XYBASIC for CP/M-80

Discussions
,
1

I’ve recently discovered this (I think as an aside from some discussion about Coherent): XYBASIC

It was written by Stephen Ness, who had been a TA to Don Knuth. Stephen also worked with Gary Kildall, mostly because Gary’s computer was the nearest one he could use with a disk drive.

The interpreter comes with full source, and instructions how to build. The default binary that’s part of the distribution includes decimal floating point. There are many other build options, including creating ROMable code and using an Am9511. It’s a smaller binary than any of the MBASIC-80 interpreters that I can find. It comes with a decent manual, too.

I rather like it, and I thought some folks here might like it too.

4 Likes
2

Great find! (I see it falls into the camp of sophisticated (rather than simple) handling of what the source called “vacuous condition” in FOR..NEXT, as we’ve been discussing in the Back To BASIC: Kemeny & Kurtz’s book from 1985 thread..)

Also of interest, a feature to suppress output while execution continues:

On the other hand, <control-O> actually suppresses console output; execution of your program continues, but output is not sent to the console until the next <control-O> is typed, or until an error occurs or the program returns to direct mode. By repeatedly toggling <control-O> you can watch a TRACE of your program intermittently without the time-consuming delay of writing all the trace information on the console.

3

But there’s a build option that executes all FOR loops, if you want: for0 equ false in version.asm.

I haven’t investigated rebuilding with my own options yet. The supplied CP/M binary works really well on my Altaid 8800.

1 Like
4

Oh, so there is! I should perhaps have noticed that. So it shows exactly how much code is needed to do this trick (around 20 instructions):

for2a:  xthl                    ;entry pointer to HL
        pop     b               ;incr pointer to BC
        if      for0
        push    h               ;save pointer to incr for FOR0 test
        endif
        call    moveb           ;copy increment to entry and return
        mvi     a,csfor
        if      not for0
        jmp     cpush           ;build the new CSTACK entry
        else
        call    cpush
;must now check for vacuous condition (e.g. FOR I=1 TO 0) in FOR0 version
        lhld    lhsad
        if      float
        lda     lhsty
        cpi     sngst
        if      f9511
        cz      lod95           ;load floating value to 9511 stack
        else
        cz      fload           ;load floating value to FACC
        endif
        endif
        mov     c,m
        inx     h
        mov     b,m             ;fetch integer value to BC
        pop     d               ;incr pointer to DE
        lxi     h,-fbyts
        dad     d               ;bound pointer to HL
        call    bdtst           ;test FOR condition
        rnc                     ;ok, just return
        lhld    cstkp
        call    next6           ;purge the CSTACK entry just built
        call    eos
        mvi     b,nextt
        call    fndst           ;find matching NEXT
        jnc     next7           ;look for following comma, return or do another
        error   f, F, R         ;not found, fatal FR error
        endif
5

Udo Munk put a disk image with several builds of XYBASIC (binary, BCD, and am9511 accelerated floating point arithmetic) here:

The xycpmfp.com binary FP executable is smaller than the BCD one, and handily faster: asciiart in 38 seconds, where the BCD one takes 59 seconds. Microsoft’s mbasic is still faster at 27 seconds, but is more than 8K larger.

(all times on my 37 MHz MinZ-C Z180 thing, which is so small I only just found it on my desk again)

2 Likes