saksham

@irunonironic whats so good abt this friday 🫩

today is good friday

Just finished building tesseract, a cache-aware, domain-specific memory allocator designed for high-throughput, repeated allocation workloads (think request-style systems, tight loops, simulation-style patterns). General purpose allocators like malloc are designed to handle all kinds of workloads. It is designed for a specific one: ~repeated allocations ~predictable sizes ~high-throughput systems It uses: size classes + freelists pre allocated arenas, thread-local behavior (to reduce contention), Micro-benchmark (1M random size allocations): tesseract faster than tcmalloc (1.52x faster), jemalloc (3.57x faster), and malloc (3.68x faster). 14.19ms vs 52.21ms. Real workload (Web server, 10M http requests): Single-threaded: tesseract 247ms, malloc 962ms. 3.89x faster. Multi-threaded (8 threads): tesseract 44.85ms, malloc 184.86ms. 4.12x faster. also its optimized for small and repeated allocations, not ideal for large or highly irregular workloads not numa aware , huge page experiments didn’t show improvement for this workload. [repo link in replies]

@irunonironic 🥰

@skshmgpt haww

good evening

@Zyara_1ot career update in this economy 🫪, gf bf se zyada mushkil to vo hai 💔

@aayushman2703 kitna khrcha aaya, both amp and guitar ?

Amp acquired 🔥🔥 Now i just have to hold my horses till Monday for the actual guitar to arrive 😭😭 This thing is top of the line for starters btw, completely unnecessary. This go big or go home mentality is gonna kill me

@Zyara_1ot nice

@skshmgpt Yepp yepppp Multithr... Eacg thread gets its own freelist cache .... When ever need it pops from its own list...

Multi-threaded (8 threads): tesseract 44.85ms, malloc 184.86ms. 4.12x faster btw......... (i will post abt it later, more clrly, rn im a lil tensed abt smth 😭ahh lyff ) repo link in replies , you can take a look at the readme...

@jerkeyray @pixperk yes

@skshmgpt @pixperk you're from gzb asw? (yaha nahi karna dox)

i miss whatever i was doing back then

@jerkeyray @pixperk gzb me kidhr

@pixperk ghaziabad :)

@bit2swaz inka bhi shi hai yaar

car

Trained a single layer neural network in C to predict AND GATE from scratch using gradient descent and Batch training of dataset in almost 300K epochs github.com/aadityansha06/…

@adhyayana108 just learning or building something ?

>socketio internals >custom events with data, strings, num, obj, array >ackn with callbacks for req-res pattern >broadcasting to all, others & specific clients by socket id >room management, join, leave, msg to room, get room info >user registration & online presence tracking

@irunonironic 👍

dont talk to me

@irunonironic parcel it to me

@skshmgpt khao

aaj paneer banaya😋😋 recipe batao apni apni , kaise banate ho i'll try it next time

@irunonironic alr did, a lots of it, now we work

@skshmgpt sleep

one of those days where i don't wanna do anything and jus sleep

@irunonironic riding harleys in hawaii

you and I, I