Talk:General-purpose computing on graphics processing units: Difference between revisions
Content deleted Content added
Tags: Mobile edit Mobile web edit Advanced mobile edit Reply |
Tags: Mobile edit Mobile web edit Advanced mobile edit New topic |
||
Line 300:
"Most operations on the GPU operate in a vectorized fashion: one operation can be performed on up to four" is an uncited claim. most GPUs are proprietary. also just because the ISA exposes an API of 4-wide SIMD does '''not''' mean that internals are the same. Broadcom VideoCore IV offered 16-wide FPMAC but internally broke that down into 4of pipelined 4-wide SIMD. [[User:Lkcl|Lkcl]] ([[User talk:Lkcl|talk]]) 10:32, 13 July 2025 (UTC)
== okayyy.... great article but definitely needs citing / tidy ==
this is a really comprehensive article, some fantastic tables, really useful info BUT, it's something of a.. it's... how can I put it best... fragmented? it needs tying in to other pages, clarifying and so on. i just added a link to [[Tiled rendering]] for example and clarified on the problems associated with GPU workloads (memory latency being key). will slowly get on it. [[User:Lkcl|Lkcl]] ([[User talk:Lkcl|talk]]) 12:00, 12 August 2025 (UTC)
| |||