RFR: 8316679: C2 SuperWord: wrong result, load should not be moved before store if not comparable

[Tobias Hartmann]

On Thu, 21 Sep 2023 14:19:58 GMT, Emanuel Peter <epeter at openjdk.org> wrote:

> **Context**
> In `SuperWord::output`, we replace the scalars with vectors, pack by pack.
> 
> In [JDK-8052081](https://bugs.openjdk.org/browse/JDK-8052081) ([changeset](https://github.com/openjdk/jdk/commit/7764490363a5963212e040baa2b79946cf4487bf)) we added an optimization to move the memory state of a LoadVector as early as possible. We go up the memory chain, and skip any `StoreVector` that provably can never alias with  the `LoadVector` (provably no overlap).
> 
> **Details**
> 
> For this, we must understand the following function `VPointer::cmp`:
> 
> https://github.com/openjdk/jdk/blob/96781ba33d4717c8df2b9ba4cd5d66858cf5c2d1/src/hotspot/share/opto/vectorization.hpp#L110-L120
> 
> We have 4 possible states: `NotComparable` means that we cannot determine the relative offsets between the two memory addresses. We have no guarantee about aliasing at all. `Less` means that one address lies before the other, `Greater` means that it comes after. `Equal` means that the two memory regions have an overlap or are identical. To determine overlap, we must know the position of the pointer, as well as the `memory_size` of each memory reference. `VPointer::not_equal` is true iff we get states `Less` or `Greater` (we know there cannot be overlap). `VPointer::comparable` is true iff the state is `NotComparable`, hence we have no guarantee about aliasing.
> 
> **Problem**
> 
> Currently, there are 2 bugs in this "move-load-before-store" logic:
> 1. [JDK-8316679](https://bugs.openjdk.org/browse/JDK-8316679) ([PR](https://github.com/openjdk/jdk/pull/15864)): We do not only skip the `not_equal` but also the `!comparable` case. However, when the two pointers are not comparable, that could be due to different invariants for example (see `TestMovingLoadBeforeStore.java`). The two pointers are not comparable, because we do not know the relative position, due to the invariants. But this exactly means that we cannot know at compile-time that there will not be aliasing. Thus, we cannot allow moving a load before such a store. Otherwise we may get wrong results. (*)
> 2. [JDK-8316594](https://bugs.openjdk.org/browse/JDK-8316594) ([PR](https://github.com/openjdk/jdk/pull/15866)): Currently, we only compare the first element of the load-pack with the `StoreVector`. This is not correct, we need to compare all elements of the load-pack with the `StoreVector`. Otherwise it may be that the first element lays below/before the memory region of the `StoreVector` (Less), but the last element of the load-pack for example c...

Thanks for the nice summary. The fix looks good to me.

test/hotspot/jtreg/compiler/loopopts/superword/TestMovingLoadBeforeStore.java line 29:

> 27:  * @requires vm.compiler2.enabled
> 28:  * @bug 8316679
> 29:  * @summary In SuperWord::output, LoadVector can be moved before StoreVector, but only if it is proven to be safe.

The test should have `@key randomness`

test/hotspot/jtreg/compiler/loopopts/superword/TestMovingLoadBeforeStore.java line 96:

> 94:             b[inv + i +  1]++;
> 95:             b[inv + i +  2]++;
> 96:             b[inv + i +  3]++;

You might want to remove the extra whitespaces in the index expressions.

-------------

Marked as reviewed by thartmann (Reviewer).

PR Review: https://git.openjdk.org/jdk/pull/15864#pullrequestreview-1654395398
PR Review Comment: https://git.openjdk.org/jdk/pull/15864#discussion_r1343531899
PR Review Comment: https://git.openjdk.org/jdk/pull/15864#discussion_r1343532671