Currently we enable it only for x86 and x64, but anecdotally, newer ARM chips (Cortex A8 and later) allow unaligned access. Flipping this compile flag would generate substantially better code in some cases.
Assignee: nobody → stejohns
Status: NEW → ASSIGNED
Priority: -- → P2
Target Milestone: --- → flash10.1
Created attachment 426362 [details] [diff] [review] Patch This removes AVMPLUS_UNALIGNED_ACCESS in favor of AVMSYSTEM_UNALIGNED_ACCESS/VMCFG_UNALIGNED_ACCESS, which is handled much like little-endian detection. Note that this requires the patch from bug 542133. Note that AVMPLUS_UNALIGNED_ACCESS was unused outside of the VM proper (gasp!) so it was removed entirely!
Attachment #426362 - Flags: review?(lhansen)
Comment on attachment 426362 [details] [diff] [review] Patch Need to correct the syntax in avmfeatures.as: "... to load and assembly ..." should probably be "... to load and assemble ..."
Attachment #426362 - Flags: review?(lhansen) → review+
(In reply to comment #2) > "... to load and assembly ..." should probably be "... to load and assemble > ..." Nice catch, will do.
Status: ASSIGNED → RESOLVED
Last Resolved: 9 years ago
Resolution: --- → FIXED
Engineering work item. Marking verified fixed.
Status: RESOLVED → VERIFIED
You need to log in before you can comment on or make changes to this bug.