1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425 426 427 428 429 430 431 432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 447 448 449 450 451 452 453 454 455 456 457 458 459 460 461 462 463 464 465 466 467 468 469 470 471 472 473 474 475 476 477 478 479 480 481 482 483 484 485 486 487 488 489 490 491 492 493 494 495 496 497 498 499 500 501 502 503 504 505 506 507 508 509 510 511 512 513 514 515 516 517 518 519 520 521 522 523 524 525 526 527 528 529 530 531 532 533 534 535 536 537 538 539 540 541 542 543 544 545 546 547 548 549 550 551 552 553 554 555 556 557 558 559 560 561 562 563 564 565 566 567 568 569 570 571 572 573 574 575 576 577 578 579 580 581 582 583 584 585 586 587 588 589 590 591 592 593 594 595 596 597 598 599 600 601 602 603 604 605 606 607 608 609 610 611 612 613 614 615 616 617 618 619 620 621 622 623 624 625 626 627 628 629 630 631 632 633 634 635 636 637 638 639 640 641 642 643 644 645 646 647 648 649 650 651 652 653 654 655 656 657 658 659 660 661 662 663 664 665 666 667 668 669 670 671 672 673 674 675 676 677 678 679 680 681 682 683 684 685 686 687 688 689 690 691 692 693 694 695 696 697 698 699 700 701 702 703 704 705 706 707 708 709 710 711 712 713 714 715 716 717 718 719 720 721 722 723 724 725 726 727 728 729 730 731 732 733 734 735 736 737 738 739 740 741 742 743 744 745 746 747 748 749 750 751 752 753 754 755 756 757 758 759 760 761 762 763 764 765 766 767 768 769 770 771 772 773 774 775 776 777 778 779 780 781 782 783 784 785 786 787 788 789 790 791 792 793 794 795 796 797 798 799 800 801 802 803 804 805 806 807 808 809 810 811 812 813 814 815 816 817 818 819 820 821 822 823 824 825 826 827 828 829 830 831 832 833 834 835 836 837 838 839 840 841 842 843 844 845 846 847 848 849 850 851 852 853 854 855 856 857 858 859 860 861 862 863 864 865 866 867 868 869 870 871 872 873 874 875 876 877 878 879 880 881 882 883 884 885 886 887 888 889 890 891 892 893 894 895 896 897 898 899 900 901 902 903 904 905 906 907 908 909 910 911 912 913 914 915 916 917 918 919 920 921 922 923 924 925 926 927 928 929 930 931 932 933 934 935 936 937 938 939 940 941 942 943 944 945 946 947 948 949 950 951 952 953 954 955 956 957 958 959 960 961 962 963 964 965 966 967 968 969 970 971 972 973 974 975 976 977 978 979 980 981 982 983 984 985 986 987 988 989 990 991 992 993 994 995 996 997 998 999 1000 1001 1002 1003 1004 1005 1006 1007 1008 1009 1010 1011 1012 1013 1014 1015 1016 1017 1018 1019 1020 1021 1022 1023 1024 1025 1026 1027 1028 1029 1030 1031 1032 1033 1034 1035 1036 1037 1038 1039 1040 1041 1042 1043 1044 1045 1046 1047 1048 1049 1050 1051 1052 1053 1054 1055 1056 1057 1058 1059 1060 1061 1062 1063 1064 1065 1066 1067 1068 1069 1070 1071 1072 1073 1074 1075 1076 1077 1078 1079 1080 1081 1082 1083 1084 1085 1086 1087 1088 1089 1090 1091 1092 1093 1094 1095 1096 1097 1098 1099 1100 1101 1102 1103 1104 1105 1106 1107 1108 1109 1110 1111 1112 1113 1114 1115 1116 1117 1118 1119 1120 1121 1122 1123 1124 1125 1126 1127 1128 1129 1130 1131 1132 1133 1134 1135 1136 1137 1138 1139 1140 1141 1142 1143 1144 1145 1146 1147 1148 1149 1150 1151 1152 1153 1154 1155 1156 1157 1158 1159 1160 1161 1162 1163 1164 1165 1166 1167 1168 1169 1170 1171 1172 1173 1174 1175 1176 1177 1178 1179 1180 1181 1182 1183 1184 1185 1186 1187 1188 1189 1190 1191 1192 1193 1194 1195 1196 1197 1198 1199 1200 1201 1202 1203 1204 1205 1206 1207 1208 1209 1210 1211 1212 1213 1214 1215 1216 1217 1218 1219 1220 1221 1222 1223 1224 1225 1226 1227 1228 1229 1230 1231 1232 1233 1234 1235 1236 1237 1238 1239 1240 1241 1242 1243 1244 1245 1246 1247 1248 1249 1250 1251 1252 1253 1254 1255 1256 1257 1258 1259 1260 1261 1262 1263 1264 1265 1266 1267 1268 1269 1270 1271 1272 1273 1274 1275 1276 1277 1278 1279 1280 1281 1282 1283 1284 1285 1286 1287 1288 1289 1290 1291 1292 1293 1294 1295 1296 1297 1298 1299 1300 1301 1302 1303 1304 1305 1306 1307 1308 1309 1310 1311 1312 1313 1314 1315 1316 1317 1318 1319 1320 1321 1322 1323 1324 1325 1326 1327 1328 1329 1330 1331 1332 1333 1334 1335 1336 1337 1338 1339 1340 1341 1342 1343 1344 1345 1346 1347 1348 1349 1350 1351 1352 1353 1354 1355 1356 1357 1358 1359 1360 1361 1362 1363 1364 1365 1366 1367 1368 1369 1370 1371 1372 1373 1374 1375 1376 1377 1378 1379 1380 1381 1382 1383 1384 1385
|
Fri Nov 24 18:42:46 1995 Bill Schelter <wfs@jany.ma.utexas.edu>
* alloc.c (ONCE_MORE):
* DEFUN("STATICP",.. had accidentally been included in a section
which was '#ifdef'ed out on nexts..., making it not be there
at link time: moved it to where it is always there.
*
Nov 11 1995 bill schelter
* gcl-2.2 released
Sun Oct 1 19:52:45 1995 Bill Schelter <wfs@jany.ma.utexas.edu>
* Many changes to gcl 2.1 to support 64 bit machines (eg Dec
alpha). Layout of structures etc changed.
* a gcl-2.2 beta was released in the summer.
since then there have been several bugs fixed. One in cmpfun.lsp
affecting write, and another in init_gcl.lsp to make sure the
link array is a string array (changed from fixnum which are no
longer sufficient to hold pointers).
* changes to fix for PA risc hpux in the hp800.h
* changes to unexec-19.27.c to allow MUCH faster saving in NFS
environment.
* testing with maxima 5.1
* reworking makefiles
Sat Apr 29 08:48:06 1994 Bill Schelter (wfs at nicolas.ma.utexas.edu)
* Changed to release under GNU Public Library license. There
have been a number of other fixes including fixes to bignums.
Thu Jan 20 10:38:00 1994 Bill Schelter (wfs at nicolas.ma.utexas.edu)
* version 624 made.
* contains just changes so that compiles on solaris 5.2. The
623 version compiled on solaris 5.3, but the earlier 5.2 version of solaris
had some differences which needed patches. These are contained
in 624.
Fri Dec 10 15:02:14 1993 Bill Schelter (wfs at nicolas.ma.utexas.edu)
* version 623 made.
* the check on string-trim for a list of chars is fixed.
Much earlier the (string-trim '(2) "ab") would run forever.
Then string-trim was broken.
Sun Dec 5 15:34:44 1993 Bill Schelter (wfs at nicolas.ma.utexas.edu)
* add solaris version: changes to several C files
and new version of sfasl for elf.
* linux port added.
Thu Oct 29 13:20:17 1992 Bill Schelter (wfs at nicolas.ma.utexas.edu)
* make sure the signal stack is 8 byte aligned. Needed on sun os 4.1.2
and higher.
Thu Jun 4 06:18:20 1992 Bill Schelter (wfs at nicolas.ma.utexas.edu)
* fix allocation of copy space during an sgc for relocatable
Wed Apr 29 09:02:59 1992 Bill Schelter (wfs at nicolas.ma.utexas.edu)
*cmpnew/lfuns.lsp make load,open, error-set go through lisp
symbol, so users can redefine them (eg when loaded common lisp
condition code stuff).
* defstruct.lsp: put the conc-name in the current package,
so that programs can discover the package in effect when
the defstruct was done, in order to reconstruct accessors.
Sun Apr 26 23:28:42 1992 Bill Schelter (wfs@sonia.ma.utexas.edu)
* predicate.c: contains_sharp_comma handle non type t arrays
* co1typep optimize for the '(satifies fun) type.
* cmpinline.lsp, cmplabel.lsp, cmploc.lsp
inline-integer (set-loc unwind-exit get-inline-loc)
fixes by r harris.
Thu Apr 23 15:09:44 1992 Bill Schelter (wfs at nicolas.ma.utexas.edu)
* cmpenv.lsp fix function return integer proclamation.
Sat Apr 11 09:25:05 1992 Bill Schelter (wfs at nicolas.ma.utexas.edu)
* numlib.lsp: changes to ensure double accuracy not lost.
Thu Apr 9 11:21:25 1992 Bill Schelter (wfs at nicolas.ma.utexas.edu)
* log,exp,.. get more accuracy if given a fixnum[don't use short-float]
* cmptop.lsp empty keyword list bug (defun foo(&key) nil)
* many files add hooks for dos port.
Fri Mar 27 15:59:24 1992 Bill Schelter (wfs at nicolas.ma.utexas.edu)
* (cmpinline.lsp) inline-args of a closure var which changed later,
signalled a compiler error.
Wed Mar 25 14:38:41 1992 Bill Schelter (wfs at nicolas.ma.utexas.edu)
* (read.d) make read listen to when ':' is a read macro. in read_object
* (cmpfun.lsp,cmpopt.lsp): Fix optimizer for (typep x 'ratio)
* (predicate.c): fix equalp so (equalp (format nil "hi") "hi") -->t
Wed Mar 11 14:16:59 1992 Bill Schelter (wfs at nicolas.ma.utexas.edu)
* (read.d): fix failure of (read-string "#b1"). Handle eof in
read_constituent
Wed Feb 26 21:38:04 1992 Bill Schelter (wfs at nicolas.ma.utexas.edu)
* cmpfun.lsp change read-char optimizer so as to avoid
a C compiler bug on dec3100, which causes an unaligned access
on that machine in calls to read-char.
Wed Jan 29 08:30:09 1992 Bill Schelter (wfs at nicolas.ma.utexas.edu)
* hp800.{h,defs} changes to run incrementally loaded text in a
separate segment (at 4zillion). Without this hp will only let
you run one image at a time. Unfortunately there will be a slight
degradation in indirect calls (which includes all calls between
user functions). Why do they need to have segments....
Thu Jan 23 17:01:37 1992 Bill Schelter (wfs at nicolas.ma.utexas.edu)
* add some save and retore around the terminal interrupt, to
prevent some possible lossage when continuing from an interrupt.
You still may lose if there is a gc during an interrupt.
Wed Jan 22 19:08:11 1992 Bill Schelter (wfs at nicolas.ma.utexas.edu)
* support for mac under AUX. Files mac2.h, mac2.defs [by weigert]
* misc small changes to other files for mac.
Sun Jan 5 21:35:21 1992 Bill Schelter (wfs at nicolas.ma.utexas.edu)
* new definitions for fplus and fminus in cmac.c The ones for
the 68020 were incorrect (since the changeover of the bignum code).
These were only used by maxima.
Fri Nov 22 08:39:42 1991 Bill Schelter (wfs at nicolas.ma.utexas.edu)
* Change proclaim of function so to turn integer
proclained args into type T. It is not possible to
pass raw integer type since this would cause a problem
with gc. Also currently it would have generated C which
broke the C compilation.
Tue Oct 29 07:59:02 1991 Bill Schelter (wfs at nicolas.ma.utexas.edu)
* add misc/warn-slow.lsp. If this file is loaded, then the compiler
warns of certain slow constructs, like undeclared arithmetic,
and slow array references.
Thu Oct 24 21:34:55 1991 Bill Schelter (wfs at nicolas.ma.utexas.edu)
* funlink.c to handle misproclaimed functions better.
* big.c correct one source of "bad length" warning
Sun Oct 20 12:55:35 1991 Bill Schelter (wfs at nicolas.ma.utexas.edu)
* gbc.c change position of where the C stack is marked from.
This is relevant for sparcs with the register windows, which
are dumped at interrupts.
* cfun,gbc changes as per PCL mods for turbo closures.
* read.d use 1000000000.0 rather than 1/ this , since it gives
more accurate value in read of float.
* cmpfun.lsp add hook for (typep x 'foo) so that can expand
this differently according to a property of 'foo.
* cmpcall.lsp, add a hook so that special code can be emitted
for calls where a super_funcall_no_event would have been emitted.
This code might be used where we expect closures (cf PCL)
Wed Oct 9 21:31:03 1991 Bill Schelter (wfs at nicolas.ma.utexas.edu)
* read.d to make 0.7 and (/ 7.0 10) produce identical results.
when read (at least on 68k,sparc, rios).
We have left the default number of digits printed out as 17,
in spite of the fact that IEE has only 15.95 significant digits.
The reason is that correct rounding to 15 or 16 digits, will
commonly cause things like most-positive-long-float to print
in a form which is not even readable (since it is rounded up above
the most-positive-long-float).
Mon Oct 7 20:54:35 1991 Bill Schelter (wfs at nicolas.ma.utexas.edu)
* fix LD_COMMAND typo in hp800.h
* fix replace_array to make the new array not live
beyond the call, so that gc won't accidentally mark two
copy two identical array bodies (so maybe overstepping the
the new relblock)
Fri Oct 4 10:58:59 1991 Bill Schelter (wfs at nicolas.ma.utexas.edu)
* close to work on already closed streams.
* make_pure_array arg fixed (num_log.c)
* estack_buf put in static area on machines like rios.
Sun Sep 29 12:43:54 1991 Bill Schelter (wfs at nicolas.ma.utexas.edu)
* fix to relblock allocation in sgc_start
* fix for paths with "~/"
* change default pagesize on sun3 to 4k
Sat Sep 7 13:51:24 1991 Bill Schelter (wfs at nicolas.ma.utexas.edu)
* fix bit-array-op to make correct call to si::make-array-pure
(num_log.c)
Mon Aug 5 18:06:35 1991 Bill Schelter (wfs at nicolas.ma.utexas.edu)
* wt-cvars add declaration for VXXalloc variables
Tue Jul 30 16:48:09 1991 Bill Schelter (wfs at nicolas.ma.utexas.edu)
* interpreted mapcar not gc protecting when given more than 2 args.
Sat Jul 27 12:32:35 1991 Bill Schelter (wfs at nicolas.ma.utexas.edu)
* changes to subtypep to make (subtypep 'cons '(and t cons))
and (subtypep 'simple-array '(not vector)) correct.
Fri Jul 26 09:25:56 1991 Bill Schelter (wfs at nicolas.ma.utexas.edu)
* change equal and hash_equal not to descend into structures.
A ruling in CLTL2, clarifies that this should be the case.
* alter get-setf-method and friends to accept environment as second
argument. Alter all the complex setf methods to pass the environment
so that local functional and macro bindings can alter the behaviour
of the setf macro.
Tue Jul 23 06:25:39 1991 Bill Schelter (wfs at nicolas.ma.utexas.edu)
* sharing for fasd files for package operations was disabled.
Otherwise the symbol in a shadow would be shared with the symbol's
later occurrence in the file--It is still a good idea to put
package operations in a separate lisp file at the beginning of
system load.
* adjustable arrays brought into conformance with changes in CLTL
II. You
may now adjust non adjustable arrays, and the fill-pointer argument
to adjust-array does not change the property of having or not having
a fill pointer for the array.
Sun Jul 21 12:44:04 1991 Bill Schelter (wfs at nicolas.ma.utexas.edu)
* fix cmac.c for maxima (error in dblrem)
* sloop for v on l by 'joe
changed to allow for the possibility that joe is a macro.
[used by maxima].
Wed Jul 10 10:45:54 1991 Bill Schelter (wfs at nicolas.ma.utexas.edu)
* littleXlsp is included which provides an interface to some
simple X windows routines. It is not in the image by default
See the file lsp/littleXlsp.lsp for directions.
Tue Jul 9 16:29:23 1991 Bill Schelter (wfs at nicolas.ma.utexas.edu)
* add VOL declarations for setjmps in format.c to allow to work
with gcc on the sparc.
* add stuff to cmplam.lsp for VOL declarations of &aux variables.
Wed Jun 26 20:59:04 1991 Bill Schelter (wfs at nicolas.ma.utexas.edu)
* Change to gc mark of c stack to make sure register windows
are flushed, by calling recursively.
* add special bignum code for rs 6000 and aix 370
* fix for read suppress and the defstruct reader.
* Source level debugging improved (see doc/DOC and doc/dbl.el)
* Catch infinite recursion of proclaimed functions. Handle
segmentation faults on an alternate stack--so don't play around
to much after a segmentation fault before quitting to top level.
* support for hp800 (and hp700) added.
Wed May 22 14:54:29 1991 Bill Schelter (wfs at nicolas.ma.utexas.edu)
* added new slot to struct stream, to hold the buffer.
gbc.c and file.d had been referencing the field in FILE
directly, but that is not reliable on att Sys V 4.
Thu Apr 11 17:04:54 1991 Bill Schelter (wfs at nicolas.ma.utexas.edu)
* add conc-name to the s-data structure so users may get at
it (defstruct.lsp).
* add install_segmentation_catcher() and also in error()
(ususually arises from real segmentation faults), then turn
off sgc, if it's on, to avoid having the error handler stall
on trying to alter pages which are write protected (since
we are still within the memprotect_handler.
Thu Apr 4 08:37:50 1991 Bill Schelter (wfs at max.ma.utexas.edu)
* If HAVE_YP_UNBIND is defined, added to c/main.c the
unbinding of the default yp domain. In sun os 4.1
we were getting a segmentation fault in _yp_dobind_soft()
on a restarted system.
Sat Mar 30 09:01:57 1991 Bill Schelter (wfs at nicolas.ma.utexas.edu)
* Fix listen and clear-input (file.d and read.d).
macros LISTEN_FOR_INPUT and HAVE_IOCTL are defined in
att.h, bsd.h, and for the aix systems. I have tested
it under aix, sun, 4.3bsd, sgi4d and hpux. Not sure about
vannilla sysv (may have to be #undef'd there).
* fixed pathnames such as "~wfs/foo.lisp" to work.
(see unixfsys.c).
Mon Mar 25 12:25:34 1991 Bill Schelter (wfs at nicolas.ma.utexas.edu)
* Make sure that long float arrays get allocated on multiple of
sizeof(double ) alignment array.c,gbc.c.. Also unixfasl.c
* Add to Smakefile an initial execution of xbin/{machine}-fixes
if that file exists. Use this for correction of temporary bug
fixes, such as the bad sgi4d c compiler.
Mon Mar 18 12:06:26 1991 Bill Schelter (wfs at nicolas.ma.utexas.edu)
* The structure of bignums and the underlying code was changed
completely. This affected many files including big.c,num_co.c
print.d, read.d, predicate.c, num_arith.c,.. See the file doc/bignum
for a discussion. The compiler was also changed to include
integer as a primitive type with storage allocated on the stack.
* A notion of deducing result type from argument types was
added to the compiler. Initially we are just doing that for
the basic integer functions, but it can be extended to others
* The optimization in cmpopt.lsp have been changed to allow more
flags to accommodate things like the result-type-from-args.
The compiler will normally warn (for the time being) if you
give it old style optimizations. Because of the extensive
changes to the compiler I have changed the default safety
for compiling the cmpnew directory back to 2.
Thu Feb 14 16:06:42 1991 Bill Schelter (wfs at nicolas.ma.utexas.edu)
* debug.lsp, eval.c: fix break-step-next and break-step-into to pass
the correct environment back so that evaluating variables or
local functions will be done correctly in the debugger.
Mon Feb 11 16:24:15 1991 Bill Schelter (wfs at nicolas.ma.utexas.edu)
* Fixed *break-points* function to bind the correct enviroment
so that variables will get the right values.
Mon Jan 7 21:21:22 1991 Bill Schelter (wfs at nicolas.ma.utexas.edu)
* make inline string adjustable (cmptop.lsp)
* multiple changes to lsp/debug.lsp and lsp/top.lsp
to allow source line debugging using si::nload.
* debugger largely redone. :bt new backtrace
function (:b still there). see DOC file
* add xdr-open xdr-write xdr-read to the si package.
* akcl 532 compatible with maxima 4-153
Wed Dec 5 01:49:50 1990 Bill Schelter (wfs at nicolas.ma.utexas.edu)
* check_alist should allow nil, since the CLTL allows nil
in place of a cons in an alist.
* many changes to debug.lsp, eval.c and top.lsp to allow
source level debugging. The emacs file dbl.el was added.
It and the DOC file contain more information, but basically
there is automatic source display when broken in the debugger
(for lisp files loaded with nload).
Tue Nov 20 20:07:02 1990 Bill Schelter (wfs at nicolas.ma.utexas.edu)
* Changes to all places where function assignments are made
to allow the hook compiler-def-hook, to be run. This hook
is used by the new source code debugger dbl.
* dbl allows debugging of lisp code with a display of an
arrow in the window opposite the line currently broken at
or being executed.
* The safety on the cmpnew files has been changed to safety 0.
Please notify of any places where this causes a problem! It
should result in significantly faster compilation. The error
checks should be in the source.
Thu Nov 8 05:31:34 1990 Bill Schelter (wfs at nicolas.ma.utexas.edu)
* fix aix3_mprotect/mprotect.c [wrong calculation of overflow]
* fix memory_protect in sgbc.c
Wed Nov 7 09:56:30 1990 Bill Schelter (wfs at nicolas.ma.utexas.edu)
* Change compiler-clear-compiler-properties, to take 2 args. This
is to provide compiler-def-hook, which lets you get the code just
before it is installed.
* Misc support for IBM 370 mainframe running under AIX. (u370)
* change malloc at startup to use some static space, since
the gc may not be initialized before some startup routines need
to malloc (aix3).
* fix to cmptag.lsp. Tags in cross closure tagobdies were sometimes not
being written out. (bug had been there forever). C compiler would
fail when the tag was not there.
Fri Oct 26 15:00:37 1990 Bill Schelter (wfs at nicolas.ma.utexas.edu)
* change call_proc, the routine used to link proclaimed functions.
It was not incrementing vs_top soon enough, so that with args
(t t fixnum), with fast links off, the make_fixnum caused by
the last arg, might cause a gc which zeroed the vs stack above
vs_top, so eliminating arg1. Symptom was an <objnull> passed
as second arg. This could only happen when functions were proclaimed,
and had type not = t.
Sat Oct 6 08:08:47 1990 Bill Schelter (wfs at nicolas.ma.utexas.edu)
* change assoc in list.d so (assoc nil '(nil (nil . a))) --> (nil
. a)
also change cmpfun.lsp in the compiler for this. (bug rep mccain)
* Allow #' in the (:print-function #'(lambda (..)) defstruct option
(bug report baxter)
Thu Sep 20 19:05:05 1990 Bill Schelter (wfs at nicolas.ma.utexas.edu)
* Fix cmpthrow in cmpcatch.lsp so that handles
lexical closures correctly.
* Fix cmpeval.lsp for incrementing a structure slot
which is fixnum.
* Fix cmpmain.lsp adding support for the floating point
save ops on rios.
Sun Sep 2 16:44:24 1990 Bill Schelter (wfs at nicolas.ma.utexas.edu)
* Fix perm_writable. It was leaving out the last page! (Vignaux)
Mon Aug 20 15:15:03 1990 Bill Schelter (wfs at nicolas.ma.utexas.edu)
* fix acos and asin in numlib.lsp (bug was (acos .5) was complex!)
Wed Aug 15 14:01:06 1990 Bill Schelter (wfs at nicolas.ma.utexas.edu)
* to fasdump.c: close_fasd was munging the array handed
to it to cleanup. This could cause gc problems later.
* Several changes to gc. The alloc_relblock function fixed
to take into acount that rb_start - heap_end may not be
holepage anymore [since if sgc_enabled we have a second
`sgc' rb_start after the first]. We ensure that nrbpage
is actually the combined number of pages for relblock,
for sgc and regular gc. (alloc.c)
Tue Aug 14 16:04:19 1990 Bill Schelter (wfs at nicolas.ma.utexas.edu)
* changes to sgc to the way memory protection is turned on.
* There may be problems with saving an image with (sgc-on t)
These have not been resolved, at least under rios, and perhaps
other machines.
* many changes for aix 3, for rios
* for sgi, change the Init_links to come before doinit
* the sgi4d does not need the links stuff.
Wed Aug 8 21:59:37 1990 Bill Schelter (wfs at nicolas.ma.utexas.edu)
* Change to h/secondary_sun_magic to make it have 8 characters.
* fix use-package so that allows using a package with an external
symbol, if that symbol is shadowed by the current package.
* Several fixes for sun os 4.1 (secondary magic stuff).
Wed May 23 11:47:34 1990 Bill Schelter (wfs at nicolas.ma.utexas.edu)
* Make traced functions stay traced after redefinition
[change clear_compiler_properties, and its many callers]
* Fix two bugs introduced in trace.lsp when its functionality
was increased.
* fasdump.c (fix coercion).
Thu Apr 12 15:28:49 1990 Bill Schelter (wfs at nicolas.ma.utexas.edu)
* Change behaviour of proclaimed, compiled keyword argument functions, so that
duplicate keys are allowed, and the leftmost takes precedence.
[bind.c:parse_key_new]
Mon Apr 9 14:48:43 1990 Bill Schelter (wfs at nicolas.ma.utexas.edu)
* make subtypep (predlib.lsp) handle types of form '(not p) correctly. This
in turn influences compiler optimizer handling of (typep x '(not p))
* fix read (read.d) so that it allows eof to occur during reading of a semicolon
comment line.
* fix c1values (cmpmulti.lsp) so that (values (truncate a b)) does just
return one value.
Sun Apr 1 22:11:46 1990 Bill Schelter (wfs at nicolas.ma.utexas.edu)
* make float-digits and float-precision return the significant
digits in terms of float-radix.
* If si::*print-nans* is not nil, then the C printing of Nan's
and infinity surrounded by #< > will be used.
Fri Mar 30 10:24:35 1990 Bill Schelter (wfs at nicolas.ma.utexas.edu)
* gbc.c, sgbc.c: displaced arrays have been a real headache.
(dotimes (i 5000) (make-array 1 :element-type 'string-char :displaced-to
(format nil "0")))
would cause bad things to happen in kcl and akcl. I finally decided
that the link between an array and the array it is displaced to should
be made firm. If A is displaced to B, the user can always do adjust-array
on A to destroy the displacement. But from now on, as long as the
displacement exists then if B is marked this will cause marking of A,
and vice versa.
Wed Mar 28 16:41:17 1990 Bill Schelter (wfs at nicolas.ma.utexas.edu)
* cmpeval.lsp co1structure-predicate:
(defstruct foo a b ..) then foo-p was less efficient
than it had been in some earlier release.
Note:after (si::freeze-defstruct 'foo) you get the fastest foo-p,
since this declares the hierarchy of structures including
foo to be frozen.
Wed Mar 7 13:30:57 1990 Bill Schelter (wfs at nicolas.ma.utexas.edu)
* fixed copying of relblock in gc, so that if an array body
is allocated on C stack [the read_fasd does this for a temp array
it needs] then this array will not be copied. In some cases this
could have caused the copied relblock to exceed nrbpage size, which
is all that sbrk had provided.
* read-fasd could not be recursively entered from the lisp_eval calls which
were possible [from things like require or other package ops].
Fixed in fasdump.c
Mon Mar 5 11:54:49 1990 Bill Schelter (wfs at nicolas.ma.utexas.edu)
* Increase BDSGETA to allow some freedom in the debugger after
a bds overflow.
Tue Feb 27 09:19:49 1990 Bill Schelter (wfs at nicolas.ma.utexas.edu)
* fix printing NIL relative to a package not using LISP.
Tue Feb 20 00:10:58 1990 Bill Schelter (wfs at fireant.ma.utexas.edu)
* Add compiler::*split-files*, to allow convenient splittling of
large lisp files, for C compilers which can't handle infinitely
long C files. See doc/DOC file.
* Fix to fasdump.c. Broke when a .o file was loaded during compilation.
Fri Feb 16 09:11:08 1990 Bill Schelter (wfs at nicolas.ma.utexas.edu)
* si::*load-pathname* is now bound to the pathname of the current file
being loaded.
Sat Feb 10 13:36:44 1990 Bill Schelter (wfs at nicolas.ma.utexas.edu)
* Add the co1special prop check in c1symbol-fun, for fixing macros which
expand into declares in do, do*, prog, prog* (see also cmpfun.lsp)
* Fix level in c2call-local (bug report by harris).
Tue Jan 23 18:37:58 1990 Bill Schelter (wfs at nicolas.ma.utexas.edu)
* change to siLreplace_array, so that the first word of the old
array header is preserved, so that in case sgc is on, the
array won't be marked SGC_RECENT, and so garbage collected if
there were no pointers to it.
Sun Jan 21 20:30:48 1990 Bill Schelter (wfs at nicolas.ma.utexas.edu)
* Many changes to the compiler: Only defuns and defmacros are
compiled by default. A flag compiler::*compile-ordinaries* if
t means all forms will be compiled (use this for pcl at the moment).
See doc under compile-file.
* eval-when default behaviour changed to be in line with the X3j13
CL standard. See compile-file doc.
files: fasdump.c,cmpaux.c,cfun.c, cmptop, cmpwt,cmpenv, cmpspecial,
cmpflet.
* some advantages of the new init scheme. Some files can be substantially
smaller, and there is more flexibility in writing out the init. Things
like closures which are constant, can be set in as constants.
Thu Jan 4 23:24:46 1990 Bill Schelter (wfs at nicolas.ma.utexas.edu)
* make sublis for eql equal and eq be substantially more efficient.
[list.d, and cmpfun.lsp]
* fix file_exists for AIX (stat returns 0 there even if file
ending in slash is not a directory (ie /u/all/bmt/.login/ would
have existed.
* make boolean be a real type, so that we can distinguish between
calls to a function which want only a boolean reply or calls
which need more. An example is probe-file which can be 30
times faster when only a boolean reply is needed, not the truename.
This required changes to cmptype,cmpif cmpinline.
* add additional optimizations to cmpopt.lsp.
* fix vector-push-extend optimization [cmpfun, cmpopt]
* various fixes to sgc
Sat Dec 23 18:30:03 1989 Bill Schelter (wfs at nicolas.ma.utexas.edu)
* Fix multiple-value-prog1 in cmpmulti.lsp
(eval-when (compile) (proclaim '(function foo (t t) t)))
(defun jil () (multiple-value-prog1 (values 1 2) (foo 3 4)))
(defun foo (a b) (joe a b) (cons a b))
(defun joe (a b) (list a b))
would have (jil) --> 3 4 until this fix. This dates from original kcl.
Tue Dec 5 20:22:58 1989 Bill Schelter (wfs at nicolas.ma.utexas.edu)
* changes to alloc.c, gbc.c, page.h, object.h to allow stratified
garbage collection [SGC]. This should help systems with
a large amount of relatively static data. Only pages written
to since sgc-on need be marked and swept. See DOC file.
Sat Nov 11 07:08:27 1989 Bill Schelter (wfs at nicolas.ma.utexas.edu)
* Add various va_end(ap)'s to match unmatched va_start's.
* add in some changes for hp300's faslink.
* add additional support for cmpnew/collectfn.lsp See the
documentation in the DOC file for emit-fn and friends.
Basically it is for getting proclamation info, who-calls info,
undefined info, from a pass of the compiler on a system.
* add additional undefined warnings for undefined lisp package
functions in addition to the list-undefined-functions.
Tue Oct 31 06:02:18 1989 Bill Schelter (wfs at nicolas.ma.utexas.edu)
* add support of Iris 4d machine.
* Fix COERCE_VA_LIST for non standard machines, to
take the argument n.
* Make sure defun, defmacro,.. clear the accessor property
for defstruct slots.
Sat Oct 28 11:06:25 1989 Bill Schelter (wfs at nicolas.ma.utexas.edu)
* remove an abort() from rel_sun4.c. This type
of relocation now occurs, and our method has been tested.
Fri Oct 27 23:01:42 1989 Bill Schelter (wfs at nicolas.ma.utexas.edu)
* if the *system-directory* directory contains the file sys-init.lsp
then this file will be loaded at startup. This facility
is used for printing the warning message at startup, but
can be used for local modifications. Also by having two
system directories [via links and different commands]
different startups could be loaded. All this is in
addition to the regular init.lsp. The purpose of this
is to allow patches to be loaded, without requiring a
resaving of the image. [Recall the system directory
is the first argument to invoking a saved_kcl, or
the first part of the pathname if there is not a first arg
Typically it is the unixport directory if you use the
xbin/kcl command]
Tue Oct 24 20:23:56 1989 Bill Schelter (wfs at nicolas.ma.utexas.edu)
* Add packlib to the unixport/boots file so that help*
gets compiled using the new do-symbols macro.
* Add alternate malloc.c file from gnu emacs, if you
define GNU_MALLOC. This runs much faster (15X) if you are
incorporating lots of C code with mallocs.
Thu Oct 19 21:28:25 1989 Bill Schelter (wfs at nicolas.ma.utexas.edu)
* add si::fwrite and si::fread for doing io on file streams.
* Eliminate the need for the separate compiler file cmpinclude.h
This means that only one file is now needed for an executable,
and there won't be confusion over which cmpinclude.h goes with
which version. We store the file as a string, and write it
into the c files as they are compiled [if *cmpinclude-string* is
a string and :system-p t is not given to compile]. The time
difference for doing the extra write is not measurable, even
with tiny compiles.
Wed Oct 4 23:41:17 1989 Bill Schelter (wfs at rascal)
* co1eql was causing a double evaluation
in some special circumstances.
Mon Oct 2 08:48:30 1989 Bill Schelter (wfs at rascal)
* Many changes to the compiler to support &optional, &rest
&key args to be passed on the C stack. This is still limited
to functions proclaimed to return one value.
* Allow user to grow the stacks from top level [see *multiply-stacks*
doc]
* Code for making proclamation files and collecting
cross referencing data in cmpnew/collectfn.lisp. This
is to allow a second compile of a system to take advantage
of information obtained in the first compile.
* catch-fatal added.
* read.d float read fixed for little endian machines.
* do,prog,prog* and do* compilation fixed so that declare's
at the beginnning of the bodies which are hidden by macros, are
detected and processed.
* :dynamic-extent declaration recognized to allow &rest
* proclaimed functions with one return value, will now be
compiled to use the C stack even if they have more than 10 args.
Tue Sep 5 22:15:23 1989 Bill Schelter (wfs at rascal)
* Change add-function-proclamation to be more efficient
when handed long lists of proclaims. Maxima for example
generates 3871 proclaims of functions. This now takes
4 seconds instead of 2 minutes.
Mon Sep 4 00:25:29 1989 Bill Schelter (wfs at rascal)
* cfun_to_combined fixed for t_sfun,t_gfun.
* ihs_function_name fixed for t_sfun,t_gfun
Mainly used in error handling and printing.
* compiled-function-name fixed to handle t_sfun,t_gfun
[this is used in the error handler]
Thu Aug 31 09:57:09 1989 Bill Schelter (wfs at rascal)
* fix reduce in seqlib.lsp.
Wed Aug 30 16:44:05 1989 Bill Schelter (wfs at rascal)
* change cmpeval.lsp,cmploc.lsp to make small fixnums
write out without the indirection through VV
---------Version 206-------------
Mon Aug 28 13:44:49 1989 Bill Schelter (wfs at rascal)
* cmptag.lsp:add-reg changed to handle dotted list case
* Added new types t_sfun, t_gfun of compiled function
objects to save space and speed funcall. See doc/funcall
for details. Many c files changed as well as compiler.
* cmpeval.lsp: made #, have the c1special property,
instead of c1, so these will compile properly on safety 3
* add si::set-mv, si::mv-ref, to allow people to implement
faster version of multiple values. Someday this way will
be the default [see doc/multiple-values and lsp/fast-mv.lisp]
Thu Aug 10 11:51:05 1989 Bill Schelter (wfs at rascal)
* Fixed prev fix to print in print.d so that ok if stream = t.
Mon Aug 7 10:19:11 1989 Bill Schelter (wfs at rascal)
* Support for HP300bsd (bsd from mt xinu) added.
* print fixed to add space after printing.
* describe modified to print more information on structures
Fri Jul 28 09:18:15 1989 Bill Schelter (wfs at rascal)
* trace.lsp Add keyword args allowing special
entry, exit, conditional and other handling of
traced forms. You can for example specially print
args, or break if args or values are inappropriate.
See file doc/trace.
* c1decl-body: recognize safety level on first pass.
Sat Jul 8 15:21:42 1989 Bill Schelter (wfs at rascal)
* defstruct.lsp and read.d:
Fix patch_sharp to handle structures, and sharp-s-reader
to do its reads recursively.
Allows constructs such as #1=#S(joe a #1#) to work.
Fri Jul 7 13:40:45 1989 Bill Schelter (wfs at rascal)
* make sure the co1 properties are cleared when functions
are defined.
Mon Jul 3 15:05:51 1989 Bill Schelter (wfs at rascal)
* Fix sharp-s-reader in defstruct.lsp and remove it
from iolib.lsp
* remove =* and =- from gbc.c package.d
* fix c1value to behave correctly when one value supplied
in cmpmulti.lsp
* change #. and #, to interact correctly when called
after #+ or #-
Fri Jun 2 20:48:06 1989 Bill Schelter (wfs at rascal)
* sfasl.c gbc.c change error messages from using stderr
to stdout. Note the first file descriptor the user
opens is typically stderr, so these error messages (which
rarely occurred) did not appear on the screen but rather
caused resetting the file pointer of the user's stream!
Wed May 31 20:31:04 1989 Bill Schelter (wfs at rascal)
* package.d,symbol.d, packlib.lsp: Changes
to package hashing and intern. Allow flexibly
sized packages instead of insisting that all
internal and external packages use table with 512 elts.
I recommend a prime number as size. The size of
the internals table is automatically grown when there are 2 x
as many symbols as the table size.
For the byte-reader.im-test file (580K) the read time
went from 24 seconds to 16.8 seconds. (had been 53 seconds).
in-package and make-package take keyword args
:external and :internal to allow specification of the
size of the table for a new package.
Tue May 30 23:00:46 1989 Bill Schelter (wfs at rascal)
* fix writing of small and large double floats
in cmpeval.lsp, so that all numbers from smallest
double to largest may be included as constants
in compiled code.
Sun May 28 16:22:07 1989 Bill Schelter (wfs at rascal)
* array.c:array_allocself takes an additional argument,
specifying the default value, or NULL if no initial
value is to be given.
si::make-vector has an additional optional arg of the
default value.
* seqlib also has calls to array_allocself, which take
a 0 final arg, indicating that the initialization
is not to be done.
* array.c: siLcopy_array_portion added, to allow quick copying
from one array to another.
* In top.lsp the *eof* is changed to a local, so as not
to conflict with the si::*eof* which is the value returned
by the system when doing getc.
g
Fri May 12 10:57:39 1989 Bill Schelter (wfs at rascal)
* Add optimization for (type x 'foo) where
foo is a structure. This also affects foo-p.
If you (setq compiler::*frozen-defstructs* t)
this allows the compiler to assume that a given
defstruct will not be extended (by including
it in more structures) by new defstructs loaded
in later files. This can significantly speed
up type checking.
Thu May 11 07:43:52 1989 Bill Schelter (wfs at rascal)
* Fix omission of the extended mul for sun4 in 1.22
Now a new dependent feature EMUL lets you specify
an assembler file in the o directory, which will
be loaded in at the end.
* AKCL added to the *features* list. This is
now necessary because of internal differences in structures
between the standard kcl and akcl.
Structure changes are pretty well complete.
Wed May 3 12:11:54 1989 Bill Schelter (wfs at rascal)
* Fix ceiling and floor (num_co.c) by riley
Mon May 1 08:05:55 1989 Bill Schelter (wfs at rascal)
* fix to gc of displaced arrays. (bug report
and partial fix by riley). Files gbc.c array.c
undisplace was referrring to possibly freed list
structure.
* Structures are being completely reworked.
There are two reasons:
To use much less overhead at compile time of the
original defstruct.
To allow raw types in defstructs, and allow packing
of these. For example a slot (x 0 :type (mod 50)) will
only require 1 byte. These structs may be made to
coincide with C structs more closely. All ptr or full
fixnum fields will be aligned on a multiple of the size
of ptr however for speed of reference and portability.
Needless to say recompilation is necessary for most files.
System files affected gbc.c, defstruct.c, predicate.c, print.d
defstruct.lsp cmpfun.lsp cmputil.lsp cmpeval.lsp predlib.lsp
and maybe some other small changes.
Naturally significant speedups will be gained if one can
keep integers in raw form:
(defstruct ja1 (a #\a :type character)(b 0 :type fixnum))
(defun joe (x n) (sloop::sloop for i below n do (setf (ja1-b x) i)))
Then (joe (make-ja1) 1000000) takes less than 1 second now,
as opposed to over 50 seconds prior to these changes.
Wed Apr 26 11:59:43 1989 Bill Schelter (wfs at rascal)
* Add message feature for AKCL start up. If unixport/message
exists and unixport/message-suppress does not, then
the file unixport/message will be printed on start up. (top.lsp)
* make o/akcllib.a, a library containing bcmp .. and other
common functions which are not present in all versions of unix.
The main link puts this after -lc, so that faster implementation
dependent versions will be used if they exist.
Mon Apr 24 20:28:49 1989 Bill Schelter (wfs at rascal)
* Change to eliminate char_table.s assembler code.
This change will unfortunately require users to recompile
their object code for use with this system.
* Fix bug in cmpinline.lsp, which was allowing
(rplacd (cons a b) nil) to give the wrong code.
* Add timing for gc, and allow have the si::*notify-gbc*
flag cause printing the type of gc.
To time gc's. do (si::gbc-time 0), to set the timer to
0, (si::gbc-time -1) to reset and turn it off. It
returns an integer in internal time units, similar
to get-internal-run-time.
* Speed up the lisp reader and intern. On a large file the
read is 2.4 times faster than it was.
* Many small changes to the include files, to eliminate duplicate
definitions of symbols, (not allowed by some compilers)
and also adding COMM_LENG, which can be null for most compilers
but should be a small integer for the IBM c compiler (it
does not accept external declarations int foo[], so we do
foo[COMM_LENG];
Fri Apr 21 18:13:23 1989 Bill Schelter (wfs at rascal)
* Fix bug with the new fast read-byte,....
If the stream argument was supplied, and was not a stream
but rather T or NIL, there was a problem. If you declare
the arg to be of type stream, then you will get identical
code to before; otherwise a typecheck for type stream
will be supplied, branching into the slower code for
non streams.
Files cmpopt.lsp and cmptype.lsp
Tue Apr 18 22:56:54 1989 Bill Schelter (wfs at rascal)
* More changes to read-byte, write-byte, read-char and
write-char. I have removed the :in-file and :out-file
declarations, and the speed up for the undeclared streams
should be virtually the same as obtained with the declarations.
This will affect mainly file streams. Files cmpfun.lsp,
cmpopt.lsp, cmpinclude.h, read.d
Note that on read-byte it is still advantageous to use an
eof which is a fixnum.
reads where the eof-error-p arg is not nil are not speeded up.
It is still the case that two way streams are not speeded up,
however if (si::fp-input-stream str) returns non nil, then
you can use the resulting stream for fast input.
Thu Apr 13 00:20:11 1989 Bill Schelter (wfs at rascal)
* fixes to subtypep (as reported by riley)
* changes to allow faster operation of read-char,write-char,
read-byte and write-byte, when operating on file streams.
In order to use these you should declare the stream
to be :in-file or :out-file.
You can use (typep str :in-file) to
check if it is really valid.
(defun myread (str)
(declare (:in-file str))
(the fixnum (read-byte str nil -100)))
The above changes should work for ansi C style stdio,
in particular for unix io. At the moment it is conditionalized
for +unix.
files: cmpfun.lsp, cmpopt.lsp, cmtype.lsp, file.d, cmpenv.lsp
The difference in reading speed is substantial: eg 3 microseconds
as opposed to 60 microseconds for read-byte.
To make your code portable do
(proclaim '(declaration :in-file))
In order for the optimizations to cut in,
the read functions must be supplied with 3 args and the
write functions with 2. The second of the read args
must be nil.
Note with read-byte, using an eof value which is a fixnum,
allows you to declare the location you will pass to as a fixnum.
Wed Mar 22 17:19:03 1989 Bill Schelter (wfs at rascal)
* Change c1fmla-constant in cmpif.lsp.
(if (null 2) x y ) was yielding x not y!
Sun Mar 19 12:26:21 1989 Bill Schelter (wfs at rascal)
* fix stream_at_end, so that a stream opened for `io'
does the check, when reading, so that read can
return eof properly.
Sun Mar 12 01:40:53 1989 Bill Schelter (wfs at rascal)
* speed up the intern of symbol in pack_hash
computation.
* psetf bug in case of two args (psetf a 3)
not passing the environment. setf.lsp
Fri Feb 24 22:27:22 1989 Bill Schelter (wfs at rascal)
* add fixes for &environment to allow it to come
anywhere in the lambda list. defmacro.lsp and cmplam
(change from R. Harris).
Wed Feb 22 17:24:30 1989 Bill Schelter (wfs at rascal)
* fix tree-equal to save the previous test in list.d
(assoc '(c) '((a b) ((c) d)) :test #'tree-equal) failed.
(by Cooperman)
Sun Feb 19 17:42:31 1989 Bill Schelter (wfs at rascal)
* Fix hash_equal in hash.d. It was broken for circular
structures or lists. This would affect sxhash as well
as equal hash tables.
* Fix obscure bug in compiler in cmptop.lsp, which could
possibly leave out a sup declaration. The c compiler
would catch this.
* Fix comparison of arrays under equalp in c/predicate.c
It had been broken for rank different from 1.
* Fix the assembler for the sun4 in sun4_chtab.s for
multiply. (bug report by Harris).
Fri Dec 9 00:47:46 1988 Bill Schelter (wfs at rascal)
* Made changes to sfasl to make it more portable.
* added structures and characters to the types
handled by fasdump.c
* changed the c stack check in gbc.c to use cs_check.
Sun Nov 27 12:31:13 1988 Bill Schelter (wfs at rascal)
* Add si::fp-input-stream, si::fp-output-stream,
Which take one arg a stream, and return a stream
with an strm->sm.sm_fp slot suitable for use with fread and fwrite.
If this is not possible nil is returned.
Wed Nov 2 16:09:17 1988 Bill Schelter (wfs at rascal)
* inline-args fix:
(defun x (c s i)
(declare (optimize (safety 2)))
(declare (fixnum i))
(setf (char s (setq i (1+ i))) c))
made c, not just i a fixnum in the char compilation,
fixed in inline-args (by E. Wang) edward@ucbarpa.Berkeley.EDU
file cmpinline.lsp
Wed Oct 12 17:01:06 1988 Bill Schelter (wfs at rascal)
* Added new array types: files cmptype,array.c,typespec.c,cmpopt,
predlib.lsp, gbc.c, and maybe some others.
Purpose of the change was to allow programs like CLX which
use lots of numerical arrays, to be much more economical.
Also make-array now coerces the element-type in a reasonable
way, and the same handling is used in the compiler.
New array element types:
signed-char, unsigned-char, signed-short, unsigned-short
The ranges on a SUN are
((INTEGER -128 127) (INTEGER 0 255) (INTEGER -32768 32767)
(INTEGER 0 65535)) respectively.
Note that now
make-array will always try to find the `best' array to
accommodate the element-type specified. For example on a SUN
(mod 1) --> bit
(integer 0 10) --> unsigned-char
(integer -3 10) --> signed-char
si::best-array-element-type is the function doing this. It
is also used by the compiler, for coercing array element types.
If you are going to declare an array you should use the same
element type as was used in making it. eg
(setq my-array (make-array 4 :element-type '(integer 0 10)))
(the (array (integer 0 10)) my-array)
.. When wanting to optimize you need to make a reference:
(the fixnum (aref (the (array (integer -3 10)) ar) (the fixnum i)))
if ar were constructed using the (integer -3 10) element-type.
You could of course used signed-char, but since the ranges
may be implementation dependent it is better to use -3 10 range.
make-array needs to do some calculation with the element-type
if you don't provide a primitive data-type. One way of doing
this in a machine independent fashion:
(defvar *my-elt-type* #. (array-element-type (make-array 1
:element-type '(integer -3 10))))
Then calls to (make-array n :element-type *my-elt-type*)
will not have to go through a type inclusion computation.
Tue Oct 11 09:52:13 1988 Bill Schelter (wfs at rascal)
* When using gcc, it could happen that there was
a string in in the init function, which got placed
before the init function. Had to add -fwritable-strings
to stop this.
Sun Oct 2 13:35:11 1988 Bill Schelter (wfs at rascal)
* Fix MP386 and att port bugs, introduced with the
new mp386.h, att.h and mp386.defs files.
* Fix unixtime at least for bsd, so that get-internal-real-time
does not wrap every few hours, but will not wrap below 400 days.
* Fix directory problems with compiling in other than the current
directory, in the name of the file specified in the include.
The command on most systems cd 's to the directory to run the
cc, and so the .h file needed to use an unprefixed name.
Sat Sep 24 16:02:22 1988 Bill Schelter (wfs at rascal)
* Changes to the computation of double-float-epsilon
and more generally XX-epsilon. First they did not
satisfy the condition of
(defun fo (e) (not (= (float 1 e) (+ (float 1 e) e))))
(fo double-float-epsilon) --> t
as per CLtL. As it stands they at least satisfy the
test, although there may be floats slightly smaller which
also do.
Second on some machines (eg HP) the calculation done
in line, carried more precision than that which would
be normal when passed through eql, so that we changed
the test == in constructing the double float epsilon
to use a function call. Otherwise the epsilon was
~10^-20 instead of the correct ~10^-16.
Thu Sep 22 14:08:07 1988 Bill Schelter (wfs at rascal)
* Fixed hash_equal not to use the cast to int,
in computing hash of a symbol name
since this made sun4's unhappy (file hash.d ).
Wed Sep 14 12:02:35 1988 Bill Schelter (wfs at rascal)
* fix to setf to make sure that
(defmacro joe (x) `(progn t (car ,x)))
(defsetf joe rplaca)
work correctly. files changed are assignment.c and setf.lsp
The defsetf'd definition takes precedence in the macroexpansion.
The bad order was introduced when the evaluation of the macros including
their environments was introduced (see below).
Mon Sep 12 10:09:56 1988 Bill Schelter (wfs at rascal)
* use varargs.h for bind.c and list.d where
variable length args are passed. We only use
va_arg(ap,object) to access the next arg now.
There are no more indexed references, since that
is less portable.
* A major change for porting: Each machine type
now has a .h file of its own, and things like alloc.c,
main.c, unixsave.c no longer should be modified for
individual machines. The file config.h is a link
from `your-machine`.h and things like the VSSIZE,
are also specified in that file. There are also
files such as bsd.h , att.h (for system V) which
can be included by the various special machine files.
* c support for extended_div was added, and will
be used if USE_C_EXTENDED_DIV is defined. The only
function which does not have a C definition on the sun4
version is extended_mul. Note that the c version
of extended_div is included in big.c rather than earith.c
since the latter is not compiled with the optimize switch.
* The assembler functions of bitop.c had been being redefined
by macros in gbc.c in an earlier change (this was actually
4 times faster on a sun3), and now those functions have
dummy definitions, which will go away soon.
Wed Aug 31 23:38:08 1988 Bill Schelter (wfs at rascal)
* use varargs for the funlink of proclaimed functions,
This affects cmpcall.lsp, funlink.c and requires
<varargs.h> to be included by cmpinclude.
The old variable arg business was not portable to risc
architectures, where args are passed in registers.
* make print return a value in print.d
Hope there are not more `implicit returns' hidden away.
Fri Aug 19 15:13:06 1988 Bill Schelter (wfs at rascal)
* Make (subtypep 'string-char 'character) --> t
Wed Aug 3 11:33:20 1988 Bill Schelter (wfs at rascal)
* I have removed the dynamic growing of the special binding
stack which I had added a few months ago (cf bds.h,main.c,bds.c.)
It had ignored
the fact that some functions eg, Levalhook grab a pointer
into the bds rather than just an index. These could easily have
been changed, but until we allow the other stacks to grow
dynamically, it is of questionable value.
Fri Jul 1 09:41:36 1988 Bill Schelter (wfs at rascal)
* allow assoc to take key in c/list.d as per mail
442 23-Jun pc%linus@mitre-bedford.ar KCL Bug: assoc doesn't take :key
Wed Jun 15 16:46:40 1988 Bill Schelter (wfs at rascal)
* Added support for switch construct (see cmpnew/cmptag.lsp
for documentation)
This will allow compiling into the c switch construct
if (the test variable is declared to be a subtype of
fixnum) which can then allow much faster switching on
cases. To do:Should optimize some of case constructs into
switch, where applicable.
Wed Jun 8 11:43:00 1988 Bill Schelter (wfs at rascal)
* Altered the marking of the c stack in gbc.c.
Now the current location for the c stack is taken
in a separate function from the one where the environment
is forced onto the stack, so that we don't have to
depend on the c compilers doing things in the expected
order. See mark_stack_carefully. Also added a flag
C_GC_OFFSET which if defined to 2, will mark
the stack twice once on 4m and once on 4m+2.
* changed the initialization in main.c of bds_org
and bds_limit.
This is to fit in with the earlier change to
bds_overflow, to allow the bds to grow. bds_org
is now a pointer rather than a hardwired value.
Tue Jun 7 04:44:15 1988 Bill Schelter (wfs at rascal)
* Fix equalp to use the fill-pointer as a limit
when comparing strings,vectors and bitvectors.
Mon Jun 6 11:29:08 1988 Bill Schelter (wfs at rascal)
* Fix rotatef in setf.lsp to return NIL per CLtL
* Fix handling of &rest arg in cmplam.lsp
c2lambda-expr-without-key as per bug fix of Yuasa
12-Nov yuasa%kurims.kurims.kyoto A bug fix
* Add the fix to quick-sort as per
67 12-Jan yuasa%tutics.tut.junet%ut Re: sort bug
* Add fix for shadowing-import per
68 12-Jan yuasa%tutics.tut.junet%ut Re: SHADOWING-IMPORT doesn't
* Fix (random 0) bug, adding TSpositive_number type. as per
69 12-Jan yuasa%.. Re: (random 0) => losing error message
* Fix (make-array 2 0) bug per
70 13-Jan yuasa%tutics.tut.junet%ut Re: Nil is a sequence
* Add bogus value (*READ-DEFAULT-FLOAT-FORMAT* t) when
printing a float in a compiled file, so that it will always
add the type in the printing, and we won't get a float
type different from that used at compile time
cf. 80 28-Jan yuasa%tutics.tut.junet%ut Re: float numbers
* Add fix for bug
setf (aref x (decf i)) (aref x 1)) ->
"base[2]= aset1(base[0],V1,fix(base[1]));"
to cmpinline.lsp, as per
92 8-Feb yuasa%tutics.tut.junet%ut Re: compiler bug?
*All fixes by yuasa in the kcl-mail-archive thru Jun 88
are now incorporated in akcl
* Altered cmp-macroexpand-1 to use the local macro environment
Wed Jun 1 13:29:57 1988 Bill Schelter (wfs at rascal)
* Altered macros.c to pass the function-macro environment
if it is non nil. This is needed so that macro expansion
functions can be called in the correct environment.
Altered cmputil.lsp so that cmp-macroexpand and cmp-macro-expand
both use the current compiler macro environment.
Finally altered setf, so that it looks at the macro
environment when it expands the place.
Deleted treatment of setf as special form by the compiler,
(introduced yesterday!)
now that macros are handled correctly, and it is defined
correctly as a macro.
Things like
(macrolet ((ab nil 'a)) (setf (the fixnum (ab)) 3))
Now work to give (set 'a (the fixnum 3)) as expected.
Mon May 30 11:40:47 1988 Bill Schelter (wfs at rascal)
* Altered setf to macroexpand the place in the current
lexical environment so that
(macrolet ((joe nil 'bil))(setf (joe) n))
would behave correctly. This change was in assignment.c
in `setf' and also in cmputils.lsp adding c1setf, to
get analagous treatment for compiler.
To do:
It is still not totally correct: I don't handle
(macrolet ((joe nil 'bil))(setf (the fixnum (joe)) n))
but
(macrolet ((joe nil '(the fixnum bil)))(setf (joe) n))
is ok.
Tue May 24 09:38:58 1988 Bill Schelter (wfs at rascal)
* Changed float to return single-float if given only one
arg, as per CLtL.
* Added a dynamic growth feature the bds stack, changing
bds_org to be a variable rather than a macro, and altering
bds_overflow.
*Enforce substantial constraints on downward closures ,to be relaxed
at a future date: Currently no args, and no cross references
to other types of closures. This is done in check-downward
in t1defun.
|