Vector vs. array V2.

here are results on my home computer, VS2008 (of which the SECURE_CTL flag works)

Code:

---

x = -802947072
a: 5.756
x = -802947072
v: 6.397
x = -802947072
d: 5.885

---

Code:

---

g++ -O2 -fomit-frame-pointer -funroll-loops

---

Code:

---

x = -802947072
a: 10.9
x = -802947072
v: 11.27
x = -802947072
d: 11.24

---

GCC 4.2.3, 64-bit Linux.

I tried it 9 times, 3 each with different orders. Exact code as in the original post except changing the order in main. Using VC++ 7.1 Release build. vector wins every time no matter what I try with main. High and low examples:

Code:

---

x = -802947072
a: 11.406
x = -802947072
v: 10.157
x = -802947072
d: 10.359

---

Code:

---

x = -802947072
d: 10.281
x = -802947072
v: 9.687
x = -802947072
a: 10.094

---

I ran this main and got the results below:

Code:

---

        {
                V v;
                TIME_IT(func, v);
        }
        {
                D d;
                TIME_IT(func, d);
        }
        {
                D d;
                TIME_IT(func, d);
        }
        {
                V v;
                TIME_IT(func, v);
        }
        {
                V v;
                TIME_IT(func, v);
        }
        {
                D d;
                TIME_IT(func, d);
        }

---

Code:

---

x = -802947072
v: 10.015
x = -802947072
d: 10.328
x = -802947072
d: 10.234
x = -802947072
v: 9.907
x = -802947072
v: 10.156
x = -802947072
d: 10.234

---

VC++ 8.0 with _SECURE_SCL defined to 0:

Code:

---

x = -802947072
d: 9.718
x = -802947072
v: 9.656
x = -802947072
v: 9.688
x = -802947072
d: 9.5
x = -802947072
d: 9.578
x = -802947072
v: 9.735

---

The moral: there's little to no difference in speed. Meanwhile, look how easy it is to write the class with vector, and how easy it is to write a class with incorrect copy semantics with new[]/delete[].

No idea why, but here are my results ( VS2008 )
Release Build with full optimization.

Code:

---

x = ...
a: 11.969
x = ...
v: 113.64
x = ...
d: 18.359

---

I don't understand why the results are so nutty.

Intel Centrino Duo 1.6ghz processor, I would think it would be much faster than that.

edit: that is with #define _SECURE_SCL 0

Where'd you put the #define?

Quote:

---

Originally Posted by Raigne

I don't understand why the results are so nutty.

---

Simple. It means you screwed up the test. Something like an extra zero on the upper bound of one of the loops would do it...

Quote:

---

Originally Posted by iMalc

Simple. It means you screwed up the test. Something like an extra zero on the upper bound of one of the loops would do it...

---

But it's the same loop that does the work for all variants of the classes (unless the code has been changed from what I posted, of course).

I think the problem has to do with the range checking (but how it can be 10x slower is a bit surprising).

--
Mats

I tried adding #define _SECURE_SCL 0 before all my #include directives too and that got the vector time down by a huge amount, but it was still about 50% higher than the others.
I wonder if there's a difference between VC++ retail vs VC++ express? I no longer have a professional version that I can test it with -- only the express version.

There shouldn't be.

My results was often that the dynamic array was faster, but sometimes also slower, so I guess the results vary. Usually, it's about 10% difference in speed anyway, so not so much overhead.

Quote:

---

Originally Posted by cpjust

I tried adding #define _SECURE_SCL 0 before all my #include directives too and that got the vector time down by a huge amount, but it was still about 50% higher than the others.
I wonder if there's a difference between VC++ retail vs VC++ express? I no longer have a professional version that I can test it with -- only the express version.

---

i couldn't get it working in VS2005. it worked in VS2008, could just be a bug or something.

Quote:

---

Originally Posted by bling

i couldn't get it working in VS2005. it worked in VS2008, could just be a bug or something.

---

Why didn't it work? Compile error or runtime error...?

For my 2005 tests I used an old project that I use for random testing. There might be some settings I've tweaked to get the output I got. You may have to turn off more microsoft extensions in the project properties.

Quote:

---

Originally Posted by cpjust

Why didn't it work? Compile error or runtime error...?

---

the _SECURE_CTL flag got ignored, so the vector results sucked.

Really? Did you try to step through with the debugger?