SCM Repository

[smlnj] Diff of /sml/trunk/HISTORY

[smlnj] / sml / trunk / HISTORY

Diff of /sml/trunk/HISTORY

-revision 1133, Tue Mar 12 03:56:23 2002 UTC
+revision 1189, Fri May 10 20:40:33 2002 UTC
 Line 13
  Description:
  ----------------------------------------------------------------------
+ Name: Matthias Blume
+ Date: 2002/05/10 16:40:00 EDT
+ Tag: blume-20020510-erg-textio
+ Description:
+ Applied the following bugfix provided by Emden Gansner:
+     Output is corrupted when outputSubstr is used rather than output.
+     The problem occurs when a substring
+         ss = (s, dataStart, dataLen)
+     where dataStart > 0, fills a stream buffer with avail bytes left.
+     avail bytes of s, starting at index dataStart, are copied into the
+     buffer, the buffer is flushed, and then the remaining dataLen-avail
+     bytes of ss are copied into the beginning of the buffer. Instead of
+     starting this copy at index dataStart+avail in s, the current code
+     starts the copy at index avail.
+     Fix:
+     In text-io-fn.sml, change line 695 from
+          val needsFlush = copyVec(v, avail, dataLen-avail, buf, 0)
+     to
+          val needsFlush = copyVec(v, dataStart+avail, dataLen-avail, buf, 0)
+ ----------------------------------------------------------------------
+ Name: Matthias Blume
+ Date: 2002/04/12 13:55:00 EDT
+ Tag: blume-20020412-assyntax
+ Description:
+. Grabbed newer assyntax.h from the XFree86 project.
+. Fiddled with how to compile X86.prim.asm without warnings.
+. (Very) Minor cleanup in CM.
+ ----------------------------------------------------------------------
+ Name: Matthias Blume
+ Date: 2002/04/01 (no joke!) 17:07:00 EST
+ Tag: blume-20020401-x86div
+ Description:
+ Added full support for div/mod/rem/quot on the x86, using the machine
+ instruction's two results (without clumsily recomputing the remainder)
+ directly where appropriate.
+ Some more extensive power-of-two support was added to the x86 instruction
+ selector (avoiding expensive divs, mods, and muls where they can be
+ replaced with cheaper shifts and masks).  However, this sort of thing
+ ought to be done earlier, e.g., within the CPS optimizer so that
+ all architectures benefit from it.
+ The compiler compiles to a fixed point, but changes might be somewhat
+ fragile nevertheless.  Please, report any strange things that you might
+ see wrt. div/mod/quot/rem...
+ ----------------------------------------------------------------------
+ Name: Matthias Blume
+ Date: 2002/03/29 17:22:00
+ Tag: blume-20020329-div
+ Description:
+ Fixed my broken div/mod logic.  Unfortunately, this means that the
+ inline code for div/mod now has one more comparison than before.
+ Fast paths (quotient > 0 or remainder = 0) are not affected, though.
+ The problem was with quotient = 0, because that alone does not tell
+ us which way the rounding went.  One then has to look at whether
+ remainder and divisor have the same sign...  :(
+ Anyway, I replaced the bootfiles with fresh ones...
+ ----------------------------------------------------------------------
+ Name: Matthias Blume
+ Date: 2002/03/29 14:10:00 EST
+ Tag: blume-20020329-inlprims
+ Description:
+ NEW BOOTFILES!!!    Version number bumped to 110.39.3.
+ Primops have changed. This means that the bin/boot-file formats have
+ changed as well.
+ To make sure that there is no confusion, I made a new version.
+ CHANGES:
+ * removed REMT from mltree (remainder should never overflow).
+ * added primops to deal with divisions of all flavors to the frontend
+ * handled these primops all the way through so they map to their respective
+   MLRISC support
+ * used these primops in the implementation of Int, Int32, Word, Word32
+ * removed INLDIV, INLMOD, and INLREM as they are no longer necessary
+ * parameterized INLMIN, INLMAX, and INLABS by a numkind
+ * translate.sml now deals with all flavors of INL{MIN,MAX,ABS}, including
+   floating point
+ * used INL{MIN,MAX,ABS} in the implementation of Int, Int32, Word, Word32,
+   and Real (but Real.abs maps to a separate floating-point-only primop)
+ TODO items:
+ * Hacked Alpha32 instruction selection, disabling the selection of REMx
+   instructions because the machine instruction encoder cannot handle
+   them.  (Hppa, PPC, and Sparc instruction selection did not handle
+   REM in the first place, and REM is supported by the x86 machine coder.)
+ * Handle DIV and MOD with DIV_TO_NEGINF directly in the x86 instruction
+   selection phase.  (The two can be streamlined because the hardware
+   delivers both quotient and remainder at the same time anyway.)
+ * Think about what to do with "valOf(Int32.minInt) div ~1" and friends.
+   (Currently the behavior is inconsistent both across architectures and
+   wrt. the draft Basis spec.)
+ * Word8 should eventually be handled natively, too.
+ * There seems to be one serious bug in mltree-gen.sml.  It appears, though,
+   as if there currently is no execution path that could trigger it in
+   SML/NJ.  (The assumptions underlying functions arith and promotable do not
+   hold for things like multiplication and division.)
+ ----------------------------------------------------------------------
+ Name: Matthias Blume
+ Date: 2002/03/27 16:27:00 EST
+ Tag: blume-20020327-mlrisc-divisions
+ Description:
+ Added support for all four division operations (ML's div, mod, quot,
+ and rem) to MLRISC.  In the course of doing so, I also rationalized
+ the naming (no more annoying switch-around of DIV and QUOT), by
+ parameterizing the operation by div_rounding_mode (which can be either
+ DIV_TO_ZERO or DIV_TO_NEGINF).
+ The generic MLTreeGen functor takes care of compiling all four
+ operations down to only round-to-zero div.
+ Missing pieces:
+   * Doing something smarter than relying on MLTreeGen on architectures
+     like, e.g., the x86 where hardware division delivers both quotient and
+     remainder at the same time.  With this, the implementation of the
+     round-to-neginf operations could be further streamlined.
+   * Remove inlining support for div/mod/rem from the frontend and replace it
+     with primops that get carried through to the backend.  Do this for all
+     int and word types.
+ ----------------------------------------------------------------------
+ Name: Matthias Blume
+ Date: 2002/03/25 17:25:00 EST
+ Tag: blume-20020325-divmod
+ Description:
+ I improved (hopefully without breaking them) the implementation of Int.div,
+ Int.mod, and Int.rem.   For this, the code in translate.sml now takes
+ advantage of the following observations:
+   Let  q = x quot y      r = x rem y
+        d = x div  y      m = x mod y
+ where "quot" is the round-to-zero version of integer division that
+ hardware usually provides.  Then we have:
+      r = x - q * y        where neither the * nor the - will overflow
+      d = if q >= 0 orelse x = q * y then q else q - 1
+                           where neither the * nor the - will overflow
+      m = if q >= 0 orelse r = 0 then r else r + y
+                           where the + will not overflow
+ This results in substantial simplification of the generated code.
+ The following table shows the number of CFG nodes and edges generated
+ for
+         fun f (x, y) = x OPER y
+         (* with OPER \in div, mod, quot, rem *)
+     OPER | nodes(old) | edges(old) | nodes(new) | edges(new)
+     --------------------------------------------------------
+      div |         24 |         39 |         12 |         16
+      mod |         41 |         71 |         12 |         16
+     quot |          8 |         10 |          8 |         10
+      rem |         10 |         14 |          8 |         10
+ ----------------------------------------------------------------------
+ Name: Matthias Blume
+ Date: 2002/03/25 22:06:00 EST
+ Tag: blume-20020325-cprotobug
+ Description:
+ Fixed a bug in cproto (c prototype decoder).
+ ----------------------------------------------------------------------
+ Name: Matthias Blume
+ Date: 2002/03/25 16:00:00 EST
+ Tag: blume-20020325-raw-primops
+ Description:
+ I did some cleanup to Allen's new primop code and
+ replaced yesterday's bootfiles with new ones.
+ (But they are stored in the same place.)
+ ----------------------------------------------------------------------
+ Name: Matthias Blume
+ Date: 2002/03/24 22:40:00 EST
+ Tag: blume-20020324-bootfiles
+ Description:
+ Made the bootfiles that Allen asked for.
+ ----------------------------------------------------------------------
+ Name: Allen Leung
+ Date: 2002/03/23 15:50:00 EST
+ Tag: leunga-20020323-flint-cps-rcc-primops
+ Description:
+. Changes to FLINT primops:
+     (* make a call to a C-function;
+      * The primop carries C function prototype information and specifies
+      * which of its (ML-) arguments are floating point. C prototype
+      * information is for use by the backend, ML information is for
+      * use by the CPS converter. *)
+   | RAW_CCALL of { c_proto: CTypes.c_proto,
+                    ml_args: ccall_type list,
+                    ml_res_opt: ccall_type option,
+                    reentrant : bool
+                  } option
+    (* Allocate uninitialized storage on the heap.
+     * The record is meant to hold short-lived C objects, i.e., they
+     * are not ML pointers.  With the tag, the representation is
+     * the same as RECORD with tag tag_raw32 (sz=4), or tag_fblock (sz=8)
+     *)
+   | RAW_RECORD of {tag:bool,sz:int}
+   and ccall_type = CCALL_INT32 | CCALL_REAL64 | CCALL_ML_PTR
+.  These CPS primops are now overloaded:
+        rawload of {kind:numkind}
+        rawstore of {kind:numkind}
+       The one argument form is:
+          rawload {kind} address
+       The two argument form is:
+          rawload {kind} [ml object, byte-offset]
+. RAW_CCALL/RCC now takes two extra arguments:
+      a. The first is whether the C call is reentrant, i.e., whether
+         ML state should be saved and restored.
+      b. The second argument is a string argument specifying the name of
+         library and the C function.
+      These things are currently not handled in the code generator, yet.
+. In CProto,
+      An encoding type of "bool" means "ml object" and is mapped into
+      C prototype of PTR.  Note that "bool" is different than "string",
+      even though "string" is also mapped into PTR, because "bool"
+      is assigned an CPS type of BOGt, while "string" is assigned INT32t.
+. Pickler/unpicker
+      Changed to handle RAW_RECORD and newest RAW_CCALL
+. MLRiscGen,
+. Changed to handle the new rawload/rawstore/rawrecord operators.
+. Code for handling C Calls has been moved to a new module CPSCCalls,
+         in the file CodeGen/cpscompile/cps-c-calls.sml
+. Added the conditional move operator
+          condmove of branch
+      to cps.  Generation of this is still buggy so it is currently
+      disabled.
+ ----------------------------------------------------------------------
+ Name: Lal George
+ Date: 2002/03/22 14:18:25 EST
+ Tag: george-20020322-cps-branch-prob
+ Description:
+ Implemented the Ball-Larus branch prediction-heuristics, and
+ incorporated graphical viewers for control flow graphs.
+ Ball-Larus Heuristics:
+ ---------------------
+ See the file compiler/CodeGen/cpscompile/cpsBranchProb.sml.
+ By design it uses the Dempster-Shafer theory for combining
+ probabilities.  For example, in the function:
+     fun f(n,acc) = if n = 0 then acc else f(n-1, n*acc)
+ the ball-larus heuristics predicts that the n=0 is unlikely
+ (OH-heuristic), and the 'then' branch is unlikely because of the
+ RH-heuristic -- giving the 'then' branch an even lower combined
+ probability using the Dempster-Shafer theory.
+ Finally, John Reppy's loop analysis in MLRISC, further lowers the
+ probability of the 'then' branch because of the loop in the else
+ branch.
+ Graphical Viewing:
+ ------------------
+ I merely plugged in Allen's graphical viewers into the compiler. The
+ additional code is not much. At the top level, saying:
+         Control.MLRISC.getFlag "cfg-graphical-view" := true;
+ will display the graphical view of the control flow graph just before
+ back-patching.  daVinci must be in your path for this to work. If
+ daVinci is not available, then the default viewer can be changed
+ using:
+         Control.MLRISC.getString "viewer"
+ which can be set to "dot" or "vcg" for the corresponding viewers. Of
+ course, these viewers must be in your path.
+ The above will display the compilation unit at the level of clusters,
+ many of which are small, boring, and un-interesting. Also setting:
+         Control.MLRISC.getInt "cfg-graphical-view_size"
+ will display clusters that are larger than the value set by the above.
+ ----------------------------------------------------------------------
+ Name: Matthias Blume
+ Date: 2002/03/21 22:20:00 EST
+ Tag: blume-20020321-kmp-bugfix
+ Description:
+ Changed the interface to the KMP routine in PreString and fixed
+ a minor bug in one place where it was used.
+ ----------------------------------------------------------------------
+ Name: Allen Leung
+ Date: 2002/03/21 20:30:00 EST
+ Tag: leunga-20020321-cfg
+ Description:
+   Fixed a potential problem in cfg edge splitting.
+ ----------------------------------------------------------------------
+ Name: Allen Leung
+ Date: 2002/03/21 17:15:00 EST
+ Tag: leunga-20020321-x86-fp-cfg
+ Description:
+. Recoded the buggy parts of x86-fp.
+      a. All the block reordering code has been removed.
+         We now depend on the block placement phases to do this work.
+      b. Critical edge splitting code has been simplified and moved into the
+         CFG modules, as where they belong.
+      Both of these were quite buggy and complex.  The code is now much, much
+      simpler.
+. X86 backend.
+      a. Added instructions for 64-bit support.  Instruction selection for
+-bit has not been committed, however, since that
+         requires changes to MLTREE which haven't been approved by
+         Lal and John.
+      b. Added support for FUCOMI and FUCOMIP when generating code for
+         PentiumPro and above.  We only generate these instructions in
+         the fast-fp mode.
+      c. Added cases for JP and JNP in X86FreqProps.
+. CFG
+      CFG now has a bunch of methods for edge splitting and merging.
+. Machine description.
+      John's simplification of MLTREE_BASIS.fcond broke a few machine
+      description things:
+      rtl-build.{sig,sml} and hppa.mdl fixed.
+      NOTE: the machine description stuff in the repository is still broken.
+            Again, I can't put my fixes in because that involves
+            changes to MLTREE.
+ ----------------------------------------------------------------------
+ Name: Matthias Blume
+ Date: 2002/03/20 15:55:00 EST
+ Tag: blume-20020320-kmp
+ Description:
+ Implemented Knuth-Morris-Pratt string matching in PreString and used
+ it for String.isSubstring, Substring.isSubstring, and
+ Substring.position.
+ (Might need some stress-testing.  Simple examples worked fine.)
+ ----------------------------------------------------------------------
+ Name: Matthias Blume
+ Date: 2002/03/19 16:37:00 EST
+ Tag: blume-20020319-witnesses
+ Description:
+ Added a structure C.W and functions convert/Ptr.convert to ml-nlffi-lib.
+ This implements a generic mechanism for changing constness qualifiers
+ anywhere within big C types without resorting to outright "casts".
+ (So far, functions such as C.rw/C.ro or C.Ptr.rw/C.Ptr.ro only let you
+ modify the constness at the outermost level.)
+ The implementation of "convert" is based on the idea of "witness"
+ values -- values that are not used by the operation but whose types
+ "testify" to their applicability.  On the implementation side, "convert"
+ is simply a projection (returning its second curried argument).  With
+ cross-module inlining, it should not result in any machine code being
+ generated.
+ ----------------------------------------------------------------------
+ Name: Matthias Blume
+ Date: 2002/03/15 16:40:00 EST
+ Tag: blume-20020315-basis
+ Description:
+ Provided (preliminary?) implementations for
+   {String,Substring}.{concatWith,isSuffix,isSubstring}
+ and
+   Substring.full
+ Those are in the Basis spec but they were missing in SML/NJ.
+ ----------------------------------------------------------------------
+ Name: Matthias Blume
+ Date: 2002/03/14 21:30:00 EST
+ Tag: blume-20020314-controls
+ Description:
+ Controls:
+ ---------
+. Factored out the recently-added Controls : CONTROLS stuff and put
+    it into its own library $/controls-lib.cm.  The source tree for
+    this is under src/smlnj-lib/Controls.
+. Changed the names of types and functions in this interface, so they
+    make a bit more "sense":
+       module -> registry
+       'a registry -> 'a group
+. The interface now deals in ref cells only.  The getter/setter interface
+    is (mostly) gone.
+. Added a function that lets one register an already-existing ref cell.
+. Made the corresponding modifications to the rest of the code so that
+    everything compiles again.
+. Changed the implementation of Controls.MLRISC back to something closer
+    to the original.  In particular, this module (and therefore MLRISC)
+    does not depend on Controls.  There now is some link-time code in
+    int-sys.sml that registers the MLRISC controls with the Controls
+    module.
+ CM:
+ ---
+   * One can now specify the lambda-split aggressiveness in init.cmi.
+ ----------------------------------------------------------------------
+ Name: Allen Leung
+ Date: 2002/03/13 17:30:00 EST
+ Tag: leunga-20020313-x86-fp-unary
+ Description:
+ Bug fix for:
+ > leunga@weaselbane:~/Yale/tmp/sml-dist{21} bin/sml
+ > Standard ML of New Jersey v110.39.1 [FLINT v1.5], March 08, 2002
+ > - fun f(x,(y,z)) = Real.~ y;
+ > [autoloading]
+ > [autoloading done]
+ >       fchsl   (%eax), 184(%esp)
+ > Error: MLRisc bug: X86MCEmitter.emitInstr
+ >
+ > uncaught exception Error
+ >   raised at: ../MLRISC/control/mlriscErrormsg.sml:16.14-16.19
+ The problem was that the code generator did not generate any fp registers
+ in this case, and the ra didn't know that it needed to run the X86FP phase to
+ translate the pseudo fp instruction.   This only happened with unary fp
+ operators in certain situations.
+ ----------------------------------------------------------------------
+ Name: Matthias Blume
+ Date: 2002/03/13 14:00:00 EST
+ Tag: blume-20020313-overload-etc
+ Description:
+. Added _overload as a synonym for overload for backward compatibility.
+    (Control.overloadKW must be true for either version to be accepted.)
+. Fixed bug in install script that caused more things to be installed
+    than what was requested in config/targets.
+. Made CM aware of the (_)overload construct so that autoloading
+    works.
+ ----------------------------------------------------------------------
+ Name: Matthias Blume
+ Date: 2002/03/12 22:03:00 EST
+ Tag: blume-20020312-url
+ Description:
+ Forgot to update BOOT and srcarchiveurl.
+ ----------------------------------------------------------------------
+ Name: Matthias Blume
+ Date: 2002/03/12 17:30:00 EST
+ Tag: blume-20020312-version110392
+ Description:
+ Yet another version number bump (because of small changes to the
+ binfile format).  Version number is now 110.39.2.  NEW BOOTFILES!
+ Changes:
+   The new pid generation scheme described a few weeks ago was overly
+   complicated.  I implemented a new mechanism that is simpler and
+   provides a bit more "stability":  Once CM has seen a compilation
+   unit, it keeps its identity constant (as long as you do not delete
+   those crucial CM/GUID/* files).  This means that when you change
+   an interface, compile, then go back to the old interface, and
+   compile again, you arrive at the original pid.
+   There now also is a mechanism that instructs CM to use the plain
+   environment hash as a module's pid (effectively making its GUID
+   the empty string).  For this, "noguid" must be specified as an
+   option to the .sml file in question within its .cm file.
+   This is most useful for code that is being generated by tools such
+   as ml-nlffigen (because during development programmers tend to
+   erase the tool's entire output directory tree including CM's cached
+   GUIDs).  "noguid" is somewhat dangerous (since it can be used to locally
+   revert to the old, broken behavior of SML/NJ, but in specific cases
+   where there is no danger of interface confusion, its use is ok
+   (I think).
+   ml-nlffigen by default generates "noguid" annotations.  They can be
+   turned off by specifying -guid in its command line.
+ ----------------------------------------------------------------------
+ Name: Lal George
+ Date: 2002/03/12 12 14:42:36 EST
+ Tag: george-20020312-frequency-computation
+ Description:
+ Integrated jump chaining and static block frequency into the
+ compiler. More details and numbers later.
+ ----------------------------------------------------------------------
  Name: Lal George
  Date: 2002/03/11 11 22:38:53 EST
  Tag: george-20020311-jump-chain-elim
-Line 60
+Line 641
  ----------------------------------------------------------------------
  Name: Allen Leung
  Date: 2002/03/11 10:30:00 EST
- Tag: leunga-20020310-runtime-string0
+ Tag: leunga-20020311-runtime-string0
  Description:
     The representation of the empty string now points to a

 Legend:



Removed from v.1133
 


changed lines


 
Added in v.1189
 Legend:



Removed from v.1133
 


changed lines


 
Added in v.1189
-Removed from v.1133
+Added in v.1189

root@smlnj-gforge.cs.uchicago.edu	ViewVC Help
Powered by ViewVC 1.0.0