U++ framework: U++ Developers corner

Home » Developing U++ » U++ Developers corner » Benchmarking U++

Show: Today's Messages :: Polls :: Message Navigator
E-mail to friend

Benchmarking U++ [message #7064]

Mon, 11 December 2006 01:00

mirek
Messages: 14291
Registered: November 2005

Ultimate Member

I have spend some time providing some comparison of U++ Core vs standard C++ library:

http://www.ultimatepp.org/www$uppweb2$vsstd$en-us.html

I would like if some STL expert would have checked my STL code if it is really the optimal solution. If this first pass is OK, I will try to submit the page to alt.lang.c++.moderated to further review (and then maybe we can make it an article).

Mirek

Report message to a moderator

Re: Benchmarking U++ [message #7099 is a reply to message #7064]

Wed, 13 December 2006 22:24

mirek
Messages: 14291
Registered: November 2005

Ultimate Member

Update:

I have tried commenting out parts of benchmark to get timings for separate facilities. I get (MSC71, in ms, 4KB input file, 120 map entries):

C++ standard library, std::map:
stream: 1813
string: 530
map: 1657
sort: 0 (std::map is sorted)

U++ Core:
stream: 578
string: 422
map: 640
sort: 360

Now really surprising is the stream difference. I always thought that reimplementing stream library for U++ is the least tenable wheel-reinventing aspect of the platform. Maybe I was wrong...

I think the problem is those stupid "ascii"/"binary" modes of Windows platform, which disallow really effective inlined gets. OTOH, that was one of primary reasons to reinvent the wheel.

Mirek

Report message to a moderator

Previous Topic:	Endian specific source changes, for Universal support.
Next Topic:	U++ as web platform

Goto Forum:

-=] Back to Top [=-

[ Syndicate this forum (XML) ] [

] [

]

Current Time: Fri Oct 09 22:38:46 GMT+2 2026

Total time taken to generate the page: 0.02515 seconds