Table des matières

Here are some of the pages of the User Guide!

Quick Start

  • RUN "#BURN" (to install ROMs.)
  • |M then ESC, or |O to go directly to the editor.
  • CONTROL-H For online help, available in the editor, monitor, and debugger.
  • CONTROL-I (Import) Loads a DAMS source file, Maxam (ASCII) or TurboAss (exported to ASCII).
  • or: CONTROL-O to open an exemple source (.o files).
  • CONTROL-1 To assemble.
  • SPACE To return to the editor.
  • CONTROL-4 To cycle through the current errors.
  • CONTROL-S to save. CONTROL-O to open.
  • CONTROL-2 To assemble & run.


Like DAMS Orgams breaks down into:

  • Editor (with validation on the fly lines seizures)
  • Monitor (aka "Monogams")
  • Debugger (disassembly, trace step-by-step, etc …)

OrgamS is quite greedy and takes 4 ROMs.

  • ORGAMS.ROM: Base Rom place in any rom position 1 to 15 (or up to 31 with the proper firmware).
  • ORGEXT.ROM: Expansion Rom placed at any rom position 1-127.
  • MONOGAMS.ROM: Monitor placed at any rom position 1-127.
  • BRICBRAC.ROM: Additional additional Rom at any position 1-127.

Several banks are being used (decreasingly), basic bank (bkBase) being:

  • C7 with 128k
  • FF with 512k (allows cohabitation with 256k RAM-DISK).

The last 4 banks are entirely taken:

  • bkBase: Housekeeping and internal buffers.
  • bkBase-1: Firmware mirror and work zone.
  • bkBase-2: Reserved for &c000 screen mirror (but not exploited now, completely wasted. See todo #f2).
  • bkBase-3: Source start

Hence, with 128k (CPC stock + ROMs), there is less than 16k available for source, and only the base 64k for the program (that being said, all the memory is usable, including firmware zones, without perturbing Orgams).
If it's too little room, please use Demomaker'sDelight version.

Screenshots of monitor, editor and trace.



/!\ Delete all sources and jump to monogams (the monitor)
You should never have to use this one, expect when memory was corrupted, to start afresh.

|ORG, |M, |OM

Jump to monogams.


Install Orgams from another bank (C7, CF, D7, DF …).
You should never have to use it. Only meant for private debugging purpose (multi-orgams in one day).

|O, |OE

Jump to editor.

|O,"filename" |OE,"filename"

Load "filename" (".o" added if no extension) to free first tab, or tab 9 if no one is free, and jump to editor.
If tab 9 contains an unsaved source, it is proposed to save it (N to skip saving, ESC to cancel the whole operation).
Equivalent to: |ORGLOAD,"filename":|O


Load "filename" in indicated tab (or free first tab if tab parameter is omitted) and returns to BASIC.
Again, it won't overwrite an unsaved source.


Clear all sources (proposing to save modified ones).
Pretty useful to create load a clean workspace:

20 |ORGLOAD,"journal",5
30 |ORGLOAD,"ply",2
40 |O,"plynrt"


Jump to trace at passed rom and address. Both parameters are optional

Flow between modules

Here, "Exec" represents the assembled program running.

From Basic
See RSX section.
The current source is checked against memory overwrite than might have happened when using other tools. E.g. Parados doing a destructive memory scan.
(TODO: Do the check for all tabs, not only the current one).

From Monitor

  • ESC: Goto Ed
  • Order 'basic': (Back) Basic.
  • Command 'DEXP' Goto debugger

From Ed

  • CTRL-F2: If ok, Goto Exec
  • ESC: Goto Monitor

From Exec

  • RET: as RESTORE (but the battery must be valid)
  • RESTORE Restores system and Goto Ed.
  • BRK / rst &30 / call &BE00: Goto Trace


  • CONTROL-ESC: Goto Editor to visualized line.
  • ESC: Goto Monitor
  • J: Goto Exec

Programming session example:

  1. Tweaking opcodes via the editor. CONTROL-4 makes it possible to go directly to the reported errors during assembly (CONTROL-1).
  2. Rather than a mere RET, use RESTORE or BRK to quit out of your program.
  3. Save source.
  4. Test by CONTROL-2: it assembles the code and run the program.
  5. The monitor allows to study all memory (except the screen memory this is used by Orgams' display).


Enter the editor using |O from BASIC, or by pressing ESC in the Monitor.
Leave the editor by either pressing ESC, launching assembly + execution (CTRL 2), or by resetting the CPC.

The current session persists across resets, with the source validity check.

Status board

From version EE beta H, the status board displays the following information:

  • Line 1
    • Path in source (in which routine and IF block the cursor currently is).
    • Line number / total lines.
    • Current MMR / RRM selection (like in Monogams)
  • Line 2
    • List of opened sources (a.k.a. tabs)
  • Line 3
    • Dump of current line, with the difference with original code highlighted.
    • Block related information (e.g. machine time taken).
    • Number of Hectibytes free (like in Monogams)
  • Line 4
    • Questions and results of actions.
    • Banks used by Orgams (like in Monogams)
  • Line 5
    • Internal debug information.

It looks like something like that ('.' for border):

.. dev=PRELUDE.GenTexture.loopx.dbg!   133/ 945                                7fc2  LRAM / 0a    ..
..1 orgch     2*crtc     3*cpct     4*         5*         6 nrt    7 player  8 bulgur  9 leprous  ..
..1203: 3E 04 09 90 CA 34 FE 32         100..122: 32 bytes, 64 NOPs                123 Hib free   ..
.. Found: string                                                                  Used: 7fdf-ff   ..
..                                                                                         fc:4534..
.......... ; This is the actual source                                                    ..........
.......... ; idem                                                                         ..........

Import sources from other assemblers

The first thing you want to do is to listen to Solefald.
The second will be to import your old tatty sources (see "Write a technical manual for Dummies", how to create "user stories").
Via CONTROL-I, Orgams automatically recognises source from DAMS and other ASCII sources (TurboASS, Maxam or ORGAMS).
The various adjustments will be made automatically for example:

  • 'NbDots EQU &1000' becomes 'nbDots = &1000'
  • 'Defs 100,nbCols%3' becomes 'FILL 100,nbCols MOD 3'

The operation is very slow because much of the pre-assembly and symbol resolution is done at this time.

Possible differences. !!WARNING!!

Modified Arithmetic.


  • LD A, pal&#ff+1; (pal and #FF) + 1


  • LD A, pal AND &FF+1; pal and &100

Indeed, as &ff+1 is visually grouped, it is therefore calculated separately.
It must be corrected to either of these:

  • LD A, pal AND &FF + 1; Direction left to right
  • LD A, [pal AND &FF]+1

ORG behavior

When DAMS assemble with A2, each ORG changes the PC ($) but not the storage address ($$). However ORGAMS does
not allow this mode. To reproduce the same action it must be explicitly replaced:

  • (DAMS) ORG dest
  • (ORGAMS) ORG dest, $$

1 parameter DEFS

The 'DEFS size' DAMS is replaced by 'F size', resulting in a syntax error.
Indeed, 'F' is known as 'FILL', but only when followed by two mandatory parameters.
With one parameter, DAMS reserves space without initializing it. There is no equivalent in ORGAMS. We must
therefore explicitly write 'F size, 0' (automatically replaced by 'FILL size, 0').


In Monogams, use the X command to compare the assembled code with your referral code.


  • Editing of lines is limited to 72 characters. If asked why, you will say that you do not know.


Keyboard shortcuts are described in the online help (press CONTROL-H).
We concentrate here on more sophisticated.

CTRL + C Displays the catalog of the disk.

CTRL + L str

Go directly to the next label beginning with "str". This is equivalent to the command 'the DAMS, with the following improvements:

  • Case Insensitive
  • You can iterate through all labels beginning with "str". Just press CTRL-L again then RETURN to move to the next or just press TAB.

Starting with release FF, you can auto-complete the labels (thanks Toms and Cwiiis), with the following caveat:
- The labels are not sorted
- Ghost labels are shown (bug#129)

When you validate a suggested label with -~ (CURSOR RIGHT), the exact label is searched.
Then using TAB from the editor means "repeat last search", it will go to this label definition (rather than iterating labels which share the prefix).

CTRL * (TAB to continue the search)

Go to the next occurrence of the label under cursor (reminiscent VI).
If there is no label under cursor, it takes the first label from this position(*). If there is no such label, it takes the closest to the left.


fx_dispatch call routine; CTRL-* will go to next occurrence of 'fx_dispatch' if the cursor is over
                          ; or next occurrence of 'routine' in all other cases.

(*) Why the label to the right? An article published in "Psychological Science in the Public Interest" shows a tendency to focus on the label following the cursor rather than the one preceding it. This effect is amplified if the cursor is an instruction (e.g. 'call'), but reversed when the guinea pig just read a verse from the Koran.

CTRL Enter

Go to the definition of the label under the cursor.
If the latter is itself a definition, then we take the next.
If there is no label under the cursor, same selection mechanism than CONTROL- *

fx_dispatch call display3D ; CTRL-ENTER go to 'display3D' whatever the cursor position

To try it is to adopt, as we say in Denmark.

CTRL Return

Back to previous position (after Ctrl Enter). It works on 8 levels.
As I am not dog (*), I give you a mnemonic: "Enter routine", "Return from routine."

(*) Not that dogs be petty, but they seldom write user guides.


Cycle through the last 8 rows changed. Very handy, I find, having wandered in the source to return to where it was.


Paste the last line cut with CTRL-DEL. UNDO serves rustic but also allows you to duplicate a line faster than the blocks. In this case, must be deleted to better re-insert!

CTRL F (Find) (TAB to continue the search)

Text Search.

  • Case insensitive
  • Indifferent to the tabs (unlike TurboAss where you have to enter the exact number of spaces to search mnemonics with operands).
  • By default, exhaustive search.
    • Precede the search string with a space if it is desired to search only at the beginning of words.
    • End the search string with a space if one wishes to look only at the end of words.
    • Surround the search string with spaces if you want exact search for the word alone, akin to `\b` in regex syntax.
    • In all these cases, the space at the beginning and at the end is not wanted literally, as shown by the following examples.
V Searched / In source -> OR A LD HL,orang LD HL,Cador LD HL,morel + 4
"Or" * * * *
" Or" * *
"Or " * *
" Or " *

CTRL-% (CTRL-+ on QWERTY) Cycle through IF-ELSE-END MACRO-ENDM [-]

If the cursor points to IF, this shortcut will jump to the corresponding ELSE (if present) or END.
And vice versa.
Same applies to macros and repetitions blocs.
If used on a line without any of those directives, it will jump to the next one of them.

CTRL-B + key: Register bookmark at current line.

A maximum of 16 bookmarks are remember (and stored in .o file).

SHIFT-CONTROL + key: Go to corresponding bookmark.

Then CONTROL-RETURN will return where you were.

Assembler Shortcuts

We reserve CONTROL+number for assembler shortcuts (presumably fewer). This allows:

  1. Expand easier each other without risk of conflict.
  2. From store / used more efficiently (less cognitive interference, to speak like my concierge).

CONTROL-1: Assemble

CONTROL-2: Assemble+jump

CONTROL-4: Go to the next assembly error.

These shortcuts are achievable with one hand, freeing the other to smoke a Cuban.

Program is lunched under DI, standard RAM connection (&C0), SP set to SP C000, with the correct firmware values for AF 'and BC'.


It is pre-assembled.

  • Facilitates the detection of errors when typing.
  • Identification of the label rather than to editing the assembly: essential for super-fast assembly.
  • Much more compact code.
  • Reduces the need for file 'object' (assembled code provided with the information necessary for the relocation and / or to access its routines).
  • There is the possibility to import / export ASCII versions.
  • In the display, imposes indentation (for opcodes / directives). It is also an advantage: uniformity of code without having to format it by hand!

Compiler directives

Org X [,Y]

Or how to generate a Y address of the code to be executed by X.

We distinguish:
* Code pointer (pseudo label $), the address where the code is supposed to run (ie $ almost plays the role of PC)
* Object pointer (pseudo label $$), the address where the code is stored by the assembler.

In the most common scenario, the two coincide, and the ORG X allows to change this address.
Used with two parameters fixed ORG $ and $$ separately.

The following code:

     ORG &1000, &2000
loop JP loop

generate the code C3 00 10 at &2000.

To change $ only:

byte message "Hello dad!", 0
      ORG &C000, $$
     ; the code is stored after post
toto; this label is &C000

To change $$ only:

     ORG $,&4000

Known bugs

Assembled code in &30-&32 is overwritten by JP &BE00 (Breakpoint mechanism).

ALIGN(n) (MACRO example)

Advance $ to the closest multiple of n.
If $ is already a multiple if n, the directive doesn't change anything.

NB: $$ is incremented by the same amount

There is no ALIGN directive. It can be simulated by:

    SKIP -$ MOD n

We want to skip x bytes such that $+x = 0 [n] and 0 <= x < n
This amounts to x = -$ MOD n

ASSERT(predicate) (MACRO example)

Raise an error if predicate isn't true.

There is no ASSERT directive. It can be simulated by:

  MACRO ASSERT predicate
    IF predicate:ELSE
   !! ERROR !!

The '!! ERROR !!' part is not seen at all if predicate is true,
otherwise it triggers an assembly error.

Example 2:

  MACRO ASSERT_NOT predicate
    IF predicate
 !! ERROR !!


  ASSERT_NOT (my_table AND &ff) ; Ensure my_table is &100 aligned


Currently the error line points to the macro definition, not to macro invocation.
That's unconvenient.

LIMIT n (Not here yet!)

No such directive for now, but here is a workaround:

After your code, add:

  SKIP &A000-$  ; To protect overlapping at &A000

If $ if greater than &A000, the value is negative, raise an error.


Directive BANK n re-configures the memory similarly to OUT &7FFF,n.

   BANK &C4
   ORG &3F00
   [...]; &200 bytes of tasty code

generates code both in main RAM and "&C4".


   BANK &C2
   ORG &3F00

generates code both in banks &C4 and &C5.

Known bug!

The connections &C1 and &C3 are poorly managed.


The BYTE command accepts strings and bytes, possibly mixed:

BYTE 12,"ABC",-1 Encodes 0c 41 42 43 ff

A "BYTE" or "WORD" without a value increments pointers ($ and $$) without writing into memory.
Three possible use-cases:

  • Reuse the current value in memory (eg: setting that you want to keep from one run to another, even in the code re-assembling in between).
  • A purpose of documentation, note that the value does not need to be initialized by the assembler (eg: variable anyway initialized at run-time).
  • Easily define positions in a structure. Eg:
         ld a, (ix + pat_flags) Give ld a, (ix + 2)
         ld e, (ix + pat_value) Give ld e, (ix + 3)
         ld d, (ix + pat_value + 1) Give ld d, (ix + 4)

          ORG 0
pat_pos WORD       ; Do not write anything at 0!
pat_flags BYTE
pat_value WORD


Restore the system, the battery, the oil level before returning to the editor.
Thus, no need to save AF 'and BC', the interrupt vector in &38, etc …


Introduced 25 Sept 2021 in FuriousFlags alpha 8.

  • /!\ Warning /!\ This is currently a restricted implementation, where you cannot pass parameters to imported files. That drastically limits the usefulness of import, will be corrected soon.
  • !! Worse !! You cannot yet invoke macro defined in an imported fill (coming soon in FF alpha 9).

So, what use cases are remaining?

  • Creating a library of common and/or fixed routines. That is, code that is unlikely to change and that has a clear interface, akin to firmware’s “vector”.
  • Isolating data that is slow to generate. E.g. 256 ** BYTE SIN(#)….
  • Separating

This directive "imports" lib.o. This is not a literal inclusion of the source like with INCLUDE or READ in some other assemblers, or with the editor's import command (CONTROL-I). This is on purpose!

  • The imported source doesn't see the host's labels. If error-free in isolation, it remains error-free as an import (this good property isn’t guaranteed with textual inclusion).
  • The host will see imported source's labels as expected. The differences are:
        • If the same label exists in both sources, the host’s one takes precedence, shadowing the other one. The rationale is that private labels (that is, global labels in lib not meant to be exposed) shouldn’t impact the host. Also, it helps keeping assembling ultra fast.
        • There will be a way to explicitly access an imported label: lib.label (not implemented yet). This is required to resolve ambiguity when two libs use the same label.
  • If B and C both import D, and A imports B and C (diamond configuration), D will only be included once. No header guard gymnastics to do. Rationale:
      • If D only consists of macros, that doesn’t change anything.
      • If D contains routines, there is no reason to duplicate them.
  • Since the imported file doesn't depend on the host, we only need to assemble it once. Subsequent assemblages will re-use the cached version. To be exact, it depends on the current $ and $$ position. If the import has changed position, it will be re-assembled. That's another reason it's recommended to put all the imports at the beginning of the source and/or after an ORG directive.

As said above, for now it's not possible to pass flags or values to the imported source.
In the future, it will be allowed by explicitly passing parameters:
In lib.o

get dev_mode = 1 ; Parameter with default value
get target ; Mandatory parameter.

In host source, something like
   IMPORT "lib" [set dev_mode 0]

Current workaround: wrap your code in macro, and pass your flags as macro parameters.

In summary, the effects of IMPORT "lib" are:

  • generating code (or not, if IMPORTed source was already met).
  • changing $, $$ accordingly (*)
  • changing bank if the directive is used (in other words, you don't get back the bank selection you had before the import, very much like $ isn't equal to the address before the import !). Not recommended anyway. The main source should be piloting ORG and BANK, not the librairies.

(*) Beware, $ and $$ are either:

  • Unchanged (when IMPORTed source met for the second and subsequent times).
  • The value at the end of IMPORTed source, as if the source was textually inserted. That means the latest values, not the highest (Note: latest = highest anyway if you don't use several ORGs out of order).

In the following situation, the code at &200 will be overwritten without warning (due to bug #a9).

; lib.o
  org &200
 ; [...] some code or table
  org &100
  fill &100,&f7
; $ and $$ = &200 here
; main.o
  IMPORT "lib"
; $ and $$ = &200 here

  call init ; oups, overwrite code &200 from lib.O

Anyway, once again, it is not recommended to use ORG at all in sources meant to be imported.

Imports and editor

All imported sources are loaded in free editor tabs.


  • They must be in memory anyway, we might as well make them visible.
  • In case of errors, give the opportunity to fix them.
  • Allow to use the main source as a multi-sources "project".

Importantly, when changing something in imported source, this in-memory version is used, not the one on disc.

  • That's what we want: changes to be reflected in assembly, without having to save and reload.


  • making lib's label accessible via full qualification: lib.zolabel
  IMPORT "crtc"
  IMPORT "ui" 
  IMPORT "player"

   call init  ; will use `init` below.

; [...]
init   ; global init
   call ui.init  ; no ambiguity 
   call player.init

; [...]
     call player.iter
     crtc.SET(4, 0)  ; Use macro "SET" defined in crtc.o
     call player.psg


  • Automatically detects invalid references
  • No limitation in size (the name of labels being stored once, do not hesitate to choose long)
  • Starts with a letter, followed by letters, numbers, or the following characters "_ #"

Local Labels

Introduced in Fluffy Flages version.

When using a label in a Macro or a repetition block, it leads to double definition error.
Solution: local labels! That's just your regular label prefixed by a dot (.).
You can repeat it, as soon it's in distinct scopes.
Now, what defines a fresh scope ? Well, a macro invocation, a block iteration, or a normal (global) label.
This allows:

 4 ** [
    bit 7,h
    jr z,bit7off  ;
    jr endBit7

ou encore

; Warning, A trashed
        add l:jr nc,.nc:inc h
.nc    ld l,a

As you may have noticed, the dot is optional when the label is referenced.
It is recommend anyway: it's a bit clearer, and avoid clash if a global label with the same name happens to exist.
In summary:

        inc l:jr nz,hu:inc e  ; Ambiguous since global "hu" also exists
        inc l:jr nz,.hu:inc e  ; correct: explicitly local
        inc l:jr nz,ho:inc e. ; correct: no global "ho"

Known limitations.

  • No such thing as local a EQU. Let me know if you need that.
.toto = 10  ; Won't work
  • CONTROL-ENTER in the editor might go to the wrong instance.

Digital Expressions

True management sign

A step = -1 does not encode &ff nor &ffff, but indeed -1. Thus, LD A, step*step passes without worries.
Therefore, an error in an overflow indicates a real problem in your program.

Extended Arithmetic

Labels can take the signed 24-bit values. Authorizes for instance:

ram = &40000; yes, 256k

chunksNb = 16;
chunkSize = ram / chunksNb;

Expressions and temporary results can reach 1024 bits.

   ld, ram*ram / &100000000; ok, it's 16


  • Warning: No operator precedence!

Sequencing is made only by spaces and square brackets.

  • Examples:
     LD,1+2*3; 9 (+ and * made in order)
     LD,1+2 *3; 9
     LD,1+ 2*3; 7 (2 * 3 form an isolated group, calculated separately before adding 1)
     LD 1+[2*3]; 7 Idem, more classic.
  • Simple rule. Nothing to remember. Visually consistent.

Logical operators and modulo in full (and, xor, or, mod)

Easy to remember.


An instruction or a block can be repeated. The advantages are obvious:

  • Compact Plus (to read, write and store)
  • The number of repetitions is controlled by a label.
  • "Defs" on steroids:
     16 ** byte 1,2,3; 1,2,3,1,2,3,1,2,3 ...

For a block repetition, '[' should be on the same line:

; Good
   NbLines ** [
       INC B: OUTI
       INC B: OUTI
       OUT (c),e: OUT (c),d: OUT (c),0
; bad
   NbLines **
     INC B: OUTI
     INC B: OUTI
     OUT (c),e: OUT (c),d: OUT (c),0

Of course, any label within a block repeated more than once will throw an error (double definition).

Expressions are re-evaluated at each iteration. This facilitates the creation of tables:

             ORG &9000
rampe3 256 ** BYTE $*3 and &FF ; stores 00 03 06 09 ... FF 02 05 08 ...


     4 ** LD A, (DE): INC E; buggy. Only 1 E INC
     4 ** [LD A, (DE): INC E]; OK.

It allows up to 8 levels of nesting. For more, request exemption from the nearest clinic.

Pseudo labels #, ##, ###

The pseudo label # is the index to the containing iteration (include from 0).
The pseudo label ## is the index of the iteration at the next level (when several nested repetition, of course).

 2 ** [
 3 ** BYTE &10*## + # ; stores 00 01 02 10 11 12

Separator ":"

  • Used to group instructions that form a logical operation, and to better include the similarities. Eg:
        LD hl, awesomeness: inc (hl)
        LD hl, brightness : inc (hl)
        LD hl, ovlness    : inc (hl)
  • Less line = better overview on a routine.
  • Essential for coming pseudo-labels.

Conditional assembly

Orgams allows 8 levels of IF ELSE END. To consume without moderation, with moderation.


Error messages

(28) Ambiguous local vs global

Happens when a referenced label is both local and global.

        inc l:jr nc,hu:inc h  ; incorrect since global "hu" also exists
        inc l:jr nc,.hu:inc h  ; correct: explicitly local

       jp hu  ; correct since we are not in toto scope anymore.

To fix it:

  • Rename one of them.
  • Or prefix by a dot "." like in the exemple above.

(30) Import: ambiguous label. (Introduced in FF alpha 8)

When a label is missing in host source, Orgams searches for it in imported sources.
If the label exists in several of them, Orgams cannot know which one to pick.

How to fix?

  • For now, you'll have to rename one of them.
  • In the future, you be able to fully qualify the label. E.g:

IMPORT "threeD" ; Contains Init
IMPORT "player" ; Also contains Init

call threeD.Init ; Unambiguous! But doesn't work for now.

Reminder: if a label exist both in host and imported sources, it's not considered ambiguous.
The host label takes precedence.

(31) Ambiguous Import.

When trying to import <filename>, two or more tabs are using the same filename.
Solution: Rename or close the duplicates.

(32) Self Import.

A source is trying to import itself. That's likely a typo.




  • d: Go to the last visited address (pointer >)
  • dnn: Visit from nn.


As d, with the following nuance:

  • t: Go to the next opcode to execute (pointer $)
  • tnn: Change PC ($) in order to trace from nn.


  • This corresponds to the DAMS t command.
  • t is equivalent to d then CLR.

m [nn]

Dump memory from nn. If nothing is specified, use the previous address.
The bytes are displayed in hexa, ASCII and graphically. Useful for checking a curve or detecting anomalies.
Zones at 0 are skipped. Thanks WHO ?

xnn, nn2

Compare two memory areas. Only the differences are displayed.
So, for example, to compare banks & c6 and & c7:

b & c2
x & 8000, & c000

hnn, size

Injects 'size' bytes from 'nn' into the source as BYTE.


Graphic memory dump.
Attention unlike the Hacker, we do not get 1 non-zero byte = 1 pixel put.
The bytes are roughly ORed in packets of 8.

clear [n] / clr [n]

Filled the first 128 kb with n or 0, regardless of the current connection.

clearpage [n] / clp [n]

Fill the connected page (64 ko) with n or 0.


When Orgasm is invoked and at each assembly, a breakpoint routine is installed at &BE00, and a jump to &BE00 at &30.


Use RST 6 to jump into the monitor in trace mode. Identically, the pseudo BRK instruction (CTRL + SPACE) in the source simply puts RST 6, which jumps to &be00.
Using &BE00 directly facilitates conditional breakpoint. Eg:

break = &BE00
    LD (pattern_pos) ; Must be non-zero
    OR a
    CALL z,break; Investigate!

All registers at the time of the judgment are preserved. The only destructive manipulation is writing two words in the stack:

  1. The return address placed by the RST (or CALL &BE00), necessary to know the current PC.
  2. One PUSH AF necessary to probe IFF (state EI / DI).

In conventional use, this does not disturb the return to the program (like an interrupt). But if SP pointed to a table, it will take account of the corruption of 4 bytes.


The BRK directive add an opcode.

The RST 6 is inserted (rather than e.g. silently replacing the following instruction).
It can be an issue for tightly aligned routines we want to break into.

Without Multiface 2, bank detection isn't 100% guaranted.

Without Multiface 2, special connections (C1, C2, C3, C9…) aren't detected, and standard connections (C4, C5, C6, C7, CC ….) might be erroneous, especially if the content in the banks is similar.

Without Multiface 2, ROM paging detection isn't 100% guaranted.

ROM number detection isn't 100% guaranted.

The assembler doesn't verify the usage validity.

Example of failure:

   LD BC, &7fC2
   OUT (C), C
   BRK; Crash !! Jumps to &30 in the wrong bank.

Make sure you restore the RAM banks before using BRK!

Soft BreakPoint.

Using the quick trace allows to break without modifying the code (in ROM for example).

  • $start to place the PC on the code to be executed. Leave the trace with ESC.
  • dbreak to move to the breakpoint address, then space or t (quick trace until $ reaches >).

For example, if the entry of your program is &9000 and you want to interrupt at &BB5A:

  • $&9000, ESC, d&bb5A, space.

For this case it may be more efficient to modify the source:

   ORG &9000:ENT $

But then &BB5A is trashed!

Debugger / disassembler.

There are two types of navigation, with or without execution. Besides, the command 'of the monitor can be understood (D)ebug or (D)isassemble.

Without enforcement, this is a classic disassembly, but with all the flexibility Orgams (kiss your lap):

  • Up and down scrolling (fast with CONTROL)
  • CONTROL-ENTER and RETURN like the editor to explore subroutines (and back!) Without having to enter a single address.

With execution include all the power of a debugger step by step:

  • (S)tep executes an instruction. If there is a call, it enters the subroutine.
  • (N)ext running in fast mode
  • Sometimes you neither want to trace nor execute a routine (eg &BB06). In this case, simply move the pointer to the next instruction, and fingers enable CONTROL-G. This changes PC ($) without executing anything.
  • To avoid having to painstakingly follow the N iterations of a loop, it will place the cursor appropriately, and then build on T or SPACE: the loop will be well executed, but in fast mode.


  • The bit 3 & 5 of the register F are not reproduced correctly.
  • Interruptions are not emulated.

Source/Memory visualization

Orgams try to find source line matching the opcode under cursor '>'. It can fail, rightly or wrongly, thus displaying "source not found". This feature slow down step-by-step trace, mostly in case of failure.
Kindly press CONTROL-V switch to memory dump, to last address selected via 'M' command in monitor. This mode doesn't penalize trace's speed.

Back to source

CONTROL-ESC returns to editor at currently visualized line, contrarily to ESC-ESC which leaves editor cursor untouched.

Back to the program.

Return by pressing J (like 'J'UMP). The state of the Z80 is restored: the point of view of registers it is as if we had called the routine PUSH AF: POP AF: RET. Of course if the program was not traced by step was taken with the current values of registers.
The CRTC registers are restored with the system defaults.

Known Issues

These are the known issues for FluffyFlags releases.
For new bugs in current GG beta, see Regression section in working.
Go to bugs for complete list and description.

#15e [monogams/rsx/minor] load"toto" isn't recognize as rsx. The "|" is mandatory.
#15d [rsx] Lot of stack used to pass parameters.
#15a [minor] Internal assertions may leave Orgams in a bad state.
#155 [MOD Arithmetic] -4 % 4 gives -4 instead of 0.
#143 [Assembleur] LOAD directive: header-less files should be loaded.
#141 [Assembling] "reload" (CONTROL-5) not taken into account if source not changed.
#13e [Import] Filename comparison is too strict (not Amsdos compatible).
#13d [Editor] Tab in invalid state after import not found.
#13b [Monogams] Artefacts when label not found in current source.
#133 [Import] Far too slow when imported file has many labels.
#12C [Editor/Auto-completion] Doesn't respect "."
#129 [Editor/Auto-completion] Ghosts labels are shown.
#124 [Editor] CONTROL-L: explicit search for local label doesn't work.
#123 [Editor] Space doesn't come back to editor (random).
#122 [Debugger] Visu source lost with 2 levels of repeats.
#121 [Display] Lines > 72 characters with errors wrongly display at assembly time.
#120 [Editor/Monogams] Tilde '~' cannot be entered on Azerty keyboard.
#11f [Disassembly] Instructions eaten after invalid &dd / &fd.
#11c CONTROL-I may hang when EOF (&1A) absent. (Fixed in FF alpha 7)
#118 [Editeur] Memory full not handled when switching to a new tab.
#114 [Editeur] Wrong state after memory full while opening source.
#113 [Assembler] Too many nested if: neither handled nor detected.
#110 [Monogams] Type command doesn't handle line bigger than 80 charaters.
#106 BRK doesn't always capture right bank connection (MMR) *without Multiface Two.
#105b BRK doesn't capture ROM connection selected via VGA.
#FA [Editor] Import may hang/crash when trying to import raw binary instead of text.
#F8 [Editor] Modification lost when reset.
#F5 [Editor] CONTROL-ENTER does not work for macro parameters
#EF [Monagams] Wrong evaluation order of parameters.
#E6 [Assembler/Begnin] Cannot jr from FFxx to 00xx
#E2 [Orgams] Breakpoint reinstalled when met or when returning from exec.
#E1 [Assembler] Should warn in case of memory conflict.
#DA [Orgams] Bank directive not honored for binary save.
#D4 [Assembler] MACRO/ENDM and IF/END interleaving not supported.
#C2 [Assemblage] Code assembled in &BFCC-&BFFF is overwritten.
#BB [Editeur] very low priority 'DEFS n' from DAMS source not correctly imported.
#BA [Editeur] Wrong error message instead of "memory full" when importing.
#B7 [Trace] $ not shown if in middle of instruction.
#A9 [Assembler] Overlapping output should raise error.
#A6 [Editor/Parser] Parsing too slow
#A4 [Trace] When entering trace screen, navigation history shouldn't be reset.
#94 [Assembleur] Code assemblé écrasé par gestion Jump.
#93 [Assembleur] Code assemblé écrasé par breakpoint.
#90 [Orgams] |O,"file" ne marche pas quand HIMEM en &4000-&7FFF
#7F [Trace/Madram] Get_line_from_pc for source visu still not fast enough.
#6E [Trace/Madram] Remontée navigation (flèche HAUT) est trop lente.
#5D [Assembleur] 'IF' consistency not checked.
#4D [Trace] CONTROL-ENTER not working for RSTs
#44 [Monogams] ?1/0 renvoie #1ff
#2E [Assembleur/Madram] Sous bug : Numéro lignes erratiques quand erreur fatale.
#1B [Madram] Parser. LD A,1 + 1 not recognized.
#15 [Assembleur] BUG LD A,"AA"
#05 [Trace] Crash when tracing trace


Orgams is very stable, used several hours per day. Alas, in case of memory corruption, there is little it can do.
Actually, there is something: a lot of assumptions are checked (aka "asserts"), so, instead of crashing, it would switch to the debugger pointing to the unexpected failure.
If it ever happens, here is what to do:

  • Take a screenshot (for bug report!)
  • Reset
  • |o
  • Save your source(s).
  • Export them as well (it would be easier to recover if the source itself is corrupted)
  • |orgams to start afresh.

Under the hood

Design choice regrets

!+++ Using two phases

* Lot of duplicated code (dispatch).
* Almost twice as slow!

!+++ Source encoding: Not making newlines implicit

* Since a lot of opcode takes 1 byte (xor a, inc c, …), that almost double the source footprint! For a 4000 lines source, that's almost 2k lost (~ 5%)
* Assembly have to go through them (~10 us per NL. That's 40ms for 4000 lines).

!+++ Putting all the variables in bank.

For instance, the current source pointer (including the bank #) is stored in the base bk.
So, to read some variables, we must remember the current bk, connect bank bk, read and restore bk.

The rationale was: for the trace (Z80 emulation) to be fast enough, all RAM is mapped as in the executed program.
That means there is no place in RAM to store any variable, no ability to call or push.
The upside is that any write to 0000-FFFF can be done directly (e.g. the emulation of push is just the instruction itself).
Any read 0000-BFFF can also be done as-is. The only issue is the page C000 since the ROM is connected.

So far so good. The tragic error was to maintain this principle (all user RAM connected) to the debugger UI and assembler as well.
It proved simpler and almost unnoticeable to reserve an area for variables and stack, and swap back to user memory just before calling the Z80 emulation.
That's what is done in recent versions, yet cannot be used to its full potential: some variables are hard to move, and bank gymnastic is still needed.

Of course, variables meant to be persistent must remain in bank anyway.

Sauf mention contraire, le contenu de cette page est protégé par la licence Creative Commons Attribution-ShareAlike 3.0 License