This looks like the logical inverse of hyper threading.
With hyper threading, you had a big core with, say, multiple ALUs that went unused if the work was numerically simple. So it tried to run two threads and use all ALUs.
This is the reverse—two cores that can merge to handle a single thread that requires more resources, in order to make a single thread run faster.
With hyper threading, you had a big core with, say, multiple ALUs that went unused if the work was numerically simple. So it tried to run two threads and use all ALUs.
This is the reverse—two cores that can merge to handle a single thread that requires more resources, in order to make a single thread run faster.