1 <!doctype linuxdoc system>
4 <title>cc65 coding hints
5 <author>Ullrich von Bassewitz, <htmlurl url="mailto:uz@cc65.org" name="uz@cc65.org">
9 How to generate the most effective code with cc65.
12 <sect>Use prototypes<p>
14 This will not only help to find errors between separate modules, it will also
15 generate better code, since the compiler must not assume that a variable sized
16 parameter list is in place and must not pass the argument count to the called
17 function. This will lead to shorter and faster code.
21 <sect>Don't declare auto variables in nested function blocks<p>
23 Variable declarations in nested blocks are usually a good thing. But with
24 cc65, there is a drawback: Since the compiler generates code in one pass, it
25 must create the variables on the stack each time the block is entered and
26 destroy them when the block is left. This causes a speed penalty and larger
31 <sect>Remember that the compiler does not optimize<p>
33 The compiler needs hints from you about the code to generate. When accessing
34 indexed data structures, get a pointer to the element and use this pointer
35 instead of calculating the index again and again. If you want to have your
36 loops unrolled, or loop invariant code moved outside the loop, you have to do
41 <sect>Longs are slow!<p>
43 While long support is necessary for some things, it's really, really slow on
44 the 6502. Remember that any long variable will use 4 bytes of memory, and any
45 operation works on double the data compared to an int.
49 <sect>Use unsigned types wherever possible<p>
51 The CPU has no opcodes to handle signed values greater than 8 bit. So sign
52 extension, test of signedness etc. has to be done by hand. The code to handle
53 signed operations is usually a bit slower than the same code for unsigned
58 <sect>Use chars instead of ints if possible<p>
60 While in arithmetic operations, chars are immidiately promoted to ints, they
61 are passed as chars in parameter lists and are accessed as chars in variables.
62 The code generated is usually not much smaller, but it is faster, since
63 accessing chars is faster. For several operations, the generated code may be
64 better if intermediate results that are known not to be larger than 8 bit are
75 the result of the & operator is an int because of the int promotion rules of
76 the language. So the compare is also done with 16 bits. When using
81 if ((unsigned char)(a & 0x0F) == 0)
84 the generated code is much shorter, since the operation is done with 8 bits
89 <sect>Make the size of your array elements one of 1, 2, 4, 8<p>
91 When indexing into an array, the compiler has to calculate the byte offset
92 into the array, which is the index multiplied by the size of one element. When
93 doing the multiplication, the compiler will do a strength reduction, that is,
94 replace the multiplication by a shift if possible. For the values 2, 4 and 8,
95 there are even more specialized subroutines available. So, array access is
96 fastest when using one of these sizes.
100 <sect>Expressions are evaluated from left to right<p>
102 Since cc65 is not building an explicit expression tree when parsing an
103 expression, constant subexpressions may not be detected and optimized properly
104 if you don't help. Look at this example:
112 The expression is parsed from left to right, that means, the compiler sees
113 'i', and puts it contents into the secondary register. Next is OFFS, which is
114 constant. The compiler emits code to add a constant to the secondary register.
115 Same thing again for the constant 3. So the code produced contains a fetch of
116 'i', two additions of constants, and a store (into 'i'). Unfortunately, the
117 compiler does not see, that "OFFS + 3" is a constant for itself, since it does
118 it's evaluation from left to right. There are some ways to help the compiler
119 to recognize expression like this:
123 <item>Write "i = OFFS + 3 + i;". Since the first and second operand are
124 constant, the compiler will evaluate them at compile time reducing the code to
125 a fetch, one addition (secondary + constant) and one store.
127 <item>Write "i = i + (OFFS + 3)". When seeing the opening parenthesis, the
128 compiler will start a new expression evaluation for the stuff in the braces,
129 and since all operands in the subexpression are constant, it will detect this
130 and reduce the code to one fetch, one addition and one store.
135 <sect>Use the preincrement and predecrement operators<p>
137 The compiler is not always smart enough to figure out, if the rvalue of an
138 increment is used or not. So it has to save and restore that value when
139 producing code for the postincrement and postdecrement operators, even if this
140 value is never used. To avoid the additional overhead, use the preincrement
141 and predecrement operators if you don't need the resulting value. That means,
160 <sect>Use constants to access absolute memory locations<p>
162 The compiler produces optimized code, if the value of a pointer is a constant.
163 So, to access direct memory locations, use
166 #define VDC_DATA 0xD601
167 *(char*)VDC_STATUS = 0x01;
170 That will be translated to
177 The constant value detection works also for struct pointers and arrays, if the
178 subscript is a constant. So
181 #define VDC ((unsigned char*)0xD600)
188 If you first load the constant into a variable and use that variable to access
189 an absolute memory location, the generated code will be much slower, since the
190 compiler does not know anything about the contents of the variable.
194 <sect>Use initialized local variables - but use it with care<p>
196 Initialization of local variables when declaring them gives shorter and faster
210 But beware: To maximize your savings, don't mix uninitialized and initialized
211 variables. Create one block of initialized variables and one of uniniitalized
212 ones. The reason for this is, that the compiler will sum up the space needed
213 for uninitialized variables as long as possible, and then allocate the space
214 once for all these variables. If you mix uninitialized and initialized
215 variables, you force the compiler to allocate space for the uninitialized
216 variables each time, it parses an initialized one. So do this:
233 The latter will work, but will create larger and slower code.
237 <sect>When using the ternary operator, cast values that are not ints<p>
239 The result type of the <tt/?:/ operator is a long, if one of the second or
240 third operands is a long. If the second operand has been evaluated and it was
241 of type int, and the compiler detects that the third operand is a long, it has
242 to add an additional <tt/int/ → <tt/long/ conversion for the second
243 operand. However, since the code for the second operand has already been
244 emitted, this gives much worse code.
251 return (a != 0)? 1 : a;
255 When the compiler sees the literal "1", it does not know, that the result type
256 of the <tt/?:/ operator is a long, so it will emit code to load a integer
257 constant 1. After parsing "a", which is a long, a <tt/int/ → <tt/long/
258 conversion has to be applied to the second operand. This creates one
259 additional jump, and an additional code for the conversion.
261 A better way would have been to write:
266 return (a != 0)? 1L : a;
270 By forcing the literal "1" to be of type long, the correct code is created in
271 the first place, and no additional conversion code is needed.
275 <sect>Use the array operator [] even for pointers<p>
277 When addressing an array via a pointer, don't use the plus and dereference
278 operators, but the array operator. This will generate better code in some
301 <sect>Use register variables with care<p>
303 Register variables may give faster and shorter code, but they do also have an
304 overhead. Register variables are actually zero page locations, so using them
305 saves roughly one cycle per access. Since the old values have to be saved and
306 restored, there is an overhead of about 70 cycles per 2 byte variable. It is
307 easy to see, that - apart from the additional code that is needed to save and
308 restore the values - you need to make heavy use of a variable to justify the
311 As a general rule: Use register variables only for pointers that are
312 dereferenced several times in your function, or for heavily used induction
313 variables in a loop (with several 100 accesses).
315 And remember: Register variables must be enabled with <tt/-r/ or <tt/-Or/.
319 <sect>Decimal constants greater than 0x7FFF are actually long ints<p>
321 The language rules for constant numeric values specify that decimal constants
322 without a type suffix that are not in integer range must be of type long int
323 or unsigned long int. This means that a simple constant like 40000 is of type
324 long int, and may cause an expression to be evaluated with 32 bits.
336 Here, the compare is evaluated using 32 bit precision. This makes the code
337 larger and a lot slower.
359 instead will give shorter and faster code.
362 <sect>Access to parameters in variadic functions is expensive<p>
364 Since cc65 has the "wrong" calling order, the location of the fixed parameters
365 in a variadic function (a function with a variable parameter list) depends on
366 the number and size of variable arguments passed. Since this number and size
367 is unknown at compile time, the compiler will generate code to calculate the
368 location on the stack when needed.
370 Because of this additional code, accessing the fixed parameters in a variadic
371 function is much more expensive than access to parameters in a "normal"
372 function. Unfortunately, this additional code is also invisible to the
373 programmer, so it is easy to forget.
375 As a rule of thumb, if you access such a parameter more than once, you should
376 think about copying it into a normal variable and using this variable instead.
projects / cc65 / blob