Skip to content

Commit

Permalink
[compile time] reduce templated code in deep multilevel
Browse files Browse the repository at this point in the history
  • Loading branch information
@N-Maas
N-Maas committed Sep 14, 2023
1 parent 27be0f5 commit 9d7e088
Showing 1 changed file with 66 additions and 95 deletions.
161 changes: 66 additions & 95 deletions mt-kahypar/partition/deep_multilevel.cpp
View file
Original file line number Diff line number Diff line change
Expand Up @@ -31,18 +31,17 @@
#include <limits>
#include <vector>

#include "tbb/parallel_invoke.h"
#include "tbb/parallel_for.h"

#include "mt-kahypar/definitions.h"
#include "mt-kahypar/macros.h"
#include "mt-kahypar/partition/metrics.h"
#include "mt-kahypar/partition/multilevel.h"
#include "mt-kahypar/partition/coarsening/coarsening_commons.h"
#include "mt-kahypar/partition/coarsening/multilevel_uncoarsener.h"
#include "mt-kahypar/partition/coarsening/nlevel_uncoarsener.h"
#include "mt-kahypar/partition/initial_partitioning/pool_initial_partitioner.h"
#include "mt-kahypar/partition/preprocessing/sparsification/degree_zero_hn_remover.h"
#include "mt-kahypar/partition/refinement/gains/gain_cache_ptr.h"
#include "mt-kahypar/partition/refinement/gains/bipartitioning_policy.h"
#include "mt-kahypar/utils/randomize.h"
#include "mt-kahypar/utils/utilities.h"
#include "mt-kahypar/utils/timer.h"
#include "mt-kahypar/utils/progress_bar.h"
Expand Down Expand Up @@ -279,19 +278,18 @@ void enableTimerAndStats(const Context& context, const bool was_enabled_before)
}
}

template<typename Hypergraph>
Context setupBipartitioningContext(const Hypergraph& hypergraph,
const Context& context,
Context setupBipartitioningContext(const Context& context,
const OriginalHypergraphInfo& info,
const PartitionID start_k,
const PartitionID end_k) {
const PartitionID end_k,
const HypernodeWeight total_weight,
const bool is_graph) {
ASSERT(end_k - start_k >= 2);
Context b_context(context);

b_context.partition.k = 2;
b_context.partition.objective = Objective::cut;
b_context.partition.gain_policy = Hypergraph::is_graph ?
GainPolicy::cut_for_graphs : GainPolicy::cut;
b_context.partition.gain_policy = is_graph ? GainPolicy::cut_for_graphs : GainPolicy::cut;
b_context.partition.verbose_output = false;
b_context.initial_partitioning.mode = Mode::direct;
b_context.type = ContextType::initial_partitioning;
Expand All @@ -306,7 +304,6 @@ Context setupBipartitioningContext(const Hypergraph& hypergraph,
b_context.refinement = b_context.initial_partitioning.refinement;

// Setup Part Weights
const HypernodeWeight total_weight = hypergraph.totalWeight();
const PartitionID k = end_k - start_k;
const PartitionID k0 = k / 2 + (k % 2 != 0 ? 1 : 0);
const PartitionID k1 = k / 2;
Expand Down Expand Up @@ -472,7 +469,7 @@ DeepPartitioningResult<TypeTraits> bipartition_block(typename TypeTraits::Hyperg
if ( bipartition.hypergraph.initialNumNodes() > 0 ) {
// Bipartition block
Context b_context = setupBipartitioningContext(
bipartition.hypergraph, context, info, start_k, end_k);
context, info, start_k, end_k, bipartition.hypergraph.totalWeight(), PartitionedHypergraph::is_graph);
bipartition.partitioned_hg = Multilevel<TypeTraits>::partition(
bipartition.hypergraph, b_context);
} else {
Expand All @@ -483,9 +480,61 @@ DeepPartitioningResult<TypeTraits> bipartition_block(typename TypeTraits::Hyperg
}

template<typename TypeTraits, typename GainCache>
void apply_bipartitions_to_hypergraph(typename TypeTraits::PartitionedHypergraph& partitioned_hg,
GainCache& gain_cache,
const vec<HypernodeID>& mapping,
const vec<DeepPartitioningResult<TypeTraits>>& bipartitions,
const vec<PartitionID>& block_ranges) {
partitioned_hg.doParallelForAllNodes([&](const HypernodeID& hn) {
const PartitionID from = partitioned_hg.partID(hn);
ASSERT(static_cast<size_t>(from) < bipartitions.size());
PartitionID to = kInvalidPartition;
const DeepPartitioningResult<TypeTraits>& bipartition = bipartitions[from];
if ( bipartition.valid ) {
ASSERT(static_cast<size_t>(hn) < mapping.size());
const HypernodeID mapped_hn = mapping[hn];
to = bipartition.partitioned_hg.partID(mapped_hn) == 0 ?
block_ranges[from] : block_ranges[from] + 1;
} else {
to = block_ranges[from];
}

ASSERT(to > kInvalidPartition && to < block_ranges.back());
if ( from != to ) {
if ( gain_cache.isInitialized() ) {
partitioned_hg.changeNodePart(gain_cache, hn, from, to);
} else {
partitioned_hg.changeNodePart(hn, from, to);
}
}
});

if ( GainCache::invalidates_entries && gain_cache.isInitialized() ) {
partitioned_hg.doParallelForAllNodes([&](const HypernodeID& hn) {
gain_cache.recomputeInvalidTerms(partitioned_hg, hn);
});
}

HEAVY_REFINEMENT_ASSERT(partitioned_hg.checkTrackedPartitionInformation(gain_cache));
}

template<typename TypeTraits>
void apply_bipartitions_to_hypergraph(typename TypeTraits::PartitionedHypergraph& partitioned_hg,
gain_cache_t gain_cache,
const vec<HypernodeID>& mapping,
const vec<DeepPartitioningResult<TypeTraits>>& bipartitions,
const vec<PartitionID>& block_ranges) {
using PartitionedHypergraph = typename TypeTraits::PartitionedHypergraph;

GainCachePtr::applyWithConcreteGainCacheForHG<PartitionedHypergraph>([&](auto& gain_cache) {
apply_bipartitions_to_hypergraph<TypeTraits>(partitioned_hg,gain_cache, mapping, bipartitions, block_ranges);
}, gain_cache);
}

template<typename TypeTraits>
void bipartition_each_block(typename TypeTraits::PartitionedHypergraph& partitioned_hg,
const Context& context,
GainCache& gain_cache,
gain_cache_t gain_cache,
const OriginalHypergraphInfo& info,
const RBTree& rb_tree,
vec<uint8_t>& already_cut,
Expand Down Expand Up @@ -551,30 +600,8 @@ void bipartition_each_block(typename TypeTraits::PartitionedHypergraph& partitio
timer.stop_timer("bipartition_blocks");

timer.start_timer("apply_bipartitions", "Apply Bipartition");
// Apply all bipartitions to current hypergraph
partitioned_hg.doParallelForAllNodes([&](const HypernodeID& hn) {
const PartitionID from = partitioned_hg.partID(hn);
ASSERT(static_cast<size_t>(from) < bipartitions.size());
PartitionID to = kInvalidPartition;
const DeepPartitioningResult<TypeTraits>& bipartition = bipartitions[from];
if ( bipartition.valid ) {
ASSERT(static_cast<size_t>(hn) < mapping.size());
const HypernodeID mapped_hn = mapping[hn];
to = bipartition.partitioned_hg.partID(mapped_hn) == 0 ?
block_ranges[from] : block_ranges[from] + 1;
} else {
to = block_ranges[from];
}

ASSERT(to > kInvalidPartition && to < block_ranges.back());
if ( from != to ) {
if ( gain_cache.isInitialized() ) {
partitioned_hg.changeNodePart(gain_cache, hn, from, to);
} else {
partitioned_hg.changeNodePart(hn, from, to);
}
}
});
apply_bipartitions_to_hypergraph(partitioned_hg, gain_cache, mapping, bipartitions, block_ranges);
timer.stop_timer("apply_bipartitions");

ASSERT([&] {
HyperedgeWeight expected_objective = current_objective;
Expand All @@ -592,69 +619,12 @@ void bipartition_each_block(typename TypeTraits::PartitionedHypergraph& partitio
return true;
}(), "Cut of extracted blocks does not sum up to current objective");

if ( GainCache::invalidates_entries && gain_cache.isInitialized() ) {
partitioned_hg.doParallelForAllNodes([&](const HypernodeID& hn) {
gain_cache.recomputeInvalidTerms(partitioned_hg, hn);
});
}
timer.stop_timer("apply_bipartitions");

timer.start_timer("free_hypergraphs", "Free Hypergraphs");
tbb::parallel_for(UL(0), bipartitions.size(), [&](const size_t i) {
DeepPartitioningResult<TypeTraits> tmp_res;
tmp_res = std::move(bipartitions[i]);
});
timer.stop_timer("free_hypergraphs");

HEAVY_REFINEMENT_ASSERT(partitioned_hg.checkTrackedPartitionInformation(gain_cache));
}

template<typename TypeTraits>
void bipartition_each_block(typename TypeTraits::PartitionedHypergraph& partitioned_hg,
const Context& context,
gain_cache_t gain_cache,
const OriginalHypergraphInfo& info,
const RBTree& rb_tree,
vec<uint8_t>& already_cut,
const PartitionID current_k,
const HyperedgeWeight current_objective,
const bool progress_bar_enabled) {
switch(gain_cache.type) {
case GainPolicy::cut:
bipartition_each_block<TypeTraits>(partitioned_hg, context,
GainCachePtr::cast<CutGainCache>(gain_cache), info, rb_tree,
already_cut, current_k, current_objective, progress_bar_enabled); break;
case GainPolicy::km1:
bipartition_each_block<TypeTraits>(partitioned_hg, context,
GainCachePtr::cast<Km1GainCache>(gain_cache), info, rb_tree,
already_cut, current_k, current_objective, progress_bar_enabled); break;
#ifdef KAHYPAR_ENABLE_SOED_METRIC
case GainPolicy::soed:
bipartition_each_block<TypeTraits>(partitioned_hg, context,
GainCachePtr::cast<SoedGainCache>(gain_cache), info, rb_tree,
already_cut, current_k, current_objective, progress_bar_enabled); break;
#endif
#ifdef KAHYPAR_ENABLE_STEINER_TREE_METRIC
case GainPolicy::steiner_tree:
bipartition_each_block<TypeTraits>(partitioned_hg, context,
GainCachePtr::cast<SteinerTreeGainCache>(gain_cache), info, rb_tree,
already_cut, current_k, current_objective, progress_bar_enabled); break;
#endif
#ifdef KAHYPAR_ENABLE_GRAPH_PARTITIONING_FEATURES
case GainPolicy::cut_for_graphs:
bipartition_each_block<TypeTraits>(partitioned_hg, context,
GainCachePtr::cast<GraphCutGainCache>(gain_cache), info, rb_tree,
already_cut, current_k, current_objective, progress_bar_enabled); break;
#ifdef KAHYPAR_ENABLE_STEINER_TREE_METRIC
case GainPolicy::steiner_tree_for_graphs:
bipartition_each_block<TypeTraits>(partitioned_hg, context,
GainCachePtr::cast<GraphSteinerTreeGainCache>(gain_cache), info, rb_tree,
already_cut, current_k, current_objective, progress_bar_enabled); break;
#endif
#endif
case GainPolicy::none: break;
default: break;
}
}

template<typename TypeTraits>
Expand Down Expand Up @@ -768,7 +738,8 @@ PartitionID deep_multilevel_partitioning(typename TypeTraits::PartitionedHypergr
// and continue with uncoarsening.
const auto target_blocks = rb_tree.targetBlocksInFinalPartition(1, 0);
Context b_context = setupBipartitioningContext(
hypergraph, context, info, target_blocks.first, target_blocks.second);
context, info, target_blocks.first, target_blocks.second,
hypergraph.totalWeight(), Hypergraph::is_graph);
Multilevel<TypeTraits>::partition(coarsest_phg, b_context);
current_k = 2;

Expand Down

0 comments on commit 9d7e088

Please sign in to comment.