On Fri, 20 Mar 2026 11:55:11 GMT, Frederic Parain <[email protected]> wrote:
> The main change in this patch is the use of the C++ acmp_map of the super
> class to generate the acmp_map of a sub-class instead of using the Java heap
> allocated int array. There's another fix in the computation of the shift
> applied to the payload to satisfy atomicity alignment constraints. A few
> asserts have been added to improve the robustness of the code.
>
> Testing in progress tiers 1 to 4.
>
> Thank you,
>
> Fred
LGTM!
Nit:
src/hotspot/share/classfile/fieldLayoutBuilder.cpp line 1274:
> 1272: _layout->shift_fields(shift);
> 1273: assert(_layout->first_field_block()->offset() %
> required_alignment == 0, "Fields should have been shifted to respect the
> required alignment");
> 1274: _payload_offset = _layout->first_field_block()->offset();
Suggestion:
_payload_offset = _layout->first_field_block()->offset();
assert(is_aligned(_payload_offset, required_alignment), "Fields should
have been shifted to respect the required alignment");
and the associated `#include "utilities/align.hpp"`
-------------
Marked as reviewed by iwalulya (no project role).
PR Review:
https://git.openjdk.org/valhalla/pull/2248#pullrequestreview-3990017127
PR Review Comment:
https://git.openjdk.org/valhalla/pull/2248#discussion_r2973510322
