Address Space overflow

Set the option to produce a linker map file, enable everything, do a rebuild (if necessary) and look for the information at the start that begins with something like:

BL51 BANKED LINKER/LOCATER V5.03, INVOKED BY:
C:\KEIL\C51\BIN\BL51.EXE BANK1 {.....etc.......}

Set the option to produce a linker map file, enable everything, do a rebuild (if necessary) and look for the information at the start that begins with something like:

BL51 BANKED LINKER/LOCATER V5.03, INVOKED BY:
C:\KEIL\C51\BIN\BL51.EXE BANK1 {.....etc.......}

I think you need to examine details relating to the 'common area'. I suspect the common code might be being placed at the start of each bank.

You mention that the datasheet specifies:

Common Bank Area:  0000H to 8000H
Bank1 Area:        8000H to FFFFH
Bank2 Area:        8000H to FFFFH
Bank3 Area:        8000H to F7FFH

But, there is no specification of the common area in the linker invocation.

For example, my old project (which seems to have similarities to your memory layout) has a linker invocation containing:

BANKAREA (0X8000, 0XFFEF) RAMSIZE (256) DISABLEWARNING (16) CODE (0X0000-0X7FFF) XDATA (0X0000-0XFFFF)

Note the 'code' which I did not see in your linker invocation.

My observations, conclusions and Solution to Address Space Overflow

* * * * * * *   C O D E   M E M O R Y   * * * * * * *
CODE    0000H     0003H     ABSOLUTE
CODE    0003H     0007H     UNIT
        000AH     0001H                  *** GAP ***
  .       .         .        .
  .       .         .        .
  .       .         .        .
  .       .         .        .
CODE    97F5H     001FH     UNIT
CODE    9814H     001EH     UNIT
CODE    9832H     0012H     UNIT
CODE    9844H     0010H     UNIT
CODE    9854H     000BH     UNIT
CODE    985FH     000AH     UNIT
CODE    9869H     000AH     UNIT
CODE    9873H     0007H     UNIT
CODE    987AH     0007H     UNIT      end address = starting address of function + length Refer NOTE
                                      end address = 987A + 7 = 9981 
* * * * * * *   C O D E   B A N K   1   * * * * * * *
        0000H     9881H                  *** GAP ***    Refer NOTE
BANK1   9881H     1D5BH     UNIT
BANK1   B5DCH     1756H     UNIT
BANK1   CD32H     0D51H     UNIT
BANK1   DA83H     0623H     UNIT
BANK1   E0A6H     05DDH     UNIT
BANK1   E683H     0541H     UNIT
BANK1   EBC4H     044EH     UNIT
BANK1   F012H     03D7H     UNIT
BANK1   F3E9H     03BCH     UNIT
BANK1   F7A5H     033AH     UNIT
BANK1   FADFH     0322H     UNIT
BANK1   FE01H     01B4H     UNIT
BANK1   FFB5H     004BH     UNIT

* * * * * * *   C O D E   B A N K   2   * * * * * * *
        0000H     9881H                  *** GAP ***     Refer NOTE
BANK2   9881H     03CCH     UNIT
BANK2   9C4DH     02A7H     UNIT
BANK2   9EF4H     02A7H     UNIT
BANK2   A19BH     0205H     UNIT
BANK2   A3A0H     0175H     UNIT
BANK2   A515H     00F6H     UNIT
BANK2   A60BH     00B2H     UNIT
BANK2   A6BDH     0079H     UNIT
BANK2   A736H     0053H     UNIT

* * * * * * *   C O D E   B A N K   3   * * * * * * *
        0000H     9881H                  *** GAP ***      Refer NOTE
BANK3   9881H     06ECH     UNIT
BANK3   9F6DH     0220H     UNIT

The following is my conclusion.

NOTE: The starting address of each bank = End address of common bank + 1

When I shifted some piece of the code from common bank into other banks, I found that the starting address of each bank in the M51 file (A part of which i have shown above), changed.

However, a very good solution is change the Code Optimization level to Level10. Go to:

Project -> Tool Chain Integration -> Compiler (Tab) -> Customize -> Optimization -> Level

Select "Level 10: Rearrange Code (Linker Optimization)".
Save Project and click on "Rebuild" or press "Ctrl + Shift + F7" keys.

Now write your code in any bank without worrying about linking. The linker will take care of everything. It rearranges the code in each bank and hence will generate the hex file with minimum code size.

OOps! Made a mistake in addition in my previous post.

987A + 7 = 9881

Sorry.

Dear All,
If in case you find my observations wrong, or have any other solution, kindly inform me.

Thank you for all the necessary help, guidance and valuable time you provided. Learnt a lot.
Hope the observations and conclusions helps other newbies to solve the problems related to 'Code Banking' and 'Address Space Overflow'.

Regards,
Dhaval Solanki

Glad to see you've got it going.

Regardless of what anyone else might say about banking, I think what Keil did to provide the mechanism is really pretty smart. Certainly for the application I used it on, the results were perfectly acceptable; in terms of development time, reliablility and end product costs. And, after the initial learning curve in getting it going, the fact that we were using banking was virtually forgotten about.

That said, for any new projects, I would recommend you keep an open mind and consider something different.

Dear,
I already highlighted
... Cortex-M3 is used for all the new project development. ... Never the less, I would also, as my personal opinion, always work with latest controllers and technology. .

I am glad that Keil has provided the feature "Optimization: Level 10: Rearrange Code (Linker Optimization) ". I mean they should, to be the best IDE software developers. Respect.

But i still Fail to understand

The starting address of each bank = End address of common bank + 1

This should not happen, ideally.

But i still Fail to understand

The starting address of each bank = End address of common bank + 1

This should not happen, ideally.

On the contrary: that must happen. Your believing otherwise indicates you don't really understand how banking works.

The common bank is called a "common bank" because it contains the material that has to be commonly accessible to code in all banks. The only way that can work is if the entire address range of the common bank is reserved and kept available, regardless which other banks are mapped. In a scheme like SiLabs', with a 32KiB+32KiB hardwired bank mapping, the only way the common bank can become bigger than 32 KiB is if all bytes beyond the 32 KiB boundary are copied to all 3 "upper" banks, so they'll be accessible regardless which of the upper banks is currently active.

I think your real problem was that in your original configuration you placed too much stuff into the common bank yourself, causing this spill-over mechanism to be triggered.

Dear Hans-Bernhard Broeker,
Thank you for clearing the air. But, in my opinion, there should be a document produced by the chip vendor stating this. A newbie may not understand this, the very first time he is using the IC, or may not have a clear picture unless he faces a problem like me.

Dhaval

But, in my opinion, there should be a document produced by the chip vendor stating this.

Though it might be helpful if the chip vendor produced application notes or other documentation relating to such things, I don't think it is really something that they should be expected to do. Which compiler vendors would they document for? Where would they stop?

The Keil documentation relating to banking is reasonably good and:

http://www.keil.com/support/man/docs/bl51/bl51_codebanking.htm

It just requires the reader to have an appreciation of what it is trying to achieve and what is possible.

But, in my opinion, there should be a document produced by the chip vendor stating this.
well, SILabs does

The Keil documentation relating to banking is reasonably good and:
but is somewhat invalid for the SILabs (and other?) chips with "internal banking"

go by the SILabs appnote, it is crystal clear

Erik

go by the SILabs appnote, it is crystal clear

I'll take your word for it :)