dissassembled code questions

This is part of a function that does this:
for(int i = 0; i < num; i++)
local_buf[i] = arg_buf[i];

//here we load the address of the local buffer into EAX
1) 0x804837d <func+33>: lea 0xffffff28(%ebp),%eax
//then we move this address into EDX
2) 0x8048383 <func+39>: mov %eax,%edx
//then we add the value of a local counter @ address 0xffffff24(%ebp) to EDX to get the offset inside the local char buffer
3) 0x8048385 <func+41>: add 0xffffff24(%ebp),%edx
//then we move the value of the local counter into EAX, OVERWRITING THE STORED ADDRESS OF THE LOCAL BUFFER
4) 0x804838b <func+47>: mov 0xffffff24(%ebp),%eax

Questions:
-in step 1 what is the point of loading the address into EAX just so we can move it into EDX? why not just do this instead:
"lea 0xffffff28(%ebp), %edx"? am i missing something? b/c it seems as tho its pointless, especially since we also overwrite
that value in EAX anyways in step 4, yes?

-in step 1 what is the point of loading the address into EAX just so we can move it into EDX? why not just do this instead:
"lea 0xffffff28(%ebp), %edx"?

Depends on the instruction set and CPU architecture. Maybe there's no instruction to load into edx directly.

i tried this,
asm("lea 0xffffff28(%ebp), %edx");
and it compiled and executed correctly, any other ideas? or is it possible that the compiler is just retarded?

Just a question, but is (%ebp) the same as [ebp] in intel?

yes its dereferencing the register

What happens when you add different optimization options. I presume you compiled this without the -O option. Try compiling it with different optimization levels -O1, -O2, -O3 and see what happens