On Wed, Dec 14, 2016 at 09:34:44AM +0100, Jakub Jelinek wrote: > Here is updated patch that does that. I had to change the combiner > in a couple of places, because combiner is very unhappy if new instructions > (with uids above the highest) are inserted into the insn stream, > e.g. FOR_EACH_LOG_LINK (link, temp_insn) for DEBUG_INSN temp_insn > will then crash. But given that DEBUG_INSNs never have non-NULL LOG_LINKS, > I think we can just avoid using LOG_LINKS or INSN_COST on DEBUG_INSNs > and allow new DEBUG_INSNs be inserted into the stream (but not regular > insns - normally combine just reuses INSN_UID of the insns it is replacing).
Could you do a gcc_checking_assert (INSN_UID (INSN) <= max_uid_known) in the LOG_LINKS and INSN_COST defines? It is hard to check if you caught every use. > @@ -2275,7 +2275,7 @@ cant_combine_insn_p (rtx_insn *insn) > /* If this isn't really an insn, we can't do anything. > This can occur when flow deletes an insn that it has merged into an > auto-increment address. */ > - if (! INSN_P (insn)) > + if (! NONDEBUG_INSN_P (insn)) > return 1; Please get rid of the extra space (and thanks for the other formatting fixes, little by little :-) ) Looks fine otherwise. Thanks, Segher
