I have created two separate programs for my AT89C51ED2: bootloader (0x0000) and normal program (0x2000). I use an absolute memory addressed variable to distinguish between the two modes of operation so that interrupts either execute code in the bootloader or call the appropriate ISRs in the program code (above 0x2000). The following is an example of the code used for interrupts in program code, in order to eliminate the usual pushes, pops and reti done by the bootloader ISR.
/** Timer 1 ISR. */ void interruptTimer1(void) interrupt 3 { #pragma asm lcall interruptTimer1Mirror ret #pragma endasm } /** Timer 1 interrupt working function. This routine is used for delays. */ void interruptTimer1Mirror() { timer1Interrupted = TRUE; }
You can get into a right mess trying to do this sort of thing. Might I suggest an alternative: Invoke the built-in bootloader when you need to upgrade the software. You then have a choice of using Flip or writing your own downloader.
Many thanks for your reply. My intention is to make the device upgradable in the field via an external UART chip (not the built-in serial port). In the stand-alone bootloader (0x0000-0x2000) I call the API routines which reprogram the device program code (0x2000-0xFFFF). I am not using a PC to download new firmware. My problem seems to be getting the two programs to work independantly (specifically the interrupt side of things) without corrupting data in the main program. Is there no way to get the interrupts to jump ONLY to my new vector table at 0x2000? What other ways are there to make the device network field-programmable ie. without going to each and every device and plugging in a serial cable?
Is this a problem with the linker/locater not knowing about the separate bootloader program/project? Could be that. Or it could be the linker not knowing about the call from that IRQ handler to the function at all --- I don't think call tree analysis looks into inline assembler blocks. What might work is to tell the compiler to place the interrupt table at 0x2000, even for the bootloader, and move the relevant parts of the bootloader code to a special section that gets written there iff there's no application in place, before you enable interrupts.
I wish to do the same thing. Is there a Flip for Cygnal C8051F021 ? Thanks.
I have successfully accomplished the exact thing that you desire. The trick is to actually use 3 programs. The first contains the interrupt vectors located at address 0. All interrupts are handled. Each interrupt handler interrogates a specific bit (I either use F0 or 20^0) to determine whether the interrupt is for the BootLoader or the real application. It then jumps to the corresponding interrupt +0x200 if in the bootloader or +0x2000 if in the application. The normal startup at location 0 points to startup code in the BootLoader (Program at address 0x200). The bootloader sets the "BootLoader" flag, determines whether the application is loaded (via checksum info saved in an unused interrupt vector address) and if so, clears the "BootLoader" flag and jumps to address 0x2000. If the application is not loaded it simply waits to do the download. When I load the bootloader initially, I merge the hex for the interrupt vector program with the bootloader hex to create a valid operating program. The application has it's interrupt vector set to 0x2000. The startup code is modified to ORG at 0x2000 also. Other than that, the app is just a normal program.
I wish to do the same thing. Is there a Flip for Cygnal C8051F021 ? The downloading software for the SILabs (nee Cygnal) chips is JTAG based. The software is free and you need a 10 pin JTAG connector on your board and an EC2 (serial) or EC3 (USB) adapter. Erik PS: you can get a complete devkit including the EC adapter for about $10 more than just the EC. I have a lot of SILabs devboards because I needed 8 EC2s for my field guys and took the chance to but devkits.
Thanks for the advice. I'll give it a try...