Lpc2000

Hi!!
I wanted to replace my current timer function with simple "for' loop
to measure time interval.So i added in my code function
void delay (volatile unsigned int j )
{
   for (;j ;--j);

   }
To obtain how much time it takes to do single loop in debbuger i tried
to count cycles in disassembly window.
After that i turned on exectution time profiling in uvision3 to
compare my otcome but my result was differ from keils.
Here are result for j=2 taken from uvision3:
0.033us void delay (volatile unsigned int j )
         {
0.150us   for (;j ;--j);

0.067us }

According to this first rotate takes 0.1us and each another
0.050us(which gives only 3 cycles Im running lpc2148 with 60MHz).
But in disassembly window i got:

          4:   for (;j ;--j);
          5:
0.050us     0x00000366  E002      B         0x0000036E
0.100us     0x00000368  9800      LDR       R0,[SP,#0x0000]
0.033us     0x0000036A  3801      SUB       R0,#0x01
0.067us     0x0000036C  9000      STR       R0,[SP,#0x0000]
0.150us     0x0000036E  9800      LDR       R0,[SP,#0x0000]
0.050us     0x00000370  2800      CMP       R0,#0x00
0.117us     0x00000372  D1F9      BNE       0x00000368
          6:   }

Did i miss something and which result should i trust?

Regards!!