dd1634e755
* $fixnum->flonum now works.
2007-06-18 14:29:39 +03:00
e51def62c1
* Fixed bugs in the just-added $fp comparison primitives.
2007-06-15 12:38:30 +03:00
3f3767800d
* Added $fl=, $fl<, $fl>, $fl<=, $fl>= primops.
2007-06-15 12:19:28 +03:00
3acdcb6c04
* Added native support for $fl+, $fl-, $fl*, $fl/
2007-06-15 08:53:34 +03:00
3798fd932a
* Added ratnum cases to binary*.
2007-06-08 10:18:36 +03:00
f3e5772e76
* Added bref to assembler
...
* Fixed bug in bytevector-s8-ref
2007-06-08 08:54:10 +03:00
a27c6e13a9
* fixed a HUGE bug in optimize-letrec.
...
* fixed two minor bugs in the code generators for - and +
* added a constant-folder for "length"
2007-06-06 10:14:07 +03:00
ae0b25ae9f
* do-spill now removes spilled vars from neighbors of conflicting
...
vars.
2007-06-06 09:24:32 +03:00
0db57fbb69
* removed the reg? argument to build-graph. Reg? was needed when
...
build-graph was used for constructing both frame conflicts and
register conflicts. Now, it's only used for register conflicts so
reg? is no longer needed.
2007-06-06 06:10:28 +03:00
39e6fc2fcf
* removed more junk.
2007-06-06 03:11:12 +03:00
665f3a0b79
* removed old one-pass compiler. chaitin is now the default
...
backend
2007-06-06 02:57:48 +03:00
7301be0ab0
* removed one instance of build-graph which reduces compile time
...
from 225 secs down to 17! Sweet!
2007-06-06 02:32:45 +03:00
feb9764645
* Cleanup of some junk code in color-by-chaitin
2007-06-06 01:16:34 +03:00
1dc9e83448
* Altcogen works. Need to rethink how to code the conflict graphs.
2007-06-04 02:55:04 +03:00
a5dbb8f45d
* Can load altcogen.
2007-06-03 04:55:40 +03:00
ca248be49a
* Libraries can export modules now.
...
* libaltcogen can now be loaded.
2007-06-02 10:21:05 +03:00
d2657bde24
* Added basic identifier-syntax (one that does not recognize the
...
set! case).
2007-06-02 05:43:24 +03:00
705e8f386b
* fixed error in check-dups of modules in internal defines.
2007-05-01 02:19:05 -04:00
f27c2df625
* assign-frame-sizes is as complete as could be (time for assigning
...
libcompile.ss dropped from ~5 secs to 1.8 secs).
2007-03-11 20:55:43 -04:00
068bc38e42
* eliminated a few instances of set-rem and set->list.
2007-03-11 20:39:21 -04:00
2b6bcc324b
* when a variable is assigned a location, it does not need to remove
...
itself from the edges of its conflicting nodes.
2007-03-11 20:07:00 -04:00
c473619084
* important realization that set->list is perhaps the most expensive
...
operation on sets. Working on eliminating most uses of set->list.
2007-03-11 20:00:08 -04:00
2b5b555bac
* Just realized that you don't need all var<->var conflicts to be
...
recorded. All is needed is for a var $x$ to know some of its var
conflicts as long as the other conflicts know about $x$. I'll be
working on this now. I'm hoping for a 50% speedup in the
assign-frame-sizes pass.
2007-03-11 18:54:15 -04:00
aa1c7e1bb7
* before changing number of bits to 28
2007-03-11 03:40:47 -04:00
3cb18c56e9
* deleting old list-based set representation.
2007-03-10 21:01:41 -05:00
fa94d2f6a1
* added a few missing instruction sequences where the dest is a
...
memory location.
* implemented integer sets (based on tree representation).
2007-03-10 19:50:24 -05:00
10bdc3dcd6
* starting with IntegerSet
2007-03-10 16:47:13 -05:00
d4559cfd4c
cleanup of repository content.
2007-03-09 19:28:38 -05:00
b326fbc07a
* completed the move to new specify-representation pass infrastructure.
2007-03-03 23:17:04 -05:00
e9740fa34f
* fixed bug in the graph construction that caused the operand of
...
idivl not to be added to the live set.
2007-03-02 00:41:28 -05:00
de7c43a16b
* optimized symbol calls using the symbol-function field.
...
* added error checks for applying nonprocedures.
2007-02-25 21:29:28 -05:00
87d8d5a5dd
* added a $set-symbol-function! primitive.
2007-02-24 15:42:57 -05:00
d99c22e8c4
* The shortcut form is now two parts: a body and a handler. The
...
evaluation of the body can jump to the handler at any point during
evaluation by means of using the interrupt primitive or by
overflow.
2007-02-22 23:02:50 -05:00
819a3ca1c1
* bug in collector causing dirty bits for code objects with younger reloc
...
vectors to be unset incorrectly. The bug was in using || instead
of | when computing the dirty pattern.
2007-02-22 21:58:38 -05:00
0bd88991cb
* looks not bad.
2007-02-19 23:33:29 -05:00
30f71b0381
* frame moves are not working as nicely as they should.
2007-02-19 18:21:35 -05:00
b6dd620b94
* chaitin can now compile the system.
2007-02-17 19:22:14 -05:00
0304c85082
* pre-spilling call-live variables works for the test cases now.
2007-02-17 18:09:03 -05:00
85bf359150
* psyntax is not working under chaitin.
2007-02-16 10:11:21 -05:00
649e7f022a
* libchezio and libhash are remaining
2007-02-15 23:54:39 -05:00
1a4cdcb7b0
* libcore now compiles with chaitin
2007-02-14 19:42:36 -05:00
434ebe9525
* The $apply macro and all the appcall forms were removed from the
...
compiler after realizing that I was stupid.
($appcall rator rand*) was just (jmpcal SL_apply rator rand*)
2007-02-14 15:50:34 -05:00
8294a8dee9
* fixed an assembler bug in "addl reg -> mem"
2007-02-13 17:24:00 -05:00
297e47db32
* adding some asm tests in asm-tests.ss
2007-02-13 05:08:48 -05:00
a123a77bbc
* nothing much
2007-02-13 02:05:58 -05:00
826adfe9dd
* can now compile libtoplevel
2007-02-12 23:19:56 -05:00
f88e3a8b65
* libnumeric can be compiled now with chaitin
2007-02-12 23:03:41 -05:00
2e477881b6
* split libcontrol into primitive libcontrol0 and extended
...
libcontrol1.
2007-02-12 19:17:31 -05:00
fad2c4e999
* fixed bug in assembler
...
* assembler now handles orl imm, mem
2007-02-12 17:59:58 -05:00
f766ca1148
* passes tests, but fails to compile psyntax.
2007-02-12 13:58:04 -05:00
Abdulaziz Ghuloum