Monday, February 20, 2017

Repost: Learn by Doing

I want to take a brief time to link to two of Mark Guzdial's recent posts.  Both including an important theme in teaching.  Students learn best by doing not hearing.  Oddly students commonly repeat this misconception.  If I structure our class time to place them as the ones doing something, rather than me "teaching" by speaking, the appraisal can be that I did not teach.  They may not dispute that they learned, but I failed to teach them.

Students learn when they do, not just hear.  And Learning in MOOCs does not take this requirement into account.

I have to regularly review these points.  So much so that I was able to give them to a group of reporters last week (part of new faculty orientation, but still).

Wednesday, February 8, 2017

Conference Attendance CGO (etc) 2017 - Day 3

Today is the last day for the conference.  I attended several more talks today and my student took 2nd in the student research competition.  So it has been a good attendance and I have received some interesting and valuable feedback on my own work, as well as finding some possible collaborations for the future.

Optimistic Loop Optimization
The compiler wants to optimize loops in the program.  However, C and LLVM IR have many complex characteristics that the Polyhedral model cannot represent, such as aliasing, wrapping, or out of bound accesses.  Rather than just assuming that these characteristics are not present, instead, the code can be analyzed to determine which violating characteristics may be present.  These assumptions are placed in the code, and can then be reduced to the set of preconditions for which the optimized loop can be executed.  Should the assumptions fail, the code instead can branch to the original version.  These optimizations can also be optimized (for example N < 127 implies N < 128).  For SPEC200x, the assumptions fail about 2% of the time and impose 4% runtime overhead.

Software Prefetching for Indirect Memory Accesses
What should we prefetch in software? A[x + N] is easy for hardware, A->next is hard for everyone, while A[B[x + N]] is easy for software and hard for hardware to predict.  So given a loop (such as exists in NAS is) that has this indirect structure, then prefetches can be inserted that will speedup the loop.  Yet, there are three key points for inserting these prefetch instructions.
- You have to prefetch both the A[B[]] and B[].  Otherwise, the prefetch will block on the B[].
- Having both prefetches requires that they are both offset from the actual access as well as each other.  Too close and they are not parallel.  Too far and the data is evicted before use.
- The first point raised that there is an actual load (not prefetch) of B[] and therefore needs to be bounds checked.

Tuesday, February 7, 2017

Conference Attendance CGO (etc) 2017 - Day 2

Several of the talks were great and very interesting.  Other talks particularly needed further presentation practice.  Unfortunately, sometimes this may come from English as a second language.  And so I am torn between wanting presenters to have practice and be open to a wider pool of researches, while also wanting to be able to easily understand the presentations.

Tapir: Embedding Fork-Join Parallelism into LLVM’s Intermediate Representation
Let's start by considering code that normalizes a vector.  This code takes 0.3s to run.  Then switch the "for" with a "cilk_for", and the execution time improves to 180s (w/ 18 cores).  When the compiler sees "cilk_for" or other parallel keywords, generally it converts these into runtime function calls that take in a function pointer for the parallel component.  (Similar to thread create routines taking in a function to execute).  With the function call, many optimization passes cannot cross the call, while previously being able to cross the "for".

Instead, let's propose three new instructions to include in the LLVM IR.  Supporting these lines required approximately 6000 lines of changes.  When the updated LLVM compiles a set of parallel programs, most can now reach 99+% work efficiency, which indicates that the parallel overhead is near 0.

Prior work would create parallel tasks symmetrically, for example each task would represent separate paths in the classic "diamond" CFG.  The problem is that the parallel program is actually taking both paths concurrently, which is not an expected behavior of the control flow.  Instead, the IR is asymmetric so that compilers can continue to reason about the basic blocks as a sequential code would appear.

Incremental Whole Program Optimization and Compilation
This covers the feature within Microsoft's Visual Studio compiler.  Each component stores hashes of the components on which it depends.  When a file is changed, it generates different hashes, which the compiler then can use to determine that its dependencies need to be re-analyzed and code gen'd.  These hash changes can then either propagate, if changed, or the compilation process will complete.

Optimizing Function Placement for Large-Scale Data-Center Applications
The common binaries for facebook are 10s-100s MBs in size.  These binaries have IPCs less than 1.0 (recall that processors can run above 2.0 and higher is better), and are experiencing frequent front-end stalls that are attributable to iTLB and I$ misses (as high as 60 per 1000, eww).  Hardware profilers can then determine the hot functions.  This information is then processed to determine the hot functions that should be clustered together.  These clusters are mapped to separate loader sessions that will load them using huge pages.

Minimizing the Cost of Iterative Compilation with Active Learning
There are too many possibilities for optimization.  Let's ask machine learning to figure this out.  The danger is always finding the right level of training to provide valuable insights without overfitting, etc.


Monday, February 6, 2017

Conference Attendance CGO (etc) 2017 - Day 1

I started the conference attendance this time on Saturday, with the LLVM-Performance workshop at which I presented an extension to my dissertation work.  I received some interesting and useful feedback from the other attendees, as well as saw additional possibilities of its usage and collaboration.  Now that it is Monday, it is time to attend some conference talks.  In the evening today, I will be being an advisor and watching one of my students present our work, which we practiced today so it should go great!

Checking Concurrent Data Structures Under the C/C++11 Memory Model
C/C++11 included additional keywords that allow specifying features of the memory model, previously covered.  In order to check data structure implementations, the data structures need to be further annotated so as to further describe valid and invalid executions.  For example, is a queue required to always return an element, or can it fail if an element was recently added?  Using these annotations, the authors were able to find issues and other identifications for the data structures.

Efficient Abortable-locking Protocol for Multi-level NUMA Systems
The impact of NUMA can be significant.  On the largest shared-memory machines, the difference between accessing lock data that is local to an SMT thread versus the farther distance is over 2000x slower.  To avoid this overhead, there is a hierarchy of locks created that mirrors the system's topology.  Each level of the hierarchy acts as a MCS-style queue lock.  How then can these threads abort from their queue?  In a single level, threads mark their status as aborted and are then skipped when handing off the lock.  In the hierarchy, the requirement of waiting on the specific level is passed along to the appropriate thread, as can be determined by using the lower level links.

The implementation was model checked and then run on a HP Superdome with 576 hardware threads.  The results showed that the lock implementation performs best when it respects all 4 levels of the NUMA (and NUCA) hierarchy.

Thread Data Sharing in Cache: Theory and Measurement
In this work, they collected memory access traces using Pin to determine the thread data sharing in parallel programs.  Particularly they worked to show how the different metrics of data sharing can be derived from a single pass of the trace, and is linear in trace size.  The concern is that the trace / analysis approaches are very slow and could potentially skew the results.  And when the results are derived from only one trace, there is additional question about their applicability.