graphics: resolve Standard4KB color targets

src/graphics/guest_gpu/tile.cpp

@@ -1351,6 +1351,7 @@ uint32_t TileGetTexturePitch(Prospero::BufferFormat format, uint32_t width,
 			}
 			break;
 		}
+		case Prospero::TileMode::kStandard4KB:
 		case Prospero::TileMode::kStandard64KB:
 		case Prospero::TileMode::kPrt:
 		case Prospero::TileMode::kDepth:

src/graphics/host_gpu/renderer/colorRenderTarget.cpp

@@ -130,12 +130,15 @@ void RenderExecutor::ResolveRenderColorTarget(uint64_t submit_id, RenderCommandB
 	if (volume && samples != 1) {
 		EXIT("multisampled 3D render targets are unsupported\n");
 	}
-	const uint32_t depth      = volume ? rt.attrib3.depth + 1u : 1u;
-	const bool     standard64 = rt.attrib3.tile_mode == Prospero::TileMode::kStandard64KB;
-	const bool     depth_tile = rt.attrib3.tile_mode == Prospero::TileMode::kDepth;
+	const uint32_t depth        = volume ? rt.attrib3.depth + 1u : 1u;
+	const bool     standard4    = rt.attrib3.tile_mode == Prospero::TileMode::kStandard4KB;
+	const bool     standard64   = rt.attrib3.tile_mode == Prospero::TileMode::kStandard64KB;
+	const bool     depth_tile   = rt.attrib3.tile_mode == Prospero::TileMode::kDepth;
+	const bool     texture_tile = standard4 || standard64 || depth_tile;
 
 	switch (rt.attrib3.tile_mode) {
 		case Prospero::TileMode::kLinear:
+		case Prospero::TileMode::kStandard4KB:
 		case Prospero::TileMode::kStandard64KB:
 		case Prospero::TileMode::kDepth:
 		case Prospero::TileMode::kRenderTarget:
@@ -149,8 +152,8 @@ void RenderExecutor::ResolveRenderColorTarget(uint64_t submit_id, RenderCommandB
 	if (samples > 1 && (!tile || levels != 1)) {
 		EXIT("multisampled render targets require a single-mip tiled surface\n");
 	}
-	if (depth_tile && samples != 1) {
-		EXIT("depth-tiled color render targets do not support multisampling\n");
+	if (texture_tile && samples != 1) {
+		EXIT("texture-tiled color render targets do not support multisampling\n");
 	}
 
 	width  = rt.attrib2.width + 1;
@@ -162,15 +165,17 @@ void RenderExecutor::ResolveRenderColorTarget(uint64_t submit_id, RenderCommandB
 		EXIT("render-target format has no valid element size\n");
 	}
 	const auto transfer_format = ImageOps::RenderTargetTransferFormat(bytes_per_element);
-	if ((standard64 || depth_tile) &&
-	    (rt.attrib3.dimension != 1 || rt.attrib3.depth != 0 || view.base_layer != 0 ||
-	     view.image_layers != 1 || samples != 1 || rt.pitch.pitch_div8_minus1 != 0 ||
-	     (rt.base.addr & 0xffffu) != 0 || rt.info.fmask_compression_enable ||
-	     rt.info.fmask_data_compression_disable || rt.info.fmask_one_frag_mode ||
-	     rt.info.cmask_fast_clear_enable || rt.info.dcc_compression_enable ||
-	     rt.info.cmask_is_linear != 0 || rt.info.cmask_addr_type != 0 || rt.info.alt_tile_mode ||
-	     rt.cmask.addr != 0 || rt.fmask.addr != 0 || rt.dcc_addr.addr != 0 ||
-	     rt.dcc.data_write_on_dcc_clear_to_reg)) {
+	TileTextureBlockLayout texture_tile_layout {};
+	if (texture_tile &&
+	    (!TileGetTextureBlockLayout(transfer_format, rt.attrib3.tile_mode, volume,
+	                                texture_tile_layout) ||
+	     rt.pitch.pitch_div8_minus1 != 0 ||
+	     (rt.base.addr & (texture_tile_layout.block.block_size - 1u)) != 0 ||
+	     rt.info.fmask_compression_enable || rt.info.fmask_data_compression_disable ||
+	     rt.info.fmask_one_frag_mode || rt.info.cmask_fast_clear_enable ||
+	     rt.info.dcc_compression_enable || rt.info.cmask_is_linear != 0 ||
+	     rt.info.cmask_addr_type != 0 || rt.info.alt_tile_mode || rt.cmask.addr != 0 ||
+	     rt.fmask.addr != 0 || rt.dcc_addr.addr != 0 || rt.dcc.data_write_on_dcc_clear_to_reg)) {
 		EXIT("unsupported texture-tiled render target: addr=0x%016" PRIx64 " tile=%u"
 		     " dimension=%u depth=%u levels=%u layer=%u/%u samples=%u fragments=%u bpe=%u"
 		     " cmask=0x%016" PRIx64 " fmask=0x%016" PRIx64 " dcc=0x%016" PRIx64 "\n",
@@ -179,12 +184,18 @@ void RenderExecutor::ResolveRenderColorTarget(uint64_t submit_id, RenderCommandB
 		     rt.attrib.num_fragments, bytes_per_element, rt.cmask.addr, rt.fmask.addr,
 		     rt.dcc_addr.addr);
 	}
+	if ((standard64 || depth_tile) && (rt.attrib3.dimension != 1 || rt.attrib3.depth != 0 ||
+	                                   view.base_layer != 0 || view.image_layers != 1)) {
+		EXIT("unsupported 64KB texture-tiled render-target view: dimension=%u depth=%u"
+		     " layer=%u/%u\n",
+		     rt.attrib3.dimension, rt.attrib3.depth, view.base_layer, view.image_layers);
+	}
 	if (rt.pitch.pitch_div8_minus1 != 0) {
 		pitch = (rt.pitch.pitch_div8_minus1 + 1u) << 3u;
 	} else if (tile) {
 		if (volume) {
 			pitch = TileGetTexturePitch(transfer_format, width, rt.attrib3.tile_mode);
-		} else if (standard64 || depth_tile) {
+		} else if (texture_tile) {
 			pitch = TileGetTexturePitch(transfer_format, width, rt.attrib3.tile_mode);
 		} else {
 			pitch = TileGetRenderTargetPitch(width, bytes_per_element, rt.attrib.num_fragments);
@@ -218,10 +229,10 @@ void RenderExecutor::ResolveRenderColorTarget(uint64_t submit_id, RenderCommandB
 	} else if (tile) {
 		TileSizeAlign layout {};
 		bool          valid_layout = false;
-		if (standard64 || depth_tile) {
+		if (texture_tile) {
 			TileGetTextureSize(transfer_format, width, height, levels, rt.attrib3.tile_mode,
 			                   &layout, mip_sizes, mip_padded);
-			valid_layout = layout.size != 0 && layout.align == 65536;
+			valid_layout = layout.size != 0 && layout.align == texture_tile_layout.block.block_size;
 		} else {
 			valid_layout =
 			    levels == 1 ? TileGetRenderTargetSize(width, height, pitch, bytes_per_element,

tests/ShaderRecompilerComputeTests.cpp

@@ -6266,6 +6266,109 @@ public:
     std::printf("[gpu]     %-32s ok\n", name);
   }
 
+  void CheckRenderExecutorColorStandard4TileDiscovery() {
+    constexpr const char *name = "RenderExecutorColorStandard4Tile";
+    constexpr uintptr_t base = 0x0000000203b00000ull;
+    constexpr uint64_t target_address = base + 0x4000;
+    constexpr uint64_t target_size = 0x180000;
+    constexpr uint64_t allocation_size = 0x184000;
+    constexpr uint64_t allocation_alignment = 0x4000;
+    constexpr auto format = Prospero::BufferFormat::k32Float;
+    constexpr auto tile = Prospero::TileMode::kStandard4KB;
+    EnsureRuntimeContext();
+
+    const auto pitch = TileGetTexturePitch(format, 256, tile);
+    TileSizeAlign slice{};
+    TileGetTextureSize(format, 256, 256, 1, tile, &slice, nullptr, nullptr);
+    const auto odd_pitch = TileGetTexturePitch(format, 257, tile);
+    TileSizeAlign odd_slice{};
+    TileGetTextureSize(format, 257, 131, 1, tile, &odd_slice, nullptr, nullptr);
+    Require(name, "tile layout",
+            pitch == 256 && slice.size == 0x40000 && slice.align == 0x1000 &&
+                odd_pitch == 288 && odd_slice.size == 0x2d000 &&
+                odd_slice.align == 0x1000,
+            "Standard4KB pitch, footprint, or block alignment is incorrect");
+
+    int64_t direct_offset = -1;
+    Require(name, "direct allocation",
+            Libs::LibKernel::Memory::KernelAllocateDirectMemory(
+                0, Libs::LibKernel::Memory::KernelGetDirectMemorySize(),
+                allocation_size, allocation_alignment, 0, &direct_offset) == 0,
+            "Standard4KB color allocation failed");
+    void *mapped = reinterpret_cast<void *>(base);
+    Require(name, "direct mapping",
+            Libs::LibKernel::Memory::KernelMapDirectMemory(
+                &mapped, allocation_size, 0x3, 0x10, direct_offset,
+                allocation_alignment) == 0 &&
+                mapped == reinterpret_cast<void *>(base),
+            "Standard4KB color mapping failed");
+    std::memset(mapped, 0, allocation_size);
+
+    {
+      RenderContext context(m_runtime_context);
+      auto &scheduler = context.GetCommandScheduler();
+      HW::Context registers{};
+      HW::UserConfig user_config{};
+      HW::Shader shaders{};
+      registers.SetColorBase(0, {.addr = target_address});
+      registers.SetColorInfo(
+          0, {.format = Prospero::ChannelLayout::k32,
+              .channel_type = Prospero::ChannelType::kFloat,
+              .channel_order = Prospero::ChannelOrder::kStandard});
+      registers.SetColorView(
+          0, {.base_array_slice_index = 5, .last_array_slice_index = 5});
+      registers.SetColorAttrib2(0, {.height = 255, .width = 255});
+      registers.SetColorAttrib3(0, {.tile_mode = tile,
+                                    .dimension = 1,
+                                    .cmask_pipe_aligned = true,
+                                    .dcc_pipe_aligned = true});
+      registers.SetRenderTargetMask(0x0f);
+      scheduler.Begin(registers, user_config, shaders);
+
+      auto &resources = context.GetGpuResources();
+      auto &texture_cache = resources.GetTextureCache();
+      auto &executor = context.GetRenderExecutor();
+      resources.MapMemory(base, allocation_size);
+
+      RenderColorInfo color{};
+      RenderExecutorTestAccess::ResolveRenderColorTarget(
+          executor, 1, scheduler.Current(), color, 0);
+      const auto attachment =
+          texture_cache.FindRenderTarget(color.image_id, color.desc);
+      const auto &image = texture_cache.GetImage(color.image_id);
+      Require(name, "captured target",
+              color.image_id && attachment != nullptr &&
+                  color.base_array_layer == 5 &&
+                  color.desc.info.data.address == target_address &&
+                  color.desc.info.data.size == target_size &&
+                  color.desc.info.tile_mode == tile &&
+                  color.desc.info.extent == vk::Extent3D{256, 256, 1} &&
+                  color.desc.info.resources == ImageSubresources{1, 6} &&
+                  color.desc.info.pitch == pitch &&
+                  color.desc.info.mip_layout[0].offset == 0 &&
+                  color.desc.info.mip_layout[0].size == target_size &&
+                  color.desc.view_info.type == vk::ImageViewType::e2D &&
+                  color.desc.view_info.base_layer == 5 &&
+                  color.desc.view_info.layer_count == 1 &&
+                  (target_address & 0xffffu) == 0x4000 &&
+                  image.usage.render_target && image.IsGpuModified(),
+              "slice 5 Standard4KB render target was not resolved");
+
+      RenderExecutorTestAccess::ResetBindings(executor);
+      resources.UnmapMemory(base, allocation_size);
+      scheduler.Finish();
+    }
+
+    Require(name, "unmap direct backing",
+            Libs::LibKernel::Memory::KernelMunmap(base, allocation_size) == 0,
+            "Standard4KB color mapping release failed");
+    Require(name, "release direct backing",
+            Libs::LibKernel::Memory::KernelReleaseDirectMemory(
+                direct_offset, allocation_size) == 0,
+            "Standard4KB color allocation release failed");
+    std::printf("[gpu]     %-32s ok\n", name);
+  }
+
   void CheckRenderExecutorColorDepthTileDiscovery() {
     constexpr const char *name = "RenderExecutorColorDepthTile";
     constexpr uintptr_t base = 0x0000000203d00000ull;
@@ -21177,6 +21280,11 @@ int main(int argc, char **argv) {
     CheckStandard64RenderTargetTileRoundTrip();
     return 0;
   }
+  if (argc == 2 && std::strcmp(argv[1], "--standard4-rt-only") == 0) {
+    VulkanHarness vulkan;
+    vulkan.CheckRenderExecutorColorStandard4TileDiscovery();
+    return 0;
+  }
   if (argc == 2 && std::strcmp(argv[1], "--image-view-only") == 0) {
     VulkanHarness vulkan;
     CheckSampledColorViews();
@@ -21286,6 +21394,7 @@ int main(int argc, char **argv) {
   vulkan.CheckGpuTilerCpuParity();
 #if KYTY_PLATFORM == KYTY_PLATFORM_WINDOWS
   vulkan.CheckRenderExecutorColorVolumeDiscovery();
+  vulkan.CheckRenderExecutorColorStandard4TileDiscovery();
   vulkan.CheckRenderExecutorColorDepthTileDiscovery();
   vulkan.CheckRenderExecutorStencilBindingDiscovery();
   vulkan.CheckUnifiedTextureCacheFlow();