Performance | Client side performance tool | Application Framework library

This issue was fixed by Google (10 February 2022).

You can now update Android Studio normally.

Thank you all for helping to bring this problem to Google's attention.

This issue was fixed by Google (10 February 2022).

You can now update Android Studio normally.

Reading a memory address does not clear the cache line.

Writing does.

So in a modern computer, there is RAM, and there are multiple layers of cache "around" the CPU (they are called L1, L2 and L3 cache, but the important part is that they are layers, and the CPU is at the middle). In a multi-core system, often the outer layers are shared; the innermost layer is usually not, and is specific to a given CPU.

Clearing the cache line means informing every other cache holding this memory "the data you own may be stale, throw it out".

Test and set writes true and atomically returns the old value. It clears the cache line, because it writes.

Test does not write. If you have another thread unsynchronized with this one, it reading the cache of this memory doesn't have to be poked.

The outer loop writes true, and exits if it replaced false. The inner loop waits until there is a false visible, then falls to outer loop. The inner loop need not clear every other cpu's cache status of the value of the atomic flag, but the outer has to (as it could change the false to true). As spinning could go on for a while, avoiding continuous cache clearing seems like a good idea.

It looks like GCC's naïveté about store-forwarding stalls is hurting its auto-vectorization strategy here. See also Store forwarding by example for some practical benchmarks on Intel with hardware performance counters, and What are the costs of failed store-to-load forwarding on x86? Also Agner Fog's x86 optimization guides.

(gcc -O3 enables -ftree-vectorize and a few other options not included by -O2, e.g. if-conversion to branchless cmov, which is another way -O3 can hurt with data patterns GCC didn't expect. By comparison, Clang enables auto-vectorization even at -O2, although some of its optimizations are still only on at -O3.)

It's doing 64-bit loads (and branching to store or not) on pairs of ints. This means, if we swapped the last iteration, this load comes half from that store, half from fresh memory, so we get a store-forwarding stall after every swap. But bubble sort often has long chains of swapping every iteration as an element bubbles far, so this is really bad.

(Bubble sort is bad in general, especially if implemented naively without keeping the previous iteration's second element around in a register. It can be interesting to analyze the asm details of exactly why it sucks, so it is fair enough for wanting to try.)

Anyway, this is pretty clearly an anti-optimization you should report on GCC Bugzilla with the "missed-optimization" keyword. Scalar loads are cheap, and store-forwarding stalls are costly. (Can modern x86 implementations store-forward from more than one prior store? no, nor can microarchitectures other than in-order Atom efficiently load when it partially overlaps with one previous store, and partially from data that has to come from the L1d cache.)

Even better would be to keep buf[x+1] in a register and use it as buf[x] in the next iteration, avoiding a store and load. (Like good hand-written asm bubble sort examples, a few of which exist on Stack Overflow.)

If it wasn't for the store-forwarding stalls (which AFAIK GCC doesn't know about in its cost model), this strategy might be about break-even. SSE 4.1 for a branchless pmind / pmaxd comparator might be interesting, but that would mean always storing and the C source doesn't do that.

If this strategy of double-width load had any merit, it would be better implemented with pure integer on a 64-bit machine like x86-64, where you can operate on just the low 32 bits with garbage (or valuable data) in the upper half. E.g.,

The original code can be re-written in the following way:

Here are some implications of calling component as function vs rendering it as element.

1. Potential violation of rules of hooks

When you call a component as a function (see TestB() below) and it contains usage of hooks inside it, in that case react thinks the hooks within that function belongs to the parent component. Now if you conditionally render that component (TestB()) you will violate one of the rules of hooks. Check the example below, click the re-render button to see the error:

Error: Rendered fewer hooks than expected. This may be caused by an accidental early return statement.

The very links you included in your question allow you to see the source code of each implementation which tells you that, no, there is no difference.

In fact elementAtOrNull literally just calls getOrNull.

I've been having the same problem (I'm on macOS Monterey), each time I try to take a screenshot the emulator crashes.

Sadly I haven't found a direct solution to this problem, that is a solution fixing the issue in the simulator. But I have learned that it is possible to take screenshots of the app from inside Android Studio, using Logcat.

Essentially, when you're running your app, if you go to the Logcat tab, there is a screenshot option which does seem to work without crashing. I've added a link to developer.android.com which explains how to do it.

Even thought this doesn't exactly fix the problem I hope it helps!

Take a screenshot (through android studio)

Edit:

I am happy to report that after a recent update for the emulator released by the developers, the issue no longer exists for me! The screenshot button has now started working again.

So if someone has the issue, I believe it can now be fixed by just updating your emulator to the latest version available.

This is the PR where they added that feat. They didn't provide a better UI due to some performance constraints. But you can find what hooks those indexes correspond to if you go to the components tab in dev tools and inspect said component; in the hooks section, you'll have a tree of the called hooks, and for each hook, a small number at the left which is the index. You'll probably need to unfold the tree of hooks to find them.

Here's a screenshot from the linked PR