*** Additions, amendments and fixes made in version 3.3.0.7 of the PROTON+ Compiler. (1st April 2008) ***

ADDITIONS
WREG tracking optimisation is now implemented as standard for all devices.
This monitors the value held within the WREG and eliminates any unnecessary MOVLW mnemonics.
For example, a sequence of ASM mnemonics such as:

MOVLW 7
MOVWF VAR1
MOVLW 7
MOVWF VAR2

will be shortened to:

MOVLW 7
MOVWF VAR1
MOVWF VAR2

Added oscillator frequency 64MHz for use with the 16 MIPS (Million Instructions Per Second) 18F(K) devices (Not yet fully tested because K devices not released at this time).
Added full interrupt context saving and restoring with 14-bit core and 16-bit core (18F) devices by the use of four new directives: -

HIGH_INT_SUB_START, HIGH_INT_SUB_END
LOW_INT_SUB_START, LOW_INT_SUB_END

When each of the above pair of directives surrounds a Hardware Generated Interrupt subroutine, any of the compiler's system variables used will be saved to reserved RAM
when the commands CONTEXT SAVE and CONTEXT RESTORE are issued.
Both high and low priority interrupts are catered for. For example: -

' High Priority Interrupt Subroutine for a 16-bit core (18F) device
HIGH_INT_SUB_START
InterruptSubHigh:
CONTEXT SAVE
' Interrupt Code goes Here !
CONTEXT RESTORE
HIGH_INT_SUB_END

' Low Priority Interrupt Subroutine for a 16-bit core (18F) device
LOW_INT_SUB_START
InterruptSubLow:
CONTEXT SAVE
' Interrupt Code goes Here !
CONTEXT RESTORE
LOW_INT_SUB_END

Several of the PICmicro's SFR's (Special Funtion Registers) will also be context saved along with the compiler's System Variables.
This is unavoidable, and essential for the seemless operation of the program.
The SFR's in mention are: -

WREG, STATUS, BSR, FSR0L\FSR0H, FSR1L\FSR1H, FSR2L\ FSR2H, PRODL\PRODH, TBLPTRL\TBLPTRH, and TABLAT

When using 14-bit core devices, the declares INT_SUB_START, INT_SUB_END can be used as well as HIGH_INT_SUB_START, HIGH_INT_SUB_END.

' Interrupt Subroutine for a 14-bit core (16F) device
INT_SUB_START
InterruptSub:
CONTEXT SAVE
' Interrupt Code goes Here !
CONTEXT RESTORE
INT_SUB_END

All of this takes RAM space, so the compiler reserves just enough RAM at the top of memory to cater for the amount of variables and SFR's used.
A hint (reminder) will be issued when this is accomplished, which can be disabled using the REMINDERS declare.
Added a simple cast for expressions within Conditions.

For example:
A comparison that contains an expression will always evaluate the expression with a DWORD (32-bit) product if the left hand variable is a WORD (16-bit).

IF WORDIN = (100 * WORDVAR) + 60000 THEN

A comparison that contains an expression will always evaluate the expression with a WORD (16-bit) product if the left hand variable is a BIT or BYTE (8-bit).

IF BYTEIN = (100 * BYTEVAR) THEN

The addition of a BYTE, WORD, DWORD or FLOAT keyword at the start of an expression within a condition will ensure that the expression will have the appropriate result. i.e.

IF DWORD WORDIN = (100 * WORDIN) + 60000 THEN

or

IF BYTE BYTEIN = BYTE (BYTEVAR - 20) THEN

Note: This will not become an official part of the language until it has been extensively tested.

The SIZEOF_TABLE directive has been added that will calculate the size of a table it is within.
For example:

CDATA SIZEOF_TABLE, 10, 20, 30, 40

The above table will contain 5 items. i.e. 5, 10, 20, 30, 40.

The same applies to LDATA and EDATA tables.

The SIZEOF_TABLE directive can be placed anywhere within the table and can be preceded by a casting directive:

LDATA as Word 10, 20, 30, 40, SIZEOF_TABLE

AMENDMENTS
The compiler's PPI and Source files are now contained in the folder "Includes" instead of "INC".
Implemented The USB Stack version 1.2 for USB 18F devices.
Tightened bank switching when using software interrupts.
For the third party Library Manager software, the macros NUM_FSR, NUM_FSR0, NUM_FSR1, NUM_FSR2, and LABEL_WORD can now be accessed from
the BASIC file without a preceding ampersand. This ensures that they follow the rules of macro generation and usage and will help the optimiser do its job.
Added support for long BASIC file names and file paths that exceed 62 characters. However, the full path should be kept less than 230 characters when using ISIS.
The compiler now produces a single ASM file containing all the information necessary for the PICmicro used.
The compiler now implements dead code removal, further tightening the ASM produced by the declare:

DEAD_CODE_REMOVE ON-OFF or TRUE-FALSE or 1-0

This means that the optimiser now has only three levels (four including 0). Note 14-bit core devices now only have two levels of optimisation. See the "KNOWN ISSUES" section
The first two levels optimise the main code. The third optimise pass (18F devices only) tightens loops and branches further, but at a cost of another assembler pass.
Optimised Boolean conditions.
The CONFIG directive can no longer be used with a following equals character with 14-bit (16F) devices. i..e. CONFIG = xxxxx
Added preceding 0 to floating point display when using the faster, but larger library routine initiated by the declare:-

FLOAT_DISPLAY_TYPE = LARGE or FAST

Word arrays can now have up to 256 elements when using 16-bit core (18F) devices.
Support for the, now obsolete, One Time Programmable (OTP) 18C devices has been removed. Chiefly due to the lack of the essential LFSR mnemonic.
Altered library ASM code for 32-bit multiply to use the hardware multiplier with 18F devices. The code produced is slightly larger but the speed increase is tremendous.
The compiler now uses a single configuration file with the extension .PPI instead of the .INF and .LPB files.
Important
The PPI files are semi-automatically created from Microchip .INC files, however, Microchip have changed some of the fuse configuration names.
To view the valid fuse names, open the relevant PPI file, found in the compiler's "Includes\PPI" folder, and scroll down to the bottom of the texts.

The names are in the [CONFIGSTART]/[CONFIGEND] section.
For 16-bit core (18F) devices, the fuse names for the new CONFIG_START/CONFIG_END block can be found in the [FUSESTART]/[FUSEEND] section.
18F devices containing built-in ethernet hardware such as the 18F97J60 have an SFR (Special Function Register) named EDATA which clashes with the compiler's eeprom data Keyword.
The EDATA SFR has to be addressed as _EDATA or PP_EDATA

Implemented support using PPI files for devices: -

12-bit core.
10F200, 10F202, 10F204, 10F206, 10F220, 10F222, 12C508, 12C508A, 12C509, 12C509A, 12CE518, 12CE519
12F508, 12F509, 12F510, 16F54, 16F57, 16C505, 16C54, 16C54A, 16C54B, 16C54C, 16C55, 16C554
16C558, 16C55A, 16C56, 16C56A, 16C57, 16C57C, 16C58A, 16C58B, RF509AF, RF509AG

14-bit core.
12C671, 12C672, 12CE673, 12CE674, 12F629, 12F635, 12F675, 12F683, RF675F, RF675H, RF675K
16C432, 16C433, 16C61, 16C62, 16C620, 16C620A, 16C621, 16C621A, 16C622, 16C622A, 16C62A, 16C62B, 16C63, 16C63A, 16C64, 16C642
16C64A, 16C65, 16C65A, 16C65B, 16C66, 16C662, 16C67, 16C71, 16C710, 16C711, 16C712, 16C715, 16C716, 16C717, 16C72, 16C72A
16C73, 16C73A, 16C73B, 16C74, 16C745, 16C74A, 16C74B, 16C76, 16C765, 16C77, 16C770, 16C771, 16C773, 16C774, 16C781, 16C782
16C84, 16C923, 16C924, 16CE623, 16CE624, 16CE625, 16CR54, 16CR54A, 16CR57A, 16CR57B, 16CR57C, 16CR58A, 16CR58B, 16CR62, 16CR63, 16CR64
16CR65, 16CR72, 16CR83, 16CR84, 16F506, 16F610, 16F616, 16F627, 16F627A, 16F628, 16F628A, 16F630, 16F636, 16F639, 16F648A, 16F676
16F684, 16F685, 16F687, 16F688, 16F689, 16F690, 16F716, 16F72, 16F73, 16F737, 16F74, 16F747, 16F76, 16F767, 16F77, 16F777
16F785, 16F818, 16F819, 16F83, 16F84, 16F84A, 16F87, 16F870, 16F871, 16F872, 16F873, 16F873A, 16F874, 16F874A, 16F876, 16F876A
16F877, 16F877A, 16F88, 16F882, 16F883, 16F884, 16F886, 16F887, 16F913, 16F914, 16F916, 16F917, 16HV610, 16HV616

16-bit core.
18F1220, 18F1230, 18F1320, 18F1330, 18F2220, 18F2320, 18F2331, 18F2410, 18F242, 18F2423, 18F2431, 18F2439, 18F2455, 18F248
18F24J10, 18F24K20, 18F2510, 18F2515, 18F252, 18F2520, 18F2523, 18F2525, 18F2539, 18F2550, 18F258, 18F2580, 18F2585, 18F25K20
18F2610, 18F2620, 18F2680, 18F4220, 18F4320, 18F4331, 18F4410, 18F442, 18F4423, 18F4431, 18F4439, 18F4455, 18F448, 18F44J10
18F44K20, 18F4510, 18F4515, 18F452, 18F4520, 18F4523, 18F4525, 18F4539, 18F4550, 18F458, 18F4580, 18F4585, 18F45J10, 18F45K20
18F4610, 18F4620, 18F4680, 18F6310, 18F6390, 18F63J90, 18F6410, 18F6490, 18F64J90, 18F6520, 18F6525, 18F6585, 18F65J10, 18F65J15
18F65J50, 18F65J90, 18F6620, 18F6621, 18F6622, 18F6627, 18F6680, 18F66J10, 18F66J15, 18F66J50, 18F66J55, 18F66J60, 18F66J65, 18F6720
18F6722, 18F67J10, 18F67J50, 18F67J60, 18F8310, 18F8390, 18F83J90, 18F8410, 18F8490, 18F84J90, 18F8520, 18F8525, 18F8585, 18F85J10
18F85J15, 18F85J50, 18F85J90, 18F8620, 18F8621, 18F8627, 18F8680, 18F86J10, 18F86J15, 18F86J50, 18F86J55, 18F86J60, 18F86J65, 18F8720
18F8722, 18F87J10, 18F87J50, 18F87J60, 18F96J60, 18F96J65, 18F97J60

For 18LF devices use the standard 18F device. This list of devices will be amended when final datasheets for a specific device are available from Microchip.

FIXES
Corrected the floating point POW command. It now traps the first parameter if 0.
Corrected STR command. It skipped an element in the array after a Quoted String was issued as an item.
Corrected ALL_DIGITAL directive for 16F88x and some 18F devices containing ANSEL0 and ANSEL1 SFRs
Corrected the DIV32 command.
Corrected POT command with a comment following it.
Corrected String slicing. i.e. STRING1[INDEX].
Corrected Integer DWORD = DWORD + WORD for 14-bit core devices where any of the variables are not in the same bank and the operator is not the same as the assignment.
Corrected RC5IN command with 16-bit core (18F) devices.
Corrected Complement and Negation within expressions.
Corrected Keyboard mapping for countries other than UK and American English when using Floating Point Constants.
Corrected BRANCHL with 16-bit core (18F) devices where the list of labels is too long to be jumped over.
Slowed down OREAD library routine when using an 8MHz crystal.
Corrected SERIN with timeout, Sometimes not loading a variable with the received data.
Corrected CREAD and CWRITE with 14-bit core devices containing SFR EECON1 in bank1 of RAM.
Corrected ADIN with devices 16F883, 16F883, 16F886 and 16F887.
Corrected WORD Array results within expressions when the Array has an expression for its index.
Corrected some obscure bank anomalies with certain 16-bit core (18F) DWORD compares.
Corrected the creation of a floating point constant when using an expression containing alias's to floating point constants: -

SYMBOL X = 2.0
SYMBOL Y = 100.0
SYMBOL Z = X * Y

Z will represent the floating point constant 200.0

If a mix of floating point constants and integer constants are in the expression, the compiler will implement a floating point calculation: -

SYMBOL X = 2 '(integer)
SYMBOL Y = 100.0
SYMBOL Z = X * Y

Z with represent the floating point constant 200.0

KNOWN ISSUES
1).
Compiler does not complain with a syntax error if a quoted string of characters is used within an expression and is not the first part. This will produce no code for the expression.

2).
STR and STRN accept multiple parameters but do not produce the correct code if STR or STRN are used as parameters before a quoted character string, or value.

3).
Important Note: The optimiser for 14-bit core devices is no longer a supported feature of the compiler and is only retained because of user pressure.

4).
A deeply nested inline command as a LOOKUPL parameter will cause an internal stack underflow error.

5).
STRING concatenations where the assignment string is also one of the operators after the first operator will produce an incorrect assignment result.
For example: -

STRING1 = "TEXT"
STRING1 = "1234" + STRING1
will leave STRING1 holding "12341234TEXT"

The workaround is to assign STRING1 to a temporary string first: -

STRING1 = "TEXT"
TEMP_STRING = STRING1
STRING1 = "1234" + TEMP_STRING