use more space efficient calling convention for calling errors

c64/basic/do_basic_error.s

@@ -1,6 +1,131 @@
-	;; Error # in X
+	;;  As many errors are triggered from many places in the BASIC
+	;; interprettor, it is important to have an efficicent means of
+	;; calling them.
+	;; LDX #$nn / jmp do_basic_error
+	;; is the simplest, but takes 5 bytes.
+	;; Having a direct JMP address for each error is preferable, as
+	;; it reduces the overhead to 3 bytes, but at the cost of having
+	;; a jump table of some sort. The .byte $2C method can be used
+	;; to have a set of addresses that each execute a different
+	;; LDX #$nn command, reducing the cost of the jump table to 3
+	;; bytes per error number.
+	;; But we can do better with a bit of scullduggery and carnal
+	;; knowledge of ZP usage.
+	;; The INC $nn opcode is $E6. ZP location $E6 is one of the screen
+	;; link addresses.  In the original KERNAL, the lower bits of those
+	;; addresses have meaning, but in our KERNAL, we ignore them, and just
+	;; have them all zeroes.  So.... We can have a sequence of $E6 bytes,
+	;; and know exactly how many of them have been executed, by checking
+	;; the value of $E6.  Then all we need to worry about is if an odd
+	;; number is required, then we have to have an exit that has a $E6
+	;; opcode with some other address as target, but where the byte value
+	;; of that argument can be safely executed as a CPU instruction.
+	;; NOP = $EA is the obvious choice, as if used as an argument to
+	;; $E6 will increment location $EA, which is another of the screen
+	;; link addresses.
 
-do_basic_error:
+	;; This gives a total overhead of 18 bytes + 1 byte per error
+	;; message.  This compares favourably to the 3 bytes per error
+	;; message that the $2C method would result in.
+
+do_NOT_IMPLEMENTED_error:
+	.byte $E6
+do_MEMORY_CORRUPT_error:
+	.byte $E6
+	;; the message "BYTES FREE"
+	.byte $E6
+	;; the word ERROR
+	.byte $E6
+	;; The word "SAVING"
+	.byte $E6
+	;; The word "VERIFYING"
+	.byte $E6
+	;; The word "LOADING"
+	.byte $E6
+	;; The word "READY."
+	.byte $E6	
+do_LOAD_error:
+	.byte $E6
+do_VERIFY_error:
+	.byte $E6
+do_UNDEFD_FUNCTION_error:
+	.byte $E6
+do_CANT_CONTINUE_error:
+	.byte $E6
+do_FORMULA_TOO_COMPLEX_error:
+	.byte $E6
+do_FILE_DATA_error:
+	.byte $E6
+do_STRING_TOO_LONG_error:
+	.byte $E6
+do_TYPE_MISMATCH_error:
+	.byte $E6
+do_ILLEGAL_DIRECT_error:
+	.byte $E6
+do_DIVISION_BY_ZERO_error:
+	.byte $E6
+do_REDIMD_ARRAY_error:
+	.byte $E6
+do_BAD_SUBSCRIPT_error:
+	.byte $E6
+do_UNDEFD_STATEMENT_error:
+	.byte $E6
+do_OUT_OF_MEMORY_error:
+	.byte $E6
+do_OVERFLOW_error:
+	.byte $E6
+do_ILLEGAL_QUANTITY_error:
+	.byte $E6
+do_OUT_OF_DATA_error:
+	.byte $E6
+do_RETURN_WITHOUT_GOSUB_error:
+	.byte $E6
+do_SYNTAX_error:
+	.byte $E6
+do_NEXT_WITHOUT_FOR_error:
+	.byte $E6
+do_ILLEGAL_DEVICE_NUMBER_error:
+	.byte $E6	
+do_MISSING_FILENAME_error:
+	.byte $E6
+do_NOT_OUTPUT_FILE_error:
+	.byte $E6
+do_NOT_INPUT_FILE_error:
+	.byte $E6
+do_DEVICE_NOT_PRESENT_error:
+	.byte $E6
+do_FILE_NOT_FOUND_error:
+	.byte $E6
+do_FILE_NOT_OPEN_error:
+	.byte $E6
+do_FILE_OPEN_error:
+	.byte $E6
+do_TO_MANY_FILES_error:
+	.byte $EA
+
+	;; Now get the error # and restore $E6 to its correct value
+	lda $E6
+	pha
+	and #$80
+	sta $E6
+	pla
+	;; $E6 has even part
+	asl
+	ora $EA
+	and #$7f
+	tax
+
+	;; Restore $EA to correct value in case the NOP above is used
+	;; as an argument to the preceeding INC $nn (opcode $E6).
+	LDA $EA
+	and #$80
+	sta $EA
+	
+	;; FALL THROUGH
+
+	;; Error # in X
+	
+;; do_basic_error:			
 	;; "?"
 	txa
 	pha