Nuts and Volts - October 2010

 else blockFlashRead( tempSize  );
}

This function — depending on whether cmd is ‘B’ or ‘g’ — calls functions to load or read the Flash. The load function is pretty much the same as the Flash load function we discussed last month for the SmileyFlash Writer. The blockFlashRead() function uses the avrlibc pgm_read_byte_near(address) function similar to what we saw in Workshop 25.

void blockFlashRead(uint16_t size)
{
uint8_t data;
do {
 // read_program_memory(address,0x00);
 data = pgm_read_byte_near(address++);
 // send byte
 sendByte(data);
 // reduce number of bytes to read
 // by one
 size—;

} while (size);

// loop through size
}

Keeping the Bootloader as Small as Possible

When we compile EduBootAVR with AVRStudio/WinAVR with our usual settings, we get a code size of 1,086 bytes. Oh, darn, the boot section boundary is at 1,024, so it is 62 bytes too large. Can we get the code smaller? Could we rewrite the program hoping to find something that will save us 62 piddly bytes? Well, we might, but there are other ways to save some space with a bootloader.

Eliminate Some of the Start Code

When we use AVRStudio with WinAVR to compile files, we get a gift of having necessary startup code added for us. This includes the interrupt jump table. But what if you aren’t using the interrupts? Well, you can eliminate it. Since we aren’t using interrupts in the bootloader, let’s dump the jump table! We do this by opening the Project Configuration Options window Custom Options pane, as shown in Figure 2. We write ‘-nostartfiles’ in the ‘Add’ test box, then click ‘Add’ so that ‘-nostartfiles’ appears as shown in the figure. This also removes some startup code that we do need, so we have to add it back by putting the following code snippet right before the main() function:

■ FIGURE 4. Connect Dialog.

SMILEY’S WORKSHOP ☺

■ FIGURE 3. Project Options General.

// From Peter Fluery AVRFreaks Aug 10 2005 -
// to remove interrupt Vector table
// put -nostartfiles in LDFlags, add the
// following function saves wasted space
void __jumpMain (void) __attribute__ ((naked))
__attribute__ ((section (“.init9”)));

void __jumpMain(void)
{
 asm volatile ( “.set __stack, %0” :: “i”
 (RAMEND) );
 asm volatile ( “clr __zero_reg__” );
 // r1 set to 0
 asm volatile ( “rjmp main”);
 // jump to main()

Now when we compile the code, we get 958 bytes which will fit in the desired boot section. We could stop here, but there is also another way to save some space.

Use a Different Optimization Level

We can keep the code unchanged and instead of using the default optimization of –Os, we can use –O1. You change this as shown in Figure 3. Just select –O1 from the Optimizations drop box. When you compile using this optimization, you get 986 bytes which would also get you inside the limits.