SISD, SIMD, MISD and MIMD [Series#2: I am Data!]

I am sure all these terms i.e., SISD, SIMD, MISD and MIMD, must be very unfamiliar by most of us in Data World.

Let’s decode it…..

Single Instruction Stream and Single Data Stream (SISD)

SISD works just like a traditional human. Human can read single letter or word at one time. Same way, old CPUs with SISD method used to read one data one by one.

In this technique, if there is an instruction/ query to find all the ‘A’ written on the page, it will read and return each ‘A’, sequentially. This method takes a lot of time to process 100s or 1000s of data point as each data stream get processed one by one. What if there is Gigabyte and Terabyte of pages of data 😊?

Single Instruction Stream and Multiple Data Stream (SIMD)

SIMD solves this problem of reading each letter or word one by one. It starts processing or reading in parallel where e.g., if there is an instruction/ query to find how many letters ‘A’ is written on a paper, SIMD starts reading all ‘A’ in parallel and can return all ‘A’, in much quick manner as compared to SISD.

Multiple Instruction Stream and Single Data Stream (MISD)

Next challenge was to send multiple instructions to read same pieces of information. In other words, if multiple instructions are sent to read all the ‘A’ then each instruction is processed one by one, sequentially. Means if there are 100s or 1000s of instructions then processing all the instructions to read ‘A’ would take a lot of time.

Multiple Instruction Stream and Multiple Data Stream (MIMD)

Another scenario is to read all the ‘B’ along with ‘A’, means multiple instructions/ queries to ready different set of information. MIMD supports his behavior so how multiple instructions can be processed in parallel to read multiple data element.

Cheers.