Dear all,

I am trying to setup the 32-bit timer to measure the execution time of
the program. The setting is according to the "Initialization code_hints
for the LPC2000 family" and the manual. However, the timer gives a
erratic result >_<

I setup a simple measurement as follow
tStart = T0TC;
for (i=0;i tEnd = T0TC;
t = tEnd - tStart;

I expect the for execution time to increase somewhat proportionally
with n, but the result is
n=3 -> t= 58
n=4 -> t= 79 (+21)
n=5 -> t= 78 (-1 ???)
n=6 -> t= 101(+23)
n=7 -> t= 120(+19)
n=8 -> t= 147(+27)
n=9 -> t= 130(-17 ???)
n-> t= 161(+31)

Have anyone tried to measure the execution time on LPC2xxx? Please give
me some advices

Thank you very much,
Richard

PS:
This is the chip initialization code as taken from the "Initialization
code_hints for the LPC2000 family" and the manual.

//set CCLK = 60Mhz
PLLCFG=0x25;
feed();
PLLCON=0x1;// Enabling the PLL
feed();
while(!(PLLSTAT & PLOCK)){}// Wait for the PLL to lock to set frequency

// Connect the PLL as the clock source
PLLCON=0x3;
feed();

//set PCLK = CCLK
VPBDIV = 0x1;

//Reset timer
T0TCR = (unsigned long)0x2; //reset timer
T0CTCR = (unsigned long)0x00000000; //set control reg to timer mode,
increase every rising PCLK edge
T0PR = (unsigned long)0x00000000;//Timer0 prescale = 0
T0TCR = (unsigned long)0x00000001; //enable timer 01

Start a New Thread

Hi,

what is the width of t? If it is not 32 bits wide and unsigned, then you
need to handle timer overflows.

Foltos

bachkhoa85 wrote:
> Dear all,
>
> I am trying to setup the 32-bit timer to measure the execution time of
> the program. The setting is according to the "Initialization code_hints
> for the LPC2000 family" and the manual. However, the timer gives a
> erratic result >_<
>
> I setup a simple measurement as follow
> tStart = T0TC;
> for (i=0;i > tEnd = T0TC;
> t = tEnd - tStart;
>
> I expect the for execution time to increase somewhat proportionally
> with n, but the result is
> n=3 -> t= 58
> n=4 -> t= 79 (+21)
> n=5 -> t= 78 (-1 ???)
> n=6 -> t= 101(+23)
> n=7 -> t= 120(+19)
> n=8 -> t= 147(+27)
> n=9 -> t= 130(-17 ???)
> n-> t= 161(+31)
>
> Have anyone tried to measure the execution time on LPC2xxx? Please give
> me some advices
>
> Thank you very much,
> Richard
>
> PS:
> This is the chip initialization code as taken from the "Initialization
> code_hints for the LPC2000 family" and the manual.
>
> //set CCLK = 60Mhz
> PLLCFG=0x25;
> feed();
> PLLCON=0x1;// Enabling the PLL
> feed();
> while(!(PLLSTAT & PLOCK)){}// Wait for the PLL to lock to set frequency
>
> // Connect the PLL as the clock source
> PLLCON=0x3;
> feed();
>
> //set PCLK = CCLK
> VPBDIV = 0x1;
>
> //Reset timer
> T0TCR = (unsigned long)0x2; //reset timer
> T0CTCR = (unsigned long)0x00000000; //set control reg to timer mode,
> increase every rising PCLK edge
> T0PR = (unsigned long)0x00000000;//Timer0 prescale = 0
> T0TCR = (unsigned long)0x00000001; //enable timer 01
>
> Start a New Thread

On Fri, 09 May 2008 12:09:33 +0200, you wrote:

>Hi,
>
>what is the width of t? If it is not 32 bits wide and unsigned, then you
>need to handle timer overflows.
>
>Foltos
>
>bachkhoa85 wrote:
>> Dear all,
>>
>> I am trying to setup the 32-bit timer to measure the execution time of
>> the program. The setting is according to the "Initialization code_hints
>> for the LPC2000 family" and the manual. However, the timer gives a
>> erratic result >_<
>>
>> I setup a simple measurement as follow
>> tStart = T0TC;
>> for (i=0;i >> tEnd = T0TC;
>> t = tEnd - tStart;
>>
>> I expect the for execution time to increase somewhat proportionally
>> with n, but the result is
>> n=3 -> t= 58
>> n=4 -> t= 79 (+21)
>> n=5 -> t= 78 (-1 ???)
>> n=6 -> t= 101(+23)
>> n=7 -> t= 120(+19)
>> n=8 -> t= 147(+27)
>> n=9 -> t= 130(-17 ???)
>> n-> t= 161(+31)

This method should work.
is n a constant or a variable?
If the former, it's possible that the compiler may be generating different code - look at the
assembler output, or check the interval on a scope.
The MAM may also be influencing things.
>> Have anyone tried to measure the execution time on LPC2xxx? Please give
>> me some advices
>>
>> Thank you very much,
>> Richard
>>
>> PS:
>> This is the chip initialization code as taken from the "Initialization
>> code_hints for the LPC2000 family" and the manual.
>>
>> //set CCLK = 60Mhz
>> PLLCFG=0x25;
>> feed();
>> PLLCON=0x1;// Enabling the PLL
>> feed();
>> while(!(PLLSTAT & PLOCK)){}// Wait for the PLL to lock to set frequency
>>
>> // Connect the PLL as the clock source
>> PLLCON=0x3;
>> feed();
>>
>> //set PCLK = CCLK
>> VPBDIV = 0x1;
>>
>> //Reset timer
>> T0TCR = (unsigned long)0x2; //reset timer
>> T0CTCR = (unsigned long)0x00000000; //set control reg to timer mode,
>> increase every rising PCLK edge
>> T0PR = (unsigned long)0x00000000;//Timer0 prescale = 0
>> T0TCR = (unsigned long)0x00000001; //enable timer 01
>>
>> Start a New Thread

Thank a lot for reply, I have checked everything

The method works well for MAM mode 0. When enable the MAM, the
problem happens.

I tried to measure time for each individual loop like this

Tstart = T0CT;
for (i=0;i Tend = T0CT;
t[i] = Tend - Tstart;
Tstart = T0CT;
}

And I tried to measure them combined
Tstart = T0CT;
for (i=0;i Tend = T0CT;
Total = Tend - Tstart;

I find that Total >> sumAll(t[i]).

In example
Mode 1: n = 10 --> Total = 200, sumAll(t[i]) = 172

Can anyone please give me some suggests why there is such a
difference?

Thank you very much
--- In l..., Mike Harrison wrote:
>
> On Fri, 09 May 2008 12:09:33 +0200, you wrote:
>
> >Hi,
> >
> >what is the width of t? If it is not 32 bits wide and unsigned,
then you
> >need to handle timer overflows.
> >
> >Foltos
> >
> >bachkhoa85 wrote:
> >> Dear all,
> >>
> >> I am trying to setup the 32-bit timer to measure the execution
time of
> >> the program. The setting is according to the "Initialization
code_hints
> >> for the LPC2000 family" and the manual. However, the timer gives
a
> >> erratic result >_<
> >>
> >> I setup a simple measurement as follow
> >> tStart = T0TC;
> >> for (i=0;i > >> tEnd = T0TC;
> >> t = tEnd - tStart;
> >>
> >> I expect the for execution time to increase somewhat
proportionally
> >> with n, but the result is
> >> n=3 -> t= 58
> >> n=4 -> t= 79 (+21)
> >> n=5 -> t= 78 (-1 ???)
> >> n=6 -> t= 101(+23)
> >> n=7 -> t= 120(+19)
> >> n=8 -> t= 147(+27)
> >> n=9 -> t= 130(-17 ???)
> >> n-> t= 161(+31)
>
> This method should work.
> is n a constant or a variable?
> If the former, it's possible that the compiler may be generating
different code - look at the
> assembler output, or check the interval on a scope.
> The MAM may also be influencing things.
> >> Have anyone tried to measure the execution time on LPC2xxx?
Please give
> >> me some advices
> >>
> >> Thank you very much,
> >> Richard
> >>
> >> PS:
> >> This is the chip initialization code as taken from
the "Initialization
> >> code_hints for the LPC2000 family" and the manual.
> >>
> >> //set CCLK = 60Mhz
> >> PLLCFG=0x25;
> >> feed();
> >> PLLCON=0x1;// Enabling the PLL
> >> feed();
> >> while(!(PLLSTAT & PLOCK)){}// Wait for the PLL to lock to set
frequency
> >>
> >> // Connect the PLL as the clock source
> >> PLLCON=0x3;
> >> feed();
> >>
> >> //set PCLK = CCLK
> >> VPBDIV = 0x1;
> >>
> >> //Reset timer
> >> T0TCR = (unsigned long)0x2; //reset timer
> >> T0CTCR = (unsigned long)0x00000000; //set control reg to timer
mode,
> >> increase every rising PCLK edge
> >> T0PR = (unsigned long)0x00000000;//Timer0 prescale = 0
> >> T0TCR = (unsigned long)0x00000001; //enable timer 01
> >>
> >>
> >>
> >>
> >>
> >> Start a New Thread