Each Monday I answer questions from the comments on the prior week’s videos. Transcripts are not available for Q&A videos due to length. I do produce closed captions for them, but Substack still has not enabled closed captions on videos :(
Questions addressed in this video:
[00:17] “I have one question, I heard the older CPUs were more efficient doing calculations for smaller number sizes. So for instance if we used 8 bit values the operations would be faster than using for instance a 32 bit or 16 bit value. In the most recent modern CPUs is there any gain for using smaller data types or are they optimized for doing work on 32 bit or 64 bit values?”
[06:33] “I wonder if we should simulate how the EU and BIU work in tandem to get a better estimate? I guess the EA and JMP instruction clocks count hide some part of that. For example, a JMP will empty the Instruction Queue, so the BIU needs some extra clocks to get the new instructions bytes from memory, and that time is added to the total clock count of the JMP instruction (though the manual says the BIU passes those bytes directly to the EU before filling its queue, but nonetheless if a memory access is done right after that, that should cause some contention over the bus)”
[14:24] “If I understand correctly, adding (for example) with memory as a destination is much slower because it has to transfer first from memory, do the add somewhere in the cpu, then transfer the result back out. And so it would generally be better to avoid that if you're going to be doing anything else with that memory value, and prefer to keep it in the register.”
[19:04] “Looking at a glance on Table-2.21 Reference, why does writing to memory from register generally have higher clocks than loading from memory to register? I assume since register and memory involved on both they must have similar clocks, is there something hidden at work when dest and src are different?”