This is the mail archive of the gcc-patches@gcc.gnu.org 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: [trans-mem] PR 46567


On 01/21/2011 12:59 PM, Richard Henderson wrote:

You wouldn't need the IOR if we make the parameter visited_blocks, and if it's null allocate the thing. That should work since TM regions do not overlap. Although perhaps that makes the interface a bit confusing...


Thats exactly the way I had it originally, and decided this was cleaner unless it became a performance issue. I wasn't sure that someone might not be making a call in someday on a sub region or something else for some reason and not get their expected result. it sounded difficult to find. I dont think IORs are very expensive.
     bitmap_obstack_initialize (&tm_obstack);
+   bb_in_TM_region = BITMAP_ALLOC (NULL);
BITMAP_ALLOC (&tm_obstack).

and then you don't need the free at the end, since we junk the entire
obstack all at once.


/me is always suspicious of obstacks, but OK :-)


Andrew


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