malivinayak

How an LSTM updates its memory

Vinayak Mali — Tue, 02 Jun 2026 15:54:54 GMT

What we are going to see

Today we are going to deep dive into the LSTM circuit. How exactly the calculations work and how LSTM forget the old information and update the new information.

For the sec of clear understanding we are going to go as per 2 sentences

John is a doctor.
Mary is a teacher

So complete sentence became John is a doctor. Mary is a teacher.

Our job is to predict the gender on each word using LSTM

Why I am writing this

One of my junior ask me a question

Heyy I'm reading this blog by colah on LSTM...I have some doubts...

                (LSTM Diagram is attached)

So here we have 3 things 
1) C(t-1) cell state long term memory 
2) H(t-1) info from last iteration 
3) Xt current input

Lets say we have 2 sentence 
John is a doctor 
Mary is a teacher 

Now let's say C(t-1) had stored things like 
    subject=john, gender = male
Current input is Mary...
now model wants to forget gender = male... 

So we take f(W(h(t-1),Xt))...
means we applied sigmoid on current input and past iteration info...
which will generate vector.. let's say [0,1,0,0,1] 
then we multiply it with Cell state...
and it will forget some things, remember something based on 0 and 1...
now my que is...
how the hell... it calculated what to remove in cell state 
by looking at current input

Understanding how it works is a one thing and explaining it effectively is completely different thing.

So today I am going to try to explain this with the help of this blog and same example. I am also in learning phase, so if any there is any mistake please feel free to address in comment section.

Some thermotical stuff before mathematical

☑️ Pre-requisite: I am assuming you have basic understanding of how LSTM work.

In LSTM we have 3 gates:

Forget Gate - What information should be forget/erase from cell state
Input Gate - What new information should we update in input state
Output Gate - What information from current cell state pass to new hidden state

So, just to be clear output gate doesn't play any role in updating cell state. Cell state is affected by Forget Gate and Input Gate.

Does Forget gate and Input gate have idea what's in cell state ❓

➡️ NOT REALLY ❌

Forget gate never looks at cell state. Lets take a look at equation of forget gate

$$f_t = \sigma \left( W_f \cdot [h_{t-1}, x_t] + b_f \right)$$

If we observe equation carefully C_t-1 previous cell state is not present. It means forget gate does not have any idea about where the gender value stored in memory. It is blind for what present in cell state.

Then question remain same "how did it know what to remove by looking at the input❓"

➡️The answer is it didn't know. It's not reading the cell state and reasoning "ah, gender is in there, let me clear it." It's just emitting a vector of gates from the input alone.

Then why does the mask happen to zero out exactly the gender slot?
Because the weights W_f are learned during training.

The read and write for particular cell is agreed while training as this will be a gender cell. It is not mention anywhere its just understanding due to weights calculation and backpropagation. When gender is predicted for "Mary is a teacher" and if output came as male then that loss is back propagated, weights are updated. LSTM learns that gender slot should be cleared when input word Mary a new subject came. The embeddings of input carry information of, if the current input is subject or not. If yes then what is its gender male or female.
So the next time input patterns resemble LSTM forget gate knows I have to forget the gender information and input gate knows I have to write the new information there.

Small Mathematical Example before actual calculation

Lets say Mary word came. In LSTM

0: Forget
1: Keep the information

Assuming 3rd cell is as gender cell.
Positive value represent male gender and
Negative value represent female gender.

$$C_t = {f_t \odot C_{t-1}} + {i_t \odot \tilde{C}_t}$$

state	1	2	3 (gender)	4
Old Cell : C_t-1	0.6	-0.3	+0.9 (male)	0.5
x Forget	1	1	0	1
= erased	0.6	-0.3	0	0.5

+ write	0	0	-0.8	0
= new cell : C_t	0.6	-0.3	-0.8 (Female)	0.5

So here you can see forget gave value 0 for 3rd cell and the previous value get erased from cell state.
Later at input gate:
i_t : provide at which cell the value should be updated and with what
C~_t : proposed value to update

Mathematical stuff for complete sentence

Concerted input z

$$z = \left[ h_{\text{prev}},(3\ \text{dims}) \middle| x_{\text{embedding}},(3\ \text{dims}) \right]$$

So, h has h1, h2, h3 and x_embeddins are x1, x2, x3 for input word

Example:

Each gate has one weight row per memory slot. column 3 controls the gender slot. Therefore h3 will represent gender value upto the current word. In this example h_3 is 0.58 it means the current statement has male subject so 'He' can be used. if value is <0.5 then gender will be predicted as female so She can be used.

Now let's see weight vectors for respective gates along with bias

W_f : Forget Gate with
Bias: +3
Here x_1 will be -5 as we want to forget gender cell

0
0
0
-5
0
0
h1
h2
h3
x1
x2
x3
W_i: Input Gate
Bias: -2
Here x_1 will be 5 as we want to update gender cell

0
0
0
5
0
0
h1
h2
h3
x1
x2
x3
W_c: Candidate
Bias: +0
Here x_2 will be 5 and x_3 will be -5 as male/female value will be decided

0
0
0
0
5
-5
h1
h2
h3
x1
x2
x3
W_o: Output Gate
Bias: +2

0
0
0
0
0
0
h1
h2
h3
x1
x2
x3

For consistency and for better understanding I have generated interactive html file from Claude by giving rough calculations.

Sentence : John is a doctor. Mary is a teacher.

Now lets see how exactly gender cell value is calculated.

Website Link: Click Here

URL: https://malivinayak.com/LSTM-Gender-Prediction/

THANK YOU 😊

Resources:

20 LLM Interview Questions to Test Your Readiness

Vinayak Mali — Tue, 02 Jun 2026 04:17:44 GMT

If you can’t answer half, you’re not ready 👇

LLM Fundamentals

1. Explain tokenization beyond “splitting text”.

2. Why do decoder-only models dominate? When would you prefer encoder-decoder?

3. Walk me through attention. Where do queries, keys, values actually get used

4. Sampling strategies: top-K vs top-P vs temperature. When do you pick which?

Prompting & Context Engineering

5. Give me an example of a zero-shot prompt failing. How did you fix it?

6. How do you version and test prompts for stability?

7. Explain “context window waste.” How do you mitigate it?

Fine-Tuning & Alignment

8. LoRA vs QLoRA vs PEFT - where do you trade off memory, speed, accuracy?

9. What are the limits of RLHF? Give one real failure mode.

10. When would you choose open-source over proprietary LLMs for fine-tuning?

RAG (Retrieval-Augmented Generation)

11. Walk through how embeddings + similarity search actually work.

12. Chroma vs Pinecone vs Weaviate - what’s your POV?

13. When does hybrid retrieval outperform vanilla vector search?

Agentic AI

14. Define an agent without buzzwords.

15. How do you handle tool-use errors in an agent loop?

16. What’s the hardest part of multi-agent orchestration?

MLOps / LLMOps

17. How would you monitor hallucinations in production?

18. What metrics do you track in an eval pipeline?

19. How do you deploy a GenAI service (FastAPI + Docker + K8s) with rollback safety?

Scaling & Risk

20. Your system just blew the budget because of model size. Walk me through trade-off options in 3 minutes.

🔗

Source: LinkedIn Post

👉 Self-test: Can you answer these out loud, clearly, without hand-waving? If not, that’s your prep gap.

Journey of ML Algorithms

Vinayak Mali — Tue, 02 Jun 2026 04:12:42 GMT

Warm welcome to everyone. You are going to join one of the best journey in ML Algorithms.

If you are already a companion of compiler journey then you already know the stop concept in our journey. For those who are joining us first time lets start with the journey manual. In this journey we are going to take a time to time stops. At each stop we are going to learn a machine learning algorithm. Feel free to ask question at any point of time in the journey. For this you can use the comment section present below. I will try to whisper there the answers and if required then we will take a small stop (blog) to properly resolve the query.

In this journey at each stop we are going to either understand Discriminative algorithm or Generative algorithm. Each algorithm may required one or multiple stops.

💡

Small visiting points list of our journey

ML Algorithms
    - Discriminative
        - Classification
            - Logistic Regression
            - Decision Trees
            - Neural Networks
            - SVM
            - k-Nearest Neighbor
        - Regression
            - Linear Regression
            - Decision Trees
            - k-Nearest Neighbor
            - Neural Networks
    - Generative
        - Text Generation
            - RNNs (GRU, LSTM)
        - Image Generation
            - VAE
            - GANs
            - Diffusion Models
            - Autoregressive Models
            - Flow-based Models
        - Audio Generation
            - VAE
            - Autoregressive Models
        - Video Generation
            - Diffusion Models
            - Autoregressive Models

Note: I may update the list time to time as per requirements current world as well as requirements of you as a traveler

Wait but haven't we all learn ML algorithms have mainly 3 types:

Supervised Learning
Unsupervised Learning
Reinforcement Learning

Then why aren't we seeing this kind of type in our journey.

Ok then, I will tell you how exactly ML algorithms are

ML Algorithms
    - Supervised and Unsupervised
        - Discriminative
        - Generative
    - Reinforcement Learning

So, In this journey we are going to learn algorithms one by one and every stop. At some stops we are going to take numerical drinks or coded food so that we can learn not just theory but also what's happening behind it.

🚍 LET'S BEGIN THE JOURNEY TOGETHER.

See you at the next stop, till then stay tuned 😊

What is Agentic AI ?

Vinayak Mali — Fri, 17 Apr 2026 17:42:32 GMT

Lets first understand how you complete your task using LLM

→ In order to get some task done you give repetitive prompts to LLM model, may be to improve the answer or to change the approach or to correct the mistake.

In agentic AI rather than repetitively prompting LLM you gave it a goal and then next important thing you give is tools by which it can do/test things by its own.

So in simple words:

Agentic = LLM + Goal + Tools

This will start doing the task if fail then automatically change or improve the approach. It will fix the bug created during process. It will continue self repetitive improvement till goal gets achieved.

Agentic AI is a dream of AUTONOMY.

😊-X- THANK YOU -X- 😊

CUDA Configuration for Windows

Vinayak Mali — Wed, 08 Apr 2026 09:43:19 GMT

Step 1: NVIDIA Video Driver

You should install the latest version of your GPUs driver. - Check which GPU is present in your system

You can download drivers here:

NVIDIA GPU Drive Download

How to check if already present

1. Device Manager Press Win + X
2. click Device Manager 
3. Expand Display adapters 
4. Look for something like: NVIDIA GeForce / RTX / GTX…

Step 2: Visual Studio C++

You will need Visual Studio, with C++ installed.

🛑IMP: By default, C++ is not installed with Visual Studio, so make sure you select ALL of the C++ options.

Visual Studio Community Edition

Step 3: Anaconda/Miniconda

You will need anaconda to install all deep learning packages

Download Anaconda

Step 4: CUDA Toolkit

Which version to choice:

Go to https://pytorch.org/get-started/locally/

Check which stable version is available for pytorch: example CUDA 11.8 and CUDA 12.1 but not CUDA 12.4

Recommended as per following image: So you can ether install 11.8 or 12.1 [If you plan to use pytorch]

Download CUDA Toolkit

How to verify if downloaded or not: C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA Here you will see the downloaded version folder

Step 5: cuDNN

GPU accelerated library of primitives for deep neural networks

Download cuDNN

Download the version which is suitable for your CUDA version

Next Step:

Unzip the downlaoded folder

Copy all bin folder files from cuDNN to C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v12.8 bin folder --- [Note I have CUDA 12.8 so by dir path include v12.8 check yours]

same copy paste all files for include and lib folder too

How to verify if everything is correctly done

Check for the environment variable set or not

If it is set automatically then everything is perfect if not then set manually

Step 6 Verification if system is using CUDA

Install PyTorch

select the cuda version you installed and all other options as per requirement

Recommendation: You can test this in virtual env too

Install PyTorch

Run the following script to test your GPU

import torch

print("Number of GPU: ", torch.cuda.device_count())
print("GPU Name: ", torch.cuda.get_device_name())


device = torch.device('cuda' if torch.cuda.is_available() else 'cpu')
print('Using device:', device)

-x- 😊 THANK YOU 😊 -x-

How to move Towards PhD Life 😊

Vinayak Mali — Sun, 11 Jan 2026 11:49:41 GMT

Hello readers, today I am writing the article for you considering my point of view. As I am going to complete my masters form IIT Kharagpur in few months, I am currently collecting the information to PhD studies.

“Moving Towards our beautiful PhD Life“ is raw information article and this is collection of information after listening multiple IIT students/videos.

Sorry for the unstructured format. I am watching multiple videos and listing to multiple IIT friends to collect the information and updating in one place. Once enough data is get collected.

Now lets move towards actual content but before that

Please Not that the information is not certain as different things are coming in view from different people. This is blog is genuine attempt to generalize and collect things together. THANK YOU 😊

Requirements

CGPA matters in academics. It directly reflects how good you are in studies atleast in academics. It is also related to seriousness. If your CGPA is not that good then be serious make a strong research profile and then start looking for the professors. Important is research.

CGPA > 8 will be fine

PhD from India or foreign?

If you get PMRF scholarship then getting PhD from India is like more beneficial. With the PMRF scheme you will be reviewed by all the professors from across IIT’s.

PhD from wither India or Foreign doesn’t matter when you have very strong research.

If foreign then which place is good?

Most best place is Europe.

Germany is best as there is less researchers and they are looking for researcher. [Start learning German from now]. First look for guide. German government is more supportive. Tuition fee is less than other countries. German is must and you have to complete A2 level course.

Switzerland is also good place.

Why PhD ?

I want to do PHD
Want deeper research in domain and enhance the knowledge of that domain

Benefits of PhD ?

More Growth opportunities than Masters Degree
Tremendous knowledge of particular domain
Patience
To join top research institute like IIT’s as a Professor
In senior positions highest educational degrees contribute for promotion

PhD Fee Structure ?

25,000 ₹ per semester
- In Part-Time PhD fees
  - Self-Sponsored: You yourself pay the fees
  - Company-Sponsored: Company where you are employed will pay the fees
    [Need to sign bond as you will serve to company for particular years after degree completion]

Types of PhD ?

Full Time PhD
Part-Time PhD
1. Self Sponsored
2. Company Sponsored

Full Time PhD - Student staying in campus and delicately doing research under supervision of professor

Stiped is getting from government of India or MHRD (Ministry of Human Resource Development)
Or you can join as project PhD also. Then you are part project. Your thesis will be part of project. Your stipend will be coming from that project funding only.

Part Time PhD - Working in the company and at the same time you are pursuing your PhD parallelly under supervision of some guides

In College:

Apply with research proposal and application along with NoC.
Appear for written exam interview in campus.
Requirements and Procedure:
1. You should be in R&D role in company and minimum 2 years in company
2. Company should be well known and should be existing more than 10 years in market
  Why this is matters is PhD is going to take next 4-6 years and company is going to pay indirectly complete fees. [Company gives to you and you pays fees]
  There is agreement between institute and company like you will be working for company and you are doing part time PhD
3. Managers approval is very very important
4. Talk with professor and check which professor is interested in part time PhD

Application Process

Recruitment happens twice in the year/Application opens twice year 1. July 2. December

Documents required
- Sponsorship Certificate/Agreement from company [If Part Time PhD]
  States that company is going to take care of all expense. e.g. college fees, if paper is going for conference then that expense, travelling cost, staying cost
  There is pre defined form for all institutes we have to take sign on that from company
- Industry guide CV
  There is another guide from company end. He is the most of the time manager your project
Research Profile. You should be in research profile in company
Research proposal
Most if the time guides are not interested in reading your proposal so you should talk with them or there research scholars.
Also relate and mention how the professors work or part of work is related in you work/proposal

Written Exam:
[For part time students most of the times no written test]

20 Subjective questions all from gate syllabus. After exam Interview.

Interview:

Multiple panels, each with 8 to 10 professors. [with your guide too]
Answer back to back questions on board.

What after selection

First 6 to 10 month we have to do course work.

IIT KGP - 4 courses
ISSC - 3 courses

For part time students NPTEL courses
Some students take 6 months of leave, complete the course and then join the company again. They talk with supervisor to setup with MTech and BTech students. Being an PhD student you just formulate the problem and then the MTech BTech student work on it, implement, generate result and then you write then paper. It is kind of team work thing.

How to choice guide ?

Shortlist the professor which align with your research domain. Connect with the student who are working with that professor. Ask the students for the feedback about guide.

Que to ask to students working with professor:

How much support they give
How much time he spent/talk with you about project
DO they are asking for extra work which is not at all research work
are they inspiring you or torturing you

Choice the guide only and only all are green ticks.

If any professor is reading this article then my senior apologies if you are getting offended in any way. This article is with perspective of students and everything mentioned is common in multiple PhD students perspective.

You should have complete information about the guide. Remember the guide should be very supportive. The bonding between guide and student is very important.

Guide can either make you or break you.

Once you finalized one or two guides, mail them like sir this is my PhD perceptive, if they give some positive feedback then only it make sense to choice the guide.

What requires to complete the PhD ?

minimum 2-4 research paper in A* conference
Quality matters. Incremental changes are not good.
Thesis Defense

How to complete PhD on time ?

First of your goal should be clear. You should have clear path like you have to do in first year this , in second year this. like that.

Also keep the guide in loop like you are going to do this, are there any suggestion or shall I move forward. You need the clear communication that this is my AIM and after that I am expecting the PhD.

Make sure your synopsis is completed in 3 or 3½ then another 6 months to write thesis.

What require to become IIT Professor

PhD Degree is mandatory. If you have post doc experience which is from outside India then it add more value in your application.

Complete PhD with certificate
Post Doc around 1 year

The detailed information will be some other article/blog.

Resources:

My IITian Friends, PhD scholars, Friends and every person who helped be getting any part of information.
PhD from IIT Kharagpur while working at Qualcomm || PhD for working Professional || Part time PhD
RF Engineer at Qualcomm : How to become IIT'S Professor :-with Priyansha , Assistant Professor @ IIT Roorkee
Completed PhD in 3.1 year and then start postdoc in 3½ years then defense and convocation is completed in 4 years.
Selection Procedure for IIT'S Professor :- with Priyansha , Assistant Professor @ IIT Roorkee
HOW to Pursue a PhD While Working - Kishalay Das

GenAI-1: What is a Language Model?

Vinayak Mali — Thu, 18 Sep 2025 04:25:23 GMT

We all heard about LLMs (Large Language Models), didn’t we ? but before understanding Large Language Models What is Language Model ?

The language model is probability distribution over a sequence of tokens. The tokens can be anything like words, characters or subwords in language. What it mean by probability distribution over sequence of tokens. Let’s understand it with simple example. Consider the words ‘the’, ‘flat’, ‘rat’, ‘ran’, ‘on’. The language model would give higher probability to the more sensible sentence due to greater likelihood of real-world usage and logical coherence.

P(‘the’, ‘rat’, ‘ran’, ‘on’, ‘flat’) = 0.03 P(‘the’, ‘rat’, ‘on’, ‘flat’, ‘ran’) = 0.02 P(‘on’, ‘flat’, ‘the’, ‘rat’, ‘ran’) = 0.01

Language models are applied to a wide range of tasks, from classification to generation.

Now, its time to explore generative language models more. So, the next question is

What is Generative Language Models ?

Thank You 😊, fellow traveler, for joining us on this Que Ans journey.

Adventures Journey of Question and Answer

Vinayak Mali — Thu, 18 Sep 2025 04:17:56 GMT

Hello Fellow travelers, You all have decided to join this journey. This will be really long journey, we will understand every spot of concepts one by one.

You might catch a little sleep, but don’t worry—I’ll keep you notified at every adventures Que, if you'd like.

You can subscribe to get notified. 😊

How to install conda in Ubuntu

Vinayak Mali — Mon, 18 Aug 2025 04:56:23 GMT

Install Miniconda (lightweight Conda)

1. Download the Miniconda installer:

wget https://repo.anaconda.com/miniconda/Miniconda3-latest-Linux-x86_64.sh

2. Run the installer:

bash Miniconda3-latest-Linux-x86_64.sh

⚠️ Follow the prompts — accept the license, choose install location, and allow it to initialize Conda (recommended).

3. Restart your shell (or run this now to enable conda):

source ~/.bashrc

4. Check where Miniconda is installed

By default, it installs to your home directory, under:

~/miniconda3

Verify it exists:

ls ~/miniconda3

If you see folders like bin, etc, envs, then it’s installed correctly.

5. Manually add Conda to your PATH

Run this command:

export PATH="$HOME/miniconda3/bin:$PATH"

Then test:

conda --version

If it works, proceed to initialize Conda:

conda init

6. Make the PATH change permanent

To make the PATH available every time you open a terminal, add it to your .bashrc or .zshrc:

echo 'export PATH="$HOME/miniconda3/bin:$PATH"' >> ~/.bashrc
source ~/.bashrc

😊 Thank You 😊

Compiler - Why and Where

Vinayak Mali — Fri, 19 Apr 2024 05:11:26 GMT

So finally we are here, this is our First Stop of compiler journey. Are you excited for this stop ? Off course you are, RIGHT ! ! !

I guess title had already gave you hint of todays stop. If yes then you got me there otherwise let me explain it first then we will continue our talk. Also in subtitle why speedbreaker, because in order to understand the concepts clearly we need to slow down our speed and learn everything patiently. In this way we will not get a sudden jerk of new things.

Now, at this moment we are going to understand why exactly there is need of compiler and If we consider language processing diagram where is the compiler located . . .

Following is the plan for our stop :

Why and where
Understanding semantic gap
How to overcome it

Why and where

If you are aware about steps of language processing then you might already know that compiler take pure high level language as input and generates assembly language as output. So, we can conclude that compiler is located after pre-processor and before assembler. Now, we got the answer about WHERE but what about WHY ?

Lets travel back to get some insights . . .

Semantic gap

Grace Hopper is the one who first designed the compiler for A-O programming language in 1952. and if you are aware then he is often credited to pioneering the concept of compiler.

If you are thinking I am going to give you some history lectures then it's not going to happen, as I also get bored of it too. But it is good to know some of it and we will get back to the above sentence may be in future stop.

Imagine you want to design the software application, all you worried about is its behaviour. In order to execute that software or machine we need to write it into the machine language which is quite challenging.

HOLD TIGHT . . .

It may start confusing, so what we are going to do is analysing everything slowly one by one:

You have an idea and want to develop an application/software. 2) This application should run/execute on computer system. 3) To run this application correctly it should have certain rules, cases or conditions. Now, let's break it down...

Application domain - The ideas and behaviour of the application that you want to develop
Execution domain - To implement ideas and execute, ideas should be interpreted for execution and that is execution domain
Semantics - Semantics means the rules of meaning of domain
Semantic gap - There is always a difference between our ideas and what we have implemented. this difference is nothing but semantic gap.

🖥

Let's take a look on example:

🛑 WARNING: Don't try to understand the code just take a look

You want to develop simple addition program:

Application domain
Take two numbers, add them and then stop

Execution domain
I am simply trying to visualize the machine code to you as per the architecture.

  Instruction: Load the first number into register 1
  Binary: 0001 0000 (Assume 0001 represents the load operation and 0000 represents register 1)
  Decimal: 16

  Instruction: Load the second number into register 2
  Binary: 0010 0001 (Assume 0010 represents the load operation and 0001 represents register 2)
  Decimal: 33

  Instruction: Add the contents of register 1 and register 2, and store the result in register 1
  Binary: 0011 0001 (Assume 0011 represents the add operation and 0001 represents register 1)
  Decimal: 49

  Instruction: Halt the program
  Binary: 1111 1111 (Assume 1111 represents the halt operation)
  Decimal: 255

  section .data
      num1 dd 10       ; First number (32-bit integer)
      num2 dd 20       ; Second number (32-bit integer)

  section .text
      global _start

  _start:
      ; Load num1 into eax
      mov eax, [num1]

      ; Add num2 to eax
      add eax, [num2]

      ; At this point, eax contains the sum of num1 and num2

      ; Exit program
      mov eax, 1       ; syscall number for exit
      xor ebx, ebx     ; exit code 0
      int 0x80         ; invoke syscall

This is not the machine code its just for visualization purpose

Because of Semantic gap you need to face many consequences, I am listing some important ones only:

It take very large development time
It also need large development efforts
Final software quality will be poor

The consequences arise because we try to develop application directly in machine language. and even if we achieve developing the executable software along with above issues it is really hard to make even a small change.

Solution

In order to tackle this issue we use software engineering approach by using methodologies and Programming Languages (PLs). So the the process will be in two steps:

Specification, design and coding step (software developer who will use programming language)
PL implementation step (Programming language designer)

Now we are going to take new domain which is ' PL domain '
Also by introducing new domain we bridged the semantic gap by engineering steps. We will understand what are this gap's are but first try to visualize what just happened.

The first step bridges the gap between application and PL domain and the second step bridges the gap between PL domain and execution domain.

Gap between Application domain and PL domain is specification and design gap or simply Specification Gap
Gap between PL domain and Execution domain is execution gap.

PL stands for Programming language

How it works

Software developer team will bridge the specification gap by taking specifications and requirements required for application. Then developer will develop the software with the help of any programming language. Now the important part is what about execution gap. But first what it mean by execution gap.
Execution gap is exist because developer will develop the application in programming language like c, cpp, python, ... etc. But machine can only understand only machine language then how we are going to resolve this issue. This issue is comes under execution gap.
The execution gap is bridged by the designer of the programming language processor, viz a translator or an interpreter.

Specification gap - semantic gap between two specification of same task
Execution gap - semantics of program(that perform same task) written in different programming language.

Semantics can be referred as rules

We are assuming a specification language (SL) for every domain. So specification written in SL is nothing but program in SL. e.g. specification written in C language is nothing but program in C.

Domain	Specification Language
PL Domain	PL (Programming Language) e.g. C, CPP, . . .
Execution Domain	Machine Language of computer system

💡

Let's take a look on example with PL domain:

🛑 WARNING: Don't try to understand the code just take a look

You want to develop simple addition program:

Application domain
Take two numbers, add them and then stop
PL domain
We will use C programming language

#include

int main(){
    int a,b;
    scanf("%d",&a);
    scanf("%d",&b);
    a = a + b;
    printf("%d",a);
    return 0;
}

Execution domain
Machine code based on compiler of C programming.

Now you can see that we can easily make modifications with PL domain.

So far we have understand that, to develop an application we need to take help of programming language which further converted to machine language somehow.

Our task is to understand that somehow.

🚏

THAT'S IT, WE HAVE TO GO NOW FOR NEXT STOP IN ORDER TO CONTINUE OUR JOURNEY.

#FAQ

How does machine code is generated ?

In a real-world scenario, machine code is typically generated by an assembler, which translates assembly code into machine code specific to the target architecture. Writing machine code directly is extremely complex and error-prone. We already know from language processing diagram that compiler generates assembly language. This assembly language is used by assembler to generate machine code.

As of today 19-04-2024 who are enjoying the compiler journey ?

As of today's date 19-04-2024 Aishwarya Patil & Paul Bogatyr are enjoying our compiler journey. I would like to thank them for being my companion.

Where can I find the complete journey of compiler design ?

You can go through the complete compiler journey by clicking HERE or simply pasting following url in any browser.

https://malivinayak.hashnode.dev/series/compiler-journey

Is there anyway I could get notified at every STOP of journey ?

There is always a way. You can subscribe to newsletter service in-order to get notified at every STOP. You can find it into the navbar on top right or at the bottom of page.

Journey of Compiler

Vinayak Mali — Tue, 16 Apr 2024 04:57:22 GMT

🚏

We are waiting at the platform to board the compiler bus to start the journey.

Before starting the journey let's take a moment to understand what is this journey, why we are going anyway, where we will end, . . .

What is this Journey of Compiler ?

Journey of Compiler Design is going to be the beautiful tour for all those who want to understand and learn about compiler. Who can travel with me - All of you who may have already learnt about compiler design or compiler construction subject in your bachelors' or may be those who are going to learn these subjects in future or maybe you are just enthusiastic and want to gain knowledge or just want to travel along with me for cheering me up.

On this journey we will get multiple stops.

WAIT.... WHAT

Did I just heard STOP

"YES", you heard me right. We are on journey, Journey to explore the world of compiler from inside, outside and around. For this we have to stop from time to time to gain experience via eating some knowledge, drinking some concepts and digesting some applications. After this we will be energized and ready to proceed to the next stop where we will repeat the same with some progress.

Why we are going anyway ?

As we discussed earlier this journey is for many types of companions. While travelling with me you can learn new concepts or revise it if you already know. The beautiful part of this journey is you will get to understand the application based conceptual knowledge step by step on each and every stop.

Where will it end ?

I am not sure about ending stop as we are on never ending journey. In this computer and AI era there is enhancement in all part at every moment even at this moment also someone somewhere on something is working to innovate new or enhance old. In order to remain upto date we can't end this journey. We will be together from prerequisites to basic, from basics to advance, and from advance to emerging technologies related to compiler.

🚍

LET'S BEGIN THE JOURNEY TOGETHER.

Characteristics of C Programming

Vinayak Mali — Mon, 09 Oct 2023 11:40:43 GMT

There are 6 characteristics:

Keywords
Operators
Separators
Constant
Predefined Functions
Syntax

Keywords

Keywords are words that have special meaning to the C compiler.
An identifier can't have the same spelling and case as a C keyword. We can't use keywords for our own purposes.
There are a total of 32 keywords in C which are as follows:

auto	break	case	char
const	continue	default	do
double	else	enum	extern
float	for	goto	if
int	long	register	return
short	signed	sizeof	static
struct	switch	typedef	union
unsigned	void	volatile	while

Operators

Special types of symbols i.e. are used to perform specific task
Types: Binary operator, Unary operator, Ternary operator, etc.
Total 44 operators are present

Separators:

Separate things called as a token
we can separate individual unit -> token

What is Token