This is the mail archive of the mailing list for the GCC project.

Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]
Other format: [Raw text]

Re: [PATCH] New Optimization: Partitioning hot & cold basic blocks

On Friday, October 10, 2003, at 05:11 PM, Richard Henderson wrote:
Perhaps I'm not looking hard enough, but how do you handle branches between the hot and cold sections?

Full width pointer indirect jumps as I recall. Should get us anywhere we need to go.

Most targets don't have and conditional branche that can span arbitrary distances, directly or indirectly. Similarly, direct branches may need to be turned into indirect branches in order to reach.

ppc is infamous for this, but you already knew that.

I'd also like to see changes to dwarf2 output so that this split is
properly represented in the debug info.  Given that this feature is
not enabled except via explicit switch, this isn't imperitive, but
it should be done once this patch is accepted.

Agreed. We are interested in increasing gcc's ability to debug optimized code, and this would probably be one aspect of that work. We'd like to hand wave the debugging issues for now, as we know that the transformation may scramble debugging information. I suspect we may have different people doing that work, Hi Devang.

Once we get it in, it'd be great if some that just `knows' the bugs could file a PR on them to help guide and ensure that all of them do get fixed. I know we've talked about this in the past on the list, so maybe just including a pointer to the main archive were we talked about it would be enough.

Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]