which one runs faster ??

[onkar]

which one runs faster ??

for(i=0;i<100;i++)
for(j=0;j<10;j++)
a[j]=0;

OR

for(j=0;j<10;j++)
for(i=0;i<100;i++)
a[j]=0;

 

[Joachim Schmitz]

which one runs faster ??

for(i=0;i<100;i++)
for(j=0;j<10;j++)
a[j]=0;

OR

for(j=0;j<10;j++)
for(i=0;i<100;i++)
a[j]=0;

time it and you'll know, for your platform, compiler and optimazation
options

Bye, Jojo

 

[Tor Rustad]

which one runs faster ??

for(i=0;i<100;i++)
for(j=0;j<10;j++)
a[j]=0;

OR

for(j=0;j<10;j++)
for(i=0;i<100;i++)
a[j]=0;

Why do you expect there is a difference? The inner loop looks identical.

 

[keith]

which one runs faster ??

for(i=0;i<100;i++)
for(j=0;j<10;j++)
a[j]=0;

OR

for(j=0;j<10;j++)
for(i=0;i<100;i++)
a[j]=0;

Short Answer: It Depends.

Less Short Answer: It depends on a bunch of things, none of which is
directly related to standard C.

Pragmatic Answer: Try both and see.

What I'd Probably Do<tm>: memset(a, 0x00, sizeof a[0][0] * 1000);

Thought Experiment: What would each of your methods do if every array
element happened to be around half a memory page in size? Assuming
you are on a system with virtual memory, of course.

 

[mark_bluemel]

which one runs faster ??
Yes.

for(i=0;i<100;i++)
for(j=0;j<10;j++)
a[j]=0;

OR

for(j=0;j<10;j++)
for(i=0;i<100;i++)
a[j]=0;

1. You haven't told me what a[rows][columns] is or what the cache line
size on your machine. If a is a 2-D array of bytes, and the cache line
was > 1000 bytes, then it probably won't matter.

2. This is not really about C as such. However, as C is a Row-Major
language (see http://en.wikipedia.org/wiki/Row-major_order), it's
normally taken to be better to vary the column index faster than the
row index, or more generally to order index variation from the right.

 

[jacob navia]

which one runs faster ??

for(i=0;i<100;i++)
for(j=0;j<10;j++)
a[j]=0;

OR

for(j=0;j<10;j++)
for(i=0;i<100;i++)
a[j]=0;

memset(a,0,1000*sizeof(a[0]));

That will be optimized in assembly by the
system provider, and it will be faster than
your loops.

 

[Eric Sosman]

which one runs faster ??

for(i=0;i<100;i++)
for(j=0;j<10;j++)
a[j]=0;

OR

for(j=0;j<10;j++)
for(i=0;i<100;i++)
a[j]=0;

memset(a,0,1000*sizeof(a[0]));

That will be optimized in assembly by the
system provider, and it will be faster than
your loops.

It will also be wrong. R-O-N-G, wrong. (Hint: On the
assumption that the original loops do not invoke undefined
behavior, is sizeof a[0] == sizeof a[0][0] possible?)

Even after the obvious repair it could still be wrong.
No, I'm not talking about exotic machines where all-bits-zero
is not "zero" for the type in question. (Hint: which elements
of `unsigned char a[120][150]' are affected by the loops and
which are affected by memset?)

 

[Robbie Hatley]

which one runs faster ??

for(i=0;i<100;i++)
for(j=0;j<10;j++)
a[j]=0;

OR

for(j=0;j<10;j++)
for(i=0;i<100;i++)
a[j]=0;

(The following assumes that "a" has been allocated as a
single contiguous block of memory, and that the dimensions
of a are [100][10]. You didn't state either. If these
assumptions are not true, the following does not hold.)

Your first set of nested loops might run faster, because it
involves a single traverse through the memory, one byte
increments in the same direction. The second jumps back
and forth.

A compiler might translate the first to machine laguage as
"rep stosl" (writing all the bytes in a single rapid-fire
burst).

The second, however, might get tranlated to nested loops,
which would be slower.

--
Cheers,
Robbie Hatley
East Tustin, CA, USA
lonewolf aatt well dott com
triple-dubya dott tustinfreezone dott org

 

[Guru Jois]

which one runs faster ??

for(i=0;i<100;i++)
for(j=0;j<10;j++)
a[j]=0;

OR

for(j=0;j<10;j++)
for(i=0;i<100;i++)
a[j]=0;

check it yourself using clock()
help - use man page for clock

Bye
Guru Jois

 

[Tak]

which one runs faster ??

for(i=0;i<100;i++)
for(j=0;j<10;j++)
a[j]=0;

OR

for(j=0;j<10;j++)
for(i=0;i<100;i++)
a[j]=0;

I thought them run in the same time, haha , I am new at c/c++;

 

[Richard Heathfield]

onkar said:

which one runs faster ??

T a[100][10] = {0};

 

[Tor Rustad]

which one runs faster ??
Yes.
for(i=0;i<100;i++)
for(j=0;j<10;j++)
a[j]=0;

OR

for(j=0;j<10;j++)
for(i=0;i<100;i++)
a[j]=0;

1. You haven't told me what a[rows][columns] is or what the cache line
size on your machine. If a is a 2-D array of bytes, and the cache line
was > 1000 bytes, then it probably won't matter.

For tuning, yes cache issues are important. There is a technique called
*loop blocking*, which address this. For example, consider a more
interesting loop:

for (i=0; i<MAX; i++)
for (j=0; j<MAX; j++)
A[j] = A[j] * B[j];


Now let us rewrite the loop, and select 'block_size' such that the A and
B memory chunks fit a cache line:

for (i=0; i<MAX; i+=block_size)
for (j=0; j<MAX; j+=block_size)
for (k=i; k<i+block_size; k++)
for (l=j; l<j+block_size; l++)
A[k][l] = A[k][l] * B[l][k];

This sort of cache optimization, is what a optimizing C compiler might
do behind the scene.

2. This is not really about C as such. However, as C is a Row-Major
language (see http://en.wikipedia.org/wiki/Row-major_order), it's
normally taken to be better to vary the column index faster than the
row index, or more generally to order index variation from the right.

IF you want to make the job easier for the C optimizer, yes then the
inner index should be over 'j'. However, for OP example, I be very
surprised if a modern optimizing C compiler really care which order the
programmer arrange 'i' and 'j'.

More important is the *readability* of the code, since the "standard"
way to code such a loop is:

for (i=0; i<ROW_MAX; i++)
for (j=0; j<COL_MAX; j++)

I like to have a very good reason before doing it the other way around.

 

[Tor Rustad]

onkar said:

which one runs faster ??

T a[100][10] = {0};

Chapter and verse please.

 

[pete]

onkar said:

which one runs faster ??

T a[100][10] = {0};

Chapter and verse please.

It doesn't matter which one of those two, runs faster.
They're not the same.
This one is a declaration.
The other one was a statement.

 

[user923005]

Chapter and verse please.

Let's make it:
static T a[100][10] = {0};
Then we have:
"5.1.2 Execution environments
1 Two execution environments are defined: freestanding and hosted. In
both cases, program startup occurs when a designated C function is
called by the execution environment. All objects with static storage
duration shall be initialized (set to their initial values) before
program startup. The manner and timing of such initialization are
otherwise unspecified. Program termination returns control to the
execution environment."

Giving a time of zero seconds.

 

[Keith Thompson]

which one runs faster ??

for(i=0;i<100;i++)
for(j=0;j<10;j++)
a[j]=0;

OR

for(j=0;j<10;j++)
for(i=0;i<100;i++)
a[j]=0;

check it yourself using clock()
help - use man page for clock

Better yet, use a profiler if your system provides one. (Details of
profiler use are off-topic.)

 

[christian.bau]

which one runs faster ??

for(i=0;i<100;i++)
for(j=0;j<10;j++)
a[j]=0;

OR

for(j=0;j<10;j++)
for(i=0;i<100;i++)
a[j]=0;

If you want to actually learn something: Write a program that executes
this code and measure the execution time.
Then vary the numbers (100 and 10). Study what happens. Check if
anything unusual happens. If anything unusual happens, write down
exactly what happens, post a complete program, and someone will
explain to you _why_ it happens.

The code that you posted is useless, because we don't actually know
what it is doing. Is a an array of arrays of int, or is it an array of
pointers to int, or is it a pointer to arrays of ints, or a pointer to
an array of pointers to int? Huge difference.

 

[Spiro Trikaliotis]

which one runs faster ??

for(i=0;i<100;i++)
for(j=0;j<10;j++)
a[j]=0;

OR

for(j=0;j<10;j++)
for(i=0;i<100;i++)
a[j]=0;

I thought them run in the same time, haha , I am new at c/c++;

leaving the topic of clc (that is: speaking OT), in practice, both can
differ significantly, but they do not need to. If the compiler
translates these "literally", you might even end up thrashing your
virtual memory (making this statement even more OT), resulting in
factors of 1000 or more between both statements.

Of course, for this to happen, it is likely that the values 100 and 10
must be much bigger than in the "samples" above.

Ok, now I'll shut my OT mouth.

Regards,
Spiro.