Image upload benchamark

73_ImageUploadBenchmark/CMakeLists.txt

@@ -0,0 +1,64 @@
+include(common RESULT_VARIABLE RES)
+if(NOT RES)
+	message(FATAL_ERROR "common.cmake not found. Should be in {repo_root}/cmake directory")
+endif()
+
+nbl_create_executable_project("" "" "" "" "${NBL_EXECUTABLE_PROJECT_CREATION_PCH_TARGET}")
+
+if(NBL_EMBED_BUILTIN_RESOURCES)
+	set(_BR_TARGET_ ${EXECUTABLE_NAME}_builtinResourceData)
+	set(RESOURCE_DIR "app_resources")
+
+	get_filename_component(_SEARCH_DIRECTORIES_ "${CMAKE_CURRENT_SOURCE_DIR}" ABSOLUTE)
+	get_filename_component(_OUTPUT_DIRECTORY_SOURCE_ "${CMAKE_CURRENT_BINARY_DIR}/src" ABSOLUTE)
+	get_filename_component(_OUTPUT_DIRECTORY_HEADER_ "${CMAKE_CURRENT_BINARY_DIR}/include" ABSOLUTE)
+
+	file(GLOB_RECURSE BUILTIN_RESOURCE_FILES RELATIVE "${CMAKE_CURRENT_SOURCE_DIR}/${RESOURCE_DIR}" CONFIGURE_DEPENDS "${CMAKE_CURRENT_SOURCE_DIR}/${RESOURCE_DIR}/*")
+	foreach(RES_FILE ${BUILTIN_RESOURCE_FILES})
+		LIST_BUILTIN_RESOURCE(RESOURCES_TO_EMBED "${RES_FILE}")
+	endforeach()
+
+	ADD_CUSTOM_BUILTIN_RESOURCES(${_BR_TARGET_} RESOURCES_TO_EMBED "${_SEARCH_DIRECTORIES_}" "${RESOURCE_DIR}" "nbl::this_example::builtin" "${_OUTPUT_DIRECTORY_HEADER_}" "${_OUTPUT_DIRECTORY_SOURCE_}")
+
+	LINK_BUILTIN_RESOURCES_TO_TARGET(${EXECUTABLE_NAME} ${_BR_TARGET_})
+endif()
+
+set(OUTPUT_DIRECTORY "${CMAKE_CURRENT_BINARY_DIR}/auto-gen")
+set(DEPENDS
+	app_resources/common.hlsl
+	app_resources/tile_upload.comp.hlsl
+)
+target_sources(${EXECUTABLE_NAME} PRIVATE ${DEPENDS})
+set_source_files_properties(${DEPENDS} PROPERTIES HEADER_FILE_ONLY ON)
+
+set(SM 6_8)
+set(JSON [=[
+[
+	{
+		"INPUT": "app_resources/tile_upload.comp.hlsl",
+		"KEY": "snakeStore"
+	}
+]
+]=])
+string(CONFIGURE "${JSON}" JSON)
+
+NBL_CREATE_NSC_COMPILE_RULES(
+	TARGET ${EXECUTABLE_NAME}SPIRV
+	LINK_TO ${EXECUTABLE_NAME}
+	DEPENDS ${DEPENDS}
+	BINARY_DIR ${OUTPUT_DIRECTORY}
+	MOUNT_POINT_DEFINE NBL_THIS_EXAMPLE_BUILD_MOUNT_POINT
+	COMMON_OPTIONS -I "${CMAKE_CURRENT_SOURCE_DIR}" -T lib_${SM}
+	OUTPUT_VAR KEYS
+	INCLUDE nbl/this_example/builtin/build/spirv/keys.hpp
+	NAMESPACE nbl::this_example::builtin::build
+	INPUTS ${JSON}
+)
+
+NBL_CREATE_RESOURCE_ARCHIVE(
+	NAMESPACE nbl::this_example::builtin::build
+	TARGET ${EXECUTABLE_NAME}_builtinsBuild
+	LINK_TO ${EXECUTABLE_NAME}
+	BIND ${OUTPUT_DIRECTORY}
+	BUILTINS ${KEYS}
+)

73_ImageUploadBenchmark/app_resources/common.hlsl

@@ -0,0 +1,11 @@
+#include <nbl/builtin/hlsl/morton.hlsl>
+
+struct PushConstantData
+{
+    uint64_t deviceBufferAddress;
+    uint64_t dstTileLocationsAddress;
+    uint32_t2 dstOffset;
+    uint32_t srcWidth;
+    uint32_t srcHeight;
+    uint32_t tilesPerRow;
+};

73_ImageUploadBenchmark/app_resources/tile_upload.comp.hlsl

@@ -0,0 +1,99 @@
+#include "common.hlsl"
+
+[[vk::binding(0,0)]] RWTexture2D<float32_t4> dstImage;
+[[vk::push_constant]] PushConstantData pc;
+
+using namespace nbl::hlsl;
+
+static const uint32_t TILE_SIZE = 128u;
+static const uint32_t TILE_SIZE_LOG2 = 7u;
+static const uint32_t TILE_SIZE_MASK = TILE_SIZE - 1u;
+static const uint32_t TILE_PIXELS_LOG2 = TILE_SIZE_LOG2 * 2u;
+static const uint32_t BLOCK_SIZE = 16u;
+static const uint32_t BLOCK_SIZE_LOG2 = 4u;
+static const uint32_t BLOCK_PIXELS_LOG2 = BLOCK_SIZE_LOG2 * 2u;
+static const uint32_t BLOCKS_PER_TILE_LOG2 = TILE_SIZE_LOG2 - BLOCK_SIZE_LOG2;
+static const uint32_t BLOCKS_PER_TILE = TILE_SIZE / BLOCK_SIZE;
+
+[numthreads(128, 4, 1)]
+[shader("compute")]
+void SnakeStore(uint32_t3 ID : SV_DispatchThreadID)
+{
+    const uint32_t2 globalPos = ID.xy;
+    const uint32_t2 tileCoord = globalPos >> TILE_SIZE_LOG2;
+    const uint32_t2 localPos = globalPos & TILE_SIZE_MASK;
+    const uint32_t tileIdx = tileCoord.y * pc.tilesPerRow + tileCoord.x;
+    const uint32_t localLinearIdx = (localPos.y << TILE_SIZE_LOG2) + localPos.x;
+    const uint32_t srcPixelIdx = (tileIdx << TILE_PIXELS_LOG2) + localLinearIdx;
+    const uint32_t packedDstTile = vk::RawBufferLoad<uint32_t>(pc.dstTileLocationsAddress + tileIdx * 4u);
+    const uint32_t2 dstTile = uint32_t2(packedDstTile & 0xffffu, packedDstTile >> 16u);
+    const uint32_t2 pixelPos = pc.dstOffset + (dstTile << TILE_SIZE_LOG2) + localPos;
+
+    const uint32_t packed = vk::RawBufferLoad<uint32_t>(pc.deviceBufferAddress + srcPixelIdx * 4u);
+
+    dstImage[pixelPos] = unpackUnorm4x8(int32_t(packed));
+}
+
+[numthreads(128, 4, 1)]
+[shader("compute")]
+void SnakeLoad(uint32_t3 ID : SV_DispatchThreadID)
+{
+    const uint32_t2 globalPos = ID.xy;
+    const uint32_t2 tileCoord = globalPos >> TILE_SIZE_LOG2;
+    const uint32_t2 localPos = globalPos & TILE_SIZE_MASK;
+    const uint32_t tileIdx = tileCoord.y * pc.tilesPerRow + tileCoord.x;
+    const uint32_t localLinearIdx = (localPos.y << TILE_SIZE_LOG2) + localPos.x;
+    const uint32_t dstPixelIdx = (tileIdx << TILE_PIXELS_LOG2) + localLinearIdx;
+    const uint32_t packedDstTile = vk::RawBufferLoad<uint32_t>(pc.dstTileLocationsAddress + tileIdx * 4u);
+    const uint32_t2 dstTile = uint32_t2(packedDstTile & 0xffffu, packedDstTile >> 16u);
+    const uint32_t2 pixelPos = pc.dstOffset + (dstTile << TILE_SIZE_LOG2) + localPos;
+
+    vk::RawBufferStore<uint32_t>(pc.deviceBufferAddress + dstPixelIdx * 4u, uint32_t(packUnorm4x8(dstImage[pixelPos])));
+}
+
+[numthreads(16, 16, 1)]
+[shader("compute")]
+void MortonStore(uint32_t3 ID : SV_GroupThreadID, uint32_t3 GroupID : SV_GroupID)
+{
+    const uint32_t2 globalBlock = GroupID.xy;
+    const uint32_t2 threadPos = ID.xy;
+    const uint32_t2 tileCoord = globalBlock >> BLOCKS_PER_TILE_LOG2;
+    const uint32_t2 blockCoordInTile = globalBlock & (BLOCKS_PER_TILE - 1u);
+    const uint32_t tileIdx = tileCoord.y * pc.tilesPerRow + tileCoord.x;
+    const uint32_t blockIdxInTile = (blockCoordInTile.y << BLOCKS_PER_TILE_LOG2) + blockCoordInTile.x;
+    const uint32_t localLinearIdx = (threadPos.y << BLOCK_SIZE_LOG2) + threadPos.x;
+    const uint32_t srcPixelIdx = (tileIdx << TILE_PIXELS_LOG2) + (blockIdxInTile << BLOCK_PIXELS_LOG2) + localLinearIdx;
+    const uint32_t packedDstTile = vk::RawBufferLoad<uint32_t>(pc.dstTileLocationsAddress + tileIdx * 4u);
+    const uint32_t2 dstTile = uint32_t2(packedDstTile & 0xffffu, packedDstTile >> 16u);
+
+    morton::code<false, 4, 2> mc;
+    mc.value = uint16_t(localLinearIdx);
+    const uint32_t2 mortonLocalPos = _static_cast<uint32_t2>(mc);
+    const uint32_t2 pixelPos = pc.dstOffset + (dstTile << TILE_SIZE_LOG2) + (blockCoordInTile << BLOCK_SIZE_LOG2) + mortonLocalPos;
+
+    const uint32_t packed = vk::RawBufferLoad<uint32_t>(pc.deviceBufferAddress + srcPixelIdx * 4u);
+    dstImage[pixelPos] = unpackUnorm4x8(int32_t(packed));
+}
+
+[numthreads(16, 16, 1)]
+[shader("compute")]
+void MortonLoad(uint32_t3 ID : SV_GroupThreadID, uint32_t3 GroupID : SV_GroupID)
+{
+    const uint32_t2 globalBlock = GroupID.xy;
+    const uint32_t2 threadPos = ID.xy;
+    const uint32_t2 tileCoord = globalBlock >> BLOCKS_PER_TILE_LOG2;
+    const uint32_t2 blockCoordInTile = globalBlock & (BLOCKS_PER_TILE - 1u);
+    const uint32_t tileIdx = tileCoord.y * pc.tilesPerRow + tileCoord.x;
+    const uint32_t blockIdxInTile = (blockCoordInTile.y << BLOCKS_PER_TILE_LOG2) + blockCoordInTile.x;
+    const uint32_t localLinearIdx = (threadPos.y << BLOCK_SIZE_LOG2) + threadPos.x;
+    const uint32_t dstPixelIdx = (tileIdx << TILE_PIXELS_LOG2) + (blockIdxInTile << BLOCK_PIXELS_LOG2) + localLinearIdx;
+    const uint32_t packedDstTile = vk::RawBufferLoad<uint32_t>(pc.dstTileLocationsAddress + tileIdx * 4u);
+    const uint32_t2 dstTile = uint32_t2(packedDstTile & 0xffffu, packedDstTile >> 16u);
+
+    morton::code<false, 4, 2> mc;
+    mc.value = uint16_t(localLinearIdx);
+    const uint32_t2 mortonLocalPos = _static_cast<uint32_t2>(mc);
+    const uint32_t2 pixelPos = pc.dstOffset + (dstTile << TILE_SIZE_LOG2) + (blockCoordInTile << BLOCK_SIZE_LOG2) + mortonLocalPos;
+
+    vk::RawBufferStore<uint32_t>(pc.deviceBufferAddress + dstPixelIdx * 4u, uint32_t(packUnorm4x8(dstImage[pixelPos])));
+}

73_ImageUploadBenchmark/config.json.template

@@ -0,0 +1,28 @@
+{
+  "enableParallelBuild": true,
+  "threadsPerBuildProcess" : 2,
+  "isExecuted": false,
+  "scriptPath": "",
+  "cmake": {
+    "configurations": [ "Release", "Debug", "RelWithDebInfo" ],
+    "buildModes": [],
+    "requiredOptions": []
+  }, 
+  "profiles": [
+    {
+      "backend": "vulkan", // should be none
+      "platform": "windows",
+      "buildModes": [],
+      "runConfiguration": "Release", // we also need to run in Debug nad RWDI because foundational example
+      "gpuArchitectures": []
+    }
+  ],
+  "dependencies": [],
+  "data": [
+    {
+      "dependencies": [],
+      "command": [""],
+      "outputs": []
+    }
+  ]
+}