July 2004 - Cvs-nhc98

cvs commit: nhc98/src/runtime/Kernel Makefile
by Sven Panne 04 Jul '04

04 Jul '04

panne 2004/07/04 06:25:38 PDT Modified files: src/runtime/Kernel Makefile Log: Don't compile code which depends on adventurous storage layout assumptions with optimizations, otherwise gcc will legitimately move the code around. NOTE: This is really only a very fragile workaround and may break again anytime soon. Either nhc98's bytecode interpreter should be changed to not rely on the assumption that certain blocks are adjacent (which will probably impose some performance hit), or the mangler should be made more robust. In the latter case, sed will probably not be enough, a strategy would be: Mark the different parts of the bytecode somehow and re-arrange the assembler code in a suitable way. => Perl! :-P Somehow nhc98's mangler and the one from GHC seem to converge... Anyway, with all these changes in the last few days, nhc98 can successfully be bootstrapped to a stage-3 compiler again: GHC => nhc98 => nhc98 Revision Changes Path 1.18 +4 -3 nhc98/src/runtime/Kernel/Makefile

1 0

cvs commit: nhc98/src/compiler98 Flags.hs
by Sven Panne 04 Jul '04

04 Jul '04

panne 2004/07/04 06:17:21 PDT Modified files: src/compiler98 Flags.hs Log: Commented out flags -stg, -bcbefore, -stgcode, and -iparse, because they seem to be dead. If they really are, we should just delete them, otherwise resurrect them. Malcolm? Revision Changes Path 1.22 +8 -8 nhc98/src/compiler98/Flags.hs

1 0

cvs commit: nhc98/docs compiler-options.html
by Sven Panne 04 Jul '04

04 Jul '04

panne 2004/07/04 01:18:49 PDT Modified files: docs compiler-options.html Log: * Synched the order/amount of compiler stages with reality * Pacified W3C validator Revision Changes Path 1.14 +539 -435 nhc98/docs/compiler-options.html

1 0

cvs commit: nhc98 configure
by Sven Panne 03 Jul '04

03 Jul '04

panne 2004/07/03 05:04:03 PDT Modified files: . configure Log: Enable the "more evil" manger (for gcc 3.3.x) for installation, too Revision Changes Path 1.108 +2 -0 nhc98/configure

1 0

cvs commit: nhc98/src/runtime/Kernel collector.c ffiexport.c haskellInit.c mark.c mutator.c profile.c remark.c tprof.c xmem.c
by Sven Panne 03 Jul '04

03 Jul '04

panne 2004/07/03 04:52:24 PDT Modified files: src/runtime/Kernel collector.c ffiexport.c haskellInit.c mark.c mutator.c profile.c remark.c tprof.c xmem.c Log: More warning police... Revision Changes Path 1.21 +5 -14 nhc98/src/runtime/Kernel/collector.c 1.4 +2 -1 nhc98/src/runtime/Kernel/ffiexport.c 1.22 +1 -0 nhc98/src/runtime/Kernel/haskellInit.c 1.9 +3 -2 nhc98/src/runtime/Kernel/mark.c 1.17 +3 -3 nhc98/src/runtime/Kernel/mutator.c 1.4 +2 -2 nhc98/src/runtime/Kernel/profile.c 1.4 +3 -7 nhc98/src/runtime/Kernel/remark.c 1.6 +8 -17 nhc98/src/runtime/Kernel/tprof.c 1.2 +1 -0 nhc98/src/runtime/Kernel/xmem.c

1 0

cvs commit: nhc98 configure
by Sven Panne 03 Jul '04

03 Jul '04

panne 2004/07/03 02:25:38 PDT Modified files: . configure Log: Fixed gcc 3.3.3 detection magic Revision Changes Path 1.107 +2 -2 nhc98/configure

1 0

cvs commit: nhc98/src/runtime/Kernel newtables.c
by Sven Panne 02 Jul '04

02 Jul '04

panne 2004/07/02 12:00:30 PDT Modified files: src/runtime/Kernel newtables.c Log: OK, this seems to be the end of nhc98's cunning memory layout: gcc 3.3.3 legitimately reshuffles the arrangement of the arrays (alphabetically?), so we must not make any assumptions about their order. This is probably only the beginning of the whole story... NOTE: We should really abstract from the magic constants 10 (for ints) and 1 (for chars), but I was too lazy to do that now. Revision Changes Path 1.5 +6 -8 nhc98/src/runtime/Kernel/newtables.c

1 0

cvs commit: nhc98/include mutlib.h
by Sven Panne 02 Jul '04

02 Jul '04

panne 2004/07/02 12:00:29 PDT Modified files: include mutlib.h Log: OK, this seems to be the end of nhc98's cunning memory layout: gcc 3.3.3 legitimately reshuffles the arrangement of the arrays (alphabetically?), so we must not make any assumptions about their order. This is probably only the beginning of the whole story... NOTE: We should really abstract from the magic constants 10 (for ints) and 1 (for chars), but I was too lazy to do that now. Revision Changes Path 1.7 +2 -2 nhc98/include/mutlib.h

1 0

cvs commit: nhc98/src/runtime/Integer alloca.c
by Sven Panne 02 Jul '04

02 Jul '04

panne 2004/07/02 11:51:20 PDT Modified files: src/runtime/Integer alloca.c Log: Warning police Revision Changes Path 1.3 +1 -0 nhc98/src/runtime/Integer/alloca.c

1 0

cvs commit: nhc98/src/runtime/Kernel newtables.c
by Sven Panne 02 Jul '04

02 Jul '04

panne 2004/07/02 11:43:12 PDT Modified files: src/runtime/Kernel newtables.c Log: Use correct type from header for arrays Revision Changes Path 1.4 +5 -4 nhc98/src/runtime/Kernel/newtables.c

1 0