The results are in: I'm running Linux FC4 on x86_64 The RTL results are quite amazing with -O3 optimization 1000000 records added 2000000 records added That took just 320ms RML RTL -O3 1260ms 320ms -O2 3200 ms 780 ms w/o 21230ms 13760ms Maybe there is something hampering RML 64-bit performance, as my best RML result on a 3GHz box doesn't differ much from Calum's 1.7GHz laptop ! I would like to see a more complex test with insert, select and delete, e.g. something like the resorce reservation system in http://www.infosys.tuwien.ac.at/NewsCache/pstl.html#evaluation where you have insertion, iteration, lookup/select and deletion On 10/9/05, Arkadiy Vertleyb <vertleyb@hotmail.com> wrote:
"Jose" <jmalv04@gmail.com> wrote
===========================================
g++ -O6 arkadiy.cpp -o test2 -Irtl -I/usr/local/include/boost-1_33 -L/usr/local/lib rtl/identifiable.hpp: In member function 'unsigned int rel::identifiable::id() const': rtl/identifiable.hpp:22: error: cast from 'char*' to 'unsigned int' loses precision make: *** [test2] Error 1 ===========================================
This actually is just a warning in VC7, and never came up in MinGW, so I didn't pay much attention to it. But why wouldn't it be able to cast, is it a 32 or 64-bit system?
Could you modify your copy of identifiable.hpp, and replace:
return reinterpret_cast<unsigned int>(p.get());
with
return p.get() - (char*)0;
This may fix the problem (at least it removes the warning in vc71).
Regards, Arkadiy
PS: my results for RTL (MinGW with optimization) --3094ms. My PC has a 1.5 GHz processor.
_______________________________________________ Unsubscribe & other changes: http://lists.boost.org/mailman/listinfo.cgi/boost