On Thu, 5 Mar 2026 07:54:16 GMT, Axel Boldt-Christmas <[email protected]> 
wrote:

> For `IS_DEST_UNINITIALIZED` it would never be valid for the 
> CardTableBarrierSet to do a pre-barrier. Currently `value_copy_in_heap` and 
> `value_store_null_in_heap` both iterates the oops and virtually calls into 
> the barriers sets pre barriers (of which all implementations are no-ops for 
> uninitialized values).
> 
> Remove this as part of the `CardTableBarrierSet` interface. After this is 
> integrated I will upstream the relevant changes of this patch to mainline.

Thanks for doing this change! I have a style thought below:

src/hotspot/share/gc/shared/cardTableBarrierSet.inline.hpp line 185:

> 183: 
> 184:     if (!HasDecorator<decorators, IS_DEST_UNINITIALIZED>::value) {
> 185:       // Pre-barriers...

I'm wondering if the code doesn't read nicer if the comments were above the 
condition:
Suggestion:

    // Pre-barriers...
    if (!HasDecorator<decorators, IS_DEST_UNINITIALIZED>::value) {


Then the two comments would be at the same "level" and it might be easier to 
see the structure of the code:

    // Pre-barriers...
    conditional_pre_barrier_stuff

    copy

    // Post-barriers...
    post_barrier_stuff

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

PR Review: 
https://git.openjdk.org/valhalla/pull/2200#pullrequestreview-3896204695
PR Review Comment: 
https://git.openjdk.org/valhalla/pull/2200#discussion_r2889761066

Reply via email to