On Thu, Aug 4, 2016 at 11:32 AM, Jan Hubicka <hubi...@ucw.cz> wrote:
>> Hi Richard,
>>
>> Thanks for the review.
>>
>> On 04/08/16 17:26, Richard Biener wrote:
>> >On Thu, Aug 4, 2016 at 6:12 AM, kugan <kugan.vivekanandara...@linaro.org>
>> >wrote:
>> >>Hi,
>> >>
>> >>During IPA-VRP implementation, I realized that we don't support streaming
>> >>wide_int in LTO. Attached patch does this. Tested with IPA-VRP. Is this OK
>> >>for trunk if bootstrap and regression testing is fine.
>> >
>> >Hmm, those functions belong to data-streamer-{in,out}.c and data-streamer.h
>> >and should be named streamer_write_wide_int / streamer_read_wide_int.
>> >
>> >Note that we already have (non-exported) streamer_write_wi /
>> >streamer_read_wi
>> >which operate on widest_ints. Those also reside in lto-streamer-{in,out}.c
>> >and
>> >should be moved to data-streamer.h (and be renamed to
>> >streamer_write_widest_int).
>>
>> I have now streamer_write_wide_int and streamer_write_widest_int.
>> Similarly for reading. There is lot of similarity. I am not very
>> familiar with wide_int so kept it that way. Is this OK now?
>>
>> Thanks,
>> Kugan
>>
>> gcc/ChangeLog:
>>
>> 2016-08-04 Kugan Vivekanandarajah <kug...@linaro.org>
>>
>> * data-streamer-in.c (streamer_read_wide_int): New.
>> (streamer_read_widest_int): Renamed function.
>> * data-streamer-out.c (streamer_write_wide_int): New
>> (streamer_write_widest_int): Renamed function.
>
> I wondered, given we do C++ now, if we don't want to just have
> stream_in/stream_out member functions for our classes and/or use just one
> function name for all of them so one does not need to look up somewhat
> irregular function names.
>
> I find LTO streaming API very hard to memorize and use without constantly
> looking up existing code.
Not sure if that would help given the arguments are different even besides the
thing you want to stream.
Richard.
> Honza
