acqn
b525554bfe
Added support for parsing asm byte size expressions with a pair of parentheses.
2020-09-27 12:22:05 +02:00
acqn
688342e194
Now every code entry has its argument parsed to tell some info.
...
It fixes the compiling performance regression as well.
Built-in ZPs are recognized.
2020-09-27 12:22:05 +02:00
acqn
cc0f8422f2
More fixes and new utils to check if opcode arguments can be used elsewhere.
...
Fixed tracking with LI_RELOAD_Y and LI_DIRECT.
Fixed tracking with LI_CHECK_Y and LI_RELOAD_Y.
2020-09-27 12:22:05 +02:00
acqn
f3771a465d
Fixed various issues in the usage-tracking code.
...
Added some utility functions.
2020-09-27 12:22:05 +02:00
acqn
d379affc4b
Moved some reusable code from cc65/coptstop.c into new files.
2020-09-27 12:22:05 +02:00
acqn
28c7aa2bc8
Replaced direct CEF_NUMARG flag checks on code entries with CE_HasNumArg().
2020-09-27 12:22:05 +02:00
acqn
c45a6b3685
Utility function ParseOpcArgStr().
2020-09-27 12:22:05 +02:00
Jesse Rosenstock
d0089aef95
Fix bit-field truncation warning message
...
Fix copy & paste bug with warning message.
struct X {
signed int a : 3;
};
struct X g = { 5 };
Before:
s.c(4): Warning: Implicit truncation from 'int' to 'int : 3' in bit-field
initializer changes value from 5 to 5
After:
s.c(4): Warning: Implicit truncation from 'int' to 'int : 3' in bit-field
initializer changes value from 5 to -3
Fixes #1268
2020-09-26 17:41:45 -04:00
Greg King
81ac28ff71
Used TAY/TYA instead of PHA/PLA when extracting a bit-field.
...
Without optimization, it saves a few CPU cycles. With optimization, it saves more cycles and a few bytes.
2020-09-22 12:47:09 -04:00
Greg King
ea95728330
Avoided an avalanche of messages from bad bit-field declarations.
...
Made cc65 replace a bad bit-field type with a good one, and always parse the field width.
Shortenned a parameter name to a spelling that's consistent with other function headers.
2020-09-22 12:47:09 -04:00
Greg King
07ea5259ac
Changed a cc65 error message to say that the sizes of bit-field types (not bit-fields) are limited.
2020-09-22 12:47:08 -04:00
Greg King
bf4b195016
Added some comments that explain where the g_branch() code generator can and can't be used.
2020-09-18 15:50:26 -04:00
acqn
de630a1245
Fixed quick hack for known registers after calling certain runtime functions,
...
and new quick hack for tosshrax.
2020-09-18 09:04:15 +02:00
acqn
079f4a99dd
Added processor state info to the OPCDesc table. However, since some opcodes are affected by the addressing mode, this info is unused in codegen/optimizer but just as quick in-code documentation/hints.
2020-09-18 09:04:15 +02:00
acqn
a7d6eb9190
Added processor flag usage/change-tracking for non-JSR/RTS code entries.
...
Some existing optimizations are impacted and need fixes in order to work correctly again.
Fixed and improved OptPrecalc. Now it respects the C/V/Z/N flags.
Fixed optimizations impacted by added support of tracking processor flags.
2020-09-18 09:04:15 +02:00
acqn
d02b12fa6c
Made local code/data labels really local to the functions where they live in.
2020-09-18 09:02:16 +02:00
acqn
86ced2bd4c
Avoided unnecessarily boosting the code label numbers with boolean AND.
2020-09-18 08:59:56 +02:00
acqn
b8ae5c28fe
Added debug output support for processor flags.
2020-09-13 13:50:35 +02:00
acqn
810e17edfe
Fixed processor states tracking for the BIT/TRB/TSB opcode.
...
Added new opcode descriptions about whether and how the opcode accesses memory.
2020-09-13 13:50:35 +02:00
acqn
66c5faeb9a
Added processor flags usage tracking.
...
Added ZNRegs for tracking what register(s) Z/N flags currently reflect.
Added utility functions to check if the specified processor state is known to be a certain value.
2020-09-13 13:50:35 +02:00
acqn
fe3f267233
Added new runtime sub bcasta/bcastax/bcasteax opposing to bnega/bnegax/bnegeax.
2020-09-11 13:33:54 +02:00
acqn
142b0bf9b3
Added utility functions to get names of comparison function/transformer subroutines.
2020-09-11 13:33:54 +02:00
acqn
ab7e9f8424
Hotfix for Issue #1250 .
2020-09-08 16:28:21 +02:00
Jesse Rosenstock
4e4e4c2d21
Allow char bit-fields
...
These are not required to be supported (only int, signed int, and
unsigned int are required), but most compilers support it.
https://port70.net/~nsz/c/c89/c89-draft.html#3.5.2.1
https://port70.net/~nsz/c/c89/c89-draft.html#A.6.5.8
For consistency with other integral types, plain `char` bit-fields
are unsigned, regardless of the `--signed-chars` option.
Fixes #1047
2020-09-08 14:24:04 +02:00
acqn
41cee0eb44
Extended support for more addressing modes in tos* optimizations.
2020-09-08 14:19:48 +02:00
acqn
7553b60ef0
Improved OptStackOps for optimizating further when operands have equal hi-bytes.
2020-09-08 14:19:48 +02:00
acqn
57117fa687
Utility functions about compare conditions.
2020-09-08 14:19:48 +02:00
acqn
64ef562fa7
Moved all optimization steps from codeopt.c and some optimization steps from coptind.c into new separate files.
2020-09-08 14:18:22 +02:00
Jesse Rosenstock
9a0e4a35e1
Fix enum bit-field ICE #1244
...
This previously resulted in an ICE:
cc65: Check failed: (Entry->Type->C & T_MASK_SIGN) == T_SIGN_SIGNED,
file 'cc65/symtab.c', line 874
This CHECK is in the code that deals with changing `int` bitfields to
`unsigned int`.
Work around this by treating enum bit-fields as having their signedness
specified, so this type change code does not get called.
Fixes #1244 .
2020-09-07 11:21:23 +02:00
Tevo
1e7a9e44af
Update comment to reflect addition of integer boundary constants
2020-09-06 23:32:00 +02:00
Tevo
23a8b2c303
Define integer size macros for lacking systems
...
Define integer size macros for lacking systems
2020-09-06 23:32:00 +02:00
acqn
6bb2d1d5d1
Fixed a comment on struct/union types in DoCompare().
2020-09-06 22:47:11 +02:00
acqn
1cde952cf5
Fixed comparing an enum type with a non-enum type in DoCompare().
2020-09-06 22:47:11 +02:00
acqn
6e0fb630d7
Fixed check for processor flags usage in case of PHP.
2020-09-06 13:32:38 +02:00
acqn
e2f950b15e
Avoided referencing string literals with sizeof so that they are not output if unused elsewhere.
2020-09-05 13:21:21 +02:00
acqn
903e84c24c
Std-functions are no longer inlined if they are unevaluated.
2020-09-05 13:21:21 +02:00
acqn
df07e23f2c
Fixed bnegeax funcinfo on register usage.
2020-09-05 02:25:56 -04:00
acqn
f849de6769
Object addresses as non-NULL for comparison with NULL.
2020-09-04 19:50:32 +02:00
acqn
8e0b2f0833
Object addresses as non-NULL in boolean context.
2020-09-04 19:50:32 +02:00
acqn
bc5570b708
Fixed logical-NOT in constant context.
2020-09-04 19:50:32 +02:00
acqn
676b14429d
Better opt chance for certain optimization steps e.g. OptPtrStore1 etc.
2020-09-04 12:00:10 +02:00
acqn
29d1999947
Rearranged the OptFunc's in alphabetic order.
2020-09-04 12:00:10 +02:00
acqn
761d00b7dc
Ignore running removed old optimization steps. Neither list them in help info.
2020-09-04 12:00:10 +02:00
acqn
ae340703f2
OptDupLoads shouldn't silently do optimizations.
2020-09-04 11:58:25 +02:00
acqn
f45d2515eb
Fixed OptPush1 in case later code would rely on that pushax zeroes register Y.
2020-09-04 11:58:25 +02:00
acqn
fb6bc275bc
Fixed evaluation flags propagation to subexpressions in assignments and function calls.
2020-09-02 21:36:59 +02:00
acqn
2379504449
Fixed AND/OR in certain cases where the 'E_NEED_TEST' flag set for usage only in subexpressions should be cleared.
2020-09-02 21:36:59 +02:00
acqn
28de3423eb
Merged some switch cases in code generation subroutines.
2020-08-31 09:12:03 +02:00
acqn
6b64b43395
Made local static data use a separated label pool from the code label pool.
2020-08-31 09:12:03 +02:00
acqn
9398e1cd33
Use a dedicated label pool for literals.
2020-08-31 09:12:03 +02:00
