Performance of static array in record on GNAT / gcc 4.6

Performance of static array in record on GNAT / gcc 4.6

[john]

Before I do the measurement myself, I wondered whether someone else has already done it. Question: If I pass a record with two large, fixed size arrays with definite values to a procedure as in parameter, is there any performance penalty for accessing elements of the arrays inside the record and procedure as opposed to defining the arrays directly inside the procedure and without wrapping them into a record? 

I'm using GNAT/gcc 4.6 with -O3 and Pragma Optimize (time)

I'm working on a (toy) virtual machine/interpreter where every cycle counts. 

Thanks for any advice!

P.S.: In case someone asks, yes, I'm considering using assembler but for the time being would like to keep it portable.

Re: Performance of static array in record on GNAT / gcc 4.6

[Florian Weimer]

> Before I do the measurement myself, I wondered whether someone else
> has already done it. Question: If I pass a record with two large,
> fixed size arrays with definite values to a procedure as in
> parameter, is there any performance penalty for accessing elements
> of the arrays inside the record and procedure as opposed to defining
> the arrays directly inside the procedure and without wrapping them
> into a record?

There shouldn't be a difference in performance.  You could verify this
by comparing the generated machine code.