InPlaceSortImpl.cu File Reference

#include "DataMgr/Allocators/ThrustAllocator.h"
#include "InPlaceSortImpl.h"

Include dependency graph for InPlaceSortImpl.cu:

Go to the source code of this file.

Functions
void	sort_on_gpu (int64_t *val_buff, int32_t *idx_buff, const uint64_t entry_count, const bool desc, const uint32_t chosen_bytes, ThrustAllocator &alloc, const int device_id)
void	sort_on_cpu (int64_t *val_buff, int32_t *idx_buff, const uint64_t entry_count, const bool desc, const uint32_t chosen_bytes)
void	apply_permutation_on_gpu (int64_t *val_buff, int32_t *idx_buff, const uint64_t entry_count, const uint32_t chosen_bytes, ThrustAllocator &alloc, const int device_id)
void	apply_permutation_on_cpu (int64_t *val_buff, int32_t *idx_buff, const uint64_t entry_count, int64_t *tmp_buff, const uint32_t chosen_bytes)

Function Documentation

void apply_permutation_on_cpu	(	int64_t *	val_buff,
		int32_t *	idx_buff,
		const uint64_t	entry_count,
		int64_t *	tmp_buff,
		const uint32_t	chosen_bytes
	)

Definition at line 193 of file InPlaceSortImpl.cu.

References apply_permutation_on_cpu().

Referenced by apply_permutation_cpu(), and apply_permutation_on_cpu().

                                                           {
#ifdef HAVE_CUDA
  switch (chosen_bytes) {
    case 1:
      apply_permutation_on_cpu(reinterpret_cast<int8_t*>(val_buff),
                               idx_buff,
                               entry_count,
                               reinterpret_cast<int8_t*>(tmp_buff));
      break;
    case 2:
      apply_permutation_on_cpu(reinterpret_cast<int16_t*>(val_buff),
                               idx_buff,
                               entry_count,
                               reinterpret_cast<int16_t*>(tmp_buff));
      break;
    case 4:
      apply_permutation_on_cpu(reinterpret_cast<int32_t*>(val_buff),
                               idx_buff,
                               entry_count,
                               reinterpret_cast<int32_t*>(tmp_buff));
      break;
    case 8:
      apply_permutation_on_cpu(val_buff, idx_buff, entry_count, tmp_buff);
      break;
    default:
      // FIXME(miyu): CUDA linker doesn't accept assertion on GPU yet right now.
      break;
  }
#endif
}

Here is the call graph for this function:

Here is the caller graph for this function:

void apply_permutation_on_gpu	(	int64_t *	val_buff,
		int32_t *	idx_buff,
		const uint64_t	entry_count,
		const uint32_t	chosen_bytes,
		ThrustAllocator &	alloc,
		const int	device_id
	)

Definition at line 163 of file InPlaceSortImpl.cu.

References apply_permutation_on_gpu().

Referenced by anonymous_namespace{InPlaceSort.cpp}::apply_permutation_gpu(), and apply_permutation_on_gpu().

                                                   {
#ifdef HAVE_CUDA
  switch (chosen_bytes) {
    case 1:
      apply_permutation_on_gpu(
          reinterpret_cast<int8_t*>(val_buff), idx_buff, entry_count, alloc, device_id);
      break;
    case 2:
      apply_permutation_on_gpu(
          reinterpret_cast<int16_t*>(val_buff), idx_buff, entry_count, alloc, device_id);
      break;
    case 4:
      apply_permutation_on_gpu(
          reinterpret_cast<int32_t*>(val_buff), idx_buff, entry_count, alloc, device_id);
      break;
    case 8:
      apply_permutation_on_gpu(val_buff, idx_buff, entry_count, alloc, device_id);
      break;
    default:
      // FIXME(miyu): CUDA linker doesn't accept assertion on GPU yet right now.
      break;
  }
#endif
}

Here is the call graph for this function:

Here is the caller graph for this function:

void sort_on_cpu	(	int64_t *	val_buff,
		int32_t *	idx_buff,
		const uint64_t	entry_count,
		const bool	desc,
		const uint32_t	chosen_bytes
	)

Definition at line 137 of file InPlaceSortImpl.cu.

References sort_on_cpu().

Referenced by sort_groups_cpu(), and sort_on_cpu().

                                              {
#ifdef HAVE_CUDA
  switch (chosen_bytes) {
    case 1:
      sort_on_cpu(reinterpret_cast<int8_t*>(val_buff), idx_buff, entry_count, desc);
      break;
    case 2:
      sort_on_cpu(reinterpret_cast<int16_t*>(val_buff), idx_buff, entry_count, desc);
      break;
    case 4:
      sort_on_cpu(reinterpret_cast<int32_t*>(val_buff), idx_buff, entry_count, desc);
      break;
    case 8:
      sort_on_cpu(val_buff, idx_buff, entry_count, desc);
      break;
    default:
      // FIXME(miyu): CUDA linker doesn't accept assertion on GPU yet right now.
      break;
  }
#endif
}

Here is the call graph for this function:

Here is the caller graph for this function:

void sort_on_gpu	(	int64_t *	val_buff,
		int32_t *	idx_buff,
		const uint64_t	entry_count,
		const bool	desc,
		const uint32_t	chosen_bytes,
		ThrustAllocator &	alloc,
		const int	device_id
	)

Definition at line 94 of file InPlaceSortImpl.cu.

References sort_on_gpu().

Referenced by QueryExecutionContext::QueryExecutionContext(), anonymous_namespace{InPlaceSort.cpp}::sort_groups_gpu(), and sort_on_gpu().

                                      {
#ifdef HAVE_CUDA
  switch (chosen_bytes) {
    case 1:
      sort_on_gpu(reinterpret_cast<int8_t*>(val_buff),
                  idx_buff,
                  entry_count,
                  desc,
                  alloc,
                  device_id);
      break;
    case 2:
      sort_on_gpu(reinterpret_cast<int16_t*>(val_buff),
                  idx_buff,
                  entry_count,
                  desc,
                  alloc,
                  device_id);
      break;
    case 4:
      sort_on_gpu(reinterpret_cast<int32_t*>(val_buff),
                  idx_buff,
                  entry_count,
                  desc,
                  alloc,
                  device_id);
      break;
    case 8:
      sort_on_gpu(val_buff, idx_buff, entry_count, desc, alloc, device_id);
      break;
    default:
      // FIXME(miyu): CUDA linker doesn't accept assertion on GPU yet right now.
      break;
  }
#endif
}

Here is the call graph for this function:

Here is the caller graph for this function: