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/HISTORY
 [smlnj] / sml / trunk / HISTORY

# Annotation of /sml/trunk/HISTORY

Revision 1222 - (view) (download)

 1 : dbm 570 This is the HISTORY file for the Yale SML/NJ CVS repository. 2 : 3 : An entry should be made for _every_ commit to the repository. 4 : The entries in this file will be used when creating the README 5 : for new versions, so keep that in mind when writing the 6 : description. 7 : 8 : The form of an entry should be: 9 : 10 : Name: 11 : george 1003 Date: yyyy/mm/dd 12 : dbm 570 Tag: 13 : Description: 14 : leunga 1142 15 : george 1136 ---------------------------------------------------------------------- 16 : george 1222 Name: Lal George 17 : Date: 2002/05/23 12:21:40 EDT 18 : Tag: george-20020523-visual-labels 19 : Description: 20 : 21 : Labels are now displayed in the graphical output to make 22 : the fall-through and target blocks obvious. 23 : 24 : ---------------------------------------------------------------------- 25 : blume 1204 Name: Matthias Blume 26 : blume 1220 Date: 2002/05/22 11:03:00 EDT 27 : Tag: blume-20020522-shrink 28 : Description: 29 : 30 : John tweaked yesterday's fix for 1131 to handle an out-of-memory 31 : situation that comes up when allocating huge arrays. 32 : 33 : ---------------------------------------------------------------------- 34 : Name: Matthias Blume 35 : blume 1215 Date: 2002/05/21 16:00:00 EDT 36 : Tag: Release_110_40 37 : Description: 38 : 39 : New working release (110.40). New bootfiles. 40 : 41 : [Also: John Reppy fixed GC bug 1131.] 42 : 43 : ---------------------------------------------------------------------- 44 : Name: Matthias Blume 45 : blume 1212 Date: 2002/05/21 12:35:00 EDT 46 : Tag: blume-20020521-cmdoc 47 : Description: 48 : 49 : CM documentation update. 50 : 51 : ---------------------------------------------------------------------- 52 : Name: Matthias Blume 53 : blume 1211 Date: 2002/05/21 10:55:00 EDT 54 : Tag: blume-20020521-misc 55 : Description: 56 : 57 : - John tweaked runtime to be silent on heap export (except when 58 : GC messages are on). 59 : - I added a few more things (cross-compiling versions of CMB) to 60 : config/preloads (as suggestions). 61 : 62 : ---------------------------------------------------------------------- 63 : Name: Matthias Blume 64 : blume 1208 Date: 2002/05/20 22:25:00 EDT 65 : Tag: blume-20020520-controls 66 : Description: 67 : 68 : - Added ControlUtil structure to control-lib.cm. 69 : - Use it throughout. 70 : - Used Controls facility to define MLRISC controls (as opposed to 71 : registering MLRISC control ref cells with Controls after the fact) 72 : - Fixed messed-up controls priorities. 73 : 74 : * Removed again all the stuff from config/preloads that one wouldn't 75 : be able to preload at the time the initial heap image is built. 76 : (Many libraries, e.g., CML, do not exist yet at this time. The 77 : only libraries that can be preloaded via config/preloads are those 78 : that come bundled with the bootfiles.) 79 : 80 : ---------------------------------------------------------------------- 81 : Name: Matthias Blume 82 : blume 1204 Date: 2002/05/20 10:59:00 EDT 83 : Tag: blume-20020520-preloads 84 : Description: 85 : 86 : Added a lot of commented-out suggestions for things to be included 87 : in config/preloads. 88 : 89 : ---------------------------------------------------------------------- 90 : leunga 1203 Name: Allen Leung 91 : Date: 2002/05/18 14:20:00 EDT 92 : Tag: leunga-20020518-mdl 93 : Description: 94 : 95 : o Made the mdl tool stuff compile and run again. 96 : o I've disabled all the stuff that depends on RTL specifications; they 97 : are all badly broken anyway. 98 : 99 : ---------------------------------------------------------------------- 100 : blume 1201 Name: Matthias Blume 101 : Date: 2002/05/17 16:49:00 EDT 102 : Tag: blume-20020517-controls 103 : Description: 104 : 105 : 0. John Reppy made several modifications to the SML/NJ library. 106 : In particular, there is a shiny new controls-lib.cm. 107 : 108 : 1. Pushed new controls interface through compiler so that everything 109 : compiles again. 110 : 111 : 2. Added FormatComb and FORMAT_COMB to the CML version of the 112 : SML/NJ library (so that CML compiles again). 113 : 114 : 3. Modified init scripts because XXX_DEFAULT environment variables 115 : are no longer with us. (Boot-time initialization is now done 116 : using the same environment variables that are also used for 117 : startup-time initialization of controls.) 118 : 119 : ---------------------------------------------------------------------- 120 : george 1192 Name: Lal George 121 : Date: 2002/05/15 09:20:10 EDT 122 : Tag: george-20020515-pseudo-op-decls 123 : Description: 124 : 125 : All pseudo-ops emitted before the first segment declaration 126 : such as TEXT, DATA, and BSS directives are assumed to be global 127 : declarations and are emitted first in the assembly file. This is 128 : useful in a number of situations where one has pseudo-ops that are not 129 : specific to any segment, and also works around the constraint that one 130 : cannot have client pseudo-ops in the TEXT segment. 131 : 132 : Because no segment is associated with these declarations it is 133 : an error to allocate any space or objects before the first segment 134 : directive and an exception will be raised. However, we cannot make 135 : this check for client pseudo-ops. 136 : 137 : These top level declarations are a field in the CFG graph_info. 138 : In theory you can continue to add to this field after the CFG has been 139 : built -- provided you know what you are doing;-) 140 : 141 : ---------------------------------------------------------------------- 142 : blume 1175 Name: Matthias Blume 143 : blume 1190 Date: 2002/05/13 16:40:00 EDT 144 : Tag: blume-20020513-pp-etc 145 : Description: 146 : 147 : A few minor bugfixes: 148 : 149 : - Stopgap measure for bug recently reported by Elsa Gunter (ppDec). 150 : (Bogus printouts for redefined bindings still occur. Compiler 151 : bug should no longer occur now. We need to redo the prettyprinter 152 : from scratch.) 153 : 154 : - CM pathname printer now also adds escape sequences for ( and ) 155 : 156 : - commend and docu fixes for ml-nlffi 157 : 158 : ---------------------------------------------------------------------- 159 : Name: Matthias Blume 160 : blume 1189 Date: 2002/05/10 16:40:00 EDT 161 : Tag: blume-20020510-erg-textio 162 : Description: 163 : 164 : Applied the following bugfix provided by Emden Gansner: 165 : 166 : Output is corrupted when outputSubstr is used rather than output. 167 : 168 : The problem occurs when a substring 169 : 170 : ss = (s, dataStart, dataLen) 171 : 172 : where dataStart > 0, fills a stream buffer with avail bytes left. 173 : avail bytes of s, starting at index dataStart, are copied into the 174 : buffer, the buffer is flushed, and then the remaining dataLen-avail 175 : bytes of ss are copied into the beginning of the buffer. Instead of 176 : starting this copy at index dataStart+avail in s, the current code 177 : starts the copy at index avail. 178 : 179 : Fix: 180 : In text-io-fn.sml, change line 695 from 181 : val needsFlush = copyVec(v, avail, dataLen-avail, buf, 0) 182 : to 183 : val needsFlush = copyVec(v, dataStart+avail, dataLen-avail, buf, 0) 184 : 185 : ---------------------------------------------------------------------- 186 : Name: Matthias Blume 187 : blume 1186 Date: 2002/04/12 13:55:00 EDT 188 : Tag: blume-20020412-assyntax 189 : Description: 190 : 191 : 1. Grabbed newer assyntax.h from the XFree86 project. 192 : 2. Fiddled with how to compile X86.prim.asm without warnings. 193 : 3. (Very) Minor cleanup in CM. 194 : 195 : ---------------------------------------------------------------------- 196 : Name: Matthias Blume 197 : blume 1185 Date: 2002/04/01 (no joke!) 17:07:00 EST 198 : Tag: blume-20020401-x86div 199 : Description: 200 : 201 : Added full support for div/mod/rem/quot on the x86, using the machine 202 : instruction's two results (without clumsily recomputing the remainder) 203 : directly where appropriate. 204 : 205 : Some more extensive power-of-two support was added to the x86 instruction 206 : selector (avoiding expensive divs, mods, and muls where they can be 207 : replaced with cheaper shifts and masks). However, this sort of thing 208 : ought to be done earlier, e.g., within the CPS optimizer so that 209 : all architectures benefit from it. 210 : 211 : The compiler compiles to a fixed point, but changes might be somewhat 212 : fragile nevertheless. Please, report any strange things that you might 213 : see wrt. div/mod/quot/rem... 214 : 215 : ---------------------------------------------------------------------- 216 : Name: Matthias Blume 217 : blume 1184 Date: 2002/03/29 17:22:00 218 : Tag: blume-20020329-div 219 : Description: 220 : 221 : Fixed my broken div/mod logic. Unfortunately, this means that the 222 : inline code for div/mod now has one more comparison than before. 223 : Fast paths (quotient > 0 or remainder = 0) are not affected, though. 224 : The problem was with quotient = 0, because that alone does not tell 225 : us which way the rounding went. One then has to look at whether 226 : remainder and divisor have the same sign... :( 227 : 228 : Anyway, I replaced the bootfiles with fresh ones... 229 : 230 : ---------------------------------------------------------------------- 231 : Name: Matthias Blume 232 : blume 1183 Date: 2002/03/29 14:10:00 EST 233 : Tag: blume-20020329-inlprims 234 : Description: 235 : 236 : NEW BOOTFILES!!! Version number bumped to 110.39.3. 237 : 238 : Primops have changed. This means that the bin/boot-file formats have 239 : changed as well. 240 : 241 : To make sure that there is no confusion, I made a new version. 242 : 243 : 244 : CHANGES: 245 : 246 : * removed REMT from mltree (remainder should never overflow). 247 : 248 : * added primops to deal with divisions of all flavors to the frontend 249 : 250 : * handled these primops all the way through so they map to their respective 251 : MLRISC support 252 : 253 : * used these primops in the implementation of Int, Int32, Word, Word32 254 : 255 : * removed INLDIV, INLMOD, and INLREM as they are no longer necessary 256 : 257 : * parameterized INLMIN, INLMAX, and INLABS by a numkind 258 : 259 : * translate.sml now deals with all flavors of INL{MIN,MAX,ABS}, including 260 : floating point 261 : 262 : * used INL{MIN,MAX,ABS} in the implementation of Int, Int32, Word, Word32, 263 : and Real (but Real.abs maps to a separate floating-point-only primop) 264 : 265 : 266 : TODO items: 267 : 268 : * Hacked Alpha32 instruction selection, disabling the selection of REMx 269 : instructions because the machine instruction encoder cannot handle 270 : them. (Hppa, PPC, and Sparc instruction selection did not handle 271 : REM in the first place, and REM is supported by the x86 machine coder.) 272 : 273 : * Handle DIV and MOD with DIV_TO_NEGINF directly in the x86 instruction 274 : selection phase. (The two can be streamlined because the hardware 275 : delivers both quotient and remainder at the same time anyway.) 276 : 277 : * Think about what to do with "valOf(Int32.minInt) div ~1" and friends. 278 : (Currently the behavior is inconsistent both across architectures and 279 : wrt. the draft Basis spec.) 280 : 281 : * Word8 should eventually be handled natively, too. 282 : 283 : * There seems to be one serious bug in mltree-gen.sml. It appears, though, 284 : as if there currently is no execution path that could trigger it in 285 : SML/NJ. (The assumptions underlying functions arith and promotable do not 286 : hold for things like multiplication and division.) 287 : 288 : ---------------------------------------------------------------------- 289 : Name: Matthias Blume 290 : blume 1181 Date: 2002/03/27 16:27:00 EST 291 : Tag: blume-20020327-mlrisc-divisions 292 : Description: 293 : 294 : Added support for all four division operations (ML's div, mod, quot, 295 : and rem) to MLRISC. In the course of doing so, I also rationalized 296 : the naming (no more annoying switch-around of DIV and QUOT), by 297 : parameterizing the operation by div_rounding_mode (which can be either 298 : DIV_TO_ZERO or DIV_TO_NEGINF). 299 : 300 : The generic MLTreeGen functor takes care of compiling all four 301 : operations down to only round-to-zero div. 302 : 303 : Missing pieces: 304 : 305 : * Doing something smarter than relying on MLTreeGen on architectures 306 : like, e.g., the x86 where hardware division delivers both quotient and 307 : remainder at the same time. With this, the implementation of the 308 : round-to-neginf operations could be further streamlined. 309 : 310 : * Remove inlining support for div/mod/rem from the frontend and replace it 311 : with primops that get carried through to the backend. Do this for all 312 : int and word types. 313 : 314 : ---------------------------------------------------------------------- 315 : Name: Matthias Blume 316 : blume 1180 Date: 2002/03/25 17:25:00 EST 317 : Tag: blume-20020325-divmod 318 : Description: 319 : 320 : I improved (hopefully without breaking them) the implementation of Int.div, 321 : Int.mod, and Int.rem. For this, the code in translate.sml now takes 322 : advantage of the following observations: 323 : 324 : Let q = x quot y r = x rem y 325 : d = x div y m = x mod y 326 : 327 : where "quot" is the round-to-zero version of integer division that 328 : hardware usually provides. Then we have: 329 : 330 : r = x - q * y where neither the * nor the - will overflow 331 : d = if q >= 0 orelse x = q * y then q else q - 1 332 : where neither the * nor the - will overflow 333 : m = if q >= 0 orelse r = 0 then r else r + y 334 : where the + will not overflow 335 : 336 : This results in substantial simplification of the generated code. 337 : The following table shows the number of CFG nodes and edges generated 338 : for 339 : fun f (x, y) = x OPER y 340 : (* with OPER \in div, mod, quot, rem *) 341 : 342 : 343 : OPER | nodes(old) | edges(old) | nodes(new) | edges(new) 344 : -------------------------------------------------------- 345 : div | 24 | 39 | 12 | 16 346 : mod | 41 | 71 | 12 | 16 347 : quot | 8 | 10 | 8 | 10 348 : rem | 10 | 14 | 8 | 10 349 : 350 : 351 : ---------------------------------------------------------------------- 352 : Name: Matthias Blume 353 : blume 1179 Date: 2002/03/25 22:06:00 EST 354 : Tag: blume-20020325-cprotobug 355 : Description: 356 : 357 : Fixed a bug in cproto (c prototype decoder). 358 : 359 : ---------------------------------------------------------------------- 360 : Name: Matthias Blume 361 : blume 1178 Date: 2002/03/25 16:00:00 EST 362 : Tag: blume-20020325-raw-primops 363 : Description: 364 : 365 : I did some cleanup to Allen's new primop code and 366 : replaced yesterday's bootfiles with new ones. 367 : (But they are stored in the same place.) 368 : 369 : ---------------------------------------------------------------------- 370 : Name: Matthias Blume 371 : blume 1175 Date: 2002/03/24 22:40:00 EST 372 : Tag: blume-20020324-bootfiles 373 : Description: 374 : 375 : Made the bootfiles that Allen asked for. 376 : 377 : ---------------------------------------------------------------------- 378 : leunga 1174 Name: Allen Leung 379 : Date: 2002/03/23 15:50:00 EST 380 : Tag: leunga-20020323-flint-cps-rcc-primops 381 : Description: 382 : 383 : 1. Changes to FLINT primops: 384 : 385 : (* make a call to a C-function; 386 : * The primop carries C function prototype information and specifies 387 : * which of its (ML-) arguments are floating point. C prototype 388 : * information is for use by the backend, ML information is for 389 : * use by the CPS converter. *) 390 : | RAW_CCALL of { c_proto: CTypes.c_proto, 391 : ml_args: ccall_type list, 392 : ml_res_opt: ccall_type option, 393 : reentrant : bool 394 : } option 395 : (* Allocate uninitialized storage on the heap. 396 : * The record is meant to hold short-lived C objects, i.e., they 397 : * are not ML pointers. With the tag, the representation is 398 : * the same as RECORD with tag tag_raw32 (sz=4), or tag_fblock (sz=8) 399 : *) 400 : | RAW_RECORD of {tag:bool,sz:int} 401 : and ccall_type = CCALL_INT32 | CCALL_REAL64 | CCALL_ML_PTR 402 : 403 : 2. These CPS primops are now overloaded: 404 : 405 : rawload of {kind:numkind} 406 : rawstore of {kind:numkind} 407 : 408 : The one argument form is: 409 : 410 : rawload {kind} address 411 : 412 : The two argument form is: 413 : 414 : rawload {kind} [ml object, byte-offset] 415 : 416 : 3. RAW_CCALL/RCC now takes two extra arguments: 417 : 418 : a. The first is whether the C call is reentrant, i.e., whether 419 : ML state should be saved and restored. 420 : b. The second argument is a string argument specifying the name of 421 : library and the C function. 422 : 423 : These things are currently not handled in the code generator, yet. 424 : 425 : 4. In CProto, 426 : 427 : An encoding type of "bool" means "ml object" and is mapped into 428 : C prototype of PTR. Note that "bool" is different than "string", 429 : even though "string" is also mapped into PTR, because "bool" 430 : is assigned an CPS type of BOGt, while "string" is assigned INT32t. 431 : 432 : 5. Pickler/unpicker 433 : 434 : Changed to handle RAW_RECORD and newest RAW_CCALL 435 : 436 : 6. MLRiscGen, 437 : 438 : 1. Changed to handle the new rawload/rawstore/rawrecord operators. 439 : 2. Code for handling C Calls has been moved to a new module CPSCCalls, 440 : in the file CodeGen/cpscompile/cps-c-calls.sml 441 : 442 : 7. Added the conditional move operator 443 : 444 : condmove of branch 445 : 446 : to cps. Generation of this is still buggy so it is currently 447 : disabled. 448 : 449 : ---------------------------------------------------------------------- 450 : george 1168 Name: Lal George 451 : Date: 2002/03/22 14:18:25 EST 452 : george 1171 Tag: george-20020322-cps-branch-prob 453 : george 1168 Description: 454 : 455 : george 1171 Implemented the Ball-Larus branch prediction-heuristics, and 456 : george 1168 incorporated graphical viewers for control flow graphs. 457 : 458 : george 1171 Ball-Larus Heuristics: 459 : george 1168 --------------------- 460 : See the file compiler/CodeGen/cpscompile/cpsBranchProb.sml. 461 : 462 : By design it uses the Dempster-Shafer theory for combining 463 : probabilities. For example, in the function: 464 : 465 : fun f(n,acc) = if n = 0 then acc else f(n-1, n*acc) 466 : 467 : the ball-larus heuristics predicts that the n=0 is unlikely 468 : (OH-heuristic), and the 'then' branch is unlikely because of the 469 : RH-heuristic -- giving the 'then' branch an even lower combined 470 : george 1171 probability using the Dempster-Shafer theory. 471 : george 1168 472 : Finally, John Reppy's loop analysis in MLRISC, further lowers the 473 : probability of the 'then' branch because of the loop in the else 474 : branch. 475 : 476 : 477 : Graphical Viewing: 478 : ------------------ 479 : I merely plugged in Allen's graphical viewers into the compiler. The 480 : additional code is not much. At the top level, saying: 481 : 482 : Control.MLRISC.getFlag "cfg-graphical-view" := true; 483 : 484 : will display the graphical view of the control flow graph just before 485 : back-patching. daVinci must be in your path for this to work. If 486 : daVinci is not available, then the default viewer can be changed 487 : using: 488 : 489 : Control.MLRISC.getString "viewer" 490 : 491 : which can be set to "dot" or "vcg" for the corresponding viewers. Of 492 : course, these viewers must be in your path. 493 : 494 : The above will display the compilation unit at the level of clusters, 495 : many of which are small, boring, and un-interesting. Also setting: 496 : 497 : Control.MLRISC.getInt "cfg-graphical-view_size" 498 : 499 : will display clusters that are larger than the value set by the above. 500 : 501 : 502 : ---------------------------------------------------------------------- 503 : blume 1159 Name: Matthias Blume 504 : Date: 2002/03/21 22:20:00 EST 505 : Tag: blume-20020321-kmp-bugfix 506 : Description: 507 : 508 : Changed the interface to the KMP routine in PreString and fixed 509 : a minor bug in one place where it was used. 510 : 511 : ---------------------------------------------------------------------- 512 : leunga 1156 Name: Allen Leung 513 : leunga 1158 Date: 2002/03/21 20:30:00 EST 514 : Tag: leunga-20020321-cfg 515 : Description: 516 : 517 : Fixed a potential problem in cfg edge splitting. 518 : 519 : ---------------------------------------------------------------------- 520 : Name: Allen Leung 521 : leunga 1156 Date: 2002/03/21 17:15:00 EST 522 : Tag: leunga-20020321-x86-fp-cfg 523 : Description: 524 : 525 : 1. Recoded the buggy parts of x86-fp. 526 : 527 : a. All the block reordering code has been removed. 528 : We now depend on the block placement phases to do this work. 529 : 530 : b. Critical edge splitting code has been simplified and moved into the 531 : CFG modules, as where they belong. 532 : 533 : Both of these were quite buggy and complex. The code is now much, much 534 : simpler. 535 : 536 : 2. X86 backend. 537 : 538 : a. Added instructions for 64-bit support. Instruction selection for 539 : 64-bit has not been committed, however, since that 540 : requires changes to MLTREE which haven't been approved by 541 : Lal and John. 542 : 543 : b. Added support for FUCOMI and FUCOMIP when generating code for 544 : PentiumPro and above. We only generate these instructions in 545 : the fast-fp mode. 546 : 547 : c. Added cases for JP and JNP in X86FreqProps. 548 : 549 : 3. CFG 550 : 551 : CFG now has a bunch of methods for edge splitting and merging. 552 : 553 : 4. Machine description. 554 : 555 : John's simplification of MLTREE_BASIS.fcond broke a few machine 556 : description things: 557 : 558 : rtl-build.{sig,sml} and hppa.mdl fixed. 559 : 560 : NOTE: the machine description stuff in the repository is still broken. 561 : Again, I can't put my fixes in because that involves 562 : changes to MLTREE. 563 : 564 : ---------------------------------------------------------------------- 565 : blume 1145 Name: Matthias Blume 566 : blume 1155 Date: 2002/03/20 15:55:00 EST 567 : Tag: blume-20020320-kmp 568 : Description: 569 : 570 : Implemented Knuth-Morris-Pratt string matching in PreString and used 571 : it for String.isSubstring, Substring.isSubstring, and 572 : Substring.position. 573 : 574 : (Might need some stress-testing. Simple examples worked fine.) 575 : 576 : ---------------------------------------------------------------------- 577 : Name: Matthias Blume 578 : blume 1152 Date: 2002/03/19 16:37:00 EST 579 : Tag: blume-20020319-witnesses 580 : Description: 581 : 582 : Added a structure C.W and functions convert/Ptr.convert to ml-nlffi-lib. 583 : 584 : This implements a generic mechanism for changing constness qualifiers 585 : anywhere within big C types without resorting to outright "casts". 586 : (So far, functions such as C.rw/C.ro or C.Ptr.rw/C.Ptr.ro only let you 587 : modify the constness at the outermost level.) 588 : The implementation of "convert" is based on the idea of "witness" 589 : values -- values that are not used by the operation but whose types 590 : "testify" to their applicability. On the implementation side, "convert" 591 : is simply a projection (returning its second curried argument). With 592 : cross-module inlining, it should not result in any machine code being 593 : generated. 594 : 595 : ---------------------------------------------------------------------- 596 : Name: Matthias Blume 597 : blume 1148 Date: 2002/03/15 16:40:00 EST 598 : Tag: blume-20020315-basis 599 : Description: 600 : 601 : Provided (preliminary?) implementations for 602 : 603 : {String,Substring}.{concatWith,isSuffix,isSubstring} 604 : 605 : and 606 : 607 : Substring.full 608 : 609 : Those are in the Basis spec but they were missing in SML/NJ. 610 : 611 : ---------------------------------------------------------------------- 612 : Name: Matthias Blume 613 : blume 1145 Date: 2002/03/14 21:30:00 EST 614 : Tag: blume-20020314-controls 615 : Description: 616 : 617 : Controls: 618 : --------- 619 : 620 : 1. Factored out the recently-added Controls : CONTROLS stuff and put 621 : it into its own library /controls-lib.cm. The source tree for 622 : this is under src/smlnj-lib/Controls. 623 : 624 : 2. Changed the names of types and functions in this interface, so they 625 : make a bit more "sense": 626 : 627 : module -> registry 628 : 'a registry -> 'a group 629 : 630 : 3. The interface now deals in ref cells only. The getter/setter interface 631 : is (mostly) gone. 632 : 633 : 4. Added a function that lets one register an already-existing ref cell. 634 : 635 : 5. Made the corresponding modifications to the rest of the code so that 636 : everything compiles again. 637 : 638 : 6. Changed the implementation of Controls.MLRISC back to something closer 639 : to the original. In particular, this module (and therefore MLRISC) 640 : does not depend on Controls. There now is some link-time code in 641 : int-sys.sml that registers the MLRISC controls with the Controls 642 : module. 643 : 644 : CM: 645 : --- 646 : 647 : * One can now specify the lambda-split aggressiveness in init.cmi. 648 : 649 : ---------------------------------------------------------------------- 650 : leunga 1142 Name: Allen Leung 651 : Date: 2002/03/13 17:30:00 EST 652 : Tag: leunga-20020313-x86-fp-unary 653 : Description: 654 : 655 : Bug fix for: 656 : 657 : > leunga@weaselbane:~/Yale/tmp/sml-dist{21} bin/sml 658 : > Standard ML of New Jersey v110.39.1 [FLINT v1.5], March 08, 2002 659 : > - fun f(x,(y,z)) = Real.~ y; 660 : > [autoloading] 661 : > [autoloading done] 662 : > fchsl (%eax), 184(%esp) 663 : > Error: MLRisc bug: X86MCEmitter.emitInstr 664 : > 665 : > uncaught exception Error 666 : > raised at: ../MLRISC/control/mlriscErrormsg.sml:16.14-16.19 667 : 668 : The problem was that the code generator did not generate any fp registers 669 : in this case, and the ra didn't know that it needed to run the X86FP phase to 670 : translate the pseudo fp instruction. This only happened with unary fp 671 : operators in certain situations. 672 : 673 : ---------------------------------------------------------------------- 674 : blume 1137 Name: Matthias Blume 675 : blume 1140 Date: 2002/03/13 14:00:00 EST 676 : Tag: blume-20020313-overload-etc 677 : Description: 678 : 679 : 1. Added _overload as a synonym for overload for backward compatibility. 680 : (Control.overloadKW must be true for either version to be accepted.) 681 : 682 : 2. Fixed bug in install script that caused more things to be installed 683 : than what was requested in config/targets. 684 : 685 : 3. Made CM aware of the (_)overload construct so that autoloading 686 : works. 687 : 688 : ---------------------------------------------------------------------- 689 : Name: Matthias Blume 690 : blume 1138 Date: 2002/03/12 22:03:00 EST 691 : Tag: blume-20020312-url 692 : Description: 693 : 694 : Forgot to update BOOT and srcarchiveurl. 695 : 696 : ---------------------------------------------------------------------- 697 : Name: Matthias Blume 698 : blume 1137 Date: 2002/03/12 17:30:00 EST 699 : Tag: blume-20020312-version110392 700 : Description: 701 : 702 : Yet another version number bump (because of small changes to the 703 : binfile format). Version number is now 110.39.2. NEW BOOTFILES! 704 : 705 : Changes: 706 : 707 : The new pid generation scheme described a few weeks ago was overly 708 : complicated. I implemented a new mechanism that is simpler and 709 : provides a bit more "stability": Once CM has seen a compilation 710 : unit, it keeps its identity constant (as long as you do not delete 711 : those crucial CM/GUID/* files). This means that when you change 712 : blume 1145 an interface, compile, then go back to the old interface, and 713 : blume 1137 compile again, you arrive at the original pid. 714 : 715 : There now also is a mechanism that instructs CM to use the plain 716 : environment hash as a module's pid (effectively making its GUID 717 : the empty string). For this, "noguid" must be specified as an 718 : option to the .sml file in question within its .cm file. 719 : This is most useful for code that is being generated by tools such 720 : as ml-nlffigen (because during development programmers tend to 721 : erase the tool's entire output directory tree including CM's cached 722 : GUIDs). "noguid" is somewhat dangerous (since it can be used to locally 723 : revert to the old, broken behavior of SML/NJ, but in specific cases 724 : where there is no danger of interface confusion, its use is ok 725 : (I think). 726 : 727 : ml-nlffigen by default generates "noguid" annotations. They can be 728 : turned off by specifying -guid in its command line. 729 : 730 : ---------------------------------------------------------------------- 731 : george 1136 Name: Lal George 732 : Date: 2002/03/12 12 14:42:36 EST 733 : Tag: george-20020312-frequency-computation 734 : Description: 735 : 736 : Integrated jump chaining and static block frequency into the 737 : compiler. More details and numbers later. 738 : leunga 1025 739 : george 1003 ---------------------------------------------------------------------- 740 : george 1133 Name: Lal George 741 : Date: 2002/03/11 11 22:38:53 EST 742 : Tag: george-20020311-jump-chain-elim 743 : Description: 744 : 745 : Tested the jump chain elimination on all architectures (except the 746 : hppa). This is on by default right now and is profitable for the 747 : alpha and x86, however, it may not be profitable for the sparc and ppc 748 : when compiling the compiler. 749 : 750 : The gc test will typically jump to a label at the end of the cluster, 751 : where there is another jump to an external cluster containing the actual 752 : code to invoke gc. This is to allow factoring of common gc invocation 753 : sequences. That is to say, we generate: 754 : 755 : f: 756 : testgc 757 : ja L1 % jump if above to L1 758 : 759 : L1: 760 : jmp L2 761 : 762 : 763 : After jump chain elimination the 'ja L1' instructions is converted to 764 : 'ja L2'. On the sparc and ppc, many of the 'ja L2' instructions may end 765 : up being implemented in their long form (if L2 is far away) using: 766 : 767 : jbe L3 % jump if below or equal to L3 768 : jmp L2 769 : L3: 770 : ... 771 : 772 : 773 : For large compilation units L2 may be far away. 774 : 775 : 776 : ---------------------------------------------------------------------- 777 : blume 1132 Name: Matthias Blume 778 : Date: 2002/03/11 13:30:00 EST 779 : Tag: blume-20020311-mltreeeval 780 : Description: 781 : 782 : A functor parameter was missing. 783 : 784 : ---------------------------------------------------------------------- 785 : leunga 1130 Name: Allen Leung 786 : leunga 1131 Date: 2002/03/11 10:30:00 EST 787 : leunga 1142 Tag: leunga-20020311-runtime-string0 788 : leunga 1131 Description: 789 : 790 : The representation of the empty string now points to a 791 : legal null terminated C string instead of unit. It is now possible 792 : to convert an ML string into C string with InlineT.CharVector.getData. 793 : This compiles into one single machine instruction. 794 : 795 : ---------------------------------------------------------------------- 796 : Name: Allen Leung 797 : leunga 1130 Date: 2002/03/10 23:55:00 EST 798 : Tag: leunga-20020310-x86-call 799 : Description: 800 : 801 : Added machine generation for CALL instruction (relative displacement mode) 802 : 803 : ---------------------------------------------------------------------- 804 : blume 1128 Name: Matthias Blume 805 : Date: 2002/03/08 16:05:00 806 : Tag: blume-20020308-entrypoints 807 : Description: 808 : 809 : Version number bumped to 110.39.1. NEW BOOTFILES! 810 : 811 : Entrypoints: non-zero offset into a code object where execution should begin. 812 : 813 : - Added the notion of an entrypoint to CodeObj. 814 : - Added reading/writing of entrypoint info to Binfile. 815 : - Made runtime system bootloader aware of entrypoints. 816 : - Use the address of the label of the first function given to mlriscGen 817 : as the entrypoint. This address is currently always 0, but it will 818 : not be 0 once we turn on block placement. 819 : - Removed the linkage cluster code (which was The Other Way(tm) of dealing 820 : with entry points) from mlriscGen. 821 : 822 : ---------------------------------------------------------------------- 823 : leunga 1127 Name: Allen Leung 824 : Date: 2002/03/07 20:45:00 EST 825 : Tag: leunga-20020307-x86-cmov 826 : Description: 827 : 828 : Bug fixes for CMOVcc on x86. 829 : 830 : 1. Added machine code generation for CMOVcc 831 : 2. CMOVcc is now generated in preference over SETcc on PentiumPro or above. 832 : 3. CMOVcc cannot have an immediate operand as argument. 833 : 834 : ---------------------------------------------------------------------- 835 : blume 1126 Name: Matthias Blume 836 : Date: 2002/03/07 16:15:00 EST 837 : Tag: blume-20020307-controls 838 : Description: 839 : 840 : This is a very large but mostly boring patch which makes (almost) 841 : every tuneable compiler knob (i.e., pretty much everything under 842 : Control.* plus a few other things) configurable via both the command 843 : line and environment variables in the style CM did its configuration 844 : until now. 845 : 846 : Try starting sml with '-h' (or, if you are brave, '-H') 847 : 848 : To this end, I added a structure Controls : CONTROLS to smlnj-lib.cm which 849 : implements the underlying generic mechanism. 850 : 851 : The interface to some of the existing such facilities has changed somewhat. 852 : For example, the MLRiscControl module now provides mkFoo instead of getFoo. 853 : (The getFoo interface is still there for backward-compatibility, but its 854 : use is deprecated.) 855 : 856 : The ml-build script passes -Cxxx=yyy command-line arguments through so 857 : that one can now twiddle the compiler settings when using this "batch" 858 : compiler. 859 : 860 : TODO items: 861 : 862 : We should go through and throw out all controls that are no longer 863 : connected to anything. Moreover, we should go through and provide 864 : meaningful (and correct!) documentation strings for those controls 865 : that still are connected. 866 : 867 : Currently, multiple calls to Controls.new are accepted (only the first 868 : has any effect). Eventually we should make sure that every control 869 : is being made (via Controls.new) exactly once. Future access can then 870 : be done using Controls.acc. 871 : 872 : Finally, it would probably be a good idea to use the getter-setter 873 : interface to controls rather than ref cells. For the time being, both 874 : styles are provided by the Controls module, but getter-setter pairs are 875 : better if thread-safety is of any concern because they can be wrapped. 876 : 877 : ***************************************** 878 : 879 : One bug fix: The function blockPlacement in three of the MLRISC 880 : backpatch files used to be hard-wired to one of two possibilities at 881 : link time (according to the value of the placementFlag). But (I 882 : think) it should rather sense the flag every time. 883 : 884 : ***************************************** 885 : 886 : Other assorted changes (by other people who did not supply a HISTORY entry): 887 : 888 : 1. the cross-module inliner now works much better (Monnier) 889 : 2. representation of weights, frequencies, and probabilities in MLRISC 890 : changed in preparation of using those for weighted block placement 891 : (Reppy, George) 892 : 893 : ---------------------------------------------------------------------- 894 : george 1116 Name: Lal George 895 : george 1124 Date: 2002/03/07 14:44:24 EST 2002 896 : Tag: george-20020307-weighted-block-placement 897 : 898 : Tested the weighted block placement optimization on all architectures 899 : (except the hppa) using AMPL to generate the block and edge frequencies. 900 : Changes were required in the machine properties to correctly 901 : categorize trap instructions. There is an MLRISC flag 902 : "weighted-block-placement" that can be used to enable weighted block 903 : placement, but this will be ineffective without block/edge 904 : frequencies (coming soon). 905 : 906 : 907 : ---------------------------------------------------------------------- 908 : Name: Lal George 909 : george 1116 Date: 2002/03/05 17:24:48 EST 910 : Tag: george-20020305-linkage-cluster 911 : 912 : george 1124 In order to support the block placement optimization, a new cluster 913 : is generated as the very first cluster (called the linkage cluster). 914 : It contains a single jump to the 'real' entry point for the compilation 915 : unit. Block placement has no effect on the linkage cluster itself, but 916 : all the other clusters have full freedom in the manner in which they 917 : reorder blocks or functions. 918 : george 1116 919 : On the x86 the typical linkage code that is generated is: 920 : ---------------------- 921 : .align 2 922 : L0: 923 : addlL1-L0, 72(%esp) 924 : george 1124 jmp L1 925 : george 1116 926 : 927 : .align 2 928 : L1: 929 : ---------------------- 930 : 931 : 72(%esp) is the memory location for the stdlink register. This 932 : must contain the address of the CPS function being called. In the 933 : george 1124 above example, it contains the address of L0; before 934 : george 1116 calling L1 (the real entry point for the compilation unit), it 935 : must contain the address for L1, and hence 936 : 937 : addl $L1-L0, 72(%esp) 938 : 939 : george 1124 I have tested this on all architectures except the hppa.The increase 940 : in code size is of course negligible 941 : george 1116 942 : ---------------------------------------------------------------------- 943 : leunga 1115 Name: Allen Leung 944 : Date: 2002/03/03 13:20:00 EST 945 : Tag: leunga-20020303-mlrisc-tools 946 : 947 : Added #[ ... ] expressions to mlrisc tools 948 : 949 : ---------------------------------------------------------------------- 950 : blume 1096 Name: Matthias Blume 951 : blume 1098 Date: 2002/02/27 12:29:00 EST 952 : Tag: blume-20020227-cdebug 953 : Description: 954 : 955 : - made types in structure C and C_Debug to be equal 956 : - got rid of code duplication (c-int.sml vs. c-int-debug.sml) 957 : - there no longer is a C_Int_Debug (C_Debug is directly derived from C) 958 : 959 : ---------------------------------------------------------------------- 960 : Name: Matthias Blume 961 : blume 1096 Date: 2002/02/26 12:00:00 EST 962 : Tag: blume-20020226-ffi 963 : Description: 964 : 965 : 1. Fixed a minor bug in CM's "noweb" tool: 966 : If numbering is turned off, then truly don't number (i.e., do not 967 : supply the -L option to noweb). The previous behavior was to supply 968 : -L'' -- which caused noweb to use the "default" line numbering scheme. 969 : Thanks to Chris Richards for pointing this out (and supplying the fix). 970 : 971 : 2. Once again, I reworked some aspects of the FFI: 972 : 973 : A. The incomplete/complete type business: 974 : 975 : - Signatures POINTER_TO_INCOMPLETE_TYPE and accompanying functors are 976 : gone! 977 : - ML types representing an incomplete type are now *equal* to 978 : ML types representing their corresponding complete types (just like 979 : in C). This is still safe because ml-nlffigen will not generate 980 : RTTI for incomplete types, nor will it generate functions that 981 : require access to such RTTI. But when ML code generated from both 982 : incomplete and complete versions of the C type meet, the ML types 983 : are trivially interoperable. 984 : 985 : NOTE: These changes restore the full generality of the translation 986 : (which was previously lost when I eliminated functorization)! 987 : 988 : B. Enum types: 989 : 990 : - Structure C now has a type constructor "enum" that is similar to 991 : how the "su" constructor works. However, "enum" is not a phantom 992 : type because each "T enum" has values (and is isomorphic to 993 : MLRep.Signed.int). 994 : - There are generic access operations for enum objects (using 995 : MLRep.Signed.int). 996 : - ml-nlffigen will generate a structure E_foo for each "enum foo". 997 : * The structure contains the definition of type "mlrep" (the ML-side 998 : representation type of the enum). Normally, mlrep is the same 999 : as "MLRep.Signed.int", but if ml-nlffigen was invoked with "-ec", 1000 : then mlrep will be defined as a datatype -- thus facilitating 1001 : pattern matching on mlrep values. 1002 : ("-ec" will be suppressed if there are duplicate values in an 1003 : enumeration.) 1004 : * Constructors ("-ec") or values (no "-ec") e_xxx of type mlrep 1005 : will be generated for each C enum constant xxx. 1006 : * Conversion functions m2i and i2m convert between mlrep and 1007 : MLRep.Signed.int. (Without "-ec", these functions are identities.) 1008 : * Coversion functions c and ml convert between mlrep and "tag enum". 1009 : * Access functions (get/set) fetch and store mlrep values. 1010 : - By default (unless ml-nlffigen was invoked with "-nocollect"), unnamed 1011 : enumerations are merged into one single enumeration represented by 1012 : structure E_'. 1013 : 1014 : ---------------------------------------------------------------------- 1015 : leunga 1080 Name: Allen Leung 1016 : leunga 1094 Date: 2002/02/25 04:45:00 EST 1017 : Tag: leunga-20020225-cps-spill 1018 : 1019 : This is a new implementation of the CPS spill phase. 1020 : The new phase is in the new file compiler/CodeGen/cpscompile/spill-new.sml 1021 : In case of problems, replace it with the old file spill.sml 1022 : 1023 : The current compiler runs into some serious performance problems when 1024 : constructing a large record. This can happen when we try to compile a 1025 : structure with many items. Even a very simple structure like the following 1026 : makes the compiler slow down. 1027 : 1028 : structure Foo = struct 1029 : val x_1 = 0w1 : Word32.int 1030 : val x_2 = 0w2 : Word32.int 1031 : val x_3 = 0w3 : Word32.int 1032 : ... 1033 : val x_N = 0wN : Word32.int 1034 : end 1035 : 1036 : The following table shows the compile time, from N=1000 to N=4000, 1037 : with the old compiler: 1038 : 1039 : N 1040 : 1000 CPS 100 spill 0.04u 0.00s 0.00g 1041 : MLRISC ra 0.06u 0.00s 0.05g 1042 : (spills = 0 reloads = 0) 1043 : TOTAL 0.63u 0.07s 0.21g 1044 : 1045 : 1100 CPS 100 spill 8.25u 0.32s 0.64g 1046 : MLRISC ra 5.68u 0.59s 3.93g 1047 : (spills = 0 reloads = 0) 1048 : TOTAL 14.71u 0.99s 4.81g 1049 : 1050 : 1500 CPS 100 spill 58.55u 2.34s 1.74g 1051 : MLRISC ra 5.54u 0.65s 3.91g 1052 : (spills = 543 reloads = 1082) 1053 : TOTAL 65.40u 3.13s 6.00g 1054 : 1055 : 2000 CPS 100 spill 126.69u 4.84s 3.08g 1056 : MLRISC ra 0.80u 0.10s 0.55g 1057 : (spills = 42 reloads = 84) 1058 : TOTAL 129.42u 5.10s 4.13g 1059 : 1060 : 3000 CPS 100 spill 675.59u 19.03s 11.64g 1061 : MLRISC ra 2.69u 0.27s 1.38g 1062 : (spills = 62 reloads = 124) 1063 : TOTAL 682.48u 19.61s 13.99g 1064 : 1065 : 4000 CPS 100 spill 2362.82u 56.28s 43.60g 1066 : MLRISC ra 4.96u 0.27s 2.72g 1067 : (spills = 85 reloads = 170) 1068 : TOTAL 2375.26u 57.21s 48.00g 1069 : 1070 : As you can see the old cps spill module suffers from some serious 1071 : leunga 1115 performance problem. But since I cannot decipher the old code fully, 1072 : instead of patching the problems up, I'm reimplementing it 1073 : leunga 1094 with a different algorithm. The new code is more modular, 1074 : smaller when compiled, and substantially faster 1075 : (O(n log n) time and O(n) space). Timing of the new spill module: 1076 : 1077 : 4000 CPS 100 spill 0.02u 0.00s 0.00g 1078 : MLRISC ra 0.25u 0.02s 0.15g 1079 : (spills=1 reloads=3) 1080 : TOTAL 7.74u 0.34s 1.62g 1081 : 1082 : Implementation details: 1083 : 1084 : As far as I can tell, the purpose of the CPS spill module is to make sure the 1085 : number of live variables at any program point (the bandwidth) 1086 : does not exceed a certain limit, which is determined by the 1087 : size of the spill area. 1088 : 1089 : When the bandwidth is too large, we decrease the register pressure by 1090 : packing live variables into spill records. How we achieve this is 1091 : completely different than what we did in the old code. 1092 : 1093 : First, there is something about the MLRiscGen code generator 1094 : that we should be aware of: 1095 : 1096 : o MLRiscGen performs code motion! 1097 : 1098 : In particular, it will move floating point computations and 1099 : address computations involving only the heap pointer to 1100 : their use sites (if there is only a single use). 1101 : What this means is that if we have a CPS record construction 1102 : statement 1103 : 1104 : RECORD(k,vl,w,e) 1105 : 1106 : we should never count the new record address w as live if w 1107 : has only one use (which is often the case). 1108 : 1109 : We should do something similar to floating point, but the transformation 1110 : there is much more complex, so I won't deal with that. 1111 : 1112 : Secondly, there are now two new cps primops at our disposal: 1113 : 1114 : 1. rawrecord of record_kind option 1115 : This pure operator allocates some uninitialized storage from the heap. 1116 : There are two forms: 1117 : 1118 : rawrecord NONE [INT n] allocates a tagless record of length n 1119 : rawrecord (SOME rk) [INT n] allocates a tagged record of length n 1120 : and initializes the tag. 1121 : 1122 : 2. rawupdate of cty 1123 : rawupdate cty (v,i,x) 1124 : Assigns to x to the ith component of record v. 1125 : The storelist is not updated. 1126 : 1127 : We use these new primops for both spilling and increment record construction. 1128 : 1129 : 1. Spilling. 1130 : 1131 : This is implemented with a linear scan algorithm (but generalized 1132 : to trees). The algorithm will create a single spill record at the 1133 : beginning of the cps function and use rawupdate to spill to it, 1134 : and SELECT or SELp to reload from it. So both spills and reloads 1135 : are fine-grain operations. In contrast, in the old algorithm 1136 : "spills" have to be bundled together in records. 1137 : 1138 : Ideally, we should sink the spill record construction to where 1139 : it is needed. We can even split the spill record into multiple ones 1140 : at the places where they are needed. But CPS is not a good 1141 : representation for global code motion, so I'll keep it simple and 1142 : am not attempting this. 1143 : 1144 : 2. Incremental record construction (aka record splitting). 1145 : 1146 : Long records with many component values which are simulatenously live 1147 : (recall that single use record addresses are not considered to 1148 : be live) are constructed with rawrecord and rawupdate. 1149 : We allocate space on the heap with rawrecord first, then gradually 1150 : fill it in with rawupdate. This is the technique suggested to me 1151 : by Matthias. 1152 : 1153 : Some restrictions on when this is applicable: 1154 : 1. It is not a VECTOR record. The code generator currently does not handle 1155 : this case. VECTOR record uses double indirection like arrays. 1156 : 2. All the record component values are defined in the same "basic block" 1157 : as the record constructor. This is to prevent speculative 1158 : record construction. 1159 : 1160 : ---------------------------------------------------------------------- 1161 : Name: Allen Leung 1162 : leunga 1086 Date: 2002/02/22 01:02:00 EST 1163 : Tag: leunga-20020222-mlrisc-tools 1164 : 1165 : Minor bug fixes in the parser and rewriter 1166 : 1167 : ---------------------------------------------------------------------- 1168 : Name: Allen Leung 1169 : leunga 1085 Date: 2002/02/21 20:20:00 EST 1170 : Tag: leunga-20020221-peephole 1171 : 1172 : Regenerated the peephole files. Some contained typos in the specification 1173 : and some didn't compile because of pretty printing bugs in the old version 1174 : of 'nowhere'. 1175 : 1176 : ---------------------------------------------------------------------- 1177 : Name: Allen Leung 1178 : leunga 1079 Date: 2002/02/19 20:20:00 EST 1179 : Tag: leunga-20020219-mlrisc-tools 1180 : Description: 1181 : 1182 : Minor bug fixes to the mlrisc-tools library: 1183 : 1184 : 1. Fixed up parsing colon suffixed keywords 1185 : 2. Added the ability to shut the error messages up 1186 : 3. Reimplemented the pretty printer and fixed up/improved 1187 : the pretty printing of handle and -> types. 1188 : 4. Fixed up generation of literal symbols in the nowhere tool. 1189 : 5. Added some SML keywords to to sml.sty 1190 : 1191 : ---------------------------------------------------------------------- 1192 : Name: Matthias Blume 1193 : blume 1078 Date: 2002/02/19 16:20:00 EST 1194 : Tag: blume-20020219-cmffi 1195 : Description: 1196 : 1197 : A wild mix of changes, some minor, some major: 1198 : 1199 : * All C FFI-related libraries are now anchored under$c: 1200 : $/c.cm -->$c/c.cm 1201 : $/c-int.cm -->$c/internals/c-int.cm 1202 : $/memory.cm -->$c/memory/memory.cm 1203 : 1204 : * "make" tool (in CM) now treats its argument pathname slightly 1205 : differently: 1206 : 1. If the native expansion is an absolute name, then before invoking 1207 : the "make" command on it, CM will apply OS.Path.mkRelative 1208 : (with relativeTo = OS.FileSys.getDir()) to it. 1209 : 2. The argument will be passed through to subsequent phases of CM 1210 : processing without "going native". In particular, if the argument 1211 : was an anchored path, then "make" will not lose track of that anchor. 1212 : 1213 : * Compiler backends now "know" their respective C calling conventions 1214 : instead of having to be told about it by ml-nlffigen. This relieves 1215 : ml-nlffigen from one of its burdens. 1216 : 1217 : * The X86Backend has been split into X86CCallBackend and X86StdCallBackend. 1218 : 1219 : * Export C_DEBUG and C_Debug from $c/c.cm. 1220 : 1221 : * C type encoding in ml-nlffi-lib has been improved to model the conceptual 1222 : subtyping relationship between incomplete pointers and their complete 1223 : counterparts. For this, ('t, 'c) ptr has been changed to 'o ptr -- 1224 : with the convention of instantiating 'o with ('t, 'c) obj whenever 1225 : the pointer target type is complete. In the incomplete case, 'o 1226 : will be instantiated with some "'c iobj" -- a type obtained by 1227 : using one of the functors PointerToIncompleteType or PointerToCompleteType. 1228 : 1229 : Operations that work on both incomplete and complete pointer types are 1230 : typed as taking an 'o ptr while operations that require the target to 1231 : be known are typed as taking some ('t, 'c) obj ptr. 1232 : 1233 : voidptr is now a bit "more concrete", namely "type voidptr = void ptr'" 1234 : where void is an eqtype without any values. This makes it possible 1235 : to work on voidptr values using functions meant to operate on light 1236 : incomplete pointers. 1237 : 1238 : * As a result of the above, signature POINTER_TO_INCOMPLETE_TYPE has 1239 : been vastly simplified. 1240 : 1241 : ---------------------------------------------------------------------- 1242 : Name: Matthias Blume 1243 : blume 1076 Date: 2002/02/19 10:48:00 EST 1244 : Tag: blume-20020219-pqfix 1245 : Description: 1246 : 1247 : Applied Chris Okasaki's bug fix for priority queues. 1248 : 1249 : ---------------------------------------------------------------------- 1250 : Name: Matthias Blume 1251 : blume 1073 Date: 2002/02/15 17:05:00 1252 : blume 1069 Tag: Release_110_39 1253 : Description: 1254 : 1255 : blume 1073 Last-minute retagging is becoming a tradition... :-( 1256 : 1257 : This is the working release 110.39. 1258 : 1259 : ---------------------------------------------------------------------- 1260 : Name: Matthias Blume 1261 : Date: 2002/02/15 16:00:00 EST 1262 : Tag: Release_110_39-orig 1263 : Description: 1264 : 1265 : blume 1069 Working release 110.39. New bootfiles. 1266 : 1267 : blume 1073 (Update: There was a small bug in the installer so it wouldn't work 1268 : with all shells. So I retagged. -Matthias) 1269 : 1270 : blume 1069 ---------------------------------------------------------------------- 1271 : Name: Matthias Blume 1272 : blume 1068 Date: 2002/02/15 14:17:00 EST 1273 : Tag: blume-20020215-showbindings 1274 : Description: 1275 : 1276 : Added EnvRef.listBoundSymbols and CM.State.showBindings. Especially 1277 : the latter can be useful for exploring what bindings are available at 1278 : the interactive prompt. (The first function returns only the list 1279 : of symbols that are really bound, the second prints those but also the 1280 : ones that CM's autoloading mechanism knows about.) 1281 : 1282 : ---------------------------------------------------------------------- 1283 : Name: Matthias Blume 1284 : blume 1067 Date: 2002/02/15 12:08:00 EST 1285 : Tag: blume-20020215-iptrs 1286 : Description: 1287 : 1288 : Two improvements to ml-nlffigen: 1289 : 1290 : 1. Write files only if they do not exist or if their current contents 1291 : do not coincide with what's being written. (That is, avoid messing 1292 : with the time stamps unless absolutely necessary.) 1293 : 1294 : 2. Implement a "repository" mechanism for generated files related 1295 : to "incomplete pointer types". See the README file for details. 1296 : 1297 : ---------------------------------------------------------------------- 1298 : Name: Matthias Blume 1299 : blume 1066 Date: 2002/02/14 11:50:00 EST 1300 : Tag: blume-20020214-quote 1301 : Description: 1302 : 1303 : Added a type 't t_' to tag.sml (in ml-nlffi-lib.cm). This is required 1304 : because of the new and improved tag generation scheme. (Thanks to Allen 1305 : Leung for pointing it out.) 1306 : 1307 : ---------------------------------------------------------------------- 1308 : george 1065 Name: Lal George 1309 : Date: 2002/02/14 09:55:27 EST 2002 1310 : Tag: george-20020214-isabelle-bug 1311 : Description: 1312 : 1313 : Fixed the MLRISC bug sent by Markus Wenzel regarding the compilation 1314 : of Isabelle on the x86. 1315 : 1316 : From Allen: 1317 : ----------- 1318 : I've found the problem: 1319 : 1320 : in ra-core.sml, I use the counter "blocked" to keep track of the 1321 : true number of elements in the freeze queue. When the counter goes 1322 : to zero, I skip examining the queue. But I've messed up the 1323 : bookkeeping in combine(): 1324 : 1325 : else (); 1326 : case !ucol of 1327 : PSEUDO => (if !cntv > 0 then 1328 : (if !cntu > 0 then blocked := !blocked - 1 else (); 1329 : ^^^^^^^^^^^^^^^^^^^^^^^ 1330 : moveu := mergeMoveList(!movev, !moveu) 1331 : ) 1332 : else (); 1333 : 1334 : combine() is called to coalesce two nodes u and v. 1335 : I think I was thinking that if the move counts of u and v are both 1336 : greater than zero then after they are coalesced then one node is 1337 : removed from the freeze queue. Apparently I was thinking that 1338 : both u and v are of low degree, but that's clearly not necessarily true. 1339 : 1340 : 1341 : 02/12/2002: 1342 : Here's the patch. HOL now compiles. 1343 : 1344 : I don't know how this impact on performance (compile 1345 : time or runtime). This bug caused the RA (especially on the x86) 1346 : to go thru the potential spill phase when there are still nodes on the 1347 : freeze queue. 1348 : 1349 : 1350 : 1351 : 1352 : ---------------------------------------------------------------------- 1353 : blume 1058 Name: Matthias Blume 1354 : blume 1064 Date: 2002/02/13 22:40:00 EST 1355 : Tag: blume-20020213-fptr-rtti 1356 : Description: 1357 : 1358 : Fixed a bug in ml-nlffigen that was introduced with one of the previous 1359 : updates. 1360 : 1361 : ---------------------------------------------------------------------- 1362 : Name: Matthias Blume 1363 : blume 1063 Date: 2002/02/13 16:41:00 EST 1364 : Tag: blume-20020213-cmlpq 1365 : Description: 1366 : 1367 : Added new priority queue export symbols (which have just been added to 1368 : smlnj-lib.cm) to CML's version of smlnj-lib.cm. (Otherwise CML would 1369 : not compile and the installer would choke.) 1370 : 1371 : ---------------------------------------------------------------------- 1372 : Name: Matthias Blume 1373 : blume 1062 Date: 2002/02/13 16:15:00 EST 1374 : Tag: blume-20020213-various 1375 : Description: 1376 : 1377 : 1. More tweaks to ml-nlffigen: 1378 : 1379 : - better internal datastructures (resulting in slight speedup) 1380 : - "-match" option requires exact match 1381 : - "localized" gensym counters (untagged structs/unions nested within 1382 : other structs/unions or within typedefs get a fresh counter; their 1383 : tag will be prefixed by a concatenation of their parents' tags) 1384 : - bug fixes (related to calculation of transitive closure of types 1385 : to be included in the output) 1386 : 1387 : 2. Minor Basis updates: 1388 : 1389 : - added implementations for List.collate and Option.app 1390 : 1391 : ---------------------------------------------------------------------- 1392 : Name: Matthias Blume 1393 : blume 1060 Date: 2002/02/11 15:55:00 EST 1394 : Tag: blume-20020211-gensym 1395 : Description: 1396 : 1397 : Added a "-gensym" option to command line of ml-nlffigen. This can be 1398 : used to specify a "stem" -- a string that is inserted in all "gensym'd" 1399 : names (ML structure names that correspond to unnamed C structs, unions, 1400 : and enums), so that separate runs of ml-nlffigen do not clash. 1401 : 1402 : ---------------------------------------------------------------------- 1403 : Name: Matthias Blume 1404 : blume 1059 Date: 2002/02/11 12:05:00 EST 1405 : Tag: blume-20020211-gensml 1406 : Description: 1407 : 1408 : A quick fix for a problem with GenSML (in the pgraph-util library): 1409 : Make generation of toplevel "local" optional. (Strictly speaking, 1410 : signature definitions within "local" are not legal SML.) 1411 : 1412 : Other than that: updates to INSTALL and cm/TODO. 1413 : 1414 : ---------------------------------------------------------------------- 1415 : Name: Matthias Blume 1416 : blume 1058 Date: 2002/02/08 15:00:00 EST 1417 : Tag: blume-20020208-uniquepid 1418 : Description: 1419 : 1420 : 0. Version number has been bumped to 110.38.1. NEW BOOTFILES!!! 1421 : 1422 : 1. The installer (config/install.sh) has gotten smarter: 1423 : 1424 : - Configuration options are a bit easier to specify now 1425 : (in config/targets). 1426 : - Bug in recognizing .tar.bz2 files fixed. 1427 : - Installer automatically resolves dependencies between 1428 : configuration options (e.g., if you ask for eXene, you will 1429 : also get cml -- regardless whether you asked for it or not). 1430 : - Installer can run in "quieter mode" by setting the environment 1431 : variable INSTALL_QUIETLY to "true". "Quieter" does not mean 1432 : "completely silent", though. 1433 : - Build HashCons library as part of smlnj-lib. 1434 : 1435 : 2. A new scheme for assigning persistent identifiers to compilation 1436 : units (and, by extension, to types etc.) has been put into place. 1437 : This fixes a long-standing bug where types and even dynamic values 1438 : can get internally confused, thereby compromising type safety 1439 : (abstraction) and dynamic correctness. See 1440 : 1441 : http://cm.bell-labs.com/cm/cs/who/blume/pid-confusion.tgz 1442 : 1443 : for an example of how things could go wrong until now. 1444 : 1445 : The downside of the new scheme is that pids are not quite as 1446 : persistent as they used to be: CM will generate a fresh pid 1447 : for every compilation unit that it thinks it sees for the first 1448 : time. That means that if you compile starting from a clean, fresh 1449 : source tree at two different times, you end up with different 1450 : binaries. 1451 : 1452 : Cutoff recompilation, however, has not been compromised because 1453 : CM keeps pid information in special caches between runs. 1454 : 1455 : ---------------------------------------------------------------------- 1456 : george 1053 Name: Lal George 1457 : george 1055 Date: 2002/02/07 15:34:13 EST 2002 1458 : Tag: 1459 : Description: 1460 : 1461 : Compilers that generate assembly code may produce global labels 1462 : whose value is resolved at link time. The various peephole optimization 1463 : modules did not take this in account. 1464 : 1465 : TODO. The Labels.addrOf function should really return an option 1466 : type so that clients are forced to deal with this issue, rather 1467 : than an exception being raised. 1468 : 1469 : ---------------------------------------------------------------------- 1470 : Name: Lal George 1471 : george 1053 Date: 2002/02/06 13:55:02 EST 1472 : Tag: george-20020206-ra-breakup 1473 : Description: 1474 : 1475 : 1. A bug fix from Allen. 1476 : 1477 : A typo causes extra fstp %st(0)'s to be generated at compensation 1478 : edges, which might cause stack underflow traps at runtime. This 1479 : occurs in fft where there are extraneous fstps right before the 'into' 1480 : trap instruction (in this case they are harmless since none of the 1481 : integers overflow.) 1482 : 1483 : george 1054 2. Pulled out various utility modules that were embedded in the modules 1484 : george 1053 of the register allocator. I need these modules for other purposes, but 1485 : george 1054 they are not complete enough to put into a library (just yet). 1486 : george 1053 ---------------------------------------------------------------------- 1487 : blume 1049 Name: Matthias Blume 1488 : Date: 2002/01/31 16:05:00 EST 1489 : Tag: blume-20020131-sparc-ccalls 1490 : Description: 1491 : 1492 : 1. C-calls on Sparc needlessly allocated a huge chunk (96 bytes) 1493 : of extra stack space by mistake. Fixed. 1494 : 1495 : 2. Bug in logic of handling of command-line options in ml-nlffigen fixed. 1496 : 1497 : ---------------------------------------------------------------------- 1498 : leunga 1048 Name: Allen Leung 1499 : Date: 2002/01/30 1500 : Tag: leunga-20020130-nowhere-bug-fix 1501 : Description: 1502 : 1503 : MLRISC bug fixes: 1504 : 1. Fixed a bindings computation bug in the 'nowhere' program generator tool. 1505 : 2. MachineInt.fromString was negating its value. 1506 : 1507 : ---------------------------------------------------------------------- 1508 : blume 1039 Name: Matthias Blume 1509 : blume 1047 Date: 2002/01/29 1510 : Tag: blume-20020129-INSTALL 1511 : Description: 1512 : 1513 : - Added somewhat detailed installation instructions (file INSTALL). 1514 : - Fixed curl-detection bug in config/install.sh. 1515 : - It is now possible to select the URL getter using the URLGETTER 1516 : environment variable: 1517 : 1518 : not set / "unknown" --> automatic detection (script tries wget, 1519 : curl, and lynx) 1520 : "wget" / "curl" / "lynx" --> use the specified program (script "knows" 1521 : how to properly invoke them) 1522 : other --> use$URLGETTER directly, it must take 1523 : precisely two command-line arguments 1524 : (source URL and destination file name) 1525 : 1526 : ---------------------------------------------------------------------- 1527 : Name: Matthias Blume 1528 : blume 1044 Date: 2002/01/28 1529 : blume 1045 Tag: blume-20020128-sparc-ccalls 1530 : Description: 1531 : 1532 : - Fixed problem with calculation of "used" registers in sparc-c-calls. 1533 : - Make use of the allocParam argument in sparc-c-calls. 1534 : 1535 : ---------------------------------------------------------------------- 1536 : Name: Matthias Blume 1537 : Date: 2002/01/28 1538 : blume 1044 Tag: blume-20020128-allocParam 1539 : Description: 1540 : 1541 : John Reppy: Changes c-calls API to accept client-callback for 1542 : allocating extra stack space. 1543 : me: Corresponding changes to mlriscGen (using a dummy argument that 1544 : does not change the current behavior). 1545 : 1546 : ---------------------------------------------------------------------- 1547 : Name: Matthias Blume 1548 : blume 1040 Date: 2002/01/28 12:00:00 1549 : Tag: Release_110_38 1550 : Description: 1551 : 1552 : This time for real!!! 1553 : 1554 : ---------------------------------------------------------------------- 1555 : Name: Matthias Blume 1556 : blume 1039 Date: 2002/01/28 10:56:00 EST 1557 : Tag: blume-20020128-retraction 1558 : Description: 1559 : 1560 : 0. Retracted earlier 110.38. (The Release_110_38 tag has been replaced 1561 : with blume-Release_110_38-retracted.) 1562 : 1563 : 1. Fixed a problem with incorrect rounding modes in real64.sml. 1564 : (Thanks to Andrew Mccreight .) 1565 : 1566 : 2. A bug in ml-nlffigen related to the handling of unnamed structs, unions, 1567 : and enums fixed. The naming of corresponding ML identifiers should 1568 : now be consistent again. 1569 : 1570 : ---------------------------------------------------------------------- 1571 : leunga 1038 Name: Allen Leung 1572 : Date: 2002/01/27 1573 : Tag: leunga-20020127-nowhere 1574 : Description: 1575 : 1576 : Added a target called nowhere in the configuration scripts. 1577 : Enabling this will build the MLRISC 'nowhere' tool (for translating 1578 : programs with where-clauses into legal SML code) during installation. 1579 : 1580 : ---------------------------------------------------------------------- 1581 : blume 1036 Name: Matthias Blume 1582 : leunga 1038 Date: 2002/01/25 21:27:00 EST 1583 : blume 1039 Tag: blume-Release_110_38-retracted 1584 : blume 1037 Description: 1585 : 1586 : Call it a (working) release! Version is 110.38. Bootfiles are ready. 1587 : 1588 : README will be added later. 1589 : 1590 : blume 1039 !!! NOTE: Re-tagged as blume-Release_110_38-retracted. Original tag 1591 : (Release_110_38) removed. Reason: Last-minute bug fixes. 1592 : 1593 : blume 1037 ---------------------------------------------------------------------- 1594 : Name: Matthias Blume 1595 : blume 1036 Date: 2002/01/25 1596 : Tag: blume-20020125-ffi 1597 : Description: 1598 : 1599 : A large number of tweaks and improvements to ml-nlffi-lib and 1600 : ml-nlffigen: 1601 : 1602 : - ML represenation types have been streamlined 1603 : - getter and setter functions work with concrete values, not abstract 1604 : ones where possible 1605 : - ml-nlffigen command line more flexible (see README file there) 1606 : - some bugs have been fixed (hopefully) 1607 : 1608 : ---------------------------------------------------------------------- 1609 : george 1034 Name: Lal George 1610 : Date: 2002/01/24 1611 : Tag: george-20020124-risc-ra-interface 1612 : Description: 1613 : 1614 : There is a dramatic simplification in the interface to the 1615 : register allocator for RISC architectures as a result of making 1616 : parallel copy instructions explicit. 1617 : 1618 : ---------------------------------------------------------------------- 1619 : blume 1028 Name: Matthias Blume 1620 : blume 1032 Date: 2002/01/22 1621 : Tag: blume-20020122-x86-ccalls 1622 : Description: 1623 : 1624 : Bug fix for c-calls on x86 (having to do with how char- and 1625 : short-arguments are being handled). 1626 : 1627 : ---------------------------------------------------------------------- 1628 : Name: Matthias Blume 1629 : blume 1031 Date: 2002/01/21 1630 : Tag: blume-20020121-ff 1631 : Description: 1632 : 1633 : Another day of fiddling with the FFI... 1634 : 1635 : 1. Bug fix/workaround: CKIT does not complain about negative array 1636 : dimensions, so ml-nlffigen has to guard itself against this possibility. 1637 : (Otherwise a negative dimension would send it into an infinite loop.) 1638 : 1639 : 2. Some of the abstract types (light objects, light pointers, most "base" 1640 : types) in structure C are now eqtypes. 1641 : 1642 : 3. Added constructors and test functions for NULL function pointers. 1643 : 1644 : ---------------------------------------------------------------------- 1645 : Name: Matthias Blume 1646 : blume 1028 Date: 2002/01/18 1647 : blume 1030 Tag: blume-20020118-ready-for-new-release 1648 : Description: 1649 : 1650 : Made config/srcarchiveurl point to a new place. (Will provide boot 1651 : files shortly.) 1652 : 1653 : Maybe we christen this to be 110.38? 1654 : 1655 : ---------------------------------------------------------------------- 1656 : Name: Matthias Blume 1657 : Date: 2002/01/18 1658 : blume 1029 Tag: blume-20020118-more-ffifiddle 1659 : Description: 1660 : 1661 : Today's FFI fiddling: 1662 : 1663 : - Provided a structure CGetSet with "convenient" versions of C.Get.* and 1664 : C.Set.* that use concrete (MLRep.*) arguments and results instead 1665 : of abstract ones. 1666 : 1667 : - Provided word-style bit operations etc. for "int" representation 1668 : types in MLRep.SBitops where ranges over Char, Int, Short, 1669 : and Long. 1670 : 1671 : ---------------------------------------------------------------------- 1672 : Name: Matthias Blume 1673 : Date: 2002/01/18 1674 : blume 1028 Tag: blume-20020118-use-x86-fp 1675 : Description: 1676 : 1677 : Now that x86-fast-fp seems to be working, I turned it back on again 1678 : by default. (Seems to work fine now, even with the FFI.) 1679 : 1680 : Other than that, I added some documentation about the FFI to 1681 : src/ml-nlffigen/README and updated the FFI test examples in 1682 : src/ml-nlffi-lib/Tests/*. 1683 : 1684 : ---------------------------------------------------------------------- 1685 : leunga 1025 Name: Allen Leung 1686 : Date: 2002/01/17 1687 : Tag: leunga-20020117-x86-fast-fp-call 1688 : Description: 1689 : 1690 : 1. Fixed a problem with handling return fp values when x86's fast fp 1691 : mode is turned on. 1692 : 1693 : 2. Minor pretty printing fix for cellset. Print %st(0) as %st(0) instead 1694 : of %f32. 1695 : 1696 : 3. Added a constructor INT32lit to the ast of MLRISC tools. 1697 : 1698 : ---------------------------------------------------------------------- 1699 : blume 1021 Name: Matthias Blume 1700 : Date: 2002/01/16 1701 : Tag: blume-20020116-ffifiddle 1702 : Description: 1703 : 1704 : More fiddling with the FFI interface: 1705 : 1706 : - Make constness 'c instead of rw wherever possible. This eliminates 1707 : the need for certain explicit coercions. (However, due to ML's 1708 : value polymorphism, there will still be many cases where explicit 1709 : coercions are necessary. Phantom types are not the whole answer 1710 : to modeling a subtyping relationship in ML.) 1711 : 1712 : - ro/rw coersions for pointers added. (Avoids the detour through */&.) 1713 : 1714 : - "printf" test example added to src/ml-nlffi-lib/Tests. (Demonstrates 1715 : clumsy workaround for varargs problem.) 1716 : 1717 : ---------------------------------------------------------------------- 1718 : george 1016 Name: Lal George 1719 : Date: 2002/01/15 1720 : Tag: 1721 : Description: 1722 : 1723 : 1. Since COPY instructions are no longer native to the architecture, 1724 : a generic functor can be used to implement the expandCopies function. 1725 : 1726 : 2. Allowed EXPORT and IMPORT pseudo-op declarations to appear inside a 1727 : TEXT segment. 1728 : 1729 : ---------------------------------------------------------------------- 1730 : blume 1011 Name: Matthias Blume 1731 : blume 1015 Date: 2002/01/15 1732 : Tag: blume-20020115-ffiupdates 1733 : Description: 1734 : 1735 : 1. Fix for bug resulting in single-precision float values being returned 1736 : incorrectly from FFI calls. 1737 : 1738 : 2. Small modifications to C FFI API: 1739 : 1740 : - memory-allocation routines return straight objects (no options) 1741 : and raise an exception in out-of-memory situations 1742 : - unsafe extensions to cast between function pointers and pointers 1743 : from/to ints 1744 : - added structure C_Debug as an alternative to structure C where 1745 : pointer-dereferencing (|*| and |*!) always check for null-pointers 1746 : - added open_lib' to DynLinkage; open_lib' works like open_lib 1747 : but also takes a (possibly empty) list of existing library handles 1748 : that the current library depends on 1749 : 1750 : ---------------------------------------------------------------------- 1751 : Name: Matthias Blume 1752 : blume 1011 Date: 2002/01/10 1753 : Tag: blume-20020110-newffigen 1754 : Description: 1755 : 1756 : 1. Updates to portable graph code. 1757 : 1758 : 2. Major update to ml-nlffigen and ml-nlffi-lib. Things are much 1759 : more scalable now so that even huge interfaces such as the one 1760 : for GTK compile in finite time and space. :-) 1761 : See src/ml-nlffigen/README for details on what's new. 1762 : 1763 : ---------------------------------------------------------------------- 1764 : george 1003 Name: Lal George 1765 : george 1009 Date: 2001/01/09 14:31:35 EST 2002 1766 : Tag: george-20011206-rm-native-copy 1767 : Description: 1768 : 1769 : Removed the native COPY and FCOPY instructions 1770 : from all the architectures and replaced it with the 1771 : explicit COPY instruction from the previous commit. 1772 : 1773 : It is now possible to simplify many of the optimizations 1774 : modules that manipulate copies. This has not been 1775 : done in this change. 1776 : 1777 : ---------------------------------------------------------------------- 1778 : Name: Lal George 1779 : george 1003 Date: 2001/12/06 16:50:13 EST 2001 1780 : Tag: george-20011206-mlrisc-instruction 1781 : Description: 1782 : george 824 1783 : george 1003 Changed the representation of instructions from being fully abstract 1784 : to being partially concrete. That is to say: 1785 : 1786 : from 1787 : type instruction 1788 : 1789 : to 1790 : type instr (* machine instruction *) 1791 : 1792 : datatype instruction = 1793 : LIVE of {regs: C.cellset, spilled: C.cellset} 1794 : | KILL of {regs: C.cellset, spilled: C.cellset} 1795 : | COPYXXX of {k: CB.cellkind, dst: CB.cell list, src: CB.cell list} 1796 : | ANNOTATION of {i: instruction, a: Annotations.annotation} 1797 : | INSTR of instr 1798 : 1799 : This makes the handling of certain special instructions that appear on 1800 : all architectures easier and uniform. 1801 : 1802 : LIVE and KILL say that a list of registers are live or killed at the 1803 : program point where they appear. No spill code is generated when an 1804 : element of the 'regs' field is spilled, but the register is moved to 1805 : the 'spilled' (which is present, more for debugging than anything else). 1806 : 1807 : LIVE replaces the (now deprecated) DEFFREG instruction on the alpha. 1808 : We used to generate: 1809 : 1810 : DEFFREG f1 1811 : f1 := f2 + f3 1812 : trapb 1813 : 1814 : but now generate: 1815 : 1816 : f1 := f2 + f3 1817 : trapb 1818 : LIVE {regs=[f1,f2,f3], spilled=[]} 1819 : 1820 : Furthermore, the DEFFREG (hack) required that all floating point instruction 1821 : use all registers mentioned in the instruction. Therefore f1 := f2 + f3, 1822 : defines f1 and uses [f1,f2,f3]! This hack is no longer required resulting 1823 : in a cleaner alpha implementation. (Hopefully, intel will not get rid of 1824 : this architecture). 1825 : 1826 : COPYXXX is intended to replace the parallel COPY and FCOPY available on 1827 : all the architectures. This will result in further simplification of the 1828 : register allocator that must be aware of them for coalescing purposes, and 1829 : will also simplify certain aspects of the machine description that provides 1830 : callbacks related to parallel copies. 1831 : 1832 : ANNOTATION should be obvious, and now INSTR represents the honest to God 1833 : machine instruction set! 1834 : 1835 : The /instructions/Instr.sml files define certain utility 1836 : functions for making porting easier -- essentially converting upper case 1837 : to lower case. All machine instructions (of type instr) are in upper case, 1838 : and the lower case form generates an MLRISC instruction. For example on 1839 : the alpha we have: 1840 : 1841 : datatype instr = 1842 : LDA of {r:cell, b:cell, d:operand} 1843 : | ... 1844 : 1845 : val lda : {r:cell, b:cell, d:operand} -> instruction 1846 : ... 1847 : 1848 : where lda is just (INSTR o LDA), etc. 1849 : 1850 : leunga 815 ---------------------------------------------------------------------- 1851 : blume 986 Name: Matthias Blume 1852 : blume 993 Date: 2001/11/22 21:40:00 EST 1853 : Tag: Release_110_37 1854 : Description: 1855 : 1856 : Release 110.37. This time for real. 1857 : 1858 : ---------------------------------------------------------------------- 1859 : Name: Matthias Blume 1860 : blume 988 Date: 2001/11/21 16:35:00 EST 1861 : Tag: blume-20011121-foot-in-mouth 1862 : Description: 1863 : 1864 : Removed the "Release_110_37" tag because of a serious bug. 1865 : This will be re-tagged once the bug is fixed. 1866 : 1867 : ---------------------------------------------------------------------- 1868 : Name: Matthias Blume 1869 : blume 987 Date: 2001/11/21 16:14:00 EST 1870 : Tag: blume-20011121-forgottenfile 1871 : Description: 1872 : 1873 : Forgot to add a file. (Just a .tex-file -- part of 1874 : the CM manual source.) 1875 : 1876 : ---------------------------------------------------------------------- 1877 : Name: Matthias Blume 1878 : blume 986 Date: 2001/11/21 16:10:00 EST 1879 : blume 988 Tag: blume-20011121-invalid_110_37 1880 : blume 986 Description: 1881 : 1882 : blume 988 Note: I removed the original tag "Release_110_37" from this commit 1883 : because we found a serious bug in all non-x86 backends. 1884 : - Matthias 1885 : 1886 : blume 986 1. Modifications to the SML/NJ code generator and to the runtime system 1887 : so that code object name strings are directly inserted into code 1888 : objects at code generation time. The only business the runtime system 1889 : has with this is now to read the name strings on occasions. 1890 : (The encoding of the name string has also changed somewhat.) 1891 : 1892 : 2. CM now implements a simple "set calculus" for specifying export lists. 1893 : In particular, it is now possible to refer to the export lists of 1894 : other libraries/groups/sources and form unions as well as differences. 1895 : See the latest CM manual for details. 1896 : 1897 : 3. An separate notion of "proxy" libraries has again be eliminated from 1898 : CM's model. (Proxy libraries are now simply a special case of using 1899 : the export list calculus.) 1900 : 1901 : 4. Some of the existing libraries now take advantage of the new set 1902 : calculus. 1903 : (Notice that not all libraries have been converted because some 1904 : of the existing .cm-files are supposed to be backward compatible 1905 : with 110.0.x.) 1906 : 1907 : 5. Some cleanup in stand-alone programs. (Don't use "exnMessage" -- use 1908 : "General.exnMessage"! The former relies on a certain hook to be 1909 : initialized, and that often does not happen in the stand-alone case.) 1910 : 1911 : ---------------------------------------------------------------------- 1912 : george 984 Name: Lal George 1913 : Date: 2001/11/21 13:56:18 EST 1914 : Tag: george-2001121-pseudo-ops 1915 : Description: 1916 : 1917 : Implemented a complete redesign of MLRISC pseudo-ops. Now there 1918 : ought to never be any question of incompatabilities with 1919 : pseudo-op syntax expected by host assemblers. 1920 : 1921 : For now, only modules supporting GAS syntax are implemented 1922 : but more should follow, such as MASM, and vendor assembler 1923 : syntax, e.g. IBM as, Sun as, etc. 1924 : 1925 : ---------------------------------------------------------------------- 1926 : blume 977 Name: Matthias Blume 1927 : Date: 2001/11/14 11:52:00 EST 1928 : Tag: blume-20011114-srcname 1929 : Description: 1930 : 1931 : 1. Routed the name of the current source file to mlriscgen where it 1932 : should be directly emitted into the code object. (This last part 1933 : is yet to be done.) 1934 : 1935 : 2. Some cleanup of the pgraph code to make it match the proposal that 1936 : I put out the other day. (The proposal notwithstanding, things are 1937 : still in flux here.) 1938 : 1939 : ---------------------------------------------------------------------- 1940 : george 976 Name: Lal George 1941 : Date: 2001/11/14 09:44:04 EST 1942 : Tag: 1943 : Description: 1944 : 1945 : Fix for a backpatching bug reported by Allen. 1946 : 1947 : Because the boundary between short and long span-dependent 1948 : instructions is +/- 128, there are an astounding number of 1949 : span-dependent instructions whose size is over estimated. 1950 : 1951 : Allen came up with the idea of letting the size of span 1952 : dependent instructions be non-monotonic, for a maxIter 1953 : number of times, after which the size must be monotonically 1954 : increasing. 1955 : 1956 : This table shows the number of span-dependent instructions 1957 : whose size was over-estimated as a function of maxIter, for the 1958 : file Parse/parse/ml.grm.sml: 1959 : 1960 : maxIter # of instructions: 1961 : 10 687 1962 : 20 438 1963 : 30 198 1964 : 40 0 1965 : 1966 : In compiling the compiler, there is no significant difference in 1967 : compilation speed between maxIter=10 and maxIter=40. Actually, 1968 : my measurements showed that maxIter=40 was a tad faster than 1969 : maxIter=10! Also 96% of the files in the compiler reach a fix 1970 : point within 13 iterations, so fixing maxIter at 40, while high, 1971 : is okay. 1972 : 1973 : ---------------------------------------------------------------------- 1974 : blume 975 Name: Matthias Blume 1975 : Date: 2001/10/31 15:25:00 EST 1976 : Tag: blume-20011031-pgraph 1977 : Description: 1978 : 1979 : CKIT: 1980 : * Changed the "Function" constructor of type Ast.ctype to carry optional 1981 : argument identifiers. 1982 : * Changed the return type of TypeUtil.getFunction accordingly. 1983 : * Type equality ignores the argument names. 1984 : * TypeUtil.composite tries to preserve argument names but gives up quickly 1985 : if there is a mismatch. 1986 : 1987 : installation script: 1988 : * attempts to use "curl" if available (unless "wget" is available as well) 1989 : 1990 : CM: 1991 : * has an experimental implementation of "portable graphs" which I will 1992 : soon propose as an implementation-independent library format 1993 : * there are also new libraries $/pgraph.cm and$/pgraph-util.cm 1994 : 1995 : NLFFI-LIB: 1996 : * some cleanup (all cosmetic) 1997 : 1998 : NLFFIGEN: 1999 : * temporarily disabled the mechanism that suppresses ML output for 2000 : C definitions whose identifiers start with an underscore character 2001 : * generate val bindings for enum constants 2002 : * user can request that only one style (light or heavy) is being used; 2003 : default is to use both (command-line arguments: -heavy and -light) 2004 : * fixed bug in handling of function types involving incomplete pointers 2005 : * generate ML entry points that take record arguments (i.e., using 2006 : named arguments) for C functions that have a prototype with named 2007 : arguments 2008 : (see changes to CKIT) 2009 : 2010 : ---------------------------------------------------------------------- 2011 : leunga 972 Name: Allen Leung 2012 : Date: 2001/10/27 20:34:00 EDT 2013 : Tag: leunga-20011027-x86-fast-fp-call 2014 : Description: 2015 : 2016 : Fixed the bug described in blume-20010920-slowfp. 2017 : 2018 : The fix involves 2019 : 1. generating FCOPYs in FSTP in ia32-svid 2020 : 2. marking a CALL with the appropriate annotation 2021 : 2022 : ---------------------------------------------------------------------- 2023 : blume 961 Name: Matthias Blume 2024 : Date: 2001/10/16 11:32:00 EDT 2025 : Tag: blume-20011016-netbsd 2026 : Description: 2027 : 2028 : Underscore patch from Chris Richards (fixing problem with compiling 2029 : runtime system under recent NetBSD). 2030 : 2031 : ---------------------------------------------------------------------- 2032 : george 959 Name: Allen Leung 2033 : Date: 2001/10/12 17:18:32 EDT 2001 2034 : Tag: leung-20011012-x86-printflowgraph 2035 : Description: 2036 : 2037 : X86RA now uses a valid (instead of dummy) PrintFlowgraph module. 2038 : 2039 : ---------------------------------------------------------------------- 2040 : george 952 Name: Lal George 2041 : george 958 Date: 2001/10/11 23:51:34 EDT 2042 : Tag: george-20011011-too-many-instrs 2043 : Description: 2044 : 2045 : The representation of a program point never expected to see more 2046 : than 65536 instructions in a basic block! 2047 : 2048 : ---------------------------------------------------------------------- 2049 : Name: Lal George 2050 : george 952 Date: 2001/10/09 09:41:37 EDT 2051 : Tag: george-20011008-mlrisc-labels 2052 : Description: 2053 : 2054 : Changed the machine description files to support printing of 2055 : local and global labels in assembly code, based on host assembler 2056 : conventions. 2057 : 2058 : ---------------------------------------------------------------------- 2059 : blume 916 Name: Matthias Blume 2060 : blume 939 Date: 2001/09/25 15:25:00 EDT 2061 : Tag: blume-20010925-exninfo 2062 : Description: 2063 : 2064 : I provided a non-hook implementation of exnName (at the toplevel) and 2065 : made the "dummy" implementation of exnMessage (at the toplevel) more 2066 : useful: if nothing gets "hooked in", then at least you are going to 2067 : see the exception name and a message indicating why you don't see more. 2068 : 2069 : [For the time being, programs that need exnMessage and want to use 2070 : ml-build should either use General.exnMessage (strongly recommended) or 2071 : refer to structure General at some other point so that CM sees a 2072 : static dependency.] 2073 : 2074 : [Similar remarks go for "print" and "use": If you want to use their 2075 : functionality in stand-alone programs generated by ml-build, then use 2076 : TextIO.output and Backend.Interact.useFile (from $smlnj/compiler.cm).] 2077 : 2078 : ---------------------------------------------------------------------- 2079 : Name: Matthias Blume 2080 : blume 938 Date: 2001/09/20 17:28:00 EDT 2081 : Tag: blume-20010920-slowfp 2082 : Description: 2083 : 2084 : Allen says that x86-fast-fp is not safe yet, so I turned it off again... 2085 : 2086 : ---------------------------------------------------------------------- 2087 : Name: Matthias Blume 2088 : blume 937 Date: 2001/09/20 17:20:00 EDT 2089 : Tag: blume-20010920-canonicalpaths 2090 : Description: 2091 : 2092 : 0. Updated the BOOT file (something that I forgot to do earlier). 2093 : 2094 : 1. Small internal change to CM so that it avoids "/../" in filenames 2095 : as much as possible (but only where it is safe). 2096 : 2097 : 2. Changed config/_run-sml (resulting in a changed bin/.run-sml) so 2098 : that arguments that contain delimiters are passed through correctly. 2099 : This change also means that all "special" arguments of the form 2100 : @SMLxxx... must come first. 2101 : 2102 : 3. Changed install script to put relative anchor names for tool commands 2103 : into pathconfig. 2104 : 2105 : ---------------------------------------------------------------------- 2106 : Name: Matthias Blume 2107 : blume 932 Date: 2001/09/18 15:35:00 EDT 2108 : Tag: blume-20010918-readme11036 2109 : Description: 2110 : 2111 : Added README files. 2112 : 2113 : ---------------------------------------------------------------------- 2114 : Name: Matthias Blume 2115 : blume 929 Date: 2001/09/18 11:45:00 EDT 2116 : Tag: Release_110_36 (retag) 2117 : Description: 2118 : 2119 : Fixed mistake in config/preloads. Retagged as 110.36. 2120 : 2121 : ---------------------------------------------------------------------- 2122 : Name: Matthias Blume 2123 : blume 928 Date: 2001/09/18 09:40:00 EDT 2124 : blume 929 Tag: Release_110_36_orig (tag changed) 2125 : blume 928 Description: 2126 : 2127 : New version (110.36). New bootfiles. 2128 : 2129 : ---------------------------------------------------------------------- 2130 : Name: Matthias Blume 2131 : blume 927 Date: 2001/09/14 16:15:00 EDT 2132 : Tag: blume-20010914-x86fastfp 2133 : Description: 2134 : 2135 : John committed some changes that Allen made, in particular a (hopefully) 2136 : correctly working version of the x86-fp module. 2137 : 2138 : I changed the default setting of the Control.MLRISC.getFlag "x86-fast-fp" 2139 : flag to "true". Everything seems to compile to a fixpoint ok, and 2140 : "mandelbrot" speeds up by about 15%. 2141 : 2142 : ---------------------------------------------------------------------- 2143 : Name: Matthias Blume 2144 : blume 923 Date: 2001/09/13 11:20:00 EDT 2145 : Tag: blume-20010913-minimal 2146 : Description: 2147 : 2148 : 1. Stefan Monnier's patch to fix a miscompilation problem that 2149 : was brought to light by John Reppy's work on Moby. 2150 : 2151 : 2. Implemented a minimal "structure Compiler" that contains just 2152 : "version" and "architecture". The minimal version will be 2153 : available when the full version is not. This is for backward- 2154 : compatibility with code that wants to test Compiler.version. 2155 : 2156 : ---------------------------------------------------------------------- 2157 : Name: Matthias Blume 2158 : blume 920 Date: 2001/08/28 14:03:00 EDT 2159 : Tag: blume-20010828-ml-lex 2160 : Description: 2161 : 2162 : Fix for bug 1581, received from Neophytos Michael. 2163 : 2164 : ---------------------------------------------------------------------- 2165 : Name: Matthias Blume 2166 : blume 918 Date: 2001/08/27 11:20:00 EDT 2167 : Tag: blume-20010827-readme11035 2168 : Description: 2169 : 2170 : Fleshed out the README file for 110.35. 2171 : 2172 : ---------------------------------------------------------------------- 2173 : Name: Matthias Blume 2174 : blume 916 Date: 2001/08/24 17:10:00 EDT 2175 : Tag: Release_110_35 2176 : Description: 2177 : 2178 : New version number (110.35). New bootfiles. 2179 : 2180 : ---------------------------------------------------------------------- 2181 : george 915 Name: Lal George 2182 : george 909 Date: 2001/08/24 13:47:18 EDT 2001 2183 : Tag: george-20010824-MLRISC-graphs 2184 : Description: 2185 : 2186 : removed clusters from MLRISC completely and replaced with graphs. 2187 : 2188 : ---------------------------------------------------------------------- 2189 : Name: Matthias Blume 2190 : blume 905 Date: 2001/08/23 17:50:00 EDT 2191 : Tag: blume-20010823-toplevel 2192 : Description: 2193 : 2194 : - some reorganization of the code that implements various kinds of 2195 : environments in the compiler (static, dynamic, symbolic, combined) 2196 : - re-implemented the EnvRef module so that evalStream works properly 2197 : (if the stream contains references to "use", "CM.make", etc.) 2198 : - cleaned up evalloop.sml and interact.sml (but they need more cleaning) 2199 : 2200 : ---------------------------------------------------------------------- 2201 : Name: Matthias Blume 2202 : blume 904 Date: 2001/08/20 15:50 EDT 2203 : Tag: blume20010820-slipup 2204 : Description: 2205 : 2206 : I forgot to commit a few files. Here they are... 2207 : 2208 : ---------------------------------------------------------------------- 2209 : Name: Matthias Blume 2210 : blume 903 Date: 2001/08/20 15:35:00 EDT 2211 : Tag: blume-20010820-debugprof 2212 : Description: 2213 : 2214 : !!!! NEW BOOTFILES !!!! 2215 : 2216 : This is another round of reorganizing the compiler sources. This 2217 : time the main goal was to factor out all the "instrumentation" 2218 : passes (for profiling and backtracing) into their own library. 2219 : The difficulty was to do it in such a way that it does not depend 2220 : on elaborate.cm but only on elabdata.cm. 2221 : 2222 : Therefore there have been further changes to both elaborate.cm and 2223 : elabdata.cm -- more "generic" things have been moved from the former 2224 : to the latter. As a result, I was forced to split the assignment 2225 : of numbers indicating "primtyc"s into two portions: SML-generic and 2226 : SML/NJ-specific. Since it would have been awkward to maintain, 2227 : I bit the bullet and actually _changed_ the mapping between these 2228 : numbers and primtycs. The bottom line of this is that you need 2229 : a new set of bin- and bootfiles. 2230 : 2231 : I have built new bootfiles for all architectures, so doing a fresh 2232 : checkout and config/install.sh should be all you need. 2233 : 2234 : The newly created library's name is 2235 : 2236 :$smlnj/viscomp/debugprof.cm 2237 : 2238 : and its sources live under 2239 : 2240 : src/compiler/DebugProf 2241 : 2242 : ---------------------------------------------------------------------- 2243 : Name: Matthias Blume 2244 : blume 902 Date: 2001/08/15 17:15:00 EDT 2245 : Tag: blume-20010815-compreorg 2246 : Description: 2247 : 2248 : This is a first cut at reorganizing the CM libraries that make up the 2249 : core of the compiler. The idea is to separate out pieces that could 2250 : be used independently by tools, e.g., the parser, the typechecker, etc. 2251 : 2252 : The current status is a step in this direction, but it is not quite 2253 : satisfactory yet. Expect more changes in the future. 2254 : 2255 : Here is the current (new) organization... 2256 : 2257 : What used to be $smlnj/viscomp/core.cm is now divided into 2258 : six CM libraries: 2259 : 2260 :$smlnj/viscomp/basics.cm 2261 : /parser.cm 2262 : /elabdata.cm 2263 : /elaborate.cm 2264 : /execute.cm 2265 : /core.cm 2266 : 2267 : The CM files for these libraries live under src/system/smlnj/viscomp. 2268 : All these libraries are proxy libraries that contain precisely 2269 : one CM library component. Here are the locations of the components 2270 : (all within the src/compiler tree): 2271 : 2272 : Basics/basics.cm 2273 : Parse/parser.cm 2274 : ElabData/elabdata.cm 2275 : Elaborator/elaborate.cm 2276 : Execution/execute.cm 2277 : core.cm 2278 : 2279 : [This organization is the same that has been used already 2280 : for a while for the architecture-specific parts of the visible 2281 : compiler and for the old version of core.cm.] 2282 : 2283 : As you will notice, many source files have been moved from their 2284 : respective original locations to a new home in one of the above 2285 : subtrees. 2286 : 2287 : The division of labor between the new libraries is the following: 2288 : 2289 : basics.cm: 2290 : - Simple, basic definitions that pertain to many (or all) of 2291 : the other libraries. 2292 : parser.cm: 2293 : - The SML parser, producing output of type Ast.dec. 2294 : - The type family for Ast is also defined and exported here. 2295 : elabdata.cm: 2296 : - The datatypes that describe input and output of the elaborator. 2297 : This includes types, absyn, and static environments. 2298 : elaborator.cm: 2299 : - The SML/NJ type checker and elaborator. 2300 : This maps an Ast.dec (with a given static environment) to 2301 : an Absyn.dec (with a new static environment). 2302 : - This libraries implements certain modules that used to be 2303 : structures as functors (to remove dependencies on FLINT). 2304 : execute.cm: 2305 : - Everything having to do with executing binary code objects. 2306 : - Dynamic environments. 2307 : core.cm: 2308 : - SML/NJ-specific instantiations of the elaborator and MLRISC. 2309 : - Top-level modules. 2310 : - FLINT (this should eventually become its own library) 2311 : 2312 : Notes: 2313 : 2314 : I am not 100% happy with the way I separated the elaborator (and its 2315 : data structures) from FLINT. Two instances of the same problem: 2316 : 2317 : 1. Data structures contain certain fields that carry FLINT-specific 2318 : information. I hacked around this using exn and the property list 2319 : module from smlnj-lib. But the fact that there are middle-end 2320 : specific fields around at all is a bit annoying. 2321 : 2322 : 2. The elaborator calculates certain FLINT-related information. I tried 2323 : to make this as abstract as I could using functorization, but, again, 2324 : the fact that the elaborator has to perform calculations on behalf 2325 : of the middle-end at all is not nice. 2326 : 2327 : 3. Having to used exn and property lists is unfortunate because it 2328 : weakens type checking. The other alternative (parameterizing 2329 : nearly *everything*) is not appealing, though. 2330 : 2331 : I removed the "rebinding =" warning hack because due to the new organization 2332 : it was awkward to maintain it. As a result, the compiler now issues some of 2333 : these warnings when compiling init.cmi during bootstrap compilation. On 2334 : the plus side, you also get a warning when you do, for example: 2335 : val op = = Int32.+ 2336 : which was not the case up to now. 2337 : 2338 : I placed "assign" and "deref" into the _Core structure so that the 2339 : code that deals with the "lazy" keyword can find them there. This 2340 : removes the need for having access to the primitive environment 2341 : during elaboration. 2342 : 2343 : ---------------------------------------------------------------------- 2344 : Name: Matthias Blume 2345 : blume 898 Date: 2001/08/13 2346 : Tag: blume-20010813-closures 2347 : Description: 2348 : 2349 : This fix was sent to us by Zhong Shao. It is supposed to improve the 2350 : performance of certain loops by avoiding needless closure allocation. 2351 : 2352 : ---------------------------------------------------------------------- 2353 : george 890 Name: Lal George 2354 : george 896 Date: 2001/07/31 10:03:23 EDT 2001 2355 : Tag: george-20010731-x86-fmalloc 2356 : Description: Fixed bug in x86 calls 2357 : 2358 : There was a bug where call instructions would mysteriously 2359 : vanish. The call instruction had to be one that returned 2360 : a floating point value. 2361 : 2362 : ---------------------------------------------------------------------- 2363 : Name: Lal George 2364 : george 890 Date: 2001/07/19 16:36:29 EDT 2001 2365 : Tag: george-20010719-simple-cells 2366 : Description: 2367 : 2368 : I have dramatically simplified the interface for CELLS in MLRISC. 2369 : 2370 : In summary, the cells interface is broken up into three parts: 2371 : 2372 : 1. CellsBasis : CELLS_BASIS 2373 : 2374 : CellsBasis is a top level structure and common for all 2375 : architectures. it contains the definitions of basic datatypes 2376 : and utility functions over these types. 2377 : 2378 : 2. functor Cells() : CELLS 2379 : 2380 : Cells generates an interface for CELLS that incorporates the 2381 : specific resources on the target architecture, such as the 2382 : presence of special register classes, their number and size, 2383 : and various useful substructures. 2384 : 2385 : 3. CELLS 2386 : 2387 : e.g. SparcCells: SPARCCELLS 2388 : 2389 : CELLS usually contains additional bindings for special 2390 : registers on the architecture, such as: 2391 : 2392 : val r0 : cell (* register zero *) 2393 : val y : cell (* Y register *) 2394 : val psr : cell (* processor status register *) 2395 : ... 2396 : 2397 : The structure returned by applying the Cells functor is opened 2398 : in this interface. 2399 : 2400 : The main implication of all this is that the datatypes for cells is 2401 : split between CellsBasis and CELLS -- a fairly simple change for user 2402 : code. 2403 : 2404 : In the old scheme the CELLS interface had a definitional binding of 2405 : the form: 2406 : 2407 : signature CELLS = sig 2408 : 2409 : structure CellsBasis = CellsBasis 2410 : 2411 : ... 2412 : 2413 : end 2414 : 2415 : With all the sharing constraints that goes on in MLRISC, this old 2416 : design quickly leads to errors such as: 2417 : 2418 : "structure definition spec inside of sharing ... " 2419 : 2420 : 2421 : and appears to require an unacceptable amount of sharing and where 2422 : constraint hackery. 2423 : 2424 : I think this error message (the interaction of definitional specs and 2425 : sharing) requires more explanation on our web page. 2426 : 2427 : ---------------------------------------------------------------------- 2428 : blume 826 Name: Matthias Blume 2429 : blume 879 Date: 2001/07/19 15:00:00 EDT 2430 : Tag: blume-20010719-libreorg 2431 : Description: 2432 : 2433 : This update puts together a fairly extensive but straightforward change 2434 : to the way the libraries that implement the interactive system are 2435 : organized: 2436 : 2437 : The biggest change is the elimination of structure Compiler. As a 2438 : replacement for this structure, there is now a CM library 2439 : (known as $smlnj/compiler.cm or$smlnj/compiler/current.cm) 2440 : that exports all the substructures of the original structure Compiler 2441 : directly. So instead of saying Compiler.Foo.bar one now simply 2442 : says Foo.bar. (The CM libraries actually export a collection of 2443 : structures that is richer than the collection of substructures of 2444 : structure Compiler.) 2445 : 2446 : To make the transition smooth, there is a separate library called 2447 : $smlnj/compiler/compiler.cm which puts together and exports the 2448 : original structure Compiler (or at least something very close to it). 2449 : 2450 : There are five members of the original structure Compiler 2451 : that are not exported directly but which instead became members 2452 : of a new structure Backend (described by signature BACKEND). These are: 2453 : structure Profile (: PROFILE), structure Compile (: COMPILE), structure 2454 : Interact (: INTERACT), structure Machine (: MACHINE), and val 2455 : architecture (: string). 2456 : 2457 : Structure Compiler.Version has become structure CompilerVersion. 2458 : 2459 : Cross-compilers for alpha32, hppa, ppc, sparc, and x86 are provided 2460 : by$smlnj/compiler/.cm where is alpha32, hppa, ppc, sparc, 2461 : or x86, respectively. 2462 : Each of these exports the same frontend structures that 2463 : $smlnj/compiler.cm exports. But they do not have a structure Backend 2464 : and instead export some structure Backend where is Alpha32, 2465 : Hppa, PPC, Sparc, or X86, respectively. 2466 : 2467 : Library$smlnj/compiler/all.cm exports the union of the exports of 2468 : $smlnj/compiler/.cm 2469 : 2470 : There are no structures Compiler anymore, use 2471 :$smlnj/compiler/.cm instead. 2472 : 2473 : Library host-compiler-0.cm is gone. Instead, the internal library 2474 : that instantiates CM is now called cm0.cm. Selection of the host 2475 : compiler (backend) is no longer done here but. (Responsibility for it 2476 : now lies with $smlnj/compiler/current.cm. This seems to be more 2477 : logical.) 2478 : 2479 : Many individual files have been moved or renamed. Some files have 2480 : been split into multiple files, and some "dead" files have been deleted. 2481 : 2482 : Aside from these changes to library organization, there are also changes 2483 : to the way the code itself is organized: 2484 : 2485 : Structure Binfile has been re-implemented in such a way that it no 2486 : longer needs any knowledge of the compiler. It exclusively deals 2487 : with the details of binfile layout. It no longer invokes the 2488 : compiler (for the purpose of creating new prospective binfile 2489 : content), and it no longer has any knowledge of how to interpret 2490 : pickles. 2491 : 2492 : Structure Compile (: COMPILE) has been stripped down to the bare 2493 : essentials of compilation. It no longer deals with linking/execution. 2494 : The interface has been cleaned up considerably. 2495 : 2496 : Utility routines for dealing with linking and execution have been 2497 : moved into their own substructures. 2498 : 2499 : (The ultimate goal of these changes is to provide a light-weight 2500 : binfile loader/linker (at least for, e.g., stable libraries) that 2501 : does not require CM or the compiler to be present.) 2502 : 2503 : CM documentation has been updated to reflect the changes to library 2504 : organization. 2505 : 2506 : ---------------------------------------------------------------------- 2507 : Name: Matthias Blume 2508 : blume 875 Date: 2001/07/10 17:30:00 EDT 2509 : Tag: Release_110_34 2510 : Description: 2511 : 2512 : blume 876 Minor tweak to 110.34 (re-tagged): 2513 : 2514 : - README.html file added to CVS repository 2515 : - runtime compiles properly under FreeBSD 3.X and 4.X 2516 : 2517 : ---------------------------------------------------------------------- 2518 : Name: Matthias Blume 2519 : Date: 2001/07/10 17:30:00 EDT 2520 : Tag: Release_110_34 2521 : Description: 2522 : 2523 : blume 875 New version number (110.34). New bootfiles. 2524 : 2525 : ---------------------------------------------------------------------- 2526 : Name: Matthias Blume 2527 : blume 874 Date: 2001/07/09 16:00:00 EDT 2528 : Tag: blume-20010709-more-varargs 2529 : Description: 2530 : 2531 : I changed the handling of varargs in ml-nlffigen again: 2532 : The ellipsis ... will now simply be ignored (with an accompanying warning). 2533 : 2534 : The immediate effect is that you can actually call a varargs function 2535 : from ML -- but you can't actually supply any arguments beyond the ones 2536 : specified explicitly. (For example, you can call printf with its format 2537 : string, but you cannot pass additional arguments.) 2538 : 2539 : This behavior is only marginally more useful than the one before, but 2540 : it has the advantage that a function or, more importantly, a function 2541 : type never gets dropped on the floor, thus avoiding follow-up problems with 2542 : other types that refer to the offending one. 2543 : 2544 : ---------------------------------------------------------------------- 2545 : Name: Matthias Blume 2546 : blume 873 Date: 2001/07/09 11:25:00 EDT 2547 : Tag: blume-20010709-varargs 2548 : Description: 2549 : 2550 : 1. ckit-lib.cm now exports structure Error 2551 : 2. ml-nlffigen reports occurences of "..." (i.e., varargs function types) 2552 : with a warning accompanied by a source location. Moreover, it 2553 : merely skips the offending function or type and proceeds with the 2554 : rest of its work.u As a result, one can safely feed C code containing 2555 : "..." to ml-nlffigen. 2556 : 3. There are some internal improvements to CM, providing slightly 2557 : more general string substitutions in the tools subsystem. 2558 : 2559 : ---------------------------------------------------------------------- 2560 : Name: Matthias Blume 2561 : blume 854 Date: 2001/06/27 15:10:00 EDT 2562 : Tag: blume-20010627-concur 2563 : Description: 2564 : 2565 : Fixed a small bug in CM's handling of parallel compilation. 2566 : (You could observe the bug by Control-C-interrupting an ordinary 2567 : CMB.make or CM.stabilize and then attaching some compile servers. 2568 : The result was that all of a sudden the previously interrupted 2569 : compilation would continue on its own. This was because of 2570 : an over-optimization: CM did not bother to clean out certain queues 2571 : when no servers were attached "anyway", resulting in the contents 2572 : of these queues to grab control when new servers did get attached.) 2573 : 2574 : There is also another minor update to the CM manual. 2575 : 2576 : ---------------------------------------------------------------------- 2577 : Name: Matthias Blume 2578 : blume 853 Date: 2001/06/26 16:15:00 EDT 2579 : Tag: blume-20010626-cmdoc 2580 : Description: 2581 : 2582 : Minor typo fixed in CM manual (syntax diagram for libraries). 2583 : 2584 : ---------------------------------------------------------------------- 2585 : Name: Matthias Blume 2586 : blume 852 Date: 2001/06/25 22:55:00 EDT 2587 : Tag: blume-20010625-x86pc 2588 : Description: 2589 : 2590 : Fixed a nasty bug in the X86 assembly code that caused signal 2591 : handlers to fail (crash) randomly. 2592 : 2593 : ---------------------------------------------------------------------- 2594 : Name: Matthias Blume 2595 : blume 846 Date: 2001/06/25 12:05:00 EDT 2596 : Tag: blume-20010625-nlffigen 2597 : Description: 2598 : 2599 : This update fixes a number of minor bugs in ml-nlffigen as reported by 2600 : Nick Carter . 2601 : 2602 : 1. Silly but ok typedefs of the form "typedef void myvoid;" are now accepted. 2603 : 2. Default names for generated files are now derived from the name of 2604 : the C file *without its directory*. In particular, this causes generated 2605 : files to be placed locally even if the C file is in some system directory. 2606 : 3. Default names for generated signatures and structures are also derived 2607 : from the C file name without its directory. This avoids silly things 2608 : like "structure GL/GL". 2609 : (Other silly names are still possible because ml-nlffigen does not do 2610 : blume 847 a thorough check of whether generated names are legal ML identifiers. 2611 : When in doubt, use command line arguments to force particular names.) 2612 : blume 846 2613 : ---------------------------------------------------------------------- 2614 : Name: Matthias Blume 2615 : blume 845 Date: 2001/06/21 12:25:00 EDT 2616 : Tag: blume-20010621-eXene 2617 : Description: 2618 : 2619 : eXene now compiles and (sort of) works again. 2620 : 2621 : The library name (for version > 110.33) is$/eXene.cm. 2622 : 2623 : I also added an new example in src/eXene/examples/nbody. See the 2624 : README file there for details. 2625 : 2626 : ---------------------------------------------------------------------- 2627 : Name: Matthias Blume 2628 : blume 844 Date: 2001/06/20 16:40:00 EDT 2629 : Tag: blume-20010620-cml 2630 : Description: 2631 : 2632 : CML now compiles and works again. 2633 : 2634 : Libraries (for version > 110.33): 2635 : 2636 : $cml/cml.cm Main CML library. 2637 :$cml/basis.cm CML's version of $/basis.cm. 2638 :$cml/cml-internal.cm Internal helper library. 2639 : $cml/core-cml.cm Internal helper library. 2640 :$cml-lib/trace-cml.cm Tracing facility. 2641 : $cml-lib/smlnj-lib.cm CML's version of$/smlnj-lib.cm 2642 : 2643 : The installer (config/install.sh) has been taught how to properly 2644 : install this stuff. 2645 : 2646 : ---------------------------------------------------------------------- 2647 : Name: Matthias Blume 2648 : blume 843 Date: 2001/06/19 17:55:00 EDT 2649 : Tag: blume-20010619-instantiate 2650 : Description: 2651 : 2652 : This un-breaks the fix for bug 1432. 2653 : (The bug was originally fixed in 110.9 but I broke it again some 2654 : time after that.) 2655 : 2656 : ---------------------------------------------------------------------- 2657 : Name: Matthias Blume 2658 : blume 842 Date: 2001/06/19 17:25:00 EDT 2659 : Tag: blume-20010619-signals 2660 : Description: 2661 : 2662 : This should (hopefully) fix the long-standing signal handling bug. 2663 : (The runtime system was constructing a continuation record with an 2664 : incorrect descriptor which would cause the GC to drop data on the floor...) 2665 : 2666 : ---------------------------------------------------------------------- 2667 : Name: Matthias Blume 2668 : blume 840 Date: 2001/06/15 15:05:00 EDT 2669 : blume 841 Tag: blume-20010615-moresparc 2670 : Description: 2671 : 2672 : Here is a short late-hour update related to Sparc c-calls: 2673 : 2674 : -- made handling of double-word arguments a bit smarter 2675 : 2676 : -- instruction selection phase tries to collapse certain clumsily 2677 : constructed ML-Trees; typical example: 2678 : 2679 : ADD(ty,ADD(_,e,LI d1),LI d2) -> ADD(ty,e,LI(d1+d2)) 2680 : 2681 : This currently has no further impact on SML/NJ since mlriscGen does 2682 : not seem to generate such patterns in the first place, and c-calls 2683 : (which did generate them in the beginning) has meanwhile been fixed 2684 : so as to avoid them as well. 2685 : 2686 : ---------------------------------------------------------------------- 2687 : Name: Matthias Blume 2688 : Date: 2001/06/15 15:05:00 EDT 2689 : blume 840 Tag: blume-20010615-sparc 2690 : Description: 2691 : 2692 : The purpose of this update is to provide an implementation of NLFFI 2693 : on Sparc machines. 2694 : 2695 : Here are the changes in detail: 2696 : 2697 : * src/MLRISC/sparc/c-calls/sparc-c-calls.sml is a new file containing 2698 : the Sparc implementation of the c-calls API. 2699 : * The Sparc backend of SML/NJ has been modified to uniformely use %fp 2700 : for accessing the ML frame. Thus, we have a real frame pointer and 2701 : can freely modify %sp without need for an omit-frame-ptr phase. 2702 : The vfp logic in src/compiler/CodeGen/* has been changed to accomodate 2703 : this case. 2704 : * ml-nlffigen has been taught to produce code for different architectures 2705 : and calling conventions. 2706 : * In a way similar to what was done in the x86 case, the Sparc 2707 : backend uses its own specific extension to mltree. (For example, 2708 : it needs to be able to generate UNIMP instructions which are part 2709 : of the calling convention.) 2710 : * ml-nlffi-lib was reorganized to make it more modular (in particular, 2711 : to make it easier to plug in new machine- and os-dependent parts). 2712 : 2713 : There are some other fairly unrelated bug fixes and cleanups as well: 2714 : 2715 : * I further hacked the .cm files for MLRISC tools (like MDLGen) so 2716 : that they properly share their libraries with existing SML/NJ libraries. 2717 : * I fixed a minor cosmetic bug in CM, supressing certain spurious 2718 : follow-up error messages. 2719 : * Updates to CM/CMB documentation. 2720 : 2721 : TODO items: 2722 : 2723 : * MLRISC should use a different register as its asmTemp on the Sparc. 2724 : (The current %o2 is a really bad choice because it is part of the 2725 : calling conventions, so things might interfere in unexpected ways.) 2726 : 2727 : ---------------------------------------------------------------------- 2728 : Name: Matthias Blume 2729 : blume 839 Date: 2001/06/07 2730 : Tag: blume-20010607-calls 2731 : Description: 2732 : 2733 : A number of internal changes related to C calls and calling conventions: 2734 : 2735 : 1. ML-Tree CALL statements now carry a "pops" field. It indicates the 2736 : number of bytes popped implicitly (by the callee). In most cases 2737 : this field is 0 but on x86/win32 it is some non-zero value. This 2738 : is information provided for the benefit of the "omit-frameptr" pass. 2739 : 2. The CALL instruction on the x86 carries a similar "pops" field. 2740 : The instruction selection phase copies its value from the ML-Tree 2741 : CALL statement. 2742 : 3. On all other architectures, the instruction selection phase checks 2743 : whether "pops=0" and complains if not. 2744 : 4. The c-calls implementation for x86 now accepts two calling conventions: 2745 : "ccall" and "stdcall". When "ccall" is selected, the caller cleans 2746 : up after the call and pops is set to 0. For "stdcall", the caller 2747 : does nothing, leaving the cleanup to the callee; pops is set to 2748 : the number of bytes that were pushed onto the stack. 2749 : 5. The cproto decoder (compiler/Semant/types/cproto.sml) now can 2750 : distinguish between "ccall" and "stdcall". 2751 : 6. The UNIMP instruction has been added to the supported Sparc instruction 2752 : set. (This is needed for implementing the official C calling convention 2753 : on this architecture.) 2754 : 7. I fixed some of the .cm files under src/MLRISC/Tools to make them 2755 : work with the latest CM. 2756 : 2757 : ---------------------------------------------------------------------- 2758 : Name: Matthias Blume 2759 : blume 838 Date: 2001/06/05 15:10:00 EDT 2760 : Tag: blume-20010605-cm-index 2761 : Description: 2762 : 2763 : 0. The "lambdasplit" parameter for class "sml" in CM has been documented. 2764 : 2765 : 1. CM can now generate "index files". These are human-readable files 2766 : that list on a per-.cm-file basis each toplevel symbol defined or 2767 : imported. The location of the index file for 2768 :

/.cm is

/CM/INDEX/.cm. 2769 : To enable index-file generation, set CM.Control.generate_index to true 2770 : or export an environment-symbol: export CM_GENERATE_INDEX=true. 2771 : 2772 : The CM manual has been updated accordingly. 2773 : 2774 : 2. I made some slight modifications to the c-calls API in MLRISC. 2775 : 2776 : a) There is now a callback to support saving/restoring of 2777 : dedicated but caller-save registers around the actual call 2778 : instruction. 2779 : b) One can optionally specify a comment-annotation for the 2780 : call instruction. 2781 : 2782 : 3. SML/NJ (mlriscGen.sml) uses this new API for the rawccall primop. 2783 : (For example, the comment annotation shows the C prototype of 2784 : the function being called.) 2785 : 2786 : ---------------------------------------------------------------------- 2787 : Name: Matthias Blume 2788 : blume 837 Date: 2001/06/01 13:30:00 EDT 2789 : Tag: blume-20010601-nlffi-cleanup 2790 : Description: 2791 : 2792 : This is mostly a cleanup of MLFFI stuff: 2793 : 2794 : - some signature files have been put into a more exposed place 2795 : - the ugly 'f type parameter is gone (simplifies types tremendously!) 2796 : - ml-nlffigen changed accordingly 2797 : - tutorial updated 2798 : 2799 : Other changes: 2800 : 2801 : - author's affiliation in CM manual(s) updated 2802 : - some more recognized keywords added to Allen's sml.sty 2803 : 2804 : ---------------------------------------------------------------------- 2805 : Name: Matthias Blume 2806 : blume 836 Date: 2001/05/25 15:30:00 EDT 2807 : Tag: blume-20010525-iptr 2808 : Description: 2809 : 2810 : - put the official 110.33-README (as it appears on the ftp server) under 2811 : CVS 2812 : - fixed a small bug related to incomplete pointer types in 2813 : ml-nlffigen 2814 : - small cosmetic change to the ml-nlffi-lib's "arr" type constructor 2815 : (it does not need the 'f type parameter) 2816 : 2817 : ---------------------------------------------------------------------- 2818 : Name: Matthias Blume 2819 : blume 834 Date: 2001/05/23 14:30:00 EDT 2820 : Tag: Release_110_33 2821 : Description: 2822 : 2823 : New version number (110.33). New bootfiles. 2824 : 2825 : ---------------------------------------------------------------------- 2826 : Name: Matthias Blume 2827 : blume 833 Date: 2001/05/22 18:06:00 EDT 2828 : Tag: blume-20010522-targets 2829 : Description: 2830 : 2831 : Made install.sh use file config/targets.customized if it exists, falling 2832 : back to config/targets if it doesn't. This way one can have a customized 2833 : version of the targets file without touching the "real thing", thus 2834 : eliminating the constant fear of accidentally checking something bogus 2835 : back into the CVS repository... (File config/targets.customized must 2836 : not be added to the repository!) 2837 : 2838 : ---------------------------------------------------------------------- 2839 : Name: Matthias Blume 2840 : blume 832 Date: 2001/05/22 16:30:00 EDT 2841 : Tag: blume-20010522-minitut 2842 : Description: 2843 : 2844 : 1. Bug fix in ml-nlffigen; now (hopefully) correctly handling 2845 : struct returns. 2846 : 2. Added src/ml-nlffi-lib/Doc/mini-tutorial.txt. This is some very 2847 : incomplete, preliminary documentation for NLFFI. 2848 : 2849 : ---------------------------------------------------------------------- 2850 : Name: Matthias Blume 2851 : blume 829 Date: 2001/05/14 11:30:00 EDT 2852 : Tag: blume-20010514-script 2853 : Description: 2854 : 2855 : Some bugs in install script fixed. 2856 : 2857 : In addition to that I also made a slight change to the NLFFI API: 2858 : Functors generated by ml-nlffigen now take the dynamic library as a 2859 : straight functor argument, not as a suspended one. (The original 2860 : functor code used to force the suspension right away anyway, so there 2861 : was nothing gained by this complication of the interface.) 2862 : 2863 : ---------------------------------------------------------------------- 2864 : Name: Matthias Blume 2865 : blume 828 Date: 2001/05/11 14:35:00 EDT 2866 : Tag: blume-20010511-ml-nlffi 2867 : Description: 2868 : 2869 : I finally took the plunge and added my new FFI code to the main 2870 : repository. For x86-linux it is now ready for prime-time. 2871 : 2872 : There are two new subdirectories of "src": 2873 : 2874 : - ml-nlffi-lib: 2875 : The utility library for programs using the FFI interface. 2876 : Here is the implementation of $/c.cm and its associated low-level 2877 : partners$/c-int.cm and $/memory.cm. 2878 : - ml-nlffigen: 2879 : A stand-alone program for generating ML glue code from C source 2880 : code. 2881 : 2882 : Building ml-nlffigen requires$/ckit-lib.cm. 2883 : 2884 : The config/install.sh script has been updates to do the Right Thing 2885 : (hopefully). 2886 : 2887 : Notice that the source tree for the C-Kit will not be put under "src" 2888 : but directly under the installation root directory. (This is the 2889 : structure that currently exists on the CVS server when you check out 2890 : module "sml".) Fortunately, config/install.sh knows about this oddity. 2891 : 2892 : Bugs: No documentation yet. 2893 : 2894 : ---------------------------------------------------------------------- 2895 : Name: Matthias Blume 2896 : blume 826 Date: 2001/05/09 16:35:00 EDT 2897 : Tag: blume-20010509-cpscontract 2898 : Description: 2899 : 2900 : Fixed a bug in the accounting code in cpsopt/contract.sml. (The 2901 : wrapper/unwrapper elimination did not decrement usage counts and some 2902 : dead variables got overlooked by the dead-up logic.) 2903 : 2904 : ---------------------------------------------------------------------- 2905 : george 824 Name: Lal George 2906 : Date: 2001/05/08 17:26:09 EDT 2907 : Tag: george-20010508-omit-frameptr 2908 : Description: 2909 : 2910 : Changes to implement the omit-frame-pointer optimization to support 2911 : raw C calls. For now, there is only support on the Intel x86, but 2912 : other architectures will follow as more experience is gained with this. 2913 : 2914 : 2915 : ---------------------------------------------------------------------- 2916 : blume 818 Name: Matthias Blume 2917 : blume 819 Date: 2001/05/07 14:40:00 EDT 2918 : Tag: blume-20010507-proxies 2919 : Description: 2920 : 2921 : I made into "proxy libraries" all libraries that qualify for such a 2922 : change. (A qualifying library is a library that has another library or 2923 : groups as its sole member and repeats that member's export list 2924 : verbatim. A proxy library avoids this repetition by omitting its export 2925 : list, effectively inheriting the list that its (only) member exports. 2926 : See the CM manual for more explanation.) 2927 : The main effect is that explicit export lists for these libraries 2928 : do not have to be kepts in sync, making maintenance a bit easier. 2929 : 2930 : I also added copyright notices to many .cm-files. 2931 : 2932 : Last but not least, I made a new set of bootfiles. 2933 : 2934 : ---------------------------------------------------------------------- 2935 : Name: Matthias Blume 2936 : blume 818 Date: 2001/05/04 17:00:00 EDT 2937 : Tag: blume-20010504-cm-lsplit 2938 : Description: 2939 : 2940 : 0. John merged pending changes to $/smlnj-lib.cm 2941 : 2942 : 1. Allen's previous change accidentally backed out of one of Lal's 2943 : earlier changes. I undid this mistake (re-introducing Lal's change). 2944 : 2945 : 2. I used the new topOrder' function from graph-scc.sml (from$/smlnj-lib.cm) 2946 : within the compiler where applicable. There is some code simplification 2947 : because of that. 2948 : 2949 : 3. The "split" phase (in FLINT) is now part of the default list of phases. 2950 : Compiler.Control.LambdaSplitting.* can be used to globally control the 2951 : lambda-splitting (cross-module-inlining) engine. In addition to that, 2952 : it can now also be controlled on a per-source basis: CM has been taught 2953 : a new tool parameter applicable to ML source files. 2954 : 2955 : - To turn lambda-splitting off completely: 2956 : local open Compiler.Control.LambdaSplitting in 2957 : val _ = set Off 2958 : end 2959 : - To make "no lambda-splitting" the global default (but allow per-source 2960 : overriding); this is the initial setting: 2961 : local open Compiler.Control.LambdaSplitting in 2962 : val _ = set (Default NONE) 2963 : end 2964 : - To make "lambda-splitting with aggressiveness a" the global default 2965 : (and allow per-source overriding): 2966 : local open Compiler.Control.LambdaSplitting in 2967 : val _ = set (Default (SOME a)) 2968 : end 2969 : 2970 : - To turn lambda-splitting off for a given ML souce file (say: a.sml) 2971 : write (in the respective .cm-file): 2972 : a.sml (lambdasplitting:off) 2973 : - To turn lambda-splitting for a.sml on with minimal aggressiveness: 2974 : a.sml (lambdasplitting:on) 2975 : - To turn lambda-splitting for a.sml on with aggressiveness (where 2976 : is a decimal non-negative integer): 2977 : a.sml (lambdasplitting:) 2978 : - To turn lambda-splitting for a.sml on with maximal aggressiveness: 2979 : a.sml (lambdasplitting:infinity) 2980 : - To use the global default for a.sml: 2981 : a.sml (lambdasplitting:default) 2982 : or simply 2983 : a.sml 2984 : 2985 : ---------------------------------------------------------------------- 2986 : leunga 815 Name: Allen Leung 2987 : Date: 2001/05/04 01:57:00 EDT 2988 : Tag: leunga-20010504-sync 2989 : Description: 2990 : leunga 788 2991 : leunga 815 MLRISC features. 2992 : 2993 : 1. Fix to CMPXCHG instructions. 2994 : 2. Changed RA interface to allow annotations in callbacks. 2995 : 3. Added a new method to the stream interface to allow annotations updates. 2996 : 2997 : george 749 ---------------------------------------------------------------------- 2998 : blume 812 Name: Matthias Blume 2999 : blume 814 Date: 2001/05/01 11:45:00 EDT 3000 : Tag: blume-20010501-pcedittmp 3001 : Description: 3002 : 3003 : Changed install.sh to use the current working directory instead of 3004 : /usr/tmp for a temporary file (pcedittmp). The previous choice 3005 : of /usr/tmp caused trouble with MacOS X because of file premission 3006 : problems. 3007 : 3008 : ---------------------------------------------------------------------- 3009 : Name: Matthias Blume 3010 : blume 812 Date: 2001/04/20 11:10:00 EDT 3011 : Tag: blume-20010420-inMLflag 3012 : Description: 3013 : 3014 : - added vp_limitPtrMask to vproc-state.h 3015 : (for use by the raw-C-calls mechanism to implement proper interrupt 3016 : handling) 3017 : - made the ML compiler aware of various data-structure offsets so it 3018 : can generate code for accessing the vp_inML flag and vp_limitPtrMask 3019 : - tweaked mlriscGen.sml to have it emit interrupt-handling code for 3020 : raw C-calls 3021 : 3022 : ---------------------------------------------------------------------- 3023 : george 810 Name: Lal George 3024 : Date: 2001/04/20 09:15:28 EDT 3025 : Tag: george-20010420-macosX 3026 : Description: 3027 : 3028 : - Changes to port to Mac OS X; Darwin. 3029 : 3030 : - In the process I found that sqrt was broken on the PPC, because the 3031 : fsqrt instruction is not implemented. 3032 : 3033 : ---------------------------------------------------------------------- 3034 : blume 799 Name: Matthias Blume 3035 : blume 808 Date: 2001/04/18 12:45:00 EDT 3036 : Tag: blume-20010418-ccalls 3037 : Description: 3038 : 3039 : - fixed two off-by-4 errors in the x86-specific c-calls implementation 3040 : (this bug prevented structure arguments containing pointers from being 3041 : passed correctly) 3042 : - changed the raw-C-call code in mlriscGen.sml in such a way that 3043 : structure arguments are represented as a pointer to the beginning 3044 : of the structure (instead of having a series of synthesized arguments, 3045 : one for each structure member) 3046 : 3047 : - made makeml script's verbosity level configurable via environment 3048 : variable (MAKEML_VERBOSITY) 3049 : 3050 : blume 813 - eliminated placeholder implementations for f32l, w16s, i16s, and f32s 3051 : blume 808 in rawmem-x86.sml; we are now using the real thing 3052 : 3053 : ---------------------------------------------------------------------- 3054 : Name: Matthias Blume 3055 : blume 806 Date: 2001/03/22 16:25:00 EST 3056 : Tag: blume-20010322-bootfiles 3057 : Description: 3058 : 3059 : Created a new set of bootfiles (for your automatic installation convenience). 3060 : 3061 : ---------------------------------------------------------------------- 3062 : Name: Matthias Blume 3063 : blume 805 Date: 2001/03/22 15:10:00 EST 3064 : Tag: blume-20010322-rawmem-parcm 3065 : Description: 3066 : 3067 : 1. All "raw memory access" primitives for the new FFI are implemented now 3068 : (at least on the x86). 3069 : 2. Some further cleanup of CM's parallel make mechanism. 3070 : 3071 : ---------------------------------------------------------------------- 3072 : Name: Matthias Blume 3073 : blume 801 Date: 2001/03/19 17:53:00 EST 3074 : Tag: blume-20010319-parallel 3075 : Description: 3076 : 3077 : Parallel make (using compile servers) now works again. 3078 : 3079 : To this end, CM.stabilize and CMB.make have been modified to work in 3080 : two passes when compile servers are attached: 3081 : 1. Compile everything, do not perform stabilization; this pass 3082 : uses compile servers 3083 : 2. Stabilize everything; this pass does not use compile servers 3084 : If there are no compile servers, the two passes are combined into one 3085 : (as before). Splitting the passes increases the inherent parallelism 3086 : in the dependency graph because the entire graph including all 3087 : libraries is available at the same time. This, in turn, improves 3088 : server utilization. The downside is that the master process will 3089 : have to do some extra work after compilation is done (because for 3090 : technical reasons it must re-read all the binfiles during stabilization). 3091 : 3092 : ---------------------------------------------------------------------- 3093 : Name: Matthias Blume 3094 : blume 800 Date: 2001/03/16 12:22:00 EST 3095 : Tag: blume-20010316-bootfiles 3096 : Description: 3097 : 3098 : Created a new set of bootfiles (for your automatic installation convenience). 3099 : 3100 : ---------------------------------------------------------------------- 3101 : Name: Matthias Blume 3102 : blume 799 Date: 2001/03/16 11:00:00 EST 3103 : Tag: blume-20010316-MLTREE-fixup 3104 : Description: 3105 : 3106 : This is a minor fixup for an (untagged) earlier commit by Allen. 3107 : (A file was missing). 3108 : 3109 : ---------------------------------------------------------------------- 3110 : leunga 788 Name: Allen Leung 3111 : leunga 796 Date: Mon Mar 5 18:54:57 EST 2001 3112 : Tag: leunga-20010305-cut-support 3113 : 3114 : 1. New support for alternative control-flow in MLTREE. 3115 : Currently we support 3116 : 3117 : FLOW_TO(CALL ...., [k1,...,kn]) 3118 : 3119 : This is needed for 'cuts to' in C-- and try/handle-like constructs 3120 : in Moby 3121 : 3122 : New assembler flag "asm-show-cutsto" to turn on control-flow debugging. 3123 : 3124 : 2. Register Allocator 3125 : 3126 : Changes in interface [from Fermin, John] 3127 : 3128 : 3. Alpha 8-bit SLL support [Fermin] 3129 : 3130 : 4. All architectures 3131 : 3132 : A new module (ClusterExpandCopies) for expanding parallel copies. 3133 : 3134 : ---------------------------------------------------------------------- 3135 : Name: Allen Leung 3136 : leunga 788 Date: 2001/02/27 23:07:00 EST 3137 : Tag: leunga-20010227-minor-stuff 3138 : 3139 : 1. Alpha bug fix for CMOVNE 3140 : 2. Handle mltree COND(..,FCMP ...,...) 3141 : 3. Bug fix in simplifier 3142 : 3143 : ---------------------------------------------------------------------- 3144 : blume 777 Name: Matthias Blume 3145 : blume 781 Date: 2001/01/30 17:50:00 EST 3146 : Tag: blume-20010130-sync 3147 : Description: 3148 : 3149 : This is just a minor update to sync my devel branch with the main brach. 3150 : The only visible change is the addition of some README files. 3151 : 3152 : ---------------------------------------------------------------------- 3153 : Name: Matthias Blume 3154 : blume 778 Date: 2001/01/12 23:30:00 JST 3155 : Tag: blume-20010112-bootfiles 3156 : Description: 3157 : 3158 : Made a new set of bootfiles that goes with the current state of the 3159 : repository. 3160 : 3161 : ---------------------------------------------------------------------- 3162 : Name: Matthias Blume 3163 : blume 777 Date: 2001/01/12 21:20:00 JST 3164 : Tag: blume-20010112-sync 3165 : Description: 3166 : 3167 : I am just flushing out some minor changes that had accumulated in 3168 : my private branch in order to sync with the main tree. (This is 3169 : mainly because I had CVS trouble when trying to merge _into_ my 3170 : private branch.) 3171 : 3172 : Most people should be completely unaffected by this. 3173 : 3174 : ---------------------------------------------------------------------- 3175 : leunga 775 Name: Allen Leung 3176 : Date: Thu Jan 11 21:03:00 EST 2001 3177 : Tag: leunga-20010111-labexp=mltree 3178 : Description: 3179 : 3180 : 1. Removed the type LabelExp and replace it by MLTree. 3181 : 2. Rewritten mltree-simplify with the pattern matcher tool. 3182 : 3. There were some bugs in alpha code generator which would break 3183 : 64-bit code generation. 3184 : 4. Redo the tools to generate code with the 3185 : 5. The CM files in MLRISC (and in src/system/smlnj/MLRISC) 3186 : are now generated by perl scripts. 3187 : 3188 : ---------------------------------------------------------------------- 3189 : blume 771 Name: Matthias Blume 3190 : blume 774 Date: 2001/01/10 21:55:00 JST 3191 : Tag: blume-20010110-rcc 3192 : Description: 3193 : 3194 : The RCC stuff now seems to work (but only on the x86). 3195 : This required hacking of the c-calls interface (and -implementation) in 3196 : MLRISC. 3197 : 3198 : Normal compiler users should be unaffected. 3199 : 3200 : ---------------------------------------------------------------------- 3201 : Name: Matthias Blume 3202 : blume 773 Date: 2001/01/09 01:20:00 JST 3203 : Tag: blume-20010109-rcc 3204 : Description: 3205 : 3206 : This is a fairly big patch, flushing out a large number of pending 3207 : changes that I made to my development copy over the last couple of days. 3208 : 3209 : Of practical relevance at this moment is a workaround for a pickling 3210 : bug that Allen ran into the other day. The cause of the bug itself is 3211 : still unknown and it might be hard to fix it properly, but the 3212 : workaround has some merits of its own (namely somewhat reducing pickling 3213 : overhead for certain libraries). Therefore, I think this solution should 3214 : be satisfactory at this time. 3215 : 3216 : The rest of the changes (i.e., the vast majority) has to do with my 3217 : ongoing efforts of providing direct support for C function calls from 3218 : ML. At the moment there is a new primop "RAW_CCALL", typing magic 3219 : in types/cproto.sml (invoked from FLINT/trans/translate.sml), a new 3220 : case in the FLINT CPS datatype (RCC), changes to cps/convert.sml to 3221 : translate uses of RAW_CCALL into RCC, and changes to mlriscGen.sml to 3222 : handle RCC. 3223 : 3224 : The last part (the changes to mlriscGen.sml) are still known to be 3225 : wrong on the x86 and not implemented on all other architectures. But 3226 : the infrastructure is in place. I had to change a few functor 3227 : signatures in the backend to be able to route the CCalls interface 3228 : from MLRISC there, and I had to specialize the mltree type (on the 3229 : x86) to include the necessary extensions. (The extensions themselves 3230 : were already there and redy to go in MLRISC/x86). 3231 : 3232 : Everything should be very happy as soon as someone helps me with 3233 : mlriscGen.sml... 3234 : 3235 : In any case, nothing of this should matter to anyone as long as the 3236 : new primop is not being used (which is going to be the case unless you 3237 : find it where I hid it :). The rest of the compiler is completely 3238 : unaffected. 3239 : 3240 : ---------------------------------------------------------------------- 3241 : Name: Matthias Blume 3242 : blume 772 Date: 2001/01/05 00:30:00 JST 3243 : Tag: blume-20010105-primops 3244 : Description: 3245 : 3246 : Added some experimental support for work that I am doing right now. 3247 : These changes mostly concern added primops, but there is also a new 3248 : experimental C library in the runtime system (but currently not enabled 3249 : anywhere except on Linux/X86). 3250 : 3251 : In the course of adding primops (and playing with them), I discovered that 3252 : Zhong's INL_PRIM hack (no type info for certain primops) was, in fact, badly 3253 : broken. (Zhong was very right he labeled this stuff as "major gross hack".) 3254 : To recover, I made type information in INL_PRIM mandatory and changed 3255 : prim.sml as well as built-in.sml accordingly. The InLine structure now 3256 : has complete, correct type information (i.e., no bottom types). 3257 : 3258 : Since all these changes mean that we need new binfiles, I also bumped the 3259 : version number to 110.32.1. 3260 : 3261 : ---------------------------------------------------------------------- 3262 : Name: Matthias Blume 3263 : blume 771 Date: 2000/12/30 22:10:00 JST 3264 : Tag: blume-20001230-various 3265 : Description: 3266 : 3267 : Added proxy libraries for MLRISC and let MLRISC libraries refer 3268 : to each other using path anchors. (See CM manual for explanation.) 3269 : 3270 : Updated CM documentation. 3271 : 3272 : Fixed some bugs in CM. 3273 : 3274 : Implemented "proxy" libraries (= syntactic sugar for CM). 3275 : 3276 : Added "-quiet" option to makeml and changed runtime system accordingly. 3277 : 3278 : Added cleanup handler for exportML to reset timers and compiler stats. 3279 : 3280 : ---------------------------------------------------------------------- 3281 : george 761 Name: Lal George 3282 : Date: 2000/12/22 22:22:58 EST 2000 3283 : Tag: Release_110_32 3284 : Description: 3285 : 3286 : Infinite precision used throughout MLRISC. 3287 : see MLRISC/mltree/machine-int.sig 3288 : 3289 : ---------------------------------------------------------------------- 3290 : blume 756 Name: Matthias Blume 3291 : blume 760 Date: 2000/12/22 23:16:00 JST 3292 : Tag: blume-20001222-warn 3293 : Description: 3294 : blume 759 3295 : blume 760 Corrected wording and formatting of some CM warning message which I 3296 : broke in my previous patch. 3297 : blume 759 3298 : ---------------------------------------------------------------------- 3299 : Name: Matthias Blume 3300 : blume 763 Date: 2000/12/22 21:20:00 JST 3301 : Tag: blume-20001222-anchorenv 3302 : Description: 3303 : 3304 : Fixed CM's handling of anchor environments in connection with CMB.make. 3305 : 3306 : ---------------------------------------------------------------------- 3307 : Name: Matthias Blume 3308 : blume 758 Date: 2000/12/22 13:15:00 JST 3309 : Tag: blume-20001222-cleanup 3310 : Description: 3311 : 3312 : Removed src/cm/ffi which does not (and did not) belong here. 3313 : 3314 : ---------------------------------------------------------------------- 3315 : Name: Matthias Blume 3316 : blume 757 Date: 2000/12/21 23:55:00 JST 3317 : Tag: blume-20001221-exn 3318 : Description: 3319 : 3320 : Probably most important: CM no longer silently swallows all exceptions 3321 : in the compiler. 3322 : Plus: some other minor CM changes. For example, CM now reports some 3323 : sizes for generated binfiles (code, data, envpickle, lambdapickle). 3324 : 3325 : ---------------------------------------------------------------------- 3326 : Name: Matthias Blume 3327 : blume 756 Date: 2000/12/15 00:01:05 JST 3328 : Tag: blume-20001215-dirtool 3329 : Description: 3330 : 3331 : - "dir" tool added. 3332 : - improvements and cleanup to Tools structure 3333 : - documentation updates 3334 : 3335 : ---------------------------------------------------------------------- 3336 : leunga 752 Name: Allen Leung 3337 : leunga 755 Date: Thu Dec 14 03:45:24 EST 2000 3338 : Description: 3339 : Tag: leunga-20001214-int-inf 3340 : Description: 3341 : 3342 : In IntInf, added these standard functions, which are missing from our 3343 : implementation: 3344 : 3345 : andb : int * int -> int 3346 : xorb : int * int -> int 3347 : orb : int * int -> int 3348 : notb : int -> int 3349 : << : int * word -> int 3350 : ~>> : int * word -> int 3351 : 3352 : Not tested, I hope they are correct. 3353 : 3354 : ---------------------------------------------------------------------- 3355 : Name: Allen Leung 3356 : leunga 752 Date: Fri Dec 8 19:23:26 EST 2000 3357 : Description: 3358 : Tag: leunga-20001208-nowhere 3359 : Description: 3360 : 3361 : Slight improvements to the 'nowhere' tool to handle OR-patterns, 3362 : to generate better error messages etc. Plus a brief manual. 3363 : 3364 : ---------------------------------------------------------------------- 3365 : george 749 Name: Lal George 3366 : Date: 2000/12/08 09:54:02 EST 2000 3367 : Tag: Release_110_31 3368 : Description: 3369 : george 705 3370 : george 749 - Version 110.31 3371 : leunga 591 ---------------------------------------------------------------------- 3372 : leunga 744 Name: Allen Leung 3373 : Date: Thu Dec 7 22:01:04 EST 2000 3374 : Tag: leunga-20001207-cell-monster-hack 3375 : Description: 3376 : 3377 : Major MLRISC internal changes. Affect all clients. 3378 : Summary: 3379 : 3380 : 1. Type CELLS.cell = int is now replaced by a datatype. 3381 : As a result, the old regmap is now gone. Almost all interfaces 3382 : in MLRISC change as a consequence. 3383 : 3384 : 2. A new brand version of machine description tool (v3.0) that generates 3385 : modules expecting the new interface. The old version is removed. 3386 : 3387 : 3. The RA interface has been further abstracted into two new functors. 3388 : RISC_RA and X86RA. These functors have much simpler interfaces. 3389 : [See also directory MLRISC/demo.] 3390 : 3391 : 4. Some other new source->source code generation tools are available: 3392 : 3393 : a. MLRISC/Tools/RewriteGen -- generate rewriters from rules. 3394 : b. MLRISC/Tools/WhereGen -- expands conditional pattern matching rules. 3395 : I use this tool to generate the peephole optimizers---with the new 3396 : cell type changes, peephole rules are becoming difficult to write 3397 : without conditional pattern matching. 3398 : 3399 : 5. More Intmap -> IntHashTable change. Previous changes by Matthias didn't 3400 : cover the entire MLRISC source tree so many things broke. 3401 : 3402 : 6. CM files have been moved to the subdirectory MLRISC/cm. 3403 : They are moved because there are a lot of them and they clutter up the 3404 : root dir. 3405 : 3406 : 7. More detailed documentation to come... 3407 : 3408 : NOTE: To rebuild from 110.30 (ftp distribution), you'll have to do 3409 : a makeml -rebuild first. This is because of other other 3410 : changes that Matthias has made (see below). 3411 : 3412 : 3413 : ---------------------------------------------------------------------- 3414 : blume 733 Name: Matthias Blume 3415 : blume 742 Date: 2000/11/30 23:12:00 JST 3416 : Tag: blume-20001130-filereorg 3417 : Description: 3418 : 3419 : Some manual updates and some file reorganizations in CM. 3420 : 3421 : ---------------------------------------------------------------------- 3422 : Name: Matthias Blume 3423 : blume 737 Date: 2000/11/24 17:45:00 JST 3424 : Tag: blume-20001124-link 3425 : Description: 3426 : 3427 : Drastically improved link traversal code for the case that the dynamic 3428 : value was already loaded at bootstrap time. As a result, CM and CMB 3429 : now both load blazingly fast -- even on a very slow machine. Also, 3430 : memory consumption has been further reduced by this. 3431 : 3432 : Warning: The format of the PIDMAP file has changed. THerefore, to 3433 : bootstrap you have to do this: 3434 : 3435 : 1. Run CMB.make 3436 : 2. Make a symbolic link for the boot directory: 3437 : ln -s sml.boot.ARCH-OS xxx 3438 : 3. "Rebuild" the boot directory: 3439 : ./makeml -boot xxx -rebuild sml ; rm xxx 3440 : 4. Boot normally: 3441 : ./makeml 3442 : 3443 : ---------------------------------------------------------------------- 3444 : Name: Matthias Blume 3445 : blume 735 Date: 2000/11/21 21:20:00 JST 3446 : Tag: blume-20001121-tools 3447 : Description: 3448 : 3449 : Continued hacking on autoloading problem -- with success this time. 3450 : Also changed tool-plugin mechanism. See new CM manual. 3451 : 3452 : ---------------------------------------------------------------------- 3453 : Name: Matthias Blume 3454 : blume 734 Date: 2000/11/19 14:30:00 JST 3455 : Tag: blume-20001119-autoload 3456 : Description: 3457 : 3458 : Some hacking to make autoloading faster. Success for CMB, no success 3459 : so far for CM. There is a reduced structure CM' that autoloads faster. 3460 : (This is a temporary, non-documented hack to be eliminated again when 3461 : the general problem is solved.) 3462 : 3463 : ---------------------------------------------------------------------- 3464 : Name: Matthias Blume 3465 : blume 733 Date: 2000/11/17 14:10:00 JST 3466 : Tag: blume-20001117-pickle-lib 3467 : Description: 3468 : 3469 : 1. Eliminated comp-lib.cm 3470 : 2. Made pickle-lib.cm 3471 : 3. Eliminated all uses of intset.sml (from comp-lib.cm) 3472 : 4. Replaced all uses of intmap.{sig,sml} (from comp-lib.cm) with 3473 : equivalent constructs from smlnj-lib.cm (INtHashTable). 3474 : 5. Point 4. also goes for those uses of intmap.* in MLRISC. 3475 : Duplicated intmap modules thrown out. 3476 : 6. Hunted down all duplicated SCC code and replaced it with 3477 : equivalent stuff (GraphSCCFn from smlnj-lib.cm). 3478 : 7. Rewrote Feedback module. 3479 : 8. Moved sortedlist.sml into viscomp-lib.cm. Eventually it 3480 : should be thrown out and equivalent modules from smlnj-lib.cm 3481 : should be used (IntRedBlackSet, IntListSet, ...). 3482 : 3483 : Confirmed that compiler compiles to fixpoint. 3484 : 3485 : ---------------------------------------------------------------------- 3486 : leunga 731 Name: Allen Leung 3487 : Date: 2000/11/10 18:00:00 3488 : Tag: leunga-20001110-new-x86-fp 3489 : 3490 : A new x86 floating point code generator has been added. 3491 : By default this is turned off. To turn this on, do: 3492 : 3493 : CM.autoload "$smlnj/compiler.cm"; 3494 : Compiler.Control.MLRISC.getFlag "x86-fast-fp" := true; 3495 : 3496 : Changes: 3497 : 3498 : 1. Changed FTAN to FPTAN so that the assembly output is correct. 3499 : 2. Changed the extension callback for FTANGENT to generate: 3500 : 3501 : fptan 3502 : fstp %st(0) 3503 : instead of 3504 : fptan 3505 : fstpl ftempmem 3506 : 3507 : 3. Numerous assembly fixes for x86. 3508 : 3509 : 5. Cleaned up the machine code output module x86/x86MC.sml and added 3510 : support for a whole bunch of instructions and addressing modes: 3511 : 3512 : fadd/fsub/fsubr/fmul/fdiv/fdivr %st, %st(n) 3513 : faddp/fsubp/fsubrp/fmulp/fdivp/fdivrp %st, %st(n) 3514 : fadd/fsub/fsubr/fmul/fdiv/fdivr %st(n), %st 3515 : fiadd/fisub/fisubr/fimul/fidiv/fidivr mem 3516 : fxch %st(n) 3517 : fld %st(n) 3518 : fst %st(n) 3519 : fst mem 3520 : fstp %st(n) 3521 : fucom %st(n) 3522 : fucomp %st(n) 3523 : 3524 : All these are now generated when the fast fp mode is turned on. 3525 : 3526 : 6. Removed the dedicated registers %st(0), ..., %st(7) from X86CpsRegs 3527 : 3528 : ---------------------------------------------------------------------- 3529 : blume 729 Name: Matthias Blume 3530 : Date: 2000/11/09 11:20:00 JST 3531 : Tag: blume-20001109-scc 3532 : Description: 3533 : 3534 : Eliminated some code duplication: 3535 : 3536 : 1. Added "where" clause to GraphSCCFn in SML/NJ Library. 3537 : (Otherwise the functor is useless.) 3538 : 2. Used GraphSCCFn where SCCUtilFun was used previously. 3539 : 3. Got rid of SCCUtilFun (in comp-lib.cm). 3540 : 3541 : ---------------------------------------------------------------------- 3542 : george 721 Name: Lal George 3543 : Date: 2000/11/06 09:02:21 EST 2000 3544 : Tag: Release_110_30 3545 : Description: 3546 : 3547 : - Version 110.30 3548 : ---------------------------------------------------------------------- 3549 : blume 715 Name: Matthias Blume 3550 : blume 716 Date: 2000/11/04 14:45:00 3551 : Tag: blume-20001104-mlbuild 3552 : Description: 3553 : 3554 : - Made ml-build faster on startup. 3555 : - Documentation fixes. 3556 : 3557 : ---------------------------------------------------------------------- 3558 : Name: Matthias Blume 3559 : blume 715 Date: 2000/11/02 17:00:00 JST 3560 : Tag: blume-20001102-condcomp 3561 : Description: 3562 : 3563 : - Small tweaks to pickler -- new BOOTFILES! 3564 : - Version bumped to 110.29.2. 3565 : - Added conditional compilation facility to init.cmi (see comment there). 3566 : ---------------------------------------------------------------------- 3567 : leunga 713 Name: Allen Leung 3568 : Date: 2000/10/23 19:31:00 3569 : Tag: leunga-20001023-demo-ra 3570 : 3571 : 1. Minor RA changes that improves spilling on x86 (affects Moby and C-- only) 3572 : 2. Test programs for the graph library updated 3573 : 3. Some new MLRISC demo programs added 3574 : 3575 : ---------------------------------------------------------------------- 3576 : blume 710 Name: Matthias Blume 3577 : blume 711 Date: 2000/08/31 22:15:00 JST 3578 : Tag: blume-20001017-errmsg 3579 : Description: 3580 : 3581 : More error message grief: Where there used to be no messages, there 3582 : now were some that had bogus error regions. Fixed. 3583 : 3584 : ---------------------------------------------------------------------- 3585 : Name: Matthias Blume 3586 : blume 710 Date: 2000/08/31 17:30:00 JST 3587 : Tag: blume-20001017-v110p29p1 3588 : Description: 3589 : 3590 : I made a version 110.29.1 with new bootfiles. 3591 : 3592 : Changes: Modified pickler/unpickler for faster and leaner unpickling. 3593 : CM documentation changes and a small bugfix in CM's error reporting. 3594 : 3595 : ---------------------------------------------------------------------- 3596 : george 705 Name: Lal George 3597 : Date: 2000/09/27 14:42:35 EDT 3598 : Tag: george-20000927-nodestatus 3599 : Description: 3600 : 3601 : Changed the type of the nodestatus, so that: 3602 : 3603 : SPILLED(~1) is now SPILLED 3604 : SPILLED(m) where m>=0 is now MEMREG(m) 3605 : SPILLED(s) where s<~1 is now SPILL_LOC(~s) 3606 : 3607 : ---------------------------------------------------------------------- 3608 : blume 697 Name: Matthias Blume 3609 : blume 703 Date: 2000/09/07 14:45:00 JST 3610 : Tag: blume-20000907-cmerrmsg 3611 : Description: 3612 : 3613 : Small tweak to CM to avoid getting ML syntax error messages twice. 3614 : 3615 : ---------------------------------------------------------------------- 3616 : Name: Matthias Blume 3617 : blume 701 Date: 2000/08/31 18:00:00 JST 3618 : Tag: blume-20000831-cvsbootfiles 3619 : Description: 3620 : 3621 : New URL for boot files (because the 110.29 files on the BL server do 3622 : now work correctly with my updated install scripts for yacc and lex). 3623 : 3624 : ---------------------------------------------------------------------- 3625 : Name: Matthias Blume 3626 : blume 697 Date: 2000/08/08 12:33:00 JST 3627 : Tag: blume-20000808-manual 3628 : Description: 3629 : 3630 : Tiny update to CM manual. 3631 : 3632 : ---------------------------------------------------------------------- 3633 : leunga 695 Name: Allen Leung 3634 : Date: 2000/08/7 19:31:00 3635 : Tag: leunga-20000807-a-whole-bunch-of-stuff 3636 : 3637 : Moby, C--, SSA, x86, machine descriptions etc. Should only affect C-- 3638 : and Mobdy. 3639 : 3640 : 1. x86 3641 : 3642 : a. Fixes to peephole module by John and Dan. 3643 : b. Assembly fix to SETcc by Allen. 3644 : c. Fix to c-call by John. 3645 : d. Fix to spilling by John. (This one deals with the missing FSTPT case) 3646 : e. Instruction selection optimization to SETcc as suggested by John. 3647 : 3648 : For example, 3649 : 3650 : MV(32, x, COND(32, CMP(32, LT, a, b), LI 1, LI 0)) 3651 : 3652 : should generate: 3653 : 3654 : MOVL a, x 3655 : SUBL b, x 3656 : SHRL 31, x 3657 : 3658 : 2. IR stuff 3659 : 3660 : A bunch of new DJ-graph related algorithms added. These 3661 : speed up SSA construction. 3662 : 3663 : 3. SSA + Scheduling 3664 : 3665 : Added code for SSA and scheduling to the repository 3666 : 3667 : ---------------------------------------------------------------------- 3668 : george 694 Name: Lal George 3669 : Date: 2000/07/27 11:53:14 EDT 3670 : 3671 : Tag: lal-20000727-linux-ppc 3672 : Description: 3673 : 3674 : Made changes to support Linux PPC. 3675 : p.s. I have confirmation that the 110.29 boot files work fine. 3676 : 3677 : ---------------------------------------------------------------------- 3678 : blume 684 Name: Matthias Blume 3679 : blume 692 Date: 2000/07/27 17:40:00 JST 3680 : Tag: blume-20000727-scripts 3681 : Description: 3682 : 3683 : !!!! WARNING !!!! 3684 : You must recompile the runtime system! 3685 : !!!! WARNING !!!! 3686 : 3687 : This is basically another round of script-enhancements: 3688 : 3689 : 1. sml, ml-build, and ml-makedepend accept options -D and -U to define 3690 : and undefine CM preprocessor symbols. 3691 : 3692 : 2. ml-build avoids generating a new heap image if it finds that the 3693 : existing one is still ok. (The condition is that no ML file had to 3694 : be recompiled and all ML files are found to be older that the heap 3695 : file.) 3696 : 3697 : To make this work smoothly, I also hacked the runtime system as 3698 : well as SMLofNJ.SysInfo to get access to the heap image suffix 3699 : (.sparc-solaris, ...) that is currently being used. 3700 : 3701 : Moreover, the signature of CM.mk_standalone has changed. See the 3702 : CM manual. 3703 : 3704 : 3. ml-makedepend accepts additional options -n, -a, and -o. (See the 3705 : CM manual for details.) 3706 : 3707 : 4. More CM manual updates: 3708 : - all of the above has been documented. 3709 : - there is now a section describing the (CM-related) command line 3710 : arguments that are accepted by the "sml" command 3711 : 3712 : ---------------------------------------------------------------------- 3713 : Name: Matthias Blume 3714 : blume 691 Date: 2000/07/25 16:20:00 JST 3715 : Tag: blume-20000725-makedepend 3716 : Description: 3717 : 3718 : Added a script called ml-makedepend. This can be used in makefiles 3719 : for Unix' make in a way very similar to the "makedepend" command for 3720 : C. 3721 : 3722 : The script internally uses function CM.sources. 3723 : 3724 : Synopsis: 3725 : 3726 : ml-makedepend [-f makefile] cmfile targetname 3727 : 3728 : The default for the makefile is "makefile" (or "Makefile" should 3729 : "makefile" not exist). 3730 : 3731 : ml-makedepend adds a cmfile/targetname-specific section to this 3732 : makefile (after removing the previous version of this section). The 3733 : section contains a single dependency specification with targetname on 3734 : the LHS (targetname is an arbitrary name), and a list of files derived 3735 : from the cmfile on the RHS. Some of the files on the RHS are 3736 : ARCH/OPSYS-specific. Therefore, ml-makedepend inserts references to 3737 : "make" variables$(ARCH) and $(OPSYS) in place of the corresponding 3738 : path names. The makefile writer is responsible for making sure that 3739 : these variables have correct at the time "make" is invoked. 3740 : 3741 : ---------------------------------------------------------------------- 3742 : Name: Matthias Blume 3743 : blume 690 Date: 2000/07/22 23:30:00 JST 3744 : Tag: blume-20000722-urlupdate 3745 : Description: 3746 : 3747 : Changed BOOT and config/srcarchiveurl to point to BL server: 3748 : 3749 : ftp://ftp.research.bell-labs.com/dist/smlnj/working/110.29/ 3750 : 3751 : ---------------------------------------------------------------------- 3752 : Name: Matthias Blume 3753 : blume 686 Date: 2000/07/18 18:00:00 JST 3754 : Tag: blume-20000718-Version_110_29 3755 : Description: 3756 : 3757 : 1. Updated src/compiler/TopLevel/main/version.sml to version 110.29 3758 : 3759 : 2. Updated config/version to 110.29 3760 : 3761 : 3. Updated config/srcarchiveurl 3762 : 3763 : 3. New boot files! 3764 : ftp://ftp.cs.princeton.edu/pub/people/blume/sml/110.29-autofetch 3765 : 3766 : ---------------------------------------------------------------------- 3767 : Name: Matthias Blume 3768 : blume 684 Date: 2000/07/11 13:58:00 JST 3769 : Tag: blume-20000711-doctypo 3770 : Description: 3771 : 3772 : Fixed a few typos in CM manual. 3773 : 3774 : ---------------------------------------------------------------------- 3775 : leunga 683 Name: Allen Leung 3776 : Date: 2000/06/15 00:38:00 3777 : Tag: leunga-20000704-sparc-x86 3778 : 3779 : 1. x86 peephole improvement sp += k; sp -= k => nop [from John] 3780 : 2. fix to x86 RET bug [found by Dan Grossman] 3781 : 3. sparc assembly bug fix for ticc instructions [found by Fermin] 3782 : 3783 : Affects c-- and moby only 3784 : 3785 : ---------------------------------------------------------------------- 3786 : blume 680 Name: Matthias Blume 3787 : blume 682 Date: 2000/07/04 15:26:00 3788 : Tag: blume-20000704-trigger 3789 : Description: 3790 : 3791 : 1. Improvements to CM manual. 3792 : 2. SMLofNJ.Internals.BTrace.trigger reinstated as an alternative way 3793 : of getting a back-trace. The function, when called, raises an 3794 : internal exception which explicitly carries the full back-trace history, 3795 : so it is unaffected by any intervening handle-raise pairs ("trivial" 3796 : or not). The interactive loop will print that history once it arrives 3797 : at top level. 3798 : Short of having all exceptions implicitly carry the full history, the 3799 : recommended way of using this facility is: 3800 : - compile your program with instrumentation "on" 3801 : - run it, when it raises an exception, look at the history 3802 : - if the history is "cut off" because of some handler, go and modify 3803 : your program so that it explicitly calls BTrace.trigger 3804 : - recompile (still instrumented), and rerun; look at the full history 3805 : 3806 : ---------------------------------------------------------------------- 3807 : Name: Matthias Blume 3808 : Date: 2000/07/03 15:36:00 JST 3809 : blume 680 Tag: blume-20000702-manual 3810 : Description: 3811 : 3812 : Small corrections and updates to CM manual. 3813 : 3814 : ---------------------------------------------------------------------- 3815 : blume 682 Name: Matthias Blume 3816 : blume 679 Date: 2000/06/29 16:04:00 JST 3817 : Tag: blume-20000629-yacctool 3818 : Description: 3819 : 3820 : Changes: 3821 : 3822 : 1. Class "mlyacc" now takes separate arguments to pass options to 3823 : generated .sml- and .sig-files independently. 3824 : 2. Corresponding CM manual updates. 3825 : 3. BTrace module now also reports call sites. (However, for loop clusters 3826 : it only shows from where the cluster was entered.) There are associated 3827 : modifications to core.sml, internals.{sig,sml}, btrace.sml, and btimp.sml. 3828 : 3829 : ---------------------------------------------------------------------- 3830 : blume 675 Name: Matthias Blume 3831 : blume 678 Date: 2000/06/27 16:51:00 JST 3832 : Tag: blume-20000627-noweb 3833 : Description: 3834 : 3835 : Changes: 3836 : 3837 : 1. Implemented "subdir" and "witness" options for noweb tool. 3838 : This caused some slight internal changes in CM's tool implementation. 3839 : 2. Fixed bug in "tool plugin" mechanism. This is essentially cleaning 3840 : some remaining issues from earlier path anchor changes. 3841 : 3. Updated CM manual accordingly. 3842 : 3843 : 4. Changed implementation of back-tracing so that I now consider it 3844 : ready for prime-time. 3845 : 3846 : In particular, you don't have to explicitly trigger the back-trace 3847 : anymore. Instead, if you are running BTrace-instrumented code and 3848 : there is an uncaught exception (regardless of whether or not it was 3849 : raised in instrumented code), the top-level evalloop will print 3850 : the back-trace. 3851 : 3852 : Features: 3853 : 3854 : - Instrumented and uninstrumented code work together seemlessly. 3855 : (Of course, uninstrumented code is never mentioned in actual 3856 : back-traces.) 3857 : 3858 : - Asymptotic time- and space-complexity of instrumented code is 3859 : equal to that of uninstrumented code. (This means that 3860 : tail-recursion is preserved by the instrumentation phase.) 3861 : 3862 : - Modules whose code has been instrumented in different sessions 3863 : work together without problem. 3864 : 3865 : - There is no penalty whatsoever on uninstrumented code. 3866 : 3867 : - There is no penalty on "raise" expressions, even in 3868 : instrumented code. 3869 : 3870 : A potential bug (or perhaps it is a feature, too): 3871 : 3872 : A back-trace reaches no further than the outermost instrumented 3873 : non-trivial "raise". Here, a "trivial" raise is one that is the 3874 : sole RHS of a "handle" rule. Thus, back-traces reach trough 3875 : 3876 : handle e => raise e 3877 : 3878 : and even 3879 : 3880 : handle Foo => raise Bar 3881 : 3882 : and, of course, through 3883 : 3884 : handle Foo => ... 3885 : 3886 : if the exception was not Foo. 3887 : 3888 : Back-traces always reach right through any un-instrumented code 3889 : including any of its "handle" expressions, trivial or not. 3890 : 3891 : To try this out, do the following: 3892 : 3893 : - Erase all existing binfiles for your program. 3894 : (You may keep binfiles for those modules where you think you 3895 : definitely don't need back-tracing.) 3896 : - Turn on back-trace instrumentation: 3897 : SMLofNJ.Internals.BTrace.mode (SOME true); 3898 : - Recompile your program. (I.e., run "CM.make" or "use".) 3899 : - You may now turn instrumentation off again (if you want): 3900 : SMLofNJ.Internals.BTrace.mode (SOME false); 3901 : - Run your program as usual. If it raises an exception that 3902 : reaches the interactive toplevel, then a back-trace will 3903 : automatically be printed. After that, the toplevel loop 3904 : will print the exception history as usual. 3905 : 3906 : ---------------------------------------------------------------------- 3907 : Name: Matthias Blume 3908 : blume 677 Date: 2000/06/26 09:56:46 JST 3909 : Tag: blume-20000626-setup 3910 : Description: 3911 : 3912 : CM: - setup-parameter to "sml" added; this can be used to run arbitrary 3913 : ML code before and after compiling a file (e.g., to set compiler 3914 : flags) 3915 : 3916 : Compiler: - improved btrace API (in core.sml, internals.{sig,sml}) 3917 : - associated changes to btrace.sml (BTrace instrumentation pass) 3918 : - cleaner implementation of btimp.sml (BTrace tracing and report 3919 : module) 3920 : 3921 : CM manual: * new path encoding documented 3922 : * description of setup-parameter to "sml" added 3923 : 3924 : The biggest user-visible change to back-tracing is that it is no 3925 : longer necessary to compile all traced modules within the same 3926 : session. (This was a real limitation.) 3927 : 3928 : ---------------------------------------------------------------------- 3929 : Name: Matthias Blume 3930 : blume 676 Date: 2000/06/24 12:40:00 JST 3931 : Tag: blume-20000624-startup 3932 : Description: 3933 : 3934 : Fixes startup slowdown problem. (I was calling SrcPath.sync a _tad_ 3935 : bit too often -- to put it mildly. :) 3936 : 3937 : ---------------------------------------------------------------------- 3938 : Name: Matthias Blume 3939 : blume 675 Date: 2000/06/23 18:20:00 JST 3940 : Tag: blume-20000623-btrace 3941 : Description: 3942 : 3943 : This updates adds a backtrace facility to aid programmers in debugging 3944 : their programs. This involves the following changes: 3945 : 3946 : 1. Module system/smlnj/init/core.sml (structure _Core) now has hooks for 3947 : keeping track of the current call stack. When programs are compiled 3948 : in a special mode, the compiler will insert calls to these hooks 3949 : into the user program. 3950 : "Hook" means that it is possible for different implementations of 3951 : back-tracing to register themselves (at different times). 3952 : 3953 : 2. compiler/MiscUtil/profile/btrace.sml implements the annotation phase 3954 : as an Absyn.dec->Absyn.dec rewrite. Normally this phase is turned off. 3955 : It can be turned on using this call: 3956 : SMLofNJ.Internals.BTrace.mode (SOME true); 3957 : Turning it off again: 3958 : SMLofNJ.Internals.BTrace.mode (SOME false); 3959 : Querying the current status: 3960 : SMLofNJ.Internals.BTrace.mode NONE; 3961 : Annotated programs are about twice as big as normal ones, and they 3962 : run a factor of 2 to 4 slower with a dummy back-trace plugin (one 3963 : where all hooks do nothing). The slowdown with a plugin that is 3964 : actually useful (such as the one supplied by default) is even greater, 3965 : but in the case of the default plugin it is still only an constant 3966 : factor (amortized). 3967 : 3968 : 3. system/Basis/Implementation/NJ/internals.{sig,sml} have been augmented 3969 : with a sub-structure BTrace for controlling back-tracing. In particular, 3970 : the above-mentioned function "mode" controls whether the annotation 3971 : phase is invoked by the compiler. Another important function is 3972 : "trigger": when called it aborts the current execution and causes 3973 : the top-level loop to print a full back-trace. 3974 : 3975 : 4. compiler/MiscUtil/profile/btimp.sml is the current default plugin 3976 : for back-tracing. It keeps track of the dynamic call stack and in 3977 : addition to that it keeps a partial history at each "level" of that 3978 : stack. For example, if a tail-calls b, b tail-calls c, and c tail-calls 3979 : d and b (at separate times, dynamically), then the report will show: 3980 : 3981 : GOTO d 3982 : /c 3983 : GOTO \b 3984 : CALL a 3985 : 3986 : This shows that there was an initial non-tail call of a, then a 3987 : tail-call to b or c, looping behavior in a cluster of functions that 3988 : consist of b and c, and then a goto from that cluster (i.e., either from 3989 : b or from c) to d. 3990 : 3991 : Note that (depending on the user program) the amount of information 3992 : that the back-trace module has to keep track of at each level is bounded 3993 : by a constant. Thus, the whole implementation has the same asymptotical 3994 : complexity as the original program (both in space and in time). 3995 : 3996 : 5. compiler/TopLevel/interact/evalloop.sml has been modified to 3997 : handle the special exception SMLofNJ.Internals.BTrace.BTrace 3998 : which is raised by the "trigger" function mentioned above. 3999 : 4000 : Notes on usage: 4001 : 4002 : - Annotated code works well together with unannotated code: 4003 : Unannotated calls simply do not show up at all in the backtrace. 4004 : 4005 : - It is not a good idea to let modules that were annotated during 4006 : different sessions run at the same time. This is because the compiler 4007 : chooses small integers to identify individual functions, and there 4008 : will be clashes if different modules were compiled in separate sessions. 4009 : (Nothing will crash, and you will even be told about the clashes, but 4010 : back-trace information will in general not be useful.) 4011 : 4012 : - Back-tracing can be confused by callcc and capture. 4013 : 4014 : - The only way of getting a back-trace right now is to explicitly 4015 : invoke the "trigger" function from your user program. Eventually, we 4016 : should make every exception carry back-trace information (if 4017 : available). But since this creates more overhead at "raise"-time 4018 : (similar to the current exnHistory overhead), I have not yet 4019 : implemented this. (The implementation will be rather easy.) With 4020 : exceptions carrying back-trace information, this facility will be even 4021 : more useful because users don't need to modify their programs... 4022 : 4023 : - While it is possible to compile the compiler with back-trace 4024 : annotations turned on (I did it to get some confidence in 4025 : correctness), you must make absolutely sure that core.sml and 4026 : btimp.sml are compiled WITHOUT annotation! (core.sml cannot actually 4027 : be compiled with annotation because there is no core access yet, but 4028 : if you compile btimp.sml with annotation, then the system will go into 4029 : an infinite recursion and crash.) 4030 : Since CM currently does not know about BTrace, the only way to turn 4031 : annotations on and off for different modules of the compiler is to 4032 : interrupt CMB.make, change the settings, and re-invoke it. Of course, 4033 : this is awkward and clumsy. 4034 : 4035 : Sample sessions: 4036 : 4037 : Standard ML of New Jersey v110.28.1 [FLINT v1.5], June 5, 2000 4038 : - SMLofNJ.Internals.BTrace.mode (SOME true); 4039 : [autoloading] 4040 : [autoloading done] 4041 : val it = false : bool 4042 : - structure X = struct 4043 : - fun main n = let 4044 : - fun a (x, 0) = d x 4045 : - | a (x, n) = b (x, n - 1) 4046 : - and b (x, n) = c (x, n) 4047 : - and c (x, n) = a (x, n) 4048 : - and d x = e (x, 3) 4049 : - and e (x, 0) = f x 4050 : - | e (x, n) = e (x, n - 1) 4051 : - and f 0 = SMLofNJ.Internals.BTrace.trigger () 4052 : - | f n = n * g (n - 1) 4053 : - and g n = a (n, 3) 4054 : - in 4055 : - f n 4056 : - end 4057 : - end; 4058 : structure X : sig val main : int -> int end 4059 : - X.main 3; 4060 : *** BACK-TRACE *** 4061 : GOTO stdIn:4.2-13.20: X.main[2].f 4062 : GOTO-( stdIn:4.2-13.20: X.main[2].e 4063 : GOTO stdIn:4.2-13.20: X.main[2].d 4064 : / stdIn:4.2-13.20: X.main[2].a 4065 : | stdIn:4.2-13.20: X.main[2].b 4066 : GOTO-\ stdIn:4.2-13.20: X.main[2].c 4067 : CALL stdIn:4.2-13.20: X.main[2].g 4068 : GOTO stdIn:4.2-13.20: X.main[2].f 4069 : GOTO-( stdIn:4.2-13.20: X.main[2].e 4070 : GOTO stdIn:4.2-13.20: X.main[2].d 4071 : / stdIn:4.2-13.20: X.main[2].a 4072 : | stdIn:4.2-13.20: X.main[2].b 4073 : GOTO-\ stdIn:4.2-13.20: X.main[2].c 4074 : CALL stdIn:4.2-13.20: X.main[2].g 4075 : GOTO stdIn:4.2-13.20: X.main[2].f 4076 : GOTO-( stdIn:4.2-13.20: X.main[2].e 4077 : GOTO stdIn:4.2-13.20: X.main[2].d 4078 : / stdIn:4.2-13.20: X.main[2].a 4079 : | stdIn:4.2-13.20: X.main[2].b 4080 : GOTO-\ stdIn:4.2-13.20: X.main[2].c 4081 : CALL stdIn:4.2-13.20: X.main[2].g 4082 : GOTO stdIn:4.2-13.20: X.main[2].f 4083 : CALL stdIn:2.15-17.4: X.main[2] 4084 : - 4085 : 4086 : (Note that because of a FLINt bug the above code currently does not 4087 : compile without BTrace turned on.) 4088 : 4089 : Here is another example, using my modified Tiger compiler: 4090 : 4091 : Standard ML of New Jersey v110.28.1 [FLINT v1.5], June 5, 2000 4092 : - SMLofNJ.Internals.BTrace.mode (SOME true); 4093 : [autoloading] 4094 : [autoloading done] 4095 : val it = false : bool 4096 : - CM.make "sources.cm"; 4097 : [autoloading] 4098 : ... 4099 : [autoloading done] 4100 : [scanning sources.cm] 4101 : [parsing (sources.cm):parse.sml] 4102 : [creating directory CM/SKEL ...] 4103 : [parsing (sources.cm):tiger.lex.sml] 4104 : ... 4105 : [wrote CM/sparc-unix/semant.sml] 4106 : [compiling (sources.cm):main.sml] 4107 : [wrote CM/sparc-unix/main.sml] 4108 : [New bindings added.] 4109 : val it = true : bool 4110 : - Main.compile ("../testcases/merge.tig", "foo.out"); 4111 : *** BACK-TRACE *** 4112 : CALL lib/semant.sml:99.2-396.21: SemantFun[2].transExp.trvar 4113 : CALL lib/semant.sml:99.2-396.21: SemantFun[2].transExp.trexp 4114 : CALL lib/semant.sml:289.3-295.22: SemantFun[2].transExp.trexp.check[2] 4115 : GOTO lib/semant.sml:289.3-295.22: SemantFun[2].transExp.trexp.check[2] 4116 : CALL lib/semant.sml:99.2-396.21: SemantFun[2].transExp.trexp 4117 : CALL lib/semant.sml:99.2-396.21: SemantFun[2].transExp.trexp 4118 : CALL lib/semant.sml:488.3-505.6: SemantFun[2].transDec.trdec[2].transBody[2] 4119 : / lib/semant.sml:411.65-543.8: SemantFun[2].transDec 4120 : CALL-\ lib/semant.sml:413.2-540.9: SemantFun[2].transDec.trdec[2] 4121 : CALL lib/semant.sml:99.2-396.21: SemantFun[2].transExp.trexp 4122 : CALL lib/semant.sml:8.52-558.4: SemantFun[2].transProg[2] 4123 : CALL main.sml:1.18-118.4: Main.compile[2] 4124 : - 4125 : 4126 : ---------------------------------------------------------------------- 4127 : blume 673 Name: Matthias Blumen 4128 : Date: 2000/06/21 18:00:00 JST 4129 : Tag: blume-20000621-manual 4130 : Description: 4131 : 4132 : CM manual update: Path environments documented. 4133 : 4134 : ---------------------------------------------------------------------- 4135 : blume 666 Name: Matthias Blume 4136 : blume 672 Date: 2000/06/19 13:40:00 4137 : Tag: blume-20000619-manual 4138 : Description: 4139 : 4140 : CM manual and system/README update. This only covers the fact that 4141 : there are no more implicit anchors. (Path environments and the "bind" 4142 : option to "cm" have yet to be documented.) 4143 : 4144 : ---------------------------------------------------------------------- 4145 : Name: Matthias Blume 4146 : blume 671 Date: 2000/06/19 11:05:00 JST 4147 : Tag: blume-20000619-chdir-bugfix 4148 : Description: 4149 : 4150 : Fixed a bug in new SrcPath module that sometimes led to a bad chDir call. 4151 : 4152 : ---------------------------------------------------------------------- 4153 : Name: Matthias Blume 4154 : blume 670 Date: 2000/06/18 22:00:10 JST 4155 : Tag: blume-20000618-implicit-anchors-really-gone 4156 : Description: 4157 : 4158 : I updates the previous HISTORY entry where I forgot to mention that 4159 : implicit anchors are no longer with us. 4160 : 4161 : The current update also gets rid of the (now useless) controller 4162 : CM.Control.implicit_anchors. 4163 : 4164 : ---------------------------------------------------------------------- 4165 : Name: Matthias Blume 4166 : blume 666 Date: 2000/06/16 17:30:00 JST 4167 : Tag: blume-20000616-anchorenv 4168 : Description: 4169 : 4170 : This patch implements the long anticipated (just kidding :) "anchor 4171 : environment" mechanism. In the course of doing this, I also 4172 : re-implemented CM's internal "SrcPath" module from scratch. The new 4173 : one should be more robust in certain boundary cases. In any case, it 4174 : is a lot cleaner than its predecessor (IMHO). 4175 : 4176 : This time, although there is yet another boot file format change, I 4177 : kept the unpickler backward-compatible. As a result, no new bootfiles 4178 : are necessary and bootstrapping is straightforward. (You cannot read 4179 : new bootfiles into an old system, but the other way around is no 4180 : problem.) 4181 : 4182 : Visible changes: 4183 : 4184 : blume 670 ** 0. Implicit path anchors (without the leading$-symbol) are no 4185 : longer recognized at all. This means that such path names are not 4186 : illegal either. For example, the name basis.cm simply refers to a 4187 : local file called "basis.cm" (i.e, the name is an ordinary path 4188 : relative to .cm-files directory). Or, to put it differently, only 4189 : names that start with $are anchored paths. 4190 : blume 666 4191 : blume 670 ** 1. The$ abbreviation for $/ has finally 4192 : vanished. 4193 : 4194 : John (Reppy) had critizised this as soon as I originally proposed and 4195 : blume 666 implemented it, but at that time I did not really deeply believe 4196 : him. :) Now I came full-circle because I need the$ syntax 4197 : in another place where it cannot be seen as an abbreviation for 4198 : $/. To avoid the confusion,$ now means what it 4199 : seems to mean (i.e., it "expands" into the corresponding anchor 4200 : value). 4201 : 4202 : However, when paths are used as members in CM description files, it 4203 : continues to be true that there must be at least another arc after the 4204 : anchor. This is now enforced separately during semantic analysis 4205 : blume 670 (i.e., from a lexical/syntactical point of view, the notation is ok.) 4206 : blume 666 4207 : ** 2. The "cm" class now accepts an option "bind". The option's value 4208 : is a sub-option list of precisely two items -- one labeled "anchor" 4209 : and the other one labeled "value". As you might expect, "anchor" is 4210 : used to specify an anchor name to be bound, and "value" specifies what 4211 : the anchor is being bound to. 4212 : 4213 : The value must be a directory name and can be given in either standard 4214 : syntax (including the possibility that it is itself an anchored path) 4215 : or native syntax. 4216 : 4217 : Examples: 4218 : 4219 : foo.cm (bind:(anchor:bar value:$mystuff/bar)) 4220 : lib.cm (bind:(anchor:a value:"H:\\x\\y\\z")) (* only works under windows *) 4221 : 4222 : and so on. 4223 : 4224 : The meaning of this is that the .cm-file will be processed with an 4225 : augmented anchor environment where the given anchor(s) is/are bound to 4226 : the given values(s). 4227 : 4228 : The rationale for having this feature is this: Suppose you are trying 4229 : to use two different (already stable) libraries a.cm and b.cm (that 4230 : you perhaps didn't write yourself). Further, suppose each of these 4231 : two libraries internally uses its own auxiliary library$aux/lib.cm. 4232 : Normally you would now have a problem because the anchor "lib" can not 4233 : be bound to more than one value globally. Therefore, the project that 4234 : uses both a.cm and b.cm must locally redirect the anchor to some other 4235 : place: 4236 : 4237 : a.cm (bind:(anchor:lib value:/usr/lib/smlnj/a-stuff)) 4238 : b.cm (bind:(anchor:lib value:/usr/lib/smlnj/b-stuff)) 4239 : 4240 : This hard-wires $lib/aux.cm to /usr/lib/smlnj/a-stuff/aux.cm or 4241 : /usr/lib/smlnj/b-stuff/aux.cm, respectively. 4242 : 4243 : Hard-wiring path names is a bit inflexible (and CM will verbosely warn 4244 : you when you do so at the time of CM.stabilize). Therefore, you can 4245 : also use an anchored path as the value: 4246 : 4247 : a.cm (bind:(anchor:lib value:$a-lib)) 4248 : b.cm (bind:(anchor:lib value:$b-lib)) 4249 : 4250 : Now you can globally configure (using the usual CM.Anchor.anchor or 4251 : pathconfig machinery) bindings for "a-lib" and "b-lib". Since "lib" 4252 : itself is always locally bound, setting it globally is no longer 4253 : meaningful or necessary (but it does not hurt either). In fact, "lib" 4254 : can still be used as a global anchor for separate purposes. As a 4255 : matter of fact, one can locally define "lib" in terms of a global 4256 : "lib": 4257 : 4258 : a.cm (bind:(anchor:lib value:$lib/a)) 4259 : b.cm (bind:(anchor:lib value:$lib/b)) 4260 : 4261 : ** 3: The encoding of path names has changed. This affects the way 4262 : path names are shown in CM's progress report and also the internal 4263 : protocol encoding used for parallel make. 4264 : 4265 : The encoding now uses one or more ':'-separated segments. Each 4266 : segments corresponds to a file that has been specified relative to the 4267 : file given by its preceding segment. The first segment is either 4268 : relative to the CWD, absolute, or anchored. Each segment itself is 4269 : basically a Unix pathname; all segments but the first are relative. 4270 : 4271 : Example: 4272 : 4273 :$foo/bar/baz.cm:a/b/c.sml 4274 : 4275 : This path denotes the file bar/a/b/c.sml relative to the directory 4276 : denoted by anchor "foo". Notice that the encoding also includes 4277 : baz.cm which is the .cm-file that listed a/b/c.sml. As usual, such 4278 : paths are resolved relative to the .cm-files directory, so baz.cm must 4279 : be ignored to get the "real" pathname. 4280 : 4281 : To make this fact more obvious, CM puts the names of such "virtual 4282 : arcs" into parentheses when they appear in progress reports. (No 4283 : parentheses will appear in the internal protocol encoding.) Thus, 4284 : what you really see is: 4285 : 4286 : \$foo/bar/(baz.cm):a/b/c.sml 4287 : 4288 : I find this notation to be much more informative than before. 4289 : 4290 : Another new feature of the encoding is that special characters 4291 : including parentheses, colons, (back)slashes, and white space are 4292 : written as \ddd (where ddd is the decimal encoding of the character). 4293 : 4294 : blume 670 *** The CM manual still needs to be updated. 4295 : 4296 : blume 666 ---------------------------------------------------------------------- 4297 : leunga 660 Name: Allen Leung 4298 : Date: 2000/06/15 00:38:00 4299 : Tag: leunga-20000615-x86-peephole 4300 : 4301 : x86 Peephole fix by Fermin. Affects c-- and moby only. 4302 : 4303 : ---------------------------------------------------------------------- 4304 : blume 658 Name: Matthias Blume 4305 : Date: 2000/06/12 11:40:00 4306 : Tag: blume-20000612-parmakefix 4307 : Description: 4308 : 4309 : More cleanup after changing the file naming scheme: This time I 4310 : repaired the parallel make mechanism for CMB.make which I broke earlier. 4311 : 4312 : ---------------------------------------------------------------------- 4313 : leunga 657 Name: Allen Leung 4314 : Date: 2000/06/09 01:25:00 4315 : Tag: leunga-20000609-various 4316 : 4317 : None of these things should affect normal SML/NJ operations 4318 : 4319 : 1. Peephole improvements provided by Fermin (c--) 4320 : 2. New annotation DEFUSE for adding extra dependence (moby) 4321 : 3. New X86 LOCK instructions (moby) 4322 : 4. New machine description language for reservation tables (scheduling) 4323 : 5. Fixes to various optimization/analysis modules (branch chaining, dominator 4324 : trees etc.) 4325 : 6. I've changed the CM files so that they can work with versions 4326 : 110.0.6, 110.25 and 110.28 4327 : 4328 :