Home My Page Projects Code Snippets Project Openings SML/NJ
Summary Activity Forums Tracker Lists Tasks Docs Surveys News SCM Files

SCM Repository

[smlnj] Annotation of /sml/trunk/READMES/110.27-README
ViewVC logotype

Annotation of /sml/trunk/READMES/110.27-README

Parent Directory Parent Directory | Revision Log Revision Log


Revision 618 - (view) (download)

1 : dbm 610
2 :     S M L / N J
3 :    
4 :     1 1 0 . 2 7 N E W S
5 :    
6 :     April 10, 2000
7 :    
8 :     WARNING
9 :    
10 :     This version is intended for compiler hackers.
11 :     We are in the midst of substantial structural changes,
12 :     and this is a snapshot.
13 :    
14 :     http://cm.bell-labs.com/cm/cs/what/smlnj/index.html
15 :    
16 :     ------------------------------------------------------------------------
17 :     Summary:
18 :    
19 :     This version has some minor tweeks to FLINT (after the major merge
20 :     in 110.26). Work continues on tuning FLINT and the various optimizations
21 :     it implements.
22 :    
23 :     CM has been revised extensively, and the modmap environment mechanism
24 :     supporting stubbified pickles has been reworked completely. The pathconfig
25 :     file has been simplified. Installation scripts have been further
26 : dbm 616 modified. See src/system/README and the latest version of the
27 :     CM manual at
28 : dbm 610
29 : dbm 616 <//http://www.kurims.kyoto-u.ac.jp/~blume/SMLNJ-DEV/manual/index.html>
30 :     <//http://www.kurims.kyoto-u.ac.jp/~blume/SMLNJ-DEV/manual.ps>
31 :    
32 :     for further information about these changes.
33 :    
34 : leunga 617 MLRISC, and particularly the x86 back end have been modified extensively.
35 : dbm 610
36 :     There are a few updates to the SML/NJ Library
37 :    
38 :     Reported bug fixes:
39 :     1556. (jhr) signal race condition
40 :     Some CM bugs (not recorded)
41 :    
42 :    
43 : dbm 616 Distribution file names
44 :     Distribution file names have been simplified. They no longer start
45 :     with the version number (e.g. "110.27-config.tar.gz" is now
46 :     simply "config.tar.gz"). The boot directory tarballs are now
47 :     "boot.alpha32-unix.tar.gz", etc. (i.e. no version number and the
48 :     "sml." prefix is dropped). The new install script will restore
49 :     the usual name (e.g. "sml.boot.alpha32-unix" when the tarball is
50 :     unpacked. [We dropped the initial "sml." for the boot tarballs to
51 :     get the file names under 28 characters because of a limitation of
52 :     the Bell Labs ftp server.]
53 :     The version README file is still named 110.27-README, however.
54 :    
55 :    
56 :     ----------------------------------------------------------------------
57 : dbm 610 The following change records are extracted from the CVS repository
58 :     HISTORY file, which started in early March, 2000, so this doesn't
59 :     cover changes between mid-December, 1999, when 110.26 was created, and
60 :     early March.
61 :    
62 :     ======================================================================
63 :     FLINT:
64 :     ======================================================================
65 :    
66 :     Name: Stefan
67 :     Date: 2000/04/07 10:00:00 EDT
68 :     Tag: monnier-20000406-branch-handling
69 :     Description:
70 :    
71 :     Improved handling of branches (mostly those generated from
72 :     polymorphic equality), removed switchoff and changed the
73 :     default optimization settings (more cpsopt and less flintopt).
74 :    
75 :    
76 :     ======================================================================
77 :     MLRISC:
78 :     ======================================================================
79 :    
80 : leunga 617 1. Register Allocator
81 : dbm 610
82 : leunga 617 a. The interface and implementation of the register allocator have been
83 :     changed slightly to accommodate the possibility of skipping
84 :     the register allocation phases completely and go directly to
85 :     memory allocation. This is needed for C-- use.
86 : dbm 610
87 : leunga 617 b. I've improved the spill propagation algorithm, using an approximation
88 :     of maximal weighted independent sets. This affects only the x86
89 :     platform.
90 : dbm 610
91 : leunga 617 2. MLTREE
92 : dbm 610
93 : leunga 617 a. Renamed the constructor CALL in MLTREE by popular demand.
94 : dbm 610
95 : leunga 617 3. X86
96 : dbm 610
97 : leunga 617 a. More assembly output problems involving the indexed addressing mode
98 :     on the x86 have been found and corrected. Thanks to Fermin Reig for the
99 :     fix.
100 : dbm 610
101 : leunga 617 b. x86Rewrite bug with MUL3 (found by Lal)
102 :    
103 :     c. Added the instructions FSTS, FSTL
104 :    
105 :     d. The old code generated for SETcc was completely wrong.
106 :     The Intel optimization guide is VERY misleading.
107 :    
108 :     e. Various fixes related floating point, and extensions.
109 :    
110 :     f. Things like
111 :    
112 :     jmp %eax
113 :     jmp (%eax)
114 :    
115 :     are now output as
116 :    
117 :     jmp *%eax
118 :     jmp *(%eax)
119 :    
120 :     g. Yet another fix for x86 assembly for idivl, imull, mull and friends.
121 :    
122 :     h. I've changed andl to testl in the floating point test sequence
123 :     whenever appropriate. The Intel optimization guide states that
124 :     testl is perferable to andl.
125 :    
126 :     4. Alpha
127 :    
128 :     a. Some extra patterns related to loads with signed/zero extension
129 :     provided by Fermin.
130 :     b. Added the instructions LDBU, LDWU, STB, STW as per Fermin's suggestion.
131 :     c. Added a new mode byteWordLoadStores to the functor parameter to Alpha()
132 :     d. Added reassociation code for address computation.
133 :    
134 :     5. PA-RISC
135 :    
136 : dbm 610 a. B label should not be a delay slot candidate! Why did this work?
137 :     b. ADDT(32, REG(32, r), LI n) now generates one instruction instead of two,
138 :     as it should be.
139 :     c. The assembly syntax for fstds and fstdd was wrong.
140 :     d. Added the composite instruction COMICLR/LDO, which is the immediate
141 :     operand variant of COMCLR/LDO.
142 : leunga 617 e. Long jumps in span dependence resolution used to depend on the existence
143 :     of the base pointer in the SML/NJ runtime.
144 : dbm 610
145 : leunga 617 A jump to a long label L was expanded into the following sequence:
146 :    
147 :     LDIL %hi(L-8192), %r29
148 :     LDO %lo(L-8192)(%r29), %r29
149 :     ADD %r29, baseptr, %r29
150 :     BV,n %r0(%r29)
151 : dbm 610
152 : leunga 617 I've changed it so that the following sequence of instructions
153 :     are generated, which doesn't mention the base pointer at all:
154 :    
155 :     BL,n L', %r29 /* branch and link, L' + 4 -> %r29 */
156 :     L': ADDIL L-(L'+4), %r29 /* Compute address of L */
157 :     BV,n %r0(%r29) /* Jump */
158 :    
159 :     6. Generic MLRISC
160 :    
161 : dbm 610 a. shuffle.sml rewritten to be slightly more efficient
162 :     b. DIV bug in mltree-simplify fixed (found by Fermin)
163 :    
164 : leunga 617 7. Assembly Output
165 : dbm 610
166 : leunga 617 a. When generating assemby, resolve the value of client defined constants,
167 :     instead of generating symbolic values. This is controlled by the
168 :     new flag "asm-resolve-constants", which is default to true.
169 : dbm 610
170 : leunga 617 b. Added a new flag
171 : dbm 610
172 : leunga 617 "asm-indent-copies" (default to false)
173 : dbm 610
174 : leunga 617 When this flag is on, parallel copies will be indented an extra level.
175 : dbm 610
176 :    
177 : leunga 617 8. Machine Descriptions/Generation
178 : dbm 610
179 : leunga 617 a. The precedence parser was slightly broken when parsing infixr symbols.
180 :     b. The type generalizing code had the bound variables reversed, resulting
181 :     in a problem during arity raising.
182 :     c. Various fixes in machine descriptions.
183 : dbm 610
184 : leunga 617 ======================================================================
185 :     CPS->MLRISC Code Generation
186 :     ======================================================================
187 : dbm 610
188 : leunga 617 This release contains *MAJOR* changes to the way code is generated from CPS
189 :     in the module mlriscGen, and in various backend modules.
190 : dbm 610
191 : leunga 617 1. Forward propagation fix.
192 :    
193 : dbm 610 There was a bug in forward propagation introduced at about the same time
194 :     as the MLRISC x86 backend, which prohibits coalescing to be
195 :     performed effectively in loops.
196 :    
197 :     Effect: speed up of loops in RISC architectures.
198 :     By itself, this actually slowed down certain benchmarks on the x86.
199 :    
200 : leunga 617 2. Forward propagating addresses from consing.
201 : dbm 610
202 :     I've changed the way consing code is generated. Basically I separated
203 :     out the initialization part:
204 :    
205 :     store tag, offset(allocptr)
206 :     store elem1, offset+4(allocptr)
207 :     store elem2, offset+8(allocptr)
208 :     ...
209 :     store elemn, offset+4n(allocptr)
210 :    
211 :     and the address computation part:
212 :    
213 :     celladdr <- offset+4+alloctpr
214 :    
215 :     and move the address computation part
216 :    
217 :     Effect: register pressure is generally lower as a result. This
218 :     makes compilation of certain expressions much faster, such as
219 :     long lists with non-trivial elements.
220 :    
221 :     [(0,0), (0,0), .... (0,0)]
222 :    
223 : leunga 617 3. Base pointer elimination.
224 : dbm 610
225 :     As part of the linkage mechanism, we generate the sequence:
226 :    
227 :     L: ... <- start of the code fragment
228 :    
229 :     L1:
230 :     base pointer <- linkreg - L1 + L
231 :    
232 :     The base pointer was then used for computing relocatable addresses
233 :     in the code fragment. Frequently (such as in lots of continuations)
234 :     this is not needed. We now eliminate this sequence whenever possible.
235 :    
236 :     For compile time efficiency, I'm using a very stupid local heuristic.
237 :     But in general, this should be done as a control flow analysis.
238 :    
239 :     Effect: Smaller code size. Speed up of most programs.
240 :    
241 :    
242 : leunga 617 4. Frequency annotations
243 : dbm 610
244 :     I've added an annotation that states that all call gc blocks have zero
245 :     execution frequencies. This improves register allocation on the x86.
246 :    
247 :     BENCHMARKS
248 :     ==========
249 :    
250 :     I've only perform the comparison on 110.25.
251 :    
252 :     The platforms are:
253 :    
254 :     HPPA A four processor HP machine (E9000) with 5G of memory.
255 :     X86 A 300Hhz Pentium II with 128M of memory, and
256 :     SPARC An Ultra sparc 2 with 512M of memory.
257 :    
258 :     I used the following parameters for the SML benchmarks:
259 :    
260 :     @SMLalloc
261 :     HPPA 256k
262 :     SPARC 512k
263 :     X86 256k
264 :    
265 :     COMPILATION TIME
266 :     ----------------
267 :     Here are the numbers comparing the compilation times of the compilers.
268 :     I've only compared 110.25 compiling the new sources versus
269 :     a fixpoint version of the new compiler compiling the same.
270 :    
271 :     110.25 New
272 :     Total Time in RA Spill+Reload Total Time In RA Spill+Reload
273 :     HPPA 627s 116s 2684+3584 599s 95s 1003+1879
274 :     SPARC 892s 173s 2891+3870 708s 116s 1004+1880
275 :     X86 999s 315s 94006+130691 987s 296s 108877+141957
276 :    
277 :     110.25 New
278 :     Code Size Code Size
279 :     HPPA 8596736 8561421
280 :     SPARC 8974299 8785143
281 :     X86 9029180 8716783
282 :    
283 :     So in summary, things are at least as good as before. Dramatic
284 :     reduction in compilation is obtained on the Sparc; I can't explain it,
285 :     but it is reproducible. Perhaps someone should try to reproduce this
286 :     on their own machines.
287 :    
288 :     SML BENCHMARKS
289 :     --------------
290 :    
291 :     On the average, all benchmarks perform at least as well as before.
292 :    
293 :     HPPA Compilation Time Spill+Reload Run Time
294 :     110.25 New 110.25 New 110.25 New
295 :    
296 :     barnesHut 3.158 3.015 4.75% 1+1 0+0 2.980 2.922 2.00%
297 :     boyer 6.152 5.708 7.77% 0+0 0+0 0.218 0.213 2.34%
298 :     count-graphs 1.168 1.120 4.32% 0+0 0+0 22.705 23.073 -1.60%
299 :     fft 0.877 0.792 10.74% 1+3 1+3 0.602 0.587 2.56%
300 :     knuthBendix 3.180 2.857 11.32% 0+0 0+0 0.675 0.662 2.02%
301 :     lexgen 6.190 5.290 17.01% 0+0 0+0 0.913 0.788 15.86%
302 :     life 0.803 0.703 14.22% 25+25 0+0 0.153 0.140 9.52%
303 :     logic 2.048 2.007 2.08% 6+6 1+1 4.133 4.008 3.12%
304 :     mandelbrot 0.077 0.080 -4.17% 0+0 0+0 0.765 0.712 7.49%
305 :     mlyacc 22.932 20.937 9.53% 154+181 32+57 0.468 0.430 8.91%
306 :     nucleic 5.183 5.060 2.44% 2+2 0+0 0.125 0.120 4.17%
307 :     ratio-regions 3.357 3.142 6.84% 0+0 0+0 116.225 113.173 2.70%
308 :     ray 1.283 1.290 -0.52% 0+0 0+0 2.887 2.855 1.11%
309 :     simple 6.307 6.032 4.56% 28+30 5+7 3.705 3.658 1.28%
310 :     tsp 0.888 0.862 3.09% 0+0 0+0 7.040 6.893 2.13%
311 :     vliw 24.378 23.455 3.94% 106+127 25+45 2.758 2.707 1.91%
312 :     --------------------------------------------------------------------------
313 :     Average 6.12% 4.09%
314 :    
315 :     SPARC Compilation Time Spill+Reload Run Time
316 :     110.25 New 110.25 New 110.25 New
317 :    
318 :     barnesHut 3.778 3.592 5.20% 2+2 0+0 3.648 3.453 5.65%
319 :     boyer 6.632 6.110 8.54% 0+0 0+0 0.258 0.242 6.90%
320 :     count-graphs 1.435 1.325 8.30% 0+0 0+0 33.672 34.737 -3.07%
321 :     fft 0.980 0.940 4.26% 3+9 2+6 0.838 0.827 1.41%
322 :     knuthBendix 3.590 3.138 14.39% 0+0 0+0 0.962 0.967 -0.52%
323 :     lexgen 6.593 6.072 8.59% 1+1 0+0 1.077 1.078 -0.15%
324 :     life 0.972 0.868 11.90% 26+26 0+0 0.143 0.140 2.38%
325 :     logic 2.525 2.387 5.80% 7+7 1+1 5.625 5.158 9.05%
326 :     mandelbrot 0.090 0.093 -3.57% 0+0 0+0 0.855 0.728 17.39%
327 :     mlyacc 26.732 23.827 12.19% 162+189 32+57 0.550 0.560 -1.79%
328 :     nucleic 6.233 6.197 0.59% 3+3 0+0 0.163 0.173 -5.77%
329 :     ratio-regions 3.780 3.507 7.79% 0+0 0+0 133.993 131.035 2.26%
330 :     ray 1.595 1.550 2.90% 1+1 0+0 3.440 3.418 0.63%
331 :     simple 6.972 6.487 7.48% 29+32 5+7 3.523 3.525 -0.05%
332 :     tsp 1.115 1.063 4.86% 0+0 0+0 7.393 7.265 1.77%
333 :     vliw 27.765 24.818 11.87% 110+135 25+45 2.265 2.135 6.09%
334 :     ----------------------------------------------------------------------------
335 :     Average 6.94% 2.64%
336 :    
337 :     X86 Compilation Time Spill+Reload Run Time
338 :     110.25 New 110.25 New 110.25 New
339 :    
340 :     barnesHut 5.530 5.420 2.03% 593+893 597+915 3.532 3.440 2.66%
341 :     boyer 8.768 7.747 13.19% 493+199 301+289 0.327 0.297 10.11%
342 :     count-graphs 2.040 2.010 1.49% 298+394 315+457 26.578 28.660 -7.26%
343 :     fft 1.327 1.302 1.92% 112+209 115+210 1.055 0.962 9.71%
344 :     knuthBendix 5.218 5.475 -4.69% 451+598 510+650 0.928 0.932 -0.36%
345 :     lexgen 9.970 9.623 3.60% 1014+841 1157+885 0.947 0.928 1.97%
346 :     life 1.183 1.183 0.00% 162+182 145+148 0.127 0.103 22.58%
347 :     logic 3.285 3.512 -6.45% 514+684 591+836 5.682 5.577 1.88%
348 :     mandelbrot 0.147 0.143 2.33% 38+41 33+54 0.703 0.690 1.93%
349 :     mlyacc 35.457 32.763 8.22% 3496+4564 3611+4860 0.552 0.550 0.30%
350 :     nucleic 7.100 6.888 3.07% 239+168 201+158 0.175 0.173 0.96%
351 :     ratio-regions 6.388 6.843 -6.65% 1182+257 981+300 120.142 120.345 -0.17%
352 :     ray 2.332 2.338 -0.29% 346+398 402+494 3.593 3.540 1.51%
353 :     simple 9.912 9.903 0.08% 1475+941 1579+1168 3.057 3.178 -3.83%
354 :     tsp 1.623 1.532 5.98% 266+200 250+211 8.045 7.878 2.12%
355 :     vliw 33.947 35.470 -4.29% 2629+2774 2877+3171 2.072 1.890 9.61%
356 :     ----------------------------------------------------------------------------
357 :     Average 1.22% 3.36%
358 :    
359 :    
360 : leunga 617
361 :     Aliasing
362 :     ---------
363 : dbm 610 This update contains a rewritten (and hopefully more correct) module
364 :     for extracting aliasing information from CPS.
365 :    
366 :     To turn on this feature:
367 :    
368 :     Compiler.Control.CG.memDisambiguate := true
369 :    
370 :     To pretty print the region information with assembly
371 :    
372 :     Compiler.Control.MLRISC.getFlag "asm-show-region" := true;
373 :    
374 :     To control how many levels of aliasing information are printed, use:
375 :    
376 :     Compiler.Control.MLRISC.getInt "points-to-show-level" := n
377 :    
378 :     The default of n is 3.
379 :    
380 :     ======================================================================
381 : blume 618 Boot code and glue scripts
382 : dbm 610 ======================================================================
383 :    
384 : blume 618 Size info in BOOTLIST
385 : dbm 610
386 : blume 618 The BOOTLIST file now has an optional first line that specifies an
387 :     upper bound on the number of boot files and an upper bound on the
388 :     length of each individual name. With this, there are no longer
389 :     hard-wired restrictions on these values in the runtime system.
390 :     (If the specification is missing in BOOTLIST, the runtime system
391 :     falls back to its old behavior, i.e., hard-wired defaults.)
392 : dbm 610
393 : blume 618 Allocation-size heuristics in .run-sml
394 : dbm 610
395 : blume 618 The .run-sml scripts tries to read processor cache size from
396 :     /proc/cpuinfo. This works on Linux and is important for small-cache
397 :     Celeron systems that suffer badly when allocation size is set too
398 :     high.
399 : dbm 610
400 : blume 618 Install script
401 : dbm 610
402 : blume 618 - Written in a more modular fashion (using shell functions).
403 :     - Made more robust.
404 :     - Automagically fetches archive files over the network if they do not
405 :     exist locally. Thus, you only need to fetch config.tar.gz yourself.
406 :     Unpack it and go!
407 :     (Requires "wget" or "lynx" to be installed on the system and a
408 :     live connection to the internet. Moreover, the contents of
409 :     config/srcarchiveurl must be set properly.)
410 :     For CVS users, this may be convenient when fetching new sets of binfiles.
411 :     - Handles archive files with or without version number and compressed
412 :     with one of "gzip", "compress", or "bzip2". Recognized suffixes are
413 :     ".tar.gz", ".tgz", ".tar", ".tar.Z", and ".tar.bz2".
414 : dbm 610
415 : blume 618 PIDMAP file
416 : dbm 610
417 : blume 618 There is a file called PIDMAP in the bootfile directory.
418 :     It is used to minimize the amount of dynamic state that needs to be
419 :     stowed away for the purpose of sharing between interactive system
420 :     and user code.
421 : dbm 610
422 : blume 618 Building standalone programs
423 : dbm 610
424 : blume 618 The command ml-build can be used to build standalone programs.
425 :     ml-build takes three arguments:
426 : dbm 610
427 : blume 618 1. the name of the CM library that implements and exports the "main"
428 :     function of your program
429 :     2. the name of the "main" function of your program as exported by 1.
430 :     (The function must have a type that makes it suitable as an argument
431 :     to SMLofNJ.exportFn.)
432 :     3. the name of the heapfile to be generated
433 : dbm 610
434 : blume 618 Other build scripts
435 : dbm 610
436 : blume 618 ml-{lex,yacc} build scripts now make use of the new mechanism for
437 :     building standalone programs.
438 : dbm 610
439 : blume 618 Fixpoint script
440 : dbm 610
441 : blume 618 I added a re-written version of Dave's fixpt script to src/system.
442 :     Changes relative to the original version:
443 :     - sh-ified (not everybody has ksh)
444 :     - automatically figures out which architecture it runs on
445 :     - uses ./makeml a bit more cleverly
446 :     - never invokes ./installml (and, thus, does not clobber your
447 :     good and working installation of sml in case something goes wrong)
448 :     - accepts max iteration count using option "-iter <n>"
449 :     - accepts a "base" name using option "-base <base>"
450 : dbm 610
451 : blume 618 It does not build any extraneous heap images but directly rebuilds
452 :     bin- and boot-hierarchies using makeml's "-rebuild" switch. Finally,
453 :     it can incorporate existing bin- and boot- hierarchies. For example,
454 :     suppose the base is set to "sml" (which is the default). Then it
455 :     successively builds
456 : dbm 610
457 : blume 618 sml.bin.<arch>-unix and sml.boot.<arch>-unix
458 :     then sml1.bin.<arch>-unix and sml1.boot.<arch>-unix
459 :     then sml2.bin.<arch>-unix and sml2.boot.<arch>-unix
460 :     ...
461 :     then sml<n>.bin.<arch>-unix and sml<n>.boot.<arch>-unix
462 : dbm 610
463 : blume 618 and so on. If any of these already exist, it will just use what's
464 :     there. In particular, many people will have the initial set of bin
465 :     and boot files around, so this saves time for at least one full
466 :     rebuild. Having sets of the form <base><k>.{bin,boot}.<arch>-unix for
467 :     <k>=1,2,... is normally not a good idea when invoking fixpt. However,
468 :     they might be the result of an earlier partial run of fixpt (which
469 :     perhaps got accidentially killed). In this case, fixpt will quickly
470 :     move through what exists before continuing where it left off earlier,
471 :     and, thus, saves a lot of time.
472 : dbm 610
473 : blume 618 Runtime system code
474 : dbm 610
475 : blume 618 - fixed several gcc -Wall warnings that were caused by missing header
476 :     files, missing initializations, etc., in runtime (not all warnings
477 :     eliminated, though)
478 :     - hand to "un-fix" some of them later because they broke the HPPA compile
479 : dbm 610
480 : blume 618 ======================================================================
481 :     CM
482 :     ======================================================================
483 : dbm 610
484 : blume 618 Several manual updates
485 : dbm 610
486 : blume 618 I always try to keep the manual in sync with CM's latest features.
487 : dbm 610
488 : blume 618 Bootstrap compilation
489 : dbm 610
490 : blume 618 No more "CMB.deliver"
491 : dbm 610
492 : blume 618 - All work is done by CMB.make (as it used to be in the old CM).
493 :     - CMB.make can be used even with existing bootfiles, i.e., bootfiles do
494 :     not have to be removed beforehand.
495 :     - In "paranoid mode" CM checks a stable libraries CRC checksum to
496 :     verify that it is "valid". (In "normal mode", such checks do not
497 :     occur.) Paranoid mode is used for bootstrap compilation. This is
498 :     what makes it possible to re-use existing bootfiles.
499 : dbm 610
500 : blume 618 Initial glue code (init.cmi)
501 : dbm 610
502 : blume 618 - treated as a genuine library now
503 :     - there are no more "built-in" modules
504 : dbm 610
505 : blume 618 CM API
506 : dbm 610
507 : blume 618 CM.Anchor.anchor instead of CM.Anchor.{set,cancel}
508 :     - Upon request by Elsa. Anchors now controlled by get-set-pair
509 :     like most other CM state variables.
510 : dbm 610
511 : blume 618 CM tools:
512 :     - It is now possible to have tools that accept additional
513 :     "command line" parameters (specified in the .cm file at each
514 :     instance where the tool's class is used).
515 : dbm 610
516 : blume 618 - The parser understands named parameters and recursive options.
517 : dbm 610
518 : blume 618 - new "make" and "shell" tools added
519 :     * facilitate fairly seemless hookup to portions of code
520 :     managed using Makefiles or Shell scripts.
521 : dbm 610
522 : blume 618 - There are no classes "shared" or "private" anymore. Instead,
523 :     the sharing annotation is now a parameter to the "sml" class.
524 : dbm 610
525 : blume 618 - Tools.registerStdShellCmdTool (from smlnj/cm/tool.cm) takes an
526 :     additional argument called "template" which is an optional
527 :     string that specifiel the layout of the tool command line. See
528 :     the CM manual for explanation.
529 : dbm 610
530 : blume 618 - A special-purpose tool can be "registered" by simply dropping
531 :     the corresponding <...>-tool.cm (and/or <...>-ext.cm) into the
532 :     same directory where the .cm file lives that uses this tool.
533 :     (The behavior/misfeature until now was to look for the tool
534 :     description files in the current working directory.) As
535 :     before, tool description files could also be anchored -- in
536 :     which case they can live anywhere they like. Following the
537 :     recent e-mail discussion, this change should make it easier to
538 :     have special-purpose tools that are shipped together with the
539 :     sources of the program that uses them.
540 :     Bug: such a tool does not get un-registered after being done
541 : dbm 610
542 : blume 618 Library names
543 : dbm 610
544 : blume 618 Library names have been completely re-organized.
545 :     Many libraries have been consolidated so that they share the same
546 :     path anchor. For example, all MLRISC-related libraries are
547 :     anchored at MLRISC, most libraries that are SML/NJ-specific are
548 :     under "smlnj". Notice that names like host-cmb.cm or
549 :     host-compiler.cm no longer exist. See system/README for a
550 :     complete description of the new naming scheme. Quick reference:
551 : dbm 610
552 : blume 618 host-cmb.cm -> smlnj/cmb.cm
553 :     host-compiler.cm -> smlnj/compiler.cm
554 :     full-cm.cm -> smlnj/cm.cm
555 :     <arch>-<os>.cm -> smlnj/cmb/<arch>-<os>.cm
556 :     <arch>-compiler.cm -> smlnj/compiler/<arch>.cm
557 : dbm 610
558 : blume 618 CM bug fixes
559 : dbm 610
560 : blume 618 - exceptions in user code are being passed through (i.e., reach top level)
561 :     - more bugs in paranoia mode fixed
562 :     - bug related to checking group owners fixed
563 :     - better error handling (suppresses many followup-messages)
564 : dbm 610
565 : blume 618 Internals
566 : dbm 610
567 : blume 618 "Global" modmap:
568 :     CM now maintains one "global" modmap that is used for all stable
569 :     libraries. The use of such a global modmap maximizes sharing and
570 :     minimizes the need for re-traversing parts of environments during
571 :     modmap construction. (However, this has minor impact since modmap
572 :     construction seems to account for just one percent or less of total
573 :     compile time.)
574 : dbm 610
575 : blume 618 ======================================================================
576 :     Compiler Internals
577 :     ======================================================================
578 : dbm 610
579 : blume 618 Environment data structures: major changes
580 : dbm 610
581 : blume 618 No CMStaticEnv anymore.
582 :     - no CMEnv, no "BareEnvironment" (actually, _only_ BareEnvironment,
583 :     but it is called Environment), no conversions between different
584 :     kinds of static environments
585 : dbm 610
586 : blume 618 - There is still a notion of a "modmap", but such modmaps are generated
587 :     on demand at the time when they are needed. This sounds slow, but I
588 :     sped up the code that generates modmaps enough for this not to lead to
589 :     a slowdown of the compiler (at least I didn't detect any).
590 : dbm 610
591 : blume 618 - To facilitate rapid modmap generation, static environments now
592 :     contain an (optional) "modtree" structure. Modtree annotations are
593 :     constructed by the unpickler during unpickling. (This means that
594 :     the elaborator does not have to worry about modtrees at all.)
595 :     Modtrees have the advantage that they are compositional in the same
596 :     way as the environment data structure itself is compositional.
597 :     As a result, modtrees never hang on to parts of an environment that
598 :     has already been rendered "stale" by filtering or rebinding.
599 : dbm 610
600 : blume 618 - all files that I touched now compile without warnings (other than
601 :     "polyEqual").
602 :    
603 :     - compiler now tends to run "leaner" (i.e., ties up less memory in
604 :     redundant modmaps)
605 :    
606 :     Stats phase "genmap" added
607 :    
608 :     - measures time spent during on-the-fly modmap generation
609 :    
610 :     Changes on behalf of CM
611 :    
612 :     Compiler.CMSA eliminated
613 :     - No longer supported by CM anyway.
614 :    
615 :     Fixed bugs in pickler that kept biting Stefan
616 :     - past refs to past refs (was caused by the possibility that
617 :     ad-hoc sharing is more discriminating than hash-cons sharing)
618 :     - integer overflow on LargeInt.minInt
619 :    
620 :     Handling of "core" environment:
621 :    
622 :     I eliminated coreEnv from compInfo. Access to the "Core"
623 :     structure is now done via the ordinary static environment that is
624 :     context to each compilation unit.
625 :    
626 :     To this end, I arranged that instead of "structure Core" a
627 :     "structure _Core" is bound in the pervasive environment. Core
628 :     access is done via _Core (which can never be accidentially rebound
629 :     because _Core is not a legal surface-syntax symbol).
630 :    
631 :     The current solution is much cleaner because the core environment
632 :     is now simply part of the pervasive environment which is part of
633 :     every compilation unit's context anyway. In particular, this
634 :     eliminates all special-case handling that was necessary until now
635 :     in order to deal with dynamic and symbolic parts of the core
636 :     environment.
637 :    
638 :     Remaining hackery (to bind the "magic" symbol _Core) is localized
639 :     in the compilation mananger's bootstrap compiler (actually: in the
640 :     "init group" handling). See the comments in
641 :     src/system/smlnj/init/init.cmi for more details.
642 :    
643 :     I also tried to track down all mentions of "Core" (as string
644 :     argument to Symbol.strSymbol) in the compiler and replaced them
645 :     with a reference to the new CoreSym.coreSym. Seems cleaner since
646 :     the actual name appears in one place only.

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