info.native_tiling && info.type == ImageType::e3D) {
+        // Switch 2 3D images (lighting/fog volumes) are rare: log each one's shape.
+        LOG_INFO(HW_GPU, "3D image at {:#x}: {}x{}x{} levels {} format {} block {}x{}x{}",
+                 gpu_addr, info.size.width, info.size.height, info.size.depth, info.resources.levels,
+                 static_cast<u32>(info.format), info.block.width, info.block.height, info.block.depth);
+    }
     const ImageId image_id = JoinImages(storage_info, gpu_addr, *cpu_addr);
     const Image& image = slot_images[image_id];
     // Using "image.gpu_addr" instead of "gpu_addr" is important because it might be different