[CUDA] remove src/treelearner/kernels (#6766)

* remove src/treelearner/kernels

* Update CMakeLists.txt

* clean up

CMakeLists.txt

@@ -252,54 +252,6 @@ if(USE_CUDA)
       set(CMAKE_CUDA_STANDARD 11)
       set(CMAKE_CUDA_STANDARD_REQUIRED ON)
     endif()
-
-    set(
-      BASE_DEFINES
-      -DPOWER_FEATURE_WORKGROUPS=12
-      -DUSE_CONSTANT_BUF=0
-    )
-    set(
-      ALLFEATS_DEFINES
-      ${BASE_DEFINES}
-      -DENABLE_ALL_FEATURES
-    )
-    set(
-      FULLDATA_DEFINES
-      ${ALLFEATS_DEFINES}
-      -DIGNORE_INDICES
-    )
-
-    message(STATUS "ALLFEATS_DEFINES: ${ALLFEATS_DEFINES}")
-    message(STATUS "FULLDATA_DEFINES: ${FULLDATA_DEFINES}")
-
-    function(add_histogram hsize hname hadd hconst hdir)
-      add_library(histo${hsize}${hname} OBJECT src/treelearner/kernels/histogram${hsize}.cu)
-      set_target_properties(
-        histo${hsize}${hname}
-        PROPERTIES
-          CUDA_SEPARABLE_COMPILATION ON
-          CUDA_ARCHITECTURES ${CUDA_ARCHS}
-      )
-      if(hadd)
-        list(APPEND histograms histo${hsize}${hname})
-        set(histograms ${histograms} PARENT_SCOPE)
-      endif()
-      target_compile_definitions(
-        histo${hsize}${hname}
-        PRIVATE
-        -DCONST_HESSIAN=${hconst}
-        ${hdir}
-      )
-    endfunction()
-
-    foreach(hsize _16_64_256)
-      add_histogram("${hsize}" "_sp_const" "True" "1" "${BASE_DEFINES}")
-      add_histogram("${hsize}" "_sp" "True" "0" "${BASE_DEFINES}")
-      add_histogram("${hsize}" "-allfeats_sp_const" "False" "1" "${ALLFEATS_DEFINES}")
-      add_histogram("${hsize}" "-allfeats_sp" "False" "0" "${ALLFEATS_DEFINES}")
-      add_histogram("${hsize}" "-fulldata_sp_const" "True" "1" "${FULLDATA_DEFINES}")
-      add_histogram("${hsize}" "-fulldata_sp" "True" "0" "${FULLDATA_DEFINES}")
-    endforeach()
 endif()
 
 include(CheckCXXSourceCompiles)
@@ -634,14 +586,6 @@ if(USE_CUDA)
         CUDA_RESOLVE_DEVICE_SYMBOLS ON
     )
   endif()
-
-  # histograms are list of object libraries. Linking object library to other
-  # object libraries only gets usage requirements, the linked objects won't be
-  # used. Thus we have to call target_link_libraries on final targets here.
-  if(BUILD_CLI)
-    target_link_libraries(lightgbm PRIVATE ${histograms})
-  endif()
-  target_link_libraries(_lightgbm PRIVATE ${histograms})
 endif()
 
 if(WIN32)

src/treelearner/kernels/histogram_16_64_256.hu

-/*!
- * Copyright (c) 2020 IBM Corporation. All rights reserved.
- * Licensed under the MIT License. See LICENSE file in the project root for license information.
- */
-
-#ifndef LIGHTGBM_TREELEARNER_KERNELS_HISTOGRAM_16_64_256_HU_
-#define LIGHTGBM_TREELEARNER_KERNELS_HISTOGRAM_16_64_256_HU_
-
-#include "LightGBM/meta.h"
-
-namespace LightGBM {
-
-// use double precision or not
-#ifndef USE_DP_FLOAT
-#define USE_DP_FLOAT 1
-#endif
-
-// ignore hessian, and use the local memory for hessian as an additional bank for gradient
-#ifndef CONST_HESSIAN
-#define CONST_HESSIAN 0
-#endif
-
-typedef unsigned char uchar;
-
-template<typename T>
-__device__ double as_double(const T t) {
-  static_assert(sizeof(T) == sizeof(double), "size mismatch");
-  double d;
-  memcpy(&d, &t, sizeof(T));
-  return d;
-}
-template<typename T>
-__device__ unsigned long long as_ulong_ulong(const T t) {
-  static_assert(sizeof(T) == sizeof(unsigned long long), "size mismatch");
-  unsigned long long u;
-  memcpy(&u, &t, sizeof(T));
-  return u;
-}
-template<typename T>
-__device__ float as_float(const T t) {
-  static_assert(sizeof(T) == sizeof(float), "size mismatch");
-  float f;
-  memcpy(&f, &t, sizeof(T));
-  return f;
-}
-template<typename T>
-__device__ unsigned int as_uint(const T t) {
-  static_assert(sizeof(T) == sizeof(unsigned int), "size_mismatch");
-  unsigned int u;
-  memcpy(&u, &t, sizeof(T));
-  return u;
-}
-template<typename T>
-__device__ uchar4 as_uchar4(const T t) {
-  static_assert(sizeof(T) == sizeof(uchar4), "size mismatch");
-  uchar4 u;
-  memcpy(&u, &t, sizeof(T));
-  return u;
-}
-
-#if USE_DP_FLOAT == 1
-typedef double acc_type;
-typedef unsigned long long acc_int_type;
-#define as_acc_type as_double
-#define as_acc_int_type as_ulong_ulong
-#else
-typedef float acc_type;
-typedef unsigned int acc_int_type;
-#define as_acc_type as_float
-#define as_acc_int_type as_uint
-#endif
-
-// use all features and do not use feature mask
-#ifndef ENABLE_ALL_FEATURES
-#define ENABLE_ALL_FEATURES 1
-#endif
-
-// define all of the different kernels
-
-#define DECLARE_CONST_BUF(name) \
-__global__ void name(__global const uchar* restrict feature_data_base, \
-                     const uchar* restrict feature_masks,\
-                     const data_size_t feature_size,\
-                     const data_size_t* restrict data_indices, \
-                     const data_size_t num_data, \
-                     const score_t* restrict ordered_gradients, \
-                     const score_t* restrict ordered_hessians,\
-                     char* __restrict__ output_buf,\
-                     volatile int * sync_counters,\
-                     acc_type* __restrict__ hist_buf_base, \
-                     const size_t power_feature_workgroups);
-
-
-#define DECLARE_CONST_HES_CONST_BUF(name) \
-__global__ void name(const uchar* __restrict__ feature_data_base, \
-                     const uchar* __restrict__ feature_masks,\
-                     const data_size_t feature_size,\
-                     const data_size_t* __restrict__ data_indices, \
-                     const data_size_t num_data, \
-                     const score_t* __restrict__ ordered_gradients, \
-                     const score_t const_hessian,\
-                     char* __restrict__ output_buf,\
-                     volatile int * sync_counters,\
-                     acc_type* __restrict__ hist_buf_base, \
-                     const size_t power_feature_workgroups);
-
-
-
-#define DECLARE_CONST_HES(name) \
-__global__ void name(const uchar* feature_data_base, \
-                     const uchar* __restrict__ feature_masks,\
-                     const data_size_t feature_size,\
-                     const data_size_t* data_indices, \
-                     const data_size_t num_data, \
-                     const score_t*  ordered_gradients, \
-                     const score_t const_hessian,\
-                     char* __restrict__ output_buf, \
-                     volatile int * sync_counters,\
-                     acc_type* __restrict__ hist_buf_base, \
-                     const size_t power_feature_workgroups);
-
-
-#define DECLARE(name) \
-__global__ void name(const uchar* feature_data_base, \
-                     const uchar* __restrict__ feature_masks,\
-                     const data_size_t feature_size,\
-                     const data_size_t* data_indices, \
-                     const data_size_t num_data, \
-                     const score_t*  ordered_gradients, \
-                     const score_t*  ordered_hessians,\
-                     char* __restrict__ output_buf, \
-                     volatile int * sync_counters,\
-                     acc_type* __restrict__ hist_buf_base, \
-                     const size_t power_feature_workgroups);
-
-
-DECLARE_CONST_HES(histogram16_allfeats);
-DECLARE_CONST_HES(histogram16_fulldata);
-DECLARE_CONST_HES(histogram16);
-DECLARE(histogram16_allfeats);
-DECLARE(histogram16_fulldata);
-DECLARE(histogram16);
-
-DECLARE_CONST_HES(histogram64_allfeats);
-DECLARE_CONST_HES(histogram64_fulldata);
-DECLARE_CONST_HES(histogram64);
-DECLARE(histogram64_allfeats);
-DECLARE(histogram64_fulldata);
-DECLARE(histogram64);
-
-DECLARE_CONST_HES(histogram256_allfeats);
-DECLARE_CONST_HES(histogram256_fulldata);
-DECLARE_CONST_HES(histogram256);
-DECLARE(histogram256_allfeats);
-DECLARE(histogram256_fulldata);
-DECLARE(histogram256);
-
-}  // namespace LightGBM
-
-#endif  // LIGHTGBM_TREELEARNER_KERNELS_HISTOGRAM_16_64_256_HU_