Skip to content
Toggle navigation
P
Projects
G
Groups
S
Snippets
Help
las3_pub
/
predictable_parallel_patterns
This project
Loading...
Sign in
Toggle navigation
Go to a project
Project
Repository
Issues
0
Merge Requests
0
Pipelines
Wiki
Members
Activity
Graph
Charts
Create a new issue
Jobs
Commits
Issue Boards
Files
Commits
Branches
Tags
Contributors
Graph
Compare
Charts
Commit
72bffbb1
authored
Apr 04, 2019
by
FritzFlorian
Browse files
Options
Browse Files
Download
Email Patches
Plain Diff
Add optional easy_profiler dependency to the project.
parent
351ced3d
Pipeline
#1125
failed with stages
in 30 seconds
Changes
13
Pipelines
1
Hide whitespace changes
Inline
Side-by-side
Showing
13 changed files
with
110 additions
and
27 deletions
+110
-27
CMakeLists.txt
+1
-0
NOTES.md
+10
-1
README.md
+23
-0
app/invoke_parallel/CMakeLists.txt
+1
-1
app/invoke_parallel/main.cpp
+12
-12
cmake/SetupEasyProfiler.cmake
+18
-0
lib/pls/CMakeLists.txt
+3
-0
lib/pls/include/pls/algorithms/invoke_parallel.h
+1
-1
lib/pls/include/pls/internal/scheduling/fork_join_task.h
+6
-1
lib/pls/include/pls/internal/scheduling/root_task.h
+10
-10
lib/pls/include/pls/internal/scheduling/scheduler.h
+2
-0
lib/pls/src/internal/scheduling/abstract_task.cpp
+11
-0
lib/pls/src/internal/scheduling/fork_join_task.cpp
+12
-1
No files found.
CMakeLists.txt
View file @
72bffbb1
...
...
@@ -15,6 +15,7 @@ include(cmake/SetupOptimizationLevel.cmake)
include
(
cmake/SetupThreadingSupport.cmake
)
include
(
cmake/SetupThreadSanitizer.cmake
)
include
(
cmake/SetupAddressSanitizer.cmake
)
include
(
cmake/SetupEasyProfiler.cmake
)
# make our internal cmake script collection avaliable in the build process.
list
(
APPEND CMAKE_PREFIX_PATH
"
${
PROJECT_SOURCE_DIR
}
/cmake"
)
...
...
NOTES.md
View file @
72bffbb1
...
...
@@ -4,7 +4,16 @@ A collection of stuff that we noticed during development.
Useful later on two write a project report and to go back
in time to find out why certain decisions where made.
## 28.03.2018 - custom new operators
## 02.04.2019 - CMake Export
We built our project using CMake to make it portable and easy to setup.
To allow others to use our library we need to make it installable on
other systems. For this we use CMake's install feature and
a
[
tutorial
](
https://pabloariasal.github.io/2018/02/19/its-time-to-do-cmake-right/
)
on how to correctly configure a CMake library to be included by other
projects.
## 28.03.2019 - custom new operators
When initializing sub_tasks we want to place them on our custom
'stack like' data structure per thread. We looked at TBB's API
...
...
README.md
View file @
72bffbb1
...
...
@@ -43,3 +43,26 @@ keep the repository in a state where the sanitizer reports errors.
Consider reading
[
the section on common data races
](
https://github.com/google/sanitizers/wiki/ThreadSanitizerPopularDataRaces
)
to get an idea of what we try to avoid in our code.
### Profiling
To make profiling portable and allow us to later analyze the logs
programaticly we use
[
easy_profiler
](
https://github.com/yse/easy_profiler
)
for capturing data. To enable profiling install the library on your system
(best building it and then running
`make install`
) and set the
cmake option
`-DEASY_PROFILER=ON`
.
After that see the
`invoke_parallel`
example app for activating the
profiler. This will generate a trace file that can be viewed with
the
`profiler_gui <output.prof>`
command.
Please note that the profiler adds overhead when looking at sub millisecond
method invokations as we do and it can not replace a seperate
profiler like
`gperf`
or
`valgrind`
for detailed analysis.
We still think it makes sense to add it in as an optional feature,
as the customizable colors and fine grained events (including collection
of variables) can be used to visualize the
`big picture`
of
program execution. Also, we hope to use it to log 'events' like
successful and failed steals in the future, as the general idea of logging
information per thread efficiently might be helpful for further
analysis.
app/invoke_parallel/CMakeLists.txt
View file @
72bffbb1
add_executable
(
invoke_parallel main.cpp
)
target_link_libraries
(
invoke_parallel pls
)
target_link_libraries
(
invoke_parallel pls
easy_profiler
)
app/invoke_parallel/main.cpp
View file @
72bffbb1
#include <pls/pls.h>
#include <iostream>
static
pls
::
static_scheduler_memory
<
8
,
2
<<
10
>
my_scheduler_memory
;
#include <easy/profiler.h>
static
constexpr
int
CUTOFF
=
20
;
static
pls
::
static_scheduler_memory
<
8
,
2
<<
14
>
my_scheduler_memory
;
static
constexpr
int
CUTOFF
=
10
;
long
fib_serial
(
long
n
)
{
if
(
n
==
0
)
{
...
...
@@ -31,19 +33,17 @@ long fib(long n) {
}
int
main
()
{
EASY_PROFILER_ENABLE
;
pls
::
scheduler
scheduler
{
&
my_scheduler_memory
,
8
};
scheduler
.
perform_work
([]
{
auto
start
=
std
::
chrono
::
high_resolution_clock
::
now
();
long
result
;
scheduler
.
perform_work
([
&
]
{
EASY_MAIN_THREAD
;
// Call looks just the same, only requirement is
// the enclosure in the perform_work lambda.
long
result
=
fib
(
30
);
auto
end
=
std
::
chrono
::
high_resolution_clock
::
now
();
long
time
=
std
::
chrono
::
duration_cast
<
std
::
chrono
::
microseconds
>
(
end
-
start
).
count
();
std
::
cout
<<
"Fib(30)="
<<
result
<<
std
::
endl
;
std
::
cout
<<
"Execution time in us: "
<<
time
<<
std
::
endl
;
result
=
fib
(
30
);
});
std
::
cout
<<
"Fib(30)="
<<
result
<<
std
::
endl
;
profiler
::
dumpBlocksToFile
(
"test_profile.prof"
);
}
cmake/SetupEasyProfiler.cmake
0 → 100644
View file @
72bffbb1
# Optional external dependencies
find_package
(
easy_profiler
)
option
(
EASY_PROFILER
"Enable the profiler"
OFF
)
if
(
EASY_PROFILER
)
if
(
easy_profiler_FOUND
)
else
()
message
(
WARNING
"EasyProfiler dependency not found on system, DISABLING it!"
)
set
(
EASY_PROFILER OFF
)
endif
()
endif
()
if
(
NOT EASY_PROFILER
)
add_definitions
(
-DDISABLE_EASY_PROFILER
)
endif
()
message
(
"-- Easy Profiler:
${
EASY_PROFILER
}
"
)
lib/pls/CMakeLists.txt
View file @
72bffbb1
...
...
@@ -29,6 +29,9 @@ target_include_directories(pls
target_link_libraries
(
pls
Threads::Threads
# pthread support
)
if
(
EASY_PROFILER
)
target_link_libraries
(
pls easy_profiler
)
endif
()
# Rules for istalling the library on a system
# ...binaries
...
...
lib/pls/include/pls/algorithms/invoke_parallel.h
View file @
72bffbb1
...
...
@@ -33,7 +33,7 @@ namespace pls {
auto
internal_body
=
[
&
]
(
fork_join_sub_task
*
this_task
){
auto
sub_task_body_1
=
[
&
]
(
fork_join_sub_task
*
){
function1
();
};
fork_join_lambda
<
decltype
(
sub_task_body_1
)
>
sub_task_1
(
&
sub_task_body_1
);
auto
sub_task_1
=
fork_join_lambda
<
decltype
(
sub_task_body_1
)
>
(
&
sub_task_body_1
);
this_task
->
spawn_child
(
sub_task_1
);
function2
();
// Execute last function 'inline' without spawning a sub_task object
...
...
lib/pls/include/pls/internal/scheduling/fork_join_task.h
View file @
72bffbb1
...
...
@@ -2,6 +2,8 @@
#ifndef PLS_TBB_LIKE_TASK_H
#define PLS_TBB_LIKE_TASK_H
#include <easy/profiler.h>
#include "pls/internal/base/aligned_stack.h"
#include "pls/internal/base/deque.h"
...
...
@@ -84,6 +86,8 @@ namespace pls {
last_stolen_
{
nullptr
}
{};
void
execute
()
override
{
EASY_BLOCK
(
"execute fork_join_task"
,
profiler
::
colors
::
LightGreen
);
// Bind this instance to our OS thread
my_stack_
=
base
::
this_thread
::
state
<
thread_state
>
()
->
task_stack_
;
root_task_
->
tbb_task_
=
this
;
...
...
@@ -97,7 +101,8 @@ namespace pls {
};
template
<
typename
T
>
void
fork_join_sub_task
::
spawn_child
(
const
T
&
task
)
{
void
fork_join_sub_task
::
spawn_child
(
const
T
&
task
)
{
EASY_FUNCTION
(
profiler
::
colors
::
Blue
)
static_assert
(
std
::
is_base_of
<
fork_join_sub_task
,
T
>::
value
,
"Only pass fork_join_sub_task subclasses!"
);
T
*
new_task
=
tbb_task_
->
my_stack_
->
push
(
task
);
...
...
lib/pls/include/pls/internal/scheduling/root_task.h
View file @
72bffbb1
...
...
@@ -2,6 +2,7 @@
#ifndef PLS_ROOT_MASTER_TASK_H
#define PLS_ROOT_MASTER_TASK_H
#include <easy/profiler.h>
#include <mutex>
#include "abstract_task.h"
...
...
@@ -13,27 +14,25 @@ namespace pls {
template
<
typename
Function
>
class
root_task
:
public
abstract_task
{
Function
function_
;
bool
finished_
;
// Improvement: Remove lock and replace by atomic variable (performance)
base
::
spin_lock
finished_lock_
;
std
::
atomic_uint8_t
finished_
;
public
:
explicit
root_task
(
Function
function
)
:
abstract_task
{
0
,
id
{
0
}},
function_
{
function
},
finished_
{
false
}
{}
finished_
{
0
}
{}
root_task
(
const
root_task
&
other
)
:
abstract_task
{
0
,
id
{
0
}},
function_
{
other
.
function_
},
finished_
{
0
}
{}
bool
finished
()
{
std
::
lock_guard
<
base
::
spin_lock
>
lock
{
finished_lock_
};
return
finished_
;
}
void
execute
()
override
{
EASY_BLOCK
(
"execute root_task"
,
profiler
::
colors
::
LightGreen
);
function_
();
{
std
::
lock_guard
<
base
::
spin_lock
>
lock
{
finished_lock_
};
finished_
=
true
;
}
finished_
=
1
;
}
bool
internal_stealing
(
abstract_task
*
/*other_task*/
)
override
{
...
...
@@ -55,6 +54,7 @@ namespace pls {
master_task_
{
master_task
}
{}
void
execute
()
override
{
EASY_BLOCK
(
"execute root_task"
,
profiler
::
colors
::
LightGreen
);
do
{
steal_work
();
}
while
(
!
master_task_
->
finished
());
...
...
lib/pls/include/pls/internal/scheduling/scheduler.h
View file @
72bffbb1
...
...
@@ -2,6 +2,7 @@
#ifndef PLS_SCHEDULER_H
#define PLS_SCHEDULER_H
#include <easy/profiler.h>
#include <array>
#include <iostream>
...
...
@@ -64,6 +65,7 @@ namespace pls {
template
<
typename
Function
>
void
perform_work
(
Function
work_section
)
{
EASY_FUNCTION
();
root_task
<
Function
>
master
{
work_section
};
// Push root task on stacks
...
...
lib/pls/src/internal/scheduling/abstract_task.cpp
View file @
72bffbb1
#include <easy/profiler.h>
#include "pls/internal/scheduling/thread_state.h"
#include "pls/internal/scheduling/abstract_task.h"
#include "pls/internal/scheduling/scheduler.h"
...
...
@@ -6,6 +8,7 @@ namespace pls {
namespace
internal
{
namespace
scheduling
{
bool
abstract_task
::
steal_work
()
{
EASY_FUNCTION
(
profiler
::
colors
::
Orange
);
auto
my_state
=
base
::
this_thread
::
state
<
thread_state
>
();
auto
my_scheduler
=
my_state
->
scheduler_
;
...
...
@@ -15,14 +18,19 @@ namespace pls {
auto
target_state
=
my_scheduler
->
thread_state_for
(
target
);
// TODO: Cleaner Locking Using std::guarded_lock
EASY_BLOCK
(
"Acquire Thread Lock"
,
profiler
::
colors
::
Red
)
target_state
->
lock_
.
lock
();
EASY_END_BLOCK
;
// Dig down to our level
EASY_BLOCK
(
"Go to our level"
)
abstract_task
*
current_task
=
target_state
->
root_task_
;
while
(
current_task
!=
nullptr
&&
current_task
->
depth
()
<
depth
())
{
current_task
=
current_task
->
child_task_
;
}
EASY_END_BLOCK
;
EASY_BLOCK
(
"Internal Steal"
)
if
(
current_task
!=
nullptr
)
{
// See if it equals our type and depth of task
if
(
current_task
->
unique_id_
==
unique_id_
&&
...
...
@@ -37,10 +45,12 @@ namespace pls {
current_task
=
current_task
->
child_task_
;
}
}
EASY_END_BLOCK
;
// Execute 'top level task steal' if possible
// (only try deeper tasks to keep depth restricted stealing)
EASY_BLOCK
(
"Top Level Steal"
)
while
(
current_task
!=
nullptr
)
{
auto
lock
=
&
target_state
->
lock_
;
if
(
current_task
->
split_task
(
lock
))
{
...
...
@@ -50,6 +60,7 @@ namespace pls {
current_task
=
current_task
->
child_task_
;
}
EASY_END_BLOCK
;
target_state
->
lock_
.
unlock
();
}
...
...
lib/pls/src/internal/scheduling/fork_join_task.cpp
View file @
72bffbb1
#include <easy/profiler.h>
#include "pls/internal/scheduling/scheduler.h"
#include "pls/internal/scheduling/fork_join_task.h"
...
...
@@ -16,9 +18,11 @@ namespace pls {
}
void
fork_join_sub_task
::
execute
()
{
EASY_BLOCK
(
"execute sub_task"
,
profiler
::
colors
::
Green
);
tbb_task_
->
currently_executing_
=
this
;
execute_internal
();
tbb_task_
->
currently_executing_
=
nullptr
;
EASY_END_BLOCK
;
wait_for_all
();
if
(
parent_
!=
nullptr
)
{
...
...
@@ -40,13 +44,18 @@ namespace pls {
void
fork_join_sub_task
::
wait_for_all
()
{
while
(
ref_count_
>
0
)
{
EASY_BLOCK
(
"get local sub task"
,
profiler
::
colors
::
Blue
)
fork_join_sub_task
*
local_task
=
tbb_task_
->
get_local_sub_task
();
EASY_END_BLOCK
if
(
local_task
!=
nullptr
)
{
local_task
->
execute
();
}
else
{
// Try to steal work.
// External steal will be executed implicitly if success
if
(
tbb_task_
->
steal_work
())
{
EASY_BLOCK
(
"steal work"
,
profiler
::
colors
::
Blue
)
bool
internal_steal_success
=
tbb_task_
->
steal_work
();
EASY_END_BLOCK
if
(
internal_steal_success
)
{
tbb_task_
->
last_stolen_
->
execute
();
}
}
...
...
@@ -63,6 +72,7 @@ namespace pls {
}
bool
fork_join_task
::
internal_stealing
(
abstract_task
*
other_task
)
{
EASY_FUNCTION
(
profiler
::
colors
::
Blue
);
auto
cast_other_task
=
reinterpret_cast
<
fork_join_task
*>
(
other_task
);
auto
stolen_sub_task
=
cast_other_task
->
get_stolen_sub_task
();
...
...
@@ -80,6 +90,7 @@ namespace pls {
}
bool
fork_join_task
::
split_task
(
base
::
spin_lock
*
lock
)
{
EASY_FUNCTION
(
profiler
::
colors
::
Blue
);
fork_join_sub_task
*
stolen_sub_task
=
get_stolen_sub_task
();
if
(
stolen_sub_task
==
nullptr
)
{
return
false
;
...
...
Write
Preview
Markdown
is supported
0%
Try again
or
attach a new file
Attach a file
Cancel
You are about to add
0
people
to the discussion. Proceed with caution.
Finish editing this message first!
Cancel
Please
register
or
sign in
to comment