[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Re: Patch
Angus Leeming wrote:
>
> The attached tar file contains re-worked xdr.h and giop.h and a patch of all
> the other little changes that I've made. The patch is against xtl-1.3.pl.8.
Nice work! I'll integrate them now.
> Note 2
> =====
> All code, except examples/auto compiles with both DEC cxx and gcc-2.95.2.
> examples/auto triggers an internal compiler bug in gcc.
I'll try to isolate the problem for submission to g++ people.
> All code compiles without warnings, except for DEC cxx's declaration that long
> long is not standard. rpcgen also has warnings. See earlier post.
We probably shouldn't care a lot with the rpcgen example, as it was there
only to as a comparative benchmark.
> Note 3
> =====
> The "quick hack" in md.h was incorrect. I've replaced it with my own that does
> work!
Oops, that should teach me not to write code while packing for holidays! :-)
> Note 4
> =====
> I've created a new method in generic_format::reset_buffer()
>
> void reset_buffer() {
> buffer.rewind();
> output_start();
> }
>
> (and similarly in v_format)
>
> This is used in examples/bench/xtl and examples/bench/vxtl, replacing
> mb.rewind()
> with
> gf.reset_buffer()
> etc.
>
> I've done this because mb.rewind() loses the byte order and alignment
> information in GIOP_format.
It makes sense.
> Note 5
> =====
> I've modified alltests.h to make it a harder test for alignment.
Great.
> Note 6
> =====
> In the copyright notice at the start of each file, shouldn't the web address:
> jop@di.uminho.pt - http://gsd.di.uminho.pt/~jop
> be
> jop@di.uminho.pt - http://gsd.di.uminho.pt/jop
> ?
They're both valid and I prefer to advertise the first as it is the
one that will probably remain valid in the long run.
> Benchmark results
> ==============
> Note the times for raw!!!!
I see. Alignment faults are expensive! In Intel this should not happen.
--
Jose Orlando Pereira
* mailto:jop@di.uminho.pt * http://gsd.di.uminho.pt/~jop *
- References:
- Patch
- From: Angus Leeming <a.leeming@ic.ac.uk>