The largest element
The same program in M68K, MIPS, RISC-V, Z80.
Six numbers and the largest of them in r8. The loop keeps a best so far and replaces it whenever
an element beats it, which is how you find the largest, the smallest, the closest or the first of
anything.
The interesting decision is made before the loop starts. r8 begins holding the first element, and
the loop begins at the second. Starting r8 at zero instead would be easier to write and would answer
wrongly for an array of negative numbers, because zero would beat all of them.
There are two comparisons in every pass of this loop and they ask different kinds of question.
cmp rax, r8 followed by jle is signed, because the elements are data and data can be negative.
cmp rcx, COUNT followed by jb is unsigned, because rcx is an index and an index never is.
Writing the same flavour of condition for both is a bug that will not show up at all until somebody
puts a negative number in the array, and then it will look like the array is the problem.
Watch it happen. Change the 42 to -42 and the answer becomes 23, correctly, because -42 is the
smallest element now. Change the jle to jbe as well and the answer becomes FFFFFFFFFFFFFFD6,
which is those same bits read as an unsigned number and therefore the largest thing in the array by a
very long way.
The three lines from jle down to .skip: can be replaced by one cmovg r8, rax, which moves only
when the condition holds and never branches at all. The answer is the same and the loop body is
shorter by two instructions.