IIRC (it's been a long time) the Cray 1 could do inverse, multiply, and square root in single cycle instructions. Isqrt would take two cycles as wood divide. (Back to the rocking chair.)
Actually it didn't http://klausler.com/cray2.txt it had a command for a square root approximation and a command to iteratively get closer to the real square root.
But even if a certain computer did compute a square root (or even an approximation) in a single cycle, that would mean CPU speed was reduced by slowing down the clock to enable a more complex action to complete within the timeframe of a single cycle, and other simpler actions like adding/subtracting/moving memory we're sacrificed in its favor.
Not necessarily. The square root hardware could be heavily pipelined to get each stage short enough for one cycle. This increases latency, but lets you keep a one-per-cycle throughput without slowing down the clock.
Comments
IIRC (it's been a long time) the Cray 1 could do inverse, multiply, and square root in single cycle instructions. Isqrt would take two cycles as wood divide. (Back to the rocking chair.)
Actually it didn't http://klausler.com/cray2.txt it had a command for a square root approximation and a command to iteratively get closer to the real square root.
But even if a certain computer did compute a square root (or even an approximation) in a single cycle, that would mean CPU speed was reduced by slowing down the clock to enable a more complex action to complete within the timeframe of a single cycle, and other simpler actions like adding/subtracting/moving memory we're sacrificed in its favor.
Not necessarily. The square root hardware could be heavily pipelined to get each stage short enough for one cycle. This increases latency, but lets you keep a one-per-cycle throughput without slowing down the clock.
I didn't know that Crays could split timber!