|
|
0694a187d7
|
unsigned>0 now optimized into unsigned!=0
|
2022-12-12 20:37:57 +01:00 |
|
|
|
f2273c0acc
|
fix several FP rom routine addresses on cx16.
|
2022-12-03 19:56:54 +01:00 |
|
|
|
df1793efbf
|
fixed: word << 12 is suddenly an uword (with optimizer on)
|
2022-11-15 03:00:41 +01:00 |
|
|
|
8f3aaf77a1
|
fix optimizer hanging on uword xx :: xx >>= 8 / xx=msb(xx)
|
2022-11-15 01:40:13 +01:00 |
|
|
|
562d722ad5
|
codegen: added missing codegen for float array inplace modification
|
2022-11-03 20:08:46 +01:00 |
|
|
|
144c1ba3a6
|
ir: fix float instruction value in formatspec
|
2022-11-03 19:08:38 +01:00 |
|
|
|
5b35232ab4
|
fix "fpReg1 out of bounds" crash for vm target for in-place float array assignment. #85
|
2022-10-29 17:04:39 +02:00 |
|
|
|
6830e15b4e
|
print warning when bit shifts are too large and result in 0. #83
|
2022-10-29 15:23:39 +02:00 |
|
|
|
af6be44676
|
ir: adding register usage inspections
fix compiler problems with untrimmed inlined asm, and when only a single return statement is present in a subroutine
|
2022-09-30 20:25:00 +02:00 |
|
|
|
101b33c381
|
split intermediate representation into separate module
|
2022-09-13 23:06:05 +02:00 |
|
|
|
18790d867c
|
optimize conditional expression WORD & $ff00 to just msb(WORD)&$ff
|
2022-08-12 00:21:44 +02:00 |
|
|
|
4d840c7db8
|
optimized mkword(0, X)
|
2022-08-11 22:51:09 +02:00 |
|
|
|
4d2b21816d
|
optimized uword <<8 and >>8
|
2022-08-11 22:25:15 +02:00 |
|
|
|
68abda1219
|
fix a few small compiler errors (removing functioncall, removing block, assigning virtual register return value)
|
2022-08-09 23:38:29 +02:00 |
|
|
|
93ce74eeb1
|
removed problematic expression "simplifications" (that introduced arbitrary r9 temp register usage)
|
2022-08-07 12:26:11 +02:00 |
|
|
|
4644c9b621
|
got rid of GoSub ast node and codegen complexity related to that.
sometimes programs get smaller, sometimes bigger.
|
2022-08-07 03:24:20 +02:00 |
|
|
|
046dceb5c2
|
added optimized case for signed division by 2
|
2022-07-24 13:59:35 +02:00 |
|
|
|
dcc1f00048
|
fix rounding errors in signed divide by power-of-two
The optimized bit-shifting division is removed (for now)
|
2022-07-24 12:34:55 +02:00 |
|
|
|
35af53828a
|
fix endless loop in optimizer, fix cx16 register clobbering in psg interrupt handler, fix crash on certain arrays, fix undefined symbol when it's in another imported module
|
2022-07-13 18:42:06 +02:00 |
|
|
|
68da661edc
|
optimize comparison to true/1 into comparison to zero, optimize while/until conditions
|
2022-07-11 16:42:52 +02:00 |
|
|
|
bdb7de34be
|
added several compiler checks against weird boolean type use in expressions
|
2022-07-07 22:23:56 +02:00 |
|
|
|
41f4e22a17
|
introduce BOOL type
|
2022-07-07 22:23:56 +02:00 |
|
|
|
7fff4f249d
|
optimize msb(cx16.r0) -> cx16.r0H, lsb(cx16.r0) -> cx16.r0L
|
2022-07-02 21:38:22 +02:00 |
|
|
|
25aad8d7be
|
improve const-evaluation of builtin expressions
|
2022-07-02 16:29:01 +02:00 |
|
|
|
965340ff90
|
logical and/or/xor/not all replaced by bitwise &,|,^,~ (ast, codegens)
this also fixed some invalid outcomes of logical expressions!
|
2022-07-02 00:38:17 +02:00 |
|
|
|
97cb0cbd08
|
tweak "not" removal/rewriting
|
2022-06-30 02:16:30 +02:00 |
|
|
|
4b358abbb7
|
"not" operator removed from ast and codegen (it's been replaced with x==0 as equivalent)
|
2022-06-29 01:13:08 +02:00 |
|
|
|
435d6f6f3f
|
vm: and/or/xor/not are all bitwise operations again
|
2022-06-28 03:17:51 +02:00 |
|
|
|
775c85fc18
|
don't swap operands that would change function evaluation order + vm: fix label casing error
|
2022-06-13 00:25:45 +02:00 |
|
|
|
5a756aaed9
|
Pipe expression "|>" removed from the language
|
2022-06-12 18:41:42 +02:00 |
|
|
|
dca092fd7c
|
fix pipe expression when start term is constant number
|
2022-06-12 16:59:28 +02:00 |
|
|
|
c6e92ecac4
|
some code cleanup
|
2022-06-12 16:15:08 +02:00 |
|
|
|
f7183e38ee
|
tweak trivial subroutine inlining
|
2022-06-08 21:05:03 +02:00 |
|
|
|
46ca0ac10d
|
properly optimize X - -1 and X + -1, this also fixes type change of ubyte - 2 + 10
|
2022-06-05 15:35:29 +02:00 |
|
|
|
7f69517fd4
|
preparing optimizing pointer indexing
|
2022-06-04 16:18:27 +02:00 |
|
|
|
ba614801ee
|
cleanup
|
2022-05-22 23:11:22 +02:00 |
|
|
|
fd6eb47e68
|
added inlining certain trivial non-asm subroutine calls
|
2022-05-22 20:22:09 +02:00 |
|
|
|
627aa61184
|
clean up subroutine inlining, basis for new try
|
2022-05-09 15:42:58 +02:00 |
|
|
|
942c5cc04b
|
fix crash when optimizing pipe expression too aggressively
|
2022-05-07 17:29:36 +02:00 |
|
|
|
348b3036ff
|
now correctly accepts "xxx" * constexpr (where constexpr is not just a single const number)
|
2022-05-05 23:21:20 +02:00 |
|
|
|
220246278a
|
removed sum(), max(), min(). abs() now always returns uword type.
This greatly simplifies internal handling of builtin functions by always having one fixed return type.
|
2022-04-14 00:21:16 +02:00 |
|
|
|
207a7e5160
|
move operator lists
|
2022-04-10 13:24:17 +02:00 |
|
|
|
036d9dbe59
|
got rid of unnecessary cast of boolean expressions by making their type dynamically adjust to byte or word
|
2022-04-04 23:43:55 +02:00 |
|
|
|
ae45ce517e
|
cleanups
|
2022-04-03 17:33:50 +02:00 |
|
|
|
30cbb6c9a8
|
implementing more of the vm
|
2022-03-27 21:59:46 +02:00 |
|
|
|
ed30108961
|
removed '**' power-operator. Use floats.pow() instead.
|
2022-03-27 13:16:34 +02:00 |
|
|
|
7b27d270a2
|
gosub only uses an identifier
|
2022-03-22 20:53:06 +01:00 |
|
|
|
97b3a0b093
|
don't use the temp-variables introducing optimizations for the vm target
|
2022-03-22 20:21:32 +01:00 |
|
|
|
ff57c5e9d3
|
working on vm and new ast
|
2022-03-21 01:36:11 +01:00 |
|
|
|
3961f26635
|
consolidating modules
|
2022-03-11 20:45:39 +01:00 |
|
