Hi all, I need some help about assigning two structures.
The simple Function
void CopyStruct(struct A *sSrc, struct A *sDst) { *sDst = *sSrc; }
is translated to
; Variable 'sDst' assigned to Register 'R10/R11' ; Variable 'sSrc' assigned to Register 'R8/R9' 0010 F06A MOV R6,R10 0012 F048 MOV R4,R8 0014 F07B MOV R7,R11 0016 F059 MOV R5,R9 0018 E6F23600 MOV R2,#036H 001C DA000000 E CALLS SEG (?C_WCPYFF),?C_WCPYFF
How is the data copied? Where can I find the implementation of the library Function "C_WCPYFF"? (Keil C166 6.04a)
Thanks for your Help.
Regards, Marcus
Maybe the call is to the same code used by the C library function :
memcpy( void* , void* , int )
- Are You sure that all the struct are copied correctly ? - Whats about the sizeof( ) of the struct ? - How the function called can know it if the sizeof is not passed to registers ?
Usually I use the following code :
typedef strcut A { .... } str_A;
void CopyStruct( str_A *sSrc, str_A *sDst) { memcpy( *sSrc , *sDst , sizeof(str_A) ) }
"How the function called can know it if the sizeof is not passed to registers ?"
I'd be pretty certain that the size is being passed in registers - how about:
0018 E6F23600 MOV R2,#036H
My primary Intention is a better understandig.
What about Data Integrity of larger Objects in Multitasking Environments? I.e. do I have to protect against Interrupts (Struct consisting of 16Bit Objects @ 16Bit operations = No, @ 8Bit operations = yes)
What about Performance in time-critical projects?
Of course I could write my own Function, but the simple assignment is quite charming.
"What about Data Integrity of larger Objects in Multitasking Environments?"
Even a simple assignment is not guaranteed to be atomic - so nothing special here.
"What about Performance in time-critical projects?"
If you have a time-critical project, why are you copying big structures around anyhow?
"Of course I could write my own Function, but the simple assignment is quite charming."
That's the trade-off: either write it yourself, and know what's going on, or go for convenience and delegate the control to the compiler. You can't have it both ways, I'm afraid.
I don't think Keil give out their source code. But you don't need it: the copy subroutine is small and you can easily figure out how it works by looking at its disassembly. I suspect in some memory models and/or optimization settings the compiler could inline the copy subroutine. If you need atomicity for thread-safe operation, you should take care of it yourself. The compiler should not disable interrupts silently as this would increase interrupt latency and create a risk of side effects. As for performance, if you think that the standard copy subroutine is suboptimal, right one yourself. That's what I would have done, anyway.
Regards, - mike
Even a simple assignment is not guaranteed to be atomic - so nothing special here
What do you mean by that? In which cases can a 8 or 16 Bit assignment on a 16 Bit Platform like the XC be unsafe? Can you give me an example?
Thanks
"In which cases can a 8 or 16 Bit assignment on a 16 Bit Platform like the XC be unsafe"
I didn't say that!
All I said was, Even a simple assignment is not guaranteed to be atomic"
In fact, as you say, it probably will be safe for 8 or 16 bit assignments on a 16 Bit Platform like the XC - but that is a feature of the target environment, not a guarantee of the language. And what happens if the 16-bit value is not word-aligned...?