News for hackers. Stuff that matters.

palinkapika

Re:VirtualBox 7.0.10 download links have disappeared on Monday July 24, 2023 @07:01AM (#36845869)
UTM uses VirtioFS with Apple's virtualisation backend. Works out of the box with my Ubuntu server guests. You still need to mount the share as described here, but that's it: https://docs.getutm.app/guest-support/linux/#macos-virtiofs

What took me a bit to figure out is that this does not mount a single folder, but a virtual parent folder to which you can add and remove directories without changing the mount. This even works while the VM is running.

For example, mounting

  share /mnt/host virtiofs rw,nofail 0 0
in the guest and then sharing

  ~/Downloads
in UTM will share the Downloads folder as

  /mnt/host/Downloads
When using the QEMU backend, VirtFS is used instead, which does not use such a parent folder and requires restarting the VM to change shares iirc.
Re:Python performance: it’s not just the interpreter on Wednesday May 20, 2020 @11:02AM (#23247678)
> If we are talking about a call like Arrays.setAll(array, function) and you have one or two candidates for function, then you are mono- or bimorphic. If you have 25 candidates for function, you are highly polymorphic. I don't understand what you mean by a mono- or bimorphic scenario with up to 25 candidates.

Sorry, I phrased that poorly. What I was trying to say was that the moment we introduced a 3rd candidate the performance with C2 decreased by ~5x. Whereas with Graal the performance stayed virtually the same when introducing a 3rd or 4th candidate. And it was performing well. But it did decrease at some point after adding more and more candidates and then the performance hit was much bigger.

I will reach out to my colleagues still working on this next week (long weekend ahead where I live). If we can reproduce this with a current Graal release, I'll share the benchmark – always happy to learn something new and if it is 'just' why out benchmark is broken. :)

Re:Python performance: it’s not just the interpreter on Wednesday May 20, 2020 @07:32AM (#23245625)
The results for the parallel code are indeed interesting. However, the speedup < 2 with C2 is suspicious. Even on an old dual core / 4 thread system (common worker pool should default to 3 workers), I would expect it to be above 2 for this kind of workload.

The last time I looked into Graal (over a year ago), I ran into something I could not make sense of.

We were designing an API with methods similar to Arrays.setAll(double[], IntToDoubleFunction) which we expected to be called with large input arrays from many places in the code base.

The performance of the code generated by C2 dropped as one would expect once you were using more that two functions (lambda functions were no longer inlined and invoked for each element). For simple arithmetic operations the runtime increased by around 5x.

The performance of the code generated by Graal was competitive with C2's code for mono- and bimorphic scenarios for a much larger number of input functions. I don't recall the exact number, but I believe it was around 25. However, once we hit that threshold, performance tanked. I believe the difference was closer to 15x than than C2's 5x.

We never figured out what caused this.

Do you have an idea? Or, since this is probably outdated data, do you know what the expected behavior is nowadays? Is there maybe any documentation on this?

Re:Python performance: it’s not just the interpreter on Tuesday May 19, 2020 @07:53PM (#23241749)
Yes, this is strongly related to inlining. Without the warmup code C2 most likely inlines your lambdas into the code generated for the for each method. At least I have seen this behavior with similar APIs (you can use a tool called JITWatch to visualize compiled and inlined methods). However, this does not scale.

Last time I checked C2 inlines at most two implementations at a polymorphic call-site (in this case the line that calls your lambda). If you pass in more lambda functions, it won’t inline them and might de-optimize existing compiled code to remove previously inlined functions (there are also cases where one implementation that is heavily used gets inlined and the others will be invoked via a function call). Graal does not seem to have the same limit but when I tested it two years ago it had others averaging out to the same performance.

Thus, increasing the max inline level does not necessarily help (you can already do that in Java 8 and 11) for this particular problem. What you would want is that the compiler inlines the for each method and the local lambdas into the methods using the API (the benchmark methods in our case), i.e., you want the compiler to copy the hot loop somewhere higher up in the call stack and then optimize it. But apparently this is easier said than done.

But again, this is probably only relevant for workloads where there is very little work to do for each item.

Re:Python performance: it’s not just the interpreter on Tuesday May 19, 2020 @05:40PM (#23240617)
And of course I screwed up the warmup parameter...
Re:Python performance: it’s not just the interpreter on Tuesday May 19, 2020 @05:29PM (#23240493)
You might already know this, but there is one potential caveat with APIs like that when it comes to performance or at least measuring their performance. The hot loop (the code that actually iterates over your data points) does not live in your code base. But performance often depends on what the JIT compiler makes out of that loop. If the API is used at several locations in your program, the compiler might not be able to generate code that is optimal for your call-site and inputs. Instead, it will generate generalized code that works for all inputs but might be slower.

However, when writing benchmarks, there is often no other code around to force the JIT compiler to generate such generalized code.

The following code demonstrates this. If the parameter warmup is set, I invoke the forEach methods with different inputs first (they do the same but are different methods in the Java byte code). The purpose is to force the compiler to generate generalized code:

    @Param({"true", "false"})
    public boolean warmup;

    @Setup
    public void setup() {
        if (!warmup) {
            Interval.oneTo(20).forEach(
                    (IntProcedure) i -> Interval.oneTo(1_000_000).forEach(
                            (IntProcedure) j -> Integer.toString(j)));
            Interval.oneTo(20).forEach(
                    (IntProcedure) i -> Interval.oneTo(1_000_000).forEach(
                            (IntProcedure) j -> Integer.toString(j)));
            Interval.oneTo(20).forEach(
                    (IntProcedure) i -> Interval.oneTo(1_000_000).forEach(
                            (IntProcedure) j -> Integer.toString(j)));
        }
    }

    @Benchmark
    public void coollectionsBlackhole(Blackhole blackhole) {
        Interval.oneTo(20).forEach(
                (IntProcedure) i -> Interval.oneTo(1_000_000).forEach(
                        (IntProcedure)j -> blackhole.consume(Integer.toString(j))));
    }


    @Benchmark
    public void collectionsDeadCode() {
        Interval.oneTo(20).forEach(
                (IntProcedure) i -> Interval.oneTo(1_000_000).forEach(
                        (IntProcedure)j -> Integer.toString(j)));
    }
And the for loop implementations for reference:

    @Benchmark
    public void loopBlackhole(Blackhole blackHole) {
        for (int j = 0; j < 20; j++) {
            for (int i = 0; i < 1_000_000; i++) {
                blackHole.consume(Integer.toString(i));
            }
        }
    }

    @Benchmark
    public void loopDeadCode() {
        for (int j = 0; j < 20; j++) {
            for (int i = 0; i < 1_000_000; i++) {
                Integer.toString(i);
            }
        }
    }

    @Benchmark
    public void loopBlackholeOnly(Blackhole hole) {
        for (int j = 0; j < 20; j++) {
            for (int i = 0; i < 1_000_000; i++) {
                hole.consume(0xCAFEBABE);
            }
        }
    }
On my desktop machine, this gives me the following results (Java 11/Hotspot/C2):

  Benchmark              (warmup)  Mode  Cnt    Score   Error  Units
  collectionsDeadCode        true  avgt   99  162.482 ± 1.615  ms/op
  collectionsDeadCode       false  avgt   99  218.816 ± 4.217  ms/op
  coollectionsBlackhole      true  avgt   99  235.122 ± 1.362  ms/op
  coollectionsBlackhole     false  avgt   99  270.192 ± 1.627  ms/op
  loopBlackhole              true  avgt   99  207.214 ± 1.162  ms/op
  loopBlackhole             false  avgt   99  206.711 ± 0.932  ms/op
  loopBlackholeOnly          true  avgt   99   74.774 ± 0.180  ms/op
  loopBlackholeOnly         false  avgt   99   74.359 ± 0.180  ms/op
  loopDeadCode               true  avgt   99  143.394 ± 0.900  ms/op
  loopDeadCode              false  avgt   99  142.654 ± 0.795  ms/op
All results are for single threaded code. Now, the difference is not huge, but significant. Overall it looks like the invocation of toString(int) is not removed and accounts for most of the runtime.

Just to be clear: I am not saying one should stay way from the stream APIs. As soon as the work per item is more than just a few arithmetic operations, there is a good chance the difference in runtime is negligible. But when doing numerical work (aggregations etc.), a simple loop might be the better option.

Finally, these differences are of course compiler-dependent. For example, C2 might behave differently than Graal and who knows what the future brings.