Babyl Options:
Version:5
Append:1

0,,
*** EOOH ***
Message-ID: <8806240204.AA02799@gsi-cam.UUCP>
Date: Thursday, 23 June 1988, 22:04-EDT
From: RWK@GSI-CAM
To: BUG-FALCON-COMPILER@GSI-CAM

In LMI FALCON-COMPILER in Experimental System 124.69,
Experimental Local-File 74.2, Experimental File-Server 23.1,
Experimental Unix-Interface 12.0, Experimental ZMail 72.0,
Experimental Tape 23.6, Experimental Lambda-Diag 16.2, microcode 1756,
SDU Boot Tape 3.14, SDU ROM 102, K 124, on Claude Debussy (LAMBDA):

I don't think PROGV and PROGW are going to work.  The rewrites
are Lambda-specific.  We'll need a way to suppress lambda-oid
rewrites in favor of appropriate P2-FOR-K hackery.




0,,
*** EOOH ***
Date: Thursday, 7 July 1988, 15:42-EDT
From: Steve Haflich <smh@GSI-CAM>
Subject: suggested optimizations
To: bug-lispm@GSI-CAM
Message-ID: <[GSI-ALEX].7-Jul-88 15:42:05.smh>

The lambda compiler should be smart enough to turn MEMBER into MEMQ in
all of the following:

	(member 'foo ...)
	(member ... '(x y z))

Specifically, if the first arg to MEMBER is a constant SYMBOL, or if the
second arg is a constant list of SYMBOLs, then the transformation is
legitimate.  Numbers could be special cased as well, but this is less
frequently useful.

This would improve both lambda and K code speed.




0,,
*** EOOH ***
Message-ID: <8808191904.AA09236@gsi-cam.UUCP>
Date: Friday, 19 August 1988, 15:04-EDT
From: Steve Haflich <smh@GSI-CAM>
Subject: Lambda compiler lossage
To: BUG-LISPM@GSI-CAM

In LMI Experimental System 126.51, ZWEI 125.16, ZMail 73.2,
Local-File 75.2, File-Server 24.1, Unix-Interface 13.0, Tape 24.2,
Lambda-Diag 17.0, microcode 1762, SDU Boot Tape 3.12, SDU ROM 102,
kold 4aug88, on Harpo (LAMBDA):

Compiling the following function blows up the lambda compiler.

   (DEFUN BOO (NAME)
     (LET
       ((A A))
       1))

The problem is in P!LET when it tries to flush the `useless' binding of
the local A to itself.

This is just a note to myself to fix it someday, as it also affects
the falcon cross compiler...




0,,
*** EOOH ***
Date: Thursday, 22 September 1988, 20:30-EDT
From: Steve Haflich <smh@GSI-CAM>
Subject: compiler blows the lambda away
To: bug-lispm@GSI-CAM
Message-ID: <[GSI-ALEX].22-Sep-88 20:30:00.smh>

Compiling the following function creates bad data and *halts* a lambda:

(defun blimp ()
  (let (*print-array*)
    #'(lambda ()
	(multiple-value-setq (*print-array*)
	  (foo)))))

There are two problems.

(1) The variable scoping analysis code in pass 1 is bizzarre and almost
certainly incorrect.  I have not fixed this yet.

(2) Some lamebrain decided it was necesary to shave a couple microcode
cycles in the compiler and use subprimitives to manipulate flag bits.
Unfortunately, problem (1) causes the subprimitive to receive a
non-numeric argument, so an invalid lisp object is created, and the
whole machine crashes or halts.  I put an error check in
COMPILER:QCOMPILE0 to catch and report the error gracefully without
crashing the machine.




0,,
*** EOOH ***
Date: Thursday, 20 October 1988, 19:00-EDT
From: Steve Haflich <smh@GSI-CAM>
Subject: Another One... (Compiling QRAND function TIME-IN-60ths)
To: jim@GSI-CAM, bug-falcon-compiler@GSI-CAM
In-reply-to: <[GSI-RAVEL].19-Oct-88 15:27:54.JIM>
Message-ID: <[GSI-HARPO].20-Oct-88 19:00:55.smh>

    Date: Wednesday, 19 October 1988, 15:27-EDT
    From: jim@GSI-CAM
    Subject: Another One... (Compiling QRAND function TIME-IN-60ths)
    Message-ID: <[GSI-RAVEL].19-Oct-88 15:27:54.JIM>

    >>ERROR: ETYPECASE failure; the value of COMPILER::DEST, MICROSECOND-INCREMENT,
	     is not a return dest, EQL either IGNORE or COMPILER::D-IGNORE, a functional dest, a register dest, a new frame dest, a progdesc, a multiple values, an open frame, a var or a new var.

This is fixed.  MULTIPLE-VALUE-SETQ couldn't properly special vars.