Skip to content
Snippets Groups Projects
  1. Apr 01, 1997
  2. Mar 29, 1997
  3. Mar 26, 1997
  4. Mar 25, 1997
  5. Mar 17, 1997
  6. Mar 15, 1997
  7. Feb 23, 1997
  8. Feb 22, 1997
    • dtc's avatar
      Use a separate stack for alien value allocation on the x86. It had · 1975d13a
      dtc authored
      been sharing the control stack with the compiler which could cause
      trouble (pushes and pops out of order, holes in arguments built up on
      the stack, exposed stack that can be written over on interrupt, etc)
      
      Assigned a static symbol, x86::*alien-stack*, and by default set it to
      the old C stack. A program many rebind this if it needs more room or
      wants the allocation elsewhere.
      
      The *alien-stack* is saved with the dynamic environment so is restored
      with a non-local exist. It's not save on function entry/exit so you
      need to take care to match an allocation with a deallocation.
      
      Makes use of new VOPs which are already in place.
      
      Backout an old alien stack deallocation hack.
      1975d13a
  9. Feb 18, 1997
  10. Feb 15, 1997
  11. Feb 14, 1997
  12. Feb 13, 1997
  13. Feb 12, 1997
  14. Feb 11, 1997
  15. Feb 10, 1997
  16. Feb 08, 1997
    • dtc's avatar
      Some style cleanups. · d38bd461
      dtc authored
      Change the offset used by the sap VOPS from an unsigned-num to a
      signed-num. This helps fixes some SAP functions for negative offsets
      which are useful on the x86 with its stack growing down. The offset is
      currently declared to be a index (unsigned-byte 29) in
      compiler/saptran.lisp so this should work without change, but to use
      the negative offsets requires complementary changes to saptran (index
      -> fixnum).
      d38bd461
    • dtc's avatar
      Use a constant push not a stack relative EA when pushing the return · bc6d4dce
      dtc authored
      address for the x86-lra case in define-full-call - read somewhere it
      was faster. Old version left in source.
      
      Removed bogus uses of two "inc ret" instead of "add ret 2" which uses
      only 3 bytes not 5 or 6 as suggested in the source.
      
      Slight style cleanups. Remove some dead code.
      bc6d4dce
    • dtc's avatar
      Fix from Rob for trouble noted by Peter: · 383096e9
      dtc authored
      > o.k.  interpreted-function-lambda-expression just didn't work for anonymous
      > lambdas with non-fixed argcounts.
      >
      >   Rob
      383096e9
    • dtc's avatar
      Slight style cleanup. · 766f3561
      dtc authored
      766f3561
    • dtc's avatar
      VOPS make-unwind-block, and make-catch-block: the code slot is unused, · 65048fc5
      dtc authored
      so fill with 0 else random garbage will be seen when read.
      65048fc5
    • dtc's avatar
      Best to leave the stack SC out of the SCs, the compiler · d48a4918
      dtc authored
      will allocate a save tn if necessary and consider the cost. Changed VOPS
      cell-setf, cell-setf-function.
      d48a4918
    • dtc's avatar
      Add float-precision-control constant. · 8fa7139c
      dtc authored
      8fa7139c
    • dtc's avatar
      Slight style cleanup. · 75b786fc
      dtc authored
      75b786fc
    • dtc's avatar
      Removed unnecessary comment. · a948221d
      dtc authored
      a948221d
    • pw's avatar
      From: Douglas Thomas Crosher <dtc@scrooge.ee.swin.oz.au> · 0efcc964
      pw authored
      Message-Id: <199702041630.DAA06602@scrooge.ee.swin.oz.au>
      Subject: Re: making eval-functions FINs
      To: cmucl-imp@cons.org
      Date: Wed, 5 Feb 1997 03:30:55 +1100 (EST)
      In-Reply-To: <199702040300.EAA16744@knight.cons.org> from "Rob MacLachlan" at Feb 3, 97 09:59:02 pm
      X-Mailer: ELM [version 2.4 PL24]
      Content-Type: text
      
      
      > I seem to have lost the most recent message about fixing compiler/eval to
      > use FINs instead of closures.  But yes, I would think that the current
      > EVAL-FUNCTION  slots should become slots in the interpreted function FIN.
      > GET-EVAL-FUN when would then become IDENTITY, so it should probably be
      > flushed entirely.
      
      Done; moved eval-function data into the interpreted-function FIN, also
      added a closure slot so describe can access it easily. Revised patches
      below.
      
      Douglas
      0efcc964
  17. Feb 06, 1997
  18. Feb 05, 1997
  19. Jan 18, 1997
Loading