Memory Superscalar Problems

Today I started reading about memory access. I don't really have much to say on the matter yet, except that it is apparently not as easy to execute multiple memory access instructions at once as it is for integer/float instructions. This mostly stems from the fact that until the instruction is executed, you don't know what memory location it needs access to (this is because the address needs to be calculated by adding an offset to a register or some other such operation, as well as being translated into a physical address). You also can't use renaming to solve artificial data dependencies because there is way too much memory.

Reposted from http://invisiblegdev.blogspot.com/