Vaibhav Srivastav’s Post

to generate associative shape include each one of A[0]-A[7] inside parentheses in recursive equation from index 7 to 0 in reverse order in previous algorithm for compiler formula.

Well i haxored on TensorRT-LLM all weekend, it was fun, learned some cool new stuff, but the stop words are broken. its probably my implementation, but the tokenizer is using the wrong tokens for the stop words, i found them, and it works when i hard code them, but it's just a pita. gotta figure out what's up with that. https://lnkd.in/gnC9Pc9A i put the code up if anyone is interested in playing with it. this is setup to run on 2 gpu's, you'll have to set tp to 1 if you've only got a single gpu.

Meta releases the Meta LLM Compiler, a family of models built on Code Llama with advanced code optimization and compiler capabilities, setting new benchmarks in code size optimization and disassembly tasks. The LLM Compiler can emulate the compiler, predict optimal passes for code size, and disassemble code. Meta is releasing the LLM Compiler 7B & 13B models under a permissive license for research and commercial use to make it easier for developers and researchers to leverage this technology and propel new research in this highly impactful domain. https://lnkd.in/eu63-yxP

There's this idea thrown around: let's define all undefined behavior to be the platform's behavior. In this article, I take a serious look at it. I also discuss a vision for compiler optimization that has been very dear to me. https://lnkd.in/ghtVWSyz

Exampreptool's Compiler Design playlist provides comprehensive coverage of the essential topics making complex concepts easy to master, helping you excel in the upcoming GATE examination Link to the entire Compiler Design Playlist on our channel: https://lnkd.in/gehRPfDv #viralvideos #viral #explore #explorepage #compilerdesign #cd #psu #dsa #algorithms #gate2025 #gatecse #gateda #preparation #examination #IITs #testseries #revision #gate

This code given below print a matrix.whose element you are input to compiler.

Welcoming the new Programming Language "Bend" that takes parallel computing to next level. Make everything run in parallel automatically by just writing Code in High Level. No need to manually code for threads anymore! Extremely beneficial project, keep up the good work! https://higherorderco.com/ #Bend #higherordeco #Parallel_Computation #Automatically

#HowTo add Python3-pip package to #RK3568 #SBC Buildroot in our latest blog 👇 https://lnkd.in/eDwv9uQM 👨🏫 This guide covers everything from hardware breakpoints setting, memory access monitoring, and peripheral debugging on ARM Cortex-A processors. Optimize your development process with practical tips and in-depth methods tailored to embedded engineering! #ForlinxEmbedded #Rockchip #AI #embeddedsoftware #EmbeddedSystems #TechInnovation #nxppartner #tutorial #sbc #EmbeddedDevelopment #TechTips #developmentboard

ICYMI➡️ Faith Ekstrand explains the twists and turns taken to add a unit testing framework to Maxwell compiler back end for #NVK: http://col.la/compilenvk #NVIDIA #VulkanAPI #OpenGL #OpenSource

Yesterday I posted about how CPython generates the LOAD_FAST_LOAD_FAST instruction to load two values onto the stack in a single instruction. Let's see how that optimization happens. Initially the CPython compiler emits two LOAD_FAST instructions for loading a and b onto the stack. The optimizations are done afterwards. The instructions generated by the compiler are put together in the form of a control flow graph (CFG) to perform some optimizations. A CFG organizes the instructions in the form of a graph where the nodes are called basic blocks consisting of a sequence of instructions. A basic block ends when it reaches a control flow instruction, such as jump, return, branch (if/else) etc. And this block gets connected via an edge to the target of the jump or control flow. Once you have such a graph, you can perform a bunch of optimizations. One of the simplest ones is elimination of unreachable code, i.e. a node in the graph that is not connected to any other node. Coming back to CPython, it builds a CFG based on the above scheme and then does a bunch of optimizations One of these optimizations is generation of super instructions, such as LOAD_FAST_LOAD_FAST, which combines two LOAD_FAST instructions into a single instruction. The following code shows how it is performed. And this is the place where the instructions are actually updated.