23.5.1. Micro-Program Architecture Basics

So, whenever we want to optimize based on encoding or compressing of the signals in each of these cells or each of the words in the program control memory, we call it as a vertical micro-program. In horizontal it is very flat and no optimization is there. So now, what way you can do the only way you can do is you have to encode encoded by mean many way xyz.

In the context of micro-program architecture, vertical micro-programs focus on optimizing the storage of signals by encoding them, while horizontal micro-programs are straightforward with no additional optimizations. Vertical micro-programming reduces the memory needed by efficiently arranging control signals whereas, horizontal micro-programming does not involve such encoding, leading to a less efficient use of memory.

Imagine two different methods of packing boxes. In one method (horizontal), you shove everything into a box as-is without thought, resulting in wasted space. In the second method (vertical), you carefully arrange items, perhaps packing them tightly or stacking them for optimal use of space, thus minimizing the size of the box. This is similar to how vertical micro-programs pack control signals more efficiently.

So, basically first the most formal or the preliminary way of doing it is basically encode the signals in the control memory. So, what do you do? So, whatever signals has to be applied you just encode them and use a decoder.

In a micro-program architecture, encoding means representing control signals in a compressed format held in memory. The signals are then converted back to their usable form through decoders, which translate encoded values back into the signals needed to control the system's operations. This encoding helps reduce the number of bits required, optimizing memory usage.

Think of encoding like using shorthand in a text message. Instead of writing out 'See you soon,' you simply type 'C u soon.' This shorthand reduces the amount of space your message takes up while still conveying the same meaning.

So, how many are there? There are 8 bits which has to be controlled or 8 locations which has to control. So, I just put a 3:8 decoder then, what will be the memory size? The memory size will be 3 bits.

When controlling multiple signals, the architecture uses a decoder to map outputs back to individual control lines. For example, with 8 control lines, a 3:8 decoder can compress this down to 3 input bits. However, only one control signal can be active at a time, which influences the design and speed of the system dynamics.

Imagine a remote control for a TV. The remote has multiple buttons, but when you press one, it takes effect immediately, but only that one at a time. Similarly, the decoder allows only one corresponding control to be active, thus ensuring there’s no conflict in the control signals.

So, now, you have to observe that this is one of the drastic way of compressing. So, instead of 2^n the size will just become n or if it is n you will go for log n upper ceiling.

The significant reduction in required bits due to encoding (from 2^n to a logarithmic scale of n) shows the efficiency gained via vertical micro-programs. However, this comes with trade-offs such as increased timing complexity, meaning multiple steps may be required to activate different control signals sequentially rather than simultaneously.

Think of it like cooking multiple dishes. If you could only cook one dish at a time due to limited space on the stove (similar to the control signals), you’d have to allocate different time slots for each dish instead of having them all ready at once. Although you save time by only using what you need (the ingredients), it requires careful planning and more steps to have everything on the table.

So, what I do? I actually make a cluster 2 clusters. In 1 cluster I will put the PC, in another cluster actually I will put memory address register in.

Hybrid micro-programming combines aspects of both vertical and horizontal approaches. By clustering related control signals, it allows for some parallelism while maintaining efficient encoding. This enables simultaneous operation of specific clusters, thus increasing overall system speed while still conserving memory.

Consider a team of chefs working on a buffet. Each chef is in charge of a different dish (cluster). Some dishes need to be prepared at the same time, so they’re grouped accordingly. While they collaborate efficiently, each chef still focuses on their recipes, optimizing both time and resources.

So, if we have a lot of 0s then such problems will become in such some kinds of number of stages will be less, but if we have more number of 1s then making them simultaneously 1 is not possible.

Vertical micro-programming can save memory and promote optimization but can introduce time delays when requiring multiple active control signals. Additionally, high sparsity (more zeros) can benefit this approach, while higher density (more ones) can complicate execution and timing.

If you think of setting up a concert with a lot of empty seats (zeros), you can move people around easily, making quick adjustments. Conversely, if the seats are mostly filled (ones), making updates becomes cumbersome, as you can’t just shift people without causing a disruption.

--