On Sun, 28 Aug 2016, Bob Ammerman wrote:

>> As to the contention that the best C compilers can somehow best an
>> experienced Assembly programmer, well, I'll leave that open to debate.
>
> I like to consider myself an "experienced" assembly programmer (at least =
the
> gray hair on my head would serve to indicate so).
>
> I had a project on a PIC24 that required the implementation of TEA (tiny
> encryption algorithm). This algorithm consists of a lot of shifting,
> rotating and bit twiddling, and, with speed of the essence, I thought it
> wise to code in assembly. After spending way too much time on it, I was a=
ble
> to come up with a version that kept all the working variables in register=
s,
> using every single register except the stack pointer. It ran fast, and I =
was
> very proud of my work. Curiosity overtook me and I decided to see what th=
e
> compiler would do with it. I quickly cobbled up a C version of it, withou=
t
> much regard to efficiency. To my surprise the compiler's code was
> substantially identical to mine (modulo storing things in different
> registers). It spilled one variable to the stack, and missed a few detail=
s,
> but its run time was only 15% slower than my very carefully optimized cod=
e.

It is common for an experienced embedded C programmer to be able to tweak=20
C code so as to coax the compiler to generate better optimization on=20
successive iterations.

For example you might find that a C compiler generates better code if the=20
control variable of a for loop decrements down to 0, or if the control=20
variable is not used inside the for loop at all. You might also find that=20
indexing an array is much slower that auto incrementing a pointer.

Regards
Sergio Masci
--=20
http://www.piclist.com/techref/piclist PIC/SX FAQ & list archive
View/change your membership options at
http://mailman.mit.edu/mailman/listinfo/piclist
.