Vijay,
On Mon, 2010-05-10 at 06:17 -0400, Vijay Saraswat wrote:
> Russel Winder wrote:
> > On Mon, 2010-05-10 at 07:13 +0100, Russel Winder wrote:
[ . . . ]
> > My apologies to all on the X10 development team, I just spotted I didn't
> > have -O on when I thought I did. With -O, the X10 code seems to be
> > faster than any other language!
> >
> Right, thanks, numbers would be helpful.
|> scons run_pi_cpp_sequential
/usr/bin/python
/home/Checkouts/Bazaar/SCons/Working/bootstrap/src/script/scons.py
run_pi_cpp_sequential
scons: Reading SConscript files ...
scons: done reading SConscript files.
scons: Building targets ...
g++ -o pi_cpp_sequential.o -c -O3 -Wall
-I/home/users/russel/Progs/Applications/Timing pi_cpp_sequential.cpp
gcc -o /home/users/russel/Progs/Applications/Timing/microsecondTime.o
-c -O3 -Wall /home/users/russel/Progs/Applications/Timing/microsecondTime.c
g++ -o pi_cpp_sequential pi_cpp_sequential.o
/home/users/russel/Progs/Applications/Timing/microsecondTime.o
./pi_cpp_sequential
==== C++ Sequential pi = 3.14159265358979194159017
==== C++ Sequential iteration count = 1000000000
==== C++ Sequential elapse = 9.718674000000000035015546
scons: done building targets.
508 anglides:~/Progs/Applications/Pi_Quadrature
|> scons run_Pi_X10_Sequential_Cpp
/usr/bin/python
/home/Checkouts/Bazaar/SCons/Working/bootstrap/src/script/scons.py
run_Pi_X10_Sequential_Cpp
scons: Reading SConscript files ...
scons: done reading SConscript files.
scons: Building targets ...
x10c++ -O -o Pi_X10_Sequential.x10_cpp Pi_X10_Sequential.x10
runx10 Pi_X10_Sequential.x10_cpp
==== X10 Sequential pi = 3.141592653589971
==== X10 Sequential iteration count = 1000000000
==== X10 Sequential elapse = 8.609898401000001
scons: done building targets.
509 anglides:~/Progs/Applications/Pi_Quadrature
|>
I think the issue here is that I am using long double in the C++ code so
the above is not actually a fair comparison. Making the change:
|> scons run_pi_cpp_sequential
/usr/bin/python
/home/Checkouts/Bazaar/SCons/Working/bootstrap/src/script/scons.py
run_pi_cpp_sequential
scons: Reading SConscript files ...
scons: done reading SConscript files.
scons: Building targets ...
g++ -o pi_cpp_sequential.o -c -O3 -Wall
-I/home/users/russel/Progs/Applications/Timing pi_cpp_sequential.cpp
g++ -o pi_cpp_sequential pi_cpp_sequential.o
/home/users/russel/Progs/Applications/Timing/microsecondTime.o
./pi_cpp_sequential
==== C++ Sequential pi = 3.141592653589970751681903
==== C++ Sequential iteration count = 1000000000
==== C++ Sequential elapse = 8.595724999999999837996256
scons: done building targets.
513 anglides:~/Progs/Applications/Pi_Quadrature
|>
OK, that appears better. Only a single data point of course so just an
indication, but it shows X10 and the C++ back end are doing the right
thing for this trivially small tight loop.
> In case you haven't seen this yet, the relevant page is
>
> http://x10.codehaus.org/Performance+Tuning+an+X10+Application
I am not trying to get into the performance tuning and benchmarking game
(though if someone pays me to do so I am happy to start :-)
--
Russel.
=============================================================================
Dr Russel Winder t: +44 20 7585 2200 voip: sip:russel.win...@ekiga.net
41 Buckmaster Road m: +44 7770 465 077 xmpp: rus...@russel.org.uk
London SW11 1EN, UK w: www.russel.org.uk skype: russel_winder
------------------------------------------------------------------------------
_______________________________________________
X10-users mailing list
X10-users@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/x10-users
