Thursday, July 13, 2017

PhD Defense - Automated Data-Driven Hint Generation for Learning Programming

Kelly Rivers defended her PhD work this afternoon.  She will returning to CMU this fall as a teaching professor.

Student enrollment is increasing, so more work is needed to automate the support, as TAs / instructors are not scaling.  Prior work (The Hint Factory) developed models based on prior student submissions, and then a current student's work can be found within the model thus providing suggestions for how to proceed.  However, programming may not fit within this model due to the larger and more varied space for which students can solve the problems.

First, student code proceeds through a series of canonicalization steps - AST, anonymized, simplification.  Such that the following python code is transformed:

import string
def any_lowercase(s):
  lst = [string.ascii_lowercase]
  for elem in s:
    if (elem in lst) == True:
      return True
    return False


import string
def any_lowercase(p0):
  for v1 in p0:
    return (v1 in string.ascii_lowercase)

Studies then went over 41 different problems with hundreds of correct solutions and thousands of incorrect solutions.  The model can then generate the edits and chain these hints as necessary.  In more than 99.9% of cases, the model could successfully generate a hint chain to reach a correct solution.

To further test this model and approach, the model started with the empty space (just teacher solution) and was compared against the final model.  Ideally, the final model will propose fewer edits than the initial model.  And for 56% of problems, this was true.  40% of problems were already optimal.  And 3% are opportunities for improvement to the model.

Next, given this model exists, how do the hints impact student learning?  Select half of the students to give them access to the hint model optionally.  Using a pre / post assessment, the measurement was a wash.  Instead, a second study was designed that required the students to use the system within a two hour OLI module.  Hints would be provided with every submission and either before or after the midtest in the OLI module.  Only 1/2 of the students actually proceeded through the module in order.  However, most learning was just within the pretest->practice->midtest, so adding those students increased the population.  The results show that the hints reduce the time required to learn the equal amount.

From interviews with students, students need and want targeted help on their work.  However, the hints generated thus far were not always useful.  Proposed another study based on different styles of hints: location, next-step, structure, and solution.  This study found that participants with lower expertise wanted more detailed hints.  Hint usage would sometimes be for what is wrong versus how to solve it.  And often, students know what to do, and just need to reference (via example / prior work) how to do this, rather than hinting what to do.

Monday, July 10, 2017

Wrote my own Calling Convention

I have been listening to the Dungeon Hacks audiobook, and it has both reminded of my past joys of playing Angband, as well as some interesting little "hacks" I did in high school.

In high school, I wrote many programs on my TI-83, often to help me in other classes, which lead to an interesting conversation:
Student: "Teacher, Brian has written programs on his calculator for this class."
Teacher: calls me forward "Brian, is this true?"
Me: "Yes."
Teacher: "Did you write them yourself?"
Me: "Yes."
Teacher: "I do not see what the problem is."

And besides writing useful programs, I also wrote games.  All in the TI-83's BASIC.  However, the TI-83 only had 24kB of space for programs and data.  And eventually I started exceeding this limit.  So I started finding interesting hacks that would reduce the space of programs, such as the trailing " of a string is not required if that string ends the line.  Each would save 1 byte, but it started to add up.

Now, the TI-BASIC on the 83 was very primitive.  Particularly it had no GOSUB instruction, only GOTO.  So you cannot write functions / subroutines and reuse code, which would both improve the design and reduce the space required.  Now I already knew the basics (no pun intended) of C, so I knew that code should be able to call other functions and get values back.  But TI-BASIC would let you call another program and then return to the calling program after the called one finished.  That's like a function call, right?

Therefore, I wrote a library program.  Variables were global, so the calling program could set specific parameters in the global variables, call the library program.  The library would use one parameter to determine which functionality to execute, update the necessary globals and then exit, thus returning.  And consequently, I had a library of common functions which sped up my development and reduced the space I needed.

And so it was only in listening to the audio book last week, did I realize that long ago I had developed a simple calling convention in high school.  And now I teach, among other things, the x86-64 Linux ABI (i.e. calling convention) to college students.

A calling convention just dictates how each register is used when calling a function.  Which ones need to be preserved, arguments, return value, etc.  And it also dictates the management of the stack.

Wednesday, May 3, 2017

PhD Defense - Meeting Tail Latency SLOs in Shared Networked Storage

Today I went to Timothy Zhu's PhD thesis defense.  His work is on achieving better sharing of data center resources to improve performance, and particularly to reduce tail latency.  He also TA'd for me last fall.

Workloads are generally bursty, and their characteristics are different.  Furthermore, they may have service level objectives (SLOs), and the system needs to meet these different objectives.  And the system contains a variety of resources that must be shared in some form.  It is not sufficient to just divide the bandwidth.  Nor can the system measure the latency and try reacting, particularly as bursty workloads do not give sufficient time to react.  While each workload has deadlines, it would be too complex to tag request packets with the deadlines for queuing and routing.  However, the deadlines can be used to generate priorities for requests.

The system is architected to have storage and network enforcement components to ensure QoS.  There is also a controller that receives an initial trace to characterize each workload, and that workload's SLOs.  The controller works through a sequence of analyses to successfully place each workload into the overall system.

Effectively, each workload is assigned a "bucket" of tokens, where the bucket size provides the ability to handle bursts and the rate that tokens are added covers the request rate for the workload.  Shorter burstier workloads receive large buckets and low rates, while constant workloads with little bursts have high rates and small buckets.  In both cases, only when the bucket is empty, is the workload rate-limited in its requests, and these requests receive the lowest priority.  Deterministic Network Calculus (DNC) to model the worst-case queue scenarios.  This plots two curves: the requesting flow and the service curve, both plotted as tokens by function of window size (dt).  The maximum distance between the curves is the maximum latency.

Using three traces: DisplayAds, MSN, and LiveMaps, they tested three approaches: Cake (reactive approach), earliest deadline first, and Timothy's scheme (PriorityMeister).  His scheme did significantly better than the others at meeting the SLOs.  However, the DNC analysis was based on achieving 100% and not the SLO's 99% (or other percentile success).  Depending on the characteristics, there can be significant differences between these guarantees.  To model the latency percentiles, Stochastic Network Calculus (SNC) can achieve this; however, the math is significantly more complex.  And the math had not previously been applied to this problem.  DNC is still better when assuming that bursts are correlated or the system is in an adversarial setting.  Reducing these assumptions (uncorrelated workloads), the SNC-based analysis permitted the system to admit 3x workloads versus the DNC analysis.

Workloads have a curve of satisfying bucket sizes and token rate pairs.  Many systems require the user to provide its rate limit.  Other systems use simple heuristics to either find the "knee of the curve" or select a rate limit as a multiple of the average rate.  However, for an individual workload, all pairs are satisfying, it is only when workloads are combined in a system do the different pairs matter.  The configurations of the set of workloads on the system can be solved for using a system of linear equations.  Therefore, when placing new workloads, the controlling architecture can find successful placements, while potentially reconfiguring the workloads assigned.

One extension would be addressing failure modes.  Currently, the system is assumed to be at degraded performance when components have failed.

Monday, May 1, 2017

Book Review: Multicore and GPU Programming: An Integrated Approach

I wanted to like this book, Multicore and GPU Programming: An Integrated Approach, and be able to use it in the classroom.  I teach a class where we cover OpenMP, MPI, Cilk, and Cuda.  Yet, I would not use this book and I have told the publishers this as well, to which they acknowledged my reasoning.

First, the author elects to use QtThreads for the base parallel programming approach.  He notes that he had considered using pthreads or C++11 thread support, and comments that he rejected C++11 threads for the book as the support was incomplete at the time.  Pthreads is left without comment.  That may be, but I have heard of and used those options, while QtThreads is something that I have never considered.

Second, the book serves as an admirable API reference.  For one or a set of function calls, significant space is dedicated to how that call works and illustrating examples for it.  Structured Parallel Programming also covers many APIs, yet it maintains a feel of being about parallel programming in general rather that the calls specifically.  However, that work covers different API sets so the two are not explicitly comparable.

Third, and this issue is really more for the editors rather than the author, the typesetting on each page is poor.  There is significant white space left bordering the text on each page.  Furthermore, the code wraps, and often not for being long code, but for long comments and comments on the same line as the code.  I understand that in programming these are stylistic choices; however, the impact of finding line wraps from long comments leaves the text looking unprofessional.  I must assume that the author wrote the code separately and then provided for being included into the book, but the editor failed to make allowances for typesetting.

In conclusion, I wanted to like and use the book.  Whenever I speak with the publisher, they always direct me to it, I just have to hope for something else to come along.  Use it as a reference perhaps, but I am cautious in my recommendation.

(This book was provided free by the publisher to review for possible use in the classroom.)

Friday, April 21, 2017

Repost: What Makes a Program Elegant?

In a recent issue of the Communications of the ACM, there was a short article titled, What Makes a Program Elegant?  I found it an interesting discussion that has summarized well the characteristics in elegant programming: minimality, accomplishment, modesty, and revelation.  Revelation is one that I had not considered before, but I think it is most important of all.  There are some code sequences that I have written, which the elegance has rested most of all on its revelation.  Using and showing some aspect of computers and programming that I have never seen before, or revealing that there is a modest way to accomplish something new or differently.

Monday, March 13, 2017

Book Review: Optimized C++: Proven Techniques for Heightened Performance

I have spent significant time on performance issues and have been in search of a book that can summarize the diversity of issues and techniques well.  I hoped that Optimized C++: Proven Techniques for Heightened Performance would provide some of the guidance I want and
This book is not quite it.  There is good material here, yet I found repeatedly thinking that the author was not aware of the past 10(?) years of changes to the field.  Not an issue of the book was from the early 2000s, but it was published last year.

A key step in improving the performance of programs is measuring it.  There are a variety of techniques for doing so.  Tools based on instrumentation and tools based on sampling profiling.  I find greater value to using the sampling profiling tools (for measuring performance) due to their lower overhead and ability to pinpoint where in a function this cost exists.  Yet the book's focus is limited to gprof-esque approaches.  I tell students that this approach is best with deep call trees, which may be a greater issue for C++ programming specifically.

The author is somewhat dismissive to compiler optimizations and emphasizes that his observed benefit has been particularly limited to function inlining.  There are many more optimizations, and you should care about them.  But again, I wonder if his experience of C++ has been deep call trees that could particularly benefit from inlining.

In a take it or leave it, this work also discourages the use of dynamic libraries.  Yes, they impose a performance penalty, but they also provide valuable functionality.  It all depends on your use case for whether you should statically or dynamically link your code.  Code that is reused by separate executables should be in a dynamic library, as it reduces the memory requirements when running and reduces the effort to patch and update those executables.  Components that are only used by a single executable should be statically linked, unless the components are of significant size such that decoupling can still benefit memory usage and the updating process.

The author related that replacing printf with puts to just print a string has performance advantages, due to printf being a complicated "God function".  The basic point is valid that printf has significant functionality; however, the anecdote should be taken with a grain of salt.  Current compilers will do this optimization (replace printf with puts) automatically.

While most of the work provides small examples, the final chapters on concurrency (?) and memory management do not.  The concurrency chapter reads as a reference book, as it lists the various APIs available and what each does.  It would be better for the book to assume that the readers are familiar with these calls (as the author does with many other topics) and discuss possible optimizations within this scope.

To conclude, the book is not bad, but I also cannot say it is accurate on every point.  Especially with performance, programmers are apt to make prompt design decisions based on "their experience" or "recent publications".  Measure your code's performance.  Only then can you discern which techniques will provide value.

Saturday, March 11, 2017

Conference Time: SIGCSE 2017 - Day 2

I started my morning by attending my regular POGIL session.  I like the technique and using it in the classroom.  However, I should probably make the transition, attend the (all / multi-day) workshop, and perhaps get one of those "ask me about POGIL" pins.

Lunch was then kindly provided by the CRA for all teaching-track faculty in attendance.  There is the start of an effort to ultimately prepare a memo to departments for how to best support / utilize us (including me).  One thing for me is the recognition of how to evaluate the quality of teaching / learning.

Micro-Classes: A Structure for Improving Student Experience in Large Classes - How can we provide the personal interactions that are valuable, which enrollments are large / increasing?  We have a resource that is scaling - the students.  The class is partitioned into microclasses, where there is clear physical separation in the lecture room.  And each microclass has a dedicated TA / tutor.  Did this work in an advanced (soph/ junior) class on data structures?

Even though the same instructor taught both the micro and the control class, the students reported higher scores for the instructor for preparedness, concern for students, etc.  Yet, there was no statistical difference in learning (as measured by grades).

Impact of Class Size on Student Evaluations for Traditional and Peer Instruction Classrooms - How can we compare the effectiveness of peer instruction being using in courses of varying class sizes?  For dozens of courses, the evaluation scores for PI and non-PI classes were compared.  There was a statistical difference between the two sets and particularly for evaluating the course and instructor.  This difference exists even when splitting by course.  This difference does not stem from frequency of course, nor the role of the instructor (teaching, tenure, etc).