Skip to content
Snippets Groups Projects
  1. Apr 15, 2008
    • rtoy's avatar
      sparc/parms.lisp: · ea08893f
      rtoy authored
      o Add SINGLE-FLOAT-BYTES and DOUBLE-FLOAT-BYTES to indicate how many
        bytes are needed to hold a SINGLE-FLOAT and DOUBLE-FLOAT,
        respectively.
      
      sparc/array.lisp:
      o Use SINGLE-FLOAT-BYTES and DOUBLE-FLOAT-BYTES instead of random
        constants or incorrect uses of VM:WORD-BYTES in the float code.
        (VM:WORD-BYTES is the wrong value to use if we have a 64-bit words.)
      o Try to replace the constants in all the shift instructions used to
        compute the byte offset from an index.  The shift is really the some
        other constant minus the number of tag bits used for the fixnum.
        This makes shift values a little easier to understand.
      ea08893f
    • rtoy's avatar
      Actually implement the DATA-VECTOR-REF-C and DATA-VECTOR-SET-C vops · af966489
      rtoy authored
      for the double-double float arrays that we support.
      af966489
  2. Apr 11, 2008
  3. Apr 10, 2008
  4. Apr 08, 2008
    • rtoy's avatar
      Update from logs. · 1e90a543
      rtoy authored
      1e90a543
    • rtoy's avatar
      o Fix type-derivation for the basic DOUBLE-DOUBLE-FLOAT arithmetic · 5c34c685
      rtoy authored
        operations.  The deftransforms were causing type-derivation to
        return DOUBLE-DOUBLE-FLOAT for everything.  Use the suggestion from
        Rob: Wrap the result of the deftransform in a TRULY-THE form and use
        the node derived type for the type.
      o The FLOAT-SIGN deftransform for DOUBLE-DOUBLE-FLOAT should be
        conditioned on the DOUBLE-DOUBLE feature.
      o Remove old unused deftransform for * for two double-double args.
      5c34c685
  5. Apr 04, 2008
    • rtoy's avatar
      o Pathname printer was producing an error for (MAKE-PATHNAME :HOST NIL · 52d2dc6c
      rtoy authored
        :TYPE "foo").  This is because PATHNAME-HOST was signaling an error
        for a host of NIL.  We check for a host of NIL now, and bypass the
        call to PATHNAME-HOST.  This allows the pathname printer to print
        #P(:HOST NIL :TYPE "foo"), as we want.
      
        (The error comes from host-pathname.  NIL is a valid result of
        host-pathname, so we could change that, but I think that has other
        implications, one of which is (MAKE-PATHNAME :HOST NIL :NAME "foo"
        :TYPE "lisp") gets printed as #P"foo.lisp", which is wrong, since
        that pathname has a unix-host host.)
      
      o Fix merging of version in MAKE-PATHNAME.  CLHS MERGE-PATHNAMES says
        if the pathname name is given, the version is not affected by the
        default pathname.
      52d2dc6c
    • agoncharov's avatar
      A white-space change (reversal to the "-2" version), to test CVS · 65b0a63d
      agoncharov authored
      commit notifications.
      65b0a63d
  6. Apr 03, 2008
  7. Apr 02, 2008
  8. Apr 01, 2008
  9. Mar 28, 2008
  10. Mar 26, 2008
  11. Mar 25, 2008
    • agoncharov's avatar
      White space noise for CVS commit testing · 152bc680
      agoncharov authored
      152bc680
    • rtoy's avatar
      Make compiler and interpreter handle slot-value the same. See thread · 4e65ef5d
      rtoy authored
      http://article.gmane.org/gmane.lisp.cmucl.devel/10471.
      
      These changes make the compiler and interpreter behave the same, and,
      as a side-effect, makes Nicolas's code work as he wants.
      
      pcl/slots.lisp:
      o Add LEGAL-SLOT-NAME-P-INTERNAL that returns two values.  If the
        slot-name is legal, return T.  If illegal, return NIL and a string
        indicating why.
      o Use this new function in the compiler macros for SLOT-VALUE, (SETF
        SLOT-VALUE) and SLOT-BOUNDP so that we use the fast ACCESSOR-*
        versions only when we know the slot-name is legal.  Otherwise, the
        compiler macros give up.
      
      pcl/methods.lisp:
      o Change the method LEGAL-SLOT-NAME-P to use
        LEGAL-SLOT-NAME-P-INTERNAL.
      o Update the call to LEGAL-SLOT-NAME-P in SHARED-INITIALIZE since
        LEGAL-SLOT-NAME-P returns 2 values now.
      4e65ef5d
  12. Mar 21, 2008
  13. Mar 19, 2008
  14. Mar 18, 2008
  15. Mar 16, 2008
  16. Mar 15, 2008
    • agoncharov's avatar
      Getting rid of some compiler warnings: · 44e0351e
      agoncharov authored
        * Eliminate `printf' format string and the following arguments
          conflicts.
      
        * A variable may be used uninitialized.
      
        * A variable may be defined but not used.
      44e0351e
  17. Mar 13, 2008
    • rtoy's avatar
      Fix off-by-one bug in ~R which prevents printing numbers from 10^63 to · 971d3122
      rtoy authored
      10^66-1.  10^63 is a vigintillion.  (From SBCL).
      971d3122
    • rtoy's avatar
      (format nil "~5f" 1d-10) was printing out way too many zeroes, · a660c8de
      rtoy authored
      exceeding the specified width.  I think the correct answer is
      ".0000".
      
      So, add a parameter, allow-overflow-p, to FLONUM-TO-STRING to allow
      the width constraint to be exceeded.  This defaults to T.
      FLONUM-TO-STRING looks at this and decides whether or not to print out
      all the leading zeroes or not.  In FIXED-FORMAT-AUX, set
      allow-overflow-p to NIL so we don't exceed the width.
      
      I wish there were another way to do this, but I can't think of one
      right now.
      
      All print tests pass, and add the following test:
      
      (format nil "~5F" 1d-10) -> ".0000"
      
      print.lisp:
      o Change FLONUM-TO-STRING to use keyword parameters instead of
        optional parameters.
      o Add :ALLOW-OVERFLOW-P, defaulting to T.
      o Use :ALLOW-OVERFLOW-P to determine if all the leading zeroes should
        be printed or not.
      
      format.lisp:
      o Update calls to FLONUM-TO-STRING to use keywords parameters.
      o In FIXED-FORMAT-AUX, set :ALLOW-OVERFLOW-P to NIL
      a660c8de
  18. Mar 05, 2008
  19. Mar 03, 2008
  20. Feb 29, 2008
Loading