Devji C.’s Post

oldie but goodie! "Hash tables consume a large volume of both compute resources and memory across Google's production system. The design for hash tables in C++ traces its origins to the SGI STL implementation from 20 years ago. Over these years, computer architecture and performance has changed dramatically and we need to evolve this fundamental data structure to follow those changes. This talk describes the process of design and optimization that starts with std::unordered_map and ends with a new design we call "SwissTable", a 2-level N-way associative hash table. Our implementation of this new design gets 2-3x better performance with significant memory reductions (compared to unordered_map) and is being broadly deployed across Google. " https://lnkd.in/d4BhASz7

New way to consume LLM API calls Anthropic just released "Message Batches API" https://lnkd.in/d2dbh55i you can use the Batches API to submit groups of up to 10,000 queries and let the processing at a 50% discount. Batches will be processed within 24 hours. (High throughput at half the cost for the right workloads). I feel that Anthropic is taking the lead in the LLM Race , started with prompt caching, Anthropic artifacts, and superior RAG (this is worth another post). I have some workloads for which the "Message Batches API" will be useful. exciting time, eager to see what next

With this pull request, users should now have the ability to access all the relevant WGSL compute primitives within TSL, including the invocationLocalIndex, invocationSubgroupIndex, workgroupId, and numWorkgroups compute builtins: https://lnkd.in/gF6nP7qf. Accordingly, the bitonic sort sample has also been updated to access the workgroupId directly rather than deriving it from existing values.

This question explores how Docker images are created in a layered file system, using copy-on-write for shared libraries stored in physical RAM, and how files are shared between container-based processes via an overlay file system. More: https://lnkd.in/gwh5Xb6e

Scenario: My manager said I want to see output of a command(df -h) which will be stored in a varuable. Solution: Step1 Firstly we will store command in a variable a=$(df -h) Step2 Then with "echo" command, we will redirect output in .txt file echo $a > output.txt

Found a new use case for LLMs that I enjoy: adding log statements to production code. They determine appropriate trace levels and effectively capture error conditions. Check out the screenshot for the log statements added to my Rust code!

you can use the -f flag to specify the path to the docker file if you don't have it in root dir.

I know that I can use either of the two options below to resolve serverless variables with values f Check it out: https://lnkd.in/dhCAZv7P Join the conversation! #serverlessframework

Kubectl cp Command: How to Copy Files From Kubernetes Pods

Day 80/90: Decrypting Code Today, I tackled a problem that involves decrypting a list of integers based on a given k. Depending on the value of k, you either sum the next k elements (if k > 0) or the previous |k| elements (if k < 0), all while handling circular indexing.