News for hackers. Stuff that matters.
Re:Random Acts of Optimization by saosebastiao on Thursday October 29, 2015 @04:29PM (#10474047)
I've gotten pretty good results from PGO as well as JIT optimizations (which isn't conceptually much different), but even if I hadn't, your comment is still a fallacy on two accounts:

1) JITs were slow until they weren't. Poor results from existing experimental PGO compilers do not prove that the concept is flawed.

2) Profiling is a heuristic substitute for a real cost model. The most advanced optimizing compilers forego profiling altogether because they evaluate optimizations against a hardware/architecture cost model. In other words, PGO isn't even necessary to accomplish what I'm talking about.

And yes, yes you can.

https://justindomke.wordpress.com/2009/02/23/the-stalin-comp...

http://www.ffconsultancy.com/ocaml/ray_tracer/languages.html

Re:Random Acts of Optimization by jerf on Thursday October 29, 2015 @04:45PM (#10474148)
And in theory, dependent types are the awesomest thing ever. But they're not ready yet, and there's no guarantee they ever will be. There's a gulf between even having a few focused, carefully-chosen proofs of concept and a practically-useful technology.

(For that matter, I just grabbed that example off the top of my head as the thing I knew that most resembled the various fancy optimizers I've seen over the years, but it seems to me that dependent typing are full of things that would be darned useful for an optimizer. It would be funny if the fastest path towards whole-program optimization went through dependent types being practical.)

Re:Random Acts of Optimization by johncolanduoni on Thursday October 29, 2015 @06:06PM (#10474715)
> And in theory, dependent types are the awesomest thing ever. But they're not ready yet, and there's no guarantee they ever will be.

I think the reason dependent types haven't made it into practical programming has less to do with them being not ready and more to do with them being fundamentally challenging to learn. I have a pretty strong pure math background and it still took me a significant amount of time to wrap my head around them to the point that I could write non-trivial proofs using them, let alone programs. For programmers who have never done much formal logic or who have never heard of a category, I can't imagine how much time it would take to get the prerequisite knowledge.

Another issue is the fundamental problem of creating a new language; you have to build up a ton of extra machinery (libraries, tooling, etc.) to even think about it becoming more than a toy, even if you don't care about it being mainstream. Rust has the backing of a large open-source company and a ton of really smart people dedicated to bringing it to primetime, but it is still going to take a long time for it to get there. Dependently typed languages like F* and Idris don't have the same manpower behind them and also have to face fundamental challenges like integrating formally verified code with a world full of unverified (but useful) code.

Re:Random Acts of Optimization by saosebastiao on Thursday October 29, 2015 @04:54PM (#10474201)
I think dependent types are amazing, but even the tiny subset of dependent types known as refinement types opens up huge avenues for optimization. Some optimizations (loop nest optimizations for a simple example) might only be feasible with static range definitions or with profile guided optimization...but if you throw in refinement types, you now have static guarantees as to the cardinality or range of a variable, and loop nest optimizations can be statically determined. I think full dependent typing could be great, but compilers can still be improved with small language improvements.
Re:Random Acts of Optimization by tel on Thursday October 29, 2015 @05:18PM (#10474379)
They're ready in small forms in Haskell today. The compiler doesn't take a whole lot of advantage of these yet, but you start to see avenues toward it.
Re:Random Acts of Optimization by haberman on Thursday October 29, 2015 @04:46PM (#10474152)
> Profiling is a heuristic substitute for a real cost model.

Even a perfect cost model can't tell you the difference between hot paths and cold paths. Profiling can.

Re:Random Acts of Optimization by saosebastiao on Thursday October 29, 2015 @05:06PM (#10474285)
Actually, they can, and quite well actually. The dotnet native compiler actually does quite a few analyses for "hotness", mostly centered around maintaining cache locality. However if there is a lot of entropy in the possible code paths, it might not be computationally feasible to analyze the hotness of all of the paths. In those cases, we accept heuristic substitutes, of which both JITs and PGO have shown good results.
Re:Random Acts of Optimization by haberman on Thursday October 29, 2015 @05:51PM (#10474619)
> it might not be computationally feasible to analyze the hotness of all of the paths

It has nothing to do with computational feasibility. It has to do with the fact that it depends on the input. In many cases there simply is not enough information at compile time to determine which paths are hotter than others.

For example, error paths are generally very cold, but how is a compiler supposed to know that a path is an error path? They just look like regular conditionals.

> In those cases, we accept heuristic substitutes, of which both JITs and PGO have shown good results.

JIT and PGO aren't heuristic, they are based on measurement. By your logic, if I look at the speedometer in my car that is just a "heuristic" of my speed. It's not a heuristic, it's an empirical measurement.

Re:Random Acts of Optimization by saosebastiao on Thursday October 29, 2015 @07:22PM (#10475165)
> It has nothing to do with computational feasibility. It has to do with the fact that it depends on the input. In many cases there simply is not enough information at compile time to determine which paths are hotter than others.

That's a fair point, but still quite a bit overstated. There are a variety of analyses that compilers can and do perform to understand possible input before it ever receives the input. I'm familiar with a handful of probabilistic and set-cardinality estimation techniques used in math optimization pre-solvers (a form of compiler) that do exactly this, and I'm pretty sure both GCC and LLVM already do several similar analyses.

Furthermore, a very large subset of optimizations won't have any path dependence in the optimization that it choses, and a significant subset after that has choice thresholds that are trivially computable given an accurate cost model. When the former happens, profiling gives no advantage at all, and when the latter happens, it is typically more efficient to branch on that threshold than it is to try to guess at which choice is more efficient by measuring sample input.

> For example, error paths are generally very cold, but how is a compiler supposed to know that a path is an error path? They just look like regular conditionals.

That might be true of some languages, but not so with others. Exceptions are obviously error paths, and several strongly typed languages have error paths encoded in the type system.

> JIT and PGO aren't heuristic, they are based on measurement. By your logic, if I look at the speedometer in my car that is just a "heuristic" of my speed. It's not a heuristic, it's an empirical measurement.

But that's the thing, measurement of your input is a heuristic. If your program execution is heavily influenced by its input, and you optimize using profiling, you are assuming that past input is predictive of future input. That assumption is a heuristic. It might be a really good heuristic, but it still is a heuristic.

Re:Random Acts of Optimization by haberman on Thursday October 29, 2015 @07:57PM (#10475363)
> Furthermore, a very large subset of optimizations won't have any path dependence in the optimization that it choses

Hot code should be: optimized for speed, inlined, unrolled (where it helps), kept local with other hot code, and register-allocated with other hot code.

Cold code should be: optimized for size, not inlined, not unrolled, kept away from hot code, and should never be accommodated in register allocation for any reason (ie. should never reduce availability of registers needed by hot paths).

So I don't agree that optimizations should be performed in a path-independent way.

> But that's the thing, measurement of your input is a heuristic.

By that standard, everything is heuristic. The assumption that your cost model matches the actual CPU the code will run on is a heuristic.

Re:Random Acts of Optimization by barrkel on Friday October 30, 2015 @10:07PM (#10481527)
If we followed your advice, the differential between worst case and best case performance would be large.

There are good reasons for preferring a less extreme difference in potential performance for general computing with AOT compiled languages. In particular, we should ensure the worst case performance isn't awful. We wouldn't want e.g. an occasional exception getting thrown to act like a DoS attack on a web server.

Re:Random Acts of Optimization by haberman on Saturday October 31, 2015 @02:21PM (#10483556)
Nothing I was suggesting would create "awful" performance. Inlining and unrolling are not necessary to get reasonable performance -- just try gcc or Clang with -Os.

For an occasional exception to "act like a DoS attack", the cold path would have to be thousands or millions of times slower than the hot paths. But compiler optimizations don't create anywhere near these kinds of constant factors. Even from -O0 to -O3 is more like a 5-10x difference, maybe 100x in extreme cases. The difference between -Os and -O3 is much more modest, more like 0-50%. Nothing that is going to cause anything remotely resembling a DoS.

Re:Random Acts of Optimization by vitalyd on Thursday October 29, 2015 @06:34PM (#10474871)
>For example, error paths are generally very cold, but how is a compiler supposed to know that a path is an error path? They just look like regular conditionals.

At least for .NET that path is likely throwing an exception. If you don't have a profile, statically assuming those paths are cold is likely to be correct.

Re:Random Acts of Optimization by haberman on Thursday October 29, 2015 @07:01PM (#10475035)
That's an interesting point -- my background is more in C and C++-without-exceptions. But in exception-based languages, I can see how the presence of "throw" would be a very useful signal for static analysis.
Re:Random Acts of Optimization by vitalyd on Thursday October 29, 2015 @07:19PM (#10475145)
Yes, this is a side benefit to exceptions (or some other language's way to highlight errors and thus likely cold paths).
Re:Random Acts of Optimization by darkmighty on Thursday October 29, 2015 @07:59PM (#10475375)
> It has to do with the fact that it depends on the input.

A special case of: https://en.wikipedia.org/wiki/No_free_lunch_in_search_and_op...

This does not hold, of course, if you could automatically simulate the users themselves...

Re:Random Acts of Optimization by vitalyd on Thursday October 29, 2015 @06:31PM (#10474852)
Profiling comes with its own baggage, nothing's perfect. A lot of times one actually knows the likely/unlikely paths at development time; that's certainly true for error conditions. It makes sense to inform the compiler of those cases using the builtins.
Re:Random Acts of Optimization by haberman on Thursday October 29, 2015 @07:08PM (#10475073)
Sure I'm not saying profiling is perfect. I was just taking exception to the statement that profiling is nothing but a poor man's substitute for a cost model.

Explicit annotations for hot/cold paths are useful, I agree. I would probably be more excited about them except last time I actually tried them with GCC, they slowed my code down (admittedly this was five years ago or so). But that's just an implementation problem, the concept is sound.

Re:Random Acts of Optimization by vitalyd on Thursday October 29, 2015 @07:17PM (#10475133)
I agree - profile and cost model are orthogonal and complementary. Former tells you what to optimize, latter dictates how to optimize it.