PATCH to remove unnecessary NOTE_INSN_DELETED

Richard Kenner kenner@vlsi1.ultra.nyu.edu
Tue Nov 30 23:59:00 GMT 1999


    Renumbering insns is something we've discussed before, but we hadn't
    done it because it starts to become a pain in the ass to debug the
    optimizers -- tracking a particular insn as it moves through the
    compiler becomes much harder.

I *strongly* agree!

    At minimum we should emit translation tables into the debugging dumps
    so that we can match things up; but perhaps we shouldn't do this
    compaction at all and instead get passes to work off LUIDs.

I don't think the translation tables will be of much help.  Quite often,
you don't even *want* do print the debugging dumps (e.g., in large
programs), but just set conditional breakpoint to see when a suspicious
insn got generated or optimized.  Doing this sort of renumbering will
make it *much* harder to debug the compiler.



More information about the Gcc-patches mailing list