Re: Preprocessor list data stucture committed to CVS

From: vesa_karvonen (vesa_karvonen_at_[hidden])
Date: 2002-01-22 12:47:23

--- In boost_at_y..., "vesa_karvonen" <vesa_karvonen_at_h...> wrote:
> I'd expect that EDG based compilers may grind on the list test for 
> quite a few minutes. I'll test this when I get to the office.

On the 800MHz PIII I use at the office, it took some 20 minutes to 
compile all the preprocessor tests in the background using Comeau 
4.2.38. The list test took 13 minutes. Currently the list used has 5 
elements. I'll see what happens if I reduce the list length.

Some time ago I implemented an almost complete (it had a bug, but it 
worked for the particular list) quick sort using the preprocessor. It 
preprocessed in less than a second on both g++ and MSVC++. On Comeau 
I stopped the process after it had been compiling for over 3 days 
(over the weekend) and used over 200Mb of memory. Since the list was 
only 5 elements long, recursion depths must have been rather small 
(say less than 25). I think I'll probably implement a more robust 
version of BOOST_PP_LIST_SORT(P,L) and try it online with Comeau. :)