Copying Aldy for the jump threading expertise.
This change can cause improvements much earlier in the optimization pipeline, especially when it comes to PHI nodes. Many of those threading tests are finicky about the IL coming in. If we do a better job optimizing earlier, an opportunity that use to exist may not any more, so it may not be surprising that we only have 17 now instead of 18.
That said, maybe he can have a look to make sure its not something too out of whack and adjust the testcase if so. I seem to recall a nuber of them were adjusted to allow a range of threading opportunities instead of a specific number.
Andrew
On 8/25/23 13:31, Adhemerval Zanella Netto wrote:Copying
Hi Andrew,
Your patch caused a regression [1] on aarch64-linux-gnu. Would you please investigate? I did a quick analysis and it seems that the expected 18 for aarch64 is now 17:
$ grep "Jumps threaded" a-ssa-dom-thread-7.c.197t.thread2 Jumps threaded: 17
Let me know if you need any assistance in reproducing these.
Thanks!
[1] https://ci.linaro.org/job/tcwg_gnu_native_check_gcc--master-aarch64-build/56...