The Carry-Lookahead Adder design was way more superior to the Ripple Carry Adder design in ti saving.
Basically, instead of waiting for each adder to calculate and pass on whether their carry was a 1, 0, or -1, making the process slow, the Carry-Lookahead Adder design introduced a sort of overview moderator also made with logic gates.
Which, instead of looking at the results of the addition or subtraction of each Adder's two bits, looked at the input bits themselves to see which of them were likely to have a carry at all.
Like a quick overhead glance that tells all the Adders that would otherwise be idle and waiting for a carry, whether to calculate or wait, based on what they held.
It did this by looking at two things — the possibility of generating a carry… and the possibility of propagating a carry.
It was obvious that the only ti a carry could be generated was when both inputs were On. 1 and 1.
So, all that the Carry-Lookahead Adder's 'overhead moderator' logic needed to do was to check every single column of bits to be calculated to see where both inputs were 1.
It would be instantaneous, without any wait ti.
And every calculation that held two 1s would already know that they were passing on a 1 and leaving a 0 as their output, all within the sa mont without needing to physically receive any carry at all.
And for the others that were holding a 1 and a 0, 0 and a 1, or a 0 and a 0, they would fail to qualify for the 'generate' check, because both their inputs weren't 1s.
So instead, the next logic would take over — the propagate logic.
As long as a column could not generate a 1 by itself, it would only be able to propagate — pass on — a 1 instead.
And that could only happen when one of the two input bits was 1 and the other was 0.
So long as a 1 was present, a carry-in would cause it to pass on a carry to the next column.
If both inputs were 0s then even if a carry ca in, the output would always be 1, otherwise it would remain a 0. This was the definite "No carry scenario."
So the logic was: a 1 and 1 = generate, a 1 and 0 or 0 and 1 = propagate, and a 0 and 0 = no propagation.
This ant that within the sa instant, this Carry-Lookahead Adder logic that had an overall view, tells every one of the adders what to do without any need for the ssage to be carried across all the way from the first Adder.
In the small scale of a single group of 64 bits, while there was a difference, it might not really seem like much…
But when one factored in the amount of those 64 bit groups Zephyr would make, all working together in another higher array of 64 again…
The speed benefit would keep widening until it beca drastically evident.
Zephyr got started on it imdiately.
He had left 2 million transistors free for his ALU at the ti he was still designing his logic gates, and now was the ti to use them.
But unlike with his registers that had needed only six transistors arranged in a specific way to form logic that could hold and save a state of a bit — a single spell fragnt… this was different.
His ALU, especially with the Carry-Lookahead Adder design he'd gone for, needed a whole 70 transistors to form the chain of logic gates that would perform the arithtic operations needed to calculate just 1 bit… just 1 spell fragnt…!
It was a whole different ball ga!
It was still very sufficient though, because even at that, it ant he could still process 28,571 Bits of spell fragnts at a ti.
And to form a fully coherent tweaked spell, grouping them into grids of 64 bits, that would still be 446 groups.
Then the sa Carry-Lookahead Adder logic could be applied to these 446 groups, further grouping them by 64 arrays each, making it just 7 arrays he'd communicate with at the largest level.
This would improve optimization and efficiency drastically.
He went about the whole thing, dialing in on the process of creating all these logic gates: XOR, OR, AND, NOT…
This ti, things were different. For every other component he'd designed, he was usually very much in control because of the knowledge he'd gotten from Aegis.
Aegis had only needed to 'look' through his eyes to make slight corrections when so ntal errors were about to take place… and that had been far and few between.
But for this, Aegis was more involved.
They had already developed a blueprint for what needed to be done. It charted out every single connection down to the letter, and Zephyr had it burned in his mory…
But imprecision was still inevitable.
Hence the need for Aegis to be more involved. It created the equivalent of an overlay over Zephyr's 'sight' after Zephyr gave it the permission to do so.
This helped Zephyr visually know where he was making a mistake imdiately, and correct it.
Because of this, the process was extrely ti consuming, as they were moving at a snail's pace.
Etching, pausing to check for correctness, before moving on again for short bit, only to do the sa again… over and over.
It wasn't surprising though.
One had to know that even in the modern world, ALU fabrication, especially for current gen processors, took a very long ti to complete.
Even with all the current technology, it still took months upon months.
Though most of this was due to the sheer amount of prefabrication preparations needed.
Layout designing, extensive simulations, verifications…
It took months of preparation to finalize all this for a modern ALU.
And this was before even fabricating.
The fabrication process itself took its own separate weeks and months alone before moving on to the post production phase…
So the fact that Zephyr was slow wasn't surprising. In fact, by modern standards, he was extrely fast, because he didn't need months for preparation.
Aegis had fully prepared a complete, correct and fitting design for him like it had it beforehand.
And Zephyr didn't even know that what he'd received was not as easy to make within a mont's notice like Aegis had.
Then there was also the fact that Zephyr wasn't etching on silicon wafers. So he didn't need to go through all the trouble it required. He was simply ntally etching onto his mana node itself, actualizing his solidified knowledge into physical form…
And together with the fact that he wasn't going through post production hassles too — he only needed to do minor checks to ensure consistency — he could cut down ti by a very large chunk.
This was why he had estimated fifteen days rather than months…
Zephyr spent the next two weeks in his mana core space, going through the whole process diligently till he completed his ALU design.
After which he spent one more day to check for inconsistencies, ensuring everything was working smoothly, before he booted himself out of the space, fully ntally fatigued.
Imdiately he cracked his eyes open, he felt the ALU he'd just designed settle into his mana core very evidently.
The dark seclusion room where he stayed thrumd and vibrated with mana streams flowing into his mana core.
He felt a state of clarity that massaged his mind, relieving him significantly of his ntal fatigue… almost like a congratulatory gift from origin for the major breakthrough he'd just achieved.
The origin feedback this ti was even more than when he'd comprehended the Bit in the Grand library back in Freehold.
Only that this ti there was no feeling of ecstasy that would drive him into doing sothing mad… There was only clarity and pleasantness.
It continued for a whole minute before Zephyr slowly ca down from that state… and surprisingly he didn't feel any withdrawal symptoms. It was a clean, slow reduction until he cald.
The mana-charged air around him cald to a still and he sat there for a while longer, savouring the aftertaste of the origin gift he'd just received.
His brows furrowed slightly as he analyzed the feeling in his mana core. He didn't know how to describe it, but he felt like his ALU had been given an extra 'corporealness'... an extra depth to it…
He closed his eyes again, knowing he needed to see for himself what had happened in his mana core…
And imdiately he delved into his mana core space, what he saw stunned him.
Reviews
All reviews (0)