Moore's Law promises very soon more than a billion transistors on a chip processing at more than 10 GHz. But, what else does Moore 's Law tell us about the future of high performance microprocessors. Is there anything left to do in the microarchitecture to make use of all this capability, or should we fold our tents and go home? Many claim we have reached the end of our rope. In this talk, I hope to spend a little time discussing the rationale of the naysayer and then move on to what we can do. Per usual, I plan to tie my comments to the fundamentals, because I believe it is always the fundamentals that provide the insights to new knowledge. Finally, I will discuss what I think we will see in the microprocessor of the year 2014, including the block-structured ISA, stronger use of SSMT, greater use of microcode, dedicated infrequently used functional units, and most importantly, a stronger coupling with the compiler, language, and algorithm technologies. If time permits, I will discuss some things I do not think we will see on the chip, like qbits and machines that think. |
