Hi Pim, all, Thanks for the report and reproducer. I reproduced the failure on ARM64 against the parent of 46795 and confirmed the fix. The targeted test passes on ARM64 and x86_64.
Given the cross-VIP traffic misdirection, 46795 should be merged as a standalone correctness fix: https://gerrit.fd.io/r/c/vpp/+/46795 I also submitted 46796, replacing the SSE-specific lookup with portable vppinfra vectors. SSE and NEON now share the same implementation, while retaining the corrected scalar fallback: https://gerrit.fd.io/r/c/vpp/+/46796 ARM64 microbenchmarks show workload-dependent results, so 46796 is a refactoring rather than a general performance optimization. Details are on Gerrit. Best regards, Jerome
-=-=-=-=-=-=-=-=-=-=-=- Links: You receive all messages sent to this group. View/Reply Online (#27183): https://lists.fd.io/g/vpp-dev/message/27183 Mute This Topic: https://lists.fd.io/mt/121260774/21656 Group Owner: [email protected] Unsubscribe: https://lists.fd.io/g/vpp-dev/leave/14379924/21656/631435203/xyzzy [[email protected]] -=-=-=-=-=-=-=-=-=-=-=-
