Commit Graph

1830 Commits

Author SHA1 Message Date
Irmen de Jong
b292124f3c replaced many short/int values by unsigned types if appropriate 2021-11-21 00:55:56 +01:00
Irmen de Jong
c0035ba1a2 char encodings now use UByte type instead of short 2021-11-21 00:07:17 +01:00
Irmen de Jong
107935ed31 add some more const folding patterns 2021-11-20 22:47:49 +01:00
Irmen de Jong
adfaddbcf4 give a nicer error when given a wrong compilation target. 2021-11-20 18:30:55 +01:00
Irmen de Jong
74db5c6be7 fix referencesIdentifier() and better removal of unnecessary assignments 2021-11-20 17:41:41 +01:00
Irmen de Jong
f9399bcce7 r=(q+r)-c and r=q+(r-c) are now both also 'augmentable', and BinExprSplitter doesn't check for associativeOperator anymore 2021-11-20 02:03:32 +01:00
Irmen de Jong
87600b23db fix constvalue parent linkage for prefix and typecast 2021-11-20 00:20:35 +01:00
Irmen de Jong
cedfb17b18 fix too aggressive removal of vars that weren't completely unused 2021-11-19 22:49:35 +01:00
Irmen de Jong
fa4c83df6b added 3 tests for discovered problems 2021-11-18 23:55:20 +01:00
Irmen de Jong
42c8720e8b fix float rounding tests 2021-11-18 22:54:49 +01:00
Irmen de Jong
b334d89715 refactor and fix the way memory addresses are checked to be in IO space or regular ram 2021-11-18 22:47:58 +01:00
Irmen de Jong
4f5d36a84d optimization added: bitwise operations with a negative constant number -> replace the number by its positive 2 complement 2021-11-18 02:51:42 +01:00
Irmen de Jong
fa11a6e18b removed faulty and too aggressive assembly optimization of double-store 2021-11-18 01:43:22 +01:00
Irmen de Jong
4c82af36e6 fix improperly changed behavior about =0 initializer 2021-11-18 00:17:22 +01:00
Irmen de Jong
dafa0d9138 fix compiler crash bug due to reused ast expression nodes. Now all (relevant) Nodes have a copy() function to make a clone. 2021-11-17 23:05:59 +01:00
Irmen de Jong
2e0450d7ed fix bug where variable=0 initializer was forgotten if vardecl is followed by an augmented assignment 2021-11-17 22:31:43 +01:00
Irmen de Jong
f48d6ca9f8 simplified NumericLiteral to always just contain a Double instead of a Number for the value 2021-11-16 23:52:54 +01:00
Irmen de Jong
1f60a2d8b9 comments 2021-11-15 01:30:12 +01:00
Irmen de Jong
5fd83f2757 version 7.3 2021-11-14 22:55:13 +01:00
Irmen de Jong
c80df4140b until-loop condition now also simplified to avoid stack-eval 2021-11-14 22:51:02 +01:00
Irmen de Jong
53e1729e2f introduce option to use internal scratch variables via prog8_lib definitions (ony for compiler, not for user code!) 2021-11-14 16:01:54 +01:00
Irmen de Jong
ab2d1122a9 conditional expressions are optimized more intelligently (simple ones are not split off in separate assignments) 2021-11-14 12:38:56 +01:00
Irmen de Jong
5190594c8a added several more assembly-level optimizations to remove redundant instructions 2021-11-14 12:23:46 +01:00
Irmen de Jong
c858ceeb58 compiler shouldn't use cx16.r15 as temp var 2021-11-14 02:38:59 +01:00
Irmen de Jong
f0f52b9166 optimize typecasted binary expression to avoid even more estack use. also fix wrong parent crash in removal of unused variable's assignments. 2021-11-13 14:22:37 +01:00
Irmen de Jong
00c6f74481 tweak temp float 2021-11-13 12:56:59 +01:00
Irmen de Jong
2177ba0ed2 added signed versions of the cx16 virtual registers 2021-11-13 02:42:21 +01:00
Irmen de Jong
3483515346 preparing for more optimizations 2021-11-12 23:23:51 +01:00
Irmen de Jong
75a06d2a40 preparing for more optimizations 2021-11-12 02:17:37 +01:00
Irmen de Jong
53ac11983b better unused variable removal 2021-11-11 03:03:21 +01:00
Irmen de Jong
69f4a4d4f8 tweak expr.typecastTo() a bit 2021-11-11 00:15:09 +01:00
Irmen de Jong
222bcb808f optimize load-store-load combo in output asm 2021-11-10 23:47:35 +01:00
Irmen de Jong
8df3da11e3 add cosr8, sinr8, cosr16 and sinr16 builtin functions that take a degree 0..179 (= 0..358 in 2 degree steps)
to more easily scale halves/quarters etc of a circle than possible with the ones that take 0..255 'degrees'.
2021-11-09 23:39:26 +01:00
Irmen de Jong
84dafda0e4 fix error message for type mismatch on builtin-function parameter 2021-11-09 22:19:07 +01:00
Irmen de Jong
b909facfe5 fix compiler stackoverflow crash on certain typecasted expressions containing floats. 2021-11-09 19:31:19 +01:00
Irmen de Jong
7780d94de1 discovered crash related to float typecasting in asm assignment codegen 2021-11-09 03:45:07 +01:00
Irmen de Jong
4937e004b5 fix compiler crash where it used wrong datatype in split assignment
fixes crash for "ubyte bb ;; uword ww ;; bb = not bb or not ww"
2021-11-09 01:13:23 +01:00
Irmen de Jong
857724c7e6 attempt to make if-statement not use stack eval anymore 2021-11-08 19:07:36 +01:00
Irmen de Jong
2d1e5bbc7e remove unimportant empty tests 2021-11-08 17:00:10 +01:00
Irmen de Jong
7961a09d16 converting compiler module's testcases to kotest assertions 2021-11-08 16:14:22 +01:00
Irmen de Jong
613efcacc7 converting compiler module's testcases to kotest (ongoing) 2021-11-08 15:08:48 +01:00
Irmen de Jong
7e8db16e18 moved to kotest assertions in compilerAst module tests 2021-11-07 21:18:18 +01:00
Irmen de Jong
f4c4ee78d9 re-use global returnvalue temp var instead of duplicating it in every subroutine that needs it 2021-11-07 14:19:21 +01:00
Irmen de Jong
72a7e61fd0 version 7.2 2021-11-06 23:42:13 +01:00
Irmen de Jong
f40620aa25 "not x" as a condition (if, while, until) is optimized into "x==0", this avoids calculating the value 2021-11-06 23:25:32 +01:00
Irmen de Jong
4270c04856 don't crash but give proper error on "-X" expression where X is not a signed type 2021-11-06 18:06:01 +01:00
Irmen de Jong
62dc824bc0 tweaks 2021-11-06 17:14:07 +01:00
Irmen de Jong
1605791f1b float swap() no longer uses evaluation stack but a single temp var instead + FAC1 2021-11-06 03:36:14 +01:00
Irmen de Jong
23961f695d fixed some parse tree node position end-columns. cleanup some todo's 2021-11-05 22:48:28 +01:00
Irmen de Jong
eea3fb48a8 add command line option 'optfloatx' to explicitly re-enable float expr optimization as this can increase code size significantly.
The output size of the various example programs using floating point, when not using this optimization, has been reduced significantly.
The resulting code runs a (tiny) bit slower though.
2021-11-03 22:52:08 +01:00