SjASMPlus 1.20.3 Documentation [2023-06-23]

1. Introduction
What is it?
Main Features
What's new?
2. Where to get and how to use
Command line
Warnings with id
Source file format
3. Labels
Local labels
@ Labels
Temporary labels
4. Constants, expressions and other features
Numeric constants
Character and string constants
Assembly language
Fake instructions
Real device emulation mode
Predefined defines
5. Pseudo-ops (aka Pseudo-instructions, Directives etc)
Simple example of usage
Conditional assembly
6. Structures
What is it?
Defining structure
STRUCT instructions
Usage of defined structure
7. Lua scripting
How to use?
SjASMPlus functions bindings to Lua
Third-party embedded library(ies)
8. SAVENEX guide
NEX File Format
Detailed description of each SAVENEX command
9. Source Level Debugging (SLD) data
What is it?
How to write "non tricky" source
SLD File Format definition (version "1")

Chapter 1. Introduction


SjASMPlus is licensed under BSD license.

What is it?

SjASMPlus is Z80 Assembly Language Cross Compiler. It is available for Win32, Linux and FreeBSD (mainly 5.x) systems. It is based on SjASM source code by Sjoerd Mastijn (

Main Features

  • Full source of assembler available under BSD license, modify and extend as you wish

  • Z80/R800/Z80N/i8080/LR35902 documented and undocumented opcodes support

  • Macro language, defines, array of defines

  • Built-in Lua scripting engine

  • Conditional assembly, block repeating

  • Modules (namespaces), local and temporary labels

  • Source and binary file inclusion, include paths

  • Multi file output, file updating, various types of exports

  • Structures to work easily with structured data in memory

  • Relocation data generator to support SymbOS-like relocation of executables

  • Virtual device mode for common machines: ZX 128, ZX Next, Amstrad CPC, … (pseudo op DEVICE)

  • ZX Spectrum specific directives and pseudo ops (SAVESNA, SAVETAP, SAVEHOB, INCHOB, INCTRD…)

  • ZX Spectrum Next specific features and directives (Z80N, 8ki memory paging, SAVENEX)

  • Amstrad CPC 464/6128 specific directives (SAVECPCSNA)

  • Correctness is assured by Cirrus-CI with 400+ automated tests (that's also 400+ examples of usage!)

  • Fake instructions as LD HL,DE (LD H,D:LD L,E) and more

  • Code inlining through colon (LD A,C:INC A:PUSH AF:IFDEF FX:LD A,D:ENDIF…)

  • Very fast compilation: 1 million lines by 2-3 seconds on modern computer

  • Multiline block comments and user’s messages


Special thanks to Sjoerd Mastijn, the author of SjASM.

Aprisobal - main programming, documentation, etc.

Thanks to:

  • Kurles/HS/CPU, Alexander Kovalenko, Ped7g, Dart Alver, Oli Wilkinson - additional programming;

  • Krystian Wlosek <kwlosek(at)> - bug fix patches, Linux makefile;

  • Ric Horne <> - bug fix patches.

  • breeze <> - bug fix patches.

  • psndcj <> - bug reporting, beta-testing.

  • elfh <> - bug reporting.

  • bugsy <> - bug reporting.

  • skrju <> - bug reporting.

  • Tygrys, UB880D, Cizo, mborik, z00m, otis - compilation errors and warnings clean up, makefiles, testing.

  • Antipod, boo_boo, PulkoMandy, Busy, Liniya, LVD - bug fix patches, testing.

  • CKirby - SLD export and support in his tools.

  • Simon Brattel - for creating mother of all assemblers Zeus and bringing all the good and peace to the mankind, giving infinite inspiration to sjasmplus contributors who stole his tremendously amazing features without even crediting him - till now, fixed (not sure what the particular list of features was copied from Zeus, I guess DG/DH, I believe some other similarities are rather copied from the ALASM which had many features before Zeus, but who knows, too late to track it all precisely, what happened in 90's and when).

Big thanks to all people, who helped on development of the compiler!


WWW: (newer versions maintained by z00m and others)

WWW: (original Aprisobal's source)


What's new?

23.6.2023 - 1.20.3
- added alias `--define` for `-D`
- added string-literals suffixes Z and C to add zero or set high bit of last char
- end of line backslash continues source line (limited support, not recommended)
- Lua: minor version upgrade to 5.4.6 (from 5.4.4)
- minor updates to Makefile
14.2.2023 - 1.20.2
- added optional second argument for `DUP` to have index variable
- option `--exp` will create file even when no `EXPORT` is used
- fixing variable name-clash when compiling against musl-clib
- LuaBridge updated, CMake and Makefile updated a bit
- minor bugfixes/improvements in parser in specific edge cases
29.7.2022 - 1.20.1
- parse decimal numeric constants with warning (for easier Lua 5.4 life)
- added `SAVEAMSDOS` (like SAVEBIN with AMSDOS header)
- added "smart" SMC offset syntax for self-modify-code labels: `abc+*: or 123`
- added `DEFDEVICE` to define custom devices
- Makefile cleanup
13.6.2022 - 1.20.0
- Lua: upgrade to 5.4, replacing tolua++ bindings with LuaBridge2.6 library, extending some bindings
- Lua: bindings slightly modified (required by upgrade), refreshed docs, added test coverage
- Lua: the 3rd party extensions (BTW not working for many years) are obsolete in 5.4 and removed
- Lua: more accurate errors/warning location reported even in complex cases
- warnings: added -Wall, --help=warnings shows on/off status, rdlow off by default
- Added HIGH mode to relocation data generator (MSB-only relocation mode)
- many open-file "fatal" errors become "non-fatal", assembling will continue
- deprecated features removed: --syntax=m, label `abs` in expressions
- `--color=auto` will stay no-color when env.var. `NO_COLOR` is defined
- refactorings, improving some error messages and parsing, small fixes in parsing logic
- fix listing of Lua's sj.parse_code (eol-comments), minor memory leaks fixed
- fix relocation of temporary labels in expressions
- invalid CLI options are reported as regular errors (also changing exit code)
- errors are colored similarly to gcc (only keyword has color), console input name is `<stdin>`
31.3.2022 - 1.19.0
- added Amstrad CPC devices ("AMSTRADCPC464", "AMSTRADCPC6128") - by Oli Wilkinson
- added Amstrad CPC save snapshot and CDT (`SAVECPCSNA`, `SAVECDT`) - by Oli Wilkinson
- added `SAVE3DOS` (like SAVEBIN with +3DOS header)
- the deprecated "ok" warning suppression is removed, use "<warning-id>-ok" comment or -Wno-...
- new temporary label suffix syntax "_b" and "_f", enabling them for all expressions
- fix `--longptr` mode to keep 32b address when `DS 0` is used
- added fake instructions adc|add|sbc|sub de,bc|de|hl|sp
- dec|inc|pop|push will accept also single-comma multiarg in --syntax=a mode
- DUP/REPT will now accept also zero count (skipping the block)
- DEFL labels can be defined even as late as in last pass
- bugfixes (macros, listing, file names in errors, SLD reversepop data)
29.8.2021 - 1.18.3
- added `--color` to enable/disable ANSI coloring of errors/warnings
- added `--syntax=s` mode to disable sub-word substitutions of DEFINEs
- added at-sign prefix for macro local labels to act as non-macro local label
- `SAVETRD` accepts names containing dot ("z.x.B" is "z.x" with extension "B") - by Dart Alver
- `SAVETRD` has optional argument to save BASIC with variables (length_minus_variables)
- minor bugfixes (conditional block parser)
14.2.2021 - 1.18.2
- [may break old sources] new exist operator to check label existence
- the --syntax=i mode makes now also register parsing case insensitive
- minor bugfixes (predefined values, savenex BMP loader less strict about "colors used" content)
23.1.2021 - 1.18.1
- Big-Endian hosts support (experimental and not tested continuously)
- added "listall", "listact" commands to OPT - to switch between listing types
- `WHILE` has optional argument to set explicit guardian-counter
- `ASSERT` has optional argument (to add description/notes for expression)
- `SLOT` and `MMU` will now accept also starting address of slot instead of its number
- fix: option --sym was not exporting labels starting with underscore
- fix: SAVENEX BMP-loader bug when certain builds of sjasmplus were unable to open BMP files
- fix: after STRUCT instance the "main" label is not polluted by last field of STRUCT
- minor bugfixes in parser, windows cmake-builds have now icon
- docs: adding "Index" section
- docs: adding some missing information (__DATE__, __TIME__), fixing HTML anchor names
12.12.2020 - 1.18.0
- [may break old sources] the colon between end of EQU/DEFL/= expression and instruction is mandatory
- [may break old sources] new abs operator for absolute integer value
- new system of warnings (and suppression), the "; ok" comments are now deprecated
- `DISPLAY` has now also binary and char formatting
- `DEFINE+` added to [re]define identifier without error
- `ELSEIF` added to conditional assembling arsenal
- `WHILE` added for conditional loops
- added "NOSLOT64K" device with 2MiB of virtual memory
- `LABELSLIST` has new optional argument to dump 16bit "virtual labels"
- `CSPECTMAP` exports STRUCT symbols with more detail (instance labels with physical address)
- SMC offset syntax for self-modify-code labels for source brevity
- added exclamation mark prefix for labels to not affect following local labels
- added "listmc" command to OPT - to list only lines with machine code bytes
- added `--lstlab=sort` variant to have symbols in listings in predictable order
- minor bugfixes in parser and listing-line-numbering, refactored symbols/labels implementation
- Added example (chargfx2asm) how to use sjasmplus as byte-processor for binary files
- SLD data improvements based on Maziac's feedback and DeZog's needs
8.8.2020 - 1.17.0
- `STRUCT` has new `TEXT` pseudo-instruction to define "DB-like" data (Issue #93)
- `STRUCT` initializer block can be now multi-line (when correctly enclosed in curly braces)
- `EQU` now allows for optional override of page number assigned to the new symbol
- new $$$ and $$$$ operators to retrieve "physical" address/page inside DISP block
- instruction `out (c),0` now emits warning (can be suppressed by the "; ok" comment)
- fixed listing of structures using long BLOCK fields (machine code was correct, but listing not)
- fixed some memory leaks, undefined behaviour and unaligned memory access
27.7.2020 - 1.16.0
- `LUA` the new emit warning (v1.15.1) is now suppressible
- Predefined defines extended and renamed (following gcc/clang ones)
- Added relocation data generator (similar to WinAPE feature), check also example
- bugfixes/improvements in parser like: operators `not`, `low`, `high` can be followed also by "("

See for full list of changes.

Chapter 2. Where to get and how to use


The latest release of this sjasmplus variant is available at - there is available zip archive with windows binary (toward bottom of the page), and zip archives with full source code of the sjasmplus.

Win32 package has:

  • sjasmplus.exe - the Win32 executable.

  • examples directory - some examples of use

  • documentation directory - documentation in various formats

You may want to download also the full source package, as it contains more than 400 automated tests used to verify correctness of executable, and these can be often helpful to better understand how specific feature of sjasmplus works, and how to use it effectively.

Linux, Unix, MacOS and BSD version can be built from the full source archive. You can compile it using GCC and included Makefile. There is an option to use CMake for compilation. See for details.

Windows binaries are compiled with MinGW environment and included file.

You can grab older (up to v1.07) binaries and sources at SourceForge project page:

Command line


sjasmplus [options] sourcefile(s)

Option flags as follows:

  -h or --help[=warnings]  Help information
  --version                Basic info (with --nologo only raw version string)
  --zxnext[=cspect]        Enable ZX Next Z80 extensions (CSpect emulator has
  extra "exit" DD00 and "break" DD01 fake instructions)
  --i8080                  Limit valid instructions to i8080 only (+ no fakes)
  --lr35902                Sharp LR35902/SM83 CPU instructions mode (+ no fakes)
  --outprefix=<path>       Prefix for save/output/.. filenames in directives
  Note: if prefix is folder, it must already exist and add trailing slash. Prefix
        is applied only to filenames defined in source (not to CLI arguments).
  -i<path> or -I<path> or --inc=<path> ( --inc without "=" to empty the list)
                           Include path (later defined have higher priority)
  --lst[=<filename>]       Save listing to <filename> (<source>.lst is default)
  --lstlab[=sort]          Append [sorted] symbol table to listing
  --sym=<filename>         Save symbol table to <filename>
  --exp=<filename>         Save exports to <filename> (see EXPORT pseudo-op)
  --raw=<filename>         Machine code saved also to <filename> (- is STDOUT)
  --sld[=<filename>]       Save Source Level Debugging data to <filename>
                           Default name is: "<first input filename>.sld.txt"
  Note: use OUTPUT,LUA/ENDLUA and other pseudo-ops to control output
  --nologo                 Do not show startup message
  --msg=[all|war|err|none|lst|lstlab]  Stderr messages verbosity ("all" is default)
  Note: "lst" or "lstlab" will turn STDERR into listing file (this will clash with
  `--lst`, use only one of the options) and some diagnostic messages of "all"
  are omitted, as they are not part of listing files. "lstlab" does sort symbols.
  --fullpath               Show full path to file in errors
  Note: the "fullpath" starts from current working directory, not from file system
		root (the MS_VS build was fixed to behave this way since v1.15.0)
  --color=[on|off|auto]    Enable or disable ANSI coloring of warnings/errors
  Note: auto detection checks for existence of environment variable TERM, and its
		content is scanned for sub-string "color" (like "xterm-256color")
  -D<NAME>[=<value>] or --define <NAME>[=<value>]
                           Define <NAME> as <value> (--define since v1.20.3)
  -W[no-]<warning_id>      Disables/enabled particular warning type
  -                        Reads STDIN as source (even in between regular files)
  --longptr                No device: program counter $ can go beyond 0x10000
  --reversepop             Enable reverse POP order (as in base SjASM version)
  --dirbol                 Enable directives processing from the beginning of line
  --nofakes                Disable fake instructions (same as --syntax=F)
  --dos866                 Encode from Windows codepage to DOS 866 (Cyrillic)
  --syntax=<...>           Adjust parsing syntax, read details below.

The option --lstlab=sort will cause also other symbol dumps being sorted (along listing): --sym, CSPECTMAP and LABELSLIST. The --msg=lstlab sets always sorting "on" (no "unsorted" option).

Value for --syntax option may consist of multiple letters, omitting letter for particular feature will use the default setting:

  a      Multi-argument delimiter ",," (default is ",") (but dec|inc|pop|push accept also ",")
  b      Whole expression parentheses are legal for memory access only (default = immediate or memory)
  B      memory access brackets [] required (default = relaxed syntax, [] allowed as extra)
  f F    Fake instructions: warning / disabled (default = enabled)
  i      Case insensitive instructions/directives (default = same case required)
† l L    Keyword labels (registers, instructions, ...): warning / error (default = silent)
  M      Alias "m" and "M" for "(hl)" to cover 8080-like syntax: ADD A,M
  s      Use only "simple" whole-word substitutions, no sub-words (since v1.18.3)
  w      Warnings option: report warnings as errors
  m      *REMOVED* in v1.20.0, use -Wno-rdlow

† work in progress: options "l" and "L" are not implemented yet, following example is then not working correctly either.

I.e. --syntax=faBil will modify parser to process source line

hl:  Ld a,(hl),,de,hl

in a way to produce warnings about keyword "hl" being used for label, about fake instruction being used (ld de,hl) and assemble (hl) as numeric expression, not memory access. Warnings on fake instructions can be suppressed for particular line by adding any end-of-line comment containing string "fake", i.e. "ld de,hl ; fake DE=HL" will assemble without warning. The "F" option is identical to "--nofakes" and preferred.

The recommended setup for new projects is --syntax=abfw which makes syntax less relaxed, so some typos and mistakes are easier to catch, for example:

        OPT reset --syntax=abfw
label:  dw 15
        ld b,(label)
        sub a,b

will produce "error: Illegal instruction (can't access memory): (label)" message for the ld b,(label) and the sub a,b will produce only the sub b instruction (to give the sub multi-argument with syntax option "a" the line would have to be sub a,,b).

The assembler will also read the environment variable SJASMPLUSOPTS (if available), and process its content as part of command line options (before the actual options), so you can pre-configure certain options in your environment, for example:

export SJASMPLUSOPTS="--zxnext=cspect --msg=war"
sjasmplus --lst --lstlab example.asm

will execute the assembling as if command line "sjasmplus --zxnext=cspect --msg=war --lst --lstlab example.asm" was used. Known issue: parser of environment variable delimits each option on any white-space character, so option containing space character will be incorrectly parsed (like "-Ifile-path with space" = fails and there is no way to escape/quote the path in the SJASMPLUSOPTS variable to make it work).

Warnings with id

Some warnings have own "id" - these can be switched off or suppressed for particular line. The line suppression mechanism requires you to add end-of-line comment with "id-ok" string somewhere in the comment (exception: for fake instructions and --syntax=f mode the word "fake" is enough without -ok suffix) (the "ok" comment suppression mechanism from previous versions of sjasmplus is removed since v1.19.0).

The id-warning is emitted together with its id in brackets:

    ld a,(10-2)    ; read memory address 8 - emits warning "rdlow"
    ld a,(10-2)    ; same instruction, but suppress warning: rdlow-ok

file.asm(1): warning[rdlow]: Reading memory at low address: 8

To list all warning-ids, their on/off state and short description of each use:

<command line>$ sjasmplus --help=warnings

Most of the extra warnings are by default enabled, but to enable/disable all of them you can use (also in source with OPT directive):

<command line>$ sjasmplus -Wall
<command line>$ sjasmplus -Wno-all

To disable particular warning globally you can use option -Wno-<id>, either on command line, or from source by using OPT:

<command line>$ sjasmplus -Wno-out0

; in source code:
    OPT -Wno-trdext -Wno-trdext3 ; disable "trdext" and "trdext3" warnings
        SAVETRD "image.trd", "file.txt", 30000, 100
    OPT -Wtrdext3                ; re-enable "trdext3" warning for following code
; ...

Warnings without id can't be suppressed and generally should happen only when the source code has actual bug (in our opinion) and can be modified to fix the warning, or when you are hitting limits of sjasmplus architecture and/or implementation.

If you believe you are hitting warning without id on correct code, which can be safely ignored (the machine code produced is stable and as expected), open issue on github with example source, so we can assign the warning new id (or explain why it is problematic :) ).

Warnings with id usually signal bug in the code (in common scenarios), but may be misleading in more specialized cases - the decision is left to the author, whether to modify the code to avoid the warning or suppress it and use the resulting machine code "as is".

Source file format

Lines in the source file should have the following form:

Label Operator Operand Comment

Lines without label must start with whitespace. All fields are optional. End of line backslash continues source line (since 1.20.3, limited support, not recommended). Operators and operands can be inlined with colon:

      Operator Operand:Operator Operand:Operator Operand... Comment

Comments should start with ';' or '//'. Comment blocks start with '/*' and end with '*/' (work in "nested" way, i.e. comment block started inside comment block must be also ended, before main block ends).

Example 2.1. 

; comment
// comment
 ld /* comment */ a,80
 comment /* nested comment block */
 ld /*
 but this won't be ld a,3!
 */ a,3

Chapter 3. Labels


Labels are case-sensitive and may be of any reasonable length, that is: up to about 70 characters. Label definitions must start on the beginning of a line, but don't have to be followed by a colon ':'. Generally labels should start with a letter or a underscore ('_'), the following characters may be chosen from letters, numbers and the following special symbols: '_', '.', '!', '?', '#' and '@'. Note that the '.' has special meaning, as it is used between module names, labels and local labels. The following are all legal and distinct labels:


It is possible to use mnemonics, pseudo-ops and register names as labels but it is not advised to do so (expression operator keyword used as label will emit warning, and is also not advised). Also note that the identifiers defined with the DEFINE pseudo-op use another name space.

(since v1.18.0) Right after label (before the optional colon) you can write extra "SMC offset" (Self Modify Code offset) in the form of plus sign and single decimal digit. The defined label will become equal to current address plus the SMC offset:

    ; SMC-offset syntax is very strict, no spacing, single '+', single digit
answer+1:   ld      a,13        ; "answer" points second opcode byte
    ; this helps to keep the SMC-label together with the target instruction
            ld      hl,answer
            ld      (hl),42     ; fix immediate in `ld a,*` to correct answer
    ; compare with classic syntax:
classic equ $+1 : ld a,7        ; often even split to two lines

(since v1.20.1) Smart SMC labels: you can write "plus asterisk" to let sjasmplus adjust self-modify label by most likely offset for following instruction. This works only for instructions with "obvious" target immediate value:

    ; smart-SMC syntax is very strict, no spacing, single '+', single '*'
answer+*:   ld      a,13        ; "answer" points second opcode byte
    ; this helps to keep the SMC-label together with the target instruction
            ld      hl,answer
            ld      (hl),42     ; fix immediate in `ld a,*` to correct answer
    ; few more examples which byte the "smart SMC" will target in the opcode
jptarget+*: jp      nz,12345    ; "jptarget" points at second opcode byte (word value 12345)
ixvalue+*:  ld      ix,12345    ; "ixvalue" points at third opcode byte (word value 12345)
nxtrval+*:  nextreg $44,123     ; "nxtrval" points at fourth opcode byte (byte value 123)
    ; unsupported situations, will cause error
invalid1+*: inc     (ix+123)    ; IXY displacements are not considered (not significant enough)
invalid2+*: cpl : ld a,14       ; only first instruction after label is considered

Local labels

When there is a module definition all labels (except those starting with a '@') are local to that module. To use a label from outside the module use modulename.labelname, in this example: 'vdp.Cls'. Labels starting with a '.' are also local to the previous non-local label.

(since v1.18.0) If you start the label with an ! (exclamation mark), it will not affect following local labels (same syntax as sjasm).

(since v1.18.3) Local label inside macro prefixed with @ (at sign) acts as regular local label, otherwise (older behaviour) the local label is local to each macro's instance.

Example 3.1. docs_examples/s_local_labels.asm

    MODULE main             ; module "main"
Main:                       ; main.Main
        CALL SetScreen      ; SetScreen
        CALL vdp.Cls        ; main.vdp.Cls
.loop:                      ; main.Main.loop
        LD A,(.event)       ; main.Main.event
        CALL ProcessEvent   ; label not found: main.ProcessEvent
        DJNZ .loop          ; main.Main.loop

        MODULE vdp          ; module "main.vdp"
@SetScreen:                 ; SetScreen
.loop:                      ; main.vdp.SetScreen.loop
Cls:                        ; main.vdp.Cls
!KeepClsForLocal:           ; main.vdp.KeepClsForLocal (since v1.18.0)
.loop:      DJNZ .loop      ; main.vdp.Cls.loop

Main.event DB 0             ; main.Main.event

Example 3.2. local labels inside macro

  macro test
.kip0       ; this is local to every instance of macro
@.kip1      ; syntax added in v1.18.3

  module main
hoi test    ; produce labels: main.hoi, 0>kip0, main.hoi.kip1

@ Labels

Labels starting with a '@' are not touched by the label processing and used 'as-is'. See 'SetScreen' in the previous example code.

Example 3.3. docs_examples/s_at_labels.asm

    MODULE xxx
Label      ; xxx.Label
.Local     ; xxx.Label.Local
@Label     ; Label
.Local     ; xxx.Label.Local => duplicate label error
@Label2    ; Label2
.Local     ; xxx.Label2.Local
@yyy.Local ; yyy.Local
yyy.Local  ; xxx.yyy.Local

Temporary labels

To keep the number of used labels reasonable it is possible to use numbers as labels. These labels can only be used as labels to jump to. To jump to these labels, use the number followed by an 'F' for forward branches or a 'B' for backward branches. Temporary labels should be defined in the same order during every pass of assembling, but they can be used within macro, or repeating blocks (old sjasmplus versions didn't allow usage within macro).

Since v1.19.0 there is alternative syntax using underscore before 'B' and 'F', this alternative underscored suffix enables temporary labels also for regular instructions (in all expressions, same way as other labels).

Example 3.4. docs_examples/s_temp_labels.asm

        ADD A,E
        JR NC,1F
        INC D
1       LD E,A
2       LD B,4
        LD A,(DE)
        OUT (152),A
        DJNZ 2B

        MACRO zing
            DUP 2
                JR 1F
1               DJNZ    1B

        .4 zing

        ; since v1.19.0
        ld hl,1B        ; *old* HL = binary 1, NOT temporary label
        ld hl,1_B       ; *new* HL = previous temporary label 1
        ld hl,1_B*3     ; *new* works also in expressions

Chapter 4. Constants, expressions and other features

Numeric constants

Numeric constants should always start with a digit or $, # or %. The following formats are supported:

12     decimal
12d    decimal
0xc    hexadecimal
$c     hexadecimal
#c     hexadecimal
0ch    hexadecimal
0b1100 binary (v1.12.1)
%1100  binary
1100b  binary
0q14   octal (v1.12.1)
14q    octal
14o    octal

(v1.12.1) Optional single quotes(') may be inserted between the digits as a separator (example: ld a,%11'01'11'00 ). They are ignored by the assembler.

(v1.20.1) decimal parts are now parsed and ignored (with optional warning), instead of causing syntax error - this should make life easier when formatting values from Lua script. It is also possible to fix all Lua scripts to produce only integer values, but if you prefer to silently ignore the extra decimal part, you can.

Character and string constants

Character constants are characters surrounded by single quotes. It is possible to use double quotes in some cases, but in general it is better to use single quotes. String constants are characters surrounded by double quotes. When double quotes are used, the following escape sequences are recognized:

\\ 92
\? 63
\' 39
\" 34
\0 0     ; since v1.11
\A 7
\B 8
\D 127
\E 27
\F 12
\N 10
\R 13
\T 9
\V 11

Inside single quotes two quotes after each other are parsed as the apostrophe itself (since v1.11).

Adding Z or C after quotes/apostrophes will make string zero-terminated or set top bit of last character (since v1.20.3).

Example 4.1. 

    BYTE "stringconstant\n" ; escape sequence assembles to newline
    BYTE 'stringconstant\n' ; \n assembles literally as two bytes: '\', 'n'
    LD HL,'hl'  ; hl = 0x686C = 'l', 'h'
    LD HL,"hl"  ; hl = 0x686C = 'l', 'h'
    LD A,"7"    ; not recommended (but works)
    LD A,'8'    ; recommended
    LD A,'\E'   ; warning + truncating value to 'E' (0x45)
    LD A,'"'    ; A = 0x22
    LD A,"'"    ; A = 0x27
    LD A,''''   ; A = 0x27 ; since v1.11
    BYTE "AB"Z, "CD"C, 'E'Z, 'F'C ; hex: 41 42 00 43 C4 45 00 C6


Expressions are evaluated in signed 32 bits in this version of SjASMPlus (unless explicitly specified as unsigned, like ">>>" operator), so intermediate values have range -2147483648 to +2147483647. True for boolean results is equal to value -1, false is equal to 0 (so result of compare or logical operators can be used as all-bit-mask further).

'$' represents the current program counter. '$$' represents the current page in the current slot in the real device emulation mode, '$$label' evaluates to number of page where the "label" was defined (only regular labels have meaningful value, labels defined under DISP mode may return the page specified in DISP itself, EQU/DEFL/... will produce mostly irrelevant values). '$$$' and '$$$$' can be used inside DISP block to retrieve the "physical" memory address and page (where the displaced machine code is written to). '{address}' can be used to read WORD from virtual device memory (correct value is read only in last pass of assembling, in early passes the zero value is always returned), '{b address}' reads only BYTE.

It is possible to use parenthesis '(' and ')' to override the precedence of the operators. The following operators may be used in expressions:

norel norel L  Label/symbol L will be treated as non-relocatable
exist exist L  is label/symbol L defined in source (even when not used) (since v1.18.2)

!     !x       logical not
~     ~x       complement
+     +x       does "nothing", can be used to make "+(...)" parse as value (not as memory)
-     -x       minus
low   low x    low 8 bits of 16 bit value or lower part of register pair
high  high x   high 8 bits of 16 bit value or higher part of register pair
not   not x    logical not
abs   abs x    absolute value (since v1.18.0)

*     x*y      multiplication
/     x/y      division
%     x%y      modulo
mod   x mod y  modulo

+     x+y      addition
-     x-y      subtraction

<<    x<<y     shift left
>>    x>>y     shift right signed
>>>   x>>>y    shift right unsigned
shl   x shl y  shift left
shr   x shr y  shift right signed

<?    x<?y     minimum
>?    x>?y     maximum

<     x<y      less than
>     x>y      greater than
<=    x<=y     equal or less than
>=    x>=y     equal or greater than

=     x=y      equal
==    x==y     equal
!=    x!=y     not equal

&     x&y      bitwise and
and   x and y  bitwise and

^     x^y      bitwise xor
xor   x xor y  bitwise xor

|     x|y      bitwise or
or    x or y   bitwise or

&&    x&&y     logical and

||    x||y     logical or

$     $        current program counter
$$    $$       current page at program counter (in virtual device mode)
$$$   $$$      "physical" program counter (inside DISP block)
$$$$  $$$$     "physical" memory page (inside DISP block in virtual device mode)
label label    value of label (aka symbol), usually memory address
$$lab $$lab    page of "lab" label (in virtual device mode)
{}    {x}      reads WORD from address x (in virtual device mode, in last pass)
{b}   {b x}    reads BYTE from address x (in virtual device mode, in last pass)

Assembly language

This version only accepts Z80 mnemonics. There are some additions to what I think is standard Z80:

  • '[' and ']' can be used instead of '(' and ')' for indirection. So LD A,[HL] is the same as LD A,(HL) (does not apply to IN/OUT ports, those must use '(...)' form)

  • IN F,(C) and OUT (C),0 and SLL/SLI can be used (warning: on some CPU versions of Z80 the OUT (C),0 is working as OUT (C),255).

  • IXL (or LX, XL), IYL (or LY, YL), IXH (or HX, XH) and IYH (or HY, YH) registers are supported.

  • Can write code throught colon: ORG 100h:LD A,10:LD B,10:SUB B:RET:IFDEF AA:.....

  • JP HL, JP IX and JP IY may be used instead of JP (HL), etc.

  • EX AF,AF or EX AF or EXA may be used instead of EX AF,AF'.

  • R800's MULUB and MULUW are recognised (but won't work on Z80, of course:)

  • Z80N, i8080 and LR35902 (SM83) modes use the identical Z80 sjasmplus syntax (!), for the correct (and incorrect) syntax examples of extended opcodes, please check the test files: Z80N test 1 test 2 and LR35902 (SM83) test 1 test 2 (also for the Z80 syntax examples you can check Z80 tests).

  • RLC, RRC, RL, RR, SLA, SRA, SLL (SLI), RES, SET undocumented instructions added.

    SET 4,(IX+4),C ; (aka LD C,SET 4,(IX+4)) is LD C,(IX+4) / SET 4,C / LD (IX+4),C
    RRC (IY),A     ; (aka LD A,RRC (IY+0))   is LD A,(IY)   / RRC A   / LD (IY),A
  • PUSH and POP can take register lists:

    PUSH AF,BC  ; push af / push bc
    POP  AF,BC  ; pop  af / pop  bc
  • and many other instructions support this "multi-argument" syntax:

    LD A,B,B,D,D,H
   /* this is:
     LD A,B
     LD B,D
     LD D,H
   ;or you can write  LD A,B:LD B,D:LD D,H

   ; since v1.13.1 it is possible and *recommended* to change the multi-arg delimiter
   ; into ",,", to avoid some ambiguities with certain instructions.
   OPT --syntax=a
   LD A,B,,B,D,,D,H ; same as example above in default syntax
   SUB A,B,,C  ; = SUB B, SUB C (the default syntax does for SUB A,B two SUBs!)

   ; since v1.19.0 DEC|INC|POP|PUSH accepts also single-comma in --syntax=a mode
   POP bc,de,hl,,af ; both single/double comma works for these four instructions

Fake instructions

Of course the Z80 is only an 8 bit cpu, but sometimes ld hl,de would be nice. SjASMPlus now 'fakes' some instructions like that. This improves the readability of the source, but it might not be the fastest way to get the result. Also possibly some 'new' load instructions do affect the flags in ways you wouldn't expect. You can use option --syntax=f to get warnings when fake instruction is used, to avoid using them by accident. Here's the list:

  rl bc           ; rl c : rl b
  rl de           ; rl e : rl d
  rl hl           ; rl l : rl h
  rr bc           ; rr b : rr c
  rr de           ; rr d : rr e
  rr hl           ; rr h : rr l
  sla bc          ; sla c : rl b
  sla de          ; sla e : rl d
  sla hl          ; add hl,hl
  sll bc          ; sli c : rl b
  sll de          ; sli e : rl d
  sll hl          ; sli l : rl h
  sli bc          ; sli c : rl b
  sli de          ; sli e : rl d
  sli hl          ; sli l : rl h
  sra bc          ; sra b : rr c
  sra de          ; sra d : rr e
  sra hl          ; sra h : rr l
  srl bc          ; srl b : rr c
  srl de          ; srl d : rr e
  srl hl          ; srl h : rr l

  ld bc,bc        ; ld b,b : ld c,c
  ld bc,de        ; ld b,d : ld c,e
  ld bc,hl        ; ld b,h : ld c,l
  ld bc,ix        ; ld b,xh : ld c,xl
  ld bc,iy        ; ld b,yh : ld c,yl
  ld bc,(hl)      ; ld c,(hl) : inc hl : ld b,(hl) : dec hl
  ld bc,(ix+nn)   ; ld c,(ix+nn) : ld b,(ix+nn+1)
  ld bc,(iy+nn)   ; ld c,(iy+nn) : ld b,(iy+nn+1)

  ld de,bc        ; ld d,b : ld e,c
  ld de,de        ; ld d,d : ld e,e
  ld de,hl        ; ld d,h : ld e,l
  ld de,ix        ; ld d,xh : ld e,xl
  ld de,iy        ; ld d,yh : ld e,yl
  ld de,(hl)      ; ld e,(hl) : inc hl : ld d,(hl) : dec hl
  ld de,(ix+nn)   ; ld e,(ix+nn) : ld d,(ix+nn+1)
  ld de,(iy+nn)   ; ld e,(iy+nn) : ld d,(iy+nn+1)

  ld hl,bc        ; ld h,b : ld l,c
  ld hl,de        ; ld h,d : ld l,e
  ld hl,hl        ; ld h,h : ld l,l
  ld hl,ix        ; push ix : pop hl
  ld hl,iy        ; push iy : pop hl
  ld hl,(ix+nn)   ; ld l,(ix+nn) : ld h,(ix+nn+1)
  ld hl,(iy+nn)   ; ld l,(iy+nn) : ld h,(iy+nn+1)

  ld ix,bc        ; ld xh,b : ld xl,c
  ld ix,de        ; ld xh,d : ld xl,e
  ld ix,hl        ; push hl : pop ix
  ld ix,ix        ; ld xh,xh : ld xl,xl
  ld ix,iy        ; push iy : pop ix

  ld iy,bc        ; ld yh,b : ld yl,c
  ld iy,de        ; ld yh,d : ld yl,e
  ld iy,hl        ; push hl : pop iy
  ld iy,ix        ; push ix : pop iy
  ld iy,iy        ; ld yh,yh : ld yl,yl

  ld (hl),bc      ; ld (hl),c : inc hl : ld (hl),b : dec hl
  ld (hl),de      ; ld (hl),e : inc hl : ld (hl),d : dec hl

  ld (ix+nn),bc   ; ld (ix+nn),c : ld (ix+nn+1),b
  ld (ix+nn),de   ; ld (ix+nn),e : ld (ix+nn+1),d
  ld (ix+nn),hl   ; ld (ix+nn),l : ld (ix+nn+1),h

  ld (iy+nn),bc   ; ld (iy+nn),c : ld (iy+nn+1),b
  ld (iy+nn),de   ; ld (iy+nn),e : ld (iy+nn+1),d
  ld (iy+nn),hl   ; ld (iy+nn),l : ld (iy+nn+1),h

  ldi bc,(hl)     ; ld c,(hl) : inc hl : ld b,(hl) : inc hl
  ldi bc,(ix+nn)  ; ld c,(ix+nn) : inc ix : ld b,(ix+nn) : inc ix
  ldi bc,(iy+nn)  ; ld c,(iy+nn) : inc iy : ld b,(iy+nn) : inc iy

  ldi de,(hl)     ; ld e,(hl) : inc hl : ld d,(hl) : inc hl
  ldi de,(ix+nn)  ; ld e,(ix+nn) : inc ix : ld d,(ix+nn) : inc ix
  ldi de,(iy+nn)  ; ld e,(iy+nn) : inc iy : ld d,(iy+nn) : inc iy

  ldi hl,(ix+nn)  ; ld l,(ix+nn) : inc ix : ld h,(ix+nn) : inc ix
  ldi hl,(iy+nn)  ; ld l,(iy+nn) : inc iy : ld h,(iy+nn) : inc iy

  ldi (hl),bc     ; ld (hl),c : inc hl : ld (hl),b : inc hl
  ldi (hl),de     ; ld (hl),e : inc hl : ld (hl),d : inc hl

  ldi (ix+nn),bc  ; ld (ix+nn),c : inc ix : ld (ix+nn),b : inc ix
  ldi (ix+nn),de  ; ld (ix+nn),e : inc ix : ld (ix+nn),d : inc ix
  ldi (ix+nn),hl  ; ld (ix+nn),l : inc ix : ld (ix+nn),h : inc ix

  ldi (iy+nn),bc  ; ld (iy+nn),c : inc iy : ld (iy+nn),b : inc iy
  ldi (iy+nn),de  ; ld (iy+nn),e : inc iy : ld (iy+nn),d : inc iy
  ldi (iy+nn),hl  ; ld (iy+nn),l : inc iy : ld (iy+nn),h : inc iy

  ldi a,(bc)      ; ld a,(bc) : inc bc
  ldi a,(de)      ; ld a,(de) : inc de
  ldi a,(hl)      ; ld a,(hl) : inc hl
  ldi b,(hl)      ; ld b,(hl) : inc hl
  ldi c,(hl)      ; ld c,(hl) : inc hl
  ldi d,(hl)      ; ld d,(hl) : inc hl
  ldi e,(hl)      ; ld e,(hl) : inc hl
  ldi h,(hl)      ; ld h,(hl) : inc hl
  ldi l,(hl)      ; ld l,(hl) : inc hl
  ldi a,(ix+nn)   ; ld a,(ix+nn) : inc ix
  ldi b,(ix+nn)   ; ld b,(ix+nn) : inc ix
  ldi c,(ix+nn)   ; ld c,(ix+nn) : inc ix
  ldi d,(ix+nn)   ; ld d,(ix+nn) : inc ix
  ldi e,(ix+nn)   ; ld e,(ix+nn) : inc ix
  ldi h,(ix+nn)   ; ld h,(ix+nn) : inc ix
  ldi l,(ix+nn)   ; ld l,(ix+nn) : inc ix
  ldi a,(iy+nn)   ; ld a,(iy+nn) : inc iy
  ldi b,(iy+nn)   ; ld b,(iy+nn) : inc iy
  ldi c,(iy+nn)   ; ld c,(iy+nn) : inc iy
  ldi d,(iy+nn)   ; ld d,(iy+nn) : inc iy
  ldi e,(iy+nn)   ; ld e,(iy+nn) : inc iy
  ldi h,(iy+nn)   ; ld h,(iy+nn) : inc iy
  ldi l,(iy+nn)   ; ld l,(iy+nn) : inc iy

  ldd a,(bc)      ; ld a,(bc) : dec bc
  ldd a,(de)      ; ld a,(de) : dec de
  ldd a,(hl)      ; ld a,(hl) : dec hl
  ldd b,(hl)      ; ld b,(hl) : dec hl
  ldd c,(hl)      ; ld c,(hl) : dec hl
  ldd d,(hl)      ; ld d,(hl) : dec hl
  ldd e,(hl)      ; ld e,(hl) : dec hl
  ldd h,(hl)      ; ld h,(hl) : dec hl
  ldd l,(hl)      ; ld l,(hl) : dec hl
  ldd a,(ix+nn)   ; ld a,(ix+nn) : dec ix
  ldd b,(ix+nn)   ; ld b,(ix+nn) : dec ix
  ldd c,(ix+nn)   ; ld c,(ix+nn) : dec ix
  ldd d,(ix+nn)   ; ld d,(ix+nn) : dec ix
  ldd e,(ix+nn)   ; ld e,(ix+nn) : dec ix
  ldd h,(ix+nn)   ; ld h,(ix+nn) : dec ix
  ldd l,(ix+nn)   ; ld l,(ix+nn) : dec ix
  ldd a,(iy+nn)   ; ld a,(iy+nn) : dec iy
  ldd b,(iy+nn)   ; ld b,(iy+nn) : dec iy
  ldd c,(iy+nn)   ; ld c,(iy+nn) : dec iy
  ldd d,(iy+nn)   ; ld d,(iy+nn) : dec iy
  ldd e,(iy+nn)   ; ld e,(iy+nn) : dec iy
  ldd h,(iy+nn)   ; ld h,(iy+nn) : dec iy
  ldd l,(iy+nn)   ; ld l,(iy+nn) : dec iy

  ldi (bc),a      ; ld (bc),a : inc bc
  ldi (de),a      ; ld (de),a : inc de
  ldi (hl),a      ; ld (hl),a : inc hl
  ldi (hl),b      ; ld (hl),b : inc hl
  ldi (hl),c      ; ld (hl),c : inc hl
  ldi (hl),d      ; ld (hl),d : inc hl
  ldi (hl),e      ; ld (hl),e : inc hl
  ldi (hl),h      ; ld (hl),h : inc hl
  ldi (hl),l      ; ld (hl),l : inc hl
  ldi (ix+nn),a   ; ld (ix+nn),a : inc ix
  ldi (ix+nn),b   ; ld (ix+nn),b : inc ix
  ldi (ix+nn),c   ; ld (ix+nn),c : inc ix
  ldi (ix+nn),d   ; ld (ix+nn),d : inc ix
  ldi (ix+nn),e   ; ld (ix+nn),e : inc ix
  ldi (ix+nn),h   ; ld (ix+nn),h : inc ix
  ldi (ix+nn),l   ; ld (ix+nn),l : inc ix
  ldi (iy+nn),a   ; ld (iy+nn),a : inc iy
  ldi (iy+nn),b   ; ld (iy+nn),b : inc iy
  ldi (iy+nn),c   ; ld (iy+nn),c : inc iy
  ldi (iy+nn),d   ; ld (iy+nn),d : inc iy
  ldi (iy+nn),e   ; ld (iy+nn),e : inc iy
  ldi (iy+nn),h   ; ld (iy+nn),h : inc iy
  ldi (iy+nn),l   ; ld (iy+nn),l : inc iy

  ldd (bc),a      ; ld (bc),a : dec bc
  ldd (de),a      ; ld (de),a : dec de
  ldd (hl),a      ; ld (hl),a : dec hl
  ldd (hl),b      ; ld (hl),b : dec hl
  ldd (hl),c      ; ld (hl),c : dec hl
  ldd (hl),d      ; ld (hl),d : dec hl
  ldd (hl),e      ; ld (hl),e : dec hl
  ldd (hl),h      ; ld (hl),h : dec hl
  ldd (hl),l      ; ld (hl),l : dec hl
  ldd (ix+nn),a   ; ld (ix+nn),a : dec ix
  ldd (ix+nn),b   ; ld (ix+nn),b : dec ix
  ldd (ix+nn),c   ; ld (ix+nn),c : dec ix
  ldd (ix+nn),d   ; ld (ix+nn),d : dec ix
  ldd (ix+nn),e   ; ld (ix+nn),e : dec ix
  ldd (ix+nn),h   ; ld (ix+nn),h : dec ix
  ldd (ix+nn),l   ; ld (ix+nn),l : dec ix
  ldd (iy+nn),a   ; ld (iy+nn),a : dec iy
  ldd (iy+nn),b   ; ld (iy+nn),b : dec iy
  ldd (iy+nn),c   ; ld (iy+nn),c : dec iy
  ldd (iy+nn),d   ; ld (iy+nn),d : dec iy
  ldd (iy+nn),e   ; ld (iy+nn),e : dec iy
  ldd (iy+nn),h   ; ld (iy+nn),h : dec iy
  ldd (iy+nn),l   ; ld (iy+nn),l : dec iy

  ldi (hl),mm     ; ld (hl),mm : inc hl
  ldi (ix+nn),mm  ; ld (ix+nn),mm : inc ix
  ldi (iy+nn),mm  ; ld (iy+nn),mm : inc iy

  ldd (hl),mm     ; ld (hl),mm : dec hl
  ldd (ix+nn),mm  ; ld (ix+nn),mm : dec ix
  ldd (iy+nn),mm  ; ld (iy+nn),mm : dec iy

  adc de,bc       ; ex de,hl : adc hl,bc : ex de,hl
  adc de,de       ; ex de,hl : adc hl,hl : ex de,hl         ;; consider alternative: rl e : rl d (-7T)
  adc de,hl       ; ex de,hl : adc hl,de : ex de,hl
  adc de,sp       ; ex de,hl : adc hl,sp : ex de,hl

  add de,bc       ; ex de,hl : add hl,bc : ex de,hl
  add de,de       ; ex de,hl : add hl,hl : ex de,hl         ;; consider alternative: srl e : rl d (-3T)
  add de,hl       ; ex de,hl : add hl,de : ex de,hl
  add de,sp       ; ex de,hl : add hl,sp : ex de,hl

  sbc de,bc       ; ex de,hl : sbc hl,bc : ex de,hl
  sbc de,de       ; ex de,hl : sbc hl,hl : ex de,hl
  sbc de,hl       ; ex de,hl : sbc hl,de : ex de,hl
  sbc de,sp       ; ex de,hl : sbc hl,sp : ex de,hl

  sub de,bc       ; or a : ex de,hl : sbc hl,bc : ex de,hl
  sub de,de       ; or a : ex de,hl : sbc hl,hl : ex de,hl  ;; consider alternative: ld de,0 (-17T)
  sub de,hl       ; or a : ex de,hl : sbc hl,de : ex de,hl
  sub de,sp       ; or a : ex de,hl : sbc hl,sp : ex de,hl
  sub hl,bc       ; or a : sbc hl,bc
  sub hl,de       ; or a : sbc hl,de
  sub hl,hl       ; or a : sbc hl,hl                        ;; consider alternative: ld hl,0 (-9T)
  sub hl,sp       ; or a : sbc hl,sp

  ; Z80N only
  mul             ; mul de ; to avoid warning specify registers

LDI increases the data pointer after the data access, so LDI A,(HL) is the same as LD A,(HL):INC HL. Likewise, LDD A,(DE) is LD A,(DE):DEC DE.

Real device emulation mode

To enable this mode you must use pseudo-op DEVICE.

In this mode the assembler is assembling program into virtual memory (as MSX's WB-ASS2, ZX-Spectrum's GENS, ZEUS, ALASM etc). You can also use pseudo-ops as MMU, SLOT, PAGE, SAVEBIN, SAVEDEV, SAVETAP, SAVESNA, SAVETRD, SAVENEX, SAVECDT, SAVECPCSNA, SAVEAMSDOS, SAVE3DOS, SAVEHOB, BPLIST, SETBP, CSPECTMAP, LABELSLIST, use special functions in Lua scripts and use operators {address}, {b address} to read WORD/BYTE from the virtual memory.

If only single DEVICE is used in whole source batch, the device becomes "global" and will affect also source ahead of the DEVICE line.

Example 4.2. docs_examples/s_realdevice.asm

    ; in this device the default slot is SLOT 3 with PAGE 0 paged in.

    ORG 32768
    JP $

    ;do something, if you don't want to corrupt virtual
    ;memory with other code, for example, loader of code.

    ;return to our virtual device:

    SAVESNA "snapshotname.sna", StartProg

Predefined devices:


Disable real device emulation mode. By default.


Has 4 slots (0-3) with size 4000h, 4 pages (0-3) with size 4000h. Slot 3 (it from 0C000h) enables to current by default.


Has 8 RAM pages (0-7) with size 4000h. Default slot is 3.


Same as Russian clone Scorption 256. Has 16 RAM pages (0-15) with size 4000h.


Same as Russian clones ATM Turbo 512 and Pentagon 512. Has 32 RAM pages (0-31) with size 4000h.


Same as Russian clones ATM Turbo 2 and Pentagon 1024 SL. Has 64 RAM pages (0-63) with size 4000h.


Similar to other spectrum devices, has 128 RAM pages (0-127) with size 4000h.


Similar to other spectrum devices, has 256 RAM pages (0-255) with size 4000h.


Similar to other spectrum devices, has 512 RAM pages (0-511) with size 4000h.


ZX Spectrum Next, has 8 slots (0-7) of size 0x2000 and 224 RAM pages (0-223) totalling at 1.75MiB of memory. The default mapping is similar to ZX128, paging in: {14, 15, 10, 11, 4, 5, 0, 1} pages. Default slot is 7 (memory range 0xE000..0xFFFF). All memory is zeroed during initialization.


Has only single slot of size 64kiB (slot 0), covering full Z80 CPU addressing range. There are 32 memory pages available, by default page 0 is mapped. That makes in total 2MiB of memory available (32 * 64 ki = 2 Mi).


Has 4 slots (0-3) of size 0x4000 and 4 pages (0-3) of 0x4000, totalling 64K. These pages are mapped linearly as { 0, 1, 2, 3 }.


Has 4 slots (0-3) of size 0x4000 and 8 RAM pages (0-7) of 0x4000 (128K total). The default page mapping is { 0, 1, 2, 3 }.

If you want to see other devices you must write to us. See Feedback chapter.

Predefined defines

SjASMPlus has predefined defines.

__SJASMPLUS__ = <24bit number>

Current version split into three 8bit values, ie. 0x010F02 for "1.15.2".

__VERSION__ = "version string"

String value (with quotes around it) of current version like "1.15.2".

__ERRORS__ = <number>

Number of errors.

__WARNINGS__ = <number>

Number of warnings.


Date of assembling (timestamp before first pass).

__TIME__ = "hh:mm:ss"

Time of assembling (timestamp before first pass, 24h form).

__PASS__ = <number>

Current assembling pass (1, 2 or 3).

__INCLUDE_LEVEL__ = <number>

Current include-nesting level.

__BASE_FILE__ = <name of base file>

Name of base file (include-level zero).

This is raw value without quotes, practically unusable by asm source, but LUA provides means to operate with such string value.

__FILE__ = <name of current file>

Name of current file.

This is raw value without quotes, practically unusable by asm source, but LUA provides means to operate with such string value.

__LINE__ = <number>

Current line number.

__COUNTER__ = <incrementing number>

Does increment upon each usage (starting as value 0). Currently difficult to use for labels/symbols, because the starting `_` of the define name prevents mid-word substitution, will make more sense with new substitution rules (if ever the sjasmplus v2.x happens, would be too big change for v1.x). Right now call LUA for the rescue.

Example 4.3. __COUNTER__ usage in LUA script

    DB __COUNTER__   ; DB 0
        sj.insert_label("label_" .. sj.get_define("__COUNTER__"), sj.current_address)
                -- creates "label_1" at "$" (0x0001)
        sj.insert_label("label_" .. sj.get_define("__COUNTER__"), _c("$+10"))
                -- creates "label_2" at "$+10" (0x000B)
label__COUNTER__: ; does *NOT* substitute in current sjasmplus, sorry
    DB __COUNTER__   ; DB 3

    ; also macro arguments substitution can be used
    MACRO createLabelWithSuffix label?, suffix?
label?_suffix? ; define global label
    createLabelWithSuffix label, __COUNTER__    ; label_4
    createLabelWithSuffix label, __COUNTER__    ; label_5

Deprecated predefined values from sjasmplus till version 1.15.1:

The predefined values were renamed and extended to be more like gcc/clang pre-defines, the following ones are deprecated originals.


Deprecated, consider using similar __SJASMPLUS__

Example 4.4. 

     ;code for sjasmplus
     ;code for other compiler

_VERSION = "version"

Deprecated, renamed to __VERSION__

Example 4.5. 

   IF _VERSION = "1.07"
     ;code for 1.07
     ;code for other version

_RELEASE = releasenumber

Deprecated, consider using similar __SJASMPLUS__

Example 4.6. 

   IF _RELEASE = 1 ; 0 - is stable version
     ;code for Release Candidate 1
     ;code for other version

_ERRORS = <number>

Number of errors. Deprecated, renamed to __ERRORS__

_WARNINGS = <number>

Number of warnings. Deprecated, renamed to __WARNINGS__

Chapter 5. Pseudo-ops (aka Pseudo-instructions, Directives etc)

Simple example of usage


[Almost] complete list of pseudo-ops follows:


.<repeat-count> <single instruction>

Repeat <single instruction> <repeat-count> many times. Doesn't work in the beginning of line. The <repeat-count> must be either simple integer number or expression fully enclosed in parentheses.

Example 5.1. docs_examples/po_dot_repeat.asm

    .3          INC A    ;will be compiled to INC A:INC A:INC A
len             EQU 10
    .(12-len)   BYTE 0   ;will be compiled to BYTE 0,0
    .2 .3       RET      ;will be compiled to 6x RET

ABYTE <offset> <bytes>

Defines a byte or a string of bytes. The offset is added to each of the following bytes.

Example 5.2. 

    ABYTE 2 4,9    ; Same as BYTE 6,11
    ABYTE 3 "ABC"  ; Same as BYTE "DEF"

ABYTEC <offset> <bytes>

Defines a byte or a string of bytes, where the last byte of the string will have bit 7 set. The offset is added to each of the following bytes.

Example 5.3. 

    ABYTEC 0 "KIP"        ; Same as BYTE "KI",'P'|128
    ABYTEC 1 "ABC",0,"DE" ; Same as BYTE "BC",'D'|128,1,'E','F'|128

ABYTEZ <offset> <bytes>

Defines a byte or a string of bytes, followed by a zero. The offset is added to each of the following bytes.

Example 5.4. 

    ABYTEZ 0 "KIP"        ; Same as BYTE "KIP",0

ALIGN [<expression equal to 1,2,4,8,16,32,64,128,256,512,1024,2048,4096,8192,16384 or 32768>[, <byte>]]

Align advances to nearest address where <new address> modulo <expression> (default 4) equals zero (stays at current address if possible).

If <byte> is specified, memory advanced over is set to it.

Example 5.5. 

    ALIGN         ; => ALIGN 4 - simply align by 4
    ALIGN 2       ; by 2 (preserves value of "device" memory)
    ALIGN 2,0     ; + fills memory with zero

ASSERT <expression>[, <anything>]

An 'assertion failed' error is issued if the expression evaluates to zero. The "anything" (optional argument since v1.18.1) is then also visible in the error message, so you can add description/notes about expression there.

Example 5.6. 

    ASSERT END_OF_PROGRAM < STACKPOINTER, Program code leaks into stack area

BINARY <filename>[,<offset>[,<length>]]

Synonym of INCBIN.

BLOCK <length>[,<fill byte>]

Defines space. Has to be followed by the number of byte to reserve, optionally followed by the value to fill these bytes with.

Example 5.7. 

    BLOCK 500     ; define a block of 500 bytes of zero
    BLOCK 500,0   ; define a block of 500 bytes of zero
    BLOCK 400,-1  ; define a block of 400 bytes of 255

BPLIST <filename> [unreal|zesarux]

Works only in real device emulation mode. See DEVICE.

Opens file to export breakpoints info directly from ASM source, currently two flavours of export are supported: unreal (default) and zesarux. For latest Unreal emulators use filename "bpx.ini", you can then load the file from the emulator UI. For ZEsarUX the file will contain command-line options, which you can add to the command when launching the emulator.

(current version of ZEsarUX will not catch the very first instruction of freshly loaded snapshot/NEX/... file, or it will even disable breakpoints when changing machine parameters and ignore the --enable-breakpoints option - will be hopefully improved in the future)

Example 5.8. 

    BPLIST "bpx.ini" unreal  ; open breakpoints list in "Unreal" format
    ; or (only one file per assembling-unit can be specified)
    BPLIST "cmd_line_options.txt" zesarux ; open breakpoints list in "ZEsarUX" format

BYTE <bytes>

Defines a byte or a string of bytes. Each value should be between -129 and 256.

Example 5.9. 

    BYTE 0x56
    BYTE 1,-78,'@'
    BYTE "Format C:? ",0h

CSPECTMAP [<filename>]

Useful for ZX-Spectrum Emulator #CSpect by Mike Dailly.

Works only in real device emulation mode. See DEVICE.

Saves labels list in format:


where TYPE is: 00 = regular label, 01 = EQU or struct defined, 02 = DEFL defined, 03 = ROM page/"none" device, 04 = STRUCT definition.

If no filename is provided, default is created by appending ".map" to source name.


Defines three bytes by 24b constant. Values should be between -16777217 and 16777216.

Example 5.10. 

    D24 0x123456   ; define three bytes 0x56, 0x34, 0x12


Synonym of BYTE.


Same as BYTE, but every last character of a string will have bit 7 set. Single character in apostrophes is NOT considered as string.

Example 5.11. 

    DC "kip" ; same as BYTE "ki",'p'|128
    DC 'A'   ; not a string, will produce 0x41 like BYTE 'A'
    DC "A", 'AB', "AB" ; three strings producing: C1 41 C2 41 C2


Synonym of DWORD.

DEFARRAY <id> <replacements>

Array of DEFINEs.

Use id[#] to retrieve current size of array.

Example 5.12. docs_examples/po_defarray.asm

    DEFARRAY myarray 10*20,"A",20,</D,40>,50,70
CNT DEFL 0 ;or CNT=0
    DUP myarray[#]      ; 6
    DISPLAY myarray[CNT]

DEFARRAY+ <id> <additional replacements>

Appending more DEFINEs to already defined array

Example 5.13. 

    DEFARRAY   myarray 'A', 'B', 'C'
    DEFARRAY+  myarray 'D', 'E'            ; now "myarray" has 5 items
    DUP 3 : DEFARRAY+ myarray '!' : EDUP   ; "DEFARRAYFILL" adding 3x '!'


Synonym of BYTE.


Synonym of DWORD.

DEFDEVICE <deviceid>, <slot_size $100..$10000>, <page_count 1..1024>[, <slot_0_initial_page>[, ...]]

(since v1.20.1) Add custom device definition. The 64ki Z80 address space is divided evenly by slot_size to form N slots and the total device memory is slot_size * page_count bytes.

The initial pages are implicitly going from 0 to page_count-1 (repeating page page_count-1 for remaining slots when there's not enough pages), but you can specify initial page layout explicitly.

Use DEVICE to select the newly defined device.


Synonym of DG.


Synonym of DH.

DEFINE+ <id> <replacement>
DEFINE <id> <replacement>

The identifier <id> will be replaced with the <replacement>. The DEFINE+ (since v1.18.0) will also redefine the identifier <id> if it already exists. The replacement could be omitted, in such case it is still possible to check if the identifier was defined with IFDEF or IFNDEF.

Example 5.14. 

    DEFINE str_honderd "Honderd"
    BYTE str_honderd,0             ; BYTE "Honderd",0
    DEFINE+ do_stuff set 4,e
    do_stuff                       ; set 4,e
    DEFINE+ do_stuff set 2,e
    do_stuff                       ; set 2,e

<label> DEFL <expression>

Assigns value of <expression> to symbol <label>. New label defined by DEFL is marked internally as "modifiable", allowing to re-assign new values to it with further DEFL statements (you can use also = instead of DEFL).

Example 5.15. 

counter DEFL 0
    DUP 4
        DB 0xAA, counter
counter = counter + 1
; machine code produced: AA 00 AA 01 AA 02 AA 03


Synonym of BYTE.


Synonym of BLOCK.


Synonym of WORD.


Synonym of ENT.

DEVICE <deviceid>[, <RAMTOP>]

Enables real device emulation mode by it identifier. If there is only single DEVICE directive in whole source batch, it becomes "global" and the device affects all lines of source, otherwise the DEVICE is applied for lines following it.

For ZXSPECTRUM-like devices (except ZX Next) you can provide RAMTOP value, which will init the device memory in similar way as "CLEAR <RAMTOP>" in BASIC would do, putting top of the fake-stack at the RAMTOP address. The default RAMTOP is since v1.15.0 0x5D5B (it was 0xFF57 before).

Predefined devices' identifiers list:

 NONE ; off real device emulation mode
 ZXSPECTRUM48   ; ZX-Spectrum 48 (4 slots, 4 pages, slot/page size 0x4000, default map: 0, 1, 2, 3)
 ZXSPECTRUM128  ; ZX-Spectrum 128 (like 48 with 8 pages, default map: 7, 5, 2, 0)
 ZXSPECTRUM256  ; e.g. Scorpion 256 (exUSSR clone of ZX-Spectrum 128)
 ZXSPECTRUM512  ; e.g. ATM-Turbo 512 (another clone)
 ZXSPECTRUMNEXT ; ZX Spectrum Next (8 slots, 224 pages, slot size 0x2000 = 1.75MiB RAM)
                ; (default pages map: 14, 15, 10, 11, 4, 5, 0, 1) (default slot: 7 (0xE000..0xFFFF))
 NOSLOT64K      ; Single slot (slot 0) covering full 64kiB address range of Z80, 32 pages
 AMSTRADCPC464  ; Amstrad CPC 464 (4 slots, 4 pages, slot/page size 0x4000, default map: 0, 1, 2, 3)
 AMSTRADCPC6128 ; Amstrad CPC 6128 (like 464 with 8 pages)

DG <data encoded in bits>

Data comprises of characters in multiples of eight, each block is converted to a byte value.

A hyphen '-' (also '.' and '_') represents 0 and any other non-whitespace character represents 1. It ignores spaces, use them for formatting if you like. Warning, "DG 10100001" is value 255, because character '0' is not a dash '-'. (since v1.11)

Example 5.16. 

    DG 1-1----1  ; store 161 at the current location
    DG ...# #... .##. .... ; store two bytes: 0x18, 0x60

DH "<data>"[,"<data2>"...]

The data string comprises pairs of hexadecimal digits, each pair is converted to a byte value. You can add spaces between pairs as you like. (since v1.11)

Example 5.17. 

    DH "0123456789ABCDEF"   ; eight bytes #01 #23 ...
    DH "01 23  45 67"       ; four bytes #01 #23 #45 #67

DISP <address>[,<page_number>]

Set the address in which the part of code should work. PHASE and TEXTAREA are synonyms of DISP. ENT will restore current address. UNPHASE, DEPHASE and ENDT are synonyms of ENT. DISP blocks can NOT be nested, and to change the displacement address within current DISP block use the ordinary ORG. When in device mode, you can specify fixed value for "fake" page of emitted instructions and regular labels, but to avoid warning, you must also map-in the target page into the target memory slot (MMU). When no fixed page in DISP is specified, the current mapping of memory pages is used.

Example 5.18. docs_examples/po_disp.asm

        ORG $8000
        LD HL,BEGIN
        LD DE,SCREEN
        JP SCREEN
BEGIN   DISP SCREEN ;code will compile for address $4000, but to the current ORG
            JP NZ,MARKA

    ASSERT $800E == BEGIN && $8015 == ENDOFPROG && $4000 == MARKA
    ASSERT $76 == {B $800F}     ; HALT instruction lands at $800F (BEGIN+1)

DISPLAY <bytes>

This pseudo-op comes from ZX-Spectrum assembler ALASM.

Out to console a string of bytes. Each value should be between -129 and 256. Keys /D, /B, /C, /H and /A set format of output of numbers:

/D - out only in Decimal
/B - out only in Binary (truncated to 8 bit)
/C - out as character in apostrophes (truncated to 8 bit)
/H - out only in Hexadecimal
/A - out both in Hexadecimal and Decimal

Example 5.19. docs_examples/po_display.asm

    ORG 100h
    ;...some code...
    DISPLAY "--the some program-- by me"
will output to the console strings:
> --the some program-- by me
> TESTLABEL address is:0x100, 256


Synonym of BYTE.


Synonym of BLOCK.

DUP <count>[, <index_variable>]

DUP specifies the number of times to generate the following lines until an EDUP pseudo-op is encountered. DUP can be used in macro's.

Optional second argument (since v1.20.2) is label name to have repetition index available inside the block (values 0, 1, 2, 3, ...).

Example 5.20. 

    DUP 3 ; zero count would skip whole block

/*this will expand to:


Synonym of WORD.


Defines a so called doubleword. Values should be between -2147483649 and 4294967296.

Example 5.21. 

    DWORD 4000h,0d000h
    DWORD 4


Same as BYTE, but an extra zero will be added at the end.

Example 5.22. 

    DZ 1      ; same as BYTE 1,0
    DZ "kip"  ; same as BYTE "kip",0

EMPTYTAP <filenameoftapefile>

Useful only for ZX-Spectrum users

Create the new or truncate existing standard tape file for emulators of ZX-Spectrum. See example of SAVETAP.

EMPTYTRD <filenameoftrdimage>[,<disclabel>]

Useful only for ZX-Spectrum users

Create the empty TRD image for emulators of ZX-Spectrum. See example of SAVETRD.

ENCODING <encoding>

Useful only for non English users

Set the current encoding, i.e. if you set "DOS", SjASMPlus will automatically convert strings from ANSI to DOS-866. Encoding may be "DOS"(DOS-866) or "WIN"(ANSI/Win-1251). Default is "WIN".

Example 5.23. 

    DB "тексттекст" ;will be тексттекст
    DB "тексттекст" ;will be ⥪бв⥪бв

END [<startaddress>]

The assembler will stop at this point. The pseudo-op END does NOT work in the beginning of line (even with --dirbol). The optional argument is used by SAVESNA, SAVECPCSNA, SAVETAP and SAVENEX.


See LUA for more information.


Synonym of ENDMODULE.


To indicate the end of a module (see MODULE).


Synonym of ENT.


Restore current address. See DISP for more information.

<label> EQU <expression>[,<pageNumber>]

To give the label a value other than the current program counter. The label should not already exist (you can assign only one value to it). The optional pageNumber can enforce explicit page value for the label (for the $$label operator). For modifiable labels holding temporary values use DEFL.

Example 5.24. 

Label EQU 3
Kip   EQU 0x23*256 + low $

EXPORT <label>

The named label will be written to the export-file, in the form 'label: EQU value'. This way the export-file can be included in other sources. Order of labels in file follows the order of EXPORT statements.

Example 5.25. 

    EXPORT DRIE  ; adds into export file line: "DRIE: EQU 0x00000003"

FPOS <position>

The FPOS directive makes it possible to set the file position to anywhere in the output file. Position value without sign is used as absolute one to be set, position starting with + or - sign will be used as relative position.

In combination with OUTPUT <filename>,r it is possible to update existing files.

Example 5.26. 

; This example will result in a file with a length of one byte:
    BYTE 0
    FPOS 0
    BYTE 1


Synonym of DH, usually used without quotes around data.

INCBIN <filename>[,<offset>[,<length>]]

To include a binary file into the outputfile. The offset and length are optional. Added in v1.12.1: if negative offset or length is provided, it counts relatively from the end of the file.

Example 5.27. 

    INCBIN "gfx.scc",7        ; include gfx.scc, skip first 7 bytes
    INCBIN "",3,256 ; include 256 bytes from offset 3
    INCBIN gfx.scc ,,7        ; 7 bytes from offset 0 (unquoted filename must end with space)
    INCBIN "48.rom",-768,-256 ; include (from 16kiB file) 512 bytes 15616..16127

INCHOB <filename>[,<offset>[,<length>]]

To include a data from a hobeta file into the outputfile. The offset and length are optional.

Example 5.28. 

    INCHOB "gfx.$c",7        ; include gfx.scc, skip first 7 bytes
    INCHOB "sprs.$c",3,256   ; include 256 bytes from offset 3
    INCHOB gfx.$c ,7        ; note the space between the filename and the ',7' here :)

INCLUDE <filename>

To include another sourcefile into the current. Sourcefiles can be nested 20 levels deep. If the file cannot be found in the current directory (the current directory is the directory the current asm file comes from!) the file will be searched for in the directories specified at the commandline. When angle brackets are used, the commandline directories are searched before the current directory.

The directory used to launch the assembling process is automatically added to the list (as if "-i." was added to command line manually) (v1.14.0 and v1.14.1 don't add it, reverted back for v1.14.2). If you want to start with completely empty include-path list, use "--inc" option early (order matters) without the "=" to empty the current list, like: sjasmplus --inc --inc=path1 --inc=path2 file.asm

Example 5.29. 

    INCLUDE <VDP.I>     ; search for file "VDP.I" in the include directories, then in current
    INCLUDE MORE.I      ; search for "MORE.I" in current directory, then in include directories

INCLUDELUA <filename>

To include another LUA script in first pass(!). If the file cannot be found in the current directory (the current directory is the directory the current file comes from) the file will be searched for in the directories specified at the commandline. When angle brackets are used, the commandline directories are searched before the current directory.

Example 5.30. 

    INCLUDELUA <mylibrary1.lua>
    INCLUDELUA mylibrary2.lua
    INCLUDELUA "library_for_zx.lua"

INCTRD <filenameoftrdimage>,<filenameintrdimage>[,<offset>[,<length>]]

To include a file from a TRD image into the outputfile. The offset and length are optional.

Example 5.31. 

    INCTRD "test.trd","mygfx.C" ; include mygfx.C from test.trd
    INCTRD "test.trd","mygfx.C",12 ; include mygfx.C from test.trd, skip first 12 bytes

INSERT <filename>[,<offset>[,<length>]]

INSERT is a synonym of INCBIN. See above.

LABELSLIST <filename>[,<virtual labels>]

Useful only for ZX-Spectrum Emulator UNREALSPECCY.

Works only in real device emulation mode. See DEVICE.

Save labels list in format:


where NN is number of RAM page and ADDRESS is truncated to 0000..3FFF range

If <virtual labels> is non zero, then the page number NN is not part of output, and ADDRESS is truncated to 0000..FFFF range.

Example 5.32. 

    LABELSLIST "x:/somepath/user.l"

LUA [pass]

Using pseudo-ops LUA and ENDLUA you can insert Lua scripts. See more in the chapter "Lua scripting".

Parameter is optional. It may be:

PASS1   -  interpret Lua script in first pass only.
PASS2   -  interpret Lua script in second pass only.
PASS3   -  interpret Lua script in third pass only. By default.
ALLPASS -  interpret Lua script in all passes. It is needed if you generate some Z80 code.

Example 5.33. 

-- some comments
        print "Hi, man! This is Lua!"
; some code now:
        _pl("LABEL LD A,10")


Not available yet.

MMU <first slot number or address> [<last slot number or address>|<single slot option>], <page number>[,<address>]

Maps memory page(s) to slot(s), similar to SLOT + PAGE combination, but allows to set up whole range of consecutive slots (with consecutive memory pages). Or when only single slot is specified, extra option can be used to extend particular slot functionality. The slot behaviour will stay set in the current DEVICE until reset by another MMU specifying same slot (even as part of range, that will clear the option to "default").

The optional third argument is address for ORG functionality.

(since v1.18.1) You can also use starting address of particular slot instead of its number, ie. for ZX128: MMU $8000,3 is alias of MMU 2,3

Single slot option (default state is: no error/warning and no wrap = nothing special):

e = error on writing beyond last byte of slot
w = warning on writing beyond last byte of slot
n = wrap address back to start of slot, map next page

Example 5.34. docs_examples/po_mmu.asm

    DEVICE ZXSPECTRUM128 : LABELSLIST "po_mmu.lbl"  ; to check label pages
    MMU 1 3, 5      ; maps slots 1, 2, 3 with pages 5, 6, 7
    ORG 0xBFFF
label1_p6: scf      ; last byte of page 6 (in slot 2)
label2_p7: scf      ; first byte of page 7 (in slot 3)

    MMU 3 e, 0      ; page 0 into slot 3, write beyond slot will cause error
    ORG 0xFFFF
    ld  a,1         ; error: Write outside of memory slot: 65536 (65536 = address outside)

    MMU 3 n, 1      ; page 1 into slot 3, make it wrap + map next page automatically
    ORG 0xFFFF      ; ! also the $ address was truncated by MMU from $10001 to $0001 !
label3_p1: scf      ; last byte of page 1, then wrapping back to 0xC000 with page 2
label4_p2: scf      ; first byte of page 2 at 0xC000

MODULE <name>

Labels has to be unique only whithin the current module (module is added as prefix to them). Also note the use of '@' operator to suppress all this label-processing. Modules can be nested, and module has to be ended by ENDMODULE.

Example 5.35. docs_examples/po_module.asm

    MODULE xxx
Kip:                ; label xxx.Kip
    ld  hl,@Kip     ; global Kip
    ld  hl,@Kop     ; global Kop
    ld  hl,Kop      ; xxx.Kop
Kop:                ; label xxx.Kop
    ld  hl,Kip      ; xxx.Kip
    ld  hl,yyy.Kip  ; yyy.Kip
    ld  hl,nested.Kip   ; xxx.nested.Kip
        MODULE nested
Kip:        ret     ; label xxx.nested.Kip

    MODULE yyy
Kip:    ret         ; label yyy.Kip
@Kop:   ret         ; label Kop (global one, no module prefix)
@xxx.Kop: nop       ; ERROR: duplicate: label xxx.Kop

Kip     ret         ; global label Kip

Since v1.14.0 the module <name> can NOT contain dot character. You can use nested modules to get identical identifier as in older versions, or please rename with underscores/etc:

    ; invalid since v1.14.0
        MODULE older.version
fn1:        ret        ; final label: @older.version.fn1
    ; can be replaced in v1.14.0 with
        MODULE new
            MODULE version
fn1:            ret    ; final label: @new.version.fn1

Since v1.14.0 the MODULE and ENDMODULE also resets the current "non-local" label prefix back to "_":

Kep:    ; "Kep" label (global one), and also works as "non-local" prefix for local labels
    MODULE zzz
.local: ; in v1.14.0 this will be "zzz._.local" label, previously it was "zzz.Kep.local"
Kup:    ; this is "zzz.Kup", but also defines "non-local" prefix as "Kup"
.local  ; this is "zzz.Kup.local"
.local: ; in v1.14.0 this will be "_.local" label, previously it was "Kup.local"

OPT [push] [reset] [listoff] [liston] [listall] [listact] [listmc] [<command line options>]

OPT pop

Allows to reset and modify options affecting syntax and parsing (for lines of source following the OPT). The options allowed for OPT are: -W[no-]<warning_id>, --syntax, --zxnext, --reversepop and --dirbol.

Ahead of options you can use OPT commands: push, pop, reset, listoff, liston, listall, listact, listmc. The "push" command will make OPT to preserve current state of options. The "reset" command will reset OPT-related options to default state. The "listoff" command will suspend the listing for following lines until "liston" command is used (listing availability is part of the push/pop state, so to "nest" listing-off you can use "OPT push listoff : ... code ... : OPT pop" code sequence). The "listall", "listact" and "listmc" commands will select the filter for listing. The "listmc" lists only lines containing machine code bytes, "listact" will remove the "skipped" blocks (like false-blocks of IF/ELSE conditional assembling, or inner definitions of macros and structures), "listall" will switch back to default listing without filtering.

Then the provided options are applied. The default values are: fake instructions enabled (no warning), multi-argument delimiter is ",", both () and [] brackets can be used to access memory, labels can have any name, ZX Next instructions are OFF, POP with multiple arguments doesn't reverse them and pseudo-ops at beginning of line are OFF (to just reset to these defaults you can use OPT reset).

The "pop" command: the previously preserved state of options is restored (states are preserved in "stack" way, so further OPT with "pop" will restore older states).

The id-warning system state is NOT saved/restored by OPT push/pop/reset and can be modified only by -W option.

Example 5.36. docs_examples/po_opt.asm

    POP bc, hl   ; pops BC first
    OPT push reset --reversepop --syntax=af
    POP bc,,hl   ; pops HL first
    LD  bc,hl    ; warning about Fake instruction
    LD  bc,hl    ; warning supressed by lowercase "fake" in this comment
    OPT reset --syntax=a
    POP bc,,hl   ; pop BC first (--reversepop was reset)
    OPT pop      ; restoring syntax to original state

ORG <address>[,<page_number>]

Set the program counter to a specific address. If the second argument is provided, it will change memory page in the current slot, see PAGE and SLOT (it will warn you when <address> is outside of it).

When used inside DISP block, only the virtual "displaced" program counter is affected, but the machine code will be still sequentially emitted in the original physical location and warning is emitted.

Example 5.37. docs_examples/po_org.asm

    ORG 100h ; or 0x100, or $100, or #100

    ; useful macro that padding code
         ; add padding
         IF $ < addr
         BLOCK addr-$
         ORG addr

    MACRO PADORG2 addr ; z80asm "FORG" replacement
         ; add padding + display warning
         IF $ > addr
           ; no padding
           DISPLAY /L, "Warning! PADORG failed! ", $, " is more than ", addr
           ; add padding
           BLOCK addr-$
         ORG addr


Ends generating compiler output to file specified in OUTPUT and resets SIZE value to default "none".

OUTPUT [<filename>[,<mode>]]

With OUTPUT it is possible to create multiple files from one source. All following instructions will be assembled to this file. It will also close (OUTEND) and finalize any previously opened output.

There are three possible output modes: truncate (overwrite existing files, this is the default), rewind (open and execute FPOS 0) and append (open and leave the file pointer at the end of the file).

OUTPUT <filename>,t  ; truncate (default)
OUTPUT <filename>,r  ; rewind
OUTPUT <filename>,a  ; append

Example 5.38. bigfile.asm

    ORG 100H
    INCLUDE loader.asm
    INCLUDE bios.asm

    OUTPUT bigfile.dat
    ORG 4000H
    INCLUDE main.asm
    ORG 8000H
    INCLUDE data.asm
    INCLUDE next.asm

This will create two files: and bigfile.dat.

PAGE <number>

Works only in real device emulation mode. See DEVICE.

Set the current memory page to current slot.

Example 5.39. 

    PAGE 7 ;set 7 page
    SAVEBIN "ram7.bin",$C000,$4000 ;- save $4000 begin from $C000 of RAM to file


Synonym of DISP.


Ends the block of code for which relocation data are produced (started by RELOCATE_START). To get the relocation data themselves, use RELOCATE_TABLE and related symbols.

For more details/examples check the examples/relocation, examples/SymbOS_nslookup and the relocation tests.


Marks start of block for which relocation data are produced. Any *regular* label (not EQU/DEFL symbols) defined inside the RELOCATE_START .. RELOCATE_END block will be treated as "relocatable" label. Any instruction/directive within the block which needs resulting opcode adjusted when relocated to different address will produce relocation-data for the RELOCATE_TABLE.

The RELOCATE_START HIGH variant checks only high-byte relocation. The produced table points to high byte (MSB) of the machine code. This mode is for OS/loaders which relocate only to 256-byte aligned boundaries (keeping LSB intact).

The relocation feature of regular label is transitive and EQU value based on already relocatable label will be tagged as relocatable too. Also structures instantiated in relocatable block have their labels tagged as relocatable and their initial values may become part of relocation data if the initial value is based on relocatable value.

You can prepend operator norel to label to strip it of its relocatable property for particular expression.

The internal heuristic is not bullet-proof, some complex expression may fail to be recognised as fixed/relocatable value and instead warn about unstable result. It is recommended to assemble relocatable blocks in lower regions of address space (like ORG 0 or ORG $1000) and to keep expressions simple enough.

RELOCATE_TABLE [<subtract_offset>]

Emits relocation data gathered across all relocation blocks. There are also two built-in symbols relocate_count and relocate_size containing the size of the table (only word-type tables are supported in sjasmplus, so size is always doubled count).

The table is raw array of WORD (16 bit) values with memory offsets of machine code which should be adjusted (when the code is relocated to different than original base address). To get offsets from particular base address, use ORG ahead of relocation block to land instructions and labels from there.

The <subtract_offset> is subtracted from every offset recorded in the relocation table (same what WinAPE does with its optional argument "base_address" for RELOCATE_TABLE).

REPT <count>[, <index_variable>]

Synonym of DUP. There is also synonym ENDR to end REPT block (EDUP works too).

SAVE3DOS <filename>,<address>,<size>[,<type>[,<w2_line>[,<w3>]]]

Works only in virtual device emulation mode. See DEVICE.

Like SAVEBIN, saves the block of device RAM from "address", but with +3DOS 128 byte header prepended.

Example 5.40. 


    ; +3 BASIC header (8 bytes starting from offset 15 in +3DOS header):
    ; type is first byte (0 = basic, 1, 2 = arrays, 3 = code/screen$)
    ; size is "first word" (bytes 1 and 2)
    ; w2_line argument is "second word" (bytes 3 and 4)
    ; w3 argument is "third word" (bytes 5 and 6) (default value = size)

    ; by default the type=3 (CODE/SCREEN$) is saved
    SAVE3DOS "screen.bin", $4000, $1b00
    ; type 3: w2 argument $C000 is target load-address
    SAVE3DOS "code.bin", $8000, $1234, 3, $C000
    ; type 0: BASIC, default w2 = $8000 (no auto-run)
    SAVE3DOS "prg1.bas", $5c00, 1234, 0
    ; type 0: w2_line is LINE auto-run (30 in example)
    SAVE3DOS "prg2.bas", $5c00, 1234, 0, 30
    ; type 1: Numeric array (refer to +3 manual for these)
    SAVE3DOS "b.dat", $8000, 30, 1
    ; type 2: Character array (or maybe just ignore these)
    SAVE3DOS "c.dat", $8000, 30, 2

SAVEAMSDOS <filename>,<address>,<size>[,<start = 0>[,<type = 2>]]

Works only in virtual device emulation mode. See DEVICE.

Like SAVEBIN, saves the block of device RAM from "address", but with AMSDOS 128 byte header prepended.

Example 5.41. 

    ... some code ...

    ; by default the type=2 (BINARY) is saved
    ; "start" (entry) value is by default 0, overriden here to $1234
    SAVEAMSDOS "code.bin", $1000, $1b00, $1234

SAVEBIN <filename>,<startadress>,<lengthofcode>

Works only in real device emulation mode. See DEVICE.

Save the block of RAM.

Example 5.42. 

    PAGE 7 ;set 7 page to current slot
    SAVEBIN "ram7.bin",$C000,$4000 ;- save 4000h begin from C000h of RAM to file
    SAVEBIN "ram2.bin",$8000,$3000 ;- save 3000h begin from 8000h of RAM to file

SAVECDT FULL <cdtname>[,<startaddr>[,<screenmode>[,<border>[,<ink0>...<ink15>]]]]


SAVECDT BASIC <cdtname>,<name>,<start>,<length>

SAVECDT CODE <cdtname>,<name>,<start>,<length>[,<customstartaddress>]

SAVECDT HEADLESS <cdtname>,<start>,<length>[,<sync>[,<format>]]

Works only in Amstrad device emulation mode. See DEVICE.

Manipulate CDT file (tape/TZX file format used by Amstrad CPC emulators).

Example 5.43. 

    ; create empty CDT file (truncates already existing file to make it empty)
    SAVECDT EMPTY "output.cdt"

    ; snapshot-like dump with default loader, including loading screen, all banks, ...
    SAVECDT FULL "output.cdt" ; defaults: END start addres, screenmode=1, border=1
                              ; ink# = 1, 24, 20, 6, 26, 0, 2, 8, 10, 12, 14, 16, 18, 22, 24, 16

    ; BASIC program prepared at address "label"
    SAVECDT BASIC "output.cdt","A",label,length

    ; block of binary data at address "label" (default "entry" = "label")
    SAVECDT CODE "output.cdt","C",label,length,entry

    ; headless block with custom sync and format at address "label", formats: 0 = AMSTRAD, 1 = ZX
    SAVECDT HEADLESS "output.cdt",label,length,sync,format ; default sync = 0x16, format = 0

SAVECPCSNA <filename>[,<startadressofprogram>]

Works only in AMSTRADCPC464 / AMSTRADCPC6128 real device emulation mode. See DEVICE.

Save the snapshot for emulators of Amstrad CPC machines. (If start address is omitted, the one provided by END is used)

The snapshot header content is hard-wired in the sjasmplus binary, with the CRTC, Palette and Gate Array registers set to those of a CPC after a boot into the main ROM. Similarly to the ZX-Spectrum SAVESNA directive, interrupts are disabled and must be re-enabled by your program. A full list of the settings can be found in the source. If you absolutely need different defaults and can't afford init-code modifying the state, you can post-patch the resulting SNA file directly from source, see Issue #139 for example.

Example 5.44. 

    ORG $1200
START  .... ;some code
    SAVECPCSNA "game.sna",START ;save snapshot to file game.sna. Start address is START ($1200)

SAVEDEV <filename>,<startPage>,<startOffset>,<length>

Works only in real device emulation mode. See DEVICE.

Like SAVEBIN, saves the block of device RAM.

But it allows lengths over 64ki, and the offset value goes directly into device virtual memory (where pages are allocated consecutively), ignoring current slot "mapping". I.e. page=2,offset=0 will start saving data from page 2 at its beginning, going through pages 3, 4, 5, ... until the requested length of data is saved.

The offset is not limited to page size, i.e. arguments page=1,offset=0x500 are equal to arguments page=0,offset=0x4500 for ZXSPECTRUM128 device (has page size 0x4000).

Example 5.45. 

    DEVICE ZXSPECTRUM128 : SAVEDEV "fullram.bin",0,0,0x20000 ; save full 128kiB

SAVEHOB <filename>,<filename_in_trdos>,<startadress>,<lengthofcode>

Works only in real device emulation mode. See DEVICE.

Save the block of RAM in Hobeta format.

Example 5.46. 

    PAGE 7 ;set 7 page to current slot
    SAVEHOB "ram7.$c","myfile1.C",$C000,$4000 ;- save 4000h begin from C000h of RAM to file
    SAVEHOB "ram2.$c","myfile2.C",$8000,$3000 ;- save 3000h begin from 8000h of RAM to file

SAVENEX <command> <command arguments>

Commands to build NEX file, for details check SAVENEX chapter.

Works only in ZXSPECTRUMNEXT device emulation mode. See DEVICE.

SAVESNA <filename>[,<startadressofprogram>]

Works only in real device emulation mode. See DEVICE.

Save the snapshot for emulators of ZX-Spectrum. (If start address is omitted, the one provided by END is used)

The snapshot header content is hard-wired in sjasmplus binary (disabled interrupts may surprise some users, you can check the source to see all the values). If you absolutely need different defaults and can't afford init-code modifying the state, you can post-patch the resulting SNA file directly from source, see Issue #139 for example.

Example 5.47. 

    ORG $8000
START  .... ;something code
    SAVESNA "game.sna",START ;save snapshot to file game.sna. Start address is START ($8000)

SAVETAP <filename>,BASIC,<fileintapeheader>,<start>,<length>[,<autorunline>[,<lengthwithoutvars>]]

SAVETAP <filename>,CODE,<fileintapeheader>,<start>,<length>[,<customstartaddress>[,<optional3rdparam>]]

SAVETAP <filename>,NUMBERS,<fileintapeheader>,<start>,<length>[,<variableletter(A..Z)>]

SAVETAP <filename>,CHARS,<fileintapeheader>,<start>,<length>[,<variableletter(A..Z)>]

SAVETAP <filename>,HEADLESS,<start>,<length>[,<customblockflag(0..255)>]

Works only in real device emulation mode. See DEVICE.

Append the tape header or block of data to the end of the standard tape file for emulators of ZX-Spectrum.

Example 5.48. 

    EMPTYTAP "output.tap"

    SAVETAP "output.tap",BASIC,"noAutorun",label,100
    SAVETAP "output.tap",BASIC,"w/Autorun",label,100,9999
    SAVETAP "output.tap",BASIC,"withVars",label,123,9999,100
    SAVETAP "output.tap",CODE,"bank17",screen,6912
    SAVETAP "output.tap",CODE,"screen",demo,length,org
    SAVETAP "output.tap",NUMBERS,"dimArray",label,100      ; a()
    SAVETAP "output.tap",NUMBERS,"othernum",label,200,'b'  ; b()
    SAVETAP "output.tap",CHARS,"charArray",label,300       ; a$()
    SAVETAP "output.tap",CHARS,"nextone",label,400,'m'     ; m$()
    SAVETAP "output.tap",HEADLESS,start,length

SAVETAP <filename>[,<startadressofprogram>]

Works only in real device emulation mode. See DEVICE.

Save the tape file for emulators of ZX-Spectrum as a "snapshot" of almost-whole memory. Generated tape file supports the ZX-Spectrum clones with extended RAM such as ATM Turbo 512, etc. (If start address is omitted, the one provided by END is used)

The stored memory starts at $5E00 (with screen data at $4000 auto-detected and optionally stored into the TAP file). The stored memory is automatically trimmed of zero values at start and end of the region (put non-zero byte markers around region you want to store, if you have zero-ed bytes at beginning or end of your code).

Example 5.49. 

    ORG $7FFF
    DB $01      ; non-zero marker to store the following zero-ed data
DAT DS 1024, 0  ; zero-ed data at $8000 (1024 bytes)
START  ....     ; some code
    SAVETAP "game.tap", START ; save tape-snapshot to file game.tap. Start address is $8400

SAVETRD <filename_of_trd_image>,<filename_in_trdos>,<address>,<length>[,<autostart_BASIC_line>[,<length_minus_variables>]]
SAVETRD <filename_of_trd_image>,|<filename_in_trdos>,<address>,<length>[,<autostart_BASIC_line>]
SAVETRD <filename_of_trd_image>,&<filename_in_trdos>,<address>,<length>

Works only in real device emulation mode. See DEVICE.

Save the device memory into TRD disk image. Saving two files with same filename will emit warning, but two files (with same name) will be created in the disk directory.

Adding pipe character "|" ahead of file name will make sjasmplus to delete old file(s) with the same name before writing the new one => replace-like functionality. If the deleted file did occupy all sectors till the free space position in disc info, sjasmplus will salvage those sectors back and save new file over them (but it will not do full reshuffle/defrag in more complex cases, sjasmplus is just assembler, not full featured TRD images manipulation tool).

Adding ampersand character "&" ahead of file name will make sjasmplus to look for existing file with the requested name (last of them, any earlier duplicates are deleted). The new content is appended to the file (sector aligned append) and the catalog entry gets only number of sectors patched, up to 255 sectors at most. This is special mode for single-file big-loaders.

The unofficial three-letter extensions are also supported (official extensions are only: B, C, D and #).

Example 5.50. 

    EMPTYTRD "test.trd" ;create empty TRD image
    PAGE 7 ;set 7 page to current slot
    SAVETRD "test.trd","myfile1.C",$C000,$4000 ;- save 4000h begin from C000h of RAM to file to TRD image
    SAVETRD "test.trd","myfile2.C",$8000,$3000 ;- save 3000h begin from 8000h of RAM to file to TRD image
    SAVETRD "test.trd",|"myfile1.C",$B000,$400 ;- replace "myfile1.C" with new file
    SAVETRD "test.trd",&"myfile1.C",$9000,$734 ;- sector-append new data to "myfile1.C"

SETBP [<expression>]
SETBREAKPOINT [<expression>]

Works only in real device emulation mode. See DEVICE.

Will add execution-type breakpoint at address <expression> to the BPLIST export file. If no expression is specified, the current program counter will be used.

Example 5.51. 

    BPLIST "cmd_line.options.txt" zesarux ; open breakpoints list in "ZEsarUX" format
        nop     ; ZEsarUX will not catch very first instruction of snapshot file loaded
        SETBP   ; so breakpoint ahead of the second instruction (after first "nop")
        xor  a
        SETBP some_routine_label  ; set also some breakpoint to other code by its label
        SETBREAKPOINT $0D6B       ; alias of SETBP, works identically
        ; ... your code of app

SHELLEXEC <filename>[,<parameters>]

Execute external program <filename> using optional command line <parameters>.

Example 5.52. 

    OUTPUT "mybin.bin"
    ;some code
    IF ((_ERRORS = 0) && (_WARNINGS = 0))
        SHELLEXEC "x:/somepath/bin2tap.exe mybin.bin mytap.tap"
       ; or SHELLEXEC "x:/somepath/bin2tap.exe","mybin.bin mytap.tap"

SIZE <filesize in bytes>

If the resulting file is less than the given length, as many zero bytes are added as necessary. See OUTPUT for more.

Example 5.53. 

    SIZE 32768       ; make sure file will be 32K

SLDOPT <type> <arguments>

Type `COMMENT` - add comma delimited list of case sensitive keywords to be detected in end-of-line comments and exported to Source Level Debugging (SLD) data. The particular keywords depend on the debugger you are using to process SLD data.

Example 5.54. 

    ORG $8000
        call init ; Log("calling init") - gets exported to SLD
        ret       ; no keyword here - not exported to SLD
var1    DB 1      ; WPMEM - exported
var2    DB 2      ; WpMem - not exported, case sensitive!

SLOT <number_or_address>

Works only in real device emulation mode. See DEVICE.

Set current slot. Slots are defined by DEVICE pseudo-op. Use pseudo-op PAGE to change page in the current slot.

(since v1.18.1) You can also use starting address of particular slot instead of its number, ie. for ZX128: SLOT $8000 is alias of SLOT 2

Example 5.55. 

    SLOT 3 ;from 0C000h to 0FFFFh
    PAGE 1 ;set page 1 to slot 3
    ORG 0C000h
    ;your program here
    PAGE 2
    INCBIN "somegfx.bin"


Ends generating compiler output to tape file block specified in TAPOUT.

TAPOUT <filename>[,<flagbyte>]

Appends one tape block at the end of specified file. All following code will be assembled to this tape file block.

Default value of flagbyte is 255.

Example 5.56. bigfile.asm

    EMPTYTAP screen.tap

    TAPOUT screen.tap,0
    DB 3
    DB 'ScreenName'
    DW 6912
    DW 16384
    DW 32768

    TAPOUT screen.tap
    INCBIN screen.bin

This will create tap file with the screen.

TEXTAREA <address>

Synonym of DISP.


Removes the identifier defined by DEFINE

Example 5.57. 

    DEFINE Release 1

    IFDEF Release
      DISPLAY "Building release version"

    UNDEFINE Release

    IFNDEF Release
      DISPLAY "It's works!"

      DISPLAY "Yes, it's the sjasmplus!"

    UNDEFINE *  ; undefine all identifiers

      DISPLAY "It's not the sjasmplus??"


Synonym of ENT.

WHILE <expression>[, <guardian-counter>]

While expression evaluates to non-zero value, generates following lines until an ENDW (EDUP/ENDR alias) pseudo-op is encountered.

The optional guardian-counter value will abort the WHILE if there are more iterations, default guardian-counter is set to 100k.

Example 5.58. 

ptr = $4000
    WHILE ptr < $4020, 500 ; max 500 loops before error
        DB low ptr
ptr = ptr + 1
    ENDW    ; EDUP or ENDR works too, but ENDW is advised for WHILE

/* will emit bytes $00, $01, ..., $1F */

WORD <words>

Defines a word. Values should be between -32787 and 65536.

Example 5.59. 

    WORD 4000h,0d000h
    WORD 4,"HA"

Conditional assembly

It may be useful to assemble a part (or not) based on a certain condition.

IF <expression>

If <expression> is non-zero the following lines are assembled until an ELSE, ELSEIF or ENDIF.

Forward reference of label will cause warning - as any machine code emitted inside the IF/IFN block with such expression can lead to unstable results. If you are sure the late definition of label will not affect resulting machine code (the IF block does not emit any machine code and does not define any label), you can suppress the warning.

Example 5.60. 

    IF (aaa == 0) || (2 = aaa && ((bbb % 13) & 0x01))
        ; some code to assemble only if the symbol `aaa` is zero
        ; or (logical OR) when symbol `aaa` is two
        ; and (logical AND) modulo 13 of `bbb` is odd number

IFN <expression>

If <expression> is zero the following lines are assembled until an ELSE, ELSEIF or ENDIF.

IFDEF <id>

The condition is true if there is an id defined. These are NOT labels.

Example 5.61. 

        CALL InitOwnMM
        CALL InitDos2MemMan


The condition is true if there isn't an id defined. These are NOT labels.

Example 5.62. 

1   IN A,(0C4H)
    AND 2
        JR NC,1B

IFUSED <label>

The condition is true if there is an label used somewhere in the code. You can create libraries of useful functions using IFUSED pseudo-op

Example 5.63. (similar to tests/misc/ifused_test.asm)


    CALL LABEL3 ; LABEL3 - yes
    LD A,(LABEL1) ; LABEL1 - yes

    DB 1

    DB 2

    DB 3

    DB 4

    LD A,LABEL2 ; LABEL2 - yes


; Output will contain bytes from LABEL1 to LABEL3 (1, 2, 3), but not contain from LABEL4, because this label is not used.

; Alternative syntax:
    IFUSED ; sjasmplus will use name of previous label, i.e. LABEL5


Known bug: when code is using label inside module "moduleX", like call labelY, only usage of moduleX.labelY label is noted. Then if you define "labelY" outside of module and hide it inside IFUSED labelY block, the call from module will be unable to find the routine.

Workaround: you can use the global-label operator @: "call @labelY" to trigger usage of the global "labelY", or you can use the alternative IFUSED syntax "labelY: IFUSED" which does not only check condition, but also does define the label. Once the label is defined, the "call labelY" line inside module will find the global variant and mark it as "used" correctly.

Local labels inside the IFUSED block are not visible until the main label is used, if you want to call only local label, you can force the visibility of the routine by "ASSERT main_label" - this will trigger usage of the global "main_label", and then the IFUSED block will be assembled, and local labels will become visible too.

IFNUSED <label>

The condition is true if there is an label not used somewhere in the code.


See IF. If the IF condition is false (in all previous IF/ELSEIF blocks), the else-part is assembled.

ELSEIF <expression>

If the previous IF-block had false condition, the ELSEIF expression is evaluated and the ELSEIF acts as new IF.


Every IF should be followed by an ENDIF.


The MACRO pseudo-op defines a macro. It should be followed by the name of the macro, optionally followed by the parameters. The following lines will be stored as the macro-body until an ENDM pseudo-op is encountered. Macro's have to be defined before their use. Macro's name must be unique (even if two macros differ by required arguments).

Example 5.64. Macro without parameters (docs_examples/s_macros.asm)

    ADD A,L
    JR NC,.hup
    INC H
    LD L,A

Labels in a macro starting with a dot are local to each macro expansion.

Example 5.65. A macro with parameters (docs_examples/s_macros.asm)

  MACRO WAVEOUT reg, data
    LD A,reg
    OUT (7EH),A
    LD A,data
    OUT (7FH),A
; this macro will make
  WAVEOUT 2,17
; expand to:
  LD A,2
  OUT (7EH),A
  LD A,17
  OUT (7FH),A

Example 5.66. Another example (docs_examples/s_macros.asm)

      IF $-.lus<127
        DJNZ .lus
        DEC B
        JP NZ,.lus

    CALL DoALot
; This will expand to:
.lus                  ; Main.lus
    CALL DoALot
    DJNZ .lus         ; Main.lus

Angle brackets can be used when the arguments contain commas.

Example 5.67. Argument in angle brackets (docs_examples/s_macros.asm)

    MACRO UseLess data
      DB data

    UseLess <10,12,13,0>
; expands to:
    DB 10,12,13,0

; use '!' to include '!' and '>' in those strings.

    UseLess <5, 6 !> 3>
; expands to:
    DB 5, 6 > 3

    UseLess <"Kip!!",3>
; expands to:
    DB "Kip!",3

As compatibility convenience to make porting from different assemblers somewhat easier, there is alternative syntax, where the macro name is written at beginning of line (as if label, but MODULE part is NOT applied to macro name).

Example 5.68. Macro name at beginning of line (docs_examples/s_macros.asm)

LabelAsMacroName    MACRO  arg1?, arg2?
                        ld  a,arg1?
                        ld  hl,arg2?

                LabelAsMacroName 1,$1234
            ; expands to:
                ld a,1 : ld hl,$1234

If some macro over-shadows regular instruction or directive name, the @ character in front of instruction/directive name can be used to inhibit macro expansion.

Example 5.69. Inhibit macro expansion operator (docs_examples/s_macros.asm)

djnz    MACRO   arg1?
            dec c
            jr  nz,arg1?
            @djnz arg1? ; avoid self-reference and use real instruction

1:      djnz    1B      ; macro replacement will be used here
1:      @djnz   1B      ; original djnz instruction here

Chapter 6. Structures

What is it?

Structures can be used to define data structures in memory more easily. The name of the structure is set to the total size of the structure.

Defining structure

A structure definition starts with: STRUCT <name>[,<initial offset>] and ends with ENDS. Structure definitions are local to the current module, but, as with labels, '@' can be used to override this.

Lines between STRUCT and ENDS should have the following format:

membername pseudo-operation operands

All fields are optional. Lines without label should start with whitespace.

When non zero offset is used, it acts as if BLOCK with length equal to offset was defined as first member of structure.

STRUCT instructions

Between the STRUCT and ENDS pseudo-instructions the following instructions can be used:

BYTE [<defaultvalue>]

To define a one byte member. The defaultvalue is used when no initialisation value is given when the structure is declared. (DB and DEFB may be used instead of BYTE).

WORD [<defaultvalue>]

To define a two byte member. The defaultvalue is used when no initialisation value is given when the structure is declared. (DW and DEFW may be used instead of WORD).

D24 [<defaultvalue>]

To define a three byte member. The defaultvalue is used when no initialisation value is given when the structure is declared.

DWORD [<defaultvalue>]

To define a four byte member. The defaultvalue is used when no initialisation value is given when the structure is declared. (DD and DEFD may be used instead of DWORD).

TEXT <length>[,{<DB-like operand(s)>}]

(since v1.17) To define a member of the specified length of bytes. The operands are DB-like formatted, and the last byte explicitly defined will fill the remaining bytes reserved upon definition (zeroed when no value is specified at all).

The default value can be modified with init values when structure is used, but then only the given bytes overwrite default content, NOT filling remainder of it.

While the operands are DB-like formatted, they must be enclosed in curly braces. Also make sure the braces are not mistaken with the optional braces on the sub-structure boundaries, when using TEXT member in sub-structure, you may need to use curly braces for the sub-structure itself around of curly braces used for TEXT values (making the sub-structure braces mandatory in such case). The length can be from 1 to 8192 bytes (if you need even longer "text", use two members and split your arguments).

BLOCK <length>[,<fillbyte>]

To define a member of the specified number of bytes. Arguments are set when defining the current structure, and are not part of init values when the structure is later used. ('#', DS and DEFS may be used instead of BLOCK).

(since v1.11) If fillbyte is omitted, the device memory content in the block area is preserved (not zeroed).

ALIGN [<expression>[, <byte>]]

To align the offset. If the expression is omitted, 4 is assumed. ('##' May be used instead of ALIGN).

(since v1.11) If the byte is omitted, device memory content is preserved (not zeroed).

<structure name> [<init values>]

It is possible to nest structures, and give new defaults for the BYTE and WORD members.

Usage of defined structure

[<label>] <struct_name> [<initial values>] will emit full structure into machine code, either using default values from structure definition, or overriding them with explicit value from the <initial values> list. In initial values you can use curly brackets {} to group particular initial values for particular sub-structure, any missing values in particular sub-structure init-list are set up by default values of particular field. See "SDOT" example below or tests/struct asm files for more examples.

<label> <struct_name> = <expression> will only set up <label>.<struct_field> labels starting from designed address provided by expression, but there will be no machine code emitted (and current address "$" will not advance).

(since v1.17) The initial values inside curly braces block can now span over multiple lines with new line ending current value expression (the comma after expression is optional in such case). Please bear in mind this is pushing original sjasmplus architecture beyond its original design (multi-line definitions) and may fail in complex scenarios (for example the "dot-repeater" macro mechanics will repeat only the remainder of current line, ignoring the curly-braces block spanning across following lines). If you have example of code which fails but you believe it's reasonable use case, please open the issue on github with such example, so it is possible to improve the feature in future (the dot-repeater is "won't fix" case, but otherwise the feature should work reliably).


Example 6.1. docs_examples/c_structures.asm


This is identical to:

SCOLOR		EQU 3 ; length
SCOLOR.RED	EQU 0 ; offset
SCOLOR.BLUE	EQU 2 ; offset

Example 6.2. docs_examples/c_structures.asm

C	SCOLOR 0,0,0 ; use new default values

This is identical to:

SDOT		EQU 5 ; length
SDOT.X		EQU 0 ; offset
SDOT.Y		EQU 1 ; offset
SDOT.C		EQU 2 ; offset
SDOT.C.RED	EQU 2 ; offset
SDOT.C.GREEN	EQU 3 ; offset
SDOT.C.BLUE	EQU 4 ; offset

Example 6.3. docs_examples/c_structures.asm


This is identical to:

SPOS	EQU 10 ; length
SPOS.X	EQU  4 ; offset
SPOS.Y	EQU  6 ; offset
SPOS.AD	EQU  8 ; offset

Example 6.4. docs_examples/c_structures.asm

When a structure is defined it is possible to declare labels with it


This is identical to:


Note the default values.

Or without label:

  SCOLOR 0,0,0
  SCOLOR 1,2,3
  ; etc.

This is identical to:

  BYTE 0,0,0
  BYTE 1,2,3
  BYTE 4,2,6
  ; etc.

DOT1 SDOT 0,0, 0,0,0     ; or 0,0,0,0,0 or {0,0,{0,0,0}}

Only BYTE and WORD members can be initialised.

The resulting labels can be used as any other label:

  ld b,(ix+SCOLOR.RED)
  ld a,(COLOR.GREEN)
  ld de,COLOR
  ; etc.

Example 6.5. docs_examples/st_usage_example.asm

value1	BYTE
value2	WORD

        ORG  0x8000
binData BIN_FILE_MAP = $        ; set up label values only (no bytes)
        INCBIN "some_data.bin"  ; load the bytes from file instead

        ; using the data through struct definition
        ld  a,(binData.value1)
        ld  hl,(binData.value2)

This is identical to:

BIN_FILE_MAP            EQU 259      ; length
BIN_FILE_MAP.value1     EQU 256      ; offset
BIN_FILE_MAP.value2     EQU 257      ; offset
; labels to access binary data loaded by INCBIN
binData                 EQU 0x8000   ; address
binData.value1          EQU 0x8100   ; address
binData.value2          EQU 0x8101   ; address

Example 6.6. docs_examples/st_text.asm

length      WORD    BLOCK_HEADER
type        BYTE    $AB
name        TEXT    10, { "none", 32, '!' } ; will produce "none !!!!!" default data
                    ; because this is definition of the field, here last byte is "filler"
datastart   WORD
datalen     WORD
checksum    BYTE    $CC

        ORG  0x8000
head1   BLOCK_HEADER {      ; Multi-line initialization requires curly braces.
    , ,                     ; Keeping default length and type by specifying empty values.
    { 'New',                ; The final `name` data will be "New Name!!"
        32,                 ; overwriting only 8 bytes of default data.
        "Name" },           ; The last "e" is NOT "filler" in the non-default value.
        $8000, $1234        ; Explicit datastart and datalen values.
}                           ; End of initial data block ("checksum" keeps default value).

; machine code (struct data):
; 12 00 AB 4E 65 77 20 4E 61 6D 65 21 21 00 80 34 12 CC
; = { $0012, $AB, "New Name!!", $8000, $1234, $CC }


Do not use the offset labels in indirections like:


This will conflict with futher 'improvements' ;-)

If this is absolutely necessary (why?) use something like this:


Chapter 7. Lua scripting


Why is scripting engine as Lua embedded to the assembler? Answer is simple: it can be used to add extra features by users. And Lua is small enough, fast and powerful scripting engine.

Sjasmplus currently includes Lua 5.4.6 (since v1.20.3). The upgrade from Lua 5.1 in v1.20.0 was done along replacing tolua++ with LuaBridge2.6 and while there was lot of effort to maintain original Lua bindings to sjasmplus, there were few changes to the API.

If your old source does not work any more after upgrade, please carefully check against the bindings documentation below.

Check if your code does use only official API, and if you were using secret extra arguments, check the test files and examples how those lua scripts were adjusted to build with v1.20.0. The bindings API documentation has been cleaned up, some API extended and has better test-coverage to make it less likely to change in the future.

Also please mind the breaking changes of the Lua itself, and adjust your scripts to Lua 5.4 syntax and library functions, there is example below with common issues already encountered (open github issues with others so we can add them to docs).

How to use?

You must use LUA and ENDLUA pseudo-ops.

Example 7.1. Hello World!

        print ("Hello World!")

SjASMPlus functions bindings to Lua

From Lua you can control some variables and use functions of the compiler. Complete list:

[integer] _c("expression")

Calculate expression using calculator of the compiler (32-bit only). Example: val = _c("SOMELABEL+12")

[void] _pc("code")

Parse string of Z80 assembly. Example: _pc("ADD A,B")

[void] _pl("label code")

Parse line of Z80 assembly. Example: _pl("SOMELABEL ADD A,B")

This binding makes almost all pseudo-instructions (directives) available from Lua script with all options and features. In the future the bindings API will be mostly frozen, extended only in case where using _pl(...) is too cumbersome.

[integer] sj.calc("expression")

See _c

[void] sj.parse_code("code")

See _pc

[void] sj.parse_line("label code")

See _pl

[void] sj.error("message","bad_value" = nil)

Print error message.

[void] sj.warning("message","bad_value" = nil)

Print warning message.

[boolean] sj.file_exists("filename")

Check for file exists.

[string] sj.get_define("name", "include_macro_args" = false)

Get define value, returns nil if define is not found. Can optionally search also macro arguments (with higher priority).

[boolean] sj.insert_define("id", "value" = "")

Add new define. Returns true if the define is new, false when define already existed and value was replaced or when id is invalid.

[integer] sj.get_label("name")

Get label address. Returns -1 if name is invalid, 0 when not found or has still undefined value (you can use also _c("name") to get symbol value and do some calculations with it).

[boolean] sj.insert_label("name", address)

Add new label. Returns true when label was successfully inserted (or value modified first time in next assembling pass), false when name is invalid or label was already defined.

[integer] sj.current_address

Read-only variable. Current address.

[integer] sj.error_count

Read-only variable. Count of Errors.

[integer] sj.warning_count

Read-only variable. Count of Warnings.

[void] sj.exit(errorcode = 1)

Shutdown the compiler.

[void] sj.add_byte(byte)

Add byte to output (or to memory) and increase sj.current_address

[void] sj.add_word(word)

Add word to output (or to memory) and twice increase sj.current_address

[integer] sj.get_byte(address)

Get byte from memory. Works only in real device emulation mode and only in last assembling pass.

[integer] sj.get_word(address)

Get word from memory. Works only in real device emulation mode and only in last assembling pass.

[string] sj.get_device()

Return current emulating device's identifier. Returns "NONE" if no emulation mode.

[boolean] sj.set_device("id" = "NONE", ramtop = 0)

Set current emulating device's identifier. Returns false if requested device id is not found and current device set is lost ("NONE" is set).

[boolean] sj.set_page(number)

Set page "number" to the current slot. Works as pseudo-op PAGE.

[boolean] sj.set_slot(number_or_address)

Set current slot to "number". Works as pseudo-op SLOT.

(since v1.20.0) You can also use starting address of slot instead, ie. for ZX128 sj.set_slot(0x8000) is alias of sj.set_slot(2).

[void] sj.shellexec("programname")

See pseudo-op SHELLEXEC.

[boolean] zx.trdimage_create("trd_name", "label" = "")

Creates empty TRD image file.

[boolean] zx.trdimage_add_file("trd_name", "somenameC", startaddress, length, autostart = -1, replace = false)

Save block of memory to TRD image file. Work only in real device emulation mode.

[boolean] zx.save_snapshot_sna("filename.sna", startaddressofprogram)

Save snapshot of memory in SNA format. Works only in real device emulation mode and only for ZXSPECTRUM48 and ZXSPECTRUM128..

Third-party embedded library(ies)

None at this moment.

Since v1.20.0 the sjasmplus integrates Lua 5.4.4 (now 5.4.6).

The libraries included in older versions of sjasmplus: lpack.c, bit (bitwise operators) and hex (convert to/from hexa numbers) are now all part of the standard Lua 5.4.

Refer to Lua documentation for full list of functions and detailed help.

For small example check the test tests/lua_examples/trivial/lua_standard_libs.asm


Example 7.2. Variables doesn't clear in new passes of the compiler

       v = 1

       print (v)
-- out to console: 1
       v = v + 1

       print (v)
-- out to console: 2

Example 7.3. To generate some code you need to generate it in all passes

        _pl("ClearScreen LD (.savesp+1),SP")
        _pc("LD SP,16384+6144")
        _pc("LD HL,0")
        for i = 32768, 38912, 2 do
            _pc("PUSH HL")
        _pl(".savesp: LD SP,0")

    LUA PASS2 ; if you fully understand what you are doing
        sj.add_byte(123) -- and you need emit bytes in other mode
    ENDLUA ; luamc-ok ; you can suppress warning here or at start of block

Example 7.4. Declare function and use it

         function savetape_mytype(filename, startaddress)
             local fp
             fp = assert(, "wb"))
             for i = 16384, 32767, 4 do
                 assert(fp:write( string.pack("bbbb",
                                sj.get_byte(i+3)) ))

 ;somewhere in your program
         savetape_mytype("tapefiles/myprogram.tape", _c("StartGameLabel"))

Example 7.5. Simple sample :)

-- Function reads number from file <fname>, increases it, creates define "BUILD" with the number and saves the number to <fname>.
-- With this function you can control count of compilations.
	function increase_build(fname)
		local fp
		local build
		fp = assert(, "rb"))
		build = tonumber(fp:read("*all"))
		if type(build) == "nil" then
		    build = 0
		build = build + 1;
		sj.insert_define("BUILD", build)
		fp = assert(, "wb"))
		assert(fp:write( build ))

-- Before using you must create empty file "build.txt"!

-- Creates define "TIME" with current time
	sj.insert_define("TIME", '"' .."%Y-%m-%d %H:%M:%S") .. '"')

; print to console our time and build number
	DISPLAY "Build time: ", TIME
	DISPLAY "Build number: ", /D, BUILD

Example 7.6. Common issues with upgrade to Lua 5.4

    x = 2.3 ^ 4.5             -- old syntax: math.exp(2.3, 4.5)
    _pc("db " .. math.floor(x))
    _pc("db " .. 35//7)       -- old syntax: 35/7
      -- floating point values are now formatted rigorously with ".0" from the Lua
      -- which leads to error in sjasmplus parser: Unexpected: .0
      -- and requires to enforce integer type already at lua side
      -- you can use either `math.floor()` function, or integer
      -- operators like integer divide `//` or left shift `1<<7`

    -- the floating point values (like "12.0") are silently truncated since v1.20.1 (to "12"),
    -- but v1.20.0 errors out on these, unless you patch the lua script to produce integers

Chapter 8. SAVENEX guide

NEX File Format

NEX is binary format for ZX Spectrum Next, aiming to provide simple delivery of software for the platform. For file format details check In short it is header + loading screen + like-snapshot binary and remaining resources appended after.

Sjasmplus currently supports V1.2 and V1.3 of NEX file format (see wiki for details).

As such, the SAVENEX commands are available only in ZXSPECTRUMNEXT device emulation mode, see DEVICE.

As the file is designed for self-contained distribution of whole applications/games, its configuration and assembling is a bit too complex for single directive, and the configuration is instead divided into multiple commands, and the assembling goes through multiple stages, so some commands must be used in correct sequence.

While the format technically allows to include multiple screen types data, they are all loaded at the beginning over each other, so it makes sense to provide only single loading screen (sjasmplus enforces that).

Detailed description of each SAVENEX command

SAVENEX OPEN <filename>[,<startAddress>[,<stackAddress>[,<entryBank16k 0..111>[,<fileVersion 2..3>]]]]

Opens a NEX file, defines start address, stack address and 16k bank to be mapped at 0xC000 before code is executed (if values are omitted, start address is zero = no start, stack address is 0xFFFE, entryBank is zero, fileVersion is 2).

Only single NEX file can be open at the same time, and to finalize the header content the command CLOSE has to be used (does auto-close if source ends).

Entry bank is number of 16k bank (0..111), not native 8k page, default is zero, i.e. the default memory map is identical to ZX 128 (ROM, RAM banks 5, 2 and 0).

Make sure your new stack has at least tens of bytes available as those will be used already by the NEX loader before executing your entry point (although released back).

The fileVersion can be 2 (NEX V1.2) or 3 (NEX V1.3), which will enforce the specified version of file. Otherwise the file is V1.2 by default and will auto-switch to V1.3 when some V1.3 feature is configured/used. When version 2 is enforced, any usage of V1.3 feature will emit error.

SAVENEX CORE <major 0..15>,<minor 0..15>,<subminor 0..255>

Set minimum required Next core version, can be set any time before CLOSE.

SAVENEX CFG <border 0..7>[,<fileHandle 0/1/$4000+>[,<PreserveNextRegs 0/1>[,<2MbRamReq 0/1>]]]

Set border colour (during loading), whether the machine should be set to default state (PreserveNextRegs = 0 = default), if the app requires extended RAM (224 8k pages) and how the file handle of the NEX file should be treated: 0 = default = close, 1 = keep open and pass in BC, $4000..$FFFE = keep open, and write into memory at provided address (after entry bank is paged in). This can be set any time before CLOSE.

SAVENEX CFG3 <DoCRC 0/1>[,<PreserveExpansionBus 0/1>[,<CLIbufferAdr>,<CLIbufferSize>]]

All of these are NEX format V1.3 features and using "CFG3" command will change the version to V1.3 automatically (if not specified by OPEN).

DoCRC: default = 1, sjasmplus will checksum the file (if you know you will be further patching file afterwards, switch it off, otherwise keep 1).

PreserveExpansionBus: default = 0, 0 = the NEX loader will disable expansion bus through NextReg $80 / 1 = the NEX loader will not do anything.

CLIbufferAdr, CLIbufferSize: default = [0, 0], address and size of buffer for NEX loader to copy the arguments line into. The buffer is copied after the "entryBank" is mapped, so you can allocate buffer in your entryBank. The loader/format has hard limit 2048 bytes, the argument line will be truncated if longer. If your reserved buffer is shorter than 2048 bytes, the copy will be also truncated to your buffer size only. If not truncated, the argument line can end with any of these: Enter (13), colon or zero byte (truncated has no terminator character, just fills the full buffer).

SAVENEX BAR <loadBar 0/1>,<barColour 0..255>[,<startDelay 0..255>[,<bankDelay 0..255>[,<posY 0..255>]]]

Loading-bar related setup ("colour" usage depends on screen mode), can be set any time before CLOSE.

The posY argument will apply only to V1.3 Layer2 screens in 320x256 and 640x256 resolution (then defaults to 254 = bottom of screen, 2px tall bar).



SAVENEX PALETTE MEM <palPage8kNum 0..223>,<palOffset>


This is optional command to set palette in alternative way ahead of SCREEN command, with higher priority (if the PALETTE command is used, following SCREEN commands will ignore the palette arguments and keep the palette defined by this command). You can be use it between OPEN and SCREEN command. But appropriate SCREEN type supporting palette must be defined by SCREEN command too.

Palette consists of 512 bytes (256 palette items from index 0), in 9b colour format: first byte is %RRRGGGBB, second byte is %P000000B (P is priority flag for Layer 2 colours).

The DEFAULT palette will generate colour values from the colour index, the same way how default Layer 2 palette is initialized on the ZX Spectrum Next.

The NONE palette will force the "no palette" flag even if SCREEN command later does specify some palette (it will be still ignored).

SAVENEX SCREEN L2 [<Page8kNum 0..223>,<offset>[,<palPage8kNum 0..223>,<palOffset>]]

Layer 2 loading screen, can be used between OPEN and first AUTO/BANK command.

Palette consists of 512 bytes (256 palette items from index 0), in 9b colour format: first byte is %RRRGGGBB, second byte is %P000000B (P is priority flag for Layer 2 colours).

Image data are 48kiB block of memory, the loader will use always banks 9..11 to display it (8k pages 18..23), but if you will prepare the data there, it will be also re-saved by AUTO command, so either use other banks, and overwrite them with valid data/code after using the SCREEN command, or reset pages 18..23 to zero after SCREEN.

If no memory address is specified, the pages 18..23 are stored in file, and if no palette address is specified, no-palette flag is set in NEX file.

SAVENEX SCREEN L2_320 [<Page8kNum 0..223>,<offset>[,<palPage8kNum 0..223>,<palOffset>]]

SAVENEX SCREEN L2_640 [<Page8kNum 0..223>,<offset>[,<palPage8kNum 0..223>,<palOffset>]]

Works same way as "L2" variant, but will set up screen for new resolutions 320x256x8 and 640x256x4. The difference is that the required image data are 80kiB (five 16kiB banks; equals ten 8kiB pages), the banks used to show the screen in loader are 9..13 (8k pages 18..27 - that's also the default address of data if not specified) - avoid these in regular banks stored in the file.

These are NEX format V1.3 features and using them will change the version to V1.3 automatically (if not specified by OPEN). This command doesn't allow to set specific palette offset value (too lazy to add it, use BMP variant if you really need it).

The data has to be already in correct format and organized as if displayed (the "transposed" bitmap where +1 address goes to pixel below, and +256 goes to pixel on the right (or pair of pixels in case of 640x256x4bpp mode), this command will just dump them into file "as is".

The loading bar colour byte will be also used "as is", which in 4bpp mode means the byte does define pair of 4 bit pixels, i.e. if you want solid-colour "3" loading bar in 4bpp mode, define it as value $33. Keep also in mind the default loading bar position is Y=254, which on most of the TV/LCD displays is outside of visible range, the reasonably "safe" (visible on almost all of the screens) resolution is about 288x224 (+16px around PAPER area, +24px is visible on many displays too), you may want to organize your screen in a way to show all important information within this area, and make the rest "unimportant" so it can hide beyond the edge of screen.

SAVENEX SCREEN LR [<Page8kNum 0..223>,<offset>[,<palPage8kNum 0..223>,<palOffset>]]

LoRes (128x96) loading screen, can be used between OPEN and first AUTO/BANK command.

Palette is similar to Layer 2 mode, just LoRes mode doesn't have priority bit.

Image data are 12288 bytes memory block - either consecutive block if specific address is provided, or without address the actual bank 5 memory is stored (taking 6144 bytes from address 0x4000 and 6144 bytes from address 0x6000).

SAVENEX SCREEN BMP <filename>[,<savePalette 0/1>[,<paletteOffset 0..15>]]

Only small subset of BMP files can be used: 256x192 (Layer 2) or 128x96 (LoRes), indexed (8bit image data with palette) and palette data will be truncated to 3:3:3 color space directly (no smart colour quantization or dithering is applied). And the file must be uncompressed.

For V1.3 NEX files you can include also 320x256 and 640x256 files (Layer 2), the 640x256 should be also 8bit indexed, but only 4 bits of pixel data will be used (256 colour palette is legitimate and will be stored "as is" in NEX file). These two new modes can also include paletteOffset argument 0..15 (does not apply to V1.2 BMP files above), the offset is added to top four bits of pixel value.

The BMP will be included as loading screen, can be used between OPEN and first AUTO/BANK command.

By default the palette from BMP file is used, but you can override that by savePalette = 0.

SAVENEX SCREEN (SCR|SHC|SHR) [<hiResColour 0..7>]

ULA/Timex modes loading screen, can be used between OPEN and first AUTO/BANK command.

The actual bank 5 memory (pages 10..11) is stored as if the image is displayed, in these modes the palette can't be specified.

SCR is classic ZX 6912 bytes long screen from address 0x4000 (page 10 is used, even if the slot 1 is modified to other page, so you must prepare the image "in place").

SHC and SHR are Timex HiColor (8x1 attribute) and HiRes (512x192 bitmap) modes, prepare data "in place", i.e. 6144 bytes into page 10 and 6144 bytes into page 11 (0x4000 and 0x6000 addresses in default memory setup). For HiRes mode you should specify ink colour (the paper is complement of ink).

SAVENEX SCREEN TILE <NextReg $6B>,<NextReg $6C>,<NextReg $6E>,<NextReg $6F>[,<AlsoStoreBank5 0/1 = 1>]

NEX V1.3 tilemap loading screen, can be used between OPEN and first AUTO/BANK command.

To define palette use the PALETTE command.

The image data are stored as regular Bank 5 of the NEX file (which is the first bank to be loaded by loader), depending on AlsoStoreBank5 value (default 1), this SCREEN command will also execute BANK 5 command to store the image data.

The NextRegisters $6B, $6C, $6E and $6F should be enough to specify any variant of tilemap mode, so the precise sub-type and image data layout is under control of user, the sjasmplus doesn't enforce any particular configuration.

If you want to use also COPPER command, use it either ahead of the SCREEN TILE command, or use "AlsoStoreBank5 = 0" to delay the storage of Bank 5 data. In such case you must then later explicitly store the Bank 5 as regular bank, either with BANK or AUTO command.

SAVENEX COPPER <Page8kNum 0..223>,<offset>

Can be used after OPEN and before first AUTO or BANK command (the copper data are stored between screen and bank data).

Exactly 2048 bytes are stored (full Copper memory), and the loader will start the copper code in mode %01 (reset CPC to 0, then starts the copper), so the copper code will wrap around infinitely after the 1024th instruction executed. The copper is started after the screen block is loaded and displayed.

SAVENEX BANK <bank16k 0..111>[,...]

Can be used after OPEN or SCREEN and before CLOSE, but the 16ki banks must be saved in correct order: 5, 2, 0, 1, 3, 4, 6, 7, 8, 9, 10, ..., 111

SAVENEX AUTO [<fromBank16k 0..111>[,<toBank16k 0..111>]]

Can be used after OPEN or SCREEN and before CLOSE. The sjasmplus will save every 16k bank containing at least one non-zero byte; detected in the correct order (automatically will save every non-zero 16k bank detected in the correct order (automatically starting from first possible bank after previous BANK/AUTO commands, or from provided "fromBank").

For "fromBank" value use the specified order above in BANK command, i.e. 5, 2, 0, ...

SAVENEX CLOSE [<fileToAppend>]

Can be used after OPEN. The currently open NEX file will be finalized (header adjusted), and optional extra file just appended to the end of NEX file.


Example 8.1. docs_examples/s_savenex_examples.asm

Creating NEX file which will have Layer2 loading screen (stripes), progress bar, and will enter infinite loop with calling stack (used by IM 1 interrupt handler) visible on the Layer 2 screen.

    ORG $7E00
start:  ei : jr $           ; app code entry point, BC = NEX file handle
    ; Layer2 screen (top 1/3 defined, bottom of it will be used also as "visible" stack)
    ORG $C000 : DUP 64*32 : DB $90,$91,$92,$93,$94,$95,$96,$97 : EDUP

    ; write everything into NEX file
    SAVENEX OPEN "example.nex", start, $FFFE, 9  ; stack will go into Layer2
    SAVENEX CORE 2, 0, 0        ; Next core 2.0.0 required as minimum
    SAVENEX CFG 4, 1, 0, 1      ; green border, file handle in BC, reset NextRegs, 2MB required
    SAVENEX BAR 1, $E0, 50, 25  ; do load bar, red colour, start/load delays 50/25 frames
    SAVENEX SCREEN L2 0, 0      ; store the data from C000 (page 0, offset 0), no palette
    SAVENEX BANK 5, 100, 101    ; store the 16ki banks 5 (contains the code at 0x7E00), 100, 101
    SAVENEX CLOSE               ; (banks 100 and 101 are added just as example)

Chapter 9. Source Level Debugging (SLD) data

What is it?

SLD data are extra "tracing" data produced during assembling for debuggers and IDEs, similar to "map" files already supported by sjasmplus ( LABELSLIST and CSPECTMAP).

The debugger can read these data, and with non-tricky source producing machine code with correct device memory mapping, the debugger can trace the origins of every instruction back to the original source code line, and display the source instead/along the disassembly view (the "map" files mentioned above provide only list of labels which is usually already super helpful, but can't track the source origins of each instruction).

The original impulse and working patch for this feature came from Chris Kirby, adding the single-instruction-step feature to his development tools: Next Development System (currently in 2019 under heavy development).


On the sjasmplus side all you need is to add another command line option when starting the assembler.

prompt$ sjasmplus --sld=project.sld.txt file1.asm file2.asm

This will produce along regular files also file `project.sld.txt` containing the tracing data. The file format is text-like, so the content can be viewed in any text editor, but it's supposed to be processed by the debugger.

The SLD data are being exported only for machine code produced within one of the virtual devices (see DEVICE). And the accuracy of the data directly depends on the state of the virtual device at the moment when the particular instruction is assembled, see next section for further advice how to get best tracing data.

If the option `--sld` without explicit filename is used, the first input source filename will be copied and its extension changed to `.sld.txt`, which should work well for single-main file type of projects.

How to write "non tricky" source

For best results (tracing data covering all your instructions and making source-level debugging available all the time):

Generate machine code only with instructions, i.e. while `db 0` in source will produce `nop` instruction, the tracing data for such `nop` will be missing. Only source line containing the "nop" will emit both the zero byte into machine code and tracing data.

Keep the memory map of the virtual device as it will be set at run-time when writing particular code. While using only regular ORG to place the code in desired memory area, keeping only the modified area with memory pages mapped-in is enough to get correct SLD data - but when using DISP directive to generate code in displaced way, you need to map-in correct pages in both memory areas (where the code is currently assembled, and where it is supposed to be operational at run-time - currently impossible if the two areas share the same SLOT area).

To map-in correct pages you can use directives: SLOT, PAGE, MMU and ORG

Only labels defined in regular way get also the memory-page data in SLD file, EQU values are exported without the page information, even if they represent memory address:

regularLabel:   nop             ; current page exported to SLD data
equLabel        EQU   $8000     ; only value without page is exported
varSymbol       =     14        ; =/DEFL labels are omitted completely

Re-using the same memory area (page + address) for two different pieces of code (for example by having multiple routines targeting the same address with DISP, and run-time code uploading the correct variant dynamically at the destination) will highly likely confuse the debugger, producing two source-origins data for the identical address. You may want to avoid this, especially with early version of tools not being able to resolve such ambiguity in reasonable way (the sjasmplus will generate tracing data ok even in this case, but they are tricky to interpret).

If you are using the colons to put multiple instructions on single line, verify the debugger can cope with the tracing data containing full line + column begin/end information, the simple/early tools will likely highlight only whole line of source. For a start keeping single instruction per line may keep things simple.

The various code generators written in macros or Lua scripts will produce accurate tracing data in case of macros, and condensed (pointing at "ENDLUA" line) data for Lua generators, but it still depends on the debugger if it can display both top level source code triggering the generator and source lines containing definitions of particular instructions.

SLD File Format definition (version "1")

The file format is still under development. If you are working on support in debugger, stay in touch with us, so we can discuss any modifications to current format with you.

The SLD data is text-file (sjasmplus is using UNIX-like newlines 0x0A, but parsers should rather cope with any of common EOL scheme). The general format is CSV-like using pipe character as delimiter between fields.

If the first field is empty, the line is one of the special control lines - the second field selects type of control line. If also the second field is empty, the remaining part of line should be ignored (comment line). Currently only one type of "control line" exists, the "" (third field is integer number):

||anything ... comment-like line

The file format version line should be always first line of the SLD data file.

The regular tracing data lines have fixed amount of fields: eight. The seventh field contains single uppercase letter defining type of the line, which affects sub-variants of format for eighth field, but first six fields share the same formatting and meaning across all types of regular lines:

<source file>|<src line>|<definition file>|<def line>|<page>|<value>|<type>|<data>

<source file> - file name of top-level source emitting the instruction/label/device.

<src line> - line and characters in top-level source file, the precise format is "<line>[:<column begin>[:<column end>]]" where first number is line number (starting at 1). The two following numbers delimited by colon are optional, representing the column where the segment starts/ends at the line. The column values are starting from 1 too, the "end" column is pointing beyond the current segment. The columns are in "bytes", i.e. tabulator does increase the column value by +1 only, and the "end" may actually point well beyond "strlen(line)".

<definition file> - file name of the source defining the particular instruction (for example where the instruction inside MACRO was originally defined). If empty, the definition-file name is identical to <source file> (common case for single-source projects).

<def line> - zero when there is no extra "definition" of instruction involved. When non-zero, the format is identical to <src line>, but with regard to the file specified in <definition file> field.

<page> - number of memory page where the <value> address points to, or -1 when value is not memory address.

<value> - any 32 bit integer value (EQU), but when representing memory address, it is full 16 bit value from Z80 address space (the top bits beyond the memory page size contains the number of "slot" where the instruction/label resides).

<type> - single uppercase letter representing type of line, current types: T (instruction Trace), L (Label), F [deprecated, parse L], D [deprecated, parse L], Z (device memory model), K (keyword comment, see `SLDOPT COMMENT`).

<data> - extra data specific for particular line type. Empty for "T" lines. For "L" a list of items separated by comma, where first three items represent sub-parts of symbol name: module name, main label, local label (sub-part may be empty), optional remaining items are usage-traits giving away various internal info about usage of symbol (currently exported traits: +equ, +macro, +reloc, +reloc_high, +used, +module, +endmod, +struct_def, +struct_data). For type "K" the comment containing one of they defined keywords is present.

For type "Z" the data field contains string describing the device memory model which is being selected in source and should be applied for following SLD data lines. The memory model string format is:

pages.size:<page size>,pages.count:<page count>,slots.count:<slots count>[,slots.adr:<slot0 adr>,...,<slotLast adr>]
// unsigned <page size> is also any-slot size in current version.
// unsigned <page count> and <slots count> define how many pages/slots there are
// uint16_t <slotX adr> is starting address of slot memory region in Z80 16b addressing

Example 9.1. Example of SLD file

|| ZX Spectrum Next device description:
|| label "main" points to 32768, with page 14 mapped-in (defined in toplevel.asm:62)
|| the label "main" is used by other code (IFUSED-like), and has relocation data
|| instruction opcode at 32768 (page 14) was created by toplevel.asm:64
|| instruction opcode at 32769 (p 14) was created by toplevel.asm:67
|| (but it is a line using macro, the instruction was defined by toplevel.asm:52)
|| instruction opcode at 32770 (p 18) was created by toplevel.asm:68:12-24
|| instruction opcode at 32771 (p 18) was created by toplevel.asm:68:24 (till EOL)
|| "PORT_NUMBER EQU 254" defined at toplevel.asm:69
|| label+instruction emitted from toplevel.asm:70, but defined in include.asm:3
|| with macro label "0>macro_defined_in_include_asm" (source ".macro_defined_in_include_asm:")



Conditional assembly, Conditional assembly
CSPECTMAP, Pseudo-ops


D24, Pseudo-ops, STRUCT instructions
DB (see BYTE)
DC, Pseudo-ops
DD (see DWORD)
DEFARRAY, Pseudo-ops
DEFARRAY+, Pseudo-ops
DEFDEVICE, Pseudo-ops
(see also DEVICE)
DEFG (see DG)
DEFH (see DH)
DEFINE, Pseudo-ops
DEFINE+, Pseudo-ops
DEFL, Pseudo-ops
(see also EQU)
(see also DISP)
DEVICE, Pseudo-ops
(see also DEFDEVICE)
(see also MMU)
(see also ZXSPECTRUM128)
DG, Pseudo-ops
DH, Pseudo-ops
DISP, Pseudo-ops
(see also ENT)
DISPLAY, Pseudo-ops
DM (see BYTE)
DS (see BLOCK)
DUP, Pseudo-ops
DW (see WORD)
DWORD, Pseudo-ops, STRUCT instructions
DZ, Pseudo-ops


ELSE, Conditional assembly
(see also IF)
ELSEIF, Conditional assembly
(see also IF)
EMPTYTAP, Pseudo-ops
EMPTYTRD, Pseudo-ops
ENCODING, Pseudo-ops
END, Pseudo-ops
ENDIF, Conditional assembly
ENDLUA, Pseudo-ops
(see also LUA)
ENDMODULE, Pseudo-ops
(see also MODULE)
ENDT (see ENT)
ENT, Pseudo-ops
(see also DISP)
EQU, Pseudo-ops
(see also DEFL)
exist, Expressions
EXPORT, Pseudo-ops


FPOS, Pseudo-ops


HEX, Pseudo-ops
(see also DH)
high, Expressions


LABELSLIST, Pseudo-ops
low, Expressions
LUA, Pseudo-ops


MMU, Pseudo-ops
(see also DEVICE)
(see also PAGE)
(see also SLOT)
mod, Expressions
MODULE, Pseudo-ops


OPT, Pseudo-ops
or, Expressions
ORG, Pseudo-ops
(see also DEVICE)
OUTEND, Pseudo-ops
(see also OUTPUT)
(see also SIZE)
OUTPUT, Pseudo-ops
(see also SIZE)


PAGE, Pseudo-ops
(see also DEVICE)
(see also MMU)
(see also SLOT)


RELOCATE_END, Pseudo-ops
REPT (see DUP)


SAVE3DOS, Pseudo-ops
SAVEAMSDOS, Pseudo-ops
SAVEBIN, Pseudo-ops
SAVECDT, Pseudo-ops
SAVECPCSNA, Pseudo-ops
SAVEDEV, Pseudo-ops
SAVEHOB, Pseudo-ops
SAVENEX, Pseudo-ops
SAVESNA, Pseudo-ops
SAVETAP, Pseudo-ops
SAVETRD, Pseudo-ops
SETBP, Pseudo-ops
(see also BPLIST)
SHELLEXEC, Pseudo-ops
shl, Expressions
shr, Expressions
SIZE, Pseudo-ops
(see also OUTPUT)
sj.add_byte(..), SjASMPlus functions bindings to Lua
sj.add_word(..), SjASMPlus functions bindings to Lua
sj.calc(..) (see _c(..))
sj.current_address, SjASMPlus functions bindings to Lua
sj.error(..), SjASMPlus functions bindings to Lua
sj.error_count, SjASMPlus functions bindings to Lua
sj.exit(..), SjASMPlus functions bindings to Lua
sj.file_exists(..), SjASMPlus functions bindings to Lua
sj.get_byte(..), SjASMPlus functions bindings to Lua
sj.get_define(..), SjASMPlus functions bindings to Lua
sj.get_device(), SjASMPlus functions bindings to Lua
sj.get_label(..), SjASMPlus functions bindings to Lua
sj.get_word(..), SjASMPlus functions bindings to Lua
sj.insert_define(..), SjASMPlus functions bindings to Lua
sj.insert_label(..), SjASMPlus functions bindings to Lua
sj.parse_code(..) (see _pc(..))
sj.parse_line(..) (see _pl(..))
sj.set_device(..), SjASMPlus functions bindings to Lua
sj.set_page(..), SjASMPlus functions bindings to Lua
sj.set_slot(..), SjASMPlus functions bindings to Lua
sj.shellexec(..), SjASMPlus functions bindings to Lua
sj.warning(..), SjASMPlus functions bindings to Lua
sj.warning_count, SjASMPlus functions bindings to Lua
SLDOPT, Pseudo-ops
SLOT, Pseudo-ops
(see also DEVICE)
(see also MMU)
(see also PAGE)


TAPEND, Pseudo-ops
(see also TAPOUT)
TAPOUT, Pseudo-ops
TEXT, STRUCT instructions


UNDEFINE, Pseudo-ops
(see also DEFINE)