3 ms·
It's theoretically interesting up until the point where he starts eliminating registers and moving things to memory addresses. Main memory is significantly slow
by Kapura 9y ago
It's theoretically interesting up until the point where he starts eliminating registers and moving things to memory addresses. Main memory is significantly slower to access than in-processor registers. Granted, if you are taking the mov-only computation engine to it's logical conclusion (as Domas does) you may be able to speed up the hardware architecture to compensate for these losses, but is there anything of actionable value to gain from such a process?
(i did really enjoy the talk, fwiw)
- progval 9y ago> you may be able to speed up the hardware architecture to compensate for these losses It reminds me of the concept of Scratchpad memory: https://en.wikipedia.org/wiki/Scratchpad_memory https://en.wikipedia.org/wiki/Scratchpad_memory However, the main issue is with branches. As all code is executed all the time, the execution time is exponential in the number of basic blocks in the control-flow graph C program.
- qubex 9y agoOISC (One Instruction Set Computing) is a well-known concept in theoretical computer science. It is not considered an avenue for for performance computing. https://en.wikipedia.org/wiki/One_instruction_set_computer?wprov=sfti1 https://en.wikipedia.org/wiki/One_instruction_set_computer?w...