Re: [Boost-users] Does shared_array's get()methodintroduceruntimeoverhead?

25 Oct 2006

      ...
-----Original Message-----
From: boost-users-bounces@lists.boost.org 
[mailto:boost-users-bounces@lists.boost.org] On Behalf Of Peng Yu
Sent: Tuesday, October 24, 2006 6:31 PM
To: boost-users@lists.boost.org
Subject: Re: [Boost-users] Does shared_array's 
get()methodintroduceruntimeoverhead?
On 10/24/06, Sohail Somani  wrote:
...
...
-----Original Message-----
From: boost-users-bounces@lists.boost.org
[mailto:boost-users-bounces@lists.boost.org] On Behalf Of Peng Yu
Sent: Tuesday, October 24, 2006 5:33 PM
Does it mean that code segment 1 is as fast as segment 2, but code
segment 3 is the slowest?
///////1
boost::shared_array a(new int[100]);
for(int i = 0; i < 100; ++ i)
 a.get()[i] = i;
////////2
boost::shared_array a(new int[100]);
int a_ptr = a.get();
for(int i = 0; i < 100; ++ i)
 a_ptr[i] = i;
////////3
boost::shared_array a(new int[100]);
for(int i = 0; i < 100; ++ i)
 a[i] = i;
All I can say is profile, profile, profile!
PS: Profile.
PPS: I'd guesstimate with a proper C++ compiler + optimizer, you
wouldn't notice much of a difference between 1 and 3.
Could you recommend a profiler on linux x86?
I recall using something called callgrind. There is a gui called
kcachegrind. I hope that should give you enough search keywords. Its
been a while since I profiled on linux.

You also always have the gcc option -pg which I've used quite often.