Instant Interrupts - Revisited

[vamtbrider]

I will have to explore the Microchip bootloader a little more. I strayed away from it because I think it requires a .dll and I wanted a usb bootloader that would work with the standard microsoft drivers. The Diolan also uses encryption which is a very nice feature and the code size is slightly smaller.

Plus I like a challenge...maybe if I can get this working it will give others another option for bootloaders and DT_INTS.

So, on to what I have found today.

From section 26.2.4 of the 18f4550 manual:

Additionally, the Indexed Literal Offset Addressing
mode may create issues with legacy applications
written to the PIC18 assembler. This is because
instructions in the legacy code may attempt to address
registers in the Access Bank below 5Fh. Since these
addresses are interpreted as literal offsets to FSR2
when the instruction set extension is enabled, the
application may read or write to the wrong data
addresses.

If I look at the .LST file for the blinky led program I see this:

00037 FLAGS                           EQU     RAM_START + 000h
  00000001            00038 GOP                             EQU     RAM_START + 001h
  00000002            00039 R4                              EQU     RAM_START + 002h
  00000004            00040 R5                              EQU     RAM_START + 004h
  00000006            00041 R6                              EQU     RAM_START + 006h
  00000008            00042 R7                              EQU     RAM_START + 008h
  0000000A            00043 R8                              EQU     RAM_START + 00Ah
  0000000C            00044 INT_Flags                       EQU     RAM_START + 00Ch
  0000000D            00045 RM1                             EQU     RAM_START + 00Dh
  0000000E            00046 RM2                             EQU     RAM_START + 00Eh
  0000000F            00047 RR1                             EQU     RAM_START + 00Fh
  00000010            00048 RR2                             EQU     RAM_START + 010h
  00000011            00049 RS1                             EQU     RAM_START + 011h
  00000012            00050 RS2                             EQU     RAM_START + 012h
  00000013            00051 wsave                           EQU     RAM_START + 013h
  00000014            00052 RetAddrH                        EQU     RAM_START + 014h
  00000017            00053 RetAddrL                        EQU     RAM_START + 017h

Looking at the 2nd Capture.txt file I attached yesterday there are lots of MOVWF statements that reference the RAM areas above which are in RAM at locations under 0x5F.

So a statement like MOVWF [0x14] is really MOVWF [FSR2+0x14] with the extended instruction enabled. This may then explain why the code jumps off into la la land and resets.

I did not think to capture FSR2 last night during my debugging. I will do that tonight and see if it matches my theory. If it does then I think I will try to change the RAM_START to 0x060 so that the extended instruction engine will not use the FSR2 on the Ram addresses.

Am I on the right path?

[ScaleRobotics]

I will have to explore the Microchip bootloader a little more. I strayed away from it because I think it requires a .dll and I wanted a usb bootloader that would work with the standard microsoft drivers.

I think there are two usb bootloaders from microchip. One is a HID Bootloader. I don't think it requires any dll's. It is located inside the huge files that you download as a group. Once all the Microchip Application Libraries v2010-02-09 is installed, the .hex file is located here: C:\Microchip Solutions\USB Device - Bootloaders\HID - Bootloader

I believe you can just bring the hidbootloader exe to a different computer, and program over the usb. But I have not fully tried it yet. I have only got to the point where my chip was regognized by a computer that has never had any microchip software loaded on it.

I need to try the executeable.

[vamtbrider]

The Diolan bootloader works as a HID USB device and my end app will eventually be a HID devices as well. If I get time this weekend I will try the Microchip HID bootloader, as Bruce indicated that maybe the more solid solution.

I am looking at changing the statement in MPASM P18F4550.inc from

ACCESS  EQU  0

to

ACCESS EQU  1

This will allow the RAM definitions to stay the same and get past the extended instruction issues with f<0x5F. The draw back will be a lot of BSR changes. We'll see how that works out.

[Bruce]

The major issue you have with the Diolan loader is that it has Enhanced CPU mode
enabled. I.E. XINST = ON.

So code compiled with PBP simply is not compatible, and will not run on the PIC as
expected unless you disable Enhanced CPU mode.

Here's a simple example you can try. Make sure you set XINST = OFF, then run this;

MyVar VAR WORD
LED VAR PORTD.0 

Loops:
 FOR MyVar = 0 TO 100
   High LED        
   GOSUB BLINK  

   Low LED       
   GOSUB BLINK  
 NEXT MyVar
 
   Goto Loops 

BLINK:
    PAUSE 200
    RETURN
    
    End

Now run the exact same code, but set XINST = ON and give it a whirl.

Once you flip that switch to enable Enhanced CPU mode, ALL your code needs to be using
the extended instruction set. Try it with ACCESS EQU 1 as well. Still no go.

[vamtbrider]

What I found interesting is that PicBasicPro compiles this code correctly for XINT on. Using brackets and leaving out the access parameter for addresses under 0x5F.

The blinky light code I originally posted now runs fine with the Diolan bootloader. The trick I found was to clear the FSR2L and FSR2H registers as the first instructions in my program. I also set Access equ 1.

I think the main issue was that I never initialized the FSR2 register to zero in my code. The Diolan bootloader uses FSR2 and it could have any value in it when the app code runs.

Since that is the register used for indexed addressing when XINST is on it must be set to zero to allow legacy code to run. Setting FSR2 to zero places the access bank and SFR's in memory just as if XINST was 0. I found this info in the data sheet. Had to read it about 10 times to make sure I was following it correctly.

The next challenge I am working on is to get the USB interrupts working with DT_INTS and the Diolan Bootloader. The USB_DEV.ASM and USB_HID.ASM use FSR2/INDF2 for indirect addressing and as a general register in several routines. Those will need to be modified to use FSR0/1 and INDF0/1 instead.

[Ioannis]

OK. I am pulling my hair off. Cannot think what might be wrong here.

I get errors like Found label after column 1 (INT_CREATE) or Illegal opcode (RBC_INT) or Address label duplicated in second pass (INT_RETURN).

The PIC® is F819, with DT-INTS-14 and MPASM of course. Also commented the confg line in the *.inc file as it can be seen on the attached files.

Ioannis

[Bruce]

Did you forget these?

INCLUDE "DT_INTS-14.bas"
INCLUDE "ReEnterPBP.bas"

[Ioannis]

No Bruce. They are in pbp directory and the project too.

It is not the first time I use these excellent routines.

That is why I post here. I am almost driving myself in the la-la land with this!

Ioannis

[ScaleRobotics]

Did you forget these?

INCLUDE "DT_INTS-14.bas"
INCLUDE "ReEnterPBP.bas"

Compiles here as well, but I do get the same errors IF I leave the above lines out of the code. Are you sure the include lines are in your water?