RM57L843: The RM57L843 Machine instructions require more cycles to be executed

Hello,

I am trying to initialize an application on RM57L843 without HALCoGen, in assembly language.
After the initialization in assembly language the application run more slowly than after the initialization with HCG HALCoGen.

The clocks seems to be good and when I trace a loop test I can see that the same instructions require more cycles to be executed after the initialization in assembly language.

With HCG the loop require 32 cycles:

Disassembly:string                         Cycl        D.Cycl

ldr       r1, [r10, #0x10]                      0             2

sub       r1, r1, r0                               2             1

add       r3, r3, #1                              3             1

cmp       r2, r1                                   4              0

bpl       #0x7d74                                4             28

ldr       r1, [r10, #0x10]                      32           2

With the initialization in assembly language loop requires more than 60 cycles:

Disassembly:string                         Cycl        D.Cycl

ldr       r1, [r10, #0x10]                      0             5

sub       r1, r1, r0                               5             1

add      r4, r4, #1                               6             1

cmp       r2, r1                                   7             0

bpl       #0x1800cc                            7             53

ldr       r1, [r10, #0x10]                     60           5

What can be the cause of these differences?
Thanks for your help.

Jerome