I'm trying to access external devices on a XC161. A0-A19 are active. So are CS0-CS4. ADDRSEL4 is set to 0x0400 (start at 40000h, size 4k) Writing to 0x40000 works but CS0 is activated not CS4 ??? I have no experience with this family, so I must be doing something silly... Startup file was generated by DAVE, but checked out to be correct. Writing is done by use of a pointer. Is there an other/more direct way of accessing external memory/device? Frank
Using HVAR this way seems to trigger a lot of syntax errors...don't know why. I'll check the IO-pin settings....
Walt, I'm afraid I messed up my code...guess it's getting late over here. The device was configured with the help of Infineon's "Dave2.1"(well, you have to start somewhere) Strangely enough Dave sets the direction of the CS-pins to input when configured as CS's! I have no clue why... Well, tomorrow is an other day. Thanks so far. Frank
O.k. i think I may have found the problem.... It looks like a problem with the Start_V2.A66 file that Keil generates. I think the following macro is not accurate CALC_ADDRSEL MACRO sym, adr, size IF size <= (4*KB) sym EQU ((adr >> 8) AND 0xFFF0) OR 0 ELSEIF size <= (8*KB) sym EQU ((adr >> 8) AND 0xFFE0) OR 1 ELSEIF size <= (16*KB) sym EQU ((adr >> 8) AND 0xFFC0) OR 2 ELSEIF size <= (32*KB) sym EQU ((adr >> 8) AND 0xFF80) OR 3 ELSEIF size <= (64*KB) sym EQU ((adr >> 8) AND 0xFF00) OR 4 ELSEIF size <= (128*KB) sym EQU ((adr >> 8) AND 0xFE00) OR 5 ELSEIF size <= (256*KB) sym EQU ((adr >> 8) AND 0xFC00) OR 6 ELSEIF size <= (512*KB) sym EQU ((adr >> 8) AND 0xF800) OR 7 ELSEIF size <= (1*MB) sym EQU ((adr >> 8) AND 0xF000) OR 8 ELSEIF size <= (2*MB) sym EQU ((adr >> 8) AND 0xE000) OR 9 ELSEIF size <= (4*MB) sym EQU ((adr >> 8) AND 0xC000) OR 10 ELSEIF size <= (8*MB) sym EQU ((adr >> 8) AND 0x8000) OR 11 ENDIF ENDM Address values are shifted 8 bits to the right(?) instead of 4 bits to the left! (right 4 bits repr. size) Things seem to work fine after this correction. I hope I won't encounter more surprises like this... Frank
The startup code allows you to specify the address as plain number, i.e. 0x400000. However, the ADDRSELx registers need just the A12 .. A23 address bits (number of bits depends on the window size). So your problem must be related to the way you specify the address in the START_V2.A66 file. We never had problems before, but maybe we should clearify a comment somewhere, so that people put real address values.
Yes, I already figured that one out. The problem however, is that I did not specify the address in the START_V2.A66 file. It was generated by Dave2.1. So, I guess there must be a problem in the compatibility there! I already wrote you an email about this matter. Thanks anyway Frank
We did not got your email. Please send the DaVE project you are using to: support.intl@keil.com
Re: Activating Chip Selecet Hello, we got a similar experience with the DAvE and I inspected the startup file. There we found differences between _ADDR1 and _SIZE1 and resulting _RGSZ1. I have asked one of the authors of DAvE and it should be a misunderstanding, but we tested and it seems definitly wrong. I wrote a second mail to DAvE author with more detailed explanation of startup file, I hope I get new answer and if something wrong I will post it. Stefan
Stefan, You have to manually adjust the startup file. I don't know where the problem lies: with Keil or Dave. Please do post any useful reply you may get from them. Frank
Are you using the latest C166 and the latest version of DAVE? I brought up this question to an Infineon guy and to our support guys and all of these issues were corrected a long time ago. Jon
Frank, as I promised, I post the answer from the DAvE authors. The original message is written in german, so I try to translate. DAvE generates a *.dpt file, which is used by the KEIL IDE to generate a project. Within this *.dpt file all register bitfields and bits are listed with there actual values.'Here e.g. you find that the values of ADDRSEL1_RGSAD1=256 and ADDRSEL1_RGSZ1=5 are.Exactly the registernames as given by Infineon here are used.Now it is in the responsibility of the kEIL IDE to adapt the startup file in the right way. This is not under DAvE's influence. We are sorry that we can not help you in this case, but this problem has to be solved by KEIL . I do not know, whether original message is usefull, so I do not post here. I work with two activated CS and 2 address lines ( A16...A17 ) via EBC for an SRAM and it works. - Stefan
Thanks, stefan. My applications works ok now too. I use 5 CS's and 20 address-lines. This problem only applies to the new V2-core like in the XC161, because it handles CS-settings differently. If you're using Dave with these chips, just change the "adr >> 8" parts in the startup file into "adr << 4". (see one of the previous messages) You only have to do this once. After this, CS's will work fine. Frank
Hi Frank, thanks a lot for this info. This macro we found suspect as I got back the first mail from the DAvE's author. But at this time I had not enough time to check completely. Sounds really good, what you wrote. I will try it out asap. What I did, was to inspect DAvE's *.dpt and the project_name.asm which hold the asm formatted informations for IDE. Right was the authors say - DAvE calculate it all right. So the macro must be the bad boy. ;-} Stefan
Hi Frank, I checked the MACRO again and with the DAvE register viewer (There it is the right value ).You are completely right. It is definitly a problem with KEIL IDE's startup file. Thanks again - your workaround solves it completely. Stefan
The integration problems between uVision2 and DAvE have been corrected in the latest C166 V4.27 release. http://www.keil.com/update/sw/c166/4.27 Jon