model-writtenrun 1209 - may be wrong, links go to the source messages. Commit-by-commit LLM analysis is in the Commits dialog.
- ChangAo Chen proposed reducing the memory allocated for each CatCTup in CatalogCacheCreateEntry() and attached a patch in #1.
- The proposal used MAXALIGN(sizeof(CatCTup)) for both the allocation size and tuple offset, relying on palloc() already returning an aligned pointer #1.
- Tom Lane committed that optimization as e5a77d876d, explaining that it saves eight bytes per CatCTup on 64-bit machines.
- The committed version also replaced context switching around the single allocation with MemoryContextAlloc(), a local allocation change beyond the submitted patch #1 e5a77d876d.
- Tom Lane closed the thread by acknowledging the finding and confirming the patch was pushed, leaving no outstanding work identified in the discussion #2.