Google
 

Trailing-Edge - PDP-10 Archives - decus_20tap1_198111 - decus/20-0003/fortio.mac
There are 7 other files named fortio.mac in the archive. Click here to see a list.
	title PASFOR - interface to allow Fortran I/O from Pascal


repeat 0,<
	
   procedure forinit(s:integer); extern;
   procedure forexit; extern;
   procedure forfix; extern;

Before calling the first fortran routine, call forinit.  The
argument is the amount of buffer space to allocate to Fortran.  Under
normal circumstances, 1000 should be enough.  3000 seems an upper
bound.

After calling the last Fortran routine, call forexit.  This will
close all fortran files and kill the emulator.

Call FORFIX on tops-20 after packages that expand core to get
working space and then contract it back to where it started, e.g.
SORT.  (Actually, SORT itself can't be called directly from
Pascal.  It should be called from a Fortran subroutine.  It
checks it argument type too strictly, so it cannot accept a call
from Pascal.)

On tops-10, the argument to FORINIT and the whole function FORFIX
can be omitted.

Also on Tops-10, we call FORINIT automatically during openning the
first Pascal file, because we now get the channel number from the
fortran runtimes.

When Fortran is made native mode, I am guessing that you will need
to get rid of some of all of the code under kludge.
>

	twoseg 400000

	search pasunv,monsym

kludge==1	;Fortran-20 is running compatibility

	entry forini,forexi,forfix

	external .jbff,.jbsa,.jbrel,.jbhrl,quit
	external reset.,exit.,forer%
ife tops10,<
	external getpag,relpag
> ;ife tops10
ifn tops10,<
	entry fn.chn,lo.chn
	external alchn.,dechn.,in.use
> ;ifn tops10

;procedure forini(iobufsp)
;  iobufsp is number of words to leave for I/O buffers for Pascal

;The only interference we have to worry about between Pascal and
;  F10 is in memory allocation.  Pascal doesn't use Tops-10
;  channels, so that is no problem.  Here are the crucial assumptions:
;We assume that fortran initially gobbles the space from .JBFF to
;  .JBREL.
;On Tops-20 we have to do a dummy core uuo to allocate all of memory, in
;  order to keep the emulator from killing us when we create pages in the
;  heap and stack.

ife tops10,<
;Get space for a fixed-size Fortran work space.
;compute arguments
forini:	subi b,1		;round up to a page
	tro b,777
	addi b,1
	push p,b		;save amount of space
	lsh b,-11		;b _ number of pages for Fortran
;get the space
	push p,c		;get us two places on the stack
	push p,d		;  for returning values
    ;stack: space in wrds, pag no., addr
	movei c,-1(p)		;where to put page number returned
	movei d,(p)		;where to put address returned
	pushj p,getpag		;now allocate a block for Fortran
;pass it to Fortran as .jbff and .jbrel
	move t,.jbff		;exchange .jbff and 0(p)
	exch t,(p)		;
	movem t,.jbff		;.jbff _ start of block
    ;stack: space in wrds, pag no., pascal .jbff
	add t,-2(p)
	subi t,1
	movem t,.jbrel		;.jbrel _ end of block
	movem t,ftnrel
> ;ife tops10

ifn tops10,<
forini:	aose in.use		;prevent reentry
	popj p,
forin:
> ;ifn tops10

;.jbff - .jbrel is now block for fortran
;now we are ready to call fortran init
	movem n,acsavn		;save global AC's
	movem o,acsavo
	movem p,acsavp
	jsp o,reset.+1
	0
	move p,acsavp		;restore global AC's
	move o,acsavo
	move n,acsavn

ife tops10,<
;Return .jbff to its pascal state and check for illicit memory expansion
	pop p,.jbff		;restore pascal's .JBFF
	pop p,(p)		;clean up stack
	pop p,(p)
	move t,.jbrel		;see if fortran had to get more space
	came t,ftnrel
	pushj p,forcer		;if so, error
;now we allocate all of memory, to turn off NXM trap
forfix:	
ifn kludge,<
	push p,.jbhrl		;and this to make restartable
	move a,[xwd 677777,377777] ;allocate all of memory
	calli a,11		;for the emulator
	 0
	pop p,.jbhrl
	move t,ftnrel
	movem t,.jbrel
> ;ifn kludge
	popj p,
> ;ife tops10

ifn tops10,<
forfix:	popj p,
> ;ifn tops10

	reloc 0

acsavn:	block 1
acsavo:	block 1
acsavp:	block 1
ftnrel:	block 1
acsav:	block 20

	reloc


;procedure forexi
;  close all fortran files 
forexi:	
ife tops10,<
;See if fortran needed more space than we gave it
	move t,.jbrel		;see if fortran has run out of space
	came t,ftnrel		;same as we left it?
	pushj p,forcer		;no - core error in fortran
> ;ife tops10
	movem p,acsavp
	movem o,acsavo
	movem n,acsavn
	move t,[jrst forex1]
	movem t,forer%		;cause exit. to return here after closing files
	hrrzi o,.+3
	pushj p,exit.
	0
	0
forex1:	move p,acsavp
	move o,acsavo
	move n,acsavn
ife tops10,<
ifn kludge,<
	movei a,.fhslf		;clear compatibility mode, or somebody will
	movei b,0		;garbage .jbrel and .jbhrl
	scvec
> ;ifn kludge
> ;ife tops10
	popj p,

forcer:	hrroi a,[asciz /
% Fortran seems to have run out of space during this program
/]
	psout
	popj p,

ifn tops10,<

;special find channel routine, calls fortran's
fn.chn:	pushj p,startf		;start up fortran if needed
	push p,o		;save o
	movei o,fnarg		;say give me any channel
	pushj p,alchn.		;allocate channel
	move a,t		;pascal wants result in a
	pop p,o
	popj p,

	xwd 0,-1
fnarg:	xwd 0,.+1
	z

;special lose channel routine, calls fortran's
lo.chn:	pushj p,startf		;start up fortran if needed
	push p,o		;save o
	push p,fnarg-1		;one arg
	push p,.+1		;this is a dummy
	push p,ac1		;ac1 is the value
	hrrzm p,-1(p)		;and put the location in the dummy slot
	pushj p,dechn.		;deallocate channel
	sub p,[xwd 3,3]
	pop p,o
	popj p,

;routine to call forini implicitly if pascal open is done before the
;first explicit call.
startf:	aose in.use		;in.use is -1 if not yet initialized
	popj p,
	movem 0,acsav		;save ac's
	move 0,[xwd 1,acsav+1]
	blt 0,acsav+17
	pushj p,forin		;call real workhorse
	move 0,[xwd acsav+1,1]	;restore ac's
	blt 0,17
	movem 0,acsav
	popj p,

> ;ifn tops10	

	end