feat(qy): enable QY CUDA runtime device backend

CMakeLists.txt

@@ -39,6 +39,7 @@ option(WITH_MARS "Enable Mars backend" OFF)
 option(WITH_CAMBRICON "Enable Cambricon backend" OFF)
 option(WITH_MOORE "Enable Moore backend" OFF)
 option(WITH_ASCEND "Enable Ascend backend" OFF)
+option(WITH_QY "Enable Denglin QY GPU backend" OFF)
 
 option(AUTO_DETECT_DEVICES "Automatically detect available devices" OFF)
 option(INFINI_RT_BUILD_TESTING "Build InfiniRT tests" OFF)
@@ -121,9 +122,24 @@ if(AUTO_DETECT_DEVICES)
         message(STATUS "Auto-detected Ascend environment.")
     endif()
 
+    set(_qy_root_detect "")
+    foreach(_qy_env QY_ROOT QY_HOME QY_PATH)
+        if(NOT _qy_root_detect AND DEFINED ENV{${_qy_env}} AND
+           NOT "$ENV{${_qy_env}}" STREQUAL "")
+            set(_qy_root_detect "$ENV{${_qy_env}}")
+        endif()
+    endforeach()
+    if(NOT _qy_root_detect AND EXISTS "/usr/local/denglin/sdk/bin/dlcc")
+        set(_qy_root_detect "/usr/local/denglin")
+    endif()
+    if(_qy_root_detect AND EXISTS "${_qy_root_detect}/sdk/bin/dlcc")
+        set(WITH_QY ON)
+        message(STATUS "Auto-detected Denglin QY environment from ${_qy_root_detect}.")
+    endif()
+
     if(WITH_NVIDIA)
         set(_non_nvidia_gpu_detected FALSE)
-        foreach(_gpu_backend WITH_ILUVATAR WITH_HYGON WITH_THEAD WITH_METAX WITH_MARS WITH_CAMBRICON WITH_MOORE WITH_ASCEND)
+        foreach(_gpu_backend WITH_ILUVATAR WITH_HYGON WITH_THEAD WITH_METAX WITH_MARS WITH_CAMBRICON WITH_MOORE WITH_ASCEND WITH_QY)
             if(${_gpu_backend})
                 set(_non_nvidia_gpu_detected TRUE)
             endif()
@@ -140,7 +156,7 @@ include_directories(${CMAKE_CURRENT_SOURCE_DIR}/src)
 
 # Only one CUDA-like GPU backend can be enabled at a time.
 set(_gpu_backend_count 0)
-foreach(_gpu_backend WITH_NVIDIA WITH_ILUVATAR WITH_HYGON WITH_THEAD WITH_METAX WITH_MARS WITH_MOORE WITH_CAMBRICON WITH_ASCEND)
+foreach(_gpu_backend WITH_NVIDIA WITH_ILUVATAR WITH_HYGON WITH_THEAD WITH_METAX WITH_MARS WITH_MOORE WITH_CAMBRICON WITH_ASCEND WITH_QY)
     if(${_gpu_backend})
         math(EXPR _gpu_backend_count "${_gpu_backend_count} + 1")
     endif()
@@ -348,8 +364,39 @@ if(WITH_ASCEND)
     message(STATUS "Using Ascend from `${ASCEND_HOME}`.")
 endif()
 
+if(WITH_QY)
+    add_compile_definitions(WITH_QY=1)
+
+    set(QY_ROOT "")
+    foreach(_qy_env QY_ROOT QY_HOME QY_PATH)
+        if(NOT QY_ROOT AND DEFINED ENV{${_qy_env}} AND NOT "$ENV{${_qy_env}}" STREQUAL "")
+            set(QY_ROOT "$ENV{${_qy_env}}")
+        endif()
+    endforeach()
+    if(NOT QY_ROOT AND EXISTS "/usr/local/denglin/sdk/bin/dlcc")
+        set(QY_ROOT "/usr/local/denglin")
+    endif()
+    if(NOT QY_ROOT OR NOT EXISTS "${QY_ROOT}/sdk/bin/dlcc")
+        message(FATAL_ERROR
+            "`WITH_QY` is `ON` but Denglin QY SDK was not found. "
+            "Set `QY_ROOT` (expected `${QY_ROOT}/sdk/bin/dlcc`).")
+    endif()
+    set(QY_ROOT "${QY_ROOT}" CACHE PATH "Denglin QY SDK root" FORCE)
+    set(QY_SDK "${QY_ROOT}/sdk")
+
+    # InfiniRT QY is header-only against the Denglin CUDA-compatible runtime.
+    # Link `curt` directly (MetaX-style) rather than requiring CMake CUDA + nvcc.
+    include_directories("${QY_SDK}/include")
+    link_directories("${QY_SDK}/lib")
+    find_library(QY_CURT_LIB NAMES curt HINTS "${QY_SDK}/lib" REQUIRED)
+    set(ENV{QY_ROOT} "${QY_ROOT}")
+    set(ENV{CUDA_HOME} "${QY_SDK}")
+    set(ENV{CUDA_PATH} "${QY_SDK}")
+    message(STATUS "QY: SDK ${QY_SDK}, runtime ${QY_CURT_LIB}")
+endif()
+
 # If all other platforms are not enabled, CPU is enabled by default.
-if(NOT WITH_NVIDIA AND NOT WITH_ILUVATAR AND NOT WITH_HYGON AND NOT WITH_THEAD AND NOT WITH_METAX AND NOT WITH_MARS AND NOT WITH_MOORE AND NOT WITH_CAMBRICON AND NOT WITH_ASCEND)
+if(NOT WITH_NVIDIA AND NOT WITH_ILUVATAR AND NOT WITH_HYGON AND NOT WITH_THEAD AND NOT WITH_METAX AND NOT WITH_MARS AND NOT WITH_MOORE AND NOT WITH_CAMBRICON AND NOT WITH_ASCEND AND NOT WITH_QY)
     set(WITH_CPU ON)
     add_compile_definitions(WITH_CPU=1)
 endif()
@@ -385,6 +432,9 @@ endif()
 if(WITH_ASCEND)
     list(APPEND INFINI_RT_PUBLIC_HEADER_DEVICES ascend)
 endif()
+if(WITH_QY)
+    list(APPEND INFINI_RT_PUBLIC_HEADER_DEVICES qy)
+endif()
 
 find_package(Python COMPONENTS Interpreter REQUIRED)
 execute_process(
@@ -408,7 +458,7 @@ if(WITH_NVIDIA OR WITH_ILUVATAR OR WITH_HYGON OR WITH_THEAD)
 endif()
 
 set(INFINI_RT_PACKAGE_VENDOR_BACKEND "")
-foreach(_backend METAX MARS MOORE CAMBRICON ASCEND)
+foreach(_backend METAX MARS MOORE CAMBRICON ASCEND QY)
     if(WITH_${_backend})
         set(INFINI_RT_PACKAGE_VENDOR_BACKEND "${_backend}")
     endif()

cmake/InfiniRTConfig.cmake.in

@@ -15,6 +15,7 @@ set(InfiniRT_WITH_MARS @WITH_MARS@)
 set(InfiniRT_WITH_MOORE @WITH_MOORE@)
 set(InfiniRT_WITH_CAMBRICON @WITH_CAMBRICON@)
 set(InfiniRT_WITH_ASCEND @WITH_ASCEND@)
+set(InfiniRT_WITH_QY @WITH_QY@)
 
 if(@INFINI_RT_PACKAGE_NEEDS_CUDATOOLKIT@)
     find_dependency(CUDAToolkit)
@@ -186,6 +187,32 @@ elseif(_InfiniRT_VENDOR_BACKEND STREQUAL "ASCEND")
     _infinirt_define_backend_runtime(
         "${_InfiniRT_ASCEND_INCLUDE_DIRS}"
         "${_InfiniRT_BACKEND_RUNTIME_LIBRARY};${_InfiniRT_BACKEND_DRIVER_LIBRARY};${CMAKE_DL_LIBS}")
+elseif(_InfiniRT_VENDOR_BACKEND STREQUAL "QY")
+    set(_InfiniRT_ROOT_HINTS "${InfiniRT_QY_ROOT}")
+    foreach(_InfiniRT_ENV QY_ROOT QY_HOME QY_PATH)
+        if(DEFINED ENV{${_InfiniRT_ENV}})
+            list(APPEND _InfiniRT_ROOT_HINTS "$ENV{${_InfiniRT_ENV}}")
+        endif()
+    endforeach()
+    list(APPEND _InfiniRT_ROOT_HINTS /usr/local/denglin)
+    find_path(_InfiniRT_BACKEND_INCLUDE_DIR
+        NAMES cuda_runtime.h
+        HINTS ${_InfiniRT_ROOT_HINTS}
+        PATH_SUFFIXES sdk/include include)
+    find_library(_InfiniRT_BACKEND_RUNTIME_LIBRARY
+        NAMES curt
+        HINTS ${_InfiniRT_ROOT_HINTS}
+        PATH_SUFFIXES sdk/lib lib lib64)
+    if(NOT _InfiniRT_BACKEND_INCLUDE_DIR OR
+       NOT _InfiniRT_BACKEND_RUNTIME_LIBRARY)
+        set(InfiniRT_FOUND FALSE)
+        set(InfiniRT_NOT_FOUND_MESSAGE
+            "Denglin QY SDK not found. Set InfiniRT_QY_ROOT or QY_ROOT.")
+        return()
+    endif()
+    _infinirt_define_backend_runtime(
+        "${_InfiniRT_BACKEND_INCLUDE_DIR}"
+        "${_InfiniRT_BACKEND_RUNTIME_LIBRARY}")
 endif()
 
 include("${CMAKE_CURRENT_LIST_DIR}/InfiniRTTargets.cmake")

scripts/dlcc_wrapper.sh

@@ -0,0 +1,78 @@
+#!/bin/bash
+# Filter CMake/host flags unsupported or harmful when invoking Denglin `dlcc`
+# as CMAKE_CUDA_COMPILER. Emulate enough of `nvcc -v` for CMake 3.22 Clang
+# CUDA detection (TOP / NVVMIR_LIBRARY_DIR).
+set -euo pipefail
+
+QY_ROOT="${QY_ROOT:-${QY_HOME:-${QY_PATH:-/usr/local/denglin}}}"
+DLCC="${QY_ROOT}/sdk/bin/dlcc"
+SDK="${QY_ROOT}/sdk"
+
+if [[ ! -x "${DLCC}" ]]; then
+  # Container image layout.
+  if [[ -x /usr/local/dlgpu/sdk/bin/dlcc ]]; then
+    QY_ROOT=/usr/local/dlgpu
+    DLCC=/usr/local/dlgpu/sdk/bin/dlcc
+    SDK=/usr/local/dlgpu/sdk
+  else
+    echo "dlcc_wrapper: dlcc not found; set QY_ROOT" >&2
+    exit 1
+  fi
+fi
+
+# CMakeDetermineCUDACompiler runs `${CMAKE_CUDA_COMPILER} -v __cmake_determine_cuda`.
+if [[ "${1:-}" == "-v" ]]; then
+  echo "clang version 15.0.6"
+  echo "Cuda compilation tools, release 11.7, V11.7.0"
+  echo "#\$ TOP=${SDK}"
+  # Path need only match CMake's `nvvm/libdevice$` regex; Denglin has no nvvm tree.
+  echo "#\$ NVVMIR_LIBRARY_DIR=${SDK}/nvvm/libdevice"
+  "${DLCC}" --version || true
+  exit 0
+fi
+
+ARGS=()
+skip_next=0
+has_cuda_path=0
+replaced_cudart=0
+for arg in "$@"; do
+  if [[ "${skip_next}" -eq 1 ]]; then
+    skip_next=0
+    continue
+  fi
+  case "${arg}" in
+    -pthread) ;;
+    -B) skip_next=1 ;;
+    -B*) ;;
+    --cuda-path=*)
+      has_cuda_path=1
+      ARGS+=("${arg}")
+      ;;
+    # CMake Clang detection probes sm_20/30/52; Denglin only accepts dlgput64.
+    --cuda-gpu-arch=sm_*) ;;
+    --cuda-gpu-arch=dlgput64)
+      ARGS+=("${arg}")
+      ;;
+    --cuda-gpu-arch=*) ;;
+    -Xcompiler=*) ;;
+    # Match pre-refactor xmake/qy.lua: drop NVIDIA static CUDA libs.
+    -lcudadevrt|-lcudart_static|-lcudart)
+      if [[ "${replaced_cudart}" -eq 0 ]]; then
+        ARGS+=("-lcurt")
+        replaced_cudart=1
+      fi
+      ;;
+    *)
+      ARGS+=("${arg}")
+      ;;
+  esac
+done
+
+EXTRA=()
+if [[ "${has_cuda_path}" -eq 0 ]]; then
+  EXTRA+=("--cuda-path=${SDK}")
+fi
+# Always compile for Denglin GPU; ignore CMake sm_* probes.
+EXTRA+=("--cuda-gpu-arch=dlgput64")
+
+exec "${DLCC}" "${EXTRA[@]}" "${ARGS[@]}"

scripts/generate_public_headers.py

@@ -57,6 +57,11 @@
         ("ascend", "device_.h", "native/ascend/device_.h"),
         ("ascend", "runtime_.h", "native/ascend/runtime_.h"),
     ),
+    "qy": (
+        ("qy", "data_type_.h", "native/cuda/qy/data_type_.h"),
+        ("qy", "device_.h", "native/cuda/qy/device_.h"),
+        ("qy", "runtime_.h", "native/cuda/qy/runtime_.h"),
+    ),
 }
 
 _DEVICE_TYPES = {
@@ -70,6 +75,7 @@
     "moore": "Device::Type::kMoore",
     "cambricon": "Device::Type::kCambricon",
     "ascend": "Device::Type::kAscend",
+    "qy": "Device::Type::kQy",
 }
 
 _DEFAULT_DEVICE_PRIORITY = (
@@ -82,6 +88,7 @@
     "moore",
     "cambricon",
     "ascend",
+    "qy",
     "cpu",
 )
 

scripts/run_performance_tests.py

@@ -15,6 +15,7 @@
     ("WITH_MOORE", "moore"),
     ("WITH_CAMBRICON", "cambricon"),
     ("WITH_ASCEND", "ascend"),
+    ("WITH_QY", "qy"),
     ("WITH_CPU", "cpu"),
 )
 

src/CMakeLists.txt

@@ -105,6 +105,13 @@ if(WITH_CAMBRICON)
         "${CAMBRICON_RUNTIME_LIB}")
 endif()
 
+if(WITH_QY)
+    target_compile_definitions(infinirt PUBLIC WITH_QY=1)
+    _infinirt_use_backend_runtime(
+        "${QY_SDK}/include"
+        "${QY_CURT_LIB}")
+endif()
+
 if(WITH_ASCEND)
     # ASCEND_HOME is set by the top-level CMakeLists.txt.
     target_compile_definitions(infinirt PUBLIC WITH_ASCEND=1)

src/device.h

@@ -23,6 +23,7 @@ class Device {
     kHygon = 7,
     kThead = 8,
     kMars = 9,
+    kQy = 10,
     kCount
   };
 
@@ -68,6 +69,7 @@ class Device {
           {Type::kHygon, "hygon"},
           {Type::kThead, "thead"},
           {Type::kMars, "mars"},
+          {Type::kQy, "qy"},
       }}};
 
   static constexpr ConstexprMap<std::string_view, Device::Type,
@@ -83,6 +85,7 @@ class Device {
           {"hygon", Type::kHygon},
           {"thead", Type::kThead},
           {"mars", Type::kMars},
+          {"qy", Type::kQy},
       }}};
 
   int index_{0};
@@ -98,7 +101,7 @@ using AllDeviceTypes =
     List<Device::Type::kCpu, Device::Type::kNvidia, Device::Type::kCambricon,
          Device::Type::kAscend, Device::Type::kMetax, Device::Type::kMoore,
          Device::Type::kIluvatar, Device::Type::kHygon, Device::Type::kThead,
-         Device::Type::kMars>;
+         Device::Type::kMars, Device::Type::kQy>;
 
 // Deferred computation of active devices. The `Filter` and `FilterList`
 // evaluation are nested inside a class template so that `DeviceEnabled`

src/native/cuda/qy/data_type_.h

@@ -0,0 +1,30 @@
+#ifndef INFINI_RT_QY_DATA_TYPE__H_
+#define INFINI_RT_QY_DATA_TYPE__H_
+
+// clang-format off
+#include <cuda_bf16.h>
+#include <cuda_fp16.h>
+// clang-format on
+
+#include "data_type.h"
+#include "native/cuda/qy/device_.h"
+
+namespace infini::rt {
+
+using cuda_bfloat16 = nv_bfloat16;
+
+using cuda_bfloat162 = nv_bfloat162;
+
+template <>
+struct TypeMap<Device::Type::kQy, DataType::kFloat16> {
+  using type = half;
+};
+
+template <>
+struct TypeMap<Device::Type::kQy, DataType::kBFloat16> {
+  using type = __nv_bfloat16;
+};
+
+}  // namespace infini::rt
+
+#endif

src/native/cuda/qy/device_.h

@@ -0,0 +1,13 @@
+#ifndef INFINI_RT_QY_DEVICE__H_
+#define INFINI_RT_QY_DEVICE__H_
+
+#include "device.h"
+
+namespace infini::rt {
+
+template <>
+struct DeviceEnabled<Device::Type::kQy> : std::true_type {};
+
+}  // namespace infini::rt
+
+#endif