Hi Richard, >> The testcase you add already works, so does the original one from the PR. >> So I must be missing something?
Thanks, you are right. I retested both the added test cases against an unmodified trunk, both are already optimized. The existing compute_final_value_from_loop_phi_latch handling appears to cover these cases. Thus the patch does not currently demonstrate any additional optimization. Can you please let me know whether we can just retain the tests and close the PR? Thanks, Naveen
