Post by thread:Switch statements (was: Odd PCL behaviour on 16C84

>I use the modification of PCL extensively to implement switch statements. Is this a "standard" way to do switch statements, by messing with the PCL? I've been wondering about this. If the set of things you're switching on is consecutive, it seems a jump table arrangement would be good, but what if they're not? In my case, I need to switch on various non-consecutive ASCII characters. Currently I'm doing something like this: ----- dispatch ; compare w with each known command in turn movf savechar, w ; check for ques tion mark sublw '?' btfsc status, z goto dosomething movf savechar, w ; check for excl amation point sublw '!' btfsc status, z goto dosomethingelse ...etc... ; default: if we get here, we didn't recognize any command default call defaultaction return ----- I use gotos to call the individual routines, with return statements at the ends of those routines, so that once they're done they return to wherever dispatch was called from. This lets me skip all the other tests once I've found a valid case. Is this reasonable? Brain dead? Suggestions/comments are welcome. Dave Johnson

On Tue, Oct 13, 1998 at 07:57:31AM -0700, Dave Johnson wrote: > good, but what if they're not? In my case, I need to switch on various > non-consecutive ASCII characters. Currently I'm doing something like this: > movf savechar, w ; check for qu estion mark > sublw '?' > btfsc status, z > goto dosomething > movf savechar, w ; check for ex clamation point > sublw '!' > ...etc... > > Is this reasonable? Brain dead? Suggestions/comments are welcome. It's not unreasonable, but it can be improved, like so: movf savechar,w xorlw '?' btfsc status,z goto dosomething xorlw '!'^'?' btfsc status,z goto dosomethingelse xorlw 'A'^'!' btfsc status,z goto dosomeotherthing ...etc... This avoids having to reload the switched value each time. Of course you could j ust use a good C compiler and get this done automagically for you... Cheers, Clyde -- Clyde Smith-Stubbs | HI-TECH Software Email: spam_OUTclydeTakeThisOuThtsoft.com | Phone Fax WWW: http://www.htsoft.com/ | USA: (408) 490 2885 (408) 490 2885 PGP: finger .....clydeKILLspam@spam@htsoft.com | AUS: +61 7 3354 2411 +61 7 3354 2422 --------------------------------------------------------------------------- HI-TECH C: compiling the real world.

> Van: Dave Johnson <djohnsonKILLspamSIRIUS.COM> > Aan: .....PICLISTKILLspam.....MITVMA.MIT.EDU > Onderwerp: Switch statements (was: Odd PCL behaviour on 16C84) > Datum: dinsdag 13 oktober 1998 16:57 > > Is this a "standard" way to do switch statements, by messing with the > PCL? I've been wondering about this. If the set of things you're > switching on is consecutive, it seems a jump table arrangement would be > good, but what if they're not? In my case, I need to switch on various > non-consecutive ASCII characters. Currently I'm doing something like this: [Snip] > Is this reasonable? Brain dead? Suggestions/comments are welcome. I think that what you did was one of the best solutions for the problem, given the controller you are working with (A somewhat restricted type). You could have "searched" thru a table and used the index of the found char as an index into a Jump-table, but this would be wasting space in something like a 16?84 (or others with the same amount of memory for that matter). Both tables together would take 3 bytes (1 for the char, and 2 for the subroutine-adress) per "case" (not including the Check & Indexed-jump routines). Your solution uses 4 bytes per "case" The difference is noth worth it on the moment. You could shave of a byte per "case" by subtracting only the difference with the last "comparision". This would cause a stale-mate between the two (if we forget the surrounding code used to access both tables). Assume : W contains the current option Test: sublw '?' ;Check first choice btfss STATUS,Z ;Branch to next check if not equal goto Test2 ;/ -- Do stuff #1 -- return Test2: sublw '!'-'?' ;Subtract difference between "!" and "?" to ;check for the "!" char btfss STATUS,Z ;Branch to next check if not equal goto Test3 ;/ -- Do stuff #2 -- return Test3: etc, etc, etc ... (quoted from "the king of siam" :-) Greetings, Rudy Wieser

Dave Johnson wrote: > > I use gotos to call the individual routines, with return statements at > the ends of those routines, so that once they're done they return to > wherever dispatch was called from. This lets me skip all the other tests > once I've found a valid case. > > Is this reasonable? Brain dead? Suggestions/comments are welcome. > > Dave Johnson -- Hello Dave, As you require some comments, here are some ideas. If fact nothing new under the sun, only better 'R E A D A B I L I T Y'. Hi! My purpose is to have a one page (max) main program in a pseudo high level language. So I use macros, macros and macros... Some general rules ( of course there are exceptions) before reading: - Each identifier beginning with a capital letter, belongs to me (Constant, Variable, Label, Macro, Subroutine). - Only basic mnemonics are in lower-case letters. - In each case (except first), it's possible to avoid reloading w before testing, but I prefer 'straight programming'. Nobody is perfect. Here is an example from one programm (serial DDS synthesizer): ;__________________________________________________________________________ ;__________________________________________________________________________ ; ; MAIN PROGRAM ;__________________________________________________________________________ ;__________________________________________________________________________ noexpand Main Init Loop ; Loop FrequencyChange? ; Update frequency if need WaitNewCommand ; Get command on serialline CaseOf ; Case Of Case1 ; . Command = Frequency input Case2 ; . Command = Up Case3 ; . Command = Down Case4 ; . Command = Band change Otherwise ; . Otherwise ERROR EndCase ; End Case Forever ; Forever ; (In fact with my assembler, macros must be defined before main.) ;__________________________________________________________________________ ; ; MACROS (First level) ;__________________________________________________________________________ Case1 macro local NextCase movf NewCommand,w ; Case Of command = 'F' xorlw 'F' JNZ NextCase FrequencyInput ; Get frequency string goto EndCase NextCase ; endm ;-------------------------------------------------------------------------- Case2 macro local NextCase movf NewCommand,w ; Case Of command = '+' xorlw '+' JNZ NextCase call FreqInc ; Frequency Up goto EndCase NextCase endm ;-------------------------------------------------------------------------- Case3 macro local NextCase movf NewCommand,w ; Case Of command = '-' xorlw '-' JNZ NextCase call FreqDec ; Frequency Down goto EndCase NextCase endm ;-------------------------------------------------------------------------- Case4 macro local NextCase movf NewCommand,w ; Case Of command = 'B' xorlw 'B' JNZ NextCase GetBand ; Band change call BandChange goto EndCase NextCase endm ;-------------------------------------------------------------------------- Otherwise macro Beep ; Command does not match, so beep endm ;__________________________________________________________________________ ; Other macros .... ;JNZ is a better way to write: btfss STATUS,Z ; goto address ; and so one... ;__________________________________________________________________________ Of course to add one case all you have to do is to: - Insert a new line in main - Copy one case macro to get the structure, paste it and change what must be changed. - If you needs several CaseOf in your program, you have to number them. (My assembler does not allow local labels, except inside a macro. It's a pity!) Any comments? Francis ----------------------------------------------------------- Francis DUPONT F6HSI GQRP # 3553 E-mail : EraseMEFrancis.Dupontspam_OUTTakeThisOuTuniv-lille1.fr -----------------------------------------------------------

Regulus Berdin wrote: {Quote hidden}> > For readability, one could write a jump if equal macro: > > JIFEQ MACRO register,literal,address > movlw literal > xorwf register,w > skpnz > goto address > ENDM > > switch: > JIFEQ data,'A',Process_A > JIFEQ data,'B',Process_B > JIFEQ data,'C',Process_C > > process_A: > ... > return > -- This idea seems quite good, except for calling one process. - You must replace goto address by call address But when you return, as you have no EXIT, you make the remaining tests. No problem, only time consuming. OR - You replace each return by a goto EndSwitch. No unnecessary tests are done, so it's very fast, and you save one stack location which is sometimes useful. Don't forget some form of Otherwise before the EndSwitch. It's a good security! Cheers ----------------------------------------------------------- Francis DUPONT F6HSI GQRP # 3553 E-mail : Francis.Dupontspam_OUTuniv-lille1.fr -----------------------------------------------------------