Number: 2041

Date:  2-Sep-84 15':01':37

Submitter: Masinter

Source: Masinter, JonL

Subject: want TYPEPROP opcode for compiling FIXP, ATOM

Assigned To: 

Attn: JonL, Charnley, Herring, Purcell

Status: Open

In/By: 

Problem Type: 

Impact: Moderate

Difficulty: Hard

Frequency: Everytime

Priority: Perhaps

System: Language Support

Subsystem: Microcode

Machine: 

Disk: 

Lisp Version: 31-Aug-84 16':17':57

Source Files: 

Microcode Version: 

Memory Size: 

File Server: 

Server Software Version: 

Disposition: 

Description: Date':  1 Aug 84 16':09 PDT'
From': JonL.pa'
Subject': Need for ucode changes'
To': vanMelle,Charnley'
cc': Masinter,Purcell,Kaplan'
. . . '
'
2) add a TYPEPROP opcode -- this would take an alphabyte and simply return T or NIL depending on whether the logand of this byte and  the hibyte of the mds typetable entry is zero.  There are currently  only two bits used in this hibyte --- \TT.NOREF and \TT.LISPREF.  I would propose reserving one bit for object-oriented usage (meaning  that this type has a class structure behind it), and leaving the   others for subsets of various types; the reason for this is to enable tests like FIXP, NUMBERP, ATOM, etc, to compile into one  TYPEPROP instruction, instead of the long sequence that they now   produce.  Larry and I have briefly talked about this one in the   context of simplifying the compilation of numeric predicates, but it would be generally more applicable;  at the very least, we could  get away from explicit dependencies on the type number assignments  and depend only upon the contents of the mds type table entries.'


Workaround: 

Test Case: 

Edit-By: 

Edit-Date: