[prev in list] [next in list] [prev in thread] [next in thread] 

List:       sbcl-commits
Subject:    [Sbcl-commits] master: Improve comments about globaldb info-vectors
From:       "Douglas Katzman" <snuglas () users ! sourceforge ! net>
Date:       2014-03-20 18:05:00
Message-ID: E1WQhKy-0002hp-Tv () sfs-ml-4 ! v29 ! ch3 ! sourceforge ! com
[Download RAW message or body]

The branch "master" has been updated in SBCL:
       via  facd5ba79f3804ca465bf3afcd78aef40815d715 (commit)
      from  d8422a83be76ca31fcfb087b7c2d11c2fa499aee (commit)

- Log -----------------------------------------------------------------
commit facd5ba79f3804ca465bf3afcd78aef40815d715
Author: Douglas Katzman <dougk@google.com>
Date:   Thu Mar 20 14:04:14 2014 -0400

    Improve comments about globaldb info-vectors
---
 src/compiler/info-vector.lisp |   47 +++++++++++++++++++---------------------
 1 files changed, 22 insertions(+), 25 deletions(-)

diff --git a/src/compiler/info-vector.lisp b/src/compiler/info-vector.lisp
index 69e8550..3f65d4b 100644
--- a/src/compiler/info-vector.lisp
+++ b/src/compiler/info-vector.lisp
@@ -10,10 +10,7 @@
 (in-package "SB!C")
 
 ;;;; This file implements abstract types which map globaldb Type-Number/Name
-;;;; pairs to data values compatibly with the volatile and compact environment
-;;;; structures defined in 'globaldb.lisp'.
-;;;; Some bootstrapping glitches prevent the legacy environment storage
-;;;; containers from being totally supplanted by the new ones unfortunately.
+;;;; pairs to data values. The database itself is defined in 'globaldb'.
 
 ;;; Quasi-lockfree concurrent hashtable
 ;;; ===================================
@@ -22,8 +19,7 @@
 ;;  http://trac.clozure.com/ccl/wiki/Internals/LockFreeHashTables
 ;;  https://github.com/boundary/high-scale-lib/blob/master/src/main/java/org/cliffc/high_scale_lib/NonBlockingHashMap.java
  
-;; The basic hashable is a very lightweight one that shares much in common
-;; with the COMPACT-INFO-ENVIRONMENT table: prime number sizing strategy,
+;; The basic hashable is a lightweight one using prime number sizing strategy,
 ;; and a secondary hash for re-probing that is not necessarily co-prime with
 ;; the table size (as it would be, ideally), and no support for deletion.
 
@@ -1039,23 +1035,14 @@ This is interpreted as
                                   (rest lists))))
              'vector))))
 
-;;; Helper macros for extracting parts of a Name in globaldb.
-;;; Names that satisfy the type specifier (CONS SYMBOL (CONS SYMBOL NULL))
-;;; are treated in a more efficient way than general names.
-
-;; If NAME matches the pattern (<SYMBOL2> <SYMBOL1>), then bind KEY1 and KEY2
-;; to those symbols and execute the body; otherwise skip the body.
-;; If ALLOW-ATOM is T [the default], then NAME can be just a symbol
-;; in which case KEY1 is bound to NAME and KEY2 to +NO-AUXILLIARY-KEY+.
+;; Given a NAME naming a globaldb object, decide whether the NAME has
+;; an efficient or "simple" form, versus a general or "hairy" form.
+;; The efficient form is either a symbol or a (CONS SYMBOL (CONS SYMBOL NULL)).
 ;;
-;; It's conceivable that we extend this to allow KEY2 to be a list,
-;; and have INFO-FIND-AUX-KEY/PACKED use EQUALP in that case.
-;; Or we could do something ad-hoc for
-;;  (PCL::SLOT-ACCESSOR :GLOBAL SB-PCL::some-name SB-PCL::READER)
-;; without supporting lists of length N in a fully general way.
-;; But for PCL methods it might make sense that fdefinition be stored
-;; in PCL metaobjects, and have a hook whereby NAME-FDEFINITION understands
-;; that certain kinds of names have a custom lookup technique.
+;; If NAME is a 2-list of symbols, bind KEY2 and KEY1 to the elements
+;; in that order, and execute the SIMPLE code, otherwise execute the HAIRY code.
+;; If ALLOW-ATOM is T - the default - then NAME can be just a symbol
+;; in which case its second component is +NO-AUXILLIARY-KEY+.
 ;;
 (defmacro with-globaldb-name ((key1 key2 &optional (allow-atom t))
                               name &key simple hairy)
@@ -1069,6 +1056,7 @@ This is interpreted as
                              ,key1 (car ,rest))
                        (and (symbolp ,key1) (symbolp ,key2) ,rest)))))
            ,simple
+           ;; The KEYs remain bound, but they should not be used for anything.
            ,hairy))))
 
 ;; Perform the approximate equivalent operations of retrieving
@@ -1122,10 +1110,19 @@ This is interpreted as
 (declaim (inline info-vector-fdefinition))
 (defun info-vector-fdefinition (vect)
   (when vect
+    ;; This is safe: Info-Vector invariant requires that it have length >= 1.
     (let ((word (the fixnum (svref vect 0))))
-      ;; Require the first type-number to be +fdefn-type-num+
-      ;; and the n-infos field to be nonzero. Info-Vector invariant
-      ;; requires that it have length >= 1, so this code is safe.
+      ;; Test that the first type-number is +fdefn-type-num+ and its n-infos
+      ;; field is nonzero. Type-num is the better discriminator because it
+      ;; can fail immediately, whereas the 'n-infos' is almost always nonzero,
+      ;; so does not inform the test. The degenerate vector +nil-packed-infos+
+      ;; can only occur from deletions, which are rare. But consider this:
+      ;;  * (DEFINE-COMPILER-MACRO (SETF ZOOK) (A) `(CAR ,A))
+      ;;  * (WRITE (SYMBOL-INFO-VECTOR 'ZOOK) :BASE 8)
+      ;;     #(120100 #<FUNCTION (COMPILER-MACRO-FUNCTION #) {100367911B}> SETF)
+      ;;         --
+      ;; The underlined packed field = +fdefn-type-num+ but it isn't an fdefn.
+      ;; It's the 'n-infos' for a secondary name. The primary name has N=0.
       (when (and (eql (mask-field (byte type-number-bits type-number-bits) word)
                       (ash +fdefn-type-num+ type-number-bits))
                  (ldb-test (byte type-number-bits 0) word))

-----------------------------------------------------------------------


hooks/post-receive
-- 
SBCL

------------------------------------------------------------------------------
Learn Graph Databases - Download FREE O'Reilly Book
"Graph Databases" is the definitive new guide to graph databases and their
applications. Written by three acclaimed leaders in the field,
this first edition is now available. Download your free book today!
http://p.sf.net/sfu/13534_NeoTech
_______________________________________________
Sbcl-commits mailing list
Sbcl-commits@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/sbcl-commits


[prev in list] [next in list] [prev in thread] [next in thread]