Posted on May 17, 2006 at 08:57Hi,
I did multiple tests and I do not understand what I notice.
First of all, my str710 have a frequency of 48 MHz.
First test
===========
I plug a frequency generatior on a GPIO0/bit10.
I bind an handler itxti() on that interrupt, call by XTI_IRQHandler()
The handler generate a pulse on each interrupt.
Then, I watch the result on a oscilloscope and I notice a latency between the clock and the pulse of 15 micro secondes.
__arm void XTI_IRQHandler(void)
{
register unsigned int b=0x01; /* bigger than u8 because of condition stop i<=128 */
register unsigned int i=0;
register unsigned char h = XTI->PRH&=XTI->MRH;
register unsigned char l = XTI->PRL&=XTI->MRL;
while ( (h || l) && b<=128) {
if (h & b ) {
(XTI_handler[i+0x8])(EICXTI_data[i+0x8+32]); // test_itxti is called here
XTI->PRH &= ~b;
h &= ~b;
}
if (l & b ) {
(XTI_handler[i])(EICXTI_data[i+32]);
XTI->PRL &= ~b;
l &= ~b;
}
b<<=1;
i++;
}
}
static void test_itxti(void*d) {
XTI_PendingBitClear(XTI_Line13);
//WRITEIO(RST_CODEC,0);
//GPIO_BitWrite(GPIO0, 7,0);
for (int i=0;i
WRITEIO(RST_CODEC,1);
}
Second test:
============
I did almost the same thing except I write on the IO in the function XTI_IRQHandler()
Then I notice a latency of 7 micro secondes.
__arm void XTI_IRQHandler(void)
{
register unsigned int b=0x01; /* bigger than u8 because of condition stop i<=128 */
register unsigned int i=0;
register unsigned char h = XTI->PRH&=XTI->MRH;
register unsigned char l = XTI->PRL&=XTI->MRL;
GPIO_BitWrite(GPIO0, 7,0); //<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<
while ( (h || l) && b<=128) {
if (h & b ) {
(XTI_handler[i+0x8])(EICXTI_data[i+0x8+32]); // test_itxti is called here
XTI->PRH &= ~b;
h &= ~b;
}
if (l & b ) {
(XTI_handler[i])(EICXTI_data[i+32]);
XTI->PRL &= ~b;
l &= ~b;
}
b<<=1;
i++;
}
}
Third test :
============
I generate the pulse directly in the IRQHandler(71x_vect.s79).
I notice a latency of 2 micro secondes
addr DCD 0xE000300C
IRQHandler
SUB LR,LR,#4 ; adj for proper CMX return
STMFD SP!,{R0-R12,LR} ; save the registers
MRS R9,SPSR ; save the SPSR
STMFD SP!,{R9}
; TEST IT
LDR R0, addr
LDRH R3,[R0]
mov r4,#0x80
bic R3,R3,R4
STRH R3,[R0]
; end
...
Finaly, I did this test :
IRQHandler
SUB LR,LR,#4 ; adj for proper CMX return
STMFD SP!,{R0-R12,LR} ; save the registers
MRS R9,SPSR ; save the SPSR
STMFD SP!,{R9}
; LUDO TEST IT
LDR R0, addr1
LDRH R3,[R0]
mov r4,#0x80
ORR R3,R3,R4
STRH R3,[R0]
LDR R0, addr2
LDRH R3,[R0]
BIC R3,R3,R4
STRH R3,[R0]
LDR R0, addr3
LDRH R3,[R0]
ORR R3,R3,R4
STRH R3,[R0]
LDR R0, addr
LDRH R3,[R0]
mov r4,#0x80
DEB:
bic R3,R3,R4
STRH R3,[R0]
NOP
NOP
NOP
NOP
NOP
NOP
NOP
NOP
NOP
ORR R3,R3,R4
STRH R3,[R0]
NOP
NOP
NOP
NOP
NOP
NOP
NOP
NOP
NOP
B DEB
; LUDO
Then I see a sqare signal of .68 micro seconde period.
0.68.10^-6 / 24 instructions = 0.28.10^-7
freq = 1/0.28.10^-7 = 35 MHz
Conclusion :
I lose time when I execute simple instruction ?
I do not understand why I need 15 micro secondes to call a handler ?
15uS = 525 instructions at 35 MHz.
7uS = 250 instuctions
I think I need only 50 instructions to get in the XTI_IRQHandler() ?
Do you understand my questions ?
Did I make mistakes in my tests ?
What do you suspect ?
Thanks for your help.
