GPUDefParametersLoad.template.inc

// Copyright 2019-2020 CERN and copyright holders of ALICE O2.
// See https://alice-o2.web.cern.ch/copyright for details of the copyright holders.
// All rights not expressly granted are reserved.
//
// This software is distributed under the terms of the GNU General Public
// License v3 (GPL Version 3), copied verbatim in the file "COPYING".
//
// In applying this license CERN does not waive the privileges and immunities
// granted to it by virtue of its status as an Intergovernmental Organization
// or submit itself to any jurisdiction.

/// \file GPUDefParametersLoad.inc
/// \author David Rohr

#include "GPUDefParametersLoadPrepare.h"

#define GPUCA_M_LB_EMPTY_(...) 0
#define GPUCA_M_LB_EMPTY_1(...) __VA_ARGS__
#define GPUCA_M_LB_EMPTY0(...) GPUCA_M_CAT(GPUCA_M_LB_EMPTY_, __VA_OPT__(1))(__VA_ARGS__)

#include "GPUDefParametersRuntime.h"
#include "GPUDefMacros.h"
#include <sstream>
#include <string>

namespace o2::gpu::internal
{

static GPUDefParameters GPUDefParametersLoad()
{
  return GPUDefParameters{
    // clang-format off
    {$<JOIN:$<LIST:TRANSFORM,$<LIST:TRANSFORM,$<LIST:TRANSFORM,$<TARGET_PROPERTY:O2_GPU_KERNELS,O2_GPU_KERNEL_NAMES>,REPLACE,[^A-Za-z0-9]+,_>,PREPEND,GPUCA_M_LB_EMPTY0(GPUCA_M_FIRST(GPUCA_LB_>,APPEND,))>,$<COMMA>>},
    {$<JOIN:$<LIST:TRANSFORM,$<LIST:TRANSFORM,$<LIST:TRANSFORM,$<TARGET_PROPERTY:O2_GPU_KERNELS,O2_GPU_KERNEL_NAMES>,REPLACE,[^A-Za-z0-9]+,_>,PREPEND,GPUCA_M_FIRST(GPUCA_M_SHIFT(GPUCA_LB_>,APPEND,$<COMMA>0))>,$<COMMA>>},
    {$<JOIN:$<LIST:TRANSFORM,$<LIST:TRANSFORM,$<LIST:TRANSFORM,$<TARGET_PROPERTY:O2_GPU_KERNELS,O2_GPU_KERNEL_NAMES>,REPLACE,[^A-Za-z0-9]+,_>,PREPEND,GPUCA_M_FIRST(GPUCA_M_SHIFT(GPUCA_M_SHIFT(GPUCA_LB_>,APPEND,$<COMMA>0$<COMMA>0)))>,$<COMMA>>},
    $<JOIN:$<LIST:TRANSFORM,$<TARGET_PROPERTY:O2_GPU_KERNELS,O2_GPU_KERNEL_PARAMS>,PREPEND,GPUCA_PAR_>,$<COMMA>>,
    $<JOIN:$<LIST:TRANSFORM,$<LIST:TRANSFORM,$<TARGET_PROPERTY:O2_GPU_KERNELS,O2_GPU_KERNEL_STRING_PARAMS>,PREPEND,GPUCA_M_STR(GPUCA_PAR_>,APPEND,)>,$<COMMA>>
    // clang-format on
  };
}

#define GPUCA_EXPORT_KERNEL_LB(name)                                                                                                                     \
  if (par.par_LB_maxThreads[i] > 0) {                                                                                                                    \
    o << "#define GPUCA_LB_" GPUCA_M_STR(name) " " << par.par_LB_maxThreads[i];                                                                          \
    if (par.par_LB_minBlocks[i] > 0) {                                                                                                                   \
      o << ", " << GPUCA_CEIL_INT_DIV(par.par_LB_maxThreads[i] * par.par_LB_minBlocks[i], (minBlockFactor ? minBlockFactor : par.par_LB_maxThreads[i])); \
    }                                                                                                                                                    \
    if (!forRTC && par.par_LB_forceBlocks[i] > 0) {                                                                                                      \
      o << ", " << par.par_LB_forceBlocks[i];                                                                                                            \
    }                                                                                                                                                    \
    o << "\n";                                                                                                                                           \
  }                                                                                                                                                      \
  i++;

#define GPUCA_EXPORT_KERNEL_PARAM(name) \
  o << "#define GPUCA_PAR_" GPUCA_M_STR(name) " " << GPUCA_M_CAT(par.par_, name) << "\n";

static std::string GPUDefParametersExport(const GPUDefParameters& par, bool forRTC, int32_t minBlockFactor = 0)
{
  std::stringstream o; // clang-format off
  int32_t i = 0;
  $<JOIN:$<LIST:TRANSFORM,$<LIST:TRANSFORM,$<LIST:TRANSFORM,$<TARGET_PROPERTY:O2_GPU_KERNELS,O2_GPU_KERNEL_NAMES>,REPLACE,[^A-Za-z0-9]+,_>,PREPEND,GPUCA_EXPORT_KERNEL_LB(>,APPEND,)>,
  >
  $<JOIN:$<LIST:TRANSFORM,$<LIST:TRANSFORM,$<TARGET_PROPERTY:O2_GPU_KERNELS,O2_GPU_KERNEL_PARAMS>,PREPEND,GPUCA_EXPORT_KERNEL_PARAM(>,APPEND,)>,
  >
  $<JOIN:$<LIST:TRANSFORM,$<LIST:TRANSFORM,$<TARGET_PROPERTY:O2_GPU_KERNELS,O2_GPU_KERNEL_STRING_PARAMS>,PREPEND,GPUCA_EXPORT_KERNEL_PARAM(>,APPEND,)>,
  >
  return o.str(); // clang-format on
}

} // namespace o2::gpu::internal