{"id":115974,"date":"2024-05-21T14:59:30","date_gmt":"2024-05-21T12:59:30","guid":{"rendered":"https:\/\/prohoster.info\/blog\/novosti-interneta\/proekt-bend-razvivaet-vysokourovnevyj-yazyk-dlya-parallelnyh-vychislenij-na-gpu"},"modified":"2024-05-21T14:59:30","modified_gmt":"2024-05-21T12:59:30","slug":"proekt-bend-razvivaet-vysokourovnevyj-yazyk-dlya-parallelnyh-vychislenij-na-gpu","status":"publish","type":"post","link":"https:\/\/prohoster.info\/en\/blog\/news\/proekt-bend-razvivaet-vysokourovnevyj-yazyk-dlya-parallelnyh-vychislenij-na-gpu","title":{"rendered":"The Bend project is developing a high-level language for parallel computations on GPUs","gt_translate_keys":[{"key":"rendered","format":"text"}]},"content":{"rendered":"
The first releases of the Bend project have been published, developing a high-level programming language for parallel computing, positioned as an alternative to low-level languages like CUDA and Metal, featuring expressive syntax and development convenience typically seen in languages like Python and Haskell. The project code is written in Rust and is distributed under the Apache 2.0 license. <\/p>\n
The Bend language features quick object distribution, the use of higher-order functions, closures, continuations, unlimited recursion, pattern matching, recursive matching (fold), and loops (bend), as well as support for integer, string, and list types. It supports two syntax variants \u2014 Python-style and Haskell-style. Programs do not require the specification of concurrency annotations, explicit thread creation, or lock management. Parallelization occurs automatically; for example, in the expression \"((1 + 2) + (3 + 4))\", the operations \"1 + 2\" and \"3 + 4\" are executed in parallel. <\/p>\n
Programs written in Bend can run on hardware that supports massive parallelism, such as GPUs, demonstrating nearly linear performance growth depending on the number of computational cores. Bend code is compiled to a low-level intermediate representation called HVM2 (Higher-order Virtual Machine 2), which is then compiled to C and CUDA representation. Currently, the project only supports execution on NVIDIA GPUs. <\/p>\n
In terms of performance, a test application implementing bitonic sorting took 12.15 seconds on a single thread with an Apple M3 Max CPU, 0.96 seconds using 16 threads, and 0.21 seconds utilizing a NVIDIA RTX 4090 GPU with 16,000 threads. \ndef sort(d, s, tree):\n switch d:\n case 0:\n return tree\n case _:\n (x,y) = tree\n lft = sort(d-1, 0, x)\n rgt = sort(d-1, 1, y)\n return rots(d, s, lft, rgt)\ndef rots(d, s, tree):\n switch d:\n case 0:\n return tree\n case _:\n (x,y) = tree\n return down(d, s, warp(d-1, s, x, y))\n \u2026 \n Source: opennet.ru<\/a> <\/p>","protected":false,"gt_translate_keys":[{"key":"rendered","format":"html"}]},"excerpt":{"rendered":"