Cache coherence problems software#

Thus, in addition to copying the requested data into the cache, the system may also “prefetch” other pieces of data that may soon be of interest. For example, when a program accesses one piece of data from the main memory, it will oftentimes wish to access other pieces of data located in close proximity (that is, in nearby memory locations). There are all sorts of tricks involved here. If the required data is in the cache, we call this a “cache hit.” If the required data is not in the cache, we call this a “cache miss.” When the CPU wishes to store the result from a computation, that data may be written to both the cache and the main memory at the same time, in which case we use the term “write-through.” Alternatively, the data may be initially written only to the cache, with writing to the main store being postponed until that portion of the cache is about to be overwritten, in which case we use the term “write-back” (or “write-behind”). If the CPU requests the same piece of data a second time, it will come from the cache (assuming this entry hasn’t been overwritten by something else in the meantime). For example, when the CPU first requests a piece of data from the main memory, that data will be simultaneously loaded into both the CPU and the cache (not “through” the cache as is implied by my illustration). It should be noted that the above illustration is a gross simplification at best.

In fact, the cache is usually considered to be part of the CPU the reason I’m showing it as being separate is to simplify the diagrams that are to come. The cache is a relatively small amount of memory that’s implemented in the fastest technology available (“damn the power consumption”) and that’s located as close to the CPU as possible. Mainframe without cache memory (left) and with cache memory (right) Thus was the idea of cache memory born, where this name was derived from the French word meaning “a secure place of storage” (noun) or “to store something in a secure place” (verb). The problem is that, even if the main memory was located in the same cabinet as the CPU, it took a relatively long time for data to be read or written and to travel back and forth between these two subsystems. That technology was derived from World War II radar memory

They literally weighed a ton when filled with mercury. You can bet those things weren’t in the same cabinet. Oh, and the first UNIVAC machines used mercury delay lines and acoustic pulses as shift-register memory. I recall seeing a SAGE setup, probably at the Computer History Museum, where the magnetic core planes were presented in a separate, standalone cabinet (see the image in this column ). These were the first and probably second computers to use magnetic cores. The other two computers were the Whirlwind and the AN/FSQ-7 machines, where the latter were used in SAGE. This storage technology predated magnetic cores. In this case, storage was provided by phosphor persistence on monochrome cathode ray tubes (CRTs). The first was the Manchester Baby, which used Williams tubes for RAM. Hi Max, I can think of three examples where the main memory was not in the CPU cabinet. Lest you doubt me on this point, I just asked my chum Steve Leibson, who is Principal Analyst and Partner at TIRIAS Research, if he knew of any examples off the top of his head, and this font of all knowledge replied as follows: In some systems, the main memory was hosted in a separate cabinet that sat next to the CPU. The original use of the term “mainframe” was derived from the large cabinet, called a main frame, that housed the central processing unit (CPU) and the main memory (later, the mainframe moniker was used to distinguish high-end commercial computers from less powerful machines.) This all began with the mainframe computers of yesteryear (cue harp sound effect and imagine using your temporal transportation tool of choice to go back in time ). Let’s also start off by reminding ourselves how the whole cache conundrum came to be in the first place. I don’t know about you, but I’ve forgotten already (I don’t like to boast, but I’m getting better and better at forgetting things the older I become).

Cache coherence problems software#

Since I’m a hardware design engineer by trade, and since this column largely focuses on maintaining the coherency of hardware caches, let’s commence by forgetting that beasts such as software caches even exist. In computing, the term “cache” refers to a hardware or software component that stores data in such a way that future requests for that data can be served faster.