how to remap interrupt vector - pic18f4550

[cluster]

Hi Charles,
i created this simple program and tried to run but it didn't worked. it should blink led and remap high priority interrupt vector

adcon1 = 15
DEFINE OSC 48

asm
org 0x0008
goto 0x2008
endasm

i var byte
j var byte

main:
high porta.0
for i = 0 to 254
for j = 0 to 254
high porta.0
next j
next i
for i = 0 to 254
for j = 0 to 254
low porta.0
next j
next i
goto main

but if i comment out the goto 0x2008 line it starts to blink led. since i am not using any interrupt then why its not blinking led?
can you please help me

thank you

[Charles Linquis]

I hope there is more to your program that you show. The PIC jumps to the interrupt vector location (0x08) when some peripheral (timer, serial port, etc) generates an interrupt, at which time the program counter gets reset to 0x08.

Just use Darrel's Instant Interrupts.

[Charles Linquis]

I should add that the program counter jumps to 0x08 whenever both the a particular interrupt ENABLE and interrupt FLAG are both set. The program counter won't
jump around location 0x08 unless you tell it to. So at location 0x00, you need a jump to a location AFTER 0x08

I currently use a technique where I get data in an RS-232 input buffer at 230Kbaud, and the data must not be blocked (must be able to receive at all times), and the data must be processed
and a response given very quickly. Receiving the data in an ISR was no trouble, but processing and responding inside that ISR would block the next bytes, if they came immediately
after the first.

Processing and sending the data in the main loop (which was long), didn't work either because that would sometimes delay the response by up to .5 sec.

So, I set up the RX_INT as high priority, and just before the INT_RETURN, I set the TMR3 interrupt flag. But I don't have Timer3 running! Timer 3 interrupt is turned on, but is low priority.
The Timer 3 ISR jumps to the "process and send" routine".
If a packet comes in (I'm using an Ethernet to RS-232 converter, and I don't have flow control connected), it always gets captured. As soon as the packet is received, it triggers
the "process and send" routine. If another packet comes in immediately, the recieve ISR is entered again, since it is at a higher priority. Pointers insure that new data doesn't over-write old data.
Now the latency is 4-7 mSec, since it doesn't have to wait for the main loop to complete.

[amgen]

Along the same lines of recieving data,
I frame the data with a start(to take) and end(to trigger) char, fast enough to be handled in the interrupt routine, then flag basic looping at 10X/sec.

IE......... [ abcdxyz...COMMAND.....1234]

"[" to start taking chars to array,
"]" to stop taking and flag to basic to parse for command ete etc.

the "[" always sets the array back to start.

Don
Amgen

[Charles Linquis]

But I don't have that luxury. I'm getting SNMP packets (80-230 bytes). The host sends me a request, and has about a 20mSec timeout. I have to do an enormous amount of processing to decode the packets, grab the appropriate data, rebuild the response packet, and send. I previously did everything (receive, process, send) in one long ISR. That worked great. I was getting queries every 410mSec. The problem is - the customer added a second host that was sending me queries every 1 second. Most of the time, all was OK. But of course, there were times when the two requests hit me at virtually the same time. Since I was processing and sending in the 'receive' ISR, I missed the second request. I should mention that I don't have any flow control out of my network port. The hardware is already in the field and the handshake lines are not connected.
So I made the ISR a 'receive only' segment and made my buffer twice as long (so it could handle two incoming packets), and modified my routines so that they worked with an offset. Offset '0' is the first byte of the first packet and offset $100 is the first byte of the second packet.
When a packet came in, I set flags (packet offset 0 recieved and packet offset $100 received) and tested those flags in my main loop. Unfortunately, my main loop takes up to .5 second to execute. The host would time out before I sent my response.
So I set up Timer3 Int as a low-priority int, but didn't turn on the timer. I set PIR2.1 just before I exit the (high priority) receive ISR. As soon as it exits, it goes into my 'fake' TMR3 int and processes and sends the packet. If another packet comes in right after the first, the high priority receive interrupt stops the 'process and send' routine and grabs the packet. It then goes right back to processing and sending - both packets if it needs to. I'm running the net-port to PIC interface at 230K baud.
It all works!

[amgen]

OK, I think I follow your sequence. Sounds like your using Basic serout or HWserout ??
I use TXint to send from either ram(array) or flash(1 of several tables at the end of flash mem).
I will put together the easy change to PB18....lib for arrayread and send you. You (and me too) can be confident since Darrel checked it out.
Don

[Charles Linquis]

I'm using HWserout, but I don't think I could do much better using TXint. It isn't the sending time that is the problem, it is the large amount of processing I have to do with the packets.
SNMP requires that I do lexicographical sorts (1.2.1.3.1.1.1.4.2.1 is before 1.2.1.3.1.1.1.4.2.1.1 and 1.2.1.3.1.1.1.4.2.1 is after 1.2.1.3.1.1.1.3.9.9.9.9.9.9.9)
Even if you get sent a string that does not match one that you have (internally) in your list, you have to pick the first item in your list that (lexicographically) comes after the received string,
then grab the appropriate data, perform encryption on certain bytes (3 different types, depending on the data), re-assemble the variable-length packets, calculate the LRCs, and send.
And some of the data is a (calculated) number 40 bits long.