32
edits
Changes
→Business Point of View Comparison for STL and TBB
'''GPU621 Darth Vector: C++11 STL vs TBB Case Studies'''
''Join me, and together we can fork the problem as master and thread''
==Generic Programming==
Generic Programming is a an objective when writing code to make algorithms reusable and with the least amount of specific code. Intel describes generic programming as "''writing the best possible algorithms with the least constraints''". An example of generic code is STL's templating functions which provide generic code that can be used with many different types without requiring much specific coding for the type( an addition template could be used for int, double, float, short, etc without requiring re-coding). A non-generic library requires types to be specified, meaning more type-specific code has to be created. [[File:gputemplates.PNG |thumb|center|600px| An example of generic coding]]
==TBB Background==
==STL Background==
The STL was created as a general purpose computation library that a focus on generic programming. The STL uses templates extensively to achieve compile time polymorphism. In general the library provide four components: algorithms, containers, functions and iterators. The library was, mostly, created by Alexander Stepanov due to his ideas about generic programming and its potential to revolutionize software development. Because of the ability of C++ to provide access to storage using pointers, C++ was used by Stepanov, even though the language was still relatively young at the time. After a long period of engineering and development of the library, it obtained final approval in July 1994 to become part of the language standard. ==List of A Comparison from STL Functions:==In general, most of STL is intended for use within a serial environment. This however changes with C++17's introduction of parallel algorithms.
<u>'''Algorithms'''</u>
Are supported by STL for various algorithms such as sorting, searching and accumulation. All can be found within the header "'''<algorithm>'''". Examples include sort() and reverse functions()functions.
<u>'''STL iterators'''</u>
<u>'''Containers'''</u>
STL supports a variety of containers for data storage. Generally these containers are supported in parallel for read actions, but does not safely support writing to the container with or without reading at the same time. There are several header files that are included such as "'''<vector>'''", "'''<queue>'''", and "'''<deque>'''".'''Most STL containers do not support concurrent operations upon them.'''They are coded as:<pre>#include<vector>#include<dequeue>#include<queue> int main(){ vector<type> myVector; dequeue<type> cards; queue<type> SenecaYorkTimHortons; }</pre>
void foo(){std::allocater<type> name;}</pre> ==A Comparison from TBB=====Containers==='''<u>concurrent_queue</u>''' : This is the concurrent version of the STL container Queue. This container supports first-in-first-out data storage like its STL counterpart. Multiple threads may simultaneously push and pop elements from the queue. Queue does NOT support and front() or back() in concurrent operations(the front could change while accessing). Also supports iterations, but are slow and are only intended for debugdebugging a program. This is defined within the header "'''tbb/concurrent_queue.h'''" and is coded as: <pre>
#include <tbb/concurrent_queue.h>
//....//
tbb:concurrent_queue<typename> name; </pre>
'''<u>concurrent_vector</u>''' : This is a container class for vectors with concurrent(parallel) support. These vectors do not support insertion or erase operations but do support operations done by multiple threadssuch as push_back(). Note that when elements are inserted, they cannot be removed without calling the clear() member function on it, which removes every element in the array. The container when storing elements does not guarantee that elements will be stored in consecutive addresses in memory. This is defined within the header "'''tbb/concurrent_vector.h'''" and is coded as: <pre>
#include <tbb/concurrent_vector.h>
//...//
tbb:concurrent_vector<typename> name; </pre>
'''<u>concurrent_hash_map</u>''' : A container class that supports hashing in parallel. The generated keys are not ordered and there will always be at least 1 element for a key. Defined within "'''tbb/concurrent_hash_map.h'''"
===Algorithms===
<u>'''parallel_for:'''</u> Provides concurrent support for for loops. This allows data to be divided up into chunks that each thread can work on. The code is defined in "'''tbb/parallel_for.h'''" and takes the template of:
<pre>
foo parallel_for(firstPos, lastPos, increment { boo()}
</pre>
<u>'''parallel_scan:'''</u> Provides concurrent support for a parallel scan. Intel promises it may invoke the function up to 2 times the amount when compared to the serial algorithm. The code is defined in "'''tbb/parallel_scan.h'''" and according to intel takes the template of:
<pre>
void parallel_scan( const Range& range, Body& body [, partitioner] );
</pre>
<u>'''parallel_invoke:'''</u> Provides support for parallel calling to functions provided in the arguments. It is defined within the header "'''tbb/parallel_invoke.h'''" and is coded as: <pre>
tbb:parallel_invoke(myFuncA, myFuncB, myFuncC);
</pre>
===Allocaters===
Handles memory allocation for concurrent containers. In particular is used to help resolve issues that affect parallel programming. Called '''scalable_allocater<type>''' and '''cache_aligned_allocater<type>'''. Defined in "'''#include <tbb/scalable_allocator.h>'''"
==TBB Memory Allocation & Fixing Issues from Parallel Programming==
TBB provides memory allocation just like in STL via the '''std::allocater''' template class. Where TBB's allocater though improves, is through its expanded support for common issues experienced in parallel programming. These allocaters are called '''scalable_allocater<type>''' and '''cache_aligned_allocater<type>''' and ensure that issues like '''Scalability''' and '''False Sharing''' performance problems are reduced.
===False Sharing===
As you may have seen from the workshop "False Sharing" a major performance hit can occur in parallel when data that sits on the same cache line in memory is used by two threads. When threads are attempting operations on the same cache line the threads will compete for access and will move the cache line around. The time taken to move the line is a significant amount of clock cycles which causes the performance problem. Through TBB, Intel created an allocated known as '''cache_aligned_allocater<type>'''. When used, any objects with memory allocation from it will never encounter false sharing. Note that if only 1 object is allocated by this allocater, false sharing may still occur. For compatability's sake(so that programmers can simply use "find and replace"), the cache_aligned_allocater takes the same arguments as the STL allocater. If you wish to use the allocater with STL containers, you only need to set the 2nd argument as the cache_allocater object.
The following is an example provided by Intel to demonstrate this:
<pre>
std::vector<int,cache_aligned_allocator<int> >;
</pre>
===Scaling Issue===
When working in parallel, several threads may be required to access shared memory which causes a performance slow down from forcing a single thread to allocate memory while other threads are required to wait. Intel describes this issue in parallel programming as '''Scalability''' and answers the issue with '''scalable_allocater<type>''' which permits concurrent memory allocation and is considered ideal for "''programs the rapidly allocate and free memory''".
==Lock Convoying Problem==
===What is a Lock?===
A Lock(also called "mutex") is a method for programmers to secure code that when executing in parallel can cause multiple threads to fight for a resource/container for some operation. When threads work in parallel to complete a task with containers, there is no indication when the thread reach the container and need to perform an operation on it. This causes problems when multiple threads are accessing the same place. When doing an insertion on a container with threads, we must ensure only 1 thread is capable of pushing to it or else threads may fight for control. By "Locking" the container, we ensure only 1 thread accesses at any given time.
To use a lock, you program must be working in parallel(ex #include <thread>) and should be completing something in parallel. You can find c++11 locks with #include <mutex>
[[File:Gpulockwhat.PNG |thumb|center|700px| Mutex Example]]
Note that there can be problems with locks. If a thread is locked but it is never unlocked, any other threads will be forced to wait which may cause performance issues. Another problem is called "Dead Locking" where each thread may be waiting for another to unlock (and vice versa) and the program is forced to wait and wait .
===Parallelism Problems & Convoying in STL===
Within STL, issues arise when you attempt to access containers in parallel. With containers, when threads update the container say with push back, it is difficult to determine where the insertion occurred within the container(each thread is updating this container in any order) additionally, the size of the container is unknown as each thread may be updating the size as it goes (thread A may see a size of 4 while thread B a size of 9).
For example, in a vector we can push some data to it in parallel but knowing where that data was pushed to requires us to iterate through the vector for the exact location .
If we attempt to find the data in parallel with other operations ongoing, 1 thread could search for the data, but another could update the vector size during that time which causes problems with thread 1's search as the memory location may change should the vector need to grow(performs a deep copy to new memory).
Locks can solve the above issue but cause significant performance issues as the threads are forced to wait for each other before continuing. This performance hit is known as '''Lock Convoying'''.
[[File:DarthVector ThreadLock.PNG |thumb|center|600px| Performance issues inside STL]]
===Lock Convoying in TBB===
TBB attempts to mitigate the performance issue from parallel code when accessing or completing an operation on a container through its own containers such as concurrent_vector.
Through '''concurrent_vector''', every time an element is accessed/changed, a return of the index location is given. TBB promises that any time an element is pushed, it will always be in the same location, no matter if the size of the vector changes in memory. With a standard vector, when the size of the vector changes, the data is copied over. If any threads are currently traversing this vector when the size changes, any iterators may no longer be valid. This support also goes further for containers so that multiple threads can iterate through the container while another thread may be growing the container. An interesting catch though is that anything iterating may iterate over objects that are being constructed, ensuring construction and access remain synchronized.
[[File:Gpuconcur.PNG |thumb|center|600px| concurrent_vector use with multiple threads]]
TBB also provides its own versions of the mutex such as ''spin_mutex'' for when mutual exclusion is still required.
You can find more information on convoying and containers here: https://software.intel.com/en-us/blogs/2008/10/20/tbb-containers-vs-stl-performance-in-the-multi-core-age
==A Comparison between Serial Vector and TBB concurrent_vector==
''If only you knew the power of the Building Blocks''
Using the code below, we will test the speed at completing some operations regarding vectors using the stl library with stl's <u>'''vector'''</u> and tbb's <u>'''concurrent_vector'''</u>.
The code below will perform a "push back" operation both in serial and concurrent. Then, it measure the time taken to complete an '''n''' of push back operations.
<nowiki>
#include <iostream>
</nowiki>
[[File:Gputable.PNG |thumb|center|1200px| A speed comparison between concurrent and serial vectors]]
==Lock Convoying Problem=The Speed Improvement====What As the table suggests, completing push back operations using tbb's concurrent vector allows for increased performance against a serial connection. TBB additionally provides a benefit that it will never need to resize the vector as pushback operations are completed. In STL, the vector is a Lock?===dynamically allocated which requires it to reallocate and copy memory over which may further slow down the push back operation.
*Efficiency which is the measure of processor utilization in a parallel program
====STL and the Threading Libraries====
'''What you don’t need to worry about'''
*Making sorting, searching algorithms.
*Partitioning data.
*Array algorithms; like copying, assigning, and checking data
Note all algorithms is done in serial, and may not be thread safe
====TBBWorries and Responsibilities====*Thread Creation, terminating, and synchronizing, partitioning, thread creation, and management is managed by TBB. This make you need not to worry about the heavy constructs of threads which are close to the hardware level.
*Making a solution from close to hardware level allows Own Parallel algorithms (makes you need not to be flexible to worry about the solution you heavy constructs of threads that are wanting to make. But present in the major downside is the requirement lower levels of implementing the foundations first to make your solution workprogramming. It also simple map, scan, pipeline, or reduce TBB has the potential of making your program inefficient if not done correctly.you covered
'''Benefit'''
The downside of TBB is since much of the close to hard hardware management is done be hide the scenes, it makes you has a developer have less control on finetuning your program. Unlike how STL with the threading library allows you to do.
===Licensing===
TBB is dual-licensed as of September 2016
*COM license as part of suites products. Offers one year of technical support and products updates
*Apache v2.0 license for Open source code. Allows the user of the software the freedom to use the software for any purpose, to distribute it, to modify it, and to distribute modified versions of the software, under the terms of the license, without concern for royalties.
===Companies and Products that uses TBB===
*DreamWorks (DreamWorks Fur Shader)
*Blue Sky Studios (animation and simulation software)
*Pacific Northwest National Laboratory (Ultrasound products)
*More: https://software.intel.com/en-us/intel-tbb/reviews