Menu
▾
▴
[oprof-cvs] CVS: oprofile/pp opreport.cpp,1.15,1.16
|
From: Philippe E. <ph...@us...> - 2003-08-11 23:55:17
|
Update of /cvsroot/oprofile/oprofile/pp
In directory sc8-pr-cvs1:/tmp/cvs-serv18910/pp
Modified Files:
opreport.cpp
Log Message:
pp/opreport.cpp: re write using count_array_t
Index: opreport.cpp
===================================================================
RCS file: /cvsroot/oprofile/oprofile/pp/opreport.cpp,v
retrieving revision 1.15
retrieving revision 1.16
diff -u -p -d -r1.15 -r1.16
--- opreport.cpp 11 Aug 2003 01:46:17 -0000 1.15
+++ opreport.cpp 11 Aug 2003 23:22:45 -0000 1.16
@@ -30,29 +30,17 @@ using namespace std;
namespace {
-
-/**
- * Used to sort parellel vector of data upon one of the vector. Sorting
- * is done by first building a bijection map[i] = i then sort this array.
- * Later all array access are done using the sorted indexer. After sorting
- * the bijection ensure than:
- * for i in [1, size)
- * compare(data[index_mapper[i]], data[index_mapper[i-1]]) == true
- */
-typedef vector<size_t> index_mapper_t;
-
+static size_t pp_nr_counters;
/// storage for a merged file summary
struct summary {
- summary(string const & image_name_, string const & lib_image_)
- : count(0), image_name(image_name_), lib_image(lib_image_) {}
- size_t count;
- string image_name;
+ count_array_t counts;
+ string image;
string lib_image;
bool operator<(summary const & rhs) const {
return options::reverse_sort
- ? count < rhs.count : rhs.count < count;
+ ? counts[0] < rhs.counts[0] : rhs.counts[0] < counts[0];
}
};
@@ -63,23 +51,17 @@ struct summary {
* dependent images such as libraries.
*/
struct group_summary {
- group_summary(string const & image_name_, string const & lib_image_)
- : count(0), image_name(image_name_), lib_image(lib_image_) {}
- group_summary() : count(0) {}
-
- size_t count;
- string image_name;
+ count_array_t counts;
+ string image;
string lib_image;
vector<summary> files;
- index_mapper_t index_mapper;
- summary const & file(size_t index) const {
- return files[index_mapper[index]];
- }
+ // return the number of samples added
+ size_t add_samples(split_sample_filename const &, size_t counts_group);
bool operator<(group_summary const & rhs) const {
return options::reverse_sort
- ? count < rhs.count : rhs.count < count;
+ ? counts[0] < rhs.counts[0] : rhs.counts[0] < counts[0];
}
/// return true if the deps should not be output
@@ -87,19 +69,12 @@ struct group_summary {
};
-/// all group_summary belonging to a count group
-struct event_group_summary {
- event_group_summary() : total_count(0.0) {}
-
- group_summary const & group(size_t index) const {
- return groups[index_mapper[index]];
- }
-
+struct counts_group_summary {
+ counts_group_summary(vector<partition_files> const & sample_files);
+ /// all group summaries for a profiles set
vector<group_summary> groups;
- /// total count of samples for this count group
- double total_count;
-
- index_mapper_t index_mapper;
+ /// total count of samples for all groups
+ count_array_t total_counts;
};
@@ -127,19 +102,33 @@ void output_header()
}
-string get_filename(string const & filename)
+size_t group_summary::
+add_samples(split_sample_filename const & split, size_t counts_group)
{
- return options::long_filenames ? filename : basename(filename);
-}
+ // FIXME: linear search inneficient ?
+ vector<summary>::iterator it;
+ for (it = files.begin(); it != files.end(); ++it) {
+ if (it->image == split.image &&
+ it->lib_image == split.lib_image)
+ break;
+ }
+ size_t nr_samples = profile_t::sample_count(split.sample_filename);
-/// Output a count and a percentage
-void output_count(double total_count, size_t count)
-{
- cout << setw(9) << count << " ";
- double ratio = op_ratio(count, total_count);
- cout << format_double(ratio * 100, percent_int_width,
- percent_fract_width) << " ";
+ if (it == files.end()) {
+ summary summary;
+ summary.image = split.image;
+ summary.lib_image = split.lib_image;
+ summary.counts[counts_group] = nr_samples;
+ counts[counts_group] += nr_samples;
+ files.push_back(summary);
+ } else {
+ // assert it->counts[counts_group] == 0
+ it->counts[counts_group] = nr_samples;
+ counts[counts_group] += nr_samples;
+ }
+
+ return nr_samples;
}
@@ -148,16 +137,17 @@ bool group_summary::should_hide_deps() c
if (files.size() == 0)
return true;
- if (count == 0)
+ // can this happens ?
+ if (counts.zero())
return true;
- string image = image_name;
+ string image_name = image;
if (options::merge_by.lib && !lib_image.empty())
- image = lib_image;
+ image_name = lib_image;
summary const & first = files[0];
string const & dep_image = first.lib_image.empty()
- ? first.image_name : first.lib_image;
+ ? first.image : first.lib_image;
bool hidedep = options::exclude_dependent;
hidedep |= options::merge_by.lib;
@@ -171,334 +161,120 @@ bool group_summary::should_hide_deps() c
}
-void
-output_deps(vector<event_group_summary> const & summaries,
- vector<group_summary> const & event_group)
+counts_group_summary::
+counts_group_summary(vector<partition_files> const & sample_files)
{
- bool should_hide_deps = true;
- for (size_t i = 0 ; i < event_group.size(); ++i) {
- if (!event_group[i].should_hide_deps())
- should_hide_deps = false;
- }
-
- if (should_hide_deps)
- return;
-
- for (size_t j = 0 ; j < event_group[0].files.size(); ++j) {
- cout << "\t";
- for (size_t i = 0; i < event_group.size(); ++i) {
- group_summary const & group = event_group[i];
- summary const & file = group.file(j);
-
- double tot_count = options::global_percent
- ? summaries[i].total_count : group.count;
+ // second member is the partition file index i.e. the events/counts
+ // identifier
+ typedef pair<split_sample_filename const *, size_t> value_t;
+ typedef multimap<string, value_t> map_t;
+ map_t sample_filenames;
- output_count(tot_count, file.count);
+ for (size_t i = 0; i < sample_files.size(); ++i) {
+ partition_files const & partition = sample_files[i];
+ for (size_t j = 0; j < partition.nr_set(); ++j) {
+ partition_files::filename_set const & files =
+ partition.set(j);
+ partition_files::filename_set::const_iterator it;
+ for (it = files.begin(); it != files.end(); ++it) {
+ value_t value(&*it, i);
+ map_t::value_type val(it->image, value);
+ sample_filenames.insert(val);
+ }
}
-
- summary const & file = event_group[0].file(j);
- if (file.lib_image.empty())
- cout << " " << get_filename(file.image_name);
- else
- cout << " " << get_filename(file.lib_image);
- cout << '\n';
}
-}
+ map_t::const_iterator it = sample_filenames.begin();
+ for (; it != sample_filenames.end(); ) {
+ split_sample_filename const * cur = it->second.first;
-/**
- * Generate summaries for each of the profiles in this
- * partition set.
- */
-group_summary summarize(partition_files::filename_set const & files)
-{
- group_summary group(files.begin()->image, files.begin()->lib_image);
+ group_summary group;
+ group.image = cur->image;
+ group.lib_image = cur->lib_image;
- partition_files::filename_set::const_iterator it;
- for (it = files.begin(); it != files.end(); ++it) {
- summary dep_summary(it->image, it->lib_image);
+ pair<map_t::const_iterator, map_t::const_iterator> p_it =
+ sample_filenames.equal_range(it->first);
+ for (it = p_it.first; it != p_it.second; ++it) {
+ size_t nr_samples = group.add_samples(
+ *it->second.first, it->second.second);
+ total_counts[it->second.second] += nr_samples;
+ }
- dep_summary.count =
- profile_t::sample_count(it->sample_filename);
+ stable_sort(group.files.begin(), group.files.end());
- group.count += dep_summary.count;
- group.files.push_back(dep_summary);
+ groups.push_back(group);
}
- return group;
+ stable_sort(groups.begin(), groups.end());
}
-/// comparator used to sort a vector<summary> through a mapping index
-class build_summary_index {
-public:
- build_summary_index(vector<summary> const & files_)
- : files(files_) {}
- bool operator()(size_t lhs, size_t rhs) const {
- return files[lhs] < files[rhs];
- }
-private:
- vector<summary> const & files;
-};
-
-
-/**
- * create the index to sorted group_summary, summaries[0] is used to create
- * the sort order. Data themself remains at fixed address, we just fill an
- * index mapper.
- */
-index_mapper_t create_index_mapper(vector<group_summary> const & summaries)
+string get_filename(string const & filename)
{
- index_mapper_t result(summaries[0].files.size());
-
- for (size_t i = 0; i < result.size(); ++i) {
- result[i] = i;
- }
-
- sort(result.begin(), result.end(),
- build_summary_index(summaries[0].files));
-
- return result;
+ return options::long_filenames ? filename : basename(filename);
}
-/**
- * build each group_summary::vector<summary> in synched way
- * populating a vector<group_summary> w/o any gap between each entry
- * i.e post condition are:
- * for i in [1 result.size)
- * result[i].groups.size() == result[i-1].groups.size()
- * for i in [1 result.size)
- * result[i].groups.image_name == result[i-1].groups.image_name
- *
- * ascii art for 3 events:
- *
- * sample_file_partition array:
- * -- AA ------- BC --- CD --- EF -----------
- * -------- BA --------- CD ------------------
- * -------------- BC ----------------- FG ----
- *
- * result:
- * -- AA -- BA -- BC --- CD --- EF --- FG ----
- * -- AA -- BA -- BC --- CD --- EF --- FG ----
- * -- AA -- BA -- BC --- CD --- EF --- FG ----
- *
- * the added items contain zero samples.
- *
- * @internal a multimap<name, <data, vector entry index>> is used to partition
- * the input data (coming from all count groups). Then we iterate over
- * equivalence class, each class contain item for an unique name and
- * associated info contains the counter nr for this data.
- */
-vector<group_summary>
-populate_summaries(vector<event_group_summary> const & unfilled, size_t index)
+/// Output a count and a percentage
+void output_count(double total_count, size_t count)
{
- typedef pair<summary const *, int> value_t;
- typedef multimap<string, value_t> map_t;
- map_t map;
-
- // Partition the files set at index.
- for (size_t i = 0; i < unfilled.size(); ++i) {
- vector<summary> const & groups = unfilled[i].group(index).files;
- for (size_t j = 0; j < groups.size(); ++j) {
- string image = groups[j].image_name;
- if (!groups[j].lib_image.empty())
- image = groups[j].lib_image;
- value_t value(&groups[j], i);
- map.insert(map_t::value_type(image, value));
- }
- }
-
- vector<group_summary> result(unfilled.size());
+ cout << setw(9) << count << " ";
+ double ratio = op_ratio(count, total_count);
+ cout << format_double(ratio * 100, percent_int_width,
+ percent_fract_width) << " ";
+}
- for (size_t i = 0; i < unfilled.size(); ++i) {
- group_summary const & group = unfilled[i].group(index);
- result[i].count = group.count;
- result[i].image_name = group.image_name;
- result[i].lib_image = group.lib_image;
- }
- // for each equivalance class.
- for (map_t::const_iterator it = map.begin(); it != map.end(); ) {
- // Populate entries summary with empty summary, it->second
- // is a representant of the current equivalence class.
- for (size_t i = 0 ; i < unfilled.size() ; ++i) {
- string image_name = it->second.first->image_name;
- string lib_image = it->second.first->lib_image;
- summary summary(image_name, lib_image);
+void
+output_deps(counts_group_summary const & summaries,
+ group_summary const & groups)
+{
+ for (size_t j = 0 ; j < groups.files.size(); ++j) {
+ cout << "\t";
+ summary const & file = groups.files[j];
+ for (size_t i = 0; i < pp_nr_counters; ++i) {
+ double tot_count = options::global_percent
+ ? summaries.total_counts[i] : groups.counts[i];
- result[i].files.push_back(summary);
+ output_count(tot_count, file.counts[i]);
}
- // Overwrite empty summary create above by the existing one.
- pair<map_t::const_iterator, map_t::const_iterator> p_it =
- p_it = map.equal_range(it->first);
- for (; it != p_it.second; ++it) {
- result[it->second.second].files.back() =
- *it->second.first;
- }
+ if (file.lib_image.empty())
+ cout << " " << get_filename(file.image);
+ else
+ cout << " " << get_filename(file.lib_image);
+ cout << '\n';
}
-
- index_mapper_t index_mapper = create_index_mapper(result);
- for (size_t i = 0; i < result.size(); ++i)
- result[i].index_mapper = index_mapper;
-
- return result;
}
/**
* Display all the given summary information
*/
-void output_summaries(vector<event_group_summary> const & summaries)
+void output_summaries(counts_group_summary const & summaries)
{
- for (size_t i = 0 ; i < summaries[0].groups.size(); ++i) {
- group_summary const & group = summaries[0].group(i);
+ for (size_t i = 0; i < summaries.groups.size(); ++i) {
+ group_summary const & group = summaries.groups[i];
- if ((group.count * 100.0) / summaries[0].total_count <
+ if ((group.counts[0] * 100.0) / summaries.total_counts[0] <
options::threshold) {
continue;
}
- for (size_t j = 0; j < summaries.size(); ++j) {
- group_summary const & group = summaries[j].group(i);
- output_count(summaries[j].total_count, group.count);
+ for (size_t j = 0; j < pp_nr_counters; ++j) {
+ output_count(summaries.total_counts[j],
+ group.counts[j]);
}
- string image = group.image_name;
+ string image = group.image;
if (options::merge_by.lib && !group.lib_image.empty())
image = group.lib_image;
cout << get_filename(image) << '\n';
- vector<group_summary> filled_summaries =
- populate_summaries(summaries, i);
-
- output_deps(summaries, filled_summaries);
- }
-}
-
-
-/// comparator used to sort a vector<group_summary> through a mapping index
-class build_group_summary_index {
-public:
- build_group_summary_index(vector<group_summary> const & group_)
- : group(group_) {}
- bool operator()(size_t lhs, size_t rhs) const {
- return group[lhs] < group[rhs];
- }
-private:
- vector<group_summary> const & group;
-};
-
-
-/**
- * create the index to sorted group_summary, summaries[0] is used to create
- * the sort order. Data remains at fixed address, we just fill an index mapper
- */
-index_mapper_t
-create_index_mapper(vector<event_group_summary> const & summaries)
-{
- index_mapper_t result(summaries[0].groups.size());
-
- // no stl-ish way to generate [0 - size()) range
- for (size_t i = 0; i < result.size(); ++i) {
- result[i] = i;
- }
-
- sort(result.begin(), result.end(),
- build_group_summary_index(summaries[0].groups));
-
- return result;
-}
-
-
-/**
- * build each event_group_summary::vector<group_summary> in synched way
- * populating a vector<event_group_summary> w/o any gap between each entry
- * i.e post condition are:
- * for i in [1 result.size)
- * result[i].groups.size() == result[i-1].groups.size()
- * for i in [1 result.size)
- * result[i].groups.image_name == result[i-1].groups.image_name
- *
- * ascii art for 3 events:
- *
- * sample_file_partition array:
- * -- AA ------- BC --- CD --- EF -----------
- * -------- BA --------- CD ------------------
- * -------------- BC ----------------- FG ----
- *
- * result:
- * -- AA -- BA -- BC --- CD --- EF --- FG ----
- * -- AA -- BA -- BC --- CD --- EF --- FG ----
- * -- AA -- BA -- BC --- CD --- EF --- FG ----
- *
- * the added items contain zero samples.
- *
- * @internal a multimap<name, <data, vector entry index>> is used to partition
- * the input data (coming from all count groups). Then we iterate over equivalence
- * class, each class contain item for an unique name and associated info
- * contains the counter nr for this data.
- */
-vector<event_group_summary> populate_group_summaries()
-{
- typedef pair<group_summary, int> value_t;
- typedef multimap<string, value_t> map_t;
- map_t map;
-
- size_t const nr_events = sample_file_partition.size();
-
- vector<event_group_summary> result(nr_events);
-
- // partition the files set.
- for (size_t event = 0; event < nr_events; ++event) {
- partition_files const & files = sample_file_partition[event];
-
- for (size_t j = 0; j < files.nr_set(); ++j) {
- partition_files::filename_set const & file_set
- = files.set(j);
-
- group_summary group(summarize(file_set));
- result[event].total_count += group.count;
-
- string image = group.image_name;
- if (options::merge_by.lib && !group.lib_image.empty())
- image = group.lib_image;
-
- value_t value(group, event);
- map.insert(map_t::value_type(image, value));
- }
- }
-
- // for each equivalence class
- for (map_t::const_iterator it = map.begin(); it != map.end(); ) {
- // populate entries with empty group_summary, it->second is a
- // representant of the current equivalence class
- for (size_t i = 0 ; i < nr_events; ++i) {
- string image_name = it->second.first.image_name;
- string lib_image = it->second.first.lib_image;
- group_summary group(image_name, lib_image);
-
- result[i].groups.push_back(group);
- }
-
- // overwrite the empty group_summary create above by existing
- // group_summary
- pair<map_t::const_iterator, map_t::const_iterator> p_it =
- map.equal_range(it->first);
- for ( ; it != p_it.second; ++it) {
- result[it->second.second].groups.back() =
- it->second.first;
- }
- }
-
- index_mapper_t index_mapper = create_index_mapper(result);
- for (size_t i = 0; i < nr_events; ++i) {
- result[i].index_mapper = index_mapper;
+ if (!group.should_hide_deps())
+ output_deps(summaries, group);
}
-
- return result;
}
@@ -611,10 +387,10 @@ int opreport(vector<string> const & non_
output_header();
if (!options::symbols) {
- vector<event_group_summary> filled_summaries =
- populate_group_summaries();
+ pp_nr_counters = sample_file_partition.size();
- output_summaries(filled_summaries);
+ counts_group_summary summaries(sample_file_partition);
+ output_summaries(summaries);
return 0;
}
|