Hi Honza, > On 13 Jan 2026, at 7:58 pm, Jan Hubicka <[email protected]> wrote: > > External email: Use caution opening links or attachments > > >> Hi Andrew, >> >>> On 6 Jan 2026, at 7:10 pm, Andrew Pinski <[email protected]> >>> wrote: >>> >>> External email: Use caution opening links or attachments >>> >>> >>> On Sun, Jan 4, 2026 at 11:11 PM Kugan Vivekanandarajah >>> <[email protected]> wrote: >>>> >>>> Hi, >>>> >>>> [PR ipa/123383] shows in ICE speculative call sequence has speculative_id >>>> 256 out of range with LTO. >>>> This also shows up lot profile bootstrapping gcc. >>>> >>>> Fix by checking lto_stmt_uid in get_next_speculative_id as done in other >>>> places too. >>> >>> LGTM based on the previous similar fix (PR93318, r10-6074) where it >>> is mentioned that only one or the other will be set. So you need to >>> compare both. >>> It would be useful to add the (semi-big) testcase I added to the bug >>> report as a (semi) bigger LTO C++ example. >>> >> Thanks for the review. Attached patch adds the test case from the PR. >> Is this OK? > > The specualtive_id should be unique only within on speculative id block > which can be iterated by first_speculative_call_target and > next_speculative_call_target, so there is no need to walk the whole list > of edges >
Please see the attached to patch which changes this. Bootstrap and regression testing are ongoing. Is this OK if no regression? Thanks, Kugan
0001-Use-next_speculative_call_target-in-get_next_specula.patch
Description: 0001-Use-next_speculative_call_target-in-get_next_specula.patch
